S1 proteins C2 and D2 are novel hnRNPs similar to the transcriptional repressor, CArG box motif-binding factor A

Abstract

S1 proteins A–D are liberated from thoroughly washed nuclei by mild digestion with DNase I or RNase A, and extracted selectively at pH 4.9 from the reaction supernatants. Here, we characterized the S1 proteins, focusing on protein D2, the most abundant S1 protein in the rat liver, and on protein C2 as well. Using a specific antibody, McAb 351, they were shown to occur in the extranucleolar nucleoplasm, and to be extracted partly in the nuclear soluble fraction. We demonstrate that the S1 proteins in this fraction exist constituting heterogeneous nuclear ribonucleoproteins (hnRNPs), through direct binding to hnRNAs, as revealed by centrifugation on density gradients, immunoprecipitation, and UV cross-linking. In hnRNPs, protein D2 occurred at nuclease-hypersensitive sites and C2 in the structures that gave rise to 40 S RNP particles. By microsequencing, protein D2 was identified with a known protein, CArG box motif-binding factor A (CBF-A), which has been characterized as a transcriptional repressor, and C2 as its isoform protein. In fact, CBF-A expressed from its cDNA was indistinguishable from protein D2 in molecular size and immunoreactivity to McAb 351. Thus, the present results demonstrate that S1 proteins C2 and D2 are novel hnRNP proteins, and suggest that the proteins C2 and D2 act in both transcriptional and post-transcriptional processes in gene expression.

Abbreviations

CBF-A

CArG box motif-binding factor A

hnRNPs

heterogeneous nuclear ribonucleoproteins

DMEM

Dulbecco's modified Eagle's medium

API

Achromobacter lyticus protease I

ssDBF 1

single-strand D-box binding factor 1

CRP-1

chicken ribonucleoprotein 1.

Functionally related proteins are often extracted together under particular conditions. Examples of such nuclear proteins are histones, the high mobility group (HMG) proteins, and nuclear lamina pore complexes [1–3]. We found S1 proteins in 1983. They constitute another such group of nuclear proteins [4]. They are present at sites sensitive to RNase as well as DNase, and extracted at pH 4.9 from the reaction mixture of nuclei treated mildly with either enzyme. Under these conditions, most proteins liberated from the nuclei become aggregated, and S1 proteins are left behind in the supernatant S1 (the first supernatant termed in the original experiments) [5,6]. This protein-aggregation is largely due to isoelectric precipitation at or near the isoelectric points of associated RNPs and chromatin (Y. Fujitsuka, K. Tsugawa & A. Inoue, unpublished results). The S1 proteins are separated into four doublets, A, B, C, and D in SDS/PAGE: A1 (migrating with an apparent molecular mass of 74.5 kDa), A2 (69.5); B1 (47.4), B2 (46.5); C1 (43.9), C2 (42.8), D1 (40.8), and D2 (39.4). Proteins B1, B2, C1, C2, D1, and D2 are liberated from nuclei with very similar kinetics in DNase I digestion, suggesting that they are present at similar sites in the nucleus [4]. S1 proteins are widely found, not only in all rat tissues examined, but also in many species [4,7] (T. Watanabe & A. Inoue, unpublished results). Polyclonal antibodies R1 and R2 raised in rabbit with protein B as an immunogen both reacted with proteins B2, C1 and D1 [8,9]. Using the antiserum R1 or R2, S1 proteins were localized in the extranucleolar nucleoplasm in the euchromatin bordering heterochromatic areas [8], where most of the RNA polymerase II transcription takes place (reviewed in [10]). In molecular cloning using the polyclonal antibody R2, clone pS1-1 was isolated. This clone encoded a new RNA-binding protein of 852 amino acids, with two RNA-recognition motifs and a binding preference for G and U polyribonucleotides [11]. The characteristic of S1-1 protein is that it contains five regions of 25–109 amino-acid residues with significant homologies to known transcription factors. A hybridoma was isolated, from which a monoclonal antibody was produced (McAb 351) that specifically reacted with S1 proteins C2 and D2 [12]. The protein C2 and/or D2 is also present in the cytoplasm in addition to the nucleus, localizing on the cytoskeletal vimentin intermediate filaments in various cells in culture [12]. These findings suggested the possibility that the S1 proteins are involved in the post-transcriptional processes in gene expression. In fact, all the S1 proteins except for protein A1, which to date has not been studied fully, have been found to be characteristic of RNA-binding activities. We report here the results on S1 proteins C2 and D2.

hnRNA is a term collectively given to primary and processing-intermediate RNAs transcribed by RNA polymerase II. It is associated with proteins to form heterogeneous ribonucleoproteins (hnRNPs). The complexes sediment between 40 S and 250 S [13], and yield 40 S particles upon mild in vitro action of endogenous or exogenous RNases [14–17]. The 40 S particles contain an RNA fragment of about 700 nucleotides and hnRNP A, B, and C proteins as major core proteins with apparent molecular masses between 32 and 44 kDa [17]. More than 20 individual major hnRNP proteins designated as hnRNP A through hnRNP U have been identified in HeLa cells [18]. Some of them (A1, A2, D, E, I and K) shuttle continuously and rapidly between the nucleus and the cytoplasm [19]. In addition to their role in transcript packaging, hnRNP proteins have various and dynamic functions. Such a role is seen typically with hnRNPs C, A1 and other proteins in splicing reactions of premRNAs. Some of their other functions are in polyadenylation, transport, turnover and stability of mRNAs, and translational silencing (reviewed in [20,21]). hnRNP proteins also participate in the regulation of transcription as have been demonstrated with hnRNP K [22–25], hnRNP D, also known as AUF1 [26–28], and hnRNP U [29].

In this paper, we report our findings that S1 proteins C2 and D2 are new hnRNP proteins. We also demonstrate that the protein D2 was identified with CBF-A, which has been characterized as a transcriptional repressor [30,31], and that the C2 protein is an isoform of CBF-A.

Materials and methods

Protein preparation

S1 proteins were prepared from rat liver nuclei as described previously [4]. Briefly, nuclei were isolated with 9 vol. of 2.3 m sucrose, 3 mm MgCl₂, and 0.2 mm phenylmethanesulfonyl fluoride. They were washed thoroughly in 0.3 m sucrose, 3 mm MgCl₂, 12 mm Tris/HCl, pH 7.4, and 0.2 mm phenylmethanesulfonyl fluoride, and mildly digested with DNase I (2.5 µg·mL⁻¹) at 30 °C for 8 min. The mixture was adjusted to pH 4.9 with Na₂/EDTA, at a final concentration of 5 mm, and centrifuged. The S1 proteins in the supernatant were precipitated with 2 vols of ethanol in the presence of 0.2 m NaCl, centrifuged, rinsed with 70% ethanol, and air-dried briefly.

Nuclear proteins were extracted from rat liver nuclei, with 0.35 m NaCl containing ‘RSB/proteasin’ solution (10 mm Tris/HCl, 10 mm NaCl, 3 mm MgCl₂, pH 7.4, 0.4 mm phenylmethanesulfonyl fluoride, 10 µg·mL⁻¹ each of pepstatin and leupeptin hemisulfate, and 100 KIU·mL⁻¹ aprotinin). The supernatant was diluted with RSB/proteasin to 0.2 m NaCl, and proteins were collected as above.

Whole proteins from cultured cells were prepared by homogenization in RSB/proteasin + 0.5 mm CaCl₂, followed by incubation of the homogenate with DNase I (20 µg·mL⁻¹) for 7 min at 30 °C. Proteins were collected as above.

Cell culture

Rat liver epithelial cells (ARL J301-3) were grown in William's E medium, and Morris hepatoma (HTC) cells and HeLa cells in Dulbecco's modified Eagle's medium (DMEM) at 37 °C in a CO₂ incubator. The media were supplemented with 10% fetal bovine serum. Usually, the cells were grown in 85 mm-plates.

Indirect immunofluorescence staining

ARL cells, grown at a semiconfluency on coverslips, were fixed in 4% paraformaldehyde for 10 min, permeabilized in 0.2% Triton X-100 for 2 min, and blocked with 10% fetal bovine serum in DMEM. The cells were stained with McAb 351, and then with a secondary FITC-labeled goat anti-(mouse IgG) Ig (Cappel). All reagents other than fetal bovine serum were used in NaCl/P_i.

Immunoblotting

Proteins were run on SDS polyacrylamide gels [32], blotted onto a poly(vinylidene difluoride) membrane (Immobilon-P, Millipore Corp.), and immunostained as described previously using either aminoethylcarbazole [12] or an ECL kit (Amersham, RPN 2106). The kit was used when reprobing of the same blot with other antibodies was needed.

Labeling of hnRNA, and immunoprecipitation

ARL cells, grown at a 50–70% confluency, were incubated with [5,6-³H]uridine (NEN, 35–50 Ci·mmol⁻¹) at 20 µCi·mL⁻¹ for 15–18 min at 37 °C in the presence of 0.04 µg actinomycin D per mL, which was added 20 min prior to addition of the isotope. Preparation, partial purification on a sucrose cushion, and immunoprecipitation of [³H]RNPs were performed similarly to the previously described methods [33]. Antibodies were used in a state bound to protein A–Sepharose beads (Amersham Pharmacia, 4 Fast Flow). The [³H]RNA bound to the beads was thoroughly washed and isolated by the method of Chomczynski and Sacchi [34], starting by the addition of total rat liver RNA (2 µg) as a carrier and 0.3 mL of solution D. RNA was electrophoresed on formaldehyde-containing agarose gels (1%) [35]. The resolved RNAs were transferred to nylon membrane and analyzed on an image analyzer (Bas-2000 II, Fuji Film, Tokyo).

UV crosslinking

ARL cells pulse-labeled with [³H]uridine in the presence of actinomycin D as described above were rinsed twice with cold NaCl/P_i, and the culture plates were placed on a glass plate chilled at 4 °C. The cells were irradiated with UV (254 nm) at an energy setting of 500 mJ·cm⁻² with an 8-cm distance from the lamps and an irradiation period of 5 min (CL-1000; UVP Inc., Upland, CA, USA). The cells were homogenized with a glass/Teflon homogenizer in Buffer A (100 mm NaCl, 2.5 mm MgCl₂, 10 mm Tris/HCl, pH 7.4) containing 0.4 mm phenylmethanesulfonyl fluoride and 10 µg leupeptin hemisulfate per mL. The nuclei were pelleted at 3000 g, suspended in the same buffer solution and incubated with 75 µg RNase A per mL and 25 µg DNase I per mL at 30 °C for 20 min. From the reaction mixtures, S1 proteins were selectively extracted with 5 mm EDTA at pH 4.9 as described above. For immunoprecipitation, the pH of the preparations was returned to 7.2 with 1 m Tris/HCl, pH 8.0.

Sucrose gradient sedimentation

HTC cells or ARL cells were grown at 37 °C at a confluency of about 50–70%, and collected using NaCl/P_i. The cells from 10 plates were homogenized in a 1 : 3 mixture of RSB/proteasin and 2.3 m sucrose, 3 mm MgCl₂, and 0.2 mm phenylmethanesulfonyl fluoride, and centrifuged at 40 000 g for 50 min. hnRNPs were extracted from the pelleted nuclei three times with 0.6-mL portions of RSB/proteasin, and aliquots (0.5 mL) were centrifuged on 5–20% (w/v) sucrose density gradients in RSB/proteasin at 80 000 g in a vertical rotor (model RP55VF rotor, Hitachi, Japan) for 45 min at 4 °C. Tobacco mosaic virus (200 S) was used as a size marker. Sixteen fractions (0.75 mL each) were collected.

40 S hnRNP core particles

Core particles were prepared from isolated nuclei as previously described [15]. Briefly, washed nuclei from rat liver, HTC or ARL cells were gently stirred on ice for 30 min in 100 mm NaCl, 3 mm MgCl₂, 1 mm dithiothreitol, 0.2 mm phenylmethanesulfonyl fluoride, and 10 mm triethanolamine/HCl, pH 8.0 to allow endogenous nuclease(s) to convert hnRNPs to 40 S particles [36]. The extraction was repeated three times for a total of 1.5 h, and the combined extracts were run on 5–20% sucrose gradients in a vertical rotor at 220 000 g for 35 min.

Transfection of HeLa cells with CBF-A cDNA

HeLa cells were incubated in 3.5-cm dishes for 24 h to a confluency about 80%. The cells were rinsed with Opti-MEM I (Gibco BRL, cat. no. 31985), and transfected in 1 mL of Opti-MEM I with liposomes prepared with cDNA (1 µg pSR-CBFA-N) and 0–12 µL of lipofectAMINE (Gibco BRL cat. no. 18324). After incubation for 4 h, 3 mL of DMEM + 10% fetal bovine serum was added, and the incubation was continued for further 60 h, replacing with fresh medium at 20 h. Whole proteins were prepared as described above.

Protein sequencing

S1 protein D2 was digested with Staphylococcus aureus V8 protease (Sigma) by a modified method of Cleveland et al. [37], and the resulting peptides were analyzed as described by Matsudaira [38]: S1 proteins (1 mg) were electrophoresed on a SDS polyacrylamide gel (3% flat stacking/11% separating gel, 0.15 × 15 × 15 cm). The gel was stained briefly with Coomassie Brilliant Blue, and destained. Protein D2 in the excised band was digested with V8 protease in situ, and electrophoresed on a second SDS/gel (16%), until bromophenol blue reached the separating gel. After 1 h, electrophoresis was resumed. Peptides in the gel were blotted to poly(vinylidene difluoride) membrane, and stained with Coomassie. Bands were cut for microsequencing on an automated protein sequencer (Applied Biosystems, Model 477A/120A).

Digestion with Achromobacter lyticus protease I (API, Wako Pure Chem. Industries, Osaka, Japan) was performed as described [39]: Briefly, after SDS/PAGE, proteins were blotted to poly(vinylidene difluoride) membrane, and stained with Ponceau S in 1% acetic acid. Excised bands (> 200 pmoles of protein) were treated with dithiothreitol (1 mg), then with Na iodoacetic acid (2.6 mg), and washed successively with 2% acetonitrile (AcCN), 0.1% SDS, and water. The membrane strips were treated with polyvinylpyrrolidone, cut into pieces about 1 × 1 mm in size, and incubated with API (1 : 10–100 enzyme/substrate molar ratio) in 100 µL of 8% AcCN and 20 mm Tris/HCl, pH 8.5 at 37 °C for 14 h under vigorous shaking. After the membrane pieces were washed with 10% AcCN, peptides in the combined supernatants were fractionated on a reverse phase HPLC column (Aquapore RP-300, C-8, 2.1 × 30 mm, Applied Biosystems from PerkinElmer) with an AcCN gradient (0–70% in 0.1% trifluoroacetic acid) at a flow rate of 200 µL·min⁻¹ at 35 °C, and sequenced as above.

Results

S1 proteins and antibody McAb 351

Fig. 1A shows electrophoretic profiles of S1 proteins A–D isolated from rat liver nuclei. Among S1 proteins, D2 was most abundant in this tissue (lane 1), occurring at about 10⁷ copies or more in a single nucleus as estimated from band intensity in SDS gels or on immunoblots (not shown). Apparent molecular masses of proteins C2 and D2 were 42 800 and 39 400, respectively. Sometimes, due to an unknown electrophoretic factor, doublets of the S1 proteins, B and C in particular, are not resolved even with the same sample.

Figure 1

S1 proteins and the specificity of McAb 351. (A) S1 proteins from rat liver were run on a SDS-gel (11%), and stained with Coomassie-Brilliant Blue. Molecular weights (MWs) were estimated using protein markers (BioLabs). (B) Rat liver S1 proteins (lanes 1 and 2) and proteins extracted from isolated nuclei at 0.35 m NaCl (lanes 3 and 4) were resolved by SDS/PAGE (8.8%), and blotted to poly(vinylidene difluoride) membrane. The blots were cut longitudinally into strips, and incubated with McAb 351, followed by a peroxidase-conjugated secondary antibody (lanes 1 and 4), or with Coomassie Brilliant Blue (lanes 2 and 3). Samples on lanes 3 and 4 were run on a flat stacking gel.

Figure 1B shows the specificity of the antibody McAb 351 used in the present study. McAb 351 is specific for S1 proteins C2 and D2 (lanes 1 and 2). Its high specificity was shown here with a mixture of liver nuclear proteins extracted at 0.35 m NaCl (lanes 3 and 4). The high specificity of the antibody has been demonstrated also with whole protein extracts of the liver [12] and various other tissues of the rat, as well as with extracts from various cultured cells (A. Inoue, unpublished results).

Occurrence of soluble S1 proteins in the nucleus

Chromatin-bound S1 proteins have been investigated in the previous studies, where the proteins were released from thoroughly washed nuclei by mild digestion with DNase I or RNase A. In the present study, it was found that S1 proteins were partly extracted in the nucleoplasmic soluble fraction (Fig. 2).

Figure 2

Occurrence of S1 proteins in the soluble nucleoplasm as well as in the chromatin. Proteins were extracted from rat liver nuclei by repeated suspension (3 min) and centrifugation (10 min) in ‘RSB/proteasin’ solution containing 10 mm NaCl (lanes 2–6), then in a stepwise manner with the same solution containing 0.15, 0.4 and 0.6 m NaCl (lanes 7–9). Proteins were separated by SDS/PAGE, and analyzed by immunoblotting using antibody McAb 351 (A), or stained with Coomassie Brilliant Blue (B). Lane 1: reference S1 proteins (0.1 µg) prepared by DNase I-digestion method.

The soluble S1 proteins were gradually extracted from isolated nuclei with a low salt (10 mm NaCl) solution, and the yield decreased after the third extraction (Fig. 2A, lanes 3–6). Most of the other proteins were, however, readily extracted in the first and second extractions (Fig. 2B). From the residual nuclei, chromatin-bound S1 proteins were extracted at increasing NaCl-concentrations (lanes 7–9); the extraction of C2 and D2 was complete at 0.6 and 0.4 m NaCl, respectively. About one-fifth of the S1 proteins were recovered in the soluble fraction.

Indirect immunofluorescence staining of ARL cells indicates that proteins C2 and D2 occur in the extranucleolar region of the cell nucleus (Fig. 3). Such staining patterns are commonly observed with all examined cell types from rodents. This suggests that the proteins C2 and D2 participate more likely in the events other than the synthesis of ribosomes or rRNAs. Immunoblotting of the nuclear proteins demonstrated that both S1 proteins C2 and D2 are indeed present in the ARL cell nuclei (data not shown).

Figure 3

Extranucleolar localization of S1 proteins C2 and D2. Indirect immuno-fluorescence staining. ARL cells grown at a semiconfluency were stained with McAb 351 and a secondary FITC-labeled anti-(mouse IgG) Ig.

Association of S1 proteins C2 and D2 with hnRNP complexes

Proteins C2 and D2 were associated with structures of extremely heterogeneous sizes, as shown when the nucleoplasmic soluble fractions were centrifuged on sucrose density gradients: S1 proteins were spread on the gradient continuously from top to bottom (Fig. 4A). The fast sedimenting S1 protein-containing complexes had sedimentation coefficients up to 200 S or more as determined using tobacco mosaic virus as a marker.

Figure 4

Analyses by sucrose density gradient sedimentation. (A) Nucleoplasmic soluble fraction from HTC cells was centrifuged on a sucrose density gradient. Sixteen fractions were collected from the bottom, and analyzed by immunoblotting using McAb 351. Numerals indicate fraction numbers, and the arrow head shows the sedimentation position of marker tobacco mosaic virus particles (200 S). S1: reference S1 proteins stained with McAb 351 (left) and with Coomassie Brilliant Blue (right). (B) Prior to the centrifugation, the nuclear extract was incubated with 50 µg·mL⁻¹ of pancreatic RNase A at 20 °C for 15 min, and analyzed as above. Similar analysis using DNase I did not alter the sedimentation profile (not shown). Essentially the same results as (A) and (B) were obtained with ARL cell extracts.

The nucleoplasmic extract was treated prior to the centrifugation with DNase I or RNase A. In contrast to DNase I that had no effect on the sedimentation profile, RNase A drastically altered the sedimentation of S1 protein-containing complexes, and converged them on the top fractions of the gradients (Fig. 4B). These results indicated that S1 proteins C2 and D2 in the nuclear soluble fraction were associated with RNAs.

The smooth continuity in the size of the RNP complexes up to 200 S or larger (Fig. 4A) suggested that the complexes were hnRNPs and not pre-rRNPs: the pre-rRNPs would have given discrete peaks over the fractions 11–14 on the gradient, as estimated from the S values of pre-rRNPs (40–80 S [40,41]), and 200 S of the reference tobacco mosaic virus. It is noteworthy that the S1 proteins, the protein C2 in particular, occurred largely as RNP complexes and not as a free form in the nucleoplasmic soluble fractions.

Immunoprecipitation of hnRNAs with McAb 351

The association of S1 proteins with hnRNAs was demonstrated. ARL cells were incubated with [³H]uridine for a short time to label nascent hnRNAs, and in the presence of 0.04 µg·mL⁻¹ actinomycin D to reduce [³H]uridine-incorporation into rRNAs by about 50–60% [42]. The nucleoplasmic soluble fractions were immunoprecipitated with McAb 351, and analyzed by gel electrophoresis. The immunoprecipitated [³H]RNAs were resolved in a smear as a mixture of molecules with various sizes (Fig. 5, lane 4), indicating that the S1 protein-associated RNAs consisted of highly heterogeneous RNA molecules (hnRNAs). The sizes were approximately several hundred nucleotides to up to 8000 nucleotides or larger. With a control antibody, RNA was little precipitated (lane 3). Densitometric analysis of Fig. 5 indicated that under the conditions, the antibody 351 immunoprecipitated as much as 58% of the input [³H]RNAs.

Figure 5

Immunoprecipitation of hnRNAs. ARL cells were pulse-labeled with [5,6-³H]uridine, and the nucleoplasmic soluble fraction was immunoprecipitated with McAb 351 bound to protein A–Sepharose beads. The precipitated [³H]RNAs were eluted in the presence of unlabelled total RNA, which was used as a carrier and a size marker. RNAs were electrophoresed, blotted on nylon membrane, and analyzed on an image analyzer. Lane 1, RNA before immunoprecipitation; lanes 2–4, RNAs immunoprecipitated without antibody (lane 2), with a control monoclonal antibody against toxoplasma (lane 3), and with McAb 351 (lane 4). Twofold more RNA samples were applied on lanes 2–4 than on lane 1 with respect to the starting nucleoplasmic extract. In the blotting, unlabelled rRNAs competed with labeled hnRNAs, giving rise to bleached [³H]images, which were used as markers at 1874 nucleotides (18 S) and 4718 nucleotides (28 S). RNAs on the lane 4 seem to reveal that some degradation had occurred: the actual sizes of the S1 protein-associated hnRNAs may be larger than shown here (see text).

Direct binding of S1 proteins C2 and D2 to hnRNAs

The direct binding of S1 proteins to hnRNAs was examined by photo-crosslinking reaction. As shown below, protein D2 was identified with CBF-A, which binds to the CArG box motif element CC(A/T)₆GG of various genes [30]. CArG box has a palindromic TAATTA in the middle of the element. Consequently, we presumed that if CBF-A acted as an RNA-binding protein, it would bind to a sequence containing UAAUUA. Based on this assumption, hnRNAs in ARL cells were metabolically labeled with [³H]uridine, and the cells were irradiated with UV. S1 proteins were prepared as usual by digestion of isolated nuclei with DNase I, but together with RNase A (Fig. 6).

Figure 6

S1 proteins C2 and D2 bind directly to hnRNAs. ARL cells were pulse-labeled with [³H]uridine, and irradiated with UV (lanes 3, 4 and 5). Control cells received no UV irradiation (lanes 1 and 2). S1 proteins were prepared by digesting isolated nuclei extensively with a mixture of DNase I and RNase A. One-fourth of each S1 protein extract was saved (lanes 1 and 3); remaining 3/4 portions were immunoprecipitated with McAb 351 (lanes 2 and 4). (A) Samples were analyzed by immunoblotting using McAb 351. (B) The blot obtained in (A) was analyzed for ³H-labeled proteins on an image analyzer. UV, UV-crosslinked; IP, immunoprecipitated. Bands above and below the proteins C2 and D2 in the panel A are the heavy and light chains of McAb 351. Lane 5 in (B): proteins immunoprecipitated by a control antibody, R2, raised against S1 proteins B [8,9]. With the McAb 351, protein C2 was immunoprecipitated quantitatively, while D2 to a lesser extent (lanes 3 and 4, B), probably due to the higher affinity of McAb 351 for protein C2 than for D2. Its affinity for protein D2 may have been further reduced by uridine-conjugation.

In contrast to the control without UV irradiation (lane 1 in B), S1 proteins in the UV-irradiated cells were apparently cross-linked to ³H-labeled RNA (lane 3 in B). To have firm results, S1 proteins were immunoprecipitated with McAb 351 from the preparation. It was remarkable that the bands of the immunoprecipitated S1 proteins C2 and D2 (lane 4 in A) and their images of radioactivity resulting from the cross-linked [³H]uridine (lane 4 in B) superimposed precisely as shown on the blot. A control antibody that reacted with S1 proteins B2, C1, and D1 did not immunoprecipitate the protein C2 nor D2, but did precipitate protein B2 and very weakly C1 and D1 (lane 5). These results indicated that the association of proteins C2 and D2 with hnRNP complexes was through direct binding to RNAs.

RNase-sensitivity of S1 protein-associated structures

hnRNP complexes are susceptible to the cleavage of RNA moiety by endogenous nuclease(s) and converted to 40 S hnRNP particles [15,36]. After this cleavage reaction, the products released from the nuclei were analyzed by density gradient centrifugation. Under the conditions employed, 40 S core particles were sedimented in the middle region of the gradients, as seen with the presence of the hnRNP core particle proteins A, B, and C [17,43] (fractions 4–10 in A and B, Fig. 7). S1 protein C2 was recovered in 40 S particles (Fig. 7C). In contrast to this, protein D2 sedimented as smaller fragments. As the protein D2-containing intact hnRNPs sedimented at up to 200 S or larger (Fig. 4A), this extensive reduction in the sedimentation velocity suggested that protein D2 localized at sites hypersensitive to nuclease(s). In fact, when RNase A was added in the 40 S particle-isolation solution, protein D2 was now recovered completely in the top fractions, while protein C2 was still in larger fragments (Fig. 7D). These results indicated that the proteins C2 and D2 were distinguishable from each other by the RNase-sensitivity of the associated structures in hnRNPs.

Figure 7

RNase-sensitivity of the sites associated with S1 proteins C2 and D2. Endogenous nuclease(s) was employed to covert hnRNPs to 40 S particles. After a prolonged (a total of 1.5 h) incubation of ARL cell nuclei at 4 °C, the products in the supernatants were centrifuged on 5–20% sucrose gradients. Fractions were collected from the bottom, and analyzed by SDS/PAGE. The gels were (A) stained with Coomassie Brilliant Blue, or (B) blotted to membrane, which was immunostained with antibody 4E4 [43] specific for 40 S hnRNP core particle proteins A, B, and C. (C) The blot was reprobed with McAb351 specific for S1 proteins C2 and D2. Numerals and brackets indicate fraction numbers and 40 S core particle proteins A–C, respectively. (a) and (b): reference S1 proteins from rat liver and from ARL cells. (D) hnRNPs were extracted in the presence of a small amount (0.5 µg·mL⁻¹) of RNase A, and analyzed similarly by sucrose gradient centrifugation.

S1 proteins are similar or identical to CBF-A

S1 proteins from the rat were resolved by SDS/PAGE, and blotted on a poly(vinylidene difluoride) membrane, then protein D2 was subjected to microsequencing. As D2 had a blocked N-terminus, internal amino-acid sequences were determined after digestion with Staphylococcus aureus V8 protease or Achromobacter lyticus protease I (API). Five peptides from V8 protease digestion and three from API digestion were sequenced. The total number of amino acids determined with the eight peptides was 146 (Fig. 8). From the database, a protein with an identical sequence emerged: rat CBF-A protein with 285 residues [31]. All of the eight peptides corresponded to the sequence of rat CBF-A, except for one undetermined amino acid in peptide 6; 117 of the determined 118 amino-acid residues were identical.

Figure 8

Sequence analysis of S1 proteins. As protein D2 had a blocked N-terminus, its internal amino-acid sequence was determined using peptides obtained by V8 protease digestion (peptides 1, 2, 3, 5, and 6) and API digestion (peptides 4, 7, and 8) as described in Materials and methods. The peptide sequences identified the protein D2 with a known protein, CBF-A. Amino-acid sequences of rat CBF-A protein and protein D2 peptides are aligned. Symbols are: *, identical amino acid; and x, undetermined amino acid. The protein C2 was similarly analyzed using its three peptides (underlined) obtained by digestion with V8 protease (the peptide overlapping the peptides 6 and 7 of protein D2) and with API (the peptide overlapping peptides 4 and 5, and 8 of protein D2).

Protein C2 was similarly analyzed with three peptides. All of them matched the sequence of CBF-A perfectly (Fig. 8, underlined sequences). The sequences of S1 proteins C2 and D2 did not coincide with any known hnRNP proteins. Therefore, we concluded that they were novel hnRNP proteins.

Further evidence for the identity of S1 protein D2 with CBF-A

In order to confirm the identity between the protein D2 and CBF-A, transfection experiments were performed, taking advantage of that the antibody McAb 351 does not react with human S1 proteins C2 and D2 on immunoblots. HeLa cells were transfected with a plasmid containing mouse CBF-A cDNA, which was placed downstream of the SV 40 promoter [30]. The cells transfected with pCBF-A cDNA-liposome complexes produced a protein that strongly reacted with McAb 351 (Fig. 9, lanes 4 and 5). The product of CBF-A cDNA had the same molecular size as the S1 protein D2 as seen on the gel, and was the only protein that reacted with the antibody in the whole cell extract. Neither untreated cells (lane 1) nor cells treated with only liposome (lanes 2 and 3) produced McAb 351-reactive proteins. From the amino-acid sequencing and transfection experiments, we concluded that S1 protein D2 was CBF-A.

Figure 9

Further evidence for the identity between S1 protein D2 and CBF-A. HeLa cells were transfected with pSR-CBFA-N using lipofectAMINE. Sixty-four hours later, whole proteins were analyzed by immunoblotting using McAb 351. Proteins were from the cells: untreated (lane 1), or treated with liposomes that were formed with 4 or 12 µL of lipofectAMINE alone (lanes 2 and 3) or with liposomes containing pSR-CBFA-N (1 µg) (lanes 4 and 5). Lane 6: reference rat liver S1 proteins. S1: the reference S1 proteins stained with Coomassie Brilliant Blue. The cell density and CBF-A production became reproducibly less when a larger volume (12 µL) of cytotoxic lipofectAMINE was used (lane 5).

Discussion

S1 proteins C2 and D2 were novel hnRNP proteins. The present results are consistent with the previous observations: S1 proteins occurred at sites highly sensitive to DNase I as well as RNase A in the cell nucleus [4,5]. It is comprehended that RNase A digests the S1 protein-containing nascent RNPs, and releases the fragments to the supernatants; likewise, DNase I rapidly degrades and liberates transcriptionally active chromatin as small fragments [44], which must contain nascent RNPs as well as transcription factors. As discussed below, the S1 proteins may also act as transcription factors (see below).

About 20% of S1 proteins occurred in the soluble nucleoplasm as hnRNPs. They were extracted gradually from isolated nuclei in a low salt solution. This slow extraction was probably due to that the S1 proteins were in hnRNPs, the large multimolecular complexes, and took time to exit the nuclei. Some of these hnRNPs may have originated from the chromatin-bound nascent hnRNPs, being liberated by the action of endogenous nuclease during the extraction steps. This seems particularly true with protein D2, as it is associated with RNase-hypersensitive structures in hnRNPs (see Fig. 7). The S1 proteins left in the nuclei were extracted at increasing salt concentrations, and the extraction was complete at 0.6 m NaCl. When the nucleoplasmic soluble fraction was analyzed at 0.6 m NaCl, S1 proteins were recovered as free form in density gradient centrifugation, thereby indicating that ionic interactions are involved between S1 proteins and hnRNAs.

Previously, we reported that a fraction of chromatin-bound S1 proteins was rapidly liberated with RNase A from thoroughly washed rat liver nuclei [5]. The liberation reaches a plateau immediately, and the plateau level is more or less the same with 2.5 and 40 µg·mL⁻¹ concentrations of the enzyme. In contrast, DNase I liberates S1 proteins in a time and dose-dependent manner, and at almost twice as much as RNase A does. In this case, a half of the S1 proteins is abruptly released with a small amount of DNase I, then followed by the release of remaining S1 proteins in accord with chromatin-digestion. Noticeably, RNase A does not now liberate S1 proteins from nuclei pretreated with DNase I [5]. Accordingly, we presume that in rat liver nuclei, about 20% of S1 proteins exist bound to hnRNAs in the soluble nucleoplasm (Fig. 2), 40% in association with nascent hnRNAs bound to chromatin, and remaining 40% through binding to chromatin DNA.

The results that about 60% of the input [³H]hnRNAs in nucleoplasmic extracts was precipitated by antibody 351 (Fig. 5) suggest that S1 proteins C2 and D2 should constitute a general, if not major, RNA-binding activity in the nucleus. In fact, the S1 proteins and hnRNP core proteins are coprecipitated from the nuclear extracts both by antibody 351 and by the core protein-specific 4E4 (results will be reported elsewhere), thus suggesting that both S1 proteins and hnRNP core proteins are likely present in the same hnRNP complexes.

In hnRNP complexes, S1 protein D2 was found at the sites hypersensitive to RNase, while protein C2 was at sites less sensitive and recovered in 40 S particles after mild nuclease digestion. Close inspection of Fig. 7 (panels B and C) indicates that the S1 protein C2-containing RNP particles sedimented slightly slower than the canonical 40 S particles, although the difference was subtle. The results indicate that the 40 S particles have structural heterogeneities.

A search in the database(Swiss-PROT using BLAST of GenomeNet) revealed that S1 protein D2 (CBF-A) belongs to a family of 10 proteins. The family members share highly conserved sequences of two consecutive RNA-binding motifs of about 166 residues (RNP I and II, Fig. 10), a characteristic stretch of QPKEVYQQQQ(Y/F)(G/S), and a 22 residue C-terminus. They differ in the N-terminal region of 67–89 amino acids, and the 12–64 amino-acid-region preceding the C-terminus. (A similar size of another family with closely related structures was noticed. AUF1s also called hnRNP D proteins belong to this sibling family [45].) Interestingly, proteins of these families commonly behave as molecules larger than actual in SDS/PAGE. This was also the case with the protein D2 (its theoretical MW is 30800, while the observed value was 39 400).

Figure 10

CBF-A family members. Rat CBF-A (S1 protein D2) and proteins with similar sequences were aligned. They were human Apobec-1-binding protein (ABBP-1) and type A/B hnRNP protein (type A/B), cat DBP 40 protein, rat type A/B hnRNP-like protein p40 (p40) and AlF-C1, and chicken single-strand D-box binding factors 1 and 2 (ssDBF 1 and 2) and ribonucleoprotein-1 (CRP-1). Residues identical to those of rat CBF-A are indicated by dashes, and gaps are by dots. Their common structural features are: a N-terminal variable region, conserved two consecutive RNA-binding motifs, RNP I and RNP II (conserved core heptamer RNP2 and octamer RNP1 are overscored for each motif), the second 16–64 amino-acid-variable region, and the 22-residue C-terminus consisting a perfectly conserved sequence RRGGHQNNYKRY. Peptides 1 and 2, and 6 and 7 of protein D2 (Fig. 8) locating in the N- and C-terminal variable regions distinguished the protein D2 from other family members. The three peptides determined for protein C2, underlined along the D2 (CBF-A) sequence, as well as the immunoreactivity to McAb 351 suggested the identity of protein C2 with p40 (see text). This p40 and CBF-A are identical except for the presence or absence of the 47-residue Gly/Tyr-rich sequence in the C-terminal region. Rat pRM10 protein (298 amino-acid residues, accession no: AF108653) is similar to CBF-A, whose N-terminal 52 amino-acid region is replaced by a completely different 65 amino-acid sequence (data not shown).

Noticeably, the CBF-A family members (Fig. 10) have various functions. Type A/B hnRNP protein was purified originally as a minor hnRNP protein from 40 S hnRNP particles, and has an activity that disrupts the secondary structure of duplex RNAs and decreases their melting temperatures [46,47]. Apobec-1-binding protein participates in mRNA-editing through binding in the nucleus to both apobec-1 (the catalytic component of editosome) and its substrate apolipoprotein B mRNA [48]. On the other hand, DBP 40 protein binds to specific sequences of feline parvovirus single-stranded DNA and blocks viral DNA replication [49]. Other members of this family have been characterized to function by binding to DNA mostly as transcription factors: type A/B hnRNP-like protein p40 activates the serine protease inhibitor 2 genes (accession no: RNO238854). AlF-C1is a repressor that suppresses the aldolase B gene promoter [50]. Likewise, single-strand D-box binding factor 1 (ssDBF 1) represses estrogen-induced transcription of apoVLDL II gene by binding specifically to the lower (template) strand of the regulatory E1D site [51]. ssDBF 2 contains a Gly/Tyr-rich 51 amino-acid-insertion in the C-terminal region. Chicken ribonucleoprotein (CRP-1) was cloned with double-stranded DNA probes derived from an enhancer element of chicken αA-crystallin gene [52]. Whether it binds RNA, however, is unknown. CBF-A is also a transcription factor. It acts on CArG box elements [30,31]. CArG box regulatory elements are composed of CC(AT-rich)₆GG sequences, and required for transcription of muscle-specific genes such as α-actin gene [53,54]. CBF-A binds to the CArG boxes with high affinities, with a preference for single-stranded DNA [30,55,56]. While CBFs 1–4 and the serum response factors (SRFs) bind to these boxes as transcriptional activators [57,58], CBF-A acts on the elements as a repressor [30].

As summarized above, some of the family members participate in the post-transcriptional processes by binding to RNA, while others participate in the transcriptional regulation or in the DNA replication through binding to DNA. The extensive sequence homologies of the two RNP motifs and the C-termini as well as their similarity in overall structure (Fig. 10) suggest that, besides their evolution from a common ancestral gene, the family members are thought to share a common mechanism of action. This may be structural, such as arrangement or organization of higher structure of RNA or DNA as seen with human type A/B hnRNP protein. At present, only S1 protein D2 (CBF-A) is shown to act on both RNA and DNA among the family members.

The S1 protein C2 is about 4-kDa larger than protein D2 in SDS/PAGE. The three partial sequences determined for protein C2 matched perfectly the sequences of protein D2 (CBF-A), the type A/B hnRNP-like protein p40, or AlF-C1, the latter two of which are a transcriptional activator and a repressor, respectively, as indicated above. Noticeably, the protein p40 has an exactly identical sequence to that of CBF-A, except for an additional Gly/Tyr-rich 47 amino-acid insertion in the C-terminal region (Fig. 10). Thus, it is an isoform of CBF-A, likely being produced by an alternative splicing. The same structural relationship with or without a Gly/Tyr-rich sequence is observed with ssDBF isoforms 1 and 2 (Fig. 10). More imporatntly, the protein p40 and AlF-C1 differ in only four amino acids in the N-terminal region. However, the AlF-C1 protein (kindly given by K. Tsutsumi, Iwate Univesity, Japan) did not react with McAb 351 in immunoblotting. Based on these, we suggest that S1 protein C2 is an isoform of protein D2, and very likely to be identical to the protein p40, that is, the type A/B hnRNP-like protein p40 (accession no: RNO238854).

There are bifunctional proteins acting as a transcription factor and an RNA-binding protein. Some examples are an RNA polymerase III transcription factor, TFIIIA [59,60], the Y-box proteins in Xenopus oocytes (reviewed in [61]), bicoid protein in Drosophila oocyte (reviewed in [62]), and Wilm's tumor suppressor protein WT1 [63–65]. Some hnRNP proteins are also included in this category. They are hnRNP K, hnRNP D (AUF1), and hnRNP U. They act as transcriptional activators, or repressors [22–29]. The present results suggest that S1 protein D2 (CBF-A) is included in this group of hnRNP proteins as a new member. The protein C2 may also be the same, as discussed above. The possible in vivo multifunctional roles of S1 proteins should, however, be attested by further study.

In conclusion, the present results demonstrated that S1 proteins C2 and D2, which we isolated in 1983, are novel hnRNP proteins.

Acknowledgements

This work was supported by grants-in-aid from Osaka City University and from the Ministry of Education, Science, Sports and Culture of Japan. We thank Drs Takeshi Mita and Shinji Kamada (Osaka University) for their gift of pSR-CBF-A-N, Dr Jeffrey Wilusz (New Jersey Medical School) for monoclonal antibody 4E4, Dr Yoshimi Okada (Teikyo University) for tobacco mosaic virus, and Dr Ken-ichi Tsutsumi (Iwate University) for AlF-C1. We also thank Dr Katsuji Tsugawa (Osaka Women's University), and Takanori Watanabe (Osaka University) for their comments.

Footnotes

Note: CBF-A differs from other transcription factors of the same name (CBF), i.e. the core-binding factor also termed PEBP2, AML-1, or Runt, and the subunit A of a major CCAAT-binding factor also called NF-Y.