Identification of the nuclear localization signal of the POU domain protein Tst-1/Oct6.

POU domain proteins are important regulators of development and terminal differentiation based upon their transcriptional activity in the nucleus. Here, we analyzed the mechanism underlying the nuclear localization of Tst-1/Oct6, a member of this family that regulates events during neurogenesis and myelination. Nuclear localization of Tst-1/Oct6 was dependent on the POU domain, as its deletion prevented access to the nucleus, whereas its transfer to the amino terminus of β-galactosidase was sufficient to prompt nuclear accumulation of this normally cytosolic protein. Interestingly, nuclear localization and high affinity DNA binding were two independent functions of the POU domain and could be separated in several mutants. While specific high affinity binding to DNA required the presence of both the POU-specific and the POU homeodomain, the POU-specific domain was dispensable for nuclear localization of Tst-1/Oct6. Rather, the nuclear localization function was selectively contained within the POU homeodomain. Specifically, a basic cluster (GRKRKKRT) preceding helix 1 of the homeodomain was shown by deletion mutagenesis to be involved in the nuclear localization of Tst-1/Oct6. This sequence, which is highly conserved among POU domain proteins, was by itself capable of translocating β-galactosidase to the nucleus defining it as the bona fide nuclear localization signal of Tst-1/Oct6 and presumably other POU domain factors.

POU domain proteins constitute a class of transcription factors that all share a highly conserved DNA-binding domain and are usually expressed during various critical, tightly controlled phases of embryonic development or cellular differentiation (for review see Refs. [1][2][3][4][5]. A typical expression pattern for POU domain proteins is displayed by the class III member Tst-1, which is also known as Scip or Oct6 (6 -12). In the mouse, it is transcribed from an intronless gene at the distal end of chromosome 4 that exhibits the features of an expressed retroposon (13,14). During development, expression of this protein, which hereafter will be referred to as Tst-1/Oct6, is detected in embryonic stem cells, skin, neuronal subpopulations, and in precursors of myelinating glia (6,11,(15)(16)(17). In the adult, expression of Tst-1/Oct6 persists in pyramidal neurons of layer 5 in the cerebral cortex and in skin (1,6,18).
Schwann cells which represent the myelinating glia of the peripheral nervous system have been used extensively to study the function of Tst-1/Oct6. In these cells, transient expression of Tst-1/Oct6 can be induced by axons and by agents that raise the intracellular cAMP level (10,19). Expression of Tst-1/Oct6 correlates with a period of rapid cell division which immediately precedes the onset of myelination (17). In line with its supposed role during differentiation of myelinating glial cells, targeted expression of a dominant negative form of Tst-1/Oct6 in Schwann cells led to severe disturbances in the normal myelination program (20).
Like all other POU domain proteins characterized so far (2,3,5), Tst-1/Oct6 has to exert its function in the nucleus. Nuclear proteins obligatorily enter this cellular compartment via the nuclear pore complex after being synthesized in the cytoplasm (21). In general, they are actively transported through the nuclear pore. This active transport requires energy (22,23), soluble factors, such as importin-␣, importin-␤, and Ran/TC4 (24 -27), and a nuclear localization signal (NLS) 1 within the protein targeted for nuclear import (for review see Ref. 28). NLSs were first discovered in the yeast Mat␣2 protein (29,30) and in SV40 large T-antigen (31,32) and are recognized by NLS-binding proteins such as the importin-␣⅐importin-␤ complex. After the initial NLS recognition in the cytosol which is mainly mediated by importin-␣, the complex docks to the nuclear pore via importin-␤, before it gets translocated through the pore in a Ran/TC4-dependent manner (25). NLSs have been identified in a variety of nuclear proteins ranging in size from less than a hundred to more than a thousand amino acids (33), including polymerases (34), kinases and phosphatases (35,36), transcription factors (37)(38)(39), histones (40), growth factors (41,42), tumor suppressors (43,44), and various viral proteins (31,32,(45)(46)(47)(48). Here, we have characterized the nuclear localization signal of Tst-1/Oct6, which is localized in the protein's DNA-binding domain as a structure highly conserved among POU domain proteins in general. Its identification allows insights into the evolution of POU domain proteins and points to ways of regulating the access of POU domain proteins to the nucleus.

MATERIALS AND METHODS
Plasmid Constructs-Tst-1/Oct6 expression plasmid pCMV/Tst-1 as well as the mutants pCMV/Tst-1 ⌬N (missing amino acids 4 -240), pCMV/Tst-1 ⌬POU (missing amino acids 241-395), pCMV/Tst-1 ⌬POU HD (missing amino acids 336 -395), pCMV/Tst-1 ⌬C (missing amino acids 396 -448), pCMV/Tst-1 ⌬NC (missing amino acids 4 -240 and 396 -448), and pCMV/Tst-1 mt (carrying a double point mutation at amino acids 383 and 384 in the recognition helix of the POU homeodo-main) have been described before (49,50). Additional point mutants of Tst-1/Oct6 were created as follows. In pCMV/Tst-1 PM1 amino acids 262 and 263 were changed from two arginines to aspartic acid and serine; in pCMV/Tst-1 PM2 amino acids 269 and 270 were changed from threonine and glutamine to methionine and glutamic acid; in pCMV/ Tst-1 PM3 amino acids 286 and 287 were changed from serine and glutamine to valine and aspartic acid; in pCMV/Tst-1 PM4 amino acids 292 and 294 were changed from arginine and glutamic acid to glutamic acid and glutamine; in pCMV/Tst-1 PM5 amino acids 338 -340 were changed from two lysines and one arginine to three alanines; in pCMV/ Tst-1 PM6 amino acids 387 and 388 were changed from two arginines to valine and aspartic acid; and in pCMV/Tst-1 dn amino acids 389 and 393 were changed from a glutamine and an arginine to a glycine and a tryptophan. The control plasmid pCMV/asTst-1 contained the region coding for Tst-1/Oct6 inserted in antisense orientation relative to the cytomegalovirus promoter. pRSVluc and pHSVoct-luc which contain the firefly luciferase gene under the control of the Rous sarcoma viruslong terminal repeat and a combination of herpes simplex virus octamer element (5Ј-GCATGCTAATGATATTCTTT-3Ј) and rat prolactin minimal promoter, respectively, have been described before (50,51).
All ␤-galactosidase expression vectors were based on pCMVlacZ (a gift of Dr. G. E. DiMattia, London Regional Cancer Centre, London, Canada) which contained the lacZ gene under the control of the CMV promoter. This lacZ gene lacked the first eight nonessential amino acids and instead contained a eukaryotic translation consensus sequence at its 5Ј end, and the SV40 small t intron A and polyadenylation signal at its 3Ј end. A short sequence encompassing the nuclear localization signal from the T-antigen of SV40 (32) was added to the 5Ј end of the lacZ gene, yielding pCMVlacZ(SV40T NLS). Similarly, fragments from the tst-1/oct6 gene were added as NcoI fragments to the 5Ј end of lacZ. pCMVlacZ(POU) contained sequences coding for amino acids 239 -404 of Tst-1/Oct6; pCMVlacZ(POU S ) contained sequences coding for amino acids 239 -324; and pCMVlacZ(POU HD ) contained sequences coding for amino acids 328 -404. Plasmid pCMVlacZ(Tst-1 NLS) was obtained by inserting a short sequence corresponding to amino acids 334 -341 of Tst-1/Oct6 (GRKRKKRT) between NcoI and NruI sites directly behind the start methionine.
Tissue Culture, Transfection, and Preparation of Protein Extracts-CV1 cells were maintained in Dulbecco's modified Eagle's medium and U138 glioblastoma cells in RPMI 1640, both supplemented with 10% fetal calf serum. One day prior to transfection, CV1 or U138 cells were plated at a density of 2 ϫ 10 5 per 60-mm plate in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Cells were transfected with 2 g of Tst-1/Oct6 expression vector and 2 g of luciferase plasmid using the calcium-phosphate technique as described (49,52). Cells were harvested 48 h posttransfection for luciferase assays (49) or for the preparation of extracts. Extracts were prepared as described (53) with minor modifications (52). Cells from two 60-mm plates yielded 300 l of cytosolic extract and 150 l of nuclear extract.
Western Blot Analysis-Western blot analysis was performed as described (52). Detection of luciferase and Tst-1/Oct6 in approximately 50 g of cytoplasmic extract and 25 g of nuclear extract was achieved using rabbit polyclonal antisera against either luciferase (gift of Dr. S. Subramani, La Jolla) or against full-length Tst-1/Oct6 purified from baculovirus-infected Sf9 cells. The amount of cytoplasmic and nuclear extract corresponded to the equivalent of 1.3 ϫ 10 5 cells.
Electrophoretic Mobility Shift Assay-Double-stranded oligonucleotides containing the herpes simplex virus octamer motif (5Ј-GCAT-GCTAATGATATTCTTT) were generated and radiolabeled with Klenow enzyme and [␣-32 P]dCTP. For electrophoretic mobility shift assays, 0.5 ng of labeled probe was incubated with nuclear extracts of cells transfected with Tst1/Oct6 expression plasmids for 20 min at room temperature in a 20-l reaction containing 10 mM Hepes, pH 8.0, 5% glycerol, 100 mM NaCl, 5 mM MgCl 2 , 2 mM dithiothreitol, 0.1 mM EDTA, 2 g of poly(dI-dC), and 0.5 g of bovine serum albumin. One-third of each reaction was loaded onto a 5% nondenaturing acrylamide gel. Electrophoresis was in 0.5 ϫ TBE at 200 V for 3 h.
Immunocytochemistry and ␤-Galactosidase Histochemistry-CV1 cells were seeded on chamber slides (Lab-Tec, Nunc Inc.) and transfected as described above. 48 h posttransfection, the medium was removed, and the cells were washed twice with PBS. Cells were then fixed with 3% formaldehyde in PBS for 20 min and treated with 1% Triton X-100 in PBS for 5 min. After washing the cells twice with PBS, cells were incubated for 20 min with the polyclonal anti-Tst1 antiserum (diluted 1:1000) in PBS containing 0.1% Tween 20 (PBST). After washing three times with PBST, cells were incubated 20 min with Cy 3conjugated goat anti-rabbit antibodies (Dianova) diluted 1:500 in PBST. Cells were washed extensively with PBST, mounted, and analyzed on an Axiovert microscope (Zeiss).

Tst-1/Oct6 Is a Predominantly Nuclear
Protein-Using electrophoretic mobility shift assays as well as Western blot analysis, Tst-1/Oct6 has been shown in previous studies to be present in nuclear extracts from embryonic stem cells, embryonic carcinoma cells, proliferating Schwann cells, oligodendrocyte precursors, and stably transformed fibroblasts (9,11,13,15). To be able to study the cellular distribution of Tst-1/Oct6 in more detail, we used rabbit polyclonal antibodies raised against full-length Tst-1/Oct6 protein that contained a 6xHIStag fused to its amino terminus and was purified from baculovirus-infected Sf9 cell extracts. When tested on nuclear and cytosolic extracts prepared from embryonic stem cells or U138 glial cells transiently transfected with Tst-1/Oct6, this antibody recognized specifically and at high dilution a protein of 47,000 that was enriched in the nucleus ( Fig. 1, and data not shown). This protein was absent from U138 cells not transfected with an expression vector for Tst-1/Oct6 showing that this protein was indeed Tst-1/Oct6. A small fraction of Tst-1/ Oct6 was also found in the cytosolic extracts. It is unclear at present whether this is an artifact of the extract preparation or represents truly cytosolic Tst-1/Oct6.
The POU Domain Targets Tst-1/Oct6 to the Nucleus-To be able to map the region of Tst-1/Oct6 responsible for its predominantly nuclear localization, we used a series of mutants in which single domains of the Tst-1/Oct6 protein had been removed (Table I). Their DNA binding activity and ability for transcriptional activation were assessed in standard electrophoretic mobility shift assays and transient transfection assays. As summarized in Table I, the results correlated well with previously published data mapping DNA binding to the POU domain and transactivation function to the amino-terminal part of the protein (50,55,56).
To determine the intracellular localization of Tst-1/Oct6 deletion mutants, we cotransfected Tst-1/Oct6 and the luciferase expression plasmid pRSVluc into various cell types and performed cell fractionation studies. The results obtained in U138 glial cells are shown in Fig. 1, but identical results were also obtained in CV1 and HeLa cells (data not shown). The quality of cytoplasmic and nuclear extracts was assessed in each case by Western blot analysis with a rabbit polyclonal antiserum directed against luciferase. Because luciferase is known to be a cytoplasmic protein present in the peroxisomes, it should preferably be detected in the cytoplasmic fraction, while it should be absent from the nuclear fraction. As shown in the bottom panel of Fig. 1, most of the luciferase is indeed localized in the cytoplasm, showing that contamination of the nuclear fraction with cytosol was minimal. Contamination of cytosol with nuclear proteins was equally low as evidenced by the distribution pattern of a high molecular weight band that exhibited weak cross-reactivity with the anti-Tst-1/Oct6 antiserum (data not shown). U138 cells transfected with wild-type Tst-1/Oct6 contained a 47,000 protein not present in mock-transfected cells, which was highly enriched in the nuclear fraction (Fig. 1). A similar pattern of distribution was also found for mutant Tst-1/Oct6 proteins that either had all sequences carboxyl-terminal (⌬C) or amino-terminal (⌬N) of the POU domain deleted.
When a mutant was assayed that just contained the POU domain of Tst-1/Oct6 (⌬NC), no immunoreactivity could be observed in Western blot analysis (Fig. 1A). The fact that this mutant was not detected was explained by the specificity of the antiserum that reacted with epitopes in the regions both amino-and carboxyl-terminal of the POU domain, but not with the POU domain itself. A bacterially expressed POU domain of Tst-1/Oct6, for instance, was not recognized by the antiserum (data not shown). Electrophoretic mobility shift analysis, on the other hand, clearly revealed a specific complex of high mobility, indicative of this mutant, that was predominantly found in the nuclear extract and comigrated with the complex formed by the bacterially expressed POU domain of Tst-1/Oct6 ( Fig. 1B and data not shown).
To corroborate the importance of the POU domain for nuclear localization and address the possibility of additional nuclear localization signals outside the POU domain, we deleted the POU domain from the protein and analyzed the cellular distribution of this mutant (⌬POU; Fig. 1A). Contrary to all other mutants analyzed, this mutant was predominantly found in the cytosolic fraction. These results imply that the POU domain is both necessary and sufficient to target Tst-1/Oct6 to the nucleus.
The POU Homeodomain Is Essential for Nuclear Localization-Analysis of nuclear localization by cell fractionation has been shown in the past to be susceptible to artifacts (57). During extract preparation, nuclear proteins that were not bound to nuclear constituents leaked from the nucleus and were recovered in the cytosol despite carrying a nuclear localization signal. Thus it is conceivable that the ⌬POU mutant was retrieved in the cytosolic fraction merely because of its inability to bind to DNA not, however, because of the true absence of a nuclear localization signal. Therefore, we tried to corroborate and extend our cell fractionation studies by immunocytochemistry on transiently transfected CV1 cells. The antiserum raised against Tst-1/Oct6 yielded only a weak background staining when employed to mock-transfected cells ( Fig. 2A). Transfection with an expression plasmid for wild-type Tst-1/Oct6, on the other hand, led to an intense staining of cell nuclei (Fig. 2B). Comparable nuclear staining was also detected with mutants ⌬N and ⌬C, indicating that these regions of Tst-1/Oct6 are dispensable for nuclear localization (Fig. 2, C and F). In contrast, the ⌬POU mutant was present predominantly in the cytosol, leading to a uniform, dispersed labeling of the entire cell (Fig. 2D). This obvious agreement between cell fractionation studies and immunocytochemistry strongly confirms the role of the POU domain in targeting Tst-1/Oct6 to the cell nucleus.
Intriguingly, a comparable loss of nuclear localization was observed with a mutant, which had only the POU homeodomain removed but still retained the POU-specific domain (⌬POU HD , Fig. 2E). This clearly points to the importance of the POU homeodomain for nuclear localization.
The POU Homeodomain Is Sufficient to Direct ␤-Galactosidase to the Cell Nucleus-For a closer inspection we transferred parts of Tst-1/Oct6 to a heterologous protein and evaluated the localization of the corresponding fusions. The fusion partner was cytosolic ␤-galactosidase (Fig. 3B). By the amino-terminal addition of a short stretch of amino acids, which represents the nuclear localization signal (NLS) of SV40 T-antigen, ␤-galacto-  a Transcriptional activation of HSV oct-luc was determined in U138 cells transiently transfected with 2 g each of reporter and effector plasmid and is presented as fold stimulation Ϯ S.E. Data were determined in at least three independent experiments, each performed in triplicate, and were calculated by comparing luciferase activities of extracts from cells transfected with expression plasmids for Tst-1/Oct6 with luciferase extracts from cells transfected with pCMV/asTst-1 instead.
b DNA binding was determined in standard electrophoretic mobility shift assays with the HSV octamer element as a probe and whole cell extract from transiently transfected U138 cells as a protein source.
c Cellular localization was determined both in subfractionation studies and immunofluorescence as shown partly in Figs. 1 and 2. sidase could, however, be directed to the cell nucleus as shown before (Fig. 3C) (58). When the POU domain of Tst-1/Oct6 was placed onto LacZ instead of the NLS of SV40 T-antigen, it was also able to target the fusion protein to the cell nucleus (Fig.  3D). In contrast, nuclear localization was not achieved by fusing the POU domain coding sequences in the opposite orientation to the lacZ gene, although a protein of the correct size was expressed in transfected cells (data not shown). This transfer experiment further corroborated the importance of the POU domain in the nuclear localization of Tst-1/Oct6.
In addition to the intact POU domain, we also transferred isolated regions of it to LacZ. The LacZ(POU HD ) fusion contained the POU homeodomain of Tst-1/Oct6 as well as seven amino acids from the preceding linker region, whereas the LacZ(POU S ) fusion carried the POU-specific domain of Tst-1/ Oct6 and most of the adjacent linker region which joins it to the POU homeodomain. Nuclear localization was only observed for the fusion protein between the POU homeodomain and ␤-galactosidase (Fig. 3F). The POU-specific domain, on the contrary, proved to be incapable of directing ␤-galactosidase to the nucleus as the fusion between LacZ and the combination of POU-specific domain and adjacent linker remained cytosolic (Fig. 3E).
A Basic Cluster in the POU Homeodomain Is Involved in the Nuclear Localization of Tst-1/Oct6 -To further define the nuclear localization signal present in the POU domain, we substituted highly conserved amino acids both in the POU-specific and the POU homeodomain (Table II). These regions were chosen because of their proven relevance for the function of the POU domain (59 -62). With the exception of mutant dn which affected a glutamine and an arginine residue in helix 3 of the POU homeodomain, all other mutants had lost their ability to bind to DNA. Concomitant with their loss of DNA binding these mutants failed to stimulate transcription from an octamercontaining promoter (Table II).
All mutants that affected the POU-specific domain (PM1, PM2, PM3, and PM4) were still found to be nuclear in immunocytochemistry studies, although none of them exhibited significant binding to an octamer DNA element. This finding is particularly noteworthy for the mutation present in PM1, which led to a disruption of the cluster of basic amino acids within helix 1 of the POU-specific domain that on the basis of its positive charge would have been a good candidate for a nuclear localization signal (28). In agreement with our transfer experiments, these results argue against a participation of the POU-specific domain in the nuclear localization of Tst-1/Oct6. Similar to the mutations in the POU-specific domain, most homeodomain mutants remained unaffected in their nuclear localization despite being defective in their DNA binding activity and transcriptional activation function. None of the proteins with mutations in helix 3 of the POU homeodomain were impaired in their nuclear localization function, although at least two of them (PM6 and dn) disrupted the basic cluster present in this DNA recognition helix. Therefore, it seemed unlikely that helix 3 of the POU homeodomain was involved in nuclear localization. This conclusion was also supported by the finding that helix 3 when transferred to a cytosolic form of luciferase was not able to target this protein to the nucleus (data not shown).
In marked contrast to all other analyzed mutants, PM5 exhibited a strong cytoplasmic staining (Fig. 4A). PM5 had three positive charges removed from the basic cluster at the beginning of the POU homeodomain. Thus, while there are three basic clusters in the POU domain, only the one preceding helix 1 of the POU homeodomain was involved in nuclear localization.
A Basic Cluster from the POU Homeodomain of Tst-1/Oct6 Is Sufficient to Direct ␤-Galactosidase to the Cell Nucleus-To analyze whether this basic cluster alone represented the nuclear localization signal of Tst-1/Oct6 or functioned only in the context of the POU homeodomain as part of a more complex signal, we transferred the isolated basic cluster onto the amino terminus of LacZ and analyzed the resulting ␤-galactosidase fusion for its cellular localization. As shown in Fig. 4B, this cluster was sufficient to direct ␤-galactosidase to the nucleus of transfected cells. Indeed, this short stretch of amino acids from Tst-1/Oct6 was as efficient in targeting a heterologous protein to the cell nucleus as the NLS from SV40 T-antigen and might be regarded as the NLS of Tst-1/Oct6. DISCUSSION Most proteins that enter the nucleus do so by an active transport through the nuclear pore that requires the presence of an NLS in the transported protein. Although there is no consensus sequence, NLS are usually short sequences that contain a high proportion of positively charged amino acids. They occur at various locations within the protein in single or multiple, functionally redundant copies. They can be contained within a contiguous stretch of amino acids or may be bi-or multipartite instead (28,33). Furthermore, NLS can be strongly dependent in their function on the exact flanking sequences (63,64). Deletion of an NLS from a nuclear protein leads to its redistribution to the cytoplasm, whereas its addition to a heterologous cytoplasmic protein often results in an accumulation of this protein in the nucleus.
The present study shows that the POU domain of Tst-1/Oct6 harbors a signal that fulfills the above-mentioned criteria of an NLS. Consequently, deletion of the POU domain caused the resulting Tst-1/Oct6 mutant to localize to the cytoplasm, whereas its addition to the usually cytoplasmic ␤-galactosidase caused it to become nuclear. Thus, the POU domain is not only involved in DNA binding (62), and interaction with a whole series of cellular and viral coactivators (49,(65)(66)(67)(68)(69), it also contains the protein's nuclear localization function. Our study, therefore, lends further proof to the multifunctional character and the tremendous importance of the POU domain in this class of transcription factors.
Mapping of the NLS was refined in consecutive studies to the POU homeodomain. Transfer of the POU-specific domain to ␤-galactosidase, on the other hand, did not cause the fusion protein to become nuclear. The absence of a nuclear localization function from the POU-specific domain was emphasized by the fact that none of the amino acid changes introduced into the four ␣-helices of the POU-specific domain (59, 60) disrupted nuclear localization, although severely impeding other functions of Tst-1/Oct6. Not even the removal of two positive charges from the basic cluster within helix 1 of the POUspecific domain in PM1 prevented Tst-1/Oct6 from becoming nuclear. Our conclusions are also supported by the fact that a naturally occurring splice variant of the POU domain protein Pit-1 which did not contain a POU-specific domain was found to enter the nucleus and serve as a dominant repressor of Pit-1 function (70,71).
Taking the accumulation of positively charged amino acids within nuclear localization signals into account, the POU ho-  b DNA binding was determined in standard electrophoretic mobility shift assays with the HSV octamer element as a probe and whole cell extract from transiently transfected U138 cells as a protein source.
c Cellular localization was determined in immunocytochemistry studies on CV1 cells transiently transfected with the respective Tst-1/Oct6 mutants. meodomain contains two candidate NLS. One of these clusters of basic residues is located in front of helix 1 of the homeodomain and the other within helix 3, the DNA recognition helix. When these basic clusters were disrupted, strongly divergent results were obtained for each cluster. Whereas the basic cluster within helix 3 of the POU homeodomain was dispensable for nuclear localization, destruction of the basic cluster that precedes helix 1 of the POU homeodomain caused an effective redistribution of Tst-1/Oct6 from the nucleus into the cytosol. A peptide (GRKRKKRT) corresponding to the same basic cluster (position 334 -341) conferred nuclear localization onto a heterologous cytosolic protein. By targeting ␤-galactosidase to the nucleus, this region was clearly defined as the bona fide NLS of Tst-1/Oct6. To our knowledge, this is the first NLS mapped in any POU domain protein.
A prerequisite for the function of nuclear localization signals is their proper exposure on the surface of the protein. The basic cluster preceding helix 1 of the POU homeodomain fulfills this criterion. Although being involved in DNA binding through a series of contacts with the phosphate backbone (Lys 338 , Arg 340 , Thr 341 ), and a base in the minor groove of DNA (Arg 340 ), this region is solvent-exposed and readily accessible for NLS-binding proteins (62).
As shown in Table III, this basic cluster is highly conserved among POU domain proteins. In fact, there is only one exception known so far. In I-Pou, a Drosophila class IV POU domain protein, the number of basic residues is reduced to only three (72). That I-Pou still enters the nucleus is probably best explained by the fact that it is also the only POU domain protein that contains three consecutive basic residues in the linker region preceding the residual cluster by 12 amino acids. The resulting sequence KRRDPDAPSVLPAGEKKRT conforms to the consensus for bipartite NLS that consists of two clusters of basic residues separated by 10 -12 amino acids (28). This peculiarity of the I-Pou sequence might very well explain why I-Pou is the only POU domain protein that could afford a disruption of the basic cluster in front of helix 1 of its homeodomain.
The high degree of conservation of the basic cluster is even more striking in light of the significant differences in the flanking regions between the various classes of POU domain proteins. It strongly argues in favor of a model in which this basic cluster functions as a nuclear localization signal in most POU domain proteins. Therefore, we have most likely identified a general mechanism of nuclear entry for POU domain proteins.
We expect our results for Tst-1/Oct6 to be prototypic for other members of the whole family of POU domain proteins.
Interestingly, the high conservation of this basic region within the POU domain family contrasts sharply with its apparent absence in several classic homeodomain proteins. The yeast Mat␣2 protein, for instance, has two NLS sequences that function independently of each other and are both localized at positions different from the NLS of Tst-1/Oct6 (29,30). Thus, the NLS seems to be a recent acquisition of POU domain proteins during evolution. It deserves to be noticed that the presence of an NLS in the DNA-binding domain of the protein is a principle also realized in other classes of transcription factors, including HMG box, HLH, and bZip proteins (37)(38)(39). Although DNA binding and nuclear localization are functionally separable, colocalization of both functions might have evolved as a consequence of exon shuffling in higher eukaryotes.
The NLS of Tst-1/Oct6 is structurally very similar to the NLS of SV40 T-antigen (PKKKRKV) (31). Unlike the majority of nuclear proteins that contain bipartite NLS (28), Tst-1/Oct6 and T-antigen have single contiguous stretches of positively charged residues that serve as their NLS. The analogy reaches even further. The NLS of SV40 T-antigen is in close proximity to phosphorylation sites for casein kinase II and cdk/cdc2. Phosphorylation of these sites has been shown to increase or reduce the rate of nuclear import, respectively (73,74). Interestingly, the NLS of Tst-1/Oct6 overlaps with a potential phosphorylation site for protein kinase A and an as yet unidentified M-phase-specific kinase, which similar to the NLS itself is strongly conserved among POU proteins and has been shown to be the target of phosphorylation in the related POU domain proteins Pit-1 (75,76) and Oct-1 (77). Although not observed so far, it is intriguing to speculate that phosphorylation at this site could regulate nucleocytoplasmic transport of POU domain proteins in a manner similar to other proteins such as Tantigen (for review see Ref. 78). This would provide yet another means for regulation of POU domain protein activity.
As mentioned above, the POU domain is also the main region for interactions between members of this class of transcription factors and other cellular proteins (49,(65)(66)(67)(68)(69)79). It is easily conceivable that such interactions also influence the accessibility of the NLS of Tst-1/Oct6. The POU domain might be an excellent target for interactions with cytosolic retention factors similar to the ones identified in other classes of transcription factors (80). Our characterization of the NLS of Tst-1/Oct6 should therefore open new directions in the functional analysis of POU domain proteins.