J Biol Chem, Vol. 274, Issue 41, 29202-29210, October 8, 1999

The Quaking I-5 Protein (QKI-5) Has a Novel Nuclear Localization Signal and Shuttles between the Nucleus and the Cytoplasm^*

Jiang Wu, Li Zhou, Kathryn Tonissen, Ronald Tee^§, and Karen Artzt^¶

From the Institute for Cellular and Molecular Biology and Department of Microbiology, University of Texas, Austin, Texas 78712-1064

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The mouse quaking (qk) gene is essential in both myelination and early embryogenesis. Its product, QKI, is an RNA-binding protein belonging to a growing protein family called STAR (signal transduction and activator of RNA). All members have an ~200-amino acid STAR domain, which contains a single extended heteronuclear ribonucleoprotein K homologue domain flanked by two domains called QUA1 and QUA2. We found that QKI isoforms could associate with each other, while one of the lethal mutations qkI^kt4 with a single amino acid change in QUA1 domain, leads to a loss of QKI self-interaction. This suggests that the QUA1 domain is responsible for QKI dimerization. Three QKI isoforms have different carboxyl termini and different subcellular localization. Here, using GFP fusion protein, we identified a 7-amino acid novel nuclear localization sequence in the carboxyl terminus of QKI-5, which is conserved in a subclass of STAR proteins containing SAM68 and ETLE/T-STAR. Thus, we name this motif STAR-NLS. In addition, the effects of active transcription, RNA-binding and self-interaction on QKI-5 localization were analyzed. Furthermore, using an interspecies heterokaryon assay, we found that QKI-5, but not another STAR protein ETLE, shuttles between the nucleus and the cytoplasm, which suggests that QKI-5 functions in both cell compartments.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

QKI¹ is a KH-type RNA-binding protein that has been identified as the product of the mouse quaking (qk) locus (1). qk^v, the original viable allele described by Sidman (2) is a recessive dysmyelination mutation that affects the nervous system globally. Mutant mice start to have tremors on day 11 after birth and develop seizures as they mature. There are also several ethylnitrosourea-induced mutations to qk (qk^enu), but surprisingly, these are all early embryonic lethals (3). These mutations, along with the recent knockout of quaking,² demonstrate that this locus is pleiotropic, affecting diverse systems and implying that it defines some fundamental process employed by many tissues.

QKI, along with germ line-deficient (GLD-1) from Caenorhabditis elegans and the mammalian cell signaling protein SAM68, defines a large, evolutionarily conserved, extended KH domain family of proteins called STAR (signal transduction and activator of RNA) (1, 4). STAR proteins also contain two new domains immediately flanking the KH domain, QUA1 and QUA2, for which no function has yet been assigned (Fig. 1). All of the STAR proteins studied have been shown to bind RNA (5-7) and some to self-associate (5, 8); however, to date, there is experimental evidence for a role in cell signaling only for SAM68 (9). This is implied for some other family members from sequence motifs such as proline-rich regions containing potential Src homology 3-binding sites and a carboxyl tail rich in tyrosine residues.

Where mutants exist, the STAR proteins display many interesting phenotypes. The Drosophila qkI homologue how/who is involved in muscular function and larval development since some mutations are flightless and others are larval lethals (10-12). Some alleles of C. elegans gld-1 have an overgrowth of mitotic oocytes, whereas others have non-essential effects on male fate specification in the hermaphrodite germ line (13). Mammalian splicing factor 1 (SF1/ZMF1) is the homologue of Bbp1, the only STAR family member in yeast, which is a component of the pre-splicing complex (7, 14).

In the rapidly growing STAR family, other members include several QKI homologs identified in frog (15), chicken (16), zebrafish (17), and human (GenBank AL031781). Recently, two genes similar to SAM68 were cloned: étoile/T-STAR/slm-2 and slm-1 (18, 19). In Drosophila, which has been studied by degenerate PCR, there are at least nine STAR family members (20); in C. elegans, where the sequence is complete, using BLAST, there are eight good matches to the STAR domain.

The mouse qkI locus produces a diverse set of proteins by alternative splicing (21). The first three studied in any detail (QKI-5, -6, and -7) appear to have different roles in development. They are constructed with the same 311-amino acid body but have different carboxyl tails consisting of 30, 8, and 14 amino acids, respectively. QKI-5 is prominently expressed in early embryogenesis, and the abnormality of QKI-5^enu is therefore believed to be responsible for the lethality of qk^enu embryos around 9.5 days of gestation. In contrast, QKI-6 and -7 are only detectable by Northern blots in late development when myelination starts in earnest. Their very low expression in myelinating glia is believed to be responsible for the qk^v phenotype (1, 22).

Using antibodies specific for each carboxyl tail, Hardy et al. (22) showed that in mouse brain, all three isoforms are expressed in glial cells. Remarkably, QKI-5 is nuclear whereas QKI-6 and -7 are predominantly cytoplasmic (22). Since the only difference in these proteins is their carboxyl tail, the 30-amino acid QKI-5 tail might be responsible for nuclear localization, although there is no classical basic nuclear localization signal (NLS) (23, 24). However, the localization study of Xenopus QKI (XQUA), which is 94% identical to QKI-5, showed some differences. Using a Xenopus oocyte injection assay, XQUA was shown to be present in both the nucleus and the cytoplasm (5). In addition, in the mouse embryo, it was found that in a subset of migrating neural progenitors that later differentiate into glial cells, QKI-5 was redistributed from the nucleus to the cytoplasm (25). Thus, QKI-5 may function in both the nucleus and cytoplasm and may, in fact, shuttle between them.

In this study, we investigated the dynamic localization of QKI-5 using green fluorescent protein (GFP) as a reporter. Furthermore, the effects of self-interaction, RNA binding, and active transcription on its nuclear localization were also studied. The results show that the 30-amino acid carboxyl tail is necessary and sufficient for its nuclear localization and contained within it is a 7-amino acid novel nuclear localization sequence we name STAR-NLS. The same NLS is used by SAM68 and T-STAR/ETLE. However, using an interspecies heterokaryon assay, we find that, unlike T-STAR/ETLE, QKI-5 shuttles between the nucleus and cytoplasm in a transcription-dependent manner.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Plasmid Constructions-- The in vitro expression constructs pTkozQKI-5 and pTkozQKI-5^kt4 were generated by inserting the entire coding region of QKI-5 or QKI-5^kt4 with a Kozak start sequence into the BglII/EcoRV sites of the in vitro expression vector pT7Ts (26). The QKI-5 coding region was inserted in frame into the EcoRI/SalI site of pET28a+ (Novagen) to generate the clone pET28QKI-5 for bacterial expression. To generate the yeast two-hybrid assay constructs, the QKI isoforms and QKI-5^kt4 were inserted in frame into the EcoRI/SalI sites of pBTM116 (a kind gift from Dr. S. M. Hollenberg, Fred Hutchinson Cancer Research Center, Seattle, WA) or NcoI/SacI site of pACTII (CLONTECH). The subsequent QKI deletion constructs were made by PCR and cloned into the EcoRI site of the vectors. The primers used are: STAR1 (CATGGTCGGGGAAATGGAA) and QKCT (ATAACACACCACTGGGTTC) for QKI-body; 5kb1 (CCGGTGCGGTGGCTACTAAAG) and 5kb2 (TGTTGCCGGTGGCGGCTC) for QKI-5tail. pBTM116-Lamin (kindly provided by Dr. S. M. Hollenberg, Fred Hutchinson Cancer Research Center, Seattle, WA) and pTD1-1 (pACTII-T-antigen) (CLONTECH) were used as negative control for yeast two-hybrid assay.

To generate GFP fusion constructs, the QKI entire coding region and a series of mutated fragments: QKI-body, QKI-5^kt4, and QKI-5tail, were cloned in frame into the mammalian expression vector pEGFP-C1 (CLONTECH). The fusion proteins have GFP at the NH₂ terminus. The GFP-NLS fusion constructs were generated by annealing three sets of two oligos: NLS1 (AATTCCCGTGTCCATCCTTACCAAAGGG) and NLS2 (GATCCCCTTTGGTAAGGATGGACACGGG) for QKI-5-NLS; SAM68-NLS1 (AATTCCAGAGAGCATCCATATGGACGTG) and SAM68-NLS2 (GATCCACGTCCATATGGATGCTCTCTGG) for SAM68-NLS; ETLE-NLS1 (AATTCCAGAGACCAGCCATATGGCAGAG) and ETLE-NLS2 (GATCCTCTGCCATATGGCTGGTCTCTGG) for ETLE-NLS, and cloned into the EcoRI/BamHI site of pEGFP-C1. The QuickChange site-directed mutagenesis kit (Stratagene) was used to mutate amino acids Y to F or R to A. GFP-QKI-5KH was made by PCR using the primer dKH1 (ATCAGGGTATTCTTTTACAGGCAC) and dKH2 (GAGGAGCAAAATAGAGGCAAGCCC) and pEGFP-QKI-5 as template. ETLE coding region was obtained by PCR from original cDNA and subcloned into the TA cloning vector PCR2.1 (Invitrogen). An ETLE-EcoRI fragment was inserted in frame into pEGFP-C1 vector. FLAG-QKI-5 was made by inserting QKI-5 coding region in frame into pCMV4 vector (Eastman Kodak Co.). pcDNA3-myc-hnRNP-A1 and pcDNA3-myc-hnRNP C1 were kindly provided by Dr. G. Dreyfuss (University of Pennsylvania, Philadelphia, PA) and described elsewhere (27).

Yeast Two-hybrid Assay-- The pBTM116QKI and pACTIIQKI constructs were cotransformed in different combinations into yeast strain L40 (28), followed by 2 days of incubation at 30 °C on the selective plate lacking tryptophan and leucine. At least six independent colonies were restreaked on another plate, and the X-gal filter assay (29) was used to determine the LacZ level of these transformants. Colonies turned blue within 2 h were considered positive.

In Vitro Coimmunoprecipitation-- The bacterial expression construct pET28QKI-5 was transformed into Escherichia coli BL21 DE3 cells. A culture of 0.6 (A₆₀₀) was induced by adding isopropyl-1-thio--D-galactopyranoside to a final concentration of 1 mM. 2 h after induction, cells were collected and broken by French press in the purification buffer (50 mM KH₂PO₄, 300 mM KCl, pH 7.0, and 1 mM phenylmethylsulfonyl fluoride). After centrifugation at 12,000 × g for 15 min, the supernatant was loaded onto a Ni-NTA agarose column (Qiagen), washed extensively with purification buffer, and eluted with purification buffer containing 300 mM imidazole.

The ³⁵S-labeled QKI proteins were produced by in vitro translation. The pTkozQKI-5 or pTkozQKI-5^kt4 was in vitro transcribed and translated using TNT coupled reticulocyte lysates system (Promega). One fifth of the reactions were mixed with 0.2 µg of His-tagged QKI-5 protein and mouse anti-6×His monoclonal antibody (CLONTECH) in the IP buffer (50 mM Tris·HCl, 150 mM NaCl, 5 mM MgCl₂, 1 mM EDTA, 0.25% Nonidet P-40, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, pH 7.5). The mixture was incubated at 4 °C overnight and precipitated by adding protein A-agarose beads (Santa Cruz Biotechnology). Luciferase was used as a negative control. After four washes with IP buffer, the precipitated proteins were resolved on a 10% SDS-PAGE gel and visualized by fluorography.

Cell Culture and Transfection-- HeLa (human epithelial carcinoma cell), COS7 (green monkey kidney cell), and NIH3T3 (mouse embryonic fibroblast cell) cells were used in the study. Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37 °C with 5% CO₂. For microscopy, cells were grown in Lab-Tek chamber slides (Nunc) and were transfected with plasmid DNA using SuperFect reagent (Qiagen) or Fugene 6 (Roche Molecular Biochemicals) according to manufacturer's instruction. The transcription inhibition assay was done as described elsewhere (30). The treatment conditions were 5 µg/ml actinomycin D along with 20 µg/ml cycloheximide for 3 h.

Western Blot Analysis-- The same amount of cells or transfected cells (~1 × 10⁴) was suspended in 20 µl of Laemmli buffer and boiled for 5 min, run on 10% SDS-PAGE gels, and transferred to polyvinylidene difluoride membranes (Millipore). The primary antibody, rabbit anti-QKI-5 polyclonal antibody (22), was used at a 1:5000 dilution. The secondary antibody was horseradish peroxidase-conjugated donkey anti-rabbit IgG (Amersham Pharmacia Biotech). Blots were visualized with ECL (Amersham Pharmacia Biotech).

Fluorescence Microscopy-- Indirect immunofluorescence was used to show endogenous QKI-5, Surf-6 and transfected Flag-QKI-5, myc-tagged hnRNP A1, and hnRNP C1. Cells were fixed with 4% paraformaldehyde for 20 min and permeabilized with 0.5% Triton X-100 for 15 min. The fixed cells were then incubated with blocking solution containing 3% bovine serum albumin and 0.2% fish gelatin in phosphate-buffered saline for 15 min. After three washes with phosphate-buffered saline, they were probed with primary antibody diluted in blocking solution for 1 h at room temperature, 1:2000 for rabbit anti-QKI-5 polyclonal antibody, 1:100 for rabbit anti-mouse Surf-6 antibody (kindly provided by Dr. C. Magoulas, Imperial Cancer Research Fund, London, United Kingdom), 1:360 for mouse anti-Flag monoclonal antibody (Sigma) and 1:10 for mouse anti-myc monoclonal antibody (kindly provided by Dr. J. Fisher, University of Texas, Austin, TX). Biotinylated goat anti-rabbit IgG or goat anti-mouse IgG (Vector) was used as secondary antibody, and Texas red-conjugated streptavidin (Vector) was used to visualize the structure. Finally, cells were stained with 50 ng/ml DAPI or 5 µg/ml Hoechst 33258 for 15 min and observed on an Axioplan fluorescence microscope (Zeiss) with a 40× plan-NeoFluar objective. To observe GFP fusion proteins, cells were fixed with 4% paraformaldehyde 24 h after the transfection of GFP constructs, then stained with DAPI (50 ng/ml) or Hoechst 33258 (5 µg/ml).

Heterokaryon Assays-- Interspecies heterokaryons of HeLa and mouse NIH3T3 cells were formed as described elsewhere (31). Briefly, HeLa cells were transfected with GFP constructs, pCMV4-Flag-QKI-5, pcDNA3-myc-hnRNP A1, or pcDNA3-myc-C1 for 24 h, followed by adding equal amount of untransfected mouse NIH3T3 cells. The co-culture was incubated for 2.5 h in the presence of 75 µg/ml cycloheximide and 30 min in 100 µg/ml cycloheximide. After being fused with 50% polyethylene glycol 3350 for 2 min, cells were washed and incubated in medium containing 100 µg/ml cycloheximide for another 4 h, followed by fixation and observation described above.

Data Base Search-- The 30-amino acid QKI-5 tail, the last 23 amino acids of SAM68, and the last 23 amino acids of ETLE were used to search the current protein data base using BLASTP program (32). The setting of expectation was 1000, and no filter was used. All other parameters used are default settings. Among the hits, significant similarities were selected by eye.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The QUA1 Domain of QKI Is Responsible for Self-interaction-- QKI^kt4 was the first of the four QKI^enu mutations to have its molecular defect identified. It has a missense change from glutamic acid to glycine at position 48 in the QUA1 domain. However, the function of QUA1 domain and the molecular process that the mutation disrupts were unknown.

The study of XQUA and mouse QKI-7 have suggested that QKI can form homodimers and that RNA binding is not necessary for its self-association (5, 8). However, it is not clear which domain is responsible for oligomerization. We used the yeast two-hybrid assay to further study QKI protein interaction. QKI isoforms, deletion fragments, and QKI-5^kt4 (Fig. 1) were fused with either LexA DNA-binding domain or GAL4 activation domain, and then cotransformed into yeast in different combinations (Fig. 2A). The X-gal assay shows that all three QKI isoforms can interact with each other; in contrast, the QKI-5^kt4 mutation cannot interact with any of the QKI isoforms. pBTM116-Lamin and pACTII-T-antigen were used as negative control. QKI-5^kt4 was chosen for this analysis as opposed to QKI-6 or QKI-7 mutation because QKI-5 is the dominant form expressed in the embryo, and we were interested in the isoform prevalent in development.

View larger version (24K): [in this window] [in a new window]   Fig. 1.   Isoforms and mutations of QKI used in the study. QKI has a 206-amino acid STAR domain, which contains a single extended KH module flanked by QUA1 and QUA2 domains. Three isoforms of QKI have different carboxyl tails, which are 30 amino acids for QKI-5, 8 for QKI-6, and 14 for QKI-7. QKI-5kt4 has a Glu to Gly change in the QUA1 domain. QKI-body has all the 311-amino acid QKI common part, but not the tail. QKI-KH contains only the QUA1 domain and part of the KH domain (residue 1-134). QKI-5tail only has the 30-amino acid QKI-5 carboxyl tail (residue 312-341). QKI-5KH is QKI-5 with 41 amino acids deleted from the middle of the KH domain (residue 96-136). QKI-5 Y/F and QKI-5 RR/AA have amino acids changed in the tail of QKI-5. The figure is drawn to approximate scale.

View larger version (28K): [in this window] [in a new window]   Fig. 2.   Summary of QKI dimerization study. A, study of QKI dimerization using yeast two-hybrid assay. QKI isoforms and mutations were cloned into pBTM116 and pACTII to fuse with LexA DNA-binding domain and GAL4 activation domain. They were cotransformed into yeast strain L40 in different combinations, and X-gal assays were performed to measure interaction. pBTM116-Lamin and pACTII-T-antigen were used as negative control. Blue colonies are indicated by (+), and white by (). B, purification of 6×His-tagged QKI-5 protein (His-QKI-5). pET28QKI-5 was transformed into E. coli BL21 DE3 cells. Protein His-QKI-5 was produced and purified as indicated under "Experimental Procedures." Crude lysate and purified protein were run on SDS-PAGE gels and stained with Coomassie Blue. C, QKI-5kt4 does not form dimers with QKI-5. 35S-Labeled, in vitro translated QKI-5, QKI-5kt4, or luciferase were mixed with 0.2 µg of His-QKI-5 and immunoprecipitated with mouse anti-6×His antibody. The precipitate was run on an SDS-PAGE gel and visualized by fluorography. One-tenth of the input 35S-labeled proteins were run on the same gel. QKI-5 coprecipitates with His-QKI-5, while QKI-5kt4 and luciferase do not.

To confirm that QKI-5^kt4 proteins have lost their self-interaction ability, we performed in vitro coimmunoprecipitation experiments (Fig. 2C). His-QKI-5 was expressed in E. coli and purified using a nickel column (Fig. 2B). Equal amounts of in vitro translated ³⁵S-labeled QKI-5, QKI-5^kt4, or luciferase as a negative control were mixed with purified His-QKI-5 protein. Mouse anti-6×His monoclonal antibody was used to immunoprecipitate the mixture and the proteins were resolved on a SDS-PAGE gel. The co-immunoprecipitated ³⁵S-labeled proteins were visualized by fluorography. QKI-5 coprecipitates with His-QKI-5, while QKI-5^kt4 and luciferase do not. In vitro translated QKI-5 only was mixed with the antibody in the coimmunoprecipitation assay to show that it was specific (Fig. 2C, last lane).

These two experiments demonstrate that QKI isoforms can interact with each other, but QKI-5^kt4 has lost self-interaction ability; thus, the QUA1 domain probably is responsible for the self-interaction of QKI. To further test this hypothesis, one of the newly identified QKI isoforms, QKI-KH (21), which contains only the QUA1 domain and part of the KH domain (residues 1-134) (Fig. 1), was used in the yeast two-hybrid assay. The result showed that QKI-KH is sufficient to interact with the other QKI isoforms (Fig. 2A). These data together strongly suggest that the QUA1 domain contains self-interaction function. It is also noteworthy that this result confirms a contemporary study using QKI-7 (33).

The QKI-5 Carboxyl Tail Is Necessary and Sufficient for QKI-5 Nuclear Localization-- At the subcellular level, QKI-5 is localized predominantly in the nucleus, while QKI-6 and QKI-7 are mostly cytoplasmic (22). Since the only difference between the isoforms is the carboxyl tail, it must be responsible for the different subcellular localization.

To further analyze the subcellular localization of QKI, we used GFP. As a first step, we analyzed endogenous QKI-5 and GFP-QKI-5 fusion protein expression in cultured cells. Using a Western blot assay, we found that all three cell lines studied, HeLa, NIH3T3 and COS7, express endogenous QKI-5 (Fig. 3A). However, there is a relatively lower level in NIH3T3 cells. Because of the relative high transfection efficiency, COS7 cells were used to check the GFP-QKI-5 protein expression. pEGFP-QKI-5 construct was transfected into COS7 cells and the expressed protein migrated at approximately 70 kDa as expected (Fig. 3A, first lane). In addition, Fig. 3B shows that GFP-QKI-5 has the same localization pattern as the endogenous protein. In HeLa cells, indirect immunofluorescence staining with anti-QKI-5 antibody reveals that the endogenous protein and GFP-QKI-5 are both nuclear but were excluded from nucleoli. GFP alone is evenly distributed throughout the cell. GFP is thus a suitable tag to study QKI localization. In NIH3T3 and COS7 cells, GFP-QKI-5 has an identical localization pattern as in HeLa cells (data not shown). Thus, HeLa cells were used in the further localization study.

View larger version (28K): [in this window] [in a new window]   Fig. 3.   Expression and localization of QKI-5 in tissue culture cells. A, Western blot assay of QKI-5 in cultured cells. Equal amount of cell extracts were run on SDS-PAGE gels and immunoblotted with rabbit anti-QKI-5 antibody. Endogenous QKI-5 runs at 38 kDa; in transfected cells, GFP-QKI-5 runs at ~70 kDa. The bands at ~60 kDa are nonspecific. B, localization of QKI-5 in HeLa cells. GFP-QKI-5 has the same subcellular localization as endogenous QKI-5. Indirect immunostaining using anti-QKI-5 antibody was used to show the endogenous QKI-5 in HeLa cells. For GFP constructs, HeLa cells were transfected with pEGFP-QKI-5 or pEGFP-C1 as control. The transfected cells were fixed with paraformaldehyde, stained with DAPI, and observed.

Using GFP, the effect of the QKI-5 tail was visualized. The 30-amino acid tail was deleted and the remainder QKI-body was fused to GFP. This pEGFP-QKI-body was then transfected into HeLa cells. Without the tail, the fusion protein is located predominantly in the cytoplasm (Fig. 4C), similar to GFP-QKI-7 (Fig. 4B). These experiments show that the carboxyl tail of QKI-5 is necessary for its nuclear localization.

View larger version (38K): [in this window] [in a new window]   Fig. 4.   Nuclear localization of QKI-5. QKI isoforms and mutations were fused to the COOH terminus of GFP, and the constructs were transfected into HeLa cells. 24 h after transfection, cells were fixed and observed. The structure of constructs A-D, F, and G are shown in Fig. 1, while sequences of E, H, and I are shown in Table I. QKI-5 is localized in nucleoplasm, but not nucleoli (A). QKI-7 is located mostly in the cytoplasm (B), and so is QKI-body, which is QKI-5 without the 30-amino acid tail (C). The 30-amino acid carboxyl tail of QKI-5 can target all the GFP into the nucleus, but there are some concentrated dots (D). The 7-amino acid conserved NLS of QKI-5, guides most of the GFP into the nucleus (E). When the Tyr in the 7 amino acids was changed to Phe in QKI-5, most of the protein is still in the nucleus, while some patches can be seen in the cytoplasm (F). When 2 Arg were changed to Ala, the nuclear localization of QKI-5 is abolished (G), which has the same effect as deleting the whole 30-amino acid tail. The SAM68 and ETLE 7-amino acid NLS have similar but not identical localizations as the QKI-5-NLS. In some cells, they target all the GFP to the nucleus, in about 80% of the cells, there is some GFP in the cytoplasm as well as the nucleus (H and I).

The next question was whether this tail alone could target the protein to the nucleus. When it was attached to the COOH terminus of GFP and transfected into HeLa cells, all the GFP is visualized in the nucleus (Fig. 4D). Thus, the 30-amino acid QKI-5tail alone is also sufficient for its nuclear localization. Interestingly, in this experiment, it was evident that there are some concentrated dots within the nucleus that were later shown to be nucleoli (see below for discussion).

QKI-5 Has a Novel Nuclear Localization Sequence (NLS)-- Although the QKI-5 tail is both necessary and sufficient for its nuclear localization, it does not contain a traditional cluster of basic amino acids or bipartite NLS (34). Upon close examination and comparison of mouse and Drosophila QKI, we found a sequence of 7 conserved amino acids (Table I): RVHPYQR in mouse and REHPYQR in Drosophila. In addition, very similar sequences were located at the extreme end of SAM68 and ETLE/T-STAR. Previous work has shown that the last 23 amino acids of SAM68 containing these 7 amino acids are responsible for its nuclear localization (35). In addition, in T-STAR, disruption of the last 5 amino acids, which overlaps with these 7, causes the protein to relocate to the cytoplasm (18). We hypothesized that the 7-amino acid consensus RXhPYQ/GR is the NLS for these STAR proteins. To examine this, we made a construct which has only these 7 amino acids from QKI-5 (QKI-5-NLS) fused to GFP and transfected it into HeLa cells. Although the fusion protein is expected to be less than 30 kDa, almost all the GFP-QKI-5-NLS was detected in the nucleus (Fig. 4E). In most of the cells, GFP-QKI-5-NLS is also concentrated in dots in the nucleus although they are not as strong as those in pEGFP-QKI-5tail transfected cells. Thus, these 7 amino acids constitute the nuclear localization signal of QKI-5.

Next, we investigated which amino acids among the 7 are important for the NLS function. Since SAM68 is heavily phosphorylated by Src and both QKI-5 and SAM68 are rich in tyrosine, we thought the phosphorylation of the tyrosine residue might affect the nuclear localization. However, when the tyrosine was mutated to phenylalanine in QKI-5 (Fig. 1) and the pEGFP-QKI-5Y/F was transfected into HeLa cells, the expressed protein was still located mostly in the nucleus, with only a small amount in the cytoplasm (Fig. 4F). So if there is phosphorylation of this tyrosine residue, it is not necessary for the nuclear localization. Considering that basic residues play an important role in the classical NLS, it seemed that the 2 arginines at both ends might be important. When these residues were both changed to alanines, the nuclear localization of GFP-QKI-5 was abolished (Fig. 4G). Thus, these 2 arginines are necessary for nuclear localization and the 7 amino acids represent a novel nuclear localization signal.

Using the same method, we tested whether the slightly different sequences in SAM68 and ETLE are also capable of directing the protein to the nucleus. Notably, the results are somewhat different. In about 20% of the transfected HeLa cells, GFP-SAM68-NLS and GFP-ETLE-NLS are completely in the nucleus; while in the remaining 80% cells, although GFP proteins are concentrated in the nucleus, there is also a minor amount of fusion protein in the cytoplasm (Fig. 4, H and I). Possibly this is because these two NLSs are not as strong as the QKI-5-NLS or the transport of these two NLS is cell cycle related. Nevertheless, since most of the cells have GFP proteins more concentrated in the nucleus than in the cytoplasm, it suggests that the 7 amino acid signals in SAM68 or ETLE have NLS function. Considering all the data, we termed this consensus sequence (RXhPYQ/GR) STAR-NLS.

The Requirements for Normal QKI-5 Nuclear Localization-- Both GFP-QKI-5 and GFP-QKI-5tail are expressed in the nucleus. However, GFP-QKI-5 was excluded from the nucleoli, while GFP-QKI-5tail locates throughout the nucleus with some concentrated dots that appear to be nucleoli when compared with a phase-contrast image (data not shown). To make sure that these dots are nucleoli, mouse nucleolar matrix protein Surf-6 was used as a marker (36). Thus, mouse NIH3T3 instead of HeLa cells were used here. pEGFP-QKI-5tail was transfected into mouse NIH3T3 cells that were then immunostained with anti-Surf-6 antibody. In Fig. 5A, the localization of GFP-QKI-5tail and endogenous Surf-6 in the same NIH3T3 cell is shown. Superimposition of the two shows that QKI-5tail colocalizes with Surf-6 in the nucleoli (Fig. 5A, right panel).

View larger version (29K): [in this window] [in a new window]   Fig. 5.   Requirement for normal QKI-5 localization. A, in NIH3T3 cells, the QKI-5tail colocalizes with nucleolar matrix protein Surf-6. Rabbit anti-mouse Surf-6 antibody was used to show the nucleoli (middle panel), which colocalize with the GFP-QKI-5tail (left panel) as shown by the yellow in the superimposition of the same cell (right panel). B, the effect of self-interaction and RNA binding on the nuclear localization of QKI-5. GFP-QKI-5, GFP-QKI-5kt4, and GFP-QKI-5KH were transiently expressed in transfected HeLa cells. The localization of GFP-QKI-5kt4, which cannot form homodimers, is the same as normal QKI-5. GFP-QKI-5KH has punctate nuclear structure, but is still excluded from nucleoli. C, the effect of transcription inhibition on QKI-5 localization. pcDNA3-myc-hnRNP A1 and pcDNA3-myc-C1 were transfected into HeLa cells. After 24 h, untransfected and transfected HeLa cells were treated with actinomycin D in the presence of cycloheximide for 3 h. Rabbit anti QKI-5 antibody were used to stain untransfected cells, and endogenous QKI-5 proteins were found in both the nucleus and the cytoplasm (left panels). myc-hnRNP A1 and myc-hnRNP C1 were stained with mouse monoclonal anti-myc antibody. hnRNP A1 accumulated in the cytoplasm after the inhibition of transcription (middle panels), while hnRNP C1 stays in the nucleus as expected (right panels).

Next, we analyzed what could be responsible for the normal nuclear localization of QKI-5, and what causes the difference between it and the localization of QKI-5tail. There are several possibilities. First, GFP-QKI-5tail could interact with some components in nucleoli, but the interaction is masked when the whole protein is used; second, some other proteins in the nucleoplasm may interact with QKI-5 and prevent it from localizing in nucleoli; and third, QKI-5 binds pre-mRNA in the nucleoplasm and the RNA binding may be required for its exclusion from the nucleoli. To test these possibilities, we examined whether self-interaction and RNA binding are important for normal QKI-5 localization.

The QKI-5^kt4 mutation loses the ability to self-interact; however, it retains the ability to bind RNA (45). When GFP was fused with QKI-5^kt4 and transfected into HeLa cells, the localization of GFP-QKI-5^kt4 is the same as wild type QKI-5 (Fig. 5B). Thus the self-interaction is not required for the normal QKI-5 localization.

To determine the importance of RNA binding, we deleted the central 41 amino acids (residue 96-136) of the KH domain (Fig. 1) and fused the remainder to GFP (GFP-QKI-5KH). According to the previous study with XQUA, the RNA-binding ability of this protein would be largely abolished (5, 37). When expressed in transfected HeLa cells, GFP-QKI-5KH is still nuclear and excluded from the nucleoli. Interestingly, in most of the transfected cells, there are a varying number of punctate dots (Fig. 5B). This pattern was very similar to that of KH SAM68 in transfected cells (38). Thus, RNA binding is important for the normal diffuse localization of QKI-5 in the nucleoplasm, but it does not exclude it from the nucleoli. The reason for the accumulation of GFP-QKI-5tail in the nucleoli remains unknown. Alternatively, it is possible that the 30-amino acid tail or the 7-amino acid NLS could interact with some component in nucleoli nonspecifically. The small peptide is highly positively charged, and thus it may interact with negatively charged rRNA or other components in the nucleolus.

Another factor important for the localization of some RNA-binding proteins is active transcription by RNA polymerase II. It is known that nuclear transport of some RNA-binding proteins such as hnRNP A1 is transcription-dependent, while for others such as hnRNP C1, it is transcription-independent (30). Thus, we tested whether transcription affects the nuclear localization of QKI-5. Transcription inhibitor actinomycin D was added to untransfected and pcDNA3-myc-hnRNP A1 or pcDNA3-myc-hnRNP C1 (27) transfected HeLa cells. Three hours later, hnRNP A1 accumulated in the cytoplasm, while hnRNP C1 was still nuclear. Interestingly, endogenous QKI-5 was also partly trapped in the cytoplasm (Fig. 5C). This result suggested that the nuclear transport of QKI-5 is transcription-dependent.

QKI-5, but Not ETLE, Shuttles between the Nucleus and the Cytoplasm-- At steady state, QKI-5 is predominantly nuclear. However, its transcription-dependent nuclear transport suggested its shuttling ability (39). Notable in this context is that, during the normal embryonic neuronal cell fate decision, QKI-5 is redistributed from the nucleus to the cytoplasm in some cells (25). This suggests that QKI-5 may also function in the cytoplasm and shuttle between the nucleus and the cytoplasm. Some RNA-binding proteins, such as a subset of hnRNP and SR proteins, shuttle rapidly between the nucleus and the cytoplasm (30, 40). To test whether QKI-5 can shuttle, we performed an interspecies heterokaryon assay (31). HeLa cells were transfected with pEGFP-QKI-5, and 24 h later, an equal amount of NIH3T3 cells were fused to them to form heterokaryons. When new protein synthesis was inhibited by cycloheximide, and Hoechst 33258 was used to distinguish the mouse and human nuclei, GFP fusion protein was observed in the NIH3T3 cell nuclei 4 h after fusion (Fig. 6A). This indicated that in the heterokaryons, GFP-QKI-5 must be exported from the HeLa nucleus to the cytoplasm and then imported into the NIH3T3 cell nucleus. Thus, QKI-5 shuttles between the nucleus and the cytoplasm. At the same time and with the same assay, the shuttling ability of another STAR protein ETLE was tested. In contrast to QKI-5, GFP-ETLE was not seen in the mouse NIH3T3 nucleus in human-mouse heterokaryons (Fig. 6C). Thus, ETLE does not shuttle between the nucleus and the cytoplasm, or at least not as rapidly as QKI-5. The shuttling protein hnRNP A1 was used as a positive control and the nonshuttling protein hnRNP C1 as a negative control. pcDNA3-myc-hnRNP A1 and pcDNA3-myc-hnRNP C1 were transfected into HeLa cells, and both of the fusion proteins behaved appropriately in the heterokaryon assay (Fig. 6, D and E).

View larger version (64K): [in this window] [in a new window]   Fig. 6.   QKI-5, but not ETLE, shuttles between the nucleus and the cytoplasm. pEGFP-QKI-5, pCMV4-FLAG-QKI-5, pEGFP-ETLE, pcDNA3-myc-hnRNP A1, and pcDNA3-myc-hnRNP C1 were transiently transfected into HeLa cells. 24 h after the transfection, they were mixed with mouse NIH3T3 cells in the presence of the translation inhibitor cycloheximide. Then, cells were fused to form heterokaryons. Four hours after fusion, cells were fixed and stained with anti-FLAG antibody for FLAG-QKI-5 (B) or anti-myc antibody for myc-hnRNP A1 (D) and myc-hnRNP C1 (E). GFP-QKI-5 (A) and GFP-ETLE (C) were observed directly under fluorescent microscope. The transfected protein localizations are shown in left panels. Hoechst 33258 was used to distinguish mouse and human cell nuclei (middle panels). Phase-contrast images of the heterokaryons are shown (right panels). The arrows indicate the mouse nuclei within human-mouse heterokaryons.

Although GFP is widely used to study protein localization, there are not many studies using it in the sensitive heterokaryon assay. At this point, it is difficult to rule out the effect of the 27-kDa GFP component in the shuttling assay. Thus, to corroborate the shuttling ability of QKI-5, another smaller tag, Flag, was used to study QKI-5 shuttling with the same assay. Immunostaining with mouse monoclonal antibody against Flag verified that Flag-QKI-5 appeared in the NIH3T3 nucleus and confirmed that QKI-5 shuttles (Fig. 6B).

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

We analyzed the structure of QKI-5, and found that the QUA1 domain is likely to be the self-interaction domain responsible for homo- and hetero-dimerization. Thus, the lethal phenotype of qkI^kt4 could be caused by a failure of QKI-5^kt4 to interact with itself or other proteins. Recently, similar results were reported for QKI-7 (33).

The studies also show that the QKI-5 carboxyl tail contains a novel nuclear localization sequence STAR-NLS and that normal QKI-5 nuclear localization requires active transcription and the capacity to bind RNA, but not self-interaction. Furthermore, we showed that QKI-5, but not the SAM68-like protein, ETLE, shuttles between the nucleus and the cytoplasm.

STAR-NLS Is a Novel Nuclear Localization Sequence-- By deletion analysis, we showed that the 30-amino acid QKI-5 carboxyl tail is necessary and sufficient for QKI-5 nuclear localization. In this fragment, a 7-amino acid novel nuclear localization sequence (RVHPYQR) was identified. It is conserved in some other STAR proteins, such as SAM68 and ETLE/T-STAR, and thus we named the consensus sequence (RXhPYQ/GR) STAR-NLS. When only this peptide was fused with GFP, it can drive the reporter protein into the nucleus.

It is of interest that some non-STAR proteins also contain this STAR-NLS. We searched the current protein data base to identify other proteins that may have a similar signal (Table I). Among those identified were: Drosophila transcription factor HNF-4 homolog, Arabidopsis splicing factor RSZp21, a hypothetical protein CEF08G2.5 from the C. elegans sequencing project, and a nuclear polyhedrosis virus protein. From their names, it would seem that at least three out of the four are nuclear. These four proteins were examined for other obvious classic NLS, but none were found. Thus, they probably use this motif for nuclear localization. Further analysis of these proteins may reveal whether the STAR-NLS is used outside of the STAR family.

STAR-NLS is different from not only the classical basic cluster NLS and bipartite NLS, but also the newly defined shuttling sequence M9 of hnRNP A1 or KNS of hnRNP K (31, 41). M9 and KNS are responsible for both nuclear import and export and these two contrasting activities appear to be inseparable (31). An important question is which pathway is used to import STAR-NLS containing proteins. Import of the classical NLS is mediated by the importin / complex. M9 interacts with transportin, an importin  like protein, and uses nuclear import machinery different from the traditional importin / (42), while the nuclear import of KNS may require a third pathway (41). The receptor for STAR-NLS is not known. Currently, we are trying to identify proteins that interact with the QKI-5 carboxyl tail. However, it is also possible that STAR-NLS may not bind to a nuclear transporting receptor directly, but interact with a different protein with an authentic NLS.

The Significance of QKI-5 Shuttling-- The finding that QKI-5 shuttles between the nucleus and the cytoplasm expands our understanding of the possible function of QKI. Of the three well studied isoforms, QKI-5 is predominantly nuclear, while QKI-6 and QKI-7 are mostly cytoplasmic. The simplest interpretation suggests that QKI-5 functions in the nucleus, while the others function mostly in the cytoplasm. The shuttling of QKI-5 makes the scenario more complex; it means that QKI-5 might also function in the cytoplasm. Furthermore, QKI isoforms can associate with each other. Their heterodimerization in either the nucleus or the cytoplasm would let QKI fine tune its function. An important role for shuttling proteins is transporting RNA, and in this sense, the QKI isoforms could function as a relay team. QKI-5 might escort a specific group of RNAs out of the nucleus, while QKI-6 and QKI-7 would continue to interact with them in the cytoplasm. QKI-5 could then be recycled back to the nucleus. Similar to a group of RNA binding proteins like hnRNP A1, the reimport of QKI-5 into the nucleus is transcription dependent. A second consequence of shuttling might be carrying cell signals from the cytoplasm into the nucleus.

Interestingly, another STAR protein ETLE does not shuttle, or its shuttling is so slow that it is not detectable in our assay. This is surprising because QKI-5 and ETLE use the same STAR-NLS. This difference in their ability to shuttle suggests that there are other sequence elements besides STAR-NLS involved. SAM68 is very close in sequence to ETLE; therefore, it will be interesting to determine whether SAM68 and other STAR proteins also shuttle.

For QKI-5, it will also be important to determine what is responsible for its shuttling/export. There are two leucine-rich regions similar to the classical nuclear export sequence in the QUA1 and QUA2 domains, (34). One or both of them could be responsible for the export of QKI-5, but it is also possible that QKI-5 could use another, unidentified export sequence. The balance between export and import signals may be responsible for changes in the localization of QKI-5 at different developmental stages, such as, has been found in fetal mouse brain (25).

Speculations about QKI Function-- STAR proteins use the same RNA-binding motif, but they have different localizations and diverse functions. In addition, most of the STAR proteins are extensively alternatively spliced, generating even more variation. This variation could permit a diversity of molecular functions. For example, SF1/ZFM1 is reported to bind the branch point in the introns of pre-mRNA and function as a pre-splicing factor (7, 14). However, it can also interact with transcription activator SSAP and repress transcription (43). Another family member from C. elegans, GLD-1, which is solely cytoplasmic, has recently been shown to repress translation of specific mRNAs by binding to the tra-2 and GLI elements (TGE) in their 3'-untranslated region (44). Likewise, QKI-6 has been found to have the same function in transgenic C. elegans (45). Finally, SAM68 has been recently shown to substitute for HIV-1 Rev protein in facilitating the nuclear export of HIV-1 RNA (46).

With respect to QKI, the mutations make it clear that it is important in myelination and early development. However, the role of QKI in RNA metabolism and, if any, in signal transduction, is not clear. Relevant are the studies that show, in qk^v mutant mice, some gene products are altered at the RNA level (47); myelin-associated glycoprotein has a changed ratio of alternative splice products (48), and the myelin basic protein mRNA level is low in mouse brain at the peak of myelination but becomes normal later (49). QKI is most likely responsible for these alterations.

	ACKNOWLEDGEMENTS

We thank Drs. Gideon Dreyfuss for hnRNP A1 and hnRNP C1 constructs, C. Magoulas for anti-Surf-6 antibody; S. M. Hollenberg for yeast two-hybrid vectors and library, and J. Fisher for anti-myc antibody. We are also grateful to Drs. Clarence Chan and Gabriela Rennebeck for reading the manuscript.

	FOOTNOTES

^* This work is supported by Grants HD10668 and HD30658 from the National Institutes of Health (to K. A.).The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Present address: School of Biomolecular and Biomedical Science, Faculty of Science, Griffith University, Nathan, Queensland 4111, Australia.

^§ Present address: University of Texas Medical School, Houston, TX 77030.

^¶ To whom correspondence should be addressed: Inst. for Cellular and Molecular Biology and Dept. of Microbiology, 2500 Speedway, University of Texas, Austin, TX 78712-1064. Tel.: 512-471-1785; Fax: 512-471-2149; E-mail: artzt@uts.cc.utexas.edu.

² T. Kaname, T. Kondo, T. Suganuma, M. Suzuki, T. Ebersole, K. Artzt, K. Yamamura, and K. Abe., submitted for publication.

	ABBREVIATIONS

The abbreviations used are: QKI, quaking; hnRNP, heteronuclear ribonucleoprotein; KH, hnRNP K homologue; STAR, signal transduction and activator of RNA; NLS, nuclear localization sequence; ETLE, étoile protein; GLD-1, germ line-deficient; SF1, splicing factor 1; GFP, green fluorescent protein; PCR, polymerase chain reaction; PAGE, polyacrylamide gel electrophoresis; HIV, human immunodeficiency virus; X-gal, 5-bromo-4-chloro-3-indolyl -D-galactopyranoside; DAPI, 4,6- diamidino-2-phenylindole.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES