Heterogeneous Ribonucleoprotein M Is a Splicing Regulatory Protein That Can Enhance or Silence Splicing of Alternatively Spliced Exons*

Splicing of fibroblast growth factor receptor 2 (FGFR2) alternative exons IIIb and IIIc is regulated by the auxiliary RNA cis-element ISE/ISS-3 that promotes splicing of exon IIIb and silencing of exon IIIc. Using RNA affinity chromatography, we have identified heterogeneous nuclear ribonucleoprotein M (hnRNP M) as a splicing regulatory factor that binds to ISE/ISS-3 in a sequence-specific manner. Overexpression of hnRNP M promoted exon IIIc skipping in a cell line that normally includes it, and association of hnRNP M with ISE/ISS-3 was shown to contribute to this splicing regulatory function. Thus hnRNP M, along with other members of the hnRNP family of RNA-binding proteins, plays a combinatorial role in regulation of FGFR2 alternative splicing. We also determined that hnRNP M can affect the splicing of several other alternatively spliced exons. This activity of hnRNP M included the ability not only to induce exon skipping but also to promote exon inclusion. This is the first report demonstrating a role for this abundant hnRNP family member in alternative splicing in mammals and suggests that this protein may broadly contribute to the fidelity of splice site recognition and alternative splicing regulation.

Splicing of fibroblast growth factor receptor 2 (FGFR2) alternative exons IIIb and IIIc is regulated by the auxiliary RNA ciselement ISE/ISS-3 that promotes splicing of exon IIIb and silencing of exon IIIc. Using RNA affinity chromatography, we have identified heterogeneous nuclear ribonucleoprotein M (hnRNP M) as a splicing regulatory factor that binds to ISE/ ISS-3 in a sequence-specific manner. Overexpression of hnRNP M promoted exon IIIc skipping in a cell line that normally includes it, and association of hnRNP M with ISE/ISS-3 was shown to contribute to this splicing regulatory function. Thus hnRNP M, along with other members of the hnRNP family of RNA-binding proteins, plays a combinatorial role in regulation of FGFR2 alternative splicing. We also determined that hnRNP M can affect the splicing of several other alternatively spliced exons. This activity of hnRNP M included the ability not only to induce exon skipping but also to promote exon inclusion. This is the first report demonstrating a role for this abundant hnRNP family member in alternative splicing in mammals and suggests that this protein may broadly contribute to the fidelity of splice site recognition and alternative splicing regulation.
Alternative splicing represents an important mechanism whereby a single gene transcript can give rise to numerous spliced mRNAs, thereby greatly expanding the ribonomic and proteomic diversity that can be obtained from a limited gene number (1,2). Despite increasing recognition that the majority of metazoan gene transcripts are subject to alternative splicing, the molecular mechanisms that regulate this process remain poorly understood. This is particularly true for alternatively spliced gene transcripts that display tightly regulated expression of distinct splice variants in different cell types or at different stages of development. The constitutive splicing process is directed by consensus sequences at the 5Ј and 3Ј splice sites as well as a branch point sequence generally located 20 -40 nucle-otides (nt) 2 upstream of the 3Ј splice site. However, these sequence determinants are not sufficient to accurately specify the actual splice sites that are used for splicing of constitutive as well as alternative exons. In recent years, numerous studies have demonstrated an important role for an additional layer of cis-acting control sequences often referred to as auxiliary ciselements (2,3). These sequences have been identified in both exons and introns and can have either positive or negative effects on splicing of a given exon. As such they are commonly referred to as exonic splicing enhancers or silencers and intronic splicing enhancers or silencers (ISEs or ISSs). These elements function largely through the activity of RNA-binding proteins that associate with them and influence the efficiency with which nearby splice sites are recognized by the basal splicing machinery. Most alternatively spliced exons are associated with multiple auxiliary cis-elements that can positively or negatively affect its inclusion. These observations have led to models of combinatorial control, in which the splicing outcome is determined by the net effect of several cis-elements that interact with differentially expressed RNA-binding proteins in varied cellular milieus (1,2,4). Although several cell-type-specific factors have previously been described, differential expression of such factors alone has not proven sufficient to account for cell-type-specific splicing decisions, which also involve contributions from ubiquitously expressed regulatory factors. Thus a challenge in the field of splicing regulation has been to determine the different contributions of numerous cis-elements and RNA-binding proteins and how their combined functions yield distinct splicing patterns in different cells.
Our studies have focused on alternative splicing of fibroblast growth factor receptor 2 (FGFR2) transcripts. Mutually exclusive splicing of two exons, IIIb and IIIc, gives rise to two functionally different receptors, FGFR2-IIIb and FGFR2-IIIc, in epithelial and mesenchymal cells, respectively. These exons encode the C-terminal half of an Ig-like domain in the receptor's extracellular domain in a region that determines ligandbinding specificity. Consequently, the two different receptor isoforms display distinct binding preferences for the FGF family of ligands (5,6). Several auxiliary cis-elements that have been shown to play a role in regulation of these alternative exons are * This work was supported by National Institutes of Health Grant CA093769.
shown in Fig. 1A. Although several abundant and ubiquitously expressed factors, including polypyrimidine tract-binding protein, TIA-1, and hnRNP A1, have been shown to bind to several of these elements and play a role in splicing regulation, the mechanism by which cell-type-specific alternative splicing of this transcript occurs has not been fully elucidated (7)(8)(9)(10)(11)(12)(13)(14). We have focused our recent studies on an element we have termed ISE/ISS-3, so named because it has been shown to promote splicing of exon IIIb (as an ISE) and at the same time to repress splicing of exon IIIc (as an ISS) in epithelial cells (9,15). Thus in cells that express FGFR2-IIIb (e.g. DT3 cells) deletion of this element in transfected FGFR2 minigenes results in a partial switch from exon IIIb splicing to exon IIIc splicing, whereas the exclusive inclusion of exon IIIc in cells that express FGFR2-IIIc (e.g. AT3 or 293T cells) is unaffected by the same deletion. We have also shown that this sequence displays epithelial cell-typespecific functions in a heterologous minigene context, suggesting that it may associate with at least one regulatory factor whose expression is limited to cells expressing FGFR2-IIIb (15). In the present study we describe one of the proteins, heterogeneous ribonucleoprotein M (hnRNP M), that binds directly and specifically to the wild-type ISE/ISS-3 sequence. In a cell line that normally includes exon IIIc (293T), we determined that overexpression of hnRNP M induced exon IIIc skipping, suggesting that this factor plays a role in silencing of this exon in cells that express FGFR2 IIIb. Transfection studies using minigenes containing several unrelated alternative exons showed that hnRNP M can induce either exon skipping or exon inclusion in different contexts. Therefore, we have identified hnRNP M as a factor involved in FGFR2 splicing regulation, and we propose that this factor may also play a more general role in the regulation of diverse alternatively spliced metazoan gene transcripts.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The plasmids and minigenes used in this study were constructed using standard cloning techniques. pI-11-FS, pI-11-FS-CXS, and pI-11-FS-CXS-IIIb Mut minigenes and their derivatives were previously described (9,16). The plasmid minigene containing the R35 (28) exon was generated by PCR of the R35 (28) plasmid (kindly provided by Thomas Cooper) using primers TNT-Xba-F (5Ј-GGCCTCTAGAGTCGAAGTCCT-CACCTGGTG-3Ј) and TNT-Not-R (5Ј-GCGCGGCCGCAC-TAGTTTAGAGGGGAGAGCGC-3Ј). This amplified a SalI to SpeI segment from the original construct that contained the exon and cTNT-derived intronic sequences (17). This fragment was inserted into the XbaI and NotI sites of the pI-11(-H3)-PL vector (9), thereby positioning the exon between two adenoviral derived flanking exons. The preprotachykinin minigene pPPT-106 (kindly provided by Gil Cote) was previously described (18). For in vitro transcription we modified the pDP19 vector (Ambion) by inserting the sequence 5Ј-GAAT-TATCGATCTCGAG-3Ј between the EcoRI and HindIII sites to generate pDP19RC-⌬EE. This modification permitted insertion of ClaI/XhoI fragments containing ISE/ISS-3 (or control sequences) under the control of the T7 promoter for in vitro transcription. To obtain the 3ϫWT and 3ϫMT inserts, the two pairs of primers were kinased and annealed: 43nt-ISE3-Sal-Xho-F (5Ј-TCGACGTGTGGTGATGGGCCTGCAGAGGT-GAGCTGGCCGGTGTCTCTCC-3Ј) with 43nt-ISE3-Sal-Xho-R (5Ј-TCGAGGAGAGACACCGGCCAGCTCACCTC-TGCAGGCCCATCACCACACG-3Ј), and 43nt-3xGU Mut-Sal-Xho-F (5Ј-TCGACGTGTGACGATGGGCCTGCAGAG-ACGAGCTGGCCGACGTCTCTCC-3Ј) with 43nt-3xGU Mut-Sal-Xho-R (5Ј-TCGAGGAGAGACGTCGGCCAGCTC-GTCTCTGCAGGCCCATCGTCACACG-3Ј). The annealed oligonucleotides were ligated, digested by XhoI and SalI, and separated on polyacrylamide gels to select three tandem copies of the sequence. Thus we obtained two new constructs, one containing three repeats of wild-type 43-nt sequence from ISE/ ISS-3 (3ϫWT) and another containing three repeats of mutant 43-nt sequence from ISE/ISS-3 in which three GU-AC substitutions occurred in each repeat (3ϫMT) flanked by ClaI and XhoI sites (Fig. 1C). To create the templates for pulldown assays PCR of these sequences was performed to introduce a ClaI site at the 5Ј-end of the sequence, and the products were re-cloned as ClaI/XhoI fragments into the DP19RC-⌬EE plasmid. Similarly, the previously described "full-length" ISE/ISS-3, 5Ј-AC, and BS fragments were also inserted in pDP19RC-⌬EE (Fig.  1B). To make the construct for expression of hnRNP M, pIN-⌬Int-hnRNP M, an EcoRV-AgeI fragment from the pHCM4 construct (kindly provided by Maurice Swanson) was inserted in the previously described pIRESneo3-⌬Int vector (15,19). Generation of N-terminal FLAG-tagged hnRNP M was carried out by inserting the same EcoRV-AgeI fragment into the FLAG expression vector pINX-N-FF-B. The latter construct was derived by insertion of sequences encoding two tandem copies of the FLAG tag upstream of the multicloning site of pIRESneo3⌬Int (sequence available on request). All plasmid constructs were prepared with Qiagen MidiPrep kits. Sequences were confirmed by the University of Pennsylvania Sequencing Facility.
UV Cross-linking-To prepare RNA probes for UV-crosslinking experiments we performed in vitro transcription with T7 RNA polymerase (Ambion) after digestion of the pDP-based plasmids with XhoI using the manufacturer's recommendations. The specific activity of the 169-nt RNA transcript was 2.7 ϫ 10 13 cpm/mol. 500,000 cpm, or 19 fmol, of [ 32 P]UTPradiolabeled RNA substrates were incubated at 30°C with 12-20 g of KATO III or 293T cell nuclear extracts in a volume of 20 l for 20 min in 1.5-ml Eppendorf tubes. The nuclear extracts from 293T cell transiently expressing the FLAGtagged plasmids were generated by transfection of 5 g of plasmid in 60-mm plates using TransIT-293 Transfection Reagent (Mirus) in accordance with manufacturer recommendations. 48 h after transfection, nuclear extracts were prepared following a previously described protocol (20). After incubation, the tubes were opened, placed on ice, and UV-irradiated in a Stratalinker 2400 (Stratagene) at a distance of 5 cm with an energy of 6000 J ϫ 100. After addition of 2 l of a stock RNase mix (10 mg/ml RNase A and 10 units/l RNase T1) the samples were incubated at 37°C for 60 min. Reactions were stopped with 4ϫ LDS sample buffer (Invitrogen) and loaded onto 10% NuPAGE Bis-Tris gel (Invitrogen). The gel was run at constant 200 V for 60 -70 min, then fixed, dried, and visualized by radiography.

hnRNP M Is a Splicing Factor
Immunoprecipitation of UV Cross-linked Proteins-Immunoprecipitation of UV-cross-linked proteins was carried out following the RNase step of the UV cross-linking protocol. Precipitation of endogenous hnRNP M from KATO III nuclear extracts was performed by first pre-clearing the UV-crosslinked extracts by addition to 30 l of pre-washed Protein G-agarose (Roche Applied Science) in 450 l of NET-2 buffer (50 mM Tris-Cl, pH 7.6, 150 mM NaCl, 0.01% Nonidet P-40, 1ϫ protease inhibitor (Sigma)) at 4°C and rotation for 30 min. The pre-cleared supernatant was then harvested after centrifugation at 1000 ϫ g for 1 min. 1.5 l of hnRNP M antibody (Santa Cruz Biotechnology, Santa Cruz, CA) was added to the precleared extract and rotated at 4°C for 60 min and then added to another 30 l of pre-washed protein G-agarose and incubated for an additional 60 min. The beads were then washed three times with 500 l of NET-2 buffer, and the precipitated proteins were eluted by addition of 20 l of 2ϫ sample loading buffer and boiling for 5 min. The mix was transferred to a Bio-Rad microspin column, centrifuged at 500 ϫ g for 5 min, electrophoresed on 10% NuPAGE Bis-Tris gels, and analyzed by Western blotting. Immunoprecipitation of FLAG-tagged hnRNP M from transiently transfected 293T cell nuclear extracts was carried out by addition of the UV-cross-linked extracts to 40 l of pre-washed anti-FLAG M2-agarose slurry (Sigma) in 400 l of cold NET-2 buffer. After rotation at 4°C for 60 min, the beads were washed twice with 400 l of NET-2 buffer and eluted by incubating for 30 min with rotation at 4°C with 10 g of 3ϫ FLAG peptide in 60 l of NET-2 buffer. The eluted proteins were harvested using Bio-Rad microspin columns and analyzed by Western as described below.
Affinity Purification-To obtain microgram quantities of unlabeled RNA for affinity chromatography in vitro transcription reactions were performed in a 50-l volume containing 0.1 g/l linearized DNA template, 2 units/l T7 RNA polymerase (Ambion), 7.5 mM NTPs, 0.1 unit/l inorganic pyrophosphatase (Sigma), 0.4 unit/l SUPERase-In (Ambion) in buffer containing 80 mM Hepes-KCl, pH 7.5, 24 mM MgCl 2 , 2 mM spermidine, and 40 mM dithiothreitol. The reaction mix was incubated overnight at 37°C and RNA was gel-purified from 10% denaturing acrylamide gels. Pulldown assays were performed similar to previously described protocols (21,22), with the following modifications. A 200-l reaction mixture containing 1 nmol of RNA, 100 mM NaOAc, pH 5.0, and 20 mM sodium m-periodate (Sigma) was incubated in the dark at room temperature for 1 h. The oxidized RNA was precipitated with ethanol and resuspended in 500 l of 100 mM NaOAc, pH 5.0. 200 l of adipic acid dehydrazide bead slurry (100 l of packed beads, Sigma) were washed twice with 1 ml of 100 mM NaOAc, pH 5.0. The resuspended RNA was added to the beads and rotated overnight at 4°C. The beads with bound RNA were then washed twice with 1 ml of 2 M NaCl and three times with 1 ml of Buffer D (23) freshly supplemented with 0.5 mM dithiothreitol and 1ϫ phenylmethylsulfonyl fluoride. A 1-ml final volume of a mix containing 500 l of KATO III nuclear extract in the presence of 1.6 mM MgCl 2 , and 5 mg/ml heparin was added to the washed beads. After rotating for 30 min at room temperature the beads were washed four times with 1 ml of buffer D supplemented with 2 mM MgCl 2 and 75 g/ml hepa-rin. The proteins were eluted by addition of 60 l of 1.5ϫ SDS loading buffer and heating at 90°C for 5 min. Beads and buffer were transferred to Handee spin columns (Pierce) and centrifuged at 1000 ϫ g for 1 min. Eluted samples were loaded on 4 -12% NuPAGE Bis-Tris gel (Invitrogen). The gel was stained using either SilverSNAP (Pierce) or Colloidal Blue staining kit (Invitrogen), and the bands of interest were excised and submitted for analysis using tandem mass spectrometry performed by the University of Pennsylvania Proteomics Core Facility.
Western Blotting-To detect antibodies against hnRNP M we used the standard technique described previously (24), with modifications. The nitrocellulose filter was blocked with 2.5% Carnation milk overnight at 4°C. The primary antibody was used at a 1:200 dilution for mouse monoclonal hnRNP M1-4 (Santa Cruz Biotechnology) or 1:1,000 dilution for monoclonal anti-FLAG M2 (Sigma). The secondary anti-mouse Ig, horseradish peroxidase-linked whole antibody (Amersham Biosciences) was used at a 1:10,000 dilution.

Identification of Sequences in the 5Ј Half of ISE/ISS-3 That Are Sufficient to Promote Exon IIIb Inclusion and Exon IIIc Skipping in Cells That
Express FGFR2-IIIb-Using stable transfection minigene assays, we previously carried out extensive mutational analysis of an 85-nt sequence constituting ISE/ISS-3 and determined that mutations or deletions within the 5Ј-end of the sequence were more detrimental to its function than alterations at the 3Ј-end (Fig. 1B) (9,15). These experiments were carried out using an FGFR2 minigene, pI-11-FS-CXS, that contains both exons IIIb and IIIc and their flanking introns, but in which ISE/ISS-3 is deleted and replaced by two convenient restriction enzymes ( Fig. 2A). Transfection of the DT3 cell line, which expresses endogenous FGFR2-IIIb, with pI-11-FS-CXS demonstrated predominant splicing of exon IIIc (Fig. 2B, lanes 1-3). Analysis of results with these minigenes, performed by RT-PCR using primers complementary to sequences in the exons flanking exons IIIb and IIIc, revealed a product of 386 or 383 bp depending on whether exon IIIb or exon IIIc was spliced, respectively. Because this size difference is difficult to distinguish, we digest these RT-PCR products separately with AvaI or HincII, which digest products containing exon IIIb or IIIc, respectively. Comparison of the fraction of products digested with these enzymes can therefore be used to determine the percentage of products that contain exon IIIb or IIIc. Transfection of a minigene in which ISE/ISS-3 has been restored to its normal position results in predominant splicing of exon IIIb and skipping of exon IIIc (Fig. 2B, lanes 4 -6). Insertion of three tandem copies of the 43 nt that comprise the 5Ј half of ISE/ISS-3 (3ϫWT) (Fig. 1C) likewise restores predominant exon IIIb inclusion (Fig. 2B, lanes 7-9). However, insertion of the same tandem copies containing the GU to AC mutations shown in Fig. 1C (3ϫMT) yields predominant exon IIIc splicing in transfected DT3 cells (Fig. 2B, lanes 10 -12). Importantly, these results demonstrating the function of ISE/ISS-3 in promoting exon IIIb inclusion and exon IIIc skipping are only observed in cells that express endogenous FGFR2-IIIb. Transfection of these minigenes into cells that express endogenous FGFR2-IIIc (e.g. AT3 and 293T) results in exclusive splicing of exon IIIc, and this outcome is observed in either the presence or absence of ISE/ISS-3 (data not shown).

Identification of hnRNP M as an ISE/ISS-3-binding Pro-
tein-Given the important role of ISE/ISS-3 in mediating celltype-specific splicing of FGFR2 exon IIIb in lieu of exon IIIc, we carried out biochemical assays to identify factors that bound specifically to this regulatory element. To characterize protein factors that bind ISE/ISS-3 we initially carried out in vitro UVcross-linking studies using radiolabeled RNAs containing the three tandem copies of the 5Ј half of ISE/ISS-3 (3ϫWT), the same tandem copies containing mutations (3ϫMT), or unrelated control RNA sequences. For our studies, we used nuclear extracts prepared from a cell line, KATO III, that also expresses FGFR2-IIIb as such extracts are expected to express the full complement of factors required for expression of this isoform, including potential cell-type-specific factors (9). After UV-cross-linking and RNase digestion, SDS-PAGE analysis was carried out to identify the molecular masses of proteins that cross-linked to the 3ϫWT probe, but not to the 3ϫMT control. As shown in Fig. 3A, several proteins that cross-linked specifically to the 3ϫWT RNA were identified. These included bands of ϳ110, 80, 70, 55, 50, and 20 kDa. To identify these proteins, we performed RNA affinity chromatography by covalently linking the same RNAs to adipic acid dihydrazide-agarose beads. After binding with KATO III nuclear extracts, the beads were

hnRNP M Is a Splicing Factor
washed, and bound proteins were eluted and analyzed on silverstained SDS-PAGE gels. Although a number of proteins bound both the wild-type and mutant RNAs, several bands corresponding in size to those seen by cross-linking were noted to be more abundant in eluates from the wild-type ISE/ISS-3 RNA compared with the mutant or bead-alone controls. Among the latter were several proteins in the 70-to 80-kDa size range. This region of the gel was excised and submitted for tandem mass spectrometry (Fig. 3B). To further characterize those proteins that specifically bound the wild-type sequence, the same region of the gel containing proteins bound to the mutant control RNA was also excised and analyzed. Mass spectrometry analysis identified hnRNP M as the protein that matched the greatest number of peptides from the wild-type sequence with a total of 239 peptides that could be matched to the protein sequence. In contrast, analysis from the mutant eluates identified only 28 peptides that matched hnRNP M.
To confirm binding of hnRNP M to the full-length ISE/ISS-3, we performed Western immunoblotting of proteins bound to this RNA sequence using hnRNP M monoclonal antibodies (25). These studies confirmed that greater quantities of hnRNP M were present in eluates from the column containing the 3ϫWT RNA than the 3ϫMT RNA and no binding to the beads alone could be detected (Fig. 3C, lanes  2-4). Furthermore, hnRNP M binding to the full-length ISE/ISS-3 sequence was equivalent to that observed with the 3ϫWT sequence (Fig. 3C, lanes 1 and 2). Of note, these antibodies recognized four protein bands representing hnRNP M. These bands (M1-M4) appear to represent different alternatively spliced isoforms of the protein (25). We also performed Western immunoblotting using eluates from the full-length ISE/ ISS-3 together with another previously described mutant, 5Ј-AC, as well as an unrelated size-matched RNA, BS, that lack splicing regulatory function (15). Once again, hnRNP M binding to the mutant or to the control RNA was significantly less than that observed with the wild-type element (Fig. 3D, lanes 1-3). Interestingly, hnRNP M was previously shown to bind to poly(G) and poly(U), but not poly(A) or poly(C), homopolymers (25,26). Our results are therefore consistent with G-and U-rich sequences in the 5Ј-end of ISE/ISS-3 as comprising a site for hnRNP M binding that is disrupted when substituted by A and C nucleotides.
hnRNP M Binds Directly to ISE/ISS-3-Although the results using RNA affinity columns showed that hnRNP M bound specifically to ISE/ISS-3 as well as the 3ϫWT sequence, these results are not sufficient to prove that the interaction is direct. It is possible that hnRNP M may interact with ISE/ISS-3 indirectly through another RNA-binding protein. Furthermore, given the abundance of hnRNP proteins in nuclear RNP complexes, an indirect association via another protein could also be mediated by RNAs bound by both proteins. Because RNA-protein cross-linking by ultraviolet light is accepted as evidence of a direct association, we immunoprecipitated the cross-linked extracts with anti-hnRNP M antibodies to determine whether hnRNP M was indeed one of the proteins seen by direct UVcross-linking (27). Immunoprecipitation of proteins crosslinked to wild-type ISE/ISS-3 (WT) in KATO III nuclear extracts with the monoclonal hnRNP M antibody revealed a band that corresponded in size to hnRNP M and one of the proteins seen by direct UV-cross-linking (Fig. 4, lanes 1 and 3). This immunoprecipitated band was not observed with the 5Ј-AC mutant sequence (5Ј-AC) or with uncoupled protein G beads (Fig. 4, lanes 4 -6). However, because immunoprecipitation with the hnRNP M antibody was very inefficient, this band was faint even after prolonged film exposure. For this reason we   DECEMBER 14, 2007 • VOLUME 282 • NUMBER 50 sought more robust, independent confirmation that hnRNP M directly cross-linked to ISE/ISS-3. We therefore transfected 293T cells with a plasmid construct that expressed hnRNP M containing an N-terminal FLAG epitope tag and harvested nuclear extracts from these cells for UV-cross-linking. Western blotting confirmed efficient expression of the FLAG-tagged protein in nuclear extracts from transiently transfected 293T cells at levels equal to or exceeding those of the endogenous protein (Fig. 4B, note slower mobility of the tagged protein). Nuclear extracts from cells expressing FLAG-hnRNP M and the empty FLAG vector (FLAG-EV) control were UV-crosslinked with wild-type ISE/ISS-3 (WT) and 5Ј-AC mutant RNAs and immunoprecipitated with anti-FLAG affinity resin. Western blotting with hnRNP M antibody demonstrated that nearly all of the FLAG-tagged hnRNP M protein was efficiently immunoprecipitated, whereas most endogenous hnRNP M was present in supernatants (Fig. 4C). Direct UV cross-linking analysis of the same extracts demonstrated an ϳ80-kDa protein as observed in KATO III extracts. In addition to being specific for the wild-type ISE/ISS-3, the band appeared slightly more intense in the cells transfected with hnRNP M compared with the empty vector control, as might be expected given the higher level of hnRNP M in these extracts. Direct binding of hnRNP M was demonstrated by immunoprecipitation of the cross-linked FLAG-tagged protein with anti-FLAG resin. This cross-linked product was only observed with the wild-type ISE/ ISS-3, again showing the sequence specific nature of the binding (Fig.  4D, lanes 5 and 6). Furthermore, the immunoprecipitated/cross-linked protein was not observed in cells that did not express the tagged hnRNP M construct, confirming that it represented a cross-link between hnRNP M and ISE/ISS-3 (Fig. 4D, lane 7). Thus, taken together the results in Fig. 4 strongly support a direct and sequence-specific interaction between hnRNP M and ISE/ISS-3.

hnRNP M Binding to ISE/ISS-3 Plays a Role in FGFR2 Exon IIIc
Silencing-Given the dual role of ISE/ISS-3 in inducing exon IIIb inclusion and exon IIIc skipping, we first performed transient co-transfection experiments using plasmids expressing the cDNA for hnRNP M and an FGFR2 reporter minigene, pI-11-FS, that contains both exons IIIb and IIIc as well as wild-type ISE/ ISS-3 (Fig. 5A). For these experiments we used 293T cells, because these cells have been shown to express endogenous FGFR2-IIIc, but to show an increase in exon IIIb splicing and a decrease in exon IIIc splicing in response to exogenous overexpression of splicing factors, such as TIA-1 or Fox proteins (10,11,13,19). When the FGFR2 minigene was co-transfected with the empty expression vector and analyzed by RT-PCR using primers complementary to sequences in the exons flanking exons IIIb and IIIc, we noted a single 283-bp band that represented nearly exclusive splicing of exon IIIc (Fig. 5B, lanes 1-3). The same band was also observed when we transfected the same plasmid expressing increasing amounts of hnRNP M, and again the product represented nearly exclusive splicing of exon IIIc (Fig.  5B, lanes 4 -9, U-B/C-D product). However, in this case we noted a dose-dependent increase in the amount of an RT-PCR product of 138 bp, which we have previously determined to represent a product in which neither exon IIIb or IIIc is spliced, but the flanking exons are directed ligated (Fig. 5B,  U-D product). This result suggested that, although hnRNP M binds specifically to the wild-type ISE/ISS-3 sequence and plays a role in exon IIIc silencing, other ISE/ISS-3-binding  3 and  4) or protein G beads alone (lanes 5 and 6). The position of a cross-linked ϳ80 kDa band that was only immunoprecipitated with the WT RNA, and the hnRNP M antibody is indicated by the arrow. B, nuclear extracts from 293T cells transiently transfected with either the empty FLAG vector (FLAG-EV) or FLAG-tagged hnRNP M (FLAG-hnRNP M) analyzed by Western immunoblotting with either anti-FLAG or anti-hnRNP M antibodies. Note that the FLAG-tagged protein has slower mobility that distinguishes it from endogenous hnRNP M and that the protein levels of each are roughly similar. C, Western blotting of 293T nuclear extracts following immunoprecipitation with anti-FLAG affinity resin. Immunoprecipitated material is designated "pellet" (lanes 1 and 2) and indicates material bound to the column and eluted with 3ϫ FLAG peptide after column washes. Supernatants (SN) of unbound proteins are in lanes 3 and 4. D, UV-cross-linking results in 293T cells transfected with the indicated plasmids and incubated with the WT or 5Ј-AC RNAs (lanes 1-4). The arrow indicates an ϳ80 kDa cross-linked band seen only with the WT RNA and which is also more evident in extracts transfected with FLAG-tagged hnRNP M. A cross-linked band of the same mobility (lane 5) is only seen following immunoprecipitation with anti-FLAG affinity resin from cells transfected with FLAG-tagged hnRNP M (and not the empty FLAG vector control) and only with the wild-type RNA.

hnRNP M Is a Splicing Factor
proteins are required to carry out its role in activating exon IIIb splicing.
To further determine the role of ISE/ISS-3 in mediating the function of hnRNP M in silencing exon IIIc, we performed cotransfections using FGFR2 minigenes that contained the wildtype ISE/ISS-3, or control minigenes that did not. To simplify our analysis, we used a previously described minigene, pI-11-FS-CXS-IIIb Mut, in which the splice sites flanking exon IIIb were mutated to preclude its splicing, thus facilitating more direct analysis of exon IIIc inclusion versus skipping (Fig. 6A) (9,15). This construct was used to insert the same full-length ISE/ ISS-3, which was shown to bind hnRNP M by Western analysis as well as the 5Ј-AC mutant and unrelated RNA sequence (BS). With each minigene, the amount of IIIc skipping was quantified both when co-transfected with an empty vector control and with the same plasmid containing the hnRNP M cDNA. This analysis revealed that induction of exon IIIc skipping by hnRNP M did not absolutely require ISE/ISS-3, but that insertion of ISE/ISS-3 into the original minigene enhanced the amount of exon IIIc skipping (Fig. 6, B and C, compare lanes 1 and 2 to  lanes 3 and 4). Furthermore, the degree of exon skipping induced by hnRNP M was decreased when the ISE/ISS-3 ele-ment contained the 5Ј-AC mutations or was replaced by the unrelated control sequence (Fig. 6, B and 6C, lanes 5-8). Thus, these experiments demonstrated that hnRNP M binding to ISE/ISS-3 promotes exon IIIc skipping, but also strongly suggested that this effect on exon silencing also involves binding at other sites flanking this alternative exon. We did note that the baseline level of exon IIIc silencing obtained with the 5Ј-AC mutant sequence was somewhat higher than that with the other minigenes. This may reflect some serendipitous effect of inserting a "new" sequence in this position and is a frequent issue that can arise when engineering sequence alterations into minigenes as   DECEMBER 14, 2007 • VOLUME 282 • NUMBER 50 JOURNAL OF BIOLOGICAL CHEMISTRY 36271 control sequences are often not "splicing-neutral" (28). Nonetheless, despite this already increased level of exon skipping, the final amount of skipping induced by hnRNP M is less than that achieved in the presence of the wild-type element, thus clearly supporting a contribution of hnRNP M binding to this element.

hnRNP M Is a Splicing Factor
hnRNP M Functions to Promote Both Exon Inclusion and Exon Skipping of Other Alternatively Spliced Exons-As noted above, induction of exon IIIc skipping by hnRNP M also occurred, albeit less efficiently, in the absence of ISE/ISS-3 suggesting that it also binds to other sites within the FGFR2 transcript to influence splicing. Binding at multiple sites within regulated transcripts has been observed for a number of splicing regulators that have been shown to have general roles in splicing of diverse transcripts (1,2,8,29). We therefore considered the possibility that hnRNP M might exert a more general function within the cell to silence, or otherwise regulate alternative exon splicing. To investigate this possibility further we performed co-transfection experiments with several unrelated minigene constructs. We tested a plasmid minigene construct, pPPT 106, which contains a genomic fragment containing exons three through five of the rat preprotachykinin (PPT) gene (18). Exon four of this transcript is alternatively spliced, with skipping predominant in rat brain (30). When we co-transfected this minigene with increasing amounts of the plasmid for hnRNP M, we noted a dose-dependent increase in PPT exon four skipping (Fig. 7A). Similarly, in the case of FGFR1, we also noted an increase in skipping of the ␣-exon inclusion, but in this case the effect of hnRNP M was more modest (data not shown). To further investigate whether hnRNP M plays a general role as a splicing silencer, we tested several additional minigene constructs. Surprisingly, we discovered that in contrast with the previous results where it promoted exon skipping, hnRNP M promoted exon inclusion of an alternative exon termed R35 (28), which is flanked by cardiac troponin T mus-cle-specific intronic splicing enhancer elements (17,18). As shown in Fig. 7B, co-transfection of a minigene containing the R35 (28) exon with the hnRNP M construct revealed a dose-dependent increase in the level of exon inclusion (note that the results are shown here as exon inclusion). Thus, our results suggest that hnRNP M can affect the splicing of a number of alternatively spliced exons, consistent with a broader role in global splicing regulation. Interestingly, however, depending on the context in which it is studied, this factor can either positively or negatively influence exon definition.

DISCUSSION
We have identified hnRNP M as a factor that binds directly and specifically to the ISE/ISS-3 auxiliary cis-element. We have previously shown that this element functions to regulate splicing only in cell types that express FGFR2-IIIb and that the element functions both to activate splicing of exon IIIb and to silence exon IIIc. However, although we show that hnRNP M does indeed play a role in exon IIIc silencing, we did not find evidence that it can increase exon IIIb inclusion using the assays described here. Thus, it is likely that other factors that also bind to ISE/ISS-3 may be involved in exon IIIb activation and that the coordinated activities of several factors together achieve these dual functions. A straightforward mechanism by which the ISE/ISS-3 element might coordinate these activities would be through the function of a factor whose expression was restricted to epithelial cell types. For this reason, we used extracts from a relevant cell type that expresses FGFR2-IIIb for these experiments to identify regulatory factors that bind this element. However, expression of hnRNP M was not noted to be significantly different between cells expressing FGFR2-IIIb or FGFR2-IIIc. Therefore, further identification and characterization of other factors that associate with the element will be necessary to clarify the mechanism by which cell-type-specific regulation is achieved.
Our experiments show that binding of hnRNP M to ISE/ ISS-3 affects the efficiency of exon IIIc silencing as mutations or substitutions of the element to yield sequences that abrogate hnRNP M binding reduce the degree of exon IIIc skipping. However, it is also clear that exon IIIc silencing can occur in the absence of this element. Thus, binding of hnRNP M to this element alone only contributes a fraction of the protein's total silencing activity on exon IIIc. The most likely mechanism by which hnRNP M plays a role in exon IIIc skipping is therefore likely to involve binding at multiple sites flanking or within exon IIIc. Such a requirement of multiple binding sites for splicing regulators to achieve maximal effects on splicing has been observed in many cases and appears to be a common theme in splicing regulation (1,31). For example, the ability of other factors that have been shown to promote efficient exon skipping such as hnRNP A1 and polypyrimidine tract-binding protein (hnRNP I) involves binding at multiple sites on both sides of the regulated exon (8,29,32). Further characterization of other sites within the FGFR2 transcript that are also bound by hnRNP M should provide further insights into the mechanism by which it promotes exon IIIc skipping.
We also determined that hnRNP M can influence the splicing of several unrelated alternative exons. In the case of prepro- tachykinin exon 4 and the ␣-exon of FGFR1 it promoted exon skipping as we observed for exon IIIc. Conversely, hnRNP M promoted inclusion of a troponin-derived exon flanked by muscle specific ISEs. Thus, unlike other well described hnRNP proteins, such as hnRNP A1 and polypyrimidine tract-binding protein, which have been shown to silence splicing of a broad array of alternative exons, the repertoire of functions of hnRNP M includes the ability to both positively and negatively influence exon definition (29,32). In this regard, the function of hnRNP M is similar to the splicing regulators Nova-1, Fox-1, CUG, and Etr-like factors (CELFs), and hnRNP L, which have likewise been shown to promote exon inclusion or exclusion, depending on the context. For these splicing regulatory proteins, the position and proximity of their binding sites relative to regulated exons have been shown to determine which of these activities predominates (31,(33)(34)(35)(36). Interestingly, members of the CELF family of splicing regulatory proteins were previously shown to activate exon R35 (28) inclusion through binding the musclespecific ISE elements flanking the R35 (28) element (17). In fact, the ability of hnRNP M to activate exon R35 (28) inclusion was nearly equivalent to that achieved in a co-transfection with a plasmid expressing CELF 6 (data not shown). Thus, these results suggest that, in addition to the CELFs, hnRNP M may also contribute to the regulation of muscle-specific exons. In any case, the fact that hnRNP M affected the splicing of each minigene we tested suggests that, similar to other well characterized splicing factors, it may play a general role in alternative splicing of numerous transcripts.
The present manuscript is the first demonstration of which we are aware in which hnRNP M has been shown to regulate splicing in metazoans. Furthermore, this is the first report in which binding to a specific auxiliary cis-element has been shown to directly correlate with the ability of hnRNP M to regulate splicing in any organism. However, potential roles for hnRNP M in constitutive as well as regulated splicing have been suggested by several previous studies in diverse species. hnRNP M was originally identified as an abundant member of a core set of 24 polypeptides present in immunopurified hnRNP complexes (37). Among these proteins, hnRNP M was shown to associate with single-stranded DNA in a heparin-resistant manner, consistent with binding to RNA as a single-stranded nucleic acid-binding protein. Like other members of the core hnRNP family, hnRNP M is characterized by several copies of the classic RNA recognition motif that mediate RNA binding, as well as an unusual methionine and arginine rich motif, the MR repeat motif (25). The roles of each of the hnRNP proteins in splicing are not entirely clear for each of these abundant RNA-binding proteins. However, hnRNP M is among several that are not only present in the pre-spliceosomal H complexes that assemble in vitro, but also in mature spliceosomes. In fact several studies, including comprehensive proteomic analyses of in vitro purified spliceosomes have shown that hnRNP M is present not only in H complex, but stays associated with RNA transcripts through the stepwise progression of spliceosomal assembly from E/CC to A, B, and the relatively mature C complex (38 -41). Further support for a role of hnRNP M in splicing was provided by studies showing that antibodies against the protein inhibited splicing in vitro (42). This study also noted that heat shock-induced inhibition of splicing was associated with a loss of hnRNP M from spliceosomal complexes, which may further implicate it in constitutive splicing.
The suggestion that hnRNP M may also play a role in splicing regulation was recently proposed based on studies of an insect hnRNP M orthologue, HRP59 (43). Immunoprecipitation of mRNAs associated with Drosophila melanogaster HRP59 revealed a number of mRNAs that contained an exonic GGAGG sequence motif, which resembled the previously described GAR type of exonic splicing enhancer (44). Although some in vitro binding preference for an RNA sequence containing GGAGG (compared with two control RNAs) was shown, no direct evidence was provided by this study that these sequences within the transcripts that contained them were indeed involved in splicing (43). Nonetheless, more recently these same investigators used RNA interference to show that D. melanogaster HRP59 can regulate splicing of its own transcript as well as that of another transcript by functioning as a negative regulator of splicing (45). Although these studies, together with our findings, suggest a conserved role of hnRNP M proteins in splicing regulation, several questions remain. Perhaps most important will be the identification of an optimal sequence motif that mediates sequence-specific binding of hnRNP M. Our results showing that binding of hnRNP M to GU-rich sequences in the intronic ISE/ISS-3 are consistent with previous reports showing a binding preference for poly(G) and poly(U) homopolymers (25,46). This contrasts with the suggestion that GGAGG exonic sequences represent an optimal binding site for the insect orthologue. It may be that further studies may indeed demonstrate that HRP59 likewise binds to GU-rich sequence elements, including those present in introns. Alternatively, it may be that there has been divergence at the level of sequence binding specificity between these related hnRNP protein family members. Notably, the unique MR repeat motif in metazoan hnRNP M is not conserved in either the Chironomus tentans or D. melanogaster HRP59 protein, suggesting that these orthologues may indeed be functionally distinct. Further studies will also be needed to dissect the mechanisms whereby hnRNP M can affect splicing by either positively or negatively regulating exon inclusion in different transcripts. More detailed identification of additional binding sites for hnRNP M flanking FGFR2 exon IIIc as well as near other exons that it regulates should begin to clarify these questions.