RNA-binding Motif Protein 15 Binds to the RNA Transport Element RTE and Provides a Direct Link to the NXF1 Export Pathway*

Retroviruses/retroelements provide tools enabling the identification and dissection of basic steps for post-transcriptional regulation of cellular mRNAs. The RNA transport element (RTE) identified in mouse retrotransposons is functionally equivalent to constitutive transport element of Type D retroviruses, yet does not bind directly to the mRNA export receptor NXF1. Here, we report that the RNA-binding motif protein 15 (RBM15) recognizes RTE directly and specifically in vitro and stimulates export and expression of RTE-containing reporter mRNAs in vivo. Tethering of RBM15 to a reporter mRNA showed that RBM15 acts by promoting mRNA export from the nucleus. We also found that RBM15 binds to NXF1 and the two proteins cooperate in stimulating RTE-mediated mRNA export and expression. Thus, RBM15 is a novel mRNA export factor and is part of the NXF1 pathway. We propose that RTE evolved as a high affinity RBM15 ligand to provide a splicing-independent link to NXF1, thereby ensuring efficient nuclear export and expression of retrotransposon transcripts.

General mRNA export in eukaryotes is mediated by NXF1 protein orthologues that are conserved from yeast to humans and bind to the export-ready mRNP, targeting them to the nuclear pore complex (NPC) 2 (1)(2)(3)(4)(5)(6). NXF1 acts as part of a stable heterodimer with its cofactor p15/NXT1 (7)(8)(9). Splicing changes the mRNP protein composition, allowing NXF1-p15 to bind and export to occur, whereas the pre-mRNPs are normally retained in the nucleus until completely spliced (10 -13). In particular, a set of proteins known as exon junction complex (EJC) is deposited onto mRNP as a result of splicing (14), providing critical determinants for the subsequent metabolic steps, including nuclear export, quality control, cytoplasmic trafficking, and translation (15). EJC consists of a stably bound core composed of eIF4AIII, Y14-Magoh, and MLN51/Barentsz that serves as a platform for a multitude of other EJC and EJC-associated factors that are bound more transiently. EJC is thought to commit the spliced mRNPs to nuclear export by providing binding sites for the NXF1-p15 export receptor. In one scenario, the EJC factor UAP56 recruits Aly/REF proteins, which bind directly to NXF1-p15, which in turn tethers the export substrate to the NPC (16 -21). Alternatively, NXF1 may assemble with the spliced mRNP via interactions with non-EJC factors such as SR proteins: SRp20 and 9G8 (22,23), ASF/SF2 (24), and U2AF (25). Thus, it appears that several pathways lead to the binding of NXF1-p15 with the export-ready mRNP. Upon NXF1-p15-dependent targeting to NPC, such complexes are translocated to the cytoplasm by a yet unknown mechanism. NXF1 is a conserved export receptor for cellular mRNAs (1)(2)(3)(4)(5)(6). Proteins of the NXF family can act on nuclear as well as on cytoplasmic mRNA trafficking (26 -28).
According to the current model, general mRNA metabolism requires the acquisition of an export signal as a result of splicing, whereas pre-mRNA is generally retained in the nucleus due both to the lack of active export and to factors retaining the pre-mRNA in the nucleus. Retroviral transcripts are a notable exception from this rule, because the unspliced transcript encodes the Gag-pol polyprotein and also serves as viral genomic RNAs, and, therefore needs to be exported prior to splicing. To overcome the general requirement of splicing before export, simian Type D retroviruses and some retroelements utilize the constitutive transport element (CTE) (29 -33) that serves as a high affinity NXF1 ligand (3), thereby providing constitutive, splicing-independent export signals.
We had discovered and characterized a novel family of cis-acting RNA transport elements (RTEs) that are present abundantly in the mouse genome and are associated with intracisternal A-particle retroelements (IAP) (34,35). RTE is functionally analogous to CTE, yet structurally unrelated, and does not bind to NXF1 with high affinity. In this work, we report the identification of the factor that directly promotes RTE function.
We report that the RNA-binding motif protein 15 (RBM15) recognizes RTE RNA specifically in vitro and activates export and expression of RTE-containing reporter mRNAs in vivo. RBM15 also binds to NXF1, and these factors act cooperatively in promoting RTE-mediated expression. Thus, RBM15 is a novel component of the NXF pathway.
Isolation of Proteins Binding to RTE RNA-Biotinylated RTE and CTE RNA bound on streptavidin beads (Dynal) were used as pulldown bait (50). The immobilized RNAs were incubated with micrococcal endonuclease-treated nuclear HeLa cell extract in buffer RBB (15 mM HEPES, pH 7.9, 50 mM KCl, 0.1 mM EDTA, and 0.2% Triton X-100) with 300 mM NaCl (RBB-300) supplemented with 4 g/ml rRNA and 2 g/l tRNA in 400-l reactions at 30°C for 2 h. The beads were washed 6ϫ in RBB-300, and the bound proteins were eluted from the RNAs in 1 M NaCl, separated on a 10% SDS-PAGE, and visualized by silver staining.
In-gel Tryptic Digestion, Nano-HPLC Separation, and NSI-MS Analysis-Protein bands were excised from gels and digested with trypsin as described (51,52). The peptides were separated by nano-HPLC (LC Packings), introduced through a nanospray ionization source to an ion-trap mass spectrometer (LCQ-Deca, ThermoFinnigan) and analyzed by tandem mass spectrometry, as described (51,52). Default score values used as cut-off parameters during TurboSequest searches were as follows, X corr Ͼ 1.0, ⌬C n Ͼ 0.1, S p Ͼ 500, and R sp Ͻ 10, and the peptide mass tolerance was 1.0.
RNA Transcription and Gel-mobility Shift-The 32 P-labeled RTE and CTE RNAs (34) were prepared as in a previous study (53), and unlabeled RNAs were synthesized using MEGAScript-T7 (Ambion, Austin, TX). 10 nM cold RTE or CTE RNA was added to 10 fmol of radiolabeled RTE or CTE RNA, respectively, to keep the RNA concentration constant for the different probes. Binding reactions were performed in 10 l of RBB-250 in the presence of 2 g of tRNA. After 30 min at room temperature, 1 l of a solution containing 0.2 mg/ml heparin and 0.05% bromphenol blue was added to the reactions, following 10 min at room temperature. Samples were loaded onto a 6% (19:1 acrylamide:bisacrylamide ratio) non-denaturing polyacrylamide gel containing 50 mM Tris-Borate, pH 8.8, 0.5 mM FIGURE 1. RBM15 binds directly to RTE RNA. A, identification of RTE-binding proteins. RTE and CTE RNA bound to streptavidin-beads were incubated in vitro with nuclear HeLa extract. Proteins binding stronger to RTE RNA than CTE RNA were sequenced by nanospray mass spectrometry. The filled arrow indicates RBM15; the open arrows indicate nonspecific interactors. The identified peptides are shown for each protein. B, recombinant RBM15 binds RTE RNA in vitro. Radiolabeled RTE or CTE RNA was used in gel-mobility shift assays with recombinant GST-RBM15. The radioactive RTE and CTE RNAs were adjusted to a final concentration of 10 nM, and GST-RBM15 was used at concentrations of 1-50 nM as indicated. The positions of probes and the RNA⅐RBM15 complexes are indicated.
EDTA. Electrophoresis was carried out at 4°C, and complexes were visualized by autoradiography.
Cell Culture, Transfection, and Microscopy-Human 293 and 293T cells were transfected using calcium phosphate technique. Human HeLa-derived HLtat cells were transfected with SuperFect (Qiagen). For indirect immunofluorescence, the fixed cells (54) were incubated with murine anti-HA antibody (Covance) at a 1:1000 dilution in phosphate-buffered saline in the presence of 0.2% bovine serum albumin, followed by Alexa 594-conjugated anti-mouse IgG (Molecular Probes, Eugene, OR). For the shuttling assay, transfected HeLa cells were mixed with an excess of untransfected cells, and, after pretreatment with cycloheximide, the cells were fused using polyethylene glycol (54). For cotransfection experiments, reporter plasmids were used at 1 g, in the absence or presence of 0.5 g of plasmids expressing export factors. All transfections contained 0.2 g of GFP (pFRED143 or pFRED25) or SEAP expression plasmids serving as internal control for efficiency of transfection and poly-A RNA preparation. Two days later, Gag (HIV p24gag antigen capture assay, Zeptometrix), SEAP (Phospha-light kit, Applied Biosystems), CAT, and GFP fluorescence (54) were measured. Nuclear and cytoplasmic mRNA was prepared (25). The RNAs were analyzed on Northern blots (42) using probes spanning the GFP or CAT coding regions.
For RNA interference experiments, 2 ϫ 10 5 HeLa cells were transfected with 10 mM SMART pool siRNA (Dharmacon) targeted to RBM15 (GGACAGAGGTGATCGAGAT, GAAGAT-AGAAGCTGTGTAT, GGACACCACCCTTACTATA, and GGTGATAGTTGGGCATATA) or non-targeting siRNA control (Dharmacon) using HiPerFect (Qiagen). One day later, the cells were re-transfected with the reporter plasmids using SuperFect (Qiagen). Culture media were collected 48 h later, and Gag and SEAP were measured. To control for the efficiency of RBM15 knockdowns, the cells were transfected with an RBM15-FLAG-expressing plasmid on day 1 and then transfected with either RBM15 siRNA pools or siRNA control pool on day 2. Cells were harvested in RBB-400 buffer on day 4, and the lysates were analyzed on immunoblots using murine anti-FLAG antibody (M2, Sigma) and horseradish peroxidase-conjugated anti-murine antibodies (Amersham Biosciences). RNA interference pool-treated untransfected cells were analyzed using anti-RBM15 antiserum (ProteinTech Group, Inc., Chicago, IL) and anti-␤ actin antibody (Sigma). The proteins were visualized by enhanced chemiluminescence (ECL plus Western blotting Detection System, Amersham Biosciences) and autography. Reverse transcription-PCR of the endogenously expressed alternatively spliced RBM15 mRNAs was performed on total poly-A containing mRNAs isolated from HeLa and 293 cells using the Titan One Tube RT-PCR kit (Roche Applied Science), and the PCR products were sequenced.
Recombinant Protein Expression and Protein Analysis-Human RBM15 was produced in Escherichia coli BL21(DE3)pLysS (Novagen) from pGEX-6P-3-RBM15. Recombinant soluble GST-RBM15 protein was isolated after freezing the bacterial pellets in phosphate-buffered saline supplemented with 160 mM NaCl and protease inhibitor (Pi, Roche Applied Science) at Ϫ70°C for 30 min. The lysates were treated with DNase I (Roche Applied Science) and cleared by centrifugation. Glutathione-agarose beads (Roche Applied Science) were added, and the mixture was rotated for 1 h at 4°C. The beads were washed three times, and the recombinant protein was eluted using a standard glutathione-containing buffer. Protein concentrations were estimated on Coomassie Blue-stained SDS-polyacrylamide gels.
RNA Export from Xenopus Oocyte Nuclei-The preparation of capped RTE and CTE-containing adenovirus precursor RNA, U1⌬Sm RNA, and U6⌬ss RNA, unlabeled RTE, CTE, and CTEm36 RNAs and oocyte nuclei microinjections were described previously (34,55,56). RNA was extracted from a pool of five oocytes after proteinase K digestion, and equivalents of one-half oocyte were analyzed on 10% polyacrylamide gels containing 7 M urea.
In Vitro Protein Binding Assays-Metabolically labeled reticulocyte-produced proteins were synthesized in a coupled transcription/translation system (TNT T7 Coupled Reticulocyte Lysate System, Promega) and used in binding reactions con- taining ϳ0.5 g of E. coli-produced GST-tagged proteins (27). The binding was performed in 200 l of RBB-400 buffer in the presence of 100 g of RNase A. Following incubation for 15 min at room temperature, the beads were washed three times with binding buffer. The bound proteins were eluted by boiling in sample buffer, separated by SDS-PAGE, and detected by autoradiography.
Immunoprecipitation Assays-Complexes of epitope-tagged proteins were immunopurified from transiently transfected 293 cells. Typically, ϳ3 ϫ 10 6 cells were extracted with 500 l of RBB-400 buffer. The extracts were cleared by centrifugation at 10,000 ϫ g for 15 min at 4°C. Immunoprecipitations were performed at 4°C for 30 min in 200 l of RBB-400 buffer in the absence or presence of RNase A using anti-FLAG M2-agarose (Sigma). The precipitates were washed six times in RBB-400 buffer supplemented with 2 M urea and analyzed on Western immunoblots using horseradish peroxidase-conjugated HA-antibodies (Roche Applied Science), and the proteins were visualized by enhanced chemiluminescence (ECL plus Western blotting Detection System, Amersham Biosciences) and autography.

Identification of RBM15 as an RTE Binding
Factor-Immobilized, biotinylated RTE RNA was used to identify putative binding factors from HeLa nuclear extracts in pull-down experiments. CTE RNA was included as control in a parallel experiment (Fig. 1A). These binding conditions enabled the specific pull-down of NXF1 with CTE RNA but not with inactive mutant CTEm36 RNA, lacking the NXF1 binding sites (3,30), confirming the binding specificity (data not shown). Four candidate RTE-specific binding factors were identified by microsequencing ( Fig. 1A), which are RNA helicase A (SwissProt accession number: O70133), RBM15 (Q96T37), heterogeneous nuclear ribonucleoprotein G (P38159), and U1 small nuclear ribonucleoprotein A (P09012). An additional band at ϳ100 kDa could not be identified because of contamination with bovine serum albumin and was not further studied. Subsequent experiments did not support specific interactions of RNA helicase A, heterogeneous nuclear ribonucleoprotein G, and U1 small nuclear ribonucleoprotein A with RTE RNA (data not shown). In contrast, RBM15 showed specific and functional interaction with RTE as detailed below. RBM15 belongs to the spen (split end) protein family and is conserved in eukaryotes from Caenorhabditis elegans to humans (57), but its function had not been investigated previously. Characteristic of this gene family is the presence of three conserved RNA recognition motifs at the N terminus and the SPOC (spen paralogue and orthologue C-terminal) domain at the C terminus.
Preferential Binding of RBM15 to RTE in Vitro-We expressed recombinant RBM15 (isoform AEϩS, see Fig. 2A) in bacteria and employed electrophoresis mobility shift assays to examine binding of RBM15 to RTE RNA (Fig. 1B). Radiolabeled RTE (left panel) or CTE (right panel) RNAs were incubated with increasing amounts of bacterially produced GST-tagged RBM15. Approximately 50% binding of RTE RNA was observed at 15 nM of RBM15 (left panel, lane 6), whereas no complex formation with CTE RNA was detectable (right panel, lanes 9 -15), except for a weak band detectable by using the highest concentration of RBM15 tested (50 nM, lane 16). These data show that recombinant RBM15 binds directly to RTE RNA, in agreement with the finding from the pull-down assay (Fig. 1A), and demonstrate that RBM15 interacts preferentially with RTE.
Identification of Novel Isoforms of RBM15-Ma et al. (48,49) reported three isoforms of RBM15 AEϩS, S, and L, which share aa 1-954 and have distinct C termini due to alternative splicing the RBM15 mRNA ( Fig. 2A). The isoform used throughout this work is RBM15 AEϩS spanning aa 1-977, which has also been used by other investigators (48,49,58). An anti-RBM15 antiserum raised against the C terminus of RBM15 AEϩS (aa 677-977) became recently available, which detects all isoforms. Testing human 293 and HeLa cells, we found several barely visible bands of endogenous RBM15 migrating higher than the major band, which migrates at ϳ100 kDa (Fig. 2B, lane 1). The endogenous RBM15 is significantly smaller than our exogenously expressed RBM15 AEϩS, which migrates at ϳ110 kDa (lane 3). Upon inspection of the sequence, we noted that there are two AUGs at residues 1 and 45, respectively. Our cDNA expression plasmid contains the optimized Kozak AUG sequence precluding initiation at downstream AUG. Therefore, we generated an expression plasmid containing 900 nucleotide of the authentic RBM15 5Ј-untranslated region obtained from the cDNA clone. Interestingly, we found that two proteins were produced (lane 2), one weaker band comigrating with the RBM15 AEϩS band shown in lane 3, and one strong, shorter band, suggesting that the second AUG at position 45 is preferentially used. This protein is slightly larger than the major endogenous produced RBM15 (lane 1). Because the proteins The presence of RTE in pDM138RTE promoted increased levels (ϳ7-fold) of CAT expression compared with the parent pDM138, as expected (34). B, RBM15 activates RTE-dependent Gag expression. Gag is only expressed from the unspliced mRNA containing CTE, RTE, or RRE. HeLa cells were transfected with the pNLgag plasmids carrying the indicated RNA export elements in the absence (open bars) or presence (black bars) of RBM15. pNLgagRRE was cotransfected with 0.1 g of HIV-1 Rev expression plasmid. Mean Gag values and standard deviations are shown. The presence of RTE, CTE, or RRE/Rev protein led to an increase in Gag production (10-, 124-, and 76-fold, respectively) as expected from previous studies (30,34,36). The additional -fold activation by exogenous RBM15 is shown. C, RBM15 activates expression from unspliced Gag-reporter mRNA. pNLCgag lacks splice sites and was previously shown to produce only unspliced mRNA (36). Transfection of the pNLCgag or pNLCgag containing RTE or the CTE in the absence or presence of the RBM15 expression vector was performed as described for B. The cell extracts were analyzed for Gag expression as above. D, activity of RTE mutants correlates with the ability to respond to cotransfected RBM15. The different RTE mutants shown were reported previously (35): RTEm27 (loop deletion; open box); m20, m25, and m26 (nucleotide changes, light gray boxes); and m21 and m24 (compensatory nucleotide changes, dark gray boxes). The RTE activity was measured as -fold activation when inserted into pNLgag compared with the parent plasmid, not containing RTE, and were reported by Smulevitch et al. (35). Ϫ and ϩ signs next to the mutants indicate their activity. HeLa cells were transfected with pNLgag containing the different RTE mutants in the absence or presence of RBM15. The additional -fold activation is shown. produced in lanes 2 and 3 are based on the isoform AEϩS, they represent the longest isoforms. These data suggest that isoforms with different C termini (see Fig. 2A) are preferentially produced. Using semi-quantitative reverse transcription-PCR, we confirmed the presence of all three forms of alternatively spliced RBM15 mRNAs in 293 cells and the presence of the 5Ј-untranslated region. We then generated cDNA expression plasmids for all isoforms utilizing either the AUG initiator codon at residues 1 or 45 as outlined in Fig. 2A. Fig. 2C shows that the major endogenous form of RBM15 comigrates with RBM15-L 45-957 initiated at the internal AUG #2. In subsequent experiments (data not shown), we confirmed that all isoforms function and localize to the nucleus like RBM15 AEϩS, which was the isoform used for all our studies and is referred to as RBM15 in this work.
RBM15 Promotes Expression of RTE-containing Reporter mRNAs-It has been shown that RTE promotes nucleocytoplasmic transport of unspliced retroviral mRNA (34,35,59). We tested the hypothesis that RBM15 binding is important for RTE function. Two reporter plasmids, pDM138 (37), producing chloramphenicol acetyl transferase (CAT, Fig. 3A), and pNLgag (36), producing HIV-1 Gag (Fig. 3B), were used to test RBM15 function in cotransfection experiments in the presence or absence of exogenous RBM15. CAT or Gag are only produced from the unspliced mRNA transcripts, which require the presence of a strong RNA transport element such as CTE or RTE in cis (30,34,59,60) or the Rev-responsive element RRE and HIV-1 Rev (36,37,61), as also shown in Fig.  3 (A and B, open bars). The presence of cotransfected RBM15 (black bars) further increased CAT expression by 6-fold from plasmid DM138-RTE but did not affect expression from pDM138 lacking RTE (Fig. 3A). Cotransfection of RBM15 also increased expression of gag from pNLGag containing RTE (ϳ27-fold), when compared with the expression obtained by the same plasmid in the absence of exogenous RBM15 (Fig. 3B). RBM15 also activated the RTE-containing pNLCgagRTE (36) to similar extent like pNLgagRTE (Fig. 3C). Plasmid pNLCgagRTE lacks the splice donor site located 5Ј to gag and produces only unspliced gag mRNA. These data indicate that RBM15 acts independent of splicing.
Different control gag plasmids were tested to study the specificity of RBM15 activation. No RBM15-induced stimulation was observed using the parent plasmid without RTE (Fig. 3, A-C), or the RRE-containing pNLgagRRE in the absence or presence of Rev (Fig. 3B), supporting specific action of RBM15 on RTE-containing mRNAs. Interestingly, a gag RNA containing the CTE transport element was reproducibly activated by RBM15 (ϳ4.5-fold, Fig. 3B). Activation of CTE-containing transcripts suggested that, although RBM15 acts preferentially on the RTE-containing RNA, it may have an additional general role in mRNA metabolism and may participate in NXF1-mediated export (see below).
RBM15 Function Requires the Presence of an Active RTE-A series of characterized RTE mutants (35), previously tested for their ability to activate gag expression (Fig. 3D), was examined in cotransfection experiments with the RBM15 expressing vector. The mutants used and their ability to induce Gag expression from pNLgag are shown in Fig. 3D. Cotransfection of RBM15 stimulated only the active RTE mutants m20, m21 m24, and m26. RBM15 did not activate the inactive mutants m25 and m27. Thus, the ability of exogenous RBM15 to further activate RTE-containing mRNA correlated with the activity of RTE. The data presented in Figs. 1 and 3 demonstrate that RBM15 specifically recognizes RTE RNA both in vitro and in vivo and that exogenous RBM15 promotes increased expression of RTEcontaining mRNAs. . Knockdown of RBM15 by RNA interference preferentially inhibits RTE activity. A, RBM15 targeting siRNA pool specifically reduced RBM15 expression. HeLa cells were transfected with 0.5 g of RBM15-FLAG expression plasmid together with 0.2 g of plasmids expressing GFP and SEAP. The next day, the cells were transfected with 10 nM of a siRNA oligonucleotide pool specific to RBM15, or with nonspecific control siRNA. RBM15 expression was analyzed by Western immunoblot. GFP and SEAP were measured in cell lysates and supernatant, respectively. B, untransfected HeLa cells were treated with siRNA pool specific to RBM15 or with nonspecific control siRNA oligonucleotide pool for 2 days. Cell extracts were analyzed on Western immunoblot using anti-RBM15 antiserum (upper panel) and anti-actin antibody (lower panel). C, HeLa cells were transfected with 10 nM of pools of RBM15-specific or control siRNA oligonucleotides. One day later, the cells were retransfected with the pNLgag plasmid containing RTE(IAP92L23) or CTE together with a plasmid expressing SEAP. Gag expression was measured from the culture supernatant 2 days later. Gag and SEAP expression in the presence of control siRNA was normalized to 100%. This experiment was performed in quadruplicate plates; means and standard deviations are shown.
Functional Knockdown of RBM15 by siRNA Inhibits RTE Function-Because we found that the exogenously expressed RBM15 stimulated the mRNA expression via RTE, we further investigated the effects of RBM15 depletion, performing functional knockdowns with siRNA. We first tested whether the selected pool of siRNAs was able to reduce cotransfected RBM15 levels. Transfection of a pool of four siRNAs targeting RBM15 led to a significant reduction of cotransfected FLAGtagged RBM15 protein levels by ϳ75% (Fig. 4A). No off-target effects were observed on the expression of coexpressed GFP and SEAP transcripts, confirming the specificity of RBM15 knockdowns. Using an RBM15-specific antiserum, we further confirmed that the endogenous levels of RBM15 were also specifically and efficiently down-regulated by siRNA (Fig. 4B).
We next cotransfected HeLa cells with gag expression vectors containing either RTE or CTE together with a pool of siR-NAs targeting the endogenous RBM15 or a pool of non-targeting siRNA. The production of secreted Gag protein was measured in the culture supernatants. Fig. 4C shows that RBM15-targeted siRNA significantly (by 74%) inhibited RTEmediated gag expression. We also noticed an inhibitory effect, although to a lesser extent, on the expression of the CTE-containing gag mRNA (by 55%). This effect was expected, because we had observed (see Fig. 3B) that exogenous RBM15 activated CTE-containing gag mRNA expression. Cotransfection of SEAP plasmid, revealed only a small effect of the RBM15-specific siRNA on the expression of the SEAP transcript (ϳ6% inhibition). These data provide another line of evidence that endogenous RBM15 is involved in RTE function. In addition, these findings further support our hypothesis that RBM15 participates in the NXF1 pathway, as evidenced by its effects on CTE-containing gag mRNAs.
RBM15 Directly Stimulates Nuclear Export of mRNA-We examined whether RBM15 is able to stimulate mRNA export directly by tethering RBM15 to a CAT reporter mRNA that is normally retained in the nucleus (25,39). The tethering assay is based on the interaction of the RNA-binding N-terminal domain of phage antiterminator protein N (N) with its RNAbinding motif (boxB) (39,62). CAT plasmids containing the boxB RNA-binding motifs (pDM128/B, Fig. 5A) or lacking the element (pDM128/PL) were cotransfected with plasmids expressing the factors of interest that were fused to the phage N-peptide. N-protein fusions to the full-length RBM15, or to the regions spanning aa 1-530 and 530 -977, respectively, were shown to localize in the nucleus (Fig. 5B), indicating the presence of two independent nuclear localization signals in RBM15. Using this tethering assay, we verified (Fig. 5C) that cotransfection of plasmids, producing the N-peptide fusion to known mRNA export factors such as HIV-1 Rev and NXF1, promoted CAT expression (black bars), whereas no expression was found using a reporter mRNA lacking the RNA binding elements (pDM128/PL, open bars), in agreement with the data previously reported by Wiegand et al. (39). Importantly, cotransfection of FIGURE 5. RBM15 tethered to RNA stimulates protein expression. A, CAT is only expressed from the unspliced mRNA containing the phage boxB RNA binding elements (pDM128/B (39)), which requires the presence of the N-peptide fused to an export factor such as RBM15. B, RBM15 contains three conserved RNA recognition motifs at the N terminus and the SPOC (spen paralogue and orthologue C-terminal) domain at the C terminus. HeLa cells were transfected with the plasmids expressing the N-peptide-RBM15 (complete, 1-530, and 530 -977) fusion containing a C-terminal HA tag and were visualized by indirect immunofluorescence using the anti-HA antibody. 4Ј,6-Diamidino-2-phenylindole staining was performed to visualize nuclei. The cytoplasmic to nuclear ratio of the unspliced CAT mRNA is calculated. The stimulation of unspliced CAT mRNA export was normalized to that the level obtained in the absence of coexpressed N fusion-tagged proteins. Similar data were obtained in several independent experiments. the N-RBM15 fusion protein also strongly promoted CAT expression (Fig. 5C). The export function of RBM15 depended on binding to the mRNA via the boxB elements, because it did not activate expression of the parent DM128/PL cat mRNA, lacking these elements (open bars). No increase in CAT expression from pDM128/B was found upon cotransfection of the RBM15 expression plasmid lacking the N-peptide (data not shown). Taken together, these data demonstrate that the observed stimulation by N-RBM15 (Fig. 5C) required direct interaction with the cat reporter mRNA. Testing the N-and C-terminal portions (RBM15 aa 1-530 and 530 -977, respectively), both localizing to the nucleus (Fig. 5B), revealed that the export activity of RBM15 lies entirely within its C-terminal portion (Fig. 5C).
To verify that RBM15 acts to increase the nuclear export of cat mRNA, we analyzed the effects of RBM15 tethering on nucleocytoplasmic distribution of boxB-containing cat transcripts by Northern blots (Fig. 5D). As expected, we found that, in the absence of tethered export factors, the unspliced cat mRNA was retained in the nucleus, whereas the spliced transcript was efficiently exported to the cytoplasm (Fig. 5D), in agreement with Zolotukhin et al. (25). Thus, the ratio of unspliced (U, CAT-producing) to spliced (S, non-protein producing) cat mRNA in the cytoplasm can be used as a quantitative measure of export efficiency for unspliced cat transcript. The ratios of unspliced to spliced cat mRNA in the nuclear fractions were not affected by any of the N fusion export factors, confirming that these proteins act specifically at the nuclear export step. We found that tethering of Rev, NXF1, or RBM15 led to an increase in the ratio of unspliced to spliced mRNAs in the cytoplasm. This is in agreement with the reported properties of Rev and NXF1, the export factors for HIV RRE-and CTE-containing mRNAs, respectively, which increase the steady-state levels of cytoplasmic unspliced HIV mRNAs (29 -33, 36, 63, 64). Together, these data provide direct evidence that RBM15 is a bona fide mRNA export factor with an effector domain located in its C-terminal portion.
RTE Utilizes the NXF1 Pathway for Nuclear Export in Xenopus Oocytes-To study the mechanism of RTE-mediated nuclear export, we employed an in vivo RNA export competition assay using Xenopus laevis oocyte nuclei microinjected with radiolabeled adenovirus-derived pre-mRNA in the presence of increasing amounts of unlabeled competitor RNA (Fig.  6). Adenovirus-derived intron-lariats are efficiently exported from the nucleus (N) to the cytoplasm (C) only if they contain an active RNA export element such as RTE (34,35) or CTE (56,65), whereas the export of the spliced mRNA is not affected, as outlined in Fig. 6D. Coinjection of U1⌬Sm RNA and U6⌬ss RNA served as quality controls and demonstrated proper function of the nuclear export machinery and intactness of the nuclei, respectively. This assay was previously used to show that export of the CTE-containing lariats utilizes the NXF1 pathway (3).
Using RTE RNA as competitor, we found interference with the nuclear export of the RTE-lariat at ϳ0.6 pmol of competitor (Fig. 6A, lane 3 versus lane 2 compared with lane 6 versus 5), indicating that the RTE export pathway is saturable. The saturating dose (ϳ0.6 pmol of RTE RNA) was comparable to those previously reported for the nuclear export pathways utilized by U1 small nuclear ribonucleoprotein (0.5 pmol) and mRNA (0.1 pmol) (55). The RTE competitor did not affect the export of U1⌬Sm RNA (representative of U1 small nuclear ribonucleoprotein export pathway). At 1.2 pmol of RTE competitor, we found some interference with splicing, resulting in reduced levels of intron-lariat and increased pre-mRNA levels (lane 1 compared with lane 7). Interestingly, the RTE competitor also strongly inhibited the export of spliced mRNA (see also Fig.  6C), because it was previously observed using the CTE as competitor in a similar assay (56). Thus, RTE is exported via a saturable pathway that overlaps with that of mRNA.
We next asked whether NXF1 was required for RTE-lariat export, despite the lack of a high affinity binding of NXF1 to RTE RNA (34). CTE RNA serves as a tool to inhibit NXF1 function, because NXF1 activity is specifically out-competed upon coinjection of very low amounts of CTE RNA (3,56,65). We found that the export of RTE-lariat was strongly inhibited by CTE competitor RNA, even at very low doses (Fig. 6B, lanes 2  versus 1 compared with lanes 6 and 5 or lanes 4 and 3). Excess CTE competitor had no effect on the export of U1⌬Sm RNA, as expected (56). As an additional control, we used the mutant CTEm36, which lacks the high affinity NXF1-binding sites, but maintains the stems and the overall RNA structure, and does not compete for CTE export (3,30). CTE and CTEm36 RNAs allow distinguishing NXF1 effects from other potential interactors (i.e. RNA helicase A (66 -68)). Fig. 6B (right panel) shows that CTEm36 RNA competitor did not affect the RTE-lariat export (lanes 8 versus 7 compared with lanes 10 versus 9). These results demonstrate a role of NXF1 in the RTE-mediated nuclear export and suggest a potential interaction of RBM15 and NXF1.
In the converse experiment, we tested whether RTE RNA could compete for the export of the CTE-containing intronlariat. Fig. 6C shows that coinjection of excess RTE RNA, using doses sufficient to compete RTE-lariat export (see Fig. 6A), had no effect on the export of the CTE-lariat (lanes 2 and 1 compared with lanes 6 and 5). Thus, RTE does not interfere with CTE export, which is in agreement with its low affinity to NXF1, as revealed by our in vitro binding studies (34). In contrast, the CTE-lariat export could be out-competed efficiently using even low doses of CTE competitor (ϳ0.06 pmol, Fig. 6C, FIGURE 6. NXF1 is involved in RTE RNA export. Radiolabeled adenovirus-derived precursor mRNA containing either RTE (A and B) or CTE (C) were injected into X. laevis oocyte nuclei. Increasing amounts of competitor RTE (A and C), CTE (B and C), or inactive mutant CTEm36 (B) RNAs were coinjected as indicated. The injected RNA mixture also contained U1⌬Sm RNA that is exported to the cytoplasm (an indicator of RNA export) and U6⌬ss, which remains in the nucleus (an indicator of nuclear integrity). Cytoplasmic (C) and nuclear (N) fractions were prepared after 3-h incubation. In panel A, total RNA (T) is also shown. The positions of the precursor mRNAs, spliced mRNAs, and the intron-lariat loop containing RTE (open triangle) or CTE (filled triangle) are indicated. Asterisks indicate an aberrant RNA species. D, schematic of the RNA products generated upon splicing of the adenovirus-derived precursor mRNA (Ad) in the absence or presence of the RTE (or CTE) inserted into the intron. The intron-lariat gets only exported to the cytoplasm in the presence of RTE (or CTE), whereas the spliced mRNA is always exported. DECEMBER 1 compared with lanes 10 versus 9), as expected (56). These data show that CTE RNA competes for the export of both the CTE-lariat (Fig. 6C) as well as the RTE-lariat (Fig. 6B,  left panel) with similar efficiency. These data are consistent with an essential role of NXF1 in RTE function.

RBM15 Promotes mRNA Export
RBM15 and NXF1 Bind to Each Other and Act Cooperatively-To test whether RBM15 and NXF1 can interact, we examined whether GST-tagged RBM15 protein can bind to reticulocyte produced radiolabeled proteins using an in vitro pull-down assay (Fig. 7A). We tested for interactions of RBM15 with the human NXF1, and as negative controls, with luciferase or with UAP56, a DExD/H box helicase involved in splicing and mRNA export (20,46,69,70). NXF1, luciferase, and UAP56 were used at the same molar concentrations in the binding reactions. We found that the human NXF1 bound to RBM15 in vitro, whereas no interactions with luciferase, UAP56 or "empty" GST beads were observed. Similarly, we found that the mouse NXF2, a highly related mRNA export factor (27), interacted with RBM15 (data not shown). The binding assays were performed in the presence of RNase A, demonstrating that the identified interactions of RBM15 with the NXF proteins are RNA-independent.
We further analyzed the RBM15-NXF1 interaction in vivo using FLAG-tagged RBM15 protein and HA-tagged NXF1 in cotransfection experiments (Fig. 7B). Cotransfection of the HA-tagged UAP56, a nuclear protein that does not bind RBM15 in vitro, served as specificity control in the assay. Western blot analysis confirmed expression of HA-tagged proteins and similar levels of the FLAG-tagged RBM15. Coimmunoprecipitations using anti-FLAG antiserum confirmed the presence of HA-tagged NXF1, but not UAP56, in the RBM15-containing complex. RNase treatment of the cell extract did not affect this association, demonstrating RNA-independent interactions. Thus, both the in vitro and in vivo experiments confirmed the interaction between NXF1 and RBM15.
We then examined the RBM15-NXF1 interaction in more detail upon cotransfection of a series of FLAG-tagged RBM15 deletion mutants and HA-tagged NXF1 (Fig. 7C). The use of RBM15 deletion mutants identified the NXF1-interacting region within aa 530 -977, whereas aa 1-530 and 530 -750 did not associate. We concluded that the C-terminal portion of RBM15 contains an interaction site for NXF1 (Fig. 7C), as well as the signals necessary to promote RNA export, as revealed by the tethering assay shown in Fig. 5. Confirming the specificity of these assays, we used Western blot analysis to verify expression of HA-tagged NXF1 and of the FLAG-tagged RBM15 proteins, whereas coimmunoprecipitations using anti-FLAG antiserum verified the presence of HA-tagged NXF1 in the complex containing the intact RBM15 . Taken together, the in vitro and the in vivo data indicate a direct interaction between NXF1 and RBM15.
Cooperativity between RBM15 and NXF1-To probe the functional interaction of RBM15 and NXF1, we examined whether coexpression of RBM15 and NXF1 affects expression of RTE-containing gag mRNA. Human 293T cells were transfected with the gag reporter plasmid containing the RTE in the absence or presence of exogenous NXF1-p15, RBM15, or a combination of both factors. Fig. 8A shows that Gag expression was increased in the presence of exogenous RBM15, as expected (see also Fig. 3), as well as, by exogenous NXF1-p15, although to a lesser extent. Importantly, we found that coexpression of both factors led to a further increase of Gag production, demonstrating cooperativity between RBM15 and NXF1. Addition of both factors had a more than additive effect on Gag production, suggesting a synergistic interaction of these factors. These data are consistent with the participation of NXF1 in RTE RNA export (Fig. 6).
We next tested whether the NXF1-RBM15 interaction also affects CTE-mediated mRNA expression (Fig. 8B). We found that cotransfection of pNLgagCTE with NXF1-p15 leads to elevated Gag levels, as expected (71). Interestingly, the presence of exogenous RBM15 also activated CTE-mediated expression. Furthermore, the combination of NXF1-p15 and RBM15 led to an additional increase in CTE-mediated expression. Similar data were obtained using another reporter plasmid expressing the HIV gag-pol transcript (71), which was used to measure NXF1-p15 response in transfected 293T cells (data not shown). These data show that RBM15 participates also in the CTE-mediated reporter gene expression. This finding is consistent with the observation that functional knockdown of RBM15 also affected the CTE-mediated gag expression (Fig. 4B). Thus, these findings reveal a functional interaction of NXF1 and RBM15 promoting RTE-as well as CTE-containing reporter mRNA export, supporting a model proposed in Fig. 8C (see below).

DISCUSSION
Studies of mRNA export of retroviruses and retroelements have led to the identification and characterization of molecular steps important for understanding cellular gene expression. In this report, we show that RBM15 selectively binds to RTE RNA, a retrotransposon-derived transport element, with high affinity. RBM15 also specifically binds to NXF1, a key export receptor for cellular mRNAs. These data are consistent with a model where RBM15 links the RTE-containing mRNA to the NXF1 export pathway (Fig. 8C). Using a previously reported in vivo export assay (39), we demonstrate that RBM15 also increases the nucleocytoplasmic transport of reporter mRNAs containing the boxB RNA element when tethered to the boxB-binding phage N-peptide (Fig. 5), similar to the export factors HIV Rev and NXF1. In addition to its effect on RTE-containing mRNAs, we also found that RBM15 increases expression of the reporter mRNAs containing CTE (Figs. 3 and 8), supporting a role of RBM15 in the NXF1 export pathway. Interestingly, RBM15 and NXF1 act cooperatively to promote a further increase in expression of RTE, as well as of CTE-containing reporter mRNAs (Fig. 8, A and B, respectively). These findings are in agreement with our previous observations, where RTE and CTE, present in close proximity on reporter mRNAs (see also Fig. 8C), were found to synergize (59). This synergy was reduced when the RNA elements were placed at a distance, suggesting that there is an interaction between RTE-and CTE-binding factors. The identification of a specific interaction of RBM15 and NXF1 suggests a possible molecular mechanism mediating the observed synergy.
We further found that the presence of excess RTE RNA is able to out-compete export of the spliced mRNA from Xenopus oocyte nuclei (Fig. 6). These results are similar to those previ- ,694 relative fluorescence units, respectively. C, models for RBM15 participation in mRNA transport: RBM15 directly binds to the RTE RNA. NXF1 binds to the C-terminal portion of RBM15, and the NXF1-p15 heterodimer provides the signal for interaction with the NPC, thus RBM15 tethers the RTE-containing mRNAs to the NXF1 export pathway. Smulevitch et al. (59) reported that the presence of RTE and CTE on a reporter mRNA synergistically increased reporter gene expression. The presence of the two RNA binding sites in close proximity may facilitate efficient interaction of RBM15 and NXF1, resulting in increased reporter mRNA expression. RBM15 also acts as a NXF1 cofactor and, thereby, RBM15 interacts indirectly with CTE. Thus, the RBM15-NXF1 interaction with the RTE RNA or the CTE RNA allow for export and expression of the unspliced mRNAs. RBM15 is thought to interact via NXF1 and the components of EJC with the cellular mRNA, suggesting a role in general export of spliced mRNA.
ously obtained for NXF1 using its high affinity target, CTE RNA, as competitor (56). One possible explanation is that RBM15 is involved in critical steps of mRNA export (see Fig.  8C). Our data suggest that RBM15 function on cellular mRNA is inhibited by the addition of excess of its high affinity binding RNA (RTE). Thus, these data are consistent with a role of RBM15 as NXF1 cofactor and provide evidence of a role of RBM15 in general mRNA export.
RBM15 belongs to a conserved family of proteins present as two members in most of the species examined, whereas there are three genes in humans and mice. The human family comprises the SPEN protein SHARP (SMRT/HDAC1-associated repressor protein); RBM15, also called One Twenty-Two (OTT); and the recently identified OTT3, also called RBM15b (72). SHARP has transcriptional repressor function (57,73,74), which is not shared by RBM15 or OTT3 (72). RBM15 is expressed in many tissues (49), but no function has been attributed to this protein. In this report, we demonstrate that RBM15 acts at the post-transcriptional level, particularly in mRNA export and expression. This function of RBM15 is clearly distinct from that reported for the related protein SHARP, which is involved in transcriptional suppression (57,73,74). In agreement with Hiriart et al. (72), we found no evidence that RBM15 acts as transcriptional repressor by testing the activity of different promoters such as HIV-1 long terminal repeat, SV40, or CMV (data not shown). Therefore, despite their evolutionary relationship, RBM15 and OTT3 appear to have functions distinct from that of SHARP. Interestingly, RBM15 has been also found fused to megakaryoblastic leukemia 1 protein (MKL1) in a translocation involving chromosome 1 and 22, resulting in acute megakaryoblastic leukemia (49,58,(75)(76)(77). The fusion protein consisting of RBM15 at its N terminus and MKL1, a transcription factor (78) at its C terminus, could interact with the mRNA export machinery. Although the RBM15-MKL1 fusion protein was found to maintain the specific transactivator function of MKL1 (78), the fusion protein lost the post-transcriptional activator function of RBM15, as measured by its inability to activate RTE-mediated mRNA expression. 3 However, it is possible that the RBM15-MKL1 fusion protein has transdominant suppressor function contributing to the oncogenic properties of RBM15-MKL1. Alternatively, the possible decrease of post-transcriptionally active RBM15 due to MKL1 fusion could affect mRNA regulation, potentially contributing to leukemogenesis. Elucidation of the role of RBM15 in mRNA export provides the basis for additional testable hypotheses on the oncogenic mechanism of RBM15-MKL1.
While this work was finalized, another member of the SPEN family, OTT3, was reported to bind to the Epstein-Barr virus early protein EB2. OTT3 was further shown to participate in splicing regulation of ␤-thalassemia mRNA, supporting its role in post-transcriptional steps of gene expression (72). Thus, both RBM15 and the related OTT3 participate in post-transcriptional control of gene expression. EB2 interacts with the SPOC domains of all three human SPEN family proteins SHARP, RBM15, and OTT3 (72). However, whereas SHARP and RBM15 interact with the C-terminal portion of EB2, OTT3 interacts with its N-terminal portion, demonstrating distinct properties despite the high level of homology in the SPOC domains (72). Interestingly, it is the C-terminal portion of RBM15 containing the SPOC domain that provides the interaction site with NXF1 as well as with the EB2 (72). Although there is no recognizable motif shared among these factors, a more in-depth analysis may reveal common structural requirements. Alternatively, distinct regions within the C-terminal portion of RBM15 interact with these factors. The function of OTT3 in RTE-mediated mRNA expression and its interaction with NXF1 are currently under investigation.
We propose that two structurally distinct but functionally analogous RNA export elements (RTE and CTE) have independently evolved (Fig. 8C). RTE, present in murine intracisternal A-particle retroelements (34,35), emerged as high affinity ligand for RBM15. CTE, present in the related Type D simian retroviruses, has evolved to bind to NXF1 (29,30,33). It is believed that high affinity binding to export factors is essential to promote the export of the full-length unspliced mRNA of retroviruses or retroelements. In contrast, the vast majority of cellular mRNAs have to undergo splicing before export, resulting in a splicing-dependent deposition of export factors such as NXF1. In this report, we present evidence that RBM15 also has a role in the general mRNA export pathway, because RTE inhibits export of the spliced adenovirus mRNA from the Xenopus oocyte nuclei (Fig. 6). RBM15 also binds directly to the major export receptor for mRNAs, NXF1. We therefore propose that RBM15 plays a role in the general mRNA export pathway, by interacting with EJC via NXF1 (Fig. 8C). The molecular mechanism by which RBM15 participates in general mRNA metabolism remains the subject of further studies.