The Single-stranded DNA-binding Proteins, Purα, Purβ, and MSY1 Specifically Interact with an Exon 3-derived Mouse Vascular Smooth Muscle α-Actin Messenger RNA Sequence*

Amino acids 44–53 of mouse vascular smooth muscle α-actin are encoded by a region of exon 3 that bears structural similarity to an essential MCAT enhancer element in the 5′ promoter of the gene. The single-stranded DNA-binding proteins, Purα, Purβ, and MSY1, interact with each other and with opposite strands of the enhancer to repress transcription in fibroblasts (Sun, S., Stoflet, E. S., Cogan, J. G., Strauch, A. R., and Getz, M. J. (1995) Mol. Cell. Biol. 15, 2429–2436; Kelm, R. J., Jr., Cogan, J. G., Elder, P. K., Strauch, A. R., and Getz, M. J. (1999) J. Biol. Chem. 274, 14238–14245). In this study, we employed both recombinant and fibroblast-derived proteins to demonstrate that all three proteins specifically interact with the mRNA counterpart of the exon 3 sequence in cell-free binding assays. When placed in the 5′-untranslated region of a reporter mRNA, the exon 3-derived sequence suppressed mRNA translation in transfected fibroblasts. Translational efficiency was restored by mutations that impaired mRNA binding of Purα, Purβ, and MSY1, implying that these proteins can also participate in messenger ribonucleoprotein formation in living cells. Additionally, primary structure determinants required for interaction of Purβ with single-stranded DNA, mRNA, and protein ligands were mapped by deletion mutagenesis. These experiments reveal highly specific protein-mRNA interactions that are potentially important in regulating expression of the vascular smooth muscle α-actin gene in fibroblasts.

In eukaryotes, protein synthesis is the end result of an integrated program of gene transcription, pre-mRNA processing, and mRNA transport, translation, and metabolism. The molecular mechanisms that function to integrate these diverse processes are largely unknown, but in some cases appear to involve multifunctional proteins capable of associating with regulatory sequences in both DNA and RNA. In Xenopus for example, FRGY2 was originally identified as an oocyte-specific transcription factor that associates with a DNA regulatory element termed the Y-box (1). However, recent studies have shown that FRGY2 also functions to repress mRNA translation, in part, by binding to a specific mRNA sequence motif (2). Thus, expression of FRGY2 leads to both increased transcription from promoters containing a Y-box and translational silencing of the mRNA transcripts (3,4). Similarly, MSY1, the mouse homologue of FRGY2, has been proposed to have dual roles in activating germ cell-specific transcription in the testis and translational repression of the resulting mRNA (5). Both situations are reminiscent of the dual role of transcription factor IIIA in the synthesis and storage of 5 S rRNA in the Xenopus oocyte (6).
Evidence also exists to suggest that pre-mRNA splicing may be coupled to transcription, in part, through common regulatory proteins. In yeast, the group I intron splicing stimulatory protein, Cbp2, has been reported to enhance transcription of the mitochondrial COB gene (7), while in higher organisms, an apparent pre-mRNA splicing factor, heterogeneous nuclear ribonucleoprotein K, has been shown to stimulate RNA polymerase II transcription in vitro and to activate and repress transcription in vivo (8,9). Thus, cotranscriptional splicing, which has been observed in a variety of different systems (10 -12), may involve the participation of proteins that function as both transcription factors and pre-mRNA splicing factors.
Studies in our laboratory have centered on the participation of a group of single-stranded DNA (ssDNA) 1 -binding proteins in transcriptional regulation of the mouse vascular smooth muscle (VSM) ␣-actin gene in fibroblasts. These proteins, recently identified as the mouse Y-box protein, MSY1 (13), and the purine-rich ssDNA-binding proteins, Pur␣ and Pur␤ (14), interact with opposite strands of a polypurine-polypyrimidine tract conserved in VSM ␣-actin gene promoters to negatively regulate transcription by a mechanism not yet fully understood (13,15). This tract, designated the promoter element (PE), resides between Ϫ165 and Ϫ195 relative to the start site of transcription, contains an inverted core MCAT enhancer element (AGGAATG), and exhibits a high degree of purine-pyrimidine asymmetry (15). Interestingly, an inspection of the complete VSM ␣-actin cDNA and genomic sequences revealed an exon 3-derived sequence with strong similarity to the 5Ј PE (16). This sequence, termed the coding element (CE), encoded amino acids 44 -53 of the mature protein, exhibited a similar degree of purine-pyrimidine asymmetry, and lay immediately adjacent to the 3Ј end of intron 2. Unlike the PE, the doublestranded form of the CE exhibited no detectable affinity for transcription enhancer factor 1, despite the presence of an inverted MCAT-like motif, nor did the isolated pyrimidine-rich strand exhibit detectable affinity for MSY1, then termed VAC-ssBF1 (16). However, Pur␣ and Pur␤, then termed the C and D components of VACssBF2, interacted equally well with the purine-rich strands of both the 5Ј PE and the exon 3-derived CE (16).
In the present study, we have investigated whether MSY1, Pur␣, and Pur␤ are capable of interacting with the mRNA counterpart of the CE. These studies were prompted, in part, by the aforementioned ability of Y-box proteins to bind RNA (5,(17)(18)(19) as well as by a recent report that Pur␣ is capable of associating with a 7 SL-like RNA (20). Our results indicate that RNA binding of all three proteins is sequence-specific and does not appear to require protein-protein interaction. Moreover, mutational analyses of mouse Pur␤ indicate that the determinants of ssDNA, mRNA, and protein binding differ from each other and from those previously reported for human Pur␣. These findings suggest that MSY1, Pur␣, and Pur␤ may be multifunctional regulators of VSM ␣-actin gene expression by virtue of specifically interacting with both DNA and RNA regulatory elements.

EXPERIMENTAL PROCEDURES
Cell Culture, Transient Transfection, and Reporter Gene Assays-Cell culture and transfection of mouse embryo-derived AKR-2B fibroblasts were performed as described previously (16). Total protein in lysed transfectants was determined by BCA dye-binding assay (Pierce) using bovine serum albumin (BSA) as a standard. Chloramphenicol acetyltransferase (CAT) and ␤-galactosidase (␤-gal) reporter proteins were measured by ELISA (Roche Molecular Biochemicals). Correctly initiated CAT mRNA transcripts were detected by primer extension analysis as described previously (16).
Synthesis of DNA and RNA Oligonucleotide Probes-Oligonucleotides possessing a 3Ј biotinyl moiety were prepared by chemical synthesis using a biotin phosphoramidite containing a mixed polarity triethylene glycol spacer (BioTEG CPG, Glen Research) on an Applied Biosystems model 394 DNA/RNA synthesizer. RNA oligonucleotides were synthesized using 2Ј-OMe-RNA CE (␤-cyanoethyl) phosphoramidites (Glen Research). Following synthesis, all oligonucleotides were purified by gel filtration over a NAP-25 column in sterile water. A 120-base RNA oligonucleotide corresponding to a tetramer of VSM ␣-actin exon 3 coding element sequence, (GGGAGUAAUGGUUG-GAAUGGGCCAAAAAGA) 4 , used in cDNA expression library screening and Northwestern blotting was subjected to additional polyacrylamide gel purification.
Construction of Mouse Pur␤ Deletion Mutants-Complementary DNAs encoding amino acids 2-263, 2-236, 2-186, 2-135, and 2-87 were amplified by PCR using primers that generated 5Ј BamHI and 3Ј KpnI cloning sites. The PCR products were agarose gel-purified, digested with restriction enzymes, and subcloned into pQE30 (Qiagen) to create fusion constructs encoding an N-terminal 6 histidine tag. Plasmids were transformed into Escherichia coli strain JM109 and the fidelity of the mutant Pur␤ cDNAs was verified by DNA sequencing. Mutant proteins were expressed and purified as described previously (13).
Discontinuous Colorimeteric Nucleic Acid Binding Assay-Purified His-tagged Pur␤ (100-l application of a 50 nM solution) was passively absorbed to polystyrene microtiter wells (Costar EIA strips) as described previously for ELISA (13). After an overnight incubation at 4°C, the coating solution was aspirated and the wells washed twice with 20 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 0.05% Tween 20 (HBSET). A blocking solution of 0.2% (w/v) ultrapure BSA (Roche Molecular Biochemicals) dissolved in HBSET was then applied (250 l/well) for 2 h at room temperature. Blocking solution was removed, and a 50-l solution of ssDNA or RNA oligonucleotide competitor diluted from 1000 to 0.064 nM in assay buffer (0.1% BSA-HBSET with 10 g/ml poly(dI-dC)) was applied to selected wells in quadruplicate. This was followed immediately by the addition of a 50-l solution of 1.0 nM 3Ј-biotinyl-CE-F diluted in assay buffer. After a 1-h incubation at 37°C, the wells were washed three times with HBSET and a 100-l solution of ExtrAvidin-HRP (Sigma) diluted 1:2000 in HBSET was applied. Following a 30-min incubation at 37°C, wells were washed as above and 100 l of the HRP chromogenic substrate, ABTS (Roche Molecular Biochemicals), was applied. Absorbance readings at 405 nm were determined after 15 min using a 96-well microplate spectrophotometer.  a The location of the sequence element within the mouse VSM ␣-actin gene and its corresponding designation are indicated within parentheses. Nucleotides are numbered relative to the start site of transcription (ϩ1). For simplicity, intron sequences have been omitted from the numbering scheme. Therefore, ϩ201 to ϩ299 refers to the location of the sequence within the mouse VSM ␣-actin cDNA while G/ϩ201 denotes the intron 2/exon 3 splice site that would exist in pre-mRNA. The coding element RNA sequence (CE-RNA) as written would also be present in the mature transcript and encode amino acids 43-53 (caG-GGA-GUA-AUG-GUU-GGA-AUG-GGC-CAA-AAA-GAc, QGVMVGMGQKD). Underlined letters denote transversion mutations.

Pur␣
paramagnetic particles in an attempt to capture MSY1 and/or Pur␣/Pur␤ from an AKR-2B fibroblast nuclear extract. As shown in the left panel of Fig . Moreover, when a parallel blot was probed with a Pur protein specific antibody, both Pur␣ and Pur␤ were detected using the CE-RNA-coupled particles ( Fig. 1, right panel, lane 5). In contrast to MSY1 and consistent with previous results, both Pur proteins were also captured using the forward, purine-rich DNA strands of both the promoter (PE-F, lane 3) and coding (CE-F, lane 7) elements. Pur␣ and Pur␤ were not captured, however, using the mutant RNA oligonucleotide (lane 6). These data indicate that all three proteins can specifically associate the same purine-rich, exon 3-derived CE-RNA sequence. This result was somewhat unexpected, given that MSY1 and the Pur proteins specifically bind to opposing pyrimidine-rich and purine-rich strands of the PE sequence, respectively (13,14). Thus, while the ssDNA and mRNA sequences that bind the Pur proteins are related, the ssDNA and mRNA sequences that bind MSY1 are completely unrelated (compare PE-R with CE-RNA in Table I).
The Exon 3-derived RNA Sequence Suppresses Translation of a Heterologous Reporter Gene in a Manner Dependent on Pur␣, Pur␤, and MSY1 Binding-Analysis of translational repression conferred by a protein-binding site located within the 5Ј-UTR of a reporter mRNA has been shown to be useful as a basic strategy for detecting RNA-protein interactions in living cells (21). We previously constructed several chimeric promoters in which the double-stranded CE sequence was incorporated both 3Ј and 5Ј of the transcriptional start site of a deletionally active VSM ␣-actin promoter-CAT reporter construct (VSMP4) to create P4/CE(3Ј) and P4/CE(5Ј), respectively ( Fig.  2A) (16). When analyzed for CAT mRNA expression, the P4/ CE(3Ј) construct was found to be transcriptionally identical to VSMP4 while the P4/CE(5Ј) construct was markedly repressed (16). However, the P4/CE(3Ј) was not tested for CAT protein output (16). Because care was taken to ensure that translational initiation at an internal VSM ␣-actin ATG codon within the 5Ј-UTR of the CAT reporter would not alter the CAT reading frame, we decided to evaluate whether or not this construct would exhibit impaired CAT protein expression as a function of Pur␣/Pur␤/MSY1 binding to CE-RNA. Thus, P4/CE(3Ј) and P4/CEmu2(3Ј) were tested in parallel transfections with their parent construct, VSMP4, and two transcriptionally repressed promoters, VSMP3 (22) and P4/CE(5Ј) (16), for both CAT protein and CAT mRNA expression in AKR-2B fibroblasts. As a measure of the relative abundance of CAT mRNA, correctly initiated CAT transcripts were assayed by primer extension analysis of poly(A) ϩ RNA harvested from lysed transfectants.  (14) and 100 pmol of biotinylated DNA or RNA oligonucleotide (lanes 3-7) were incubated as described previously (14). Nucleic acid-bound proteins were captured on streptavidin-coupled paramagnetic particles, washed three times with 20 mM HEPES, pH 8.0, 500 mM NaCl, and eluted with 2ϫ Laemmli sample preparation buffer. Eluates were assayed by Western blotting using rabbit anti-MSY276 -302 (13) for detection of MSY1 (left panel) and rabbit anti-PurB42-69 (13) for detection of both Pur␣ and Pur␤ (right panel). Each lane represents the amount of Pur␣, Pur␤, or MSY1 captured from 100 g nuclear protein (i.e. 50% sample load). Lane 1 shows the immunoreactivity present in 10 g of nuclear extract. Lane 2 is a control for nonspecific binding by paramagnetic particles. To enhance mRNA accumulation, quiescent transfectants were co-stimulated for 6 h with 20% fetal bovine serum and 10 g/ml cycloheximide (16). Cycloheximide inhibits normal transcriptional attenuation that occurs after serum stimulation (50). Correctly initiated CAT mRNA transcripts were assayed by primer extension analysis of poly(A) ϩ RNA isolated and pooled from four independent transfectants. Arrows indicate the positions of p99␤Ac-CAT (␤Ac, 129 nucleotides), VSMP4 (␣Ac, 168 nucleotides), P4/CE(3Ј) and P4/CEmu2(3Ј) (␣Ac/CE, 207 nucleotides) extension products. 32 P markers (M, pBR322 MspI digest) are shown for reference. Primer extension of RNA harvested from mock transfected cells (mock) served as a control for the fidelity of the CAT mRNA primer. Consistent with our previous data (16), the two repressed promoters, P4/CE(5Ј) and VSMP3, produced clearly less CAT mRNA relative to VSMP4 while P4/CE(3Ј) produced a comparable amount (Fig. 2B, lanes 1-4). Importantly, an extension product of identical size (207 bases) and intensity was obtained from P4/CE(3Ј) and P4/CEmu2(3Ј), indicating that insertion of either the wild type CE, or a mutant thereof, into the 5Ј-UTR of VSMP4 had no substantive effect on the ability of the promoter to drive transcription of the CAT reporter gene (Fig. 2B, lanes  4 and 5).
Because quiescent, G 0 -arrested AKR-2B fibroblasts exhibit relatively low rates of protein synthesis (23), lysates of quiescent transfectants were assayed for CAT protein by quantitative immunoassay. While VSMP3, VSMP4, and P4/CE(5Ј) produced CAT protein at levels that were consistent with CAT mRNA expression, P4/CE(3Ј) produced only ϳ10% of the CAT protein of its parent construct, VSMP4 (Fig. 3). Importantly, mutation of the MSY1/Pur␣/Pur␤ CE-RNA binding site restored CAT protein production to ϳ70% of the level obtained with VSMP4 (Fig. 3, see P4/CEmu2(3Ј)). Residual MSY1 binding to the mutant CE-RNA (Fig. 1) may account for the incomplete restoration of CAT protein synthesis. Collectively, these data suggest that the presence of a MSY1/Pur␣/Pur␤ mRNA binding site in the 5Ј-UTR of the CAT reporter gene reduces the efficiency of CAT mRNA translation. Since mRNA-binding proteins with physiologic roles unrelated to translational regulation have been shown to function similarly when their RNAbinding sites were placed in the 5Ј-UTR of reporter mRNAs (21), it is premature to conclude that a MSY1/Pur␣/Pur␤ complex also impairs mRNA translation when bound to the CE sequence in its natural position in VSM ␣-actin mRNA. Nevertheless, these results reinforce the conclusions drawn from in vitro binding assays, and again implicate MSY1, Pur␣, and Pur␤ as sequence-specific mRNA-binding proteins.
MSY1, Pur␣, and Pur␤ Can Independently Bind the Exon 3-derived CE-RNA Sequence, as Evidenced by RNA Binding Site-dependent cDNA Cloning and Northwestern Blotting-Be-cause of the mutual interactions between MSY1 and the Pur proteins (13), these initial data did not allow us to determine whether all three proteins directly interacted with RNA, or whether one or more proteins interacted indirectly via proteinprotein interaction. To approach this problem, a mouse lung cDNA expression library was screened with a tetramer of the exon 3-derived CE-RNA sequence. Such a strategy was successfully used to clone cDNAs encoding Pur␣ and Pur␤ with the ssDNA counterpart of the CE-RNA sequence (CE-F) (14) and cDNAs encoding MSY1 with the pyrimidine-rich strand of the promoter element (PE-R) (13). Using a 32 P-end-labeled tetramer of the CE-RNA sequence as a probe in the presence of a 160-fold molar excess of the CE-RNAmu2, seven individual clones were purified through tertiary screens. Fig. 4 shows a representative filter-binding assay for one of these clones in which ssDNA and RNA binding capacity was compared. All seven clones demonstrated the same binding specificity for CE-RNA but not CE-F (i.e. the ssDNA counterpart of the CE-RNA sequence) suggesting that these clones did not encode the Pur proteins. Indeed, subsequent DNA sequencing revealed that all seven clones encoded overlapping sequences corresponding to MSY1. These cloning data validated the results of Fig. 1 and again indicated that MSY1 can specifically interact with both the pyrimidine-rich strand of the promoter element sequence and an unrelated, purine-rich, exon 3-derived mRNA sequence. Moreover, interaction of MSY1 with CE-RNA does not require simultaneous Pur␣ or Pur␤ binding. It is also noteworthy that this mRNA sequence is unrelated to a previously identified Y-box protein consensus RNA binding motif (5Ј-AACAUC-3Ј) (19).
The fact that no clones corresponding to Pur␣ or Pur␤ were obtained from 300,000 plaques initially screened suggested to us that the Pur cDNAs were either of low abundance relative to MSY1, or perhaps, the Pur␣/␤-CE-RNA interaction observed in Fig. 1 occurred only indirectly via association with MSY1. To test for direct interaction between mouse Pur␣ and Pur␤ with CE-RNA, we analyzed recombinant His-tagged proteins purified from E. coli (13) by Northwestern blotting using [ 32 P](CE-RNA) 4 as a probe. As a control for loading equivalence, a second Screening of a mouse lung cDNA expression library (300,000 plaques) with a 32 P-end-labeled tetramer of the exon 3-derived CE-RNA sequence yielded seven independent clones. Each clone was tested for its RNA binding specificity in a tertiary filter-binding assay. The same results were obtained with all seven clones. The data for one of the clones, 4 -2, is shown. The left half was incubated with 32 P-CE-RNA 4 (ϳ50 ng/ml, 150,000 cpm/ml) in the presence of 2.0 g/ml CE-RNAmu2, 10 g/ml poly(dI-dC), and 10 g/ml total AKR-2B RNA. The right half was incubated with the counterpart ssDNA probe, 32 P-CE-F 4, (ϳ50 ng/ml, 170,000 cpm/ml) in the presence of 2.0 g/ml CE-Fmu2 and 10 g/ml poly(dI-dC).

SDS-polyacrylamide gel was run in parallel and stained with
Coomassie Blue to visualize all wild type and mutant proteins (Fig. 5B). As shown in Fig. 5C, full-length MSY1, Pur␣, and Pur␤ (lanes 1, 2, and 3, respectively) each bound the [ 32 P](CE-RNA) 4 probe. However, C-terminal deletion mutants of mouse Pur␤ demonstrated variable RNA-binding capacity (lanes 4 -8). Specifically, deletion of amino acids 264 -324, which contains the so-called "psycho" motif (24), had no obvious effect on the ability of the ␤263 mutant to bind CE-RNA (compare lanes 3 and 4). Additional deletion of the third basic-aromatic (class I) repeat (amino acids 237-263) virtually abolished the ability of the ␤236 mutant to bind CE-RNA (lane 5). Curiously, deletion of the next 50 amino acids, which include the second acidicleucine rich (class II) repeat and a glycine-rich sequence unique to Pur␤, resulted in the restoration of RNA-binding capacity in the ␤186 mutant (lane 6). Moreover, deletion of the next 51 amino acids, containing the second basic-aromatic (class I) repeat, once again yielded a mutant protein, ␤135, that was impaired in its ability to bind CE-RNA relative to ␤186 (com- pare lanes 6 and 7). Collectively, these data suggest that 1) binding of Pur␣ and Pur␤ to the CE-RNA sequence can occur independent of interaction of the Pur proteins with each other or with MSY1; 2) sequences between amino acids 237 and 263 and amino acids 136 and 186, which include the second and third basic-aromatic (class I) repeats, respectively, are required for binding of Pur␤ to RNA; and 3) sequences between amino acids 187 and 236, which include the second acidic-leucine rich (class II) repeat, may function as an inhibitory domain.
Ligand Binding Properties of Pur␤ Deletion Mutants Reveal Primary Structure Determinants Necessary for Interaction of Pur␤ with ssDNA, mRNA, Pur␣, and MSY1-Because of the unusual pattern of RNA binding displayed by the Pur␤ deletion mutants, we also asked whether the same was true for ssDNA binding. As shown in Fig. 6, Southwestern blotting of the Pur␤ mutants using the ssDNA counterpart of the CE-RNA tetramer (CE-F 4 ) as a probe revealed a markedly different profile. For example, the ␤186 mutant, which bound CE-RNA quite efficiently (Fig. 5, lane 6), showed no evidence of ssDNA-binding capacity (Fig. 6, lane 3). In fact, only the full-length and ␤263 proteins were capable of binding ssDNA efficiently (Fig. 6,  lanes 1 and 6). Importantly, since the ␤236 mutant was impaired in its ability to bind both ssDNA (Fig. 6, lane 2) and RNA (Fig. 5, lane 5), the third basic-aromatic (class I) repeat is likely required for both ssDNA and RNA binding. However, because mutants such as ␤186 and, to a lesser extent, ␤135 also bind to RNA, additional sequences, such as the second class II repeat, or perhaps the N-terminal glycine-rich region, appear to assist in the Pur␤-RNA interaction.
The affinity and specificity of the interaction between recombinant Pur␤ and ssDNA or RNA were also evaluated quantitatively using a discontinuous, colorimetric nucleic acid binding assay. Full-length, His-tagged Pur␤ was passively immobilized on microtiter wells and tested for its ability to bind fluid-phase biotinyl-CE-F (ssDNA) in the presence of varying amounts of unlabeled wild type and mutant ssDNA and RNA oligonucleotides. Complex formation between biotinyl-CE-F and Pur␤ was detected following subsequent incubations with HRP-coupled streptavidin and a chromogenic substrate. As shown in Fig. 7, the competition curves generated with CE-F and CE-RNA oligonucleotides do not overlap. This is likely a reflection of the non-identity of the ssDNA and RNA-binding regions within Pur␤. However, despite the difference in apparent affinity between the CE-F (IC 50 ϳ8 nM) and CE-RNA (IC 50 ϳ25 nM) oligonucleotides, Pur␤ binding to these sequences is nevertheless specific, as evidenced by the right-shifted curves obtained with the CE-Fmu2 (IC 50 ϳ150 nM) and CE-RNAmu2 (IC 50 Ͼ500 nM) oligonucleotides.
To address whether there is any overlap between sequences required for ssDNA/RNA binding and those required for protein-protein interaction, we utilized an ELISA method to test the protein-binding capacity of the Pur␤ deletion mutants. This assay system was previously used to identify interactions between Pur␣ and Pur␤ and between Pur␣/Pur␤ and MSY1 in fibroblast nuclear extracts (13). As a control, the coating efficiency of each mutant was first assessed with respect to the full-length protein by ELISA using an antibody directed against an N-terminal peptide epitope present in all the mutants. As shown in Fig. 8, the coating efficiencies of ␤263, ␤186, and ␤135 did not differ substantially from the full-length protein. The ␤87 mutant displayed a somewhat higher coating efficiency, while the ␤236 mutant demonstrated a slightly lower coating efficiency. When tested for MSY1 and Pur␣ binding, only the full-length protein demonstrated high affinity binding capacity (Fig. 9). Importantly, deletion of amino acids 264 -324 substantially impaired the ability of the ␤263 mutant to complex with either fibroblast-derived MSY1 or Pur␣. This result is in stark contrast to the data obtained from analyses of nucleic acid binding, where this mutation appeared to have no adverse effect on the ability of the protein to bind either ssDNA (Fig. 6) and RNA (Fig. 5). However, because further deletion of amino acids 237-263 completely eliminates Pur␣ and MSY1 binding (Fig. 9), we cannot exclude the possibility that there may be some partial overlap in amino acid sequences, which contribute to nucleic acid and protein binding capacity.

DISCUSSION
Earlier analyses of mouse vascular smooth muscle ␣-actin gene transcription in fibroblasts and undifferentiated myoblasts led to the formulation of a molecular model in which the activity of an essential, transcription enhancer factor 1-dependent, MCAT enhancer element located within a polypurinepolypyrimidine tract (Ϫ165 to Ϫ195) was subject to negative regulation by sequence-specific, ssDNA-binding proteins (15,25). It was later discovered that the purine-rich strand binding activity, then termed vascular actin single-stranded DNA-binding factor 2 (VACssBF2), also interacted specifically with the sense strand of a protein-coding element (CE) sequence bearing similarity to the 5Ј promoter element (PE) sequence (16). This CE sequence was located at the 5Ј end of exon 3 and encoded amino acids 44 -53 of VSM ␣-actin. The sense strand of the CE sequence was subsequently used to screen a mouse lung cDNA expression library and to isolate clones encoding the purinerich strand-binding proteins, Pur␣ and Pur␤ (14). DNA-binding analyses of Pur␣ and Pur␤ expressed in mouse fibroblasts confirmed that these proteins were identical to the p46 and p44 components of VACssBF2 (14). More recently, we employed the same binding site screening strategy to identify the mouse Y-box protein, MSY1, as VACssBF1 (13). Biochemical studies conducted with peptide-specific polyclonal antibodies confirmed that fibroblast-derived Pur␣/Pur␤ and MSY1 specifically bind to opposite strands of the MCAT containing PE sequence and revealed that Pur␣ and Pur␤ associate with each other and with MSY1 via DNA-independent protein-protein interactions (13). Armed with specific immune reagents (13) and knowledge that human Pur␣ (20) and several Y-box proteins can also bind RNA (5,(17)(18)(19), we decided to test the mRNA counterpart of the aforementioned CE sequence as a target for Pur␣, Pur␤, and MSY1 binding.
In this study, we present several lines of evidence that support the conclusion that mouse Pur␣, Pur␤, and MSY1 can each specifically interact with a purine-rich mRNA sequence encoding amino acids 44 -53 of VSM ␣-actin. First, all three proteins were efficiently captured from a fibroblast nuclear extract using a biotinylated synthetic CE-RNA oligonucleotide coupled to paramagnetic particles (Fig. 1). Second, multiple cDNAs encoding MSY1 were cloned from a mouse lung cDNA expression library using CE-RNA as a probe (Fig. 4). Third, purified re- Microtiter wells coated with His-tagged Pur␤ mutants (5.0 pmol application) were incubated with varying amounts of affinity-purified rabbit anti-PurB42-69 antibody (13). After a 30-min incubation at 37°C, solid phase immune complexes were detected by ELISA using a goat antirabbit IgG-HRP conjugate as described previously (13). The top panel represents the raw data corrected for background absorbance generated with binding buffer alone. The bottom panel shows only the data points that fall on the linear portion of the curves shown in the top panel. The numbers in parentheses represent the relative deviation from the slope of the line obtained for full-length Pur␤ (␤324). combinant Pur␣, Pur␤, and MSY1 each independently bound a CE-RNA probe by Northwestern blotting (Fig. 5). Fourth, the wild-type CE sequence but not a mutant deficient in Pur␣, Pur␤, and MSY1 binding, impaired translation of a CAT reporter mRNA when placed in the 5Ј-UTR of the CAT gene (Figs.  2 and 3). These new findings, together with our previous results, suggest that there are at least two potential regulatory binding sites for Pur␣, Pur␤, and MSY1 in the mouse VSM ␣-actin gene. Although human and avian Pur␣ (20,26) as well as several Y-box protein homologues (5,(17)(18)(19) have been previously implicated in functionally relevant RNA binding, this is the first study, to our knowledge, to document the interaction of Pur␣, Pur␤, and MSY1 with distinct ssDNA and mRNA-binding sites within the promoter and protein-coding regions of the same gene.
Because RNA binding by Pur␤ had not been previously reported, we also felt it important to identify amino acid sequences that contributed to RNA binding affinity in relation to those which mediated ssDNA and protein-protein interaction. Mouse Pur␤ is ϳ70% identical to mouse Pur␣ at the amino acid level (14). Like Pur␣, Pur␤ is a modular protein characterized by a N-terminal domain (amino acids 1-39) rich in glycine residues, a central domain (amino acids 40 -263) consisting of three basic-aromatic (class I) and two acidic-leucine rich (class II) repeats, a C-terminal domain (amino acids 264 -324) con-taining a region of limited similarity to T-antigen (termed the psycho motif in Pur␣; Ref. 27), and a glutamate-rich tail (Ref. 14; see Fig. 5). Comparative ssDNA and RNA binding analyses of C-terminal deletion mutants revealed that amino acids 264 -324 were dispensable for ssDNA and RNA binding (Figs. 5 and 6), while ELISA data suggested that this region was critical for high affinity binding to Pur␣ and MSY1 (Fig. 9). Additional deletion of amino acids 237-263, which includes the third basic-aromatic repeat, eliminated ssDNA, RNA, and proteinbinding capacity, implying that there is some partial overlap in amino acid sequences that mediate nucleic acid binding and protein-protein interaction. In contrast to these results, the third basic-aromatic repeat does not seem to be required for Pur␣ binding to either ssDNA (27,28) or RNA (20). Curiously, further deletion of amino acids 187-236 restored the ability of mutant Pur␤ to bind RNA (Fig. 5) but failed to restore ssDNA binding (Fig. 6). Amino acids 187-236 encompass an acidicleucine repeat that is interrupted by a glycine-rich stretch in Pur␤ but not in Pur␣ (Ref. 14; see Fig. 5). This structural difference could explain why this region appears to function as an inhibitory domain in Pur␤-RNA binding. In Pur␣, sequences within this region appear to be required for binding to the purine-rich strand of c-myc PUR element (27) and to 7 SL-like RNA (20) but not to a single-stranded target sequence in the JC virus lytic control element (29). Thus, one cannot exclude the possibility that the relative importance of the second acidicleucine repeat module to single-stranded nucleic acid binding by the Pur proteins may be dictated by the target sequence.
What is the functional consequence of Pur␣, Pur␤, and MSY1 binding to VSM ␣-actin CE-RNA? Incorporation of this sequence into the 5Ј-UTR of a CAT reporter gene substantially reduced the efficiency of CAT protein translation in a manner that was dependent on nucleotides that contribute to high affinity Pur protein/MSY1 mRNA-binding (Figs. 1, 2, 3, and 7). Interestingly, this same CE sequence suppressed CAT mRNA transcription when positioned 5Ј and directly flanking an enhancer-protein binding site in chimeric promoter-reporter gene constructs in a manner that was again dependent on Pur␣/ Pur␤ (then termed VACssBF2) binding, but in this instance, to the CE-F DNA strand (16). While these results do not directly address the issue of the functional significance of Pur protein and/or MSY1 binding to the CE in its natural exonic position, they do, however, suggest that these proteins possess the capacity to influence gene expression at the level of transcription and translation. In this regard, it is noteworthy that Y-box protein homologues have been implicated in transcriptional activation (3,30,31), transcriptional repression (32,33), and translational repression (4,18). Similarly, Pur protein homologues have been implicated in both transcriptional activation (34 -40) and repression (14,41), and in several notable cases, appear to functionally cooperate with Y-box proteins (42,43).
Clearly, one of the most striking and unexpected findings to emerge from these studies is that MSY1 exhibits specific binding to a pyrimidine-rich ssDNA sequence and to a completely unrelated purine-rich mRNA sequence within the same gene. This mRNA sequence is also unrelated to a previously characterized Xenopus Y-box protein (FRGY2) consensus RNA binding motif identified through in vitro selection (19) and implicated in translational repression (4). Although MSY1 has been proposed to similarly participate in translational masking and mRNA packaging (5), the distinctiveness of its VSM ␣-actin CE-RNA binding site suggests an alternative functional role. It is tempting to speculate that expression of the VSM ␣-actin gene in fibroblasts could be regulated through the coordinated interaction of Pur␣, Pur␤, and MSY1 with promoter-and mRNA-derived sequences. For example, sequestration of Pur␣, FIG. 9. Amino acids 264 -324 are required for Pur␤ binding to MSY1. Microtiter wells coated with His-tagged Pur␤ mutants (5.0 pmol application) were incubated with varying amounts of AKR-2B fibroblast nuclear protein as described previously (13). Wells were aspirated and washed three times, and solid phase MSY1-Pur␤ (bottom) complexes were detected by ELISA using either rabbit anti-MSY1242-267 or rabbit anti-PurA291-313 as the primary antibody. Pur␤, and MSY1 by virtue of binding to newly transcribed mRNA might establish an autoregulatory loop to augment VSM ␣-actin transcription. An analogous mechanism involving ␤-casein mRNA-dependent sequestration of a ssDNA-binding repressor has been proposed to explain the lactogenic hormoneinduced expression of the ␤-casein gene in mammary epithelial cell (44). It is also possible that Pur␣, Pur␤, and MSY1 may function in regulating some other aspect of mRNA metabolism, such as pre-mRNA splicing. This is suggested by the similarity between the exon-derived CE sequence and a class of purinerich sequences located within the exons of several different cellular and viral genes, which affect splice site selection and kinetics of pre-mRNA processing (45)(46)(47)(48)(49). The position of the CE precisely at the intron 2/exon 3 boundary lends further credence to this notion. Clearly, the significance of Pur␣, Pur␤, and MSY1 binding to VSM ␣-actin CE-RNA will only be resolved through further experimentation.