A Cellular Protein, hnRNP H, Binds to the Negative Regulator of Splicing Element from Rous Sarcoma Virus*

Incomplete RNA splicing is a key feature of the retroviral life cycle. This is in contrast to the processing of most cellular pre-mRNAs, which are usually spliced to completion. In Rous sarcoma virus, splicing control is achieved in part through acis-acting RNA element termed the negative regulator of splicing (NRS). The NRS is functionally divided into two parts termed NRS5′ and NRS3′, which bind a number of splicing factors. The U1 and U11 small nuclear ribonucleoproteins interact with sequences in NRS3′, whereas NRS5′ binds several proteins including members of the family of proteins. Among the proteins that specifically bind NRS5′ is a previously unidentified 55-kDa protein (p55). In this report we describe the isolation and identification of p55. The p55 binding site was localized by UV cross-linking to a 31-nucleotide segment, and a protein that binds specifically to it was isolated by RNA affinity selection and identified by mass spectrometry as hnRNP H. Antibodies against hnRNP H immunoprecipitated cross-linked p55 and induced a supershift of a p55-containing complex formed in HeLa nuclear extract. Furthermore, UV cross-linking and electrophoretic mobility shift assays indicated that recombinant hnRNP H specifically interacts with the p55 binding site, confirming that hnRNP H is p55. The possible roles of hnRNP H in NRS function are discussed.

Most eukaryotic protein-encoding genes contain introns that must be removed from primary RNA transcripts by splicing to generate functional mRNA (1). Viruses that infect eukaryotic cells have evolved mechanisms to exploit the splicing process to expand their coding potential and regulate gene expression. Expression of certain viral genes requires that full-length viral RNAs undergo splicing to subgenomic forms. For example, HIV, 1 a complex retrovirus, produces ϳ40 spliced RNA forms for the expression of its multiple gene products (2,3). In contrast, a simple retrovirus such as Rous sarcoma virus (RSV) contains only four genes (gag, pol, env, and src) and produces only two spliced subgenomic RNAs (env and src) (see Fig. 1) (4). In all cases, full-length genomic RNAs are required for the expression of the gag/pol genes and for use as genome in progeny virions. The maintenance of both spliced and unspliced viral RNA forms differs from the processing of most cellular pre-mRNAs and is accomplished through incomplete splicing of the viral RNAs. RSV and other members of the avian leukosis virus family contain a unique cis-acting element termed the negative regulator of splicing (NRS) that acts globally to repress splicing to both the env and src splice sites (5,6). The NRS is located within the gag gene and is distinct from the viral splice sites, lying Ͼ300 nucleotides downstream of the 5Ј splice site and Ͼ4000 nucleotides upstream of the first 3Ј splice site (7). The importance of the NRS to splicing control and the viral life cycle is illustrated by observations that NRS deletions and mutations cause increased splicing and impaired replication (5,6,8).
The NRS can also inhibit the splicing of heterologous introns (5,7,9), demonstrating that the inhibition is a general property of the element itself. The minimal NRS RNA element (see Fig.  1) spans ϳ230 nucleotides and functionally can be divided into two parts (NRS5Ј and NRS3Ј), both of which are required for function (7). Despite its role in splicing inhibition, the NRS has been shown to associate with a number of essential cellular splicing factors (9,10). NRS3Ј contains overlapping sites that bind the U1 and U11 small nuclear ribonucleoproteins (snRNPs) (8,11), factors that recognize authentic 5Ј splice sites in U2-and U12-dependent introns and initiate the formation of the spliceosome (1,(12)(13)(14). NRS5Ј is highly purine-rich and binds members of the SR family of splicing factors, predominantly ASF/SF2 (10). SR proteins are a class of serine-argininerich splicing factors, which are required at many steps of the splicing process and among other activities aid in the recruitment of U1 snRNP to 5Ј splice sites (15)(16)(17)(18)(19). NRS splicing inhibition is dependent on SR proteins (10,20) and on the binding of U1 snRNP (8,11). It is predicted that the NRS functions as a "decoy" 5Ј splice site that sequesters the authentic 3Ј splice site in a stable but nonproductive complex, thereby preventing its interaction with the authentic 5Ј splice site.
Although considerable evidence supports a role for SR proteins in NRS-mediated splicing inhibition, deletion of the primary SR binding region in NRS5Ј does not fully abolish NRS function (7,20), suggesting that the region downstream of the SR protein binding site may contribute to NRS activity. Studies of a recombinant avian leukosis virus (EU8) that induces rapid onset B-cell lymphomas due to the presence of a 42-nucleotide deletion within NRS5Ј (21) lend further support for a role of this downstream site in splicing inhibition. The deletion not only impairs NRS inhibition (21) but also correlates with increased readthrough past the viral polyadenylation site into a downstream cellular gene, c-myb (22). Overexpression of truncated Myb protein caused by increased splicing results in the observed lymphomas (21, 22). The deletion maps largely downstream of the SR protein binding site, suggesting that NRS impairment is caused by the loss of binding of some other factor(s) within this region. In previous studies, an unidentified protein of approximately 55 kDa (termed p55) that is biochemically distinct from the classical SR proteins was shown to bind specifically to NRS5Ј (10). Using a UV cross-linking assay, we have mapped the p55 binding site to a 31-nucleotide region of NRS5Ј downstream of the primary SR protein binding site and within the region deleted in the EU8 virus, suggesting a relationship between p55 binding and NRS function. To better understand the significance of p55, we sought to isolate and characterize this protein. Using RNA affinity selection, we have identified p55 as the cellular protein hnRNP H, a member of the heterogeneous nuclear ribonucleoprotein (hnRNP) family of proteins that are involved in the processing and transport of pre-mRNAs (23). Based on reported activities for hnRNP H, this protein may be involved in NRS-mediated splicing inhibition and/or polyadenylation of viral RNAs.

EXPERIMENTAL PROCEDURES
Plasmid Construction and in Vitro RNA Transcription-RSV DNA fragments were obtained from the Prague C strain (24), and sequence coordinates are as described by Schwartz et al. (25). Plasmids p3ZKXMS, p3ZMS3Ј, p3ZBB, p3ZBB5Ј, and p3ZBB3Ј were described previously (10,11). Additional NRS fragments (nucleotides 701-753, 754 -797, 701-770, 740 -797, and 740 -770) were generated by polymerase chain reaction using primers containing KpnI and XbaI sites and subcloned into these sites in pGEM-3Z (Promega). Primer sequences are available on request. The sequence of all constructs was confirmed by DNA sequencing.
Plasmids were linearized with XbaI (with the exception of p3ZBB, which was linearized with BamHI) and transcribed in vitro using T7 RNA polymerase and [ 32 P]ATP (10,26). Control RNA was transcribed from PvuII-digested pGEM-4Z (Promega). All RNAs were gel-purified before use.
Plasmid p3Z(740 -797)x2 was generated by first digesting p3Z(740 -797) with KpnI, blunting with T4 DNA polymerase, and then digesting with HindIII. The excised fragment was then subcloned into HincII/ HindIII-digested p3Z(740 -797) to generate the final construct. To produce RNA for affinity selections, this plasmid was linearized with SalI, and large scale in vitro transcription was performed using T7 RNA polymerase (27). Control RNAs for affinity selection were transcribed from PvuII-digested pGEM-4Z. All RNAs were trace-labeled with [ 32 P]ATP to monitor yield and coupling to agarose beads (described below).
RNA Affinity Selections-250 pmol of RNA was covalently coupled to agarose beads as described (28). Affinity selections were performed by adding the RNA-agarose beads to 325 l of reactions containing ϳ2.65 mg of a 0 -65% ammonium sulfate (AS65) fraction of HeLa total cell extract in 12.7 mM Hepes (pH 7.9), 38.5 mM KCl, 20 mM creatine phosphate, 0.4 mM ATP, 3 mM MgCl 2 , 0.16 mM dithiothreitol, 1.9% glycerol, and 0.08 mM EDTA. After the reactions were incubated at 30°C for 20 min, the beads were collected, washed four times with 1 ml of 20 mM Hepes (pH 7.9), 100 mM KCl, 5% glycerol, 0.2 mM EDTA, and 4 mM MgCl 2 at 4°C, and resuspended in 30 l of sample loading buffer. Bound proteins were eluted at 90°C for 5 min, and half of the supernatant was analyzed on a 10% SDS-polyacrylamide gel. The gel was Coomassie Blue-stained, and bands of interest were excised directly from the gel and subjected to MALDI-TOF analysis (29) by John Leszyk at the Protein Microsequencing and Proteomic Mass Spectrometry Lab at the University of Massachusetts Medical School.
Extract Preparation and Purification of Recombinant Proteins-HeLa S3 cells were grown in spinner flasks, and nuclear extracts were produced as described previously (30). The AS65 fraction used (a gift from L. McNally, Medical College of Wisconsin) was produced as described for the preparation of total HeLa SR proteins (31) and was dialyzed against 20 mM Hepes (pH 7.9), 100 mM KCl, 5% glycerol, and 0.2 mM EDTA before use.
The Q fraction was produced as follows. AS65 extract was adjusted to 2 M NaCl and centrifuged at 213,000 ϫ g for 20 min at 4°C to extract p55 cross-linking activity from endogenous complexes. The supernatant was collected, concentrated using a centrifugal filter device (Centriprep YM-30, Millipore), and separated on a Sephacryl S-200 HR (Amersham Pharmacia Biotech) gel filtration column. Fractions with peak crosslinking activity were pooled, loaded on a prepacked High Q ion exchange column (Econo-Pac, Bio-Rad), and eluted using a salt gradient from 100 to 600 mM KCl. The resulting fractions containing an ϳ55-kDa cross-linking activity were pooled, diluted with 20 mM Hepes (pH 7.9), 5% glycerol, and 0.2 mM EDTA to approximately 100 mM KCl, and concentrated by centrifugation (Centriplus YM-30, Millipore) before use.
To produce recombinant hnRNP H, BL21(DE3)pLysS bacterial cells (Novagen) were transfected with a pET15b vector containing the hnRNP H cDNA (generously provided by D. Black, University of California, Los Angeles), and the protein was purified from inclusion bodies using Ni 2ϩ affinity chromatography (His-Bind Resin, Novagen) as described (32). The recombinant protein was eluted in 20 mM Hepes (pH 7.9), 100 mM KCl, 5% glycerol, and 0.2 mM EDTA containing 250 mM imidazole and dialyzed against 20 mM Hepes (pH 7.9), 100 mM KCl, 5% glycerol, and 0.2 mM EDTA. The protein was concentrated using centrifugal filter devices (Centriplus YM-30 and Centricon YM-30, Millipore) according to the manufacturer's instructions. Aliquots were stored at Ϫ80°C until use. All protein concentrations were determined using the Bio-Rad protein assay kit (Bio-Rad).
UV Cross-linking-High specific activity RNAs (10 5 -10 6 dpm) were incubated in 25-l reaction mixtures under modified in vitro splicing conditions (13 mM Hepes (pH 7.9), 100 mM KCl, 20 mM creatine phosphate, 0.4 mM ATP, 3 mM MgCl 2 , 0.16 mM dithiothreitol, 2% glycerol, 0.08 mM EDTA, and 0.4% mammalian cell and tissue extract Protease Inhibitor Cocktail (Sigma) with 25 g of nuclear extract or 2-40 pmol of recombinant hnRNP H for 10 -30 min at 30°C. The samples were spotted onto parafilm at 4°C and exposed to UV light (254 nm) for 20 min at a distance of ϳ4 cm, transferred to microcentrifuge tubes containing 20 -25 g of RNase A, and incubated at 37°C for 15 min. Sample loading buffer was then added, and samples were separated on a 10 or 12% SDS-polyacrylamide gel. Gels were Coomassie Blue-stained to visualize size standards, dried, and analyzed by autoradiography or with a PhosphorImager (Molecular Dynamic Storm 860).
Immunoprecipitations-Protein A-Sepharose beads (Amersham Pharmacia Biotech) were washed in phosphate-buffered saline and suspended as a 50% slurry. 10 l of slurry was incubated in 500 l of IP buffer (20 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.1% Triton X-100) in the presence or absence of 1 l of antiserum and rotated at 4°C for 3 h. The beads were washed three times with 500 l of IP buffer and resuspended in 100 l of IP buffer. UV cross-linking was performed as described above, and following RNase treatment, the beads were added to the samples, rotated at 4°C for 1 h, washed three times with 500 l of IP buffer, and suspended in 40 l of sample loading buffer. Bound proteins were eluted by boiling for 5 min, and the supernatants were separated on a 10% SDS-polyacrylamide gel and visualized as described above.
Electrophoretic Mobility Shift Assays-Labeled RNA (2.5 fmol) was incubated at 30°C for 30 min in 25 l of 13 mM Hepes (pH 7.9), 70 mM KCl, 20 mM creatine phosphate, 0.4 mM ATP, 3 mM MgCl 2 , 0.16 mM dithiothreitol, 2% glycerol, 0.08 mM EDTA, and 1 g of tRNA in the presence or absence of 5 g of nuclear extract or 10 -40 pmol of recombinant hnRNP H. Heparin (Sigma) was then added to a concentration of 5 mg/ml, and the reaction was incubated at 30°C for an additional 10 min. The completed reaction was separated on a 4% polyacrylamide (29:1) nondenaturing gel. After electrophoresis, the gel was dried and visualized by autoradiography or with a PhosphorImager. For supershifts, assays were performed as described above except that the reactions were initially incubated at 30°C for 15 min, 1 l of antiserum was added, and the reactions were incubated an additional 15 min at 30°C. Antisera against hnRNP H and hnRNP F were generously provided by D. Black. Control antisera (glutathione S-transferase-myelin-basic protein and affinity-purified Exo U) were gifts from R. Fritz and D. Frank, Medical College of Wisconsin.

RESULTS
Localization of the p55 Binding Site-Previous studies established that a number of protein factors in HeLa nuclear extract bind specifically to the 5Ј portion of the NRS (nucleotides 701-797) ( Fig. 1) (10). Although most of these proteins are implicated as being SR proteins (10), the identity of a 55-kDa protein that is biochemically distinct from the SR proteins has remained unclear. The binding site for the SR proteins was localized to nucleotides 715-748 of NRS5Ј ( Fig. 1) (10), but because deletion of this region does not fully eliminate NRS activity (7,20), other factor(s) that binds within nucleotides 748 -797 may contribute to splicing inhibition. Although it was shown that p55 binds NRS5Ј, its precise binding site was not determined. To identify the p55 binding site we performed UV cross-linking in HeLa nuclear extract with a panel of NRS RNA variants ( Fig. 2A). A number of nonspecific RNA-binding proteins were detected when cross-linking was performed with vector RNA (Fig. 2B, lane 1). Consistent with previous studies (10), a specific cross-link of 55 kDa was observed using the full-length NRS and NRS5Ј but not with NRS3Ј or vector RNA (Fig. 2B, lanes 1-4). Using the NRS5Ј variants, we mapped the minimal p55 binding site to nucleotides 740 -770, indicating that p55 binds primarily downstream of the SR proteins (Fig.  2B, lanes 5-9). In addition, this region maps within a 42nucleotide region of the NRS (nt 735-776) previously shown to be functionally important for proper splicing control in a recombinant avian virus (21), further implicating p55 as a potentially important factor in NRS activity. Based on these observations, we sought to purify and identify p55.
Identification of p55 as the Cellular Protein hnRNP H-Because previous data indicated that p55 was present in both nuclear and cytoplasmic HeLa cell extracts (10), we chose to use HeLa total cell extracts as a starting material and follow p55 purification by UV cross-linking. A 0-65% ammonium sulfate precipitation was performed to remove SR proteins, which might interfere with or obscure p55 isolation. Initially this AS65 extract was fractionated biochemically (see "Experimental Procedures"), and enrichment of an ϳ55-kDa cross-link was observed (data not shown). The predominant protein in this size range was identified by mass spectrometry as a 48-kDa proteolytic fragment of nucleolin (data not shown). Importantly, this protein did not show specific binding to the NRS (data not shown), and since it has not been suggested to participate in any aspect of pre-mRNA splicing (33), we concluded that this protein was unlikely to be biologically relevant to NRS function.
Because it appeared that the specific p55 cross-link was lost in the above purification scheme, we next performed RNA affinity selection directly from AS65 HeLa extracts (Fig. 3). Several proteins were selected with vector RNA, including one that comigrated with the 48-kDa nucleolin fragment (Fig. 3,  lane 2). This and other proteins were also detected using a p55-specific NRS RNA, but an additional specific protein of ϳ55 kDa was also selected (Fig. 3, lane 3). Because this protein was the correct size for p55, it was excised from the gel, digested with trypsin, and subjected to mass spectrometry. Nine of eleven identified masses corresponded to a single protein, and post-source decay analysis unambiguously confirmed this to be the cellular protein, hnRNP H (data not shown).
Antibodies against hnRNP H Specifically Immunoprecipitate A, the RSV genome is shown above with the relative positions of the long terminal repeats (LTR) and the gag, pol, env, and src genes indicated (not to scale). Alternative splicing is shown by lines between the 5Ј and 3Ј splice sites (ss). The location of the NRS is shown within the gag gene and expanded below including sequence coordinates. The overlapping binding sites for the U1 and U11 small nuclear ribonucleoproteins (snRNPs) in the 3Ј portion of the NRS (nt 913-926) are denoted by a black box. The lightly shaded region indicates the 5Ј portion of the NRS, and the site where SR proteins were previously shown to bind (nt 715-748) is dark shaded. B, a partial sequence of the purine-rich 5Ј region of the NRS is shown, and the binding site for p55, as determined in this study, is indicated by brackets. the p55 Cross-link from AS65-It was predicted that if hnRNP H is p55, antibodies against hnRNP H should immunoprecipitate the p55 cross-link. To test this, NRS RNAs were incubated in either AS65 or a fraction (Q) enriched for nucleolin but lacking hnRNP H (data not shown), the reactions were UV cross-linked, and the samples were immunoprecipitated with hnRNP H antibodies. When NRS5Ј was used as the target RNA, antibodies against hnRNP H specifically immunoprecipitated a 55-kDa cross-link from AS65 but not from the Q fraction (Fig. 4, compare lanes 2 and 10). The precipitated crosslink exhibited the same mobility pattern as the cross-link seen in the extract (Fig. 4, lanes 1 and 2), and no cross-link was immunoprecipitated using control antibodies or Protein A-Sepharose beads alone (Fig. 4, lanes 3 and 4). Identical results were obtained when the minimal p55 binding site (nt 740 -770) RNA was used (Fig. 4, lanes 5-8 and lanes 13-16). These results are consistent with hnRNP H being p55.
Antibodies against hnRNP H Supershift NRS5Ј-specific Complexes-Previous studies showed that a specific complex forms on NRS5Ј in HeLa nuclear extract as determined by mobility shift assays and that a 55-kDa protein is a component of this complex (10). To provide evidence that this 55-kDa protein is hnRNP H, we performed electrophoretic mobility supershift assays. The specific complex was formed on NRS5Ј RNA in HeLa nuclear extract (Fig. 5, lanes 1 and 2) and could be supershifted with antibodies against hnRNP H but not with control antibodies (lanes 3-6). Identical results were obtained using the minimal p55 binding site (nt 740 -770) (Fig. 5, lanes  10 -15). Additionally, the supershift was detected with only anti-hnRNP H and nuclear extract and not with antiserum alone (Fig. 5, compare lanes 3 to lanes 7-9, and lane 12 to lanes 16 -18). Furthermore, although antibodies against hnRNP H supershifted NRS complexes, antibodies against a closely related protein, hnRNP F, did not (Fig. 5, lanes 5 and 14). Taken together, these results indicate that hnRNP H is a component of the NRS5Ј-specific complex observed with mobility shift assays.
Recombinant hnRNP H Cross-links to the NRS at the p55 Binding Site-To determine whether hnRNP H interacts directly with the NRS, recombinant hnRNP H was produced in bacteria and utilized in UV cross-linking assays. The recombinant hnRNP H was essentially homogeneous and reacted strongly with anti-hnRNP H antibody (data not shown). Little cross-linking was detected to vector RNA with up to 40 pmol of recombinant hnRNP H (Fig. 6A, lanes 1-3). In contrast, crosslinking to NRS5Ј was detected with as little as 2 pmol of recombinant protein, and the cross-link was intensified with increasing hnRNP H concentration (Fig. 6A, lanes 4 -7), indicating that hnRNP H binds directly to NRS5Ј.
It was further predicted that if hnRNP H is p55, crosslinking to the NRS should be dependent on the presence of the p55 binding site (as determined in Fig. 2). To test this, UV cross-linking was performed with recombinant hnRNP H and selected RNAs that were used in Fig. 2 to determine the p55 binding site. Consistent with the results above, cross-linking of hnRNP H was seen with full-length NRS and NRS5Ј but not with vector RNA (Fig. 6B, lanes 1-6). Some cross-linking was detected using NRS3Ј (Fig. 6B, lanes 7 and 8), indicating a degree of promiscuity in the binding of purified hnRNP H. Cross-linking was not observed with a segment of NRS5Ј that does not cross-link p55 (nt 701-753) (Fig. 6B, lanes 9 and 10). In contrast, cross-linking was observed using a region of NRS5Ј which showed slight p55 cross-linking (nt 754 -797) (Fig. 6B,  lanes 11 and 12) and was clearly detected using the minimal p55 binding site (Fig. 6B, lanes 13 and 14). These results show that hnRNP H cross-linking is dependent on the presence of the p55 binding site.
Recombinant hnRNP H Induces a Mobility Shift of NRS RNAs Containing the p55 Binding Site-Because recombinant hnRNP H binds directly to the NRS, we predicted that it would shift NRS RNAs in an electrophoretic mobility shift assay. No shift was observed with vector RNA and up to 40 pmol of recombinant hnRNP H (Fig. 7A, lanes 3-6), whereas NRS5Ј RNA showed an observable shift with as little as 20 pmol (Fig.  7A, lanes 9 -12). To confirm that the binding of hnRNP H to the NRS requires the p55 binding site, mobility shifts were performed using selected RNAs described in Fig. 2. As expected, shifts were detected using full-length NRS and NRS 5Ј but not vector RNA (Fig. 7B, lanes 3, 7, and 12). The presence of secondary structure in NRS3Ј caused a portion of this RNA to migrate similarly to the hnRNP H-dependent complex, but this pattern was unchanged in the presence or absence of recombinant hnRNP H (Fig. 7B, compare lanes 14 and 17) indicating the protein does not form a complex with this RNA. In addition, no shift was detected using nucleotides 701-753, which do not cross-link p55 or hnRNP H (Fig. 7B, lane 21), but a shift was observed with NRS5Ј segments that do cross-link hnRNP H (Fig. 7B, lanes 25 and 30). Taken together, these results confirm that hnRNP H binds directly to the p55 binding site. DISCUSSION Despite its roles in splicing inhibition, a number of essential splicing factors have been shown to interact with the NRS, including the U1 and U11 snRNPs and several SR proteins. In addition to these factors, a previously unidentified 55 kDa protein (p55) was shown to interact with the NRS by UV cross-linking (10). We became interested in this protein because of its specific binding the 5Ј portion of the NRS and because its biochemical properties suggested it was not an SR protein (10). We mapped the binding site for p55 to nucleotides 740 -770 of the NRS, downstream of the primary SR protein binding site (nt 715-748). Because this region can support a degree of NRS activity in the absence of the predominant SR protein binding site (7,20) and is required for maximal NRSmediated splicing inhibition (21), we sought to identify and characterize this protein.
Through biochemical purification, we identified two proteins of approximately 50 -55 kDa that cross-link to nucleotides 740 -770 in HeLa cell extracts, only one of which was sequencespecific (data not shown). The nonspecific protein was determined to be a previously described proteolytic fragment of nucleolin containing the RNA binding domains (34). To purify the protein responsible for the specific p55 cross-link, we used RNA affinity and identified it as the cellular factor hnRNP H by mass spectrometry. Three observations support the view that hnRNP H is p55. First, antibodies against hnRNP H immunoprecipitated the p55 cross-link from HeLa cellular extracts. Second, hnRNP H antibodies supershifted complexes that form on NRS5Ј in HeLa nuclear extract. Third, recombinant hnRNP H assembled a similar complex and could be cross-linked only to NRS RNAs that contained its binding site, indicating that hnRNP H binds the NRS directly. Furthermore, these hnRNP H-induced complexes migrated identically to those observed in nuclear extract (Fig. 7B, compare lanes 7, 12, 25, and 30 to lanes 5, 10, 23, and 28) suggesting that hnRNP H is solely responsible for these shifts. Collectively, these results indicate that hnRNP H is p55.
hnRNP H is a member of a large group of proteins known as heterogeneous nuclear ribonucleoproteins that bind to RNA polymerase II transcripts and contribute to their maturation to mRNAs (23). The interaction of an hnRNP protein with a viral RNA is not unique. Recently, members of the hnRNP A/B family of proteins were found to bind to an exonic splicing silencer sequence in exon 2 of the HIV-1 tat RNA and shown to be required for splicing inhibition (28). The fact that hnRNP H may be involved in NRS activity might reflect a common role for hnRNPs in viral RNA splicing inhibition. However, the requirement for additional sequences within the NRS (i.e. the U1 site) highlights an important difference between the mechanisms of NRS and HIV tat splicing inhibition.
Although our data do not address function, the significance of hnRNP H in NRS function is suggested by studies of the EU8 recombinant avian leukosis virus (21). This virus, which induces rapid onset B-cell lymphomas through insertional activation of the c-myb gene (22), carries a gag deletion that impairs NRS function (21). This deletion removes nucleotides FIG. 5. Complexes assembled on NRS5 are supershifted with antibodies against hnRNP H. Electrophoretic mobility shift assays were performed using either NRS5Ј or p55-specific (nt 740 -770) RNAs. High specific activity RNAs (2.5 fmol) were incubated under in vitro splicing conditions in 5 g of HeLa nuclear extract (NE), treated with heparin, resolved on a 4% polyacrylamide native gel, and visualized with a PhosphorImager. For supershifts, antisera (Ab) raised against hnRNP H (H), a related protein hnRNP F (F), or control proteins Exo U (U) and glutathione S-transferase-myelin-basic protein (G) were added after the binding reaction as described under "Experimental Procedures." The positions of free RNA, the NRS-specific shifted complex (arrow), and the corresponding supershifted complex (*) are indicated. Decreases in the relative amounts of shifted complex and free RNA are caused by nucleases present in the antisera (H, F, and G) as they are not seen with affinity-purified antibodies (U). The figures were generated as described for Fig. 2. 735-776 of the NRS, which encompass the hnRNP H binding site. It is tempting to speculate that the loss of hnRNP H binding may result in the NRS defect seen in the EU8 virus. Although we cannot rule out the possibility that other proteins may functionally interact with this region, the specificity of hnRNP H makes it a prime candidate.
In contrast to previously identified NRS binding factors (SR proteins and snRNPs), which have defined roles in pre-mRNA splicing, the role of hnRNP H in cellular RNA metabolism is not clear. Recent reports have implicated hnRNP H in the regulated splicing of at least two genes. Chou et al. (32) have shown hnRNP H to be a component of a neuron-specific enhancer complex that stimulates splicing of the c-src N1 exon. hnRNP H has also been shown to bind an exonic splicing silencer element in exon 7 of the rat ␤-tropomyosin gene that represses splicing of this exon in nonmuscle cells (35). However, immunodepletion had no effect on the splicing of either an adenoviral (32) or ␤-globin (35) pre-mRNA, suggesting that hnRNP H is not a general splicing factor. Although hnRNP H is required in both these cases, it is unclear how this protein precisely contributes to either positive or negative splicing regulation. In addition, although these are both examples of tissue-specific alternative splicing, hnRNP H does not exhibit tissue-specific expression in humans (36), rats (37), or mice (35), suggesting that these effects may be mediated via proteinprotein interactions. Determination of additional factors that functionally interact with hnRNP H will be important in understanding how this protein acts to modulate pre-mRNA splicing.
hnRNP H is a member of a subfamily of distinct but structurally similar hnRNP proteins that include hnRNP H, hnRNP HЈ, and hnRNP F (36,38). hnRNP F is 78% identical to hnRNP H (36) and is also a functional component of the c-src splicing enhancer complex (39). Although hnRNP H and hnRNP F may bind as a heterodimer to the c-src enhancer (32), hnRNP F was not detected in hnRNP H-containing NRS5Ј complexes by supershift assays, suggesting that hnRNP F is not involved in NRS function. hnRNP HЈ, which is 96% identical to hnRNP H, binds a G-rich element downstream of the core SV40 late polyadenylation signal and stimulates 3Ј end processing (40). Despite their extreme homology, hnRNP HЈ was not detected by RNA affinity selection with NRS-specific RNAs (data not shown). However, because of the high homology between these two proteins it is unlikely that they possess distinct binding specificities, and the absence of hnRNP HЈ may reflect a difference in the relative levels of hnRNP H and hnRNP HЈ in HeLa cells.
How might hnRNP H function in NRS-mediated splicing inhibition? Recent evidence indicates that NRS5Ј can function as a potent splicing enhancer, which is consistent with the binding of SR proteins to NRS5Ј (20). However, enhancer activity is not restricted to the primary SR binding region (nt 715-748) and is diffusely localized throughout NRS5Ј (20). This suggests that hnRNP H may be capable of mediating splicing enhancer activity or augmenting other enhancer activities, consistent with its role in the neural c-src splicing enhancer complex (32). An alternative role for hnRNP H is suggested by the observations that hnRNP HЈ is involved in pre-mRNA 3Ј end processing (40). Based on the high identity between hnRNP H and hnRNP HЈ, it is likely that the two proteins act similarly in vivo. As discussed previously, the EU8 virus exhibits increased readthrough of viral RNAs into the downstream c-myb gene (22). This readthrough effect is likely caused by the decreased efficiency of viral RNA 3Ј end processing, which is observed upon deletion of the NRS (41). The loss of the hnRNP H binding site in EU8 suggests a role in polyadenylation. Additionally, because deletion of the hnRNP H binding site appears to lead to FIG. 7. Recombinant hnRNP H causes a mobility shift of NRS RNAs, which is dependent on the p55 binding site. A, the indicated amounts of hnRNP H were incubated with 2.5 fmol of either 4Z RNA or NRS5Ј RNA, and electrophoretic mobility shift assays were performed as described in Fig. 5. The positions of free RNA and the shifted complex seen with NRS5Ј are indicated. B, electrophoretic mobility shift assays were performed as in Fig. 5 using either 5 g of HeLa nuclear extract (NE) or 30 pmol of recombinant hnRNP H (H) in the presence or absence of hnRNP H antiserum (Ab). RNAs were as described in Fig. 2 with the exception that NRS3Ј RNA was transcribed from p3ZBB3Ј (nt 798 -932). The positions of free RNA, shifted complexes, and the corresponding supershifted complexes (*) are indicated. Note that lanes 1-18 and 19 -31 are derived from separate gels. The figures were generated as described for Fig. 2. both splicing and polyadenylation defects, it is possible that hnRNP H may participate in the coordination of the splicing and polyadenylation machinery on the viral RNAs. We are currently conducting experiments to explore these possibilities to more clearly define the role of hnRNP H in NRS-mediated splicing inhibition.