hnRNP F Influences Binding of a 64-Kilodalton Subunit of Cleavage Stimulation Factor to mRNA Precursors in Mouse B Cells

Cell culture and nuclear extract preparation.

Mouse lymphoid cell lines were grown by the Cell Culture Center (Cellex Biosciences, Inc., Minneapolis, Minn.) in “slosh” culture in Iscove's modified Dulbecco medium containing 5% horse serum, as described previously (7, 23). A20 lymphoma cells are representative of a memory B-cell line (IgG2a sec-to-mb ratio, 1:1), and the AxJ myeloma line behaves like a plasma cell line in its IgG2a (sec-to-mb ratio, >30) expression pattern (23).

Nuclear extracts were prepared as described previously (7) with the following modifications. The nuclei were extracted in buffer containing 300 mM NaCl, 0.5 mM phenylmethylsulfonyl fluoride, 100 μg of tosylsulfonyl phenylalanyl chloromethyl ketone/ml, and 1 μg of aprotinin, leupeptin, pepstatin A, and trypsin-chymotrypsin inhibitor/ml. The nuclear extracts were dialyzed against a buffer containing 20 mM HEPES (pH 7.9), 230 mM potassium glutamate, 1 mM MgCl₂, 0.2 mM EDTA, 20% (vol/vol) glycerol, and protease inhibitors as described above. Protein concentrations were determined by a Bradford assay (Bio-Rad, Hercules, Calif.) and were typically between 4 and 6 mg/ml. Extracts prepared by this method are competent for in vitro cleavage, polyadenylation, and splicing (data not shown).

Transfection of AxJ cells with recombinant hnRNP F.

Actively growing cells were transferred to Opti-MEM (Gibco/BRL), transfected with DNA containing hnRNP F cloned into the Flag tag vector (Sigma Chemical Co., St. Louis, Mo.) plus Superfect reagent (Qiagen), incubated for several hours at 37°C and 5% CO₂, and then diluted with medium containing serum and antibiotics for an overnight incubation. The next day cells were diluted about twofold to approximately 5 × 10⁵/ml and dispensed to a 24-well plate. On day 3, 1 ml of medium with serum, antibiotics, and 1 mg of G418 (Genticin; Gibco/BRL)/ml was added per well. Cells were fed for several more days with the same G418-containing medium. Individual clones arose about 10 to 14 days later and were analyzed as indicated below.

Plasmid constructs and in vitro transcription of RNA.

The IgM sec-pA site constructs were generously provided by Cathy Phillips (30). The construct H11G11 is labeled mu-sec, H00G11 is labeled mu-sec(U→G), and H11G00 is labeled mu-sec(ΔGU). The mu-sec construct contains the weak Ig mu-sec-pA site cloned into the pGEM-T vector. The mu-sec(U→G) construct mutates the two upstream AU-rich elements, including the AAUAAA polyadenylation signal, by replacing U residues with Gs (Fig. 1B). The mu-sec(ΔGU) construct deletes the two GU-rich elements downstream of the cleavage site. The clones used to produce T7 promoter-driven, ³²P-labeled antisense RNA to probe the Northern blots of the transformed AxJ cells were derived from GAPDH (glyceraldehyde-3-phosphate dehydrogenase) and a clone specific to the IgG2a CH3 exon (exon 347), both previously described (9). The mutation of the μ-mb-pA (mu-mb-pA) site was accomplished by oligonucleotide mutagenesis (38).

Plasmids were linearized and transcribed with either SP6 or T7 RNA polymerase in a 20- to 100-μl reaction mixture containing 500 μM cap analog [m7G(5′)ppp(5′)G], 250 μM recombinant GTP, 25 or 250 μM recombinant UTP, (depending on the specific activity of the RNA), 500 μM (each) recombinant ATP and CTP, 20 U of RNasin RNase inhibitor (Promega Corp., Madison, Wis.), and 20 to 50 μCi of [α-³²P]UTP (New England Nuclear, Boston, Mass.). For biotinylated transcripts, 25 to 30 μM biotin-11-UTP (ENZO Diagnostics, Inc.; supplied through Sigma Chemical Co.) was included, resulting in the incorporation of from one to three biotins per molecule. Transcription reactions were carried out at 37 to 40°C for 60 to 90 min, the template was digested by DNase I for 15 min, and the transcribed RNA was extracted with phenol-chloroform, precipitated with ethanol, and purified either over a G-50 spin column or from a urea-polyacrylamide gel as previously described (7).

An hnRNP F cDNA clone in a pET vector (His tagged) was kindly provided by Doug Black (5). The BamH1 fragment containing the entire hnRNP coding portion plus a small segment of a multicloning site was removed from this clone and inserted into the unique BamH1 site of pFlag-tag (Sigma; E1775). The sequence of the resulting clone (clone 407) was determined, and the hnRNP F protein was verified as being in frame with the vector AUG from the Flag epitope.

Western analysis and antibodies.

Protein samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), electroblotted to a polyvinylidene difluoride (PVDF) membrane (PolyScreen PVDF transfer membrane; NEN Life Science Products, Inc., Boston, Mass.), and probed with antibodies. Antipeptide antibodies against CstF-64, CstF-77, and CPSF-160 were raised in rabbit hosts as previously described (18). The 3A7 monoclonal, anti-CstF-64 antibody was generously provided by Clint MacDonald.

Rabbit antisera variously recognizing hnRNP H and H′ and/or F were raised by coupling peptides corresponding to the predicted amino acid sequences of the human hnRNP proteins through a terminal cysteine to keyhole limpet hemocyanin with gluteraldehyde as described previously (18). Injections and collections of sera were performed at Charles River Pharmservices (Southbridge, Mass.). An antiserum (NH2) that reacts with both hnRNP H and H′ as well as F was generated for human hnRNP H′ (amino acids 4 to 21 with a carboxyl-terminal cysteine; STEGREGFVVKVRGLPWSC). An antiserum recognizing predominantly hnRNP H and H′ (C) was generated for CDQVLDENSSDYQSN, corresponding to amino acids 434 to 447 of H′. An Antiserum recognizing predominantly hnRNP F (I) was generated for CTARRYIGIVKQAGLER, corresponding to amino acids 215 to 230 of the F sequence. An equal mixture of C and I antibodies was used for most of the Western blots.

Antiserum to the whole CstF-64 molecule was raised by injecting chickens with a recombinant glutathione S-transferase–CstF-64 fusion protein; the IgY was prepared from eggs. Injections and collections of antibodies were performed at Charles River Pharmservices. Rabbit polyclonal antibodies against CPSF-100 and CPSF-30 were generously provided by Silvia Barabino, Ursula Rüegsegger, and Walter Keller. Chicken anti-rat polypyrimidine tract binding protein (PTB) was obtained from Paula Grabowski. Anti-GAPDH antibodies were purchased from Chemicon Corp. Peroxidase-coupled secondary antibodies were purchased from Sigma (mouse), Roche Molecular Biochemicals (rabbit), and Jackson Laboratory (chicken) and detected by chemiluminescence. Anti-Flag epitope antibodies were purchased from Sigma.

Polyadenylation complex assembly, gel filtration chromatography, and affinity purification.

Polyadenylation complex formation was permitted to occur for 20 min at 30°C in a reaction mixture containing 15 to 25 pmol of substrate RNA with biotin and ³²P (38), 3.75 to 6.25 mg of nuclear extract, 40 g of yeast tRNA/ml, 1 mM ATP, 10 mM creatine phosphate, 0.7 mM MgCl₂, 2 mM dithiothreitol (DTT), 80 U of RNasin RNase inhibitor/ml, and 20% (vol/vol) glycerol in 20 mM HEPES, pH 7.9.

After the assembly and incubation were complete, the mixture was loaded over a 1.5- by 50-cm Sephacryl S-500HR (Amersham Pharmacia Biotech, Inc., Piscataway, N.J.) gel filtration column at 4°C and separated using a 1× PCB-1 buffer system (150 mM KCl, 0.1% Triton X-100, 0.5 mM [each] DTT and phenylmethylsulfonyl fluoride in 20 mM Tris, pH 7.6). Prior to use, the columns were blocked with nuclear extract to minimize the nonspecific interaction of complex-associated proteins with the matrix. Columns were run at a flow rate of 0.35 ml · min⁻¹. Fractions of 1.25 ml were collected until at least one bed volume was eluted from the column.

Cerenkov counts (20) were acquired for 100 μl of each collected 1.25-ml fraction to determine the elution profile of the RNA from the column. Peak fractions were pooled for subsequent purification. A portion of each pool was removed for analysis of total (unselected) protein content. The remaining samples were incubated with 15 to 20 μl of avidin-D–agarose (Vector Laboratories, Inc., Burlingame, Calif.)/ml with gentle rocking at 4°C for at least 6 h. The affinity resin was then washed at 4°C with gentle rocking once for 15 min plus three times for 10 min with 150 mM NaCl in 20 mM Tris, pH 7.6, and then once for 10 min with 50 mM NaCl in 20 mM Tris, pH 7.6. Cerenkov counts for the washed resin were measured to determine the fraction of input RNA associated with the resin. Approximately 60% of the RNA counts in pooled fractions containing biotinylated RNA bind to the avidin resin, versus less than 1% binding of nonbiotinylated RNA.

The RNA-protein complexes were released from the affinity resin by incubation with 0.6 μg of RNase A at 37°C for 20 min. The proteins were then eluted with 250 μl of elution buffer (2% SDS, 20 mM DTT in 20 mM Tris, pH 7.6) with gentle rocking at 4°C for 10 min. The resin was recounted to determine the efficiency of RNase incubation and elution; generally, 85 to 95% of bound counts were released. The eluents were heated to 65 to 70°C for 5 to 10 min, 40 μg of yeast tRNA or glycogen carrier was added, and 4 volumes of ice-cold acetone were added to precipitate the proteins. The samples were incubated on ice for 10 to 20 min and pelleted in a 4°C microcentrifuge for 20 min. The air-dried pellets were resuspended in 1.5× SDS sample buffer (0.15 M DTT, 3% SDS, 120 mM Tris [pH 6.8], 15% glycerol, 0.3% bromophenol blue) for subsequent analysis by SDS-PAGE and Western immunoblotting.

UV cross-linking assays of nuclear extract proteins, immunoprecipitations, and cleavage reactions.

In vitro reaction mixtures for UV cross-linking of nuclear extract proteins to RNA contained 60 fmol (at least 6 × 10⁵ cpm) of high-specific-activity ³²P-labeled substrate RNA, 5 μg of nuclear extract, 40 μg of yeast tRNA/ml, 1 mM ATP, 0.7 mM MgCl₂, 0.2 mM EDTA, 10 to 20% glycerol, and 230 mM potassium glutamate in 20 mM HEPES, pH 7.9. Reaction mixtures were incubated at 30°C for 5 min, placed on ice, and irradiated with 1.8 J of UV light/cm². Unprotected RNA was digested with 50 g of RNase A/ml at 37°C for 15 to 20 min; proteins tagged with radiolabeled RNA were subjected to preclearing and then to immunoprecipitation with anti-CstF-64 antibodies coupled to protein A-Sepharose. Antigen-antibody-resin was washed with 50 mM Tris, pH 7.4–50 mM NaCl–0.5% NP-40 and resuspended in 20 μl of SDS-gel loading buffer, boiled for 5 min, and resolved by SDS–7.5% PAGE. RNA-protein complexes were visualized by autoradiography.

Cleavage assays with AxJ nuclear extracts were conducted as previously described using a 32P-labeled pSVL runoff transcript (38). A 32P-labeled runoff transcript of pSVL cut with HincII, 6 nucleotides shorter than the cleavage product, was included in the 5% acrylamide–8 M urea gel as a marker.

Band shift assays.

Band shift assays with SV40 late pre-mRNA substrates were performed as previously described (1, 2). SVL-Gem, which contained a 22-base substitution at positions 202 to 224 (from the transcription start site) that removes sequences downstream of the CstF binding site on SVL pre-mRNA, was prepared by a PCR approach. PCR mixtures containing pSVL, the SP6 promoter-specific primer (5′ CATACGATTTAGGTGACACT), and a mutagenic primer (5′-ACCGAGCTCGAATTCCGTGTATTCTGAACCTGAAACAT) yielded a 247-bp DNA fragment. The substituted region contains sequences from the pGEM4 polylinker. Transcription of this template yielded a 224-base RNA.

Purification of recombinant hnRNP F.

His-tagged hnRNP F was purified as previously described (5). The bacterial cell pellet was incubated in 50 mM NaPO₄, pH 8.0–300 mM NaCl–200 μg of lysozyme/ml and then sonicated and brought to 6 M urea at room temperature. The denatured proteins were loaded onto a Ni-nitrilotriacetic acid column and refolded on the column by gradually decreasing the urea concentration to 1 M. The protein was eluted with buffer containing 0.25 M imidazole, and imidazole was removed by extensive dialysis.

GU- or U-rich and AAUAAA element-dependent binding of CstF-64 to pre-mRNA is increased in plasma cells in the absence of an increase in the amount of protein.

We analyzed the ability of CstF-64 to bind to ³²P-labeled pre-mRNAs made from Ig mu-sec constructs (Fig. 1B) by immunoprecipitation of CstF-64 from nuclear extracts prepared from A20 memory B or lymphoma cells or AxJ plasma or myeloma cells using either polyclonal or COOH-specific anti-CstF-64 antibodies. Both extracts were competent for polyadenylation and cleavage (7). The results shown in Fig. 2A for the polyclonal anti-CstF-64 antiserum confirm that there is increased UV cross-linking of the Ig mu-sec poly(A) site to CstF-64 in the plasma extract relative to that in the memory cell extract. The results further demonstrate that CstF-64 binding in the plasma cell extract is disrupted upon mutation of either the AAUAAA binding sites recognized by CPSF [mu-sec(U→G) construct] or the GU- or U-rich downstream element seen by CstF [mu-sec(ΔGU) constructs]. Levels of protein CstF-64 in the extracts representing different B-cell stages and endogenous Ig sec-to-mb mRNA ratios are identical as shown by the Western analyses using two different antibodies to CstF-64, either a serum prepared in chickens containing a polyclonal antibody to the whole molecule (Fig. 2B, lane 1 versus 2) or a rabbit antiserum to the COOH-terminal 25 amino acids (Fig. 2B lane 3 versus 4). As a loading control the blot was probed with an antibody against PTB also located in the nucleus (42); the results in Fig. 2B show that there was an equal amount of protein in the memory and plasma cell nuclear extracts. Antibodies to the proteins B, D, and D′ of the small nuclear RNPs were also used to verify equal protein loads (data not shown).

hnRNP F level versus H or H′ level is decreased in plasma cells.

The hnRNP H/H′/F subfamily consists of three highly homologous but distinct RNA-binding proteins encoded by genes mapping to separate chromosomes (12, 21). The hnRNP H′ (DSEF-1) was identified because it bound to the G-rich element downstream of the SVL poly(A) site and was subsequently shown to bind to the downstream elements of several poly(A) sites with only GU-rich elements, such as the Ig mu-mb-pA site (1, 2, 31). Given their affinity for G- and GU-rich sequences (5, 24), one or more members of the hnRNP H/H′/F subfamily, we hypothesized, might influence CstF-64 binding to the sequences found downstream of polyadenylation sites.

Using antipeptide antibodies that distinguish between the H and F members of the hnRNP H/H′/F family (see Materials and Methods) we examined the protein levels in whole-cell lysates prepared from various plasma and B-cell lines. Figure 3 shows the results of a representative Western analysis of hnRNP H, H′, and F proteins, demonstrating at least a twofold decrease in hnRNP F protein levels and at least a twofold increase in the H and H′ species in plasma cells relative to immature (70Z), mature (WEHI231), and memory (A20) B-cell lines. We conclude that the F-to-H or -H′ ratio is decreased appreciably in plasma cells.

Large complexes containing hnRNPs H and H′ and hnRNP F are associated with the mu-sec-pA site in memory B-cell extracts.

We wanted to determine if hnRNPs H and H′ and hnRNP F were differentially associated with exogenously added pre-mRNA in nuclear extracts prepared from memory B versus plasma cells. We used a two-step purification procedure to isolate and characterize RNA-protein complexes assembled in vitro (38). Nuclear extract was incubated with biotinylated mu-sec pre-mRNA (Fig. 1B) with a ³²P-labeled tracer; the resultant complexes were fractionated by gel filtration on Sephacryl S-500HR. The bulk of the labeled RNA eluted at between 60 and 75% of the bed volume of the column, as shown in Fig. 4A, in both the memory B and the plasma cells. A portion of the RNA elutes in the void volume (40% of the bed volume) in very large complexes. The RNA eluting at >80% of the bed volume contained free nucleotides, short RNAs, and nucleolytic degradation products; the amounts of these varied from RNA preparation to preparation. Figure 4A indicates the fractions between 40 and 80% of the bed volume that were pooled for affinity purification and Western analysis. Thyroglobulin, used as a protein molecular mass marker (669,000 Da), eluted from the Sephacryl column at approximately 75 to 80% of the bed volume. We analyzed the total protein levels in the 40- to 80%-bed-volume samples by removing a portion of the fractions eluted from Sephacryl before the avidin affinity step; the immunoblot of those unselected fractions is shown in Fig. 4B. Coincident elution of CstF subunits CstF-64 and -77 occurred in a broad range of fractions from 65 to 85% of the total column volume in both cell types (Fig. 4B, lanes 2 to 6 and 9 to 13). Factors CPSF-160, -100, and -30 coeluted with the CstF subunits in the same broad range of fractions (not shown). The bulk of the polyadenylation and cleavage factors must remain associated with endogenous RNA to form these complexes. The elution of hnRNP H and H′ and hnRNP F from the column without the affinity selection was somewhat variable; representative Western blots are shown in Fig. 4B and 6. While the elution of these hnRNPs was broad, the bulk of the hnRNPs eluted predominantly at 70 to 80% of the bed volume in both cell types.

As a second step of purification of the RNA-protein complexes, pooled fractions from the column, as indicated in Fig. 4A, were incubated with avidin-agarose overnight at 4°C as previously described (38). The proteins in the resin-bound, biotinylated RNA-protein complexes were released by treatment with RNase and SDS and analyzed by SDS–8% PAGE and Western immunoblotting. The CstF-64 and -77 proteins that were affinity purified with the biotinylated RNA eluted from the Sephacryl column predominantly between 60 and 70% of the bed volume (Fig. 4C). Copurifying with CstF-RNA complexes were the CPSF subunits of 160, 100, and 30 kDa (data not shown). The affinity-purified proteins elute at a position in the column commensurate with their forming a complex on the input RNA, with an apparent molecular mass on the order of 2 × 10⁶ to 4 × 10⁶ Da, in both cell types. This is slightly larger than the average size of the endogenous complexes. The amounts of the CstF-64-, CstF-77-, and CstF-50-containing complexes were significantly reduced in the memory B-cell extracts compared with plasma cells; note the relative intensities of the CstF-77 lanes in Fig. 4C. In order to show detail, the CstF-64 signal from memory cells in Fig. 4C was exposed for a much longer time than the CstF-64 signal from plasma cells.

The striking observation was that, in the memory B cells, large complexes containing hnRNP H, H′, and F proteins were formed with mu-sec pre-mRNA (Fig. 4C, lanes 3 to 5), while in plasma cells these complexes were not detectable (Fig. 4C, lanes 11 to 17) even after prolonged exposure of the blot. The hnRNP proteins were in fact present in the total fractions in each cell type (Fig. 4B, lanes 2 to 6 and 8 to 13). The affinity-purified complexes containing hnRNP H, H′, and F in the memory B cells were on average larger than those formed with the CstF subunits and eluted at about 60% of the bed volume. The absence of these complexes in nuclear extracts prepared from plasma cells might explain the observed higher efficiency with which CstF-64 binds to pre-mRNAs in those extracts (Fig. 2).

Deletion of downstream sequences, but not mutation of the AAUAAA poly(A) signal, disrupts hnRNP H, H′, and F association with mu-sec RNA.

We hypothesized that a competition between the hnRNP H, H′, and F proteins and CstF-64 for binding to the same region of the RNA could occur. We therefore determined if hnRNPs H and H′ and/or F associate with mu-sec pre-mRNA via downstream sequences. Using the Sephacryl column and an affinity purification technique we analyzed the RNA-protein complexes formed in A20 memory B-cell extracts on two mu-sec mutants (constructs are shown in Fig. 1B). In the mu-sec(U→G) construct, four U residues have been converted to Gs in the AAUAAA polyadenylation signal and adjacent upstream AU-rich element; in mu-sec(ΔGU) the two downstream GU-rich elements where CstF binds were deleted along with the adjacent downstream sequences. Disruption of either the AAUAAA polyadenylation signal or downstream sequences, respectively resulted in a loss of CstF binding to the RNA as shown in a representative experiment in Fig. 5. Association of CPSF subunits with both of these mu-sec mutants was also blocked (data not shown). However, binding of hnRNPs H, H′, and F to the RNA was retained in the mu-sec(U→G) construct (Fig. 5A), where the GU- or U-rich element and adjacent downstream sequences are undisturbed. In contrast, deletion of these downstream sequences results in a dramatic reduction of hnRNP H, H′, and F binding to the mu-sec(ΔGU) RNA (Fig. 5B). Thus, downstream sequences are required for hnRNP H, H′, and F to associate with the RNA in the memory B-cell extracts. Curiously, the region downstream of the mu-sec poly(A) site contains only limited regions of G richness. The hnRNP H, H′, and F proteins might interact with the GU-rich region either directly or through protein-protein interactions.

CPSF and CstF also failed to associate with constructs lacking downstream or AAUAA regions in nuclear extracts prepared from plasma cells (data not shown). Consistent with our observations with the wild-type mu-sec construct in plasma extracts (Fig. 4C), hnRNP H, H′, and F family proteins did not associate with either of the mu-sec poly(A) region mutants in plasma nuclear extracts (data not shown). Therefore the association of hnRNP H, H′, and F with the downstream regions of the mu-sec polyadenylation signal appears to be developmentally regulated.

Large complexes containing hnRNPs H, H′, and F are associated with the mu-mb-pA site in memory B-cell extracts.

hnRNP H, H′, and F complex formation with pre-mRNA in memory but not plasma B cells was investigated with the mu-mb pre-mRNA. Figure 6 shows the results of experiments performed using the strong mu-mb-pA site and a point mutation of the mu-mb AAUAAA hexanucleotide to AAGAAA; for simplicity, only the hnRNP profile is shown. Affinity-purified CstF eluted from the columns with mu-mb RNA at 60 to 70% of the bed volume as had been seen with mu-sec RNA in extracts from both cell types (data not shown). However, complexes of hnRNP H, H′, and F eluting at 40 to 60% of the bed volume, a position in the column larger than those of the CstF-containing fractions, were seen only in the memory B-cell extracts (compare affinity fractions in Fig. 6A and C). Mutation of the poly(A) signal hexanucleotide to AAGAAA resulted in the predicted loss of CstF-CPSF complex formation but still supported the assembly of large hnRNP H, H′, and F-containing complexes in the memory B-cell extract (Fig. 6B, affinity fraction). This observation suggests that association of the hnRNP H/H′/F family proteins with the RNA requires neither an intact hexanucleotide poly(A) signal nor assembly of a polyadenylation complex and occurs on both the mu-mbpA and mu-sec-pA sites.

In the nuclear extracts from plasma cells, the mu-mb pre-mRNA supported the formation of a polyadenylation complex containing CstF subunits and CPSF subunits eluting from the column at 65 to 75% of the column bed volume (data not shown). However, only a small amount of hnRNP H, H′, or F was capable of being affinity purified with the mu-mb RNA (Fig. 6C, affinity fraction). A similar observation was also made with the SVL poly(A) site and the Ig γ heavy-chain sites in plasma cells (data not shown). These results confirm the conclusion reached for the mu-sec site, namely, that in plasma nuclear extracts hnRNP H, H′, and F do not efficiently form complexes with input pre-mRNA. Similar results were obtained with pre-mRNA containing SVL sequences (data not shown).

Competition between hnRNP F and H′ for binding to RNA.

The data described above demonstrate that, in memory B cells, hnRNP F protein, in a large complex, can interact with the three polyadenylation signals tested. To determine if hnRNP F was capable of direct interaction with the pre-mRNA, we allowed recombinant hnRNP F to bind to ³²P-labeled SVL pre-mRNA and performed a gel retardation assay (Fig. 7A). The binding of hnRNP F to SVL pre-mRNA seen with the intact SVL pre-mRNA (lane 2) is abolished by the replacement of the G-rich region downstream of the CstF binding site with polylinker sequences (Fig. 7A, lane 4). This indicates that the binding of hnRNP F to SVL pre-mRNA is direct and in close proximity to the GU- or U-rich downstream region of the SVL pre-mRNA. Purified recombinant hnRNP F also binds to Ig μRNA and Ig γ pre-mRNAs, both at the membrane and secretory form-encoding sites, as determined by UV-cross-linking experiments (data not shown).

Since hnRNP H and H′ and hnRNP F have a preference for G-rich sequences, they may be competing for a shared binding site on the pre-mRNA. Addition of recombinant hnRNP H′ to a mixture containing hnRNP F and intact SVL pre-mRNA resulted in a shift from an F-specific complex to one that is the same size previously observed with H′ alone (1) (Fig. 7B). The ratio of recombinant H′ to F necessary to achieve this shift in complex size is approximately the same as the F/H protein ratio seen by Western analyses in memory B-cell extracts, leading us to conclude that the hnRNP F-to-H′ ratio influences the binding characteristics of F to pre-mRNA.

The amount of recombinant F protein needed for maximal binding in Fig. 7 (approximately 200 ng) is much larger than that previously measured with hnRNP H′ (1). This could be due to a low affinity of hnRNP F for polyadenylation substrates or alternatively to the relative activity of the recombinant protein preparations. Purification of His-tagged hnRNP F on a nickel column requires a denaturation-and-renaturation step (5), which may reduce its specific activity. Finally, F may be modified in eucaryotic cells, and those modifications would be lacking in the bacterially produced protein, necessitating the large amounts necessary for an effect here.

HnRNP F reduces CstF-64 RNA-binding activity and the cleavage reaction.

We wanted to test the hypothesis that excess hnRNP F, as seen in the memory cells, could interfere with CstF-64 binding to pre-mRNA containing a poly(A) site and GU- or U-rich downstream sequence. We added increasing concentrations of purified recombinant hnRNP F protein to plasma nuclear extract and then asked how efficiently CstF-64 could bind to ³²P-labeled pre-mRNA. The reaction mixture was treated with anti-CstF-64 antiserum and the amount of ³²P label UV cross-linked to CstF-64 that was immunoprecipitated was determined by autoradiography (Fig. 8). The ability of CstF-64 to bind to both the mu-sec and mu-mb pre-mRNAs was diminished when increasing amounts of recombinant hnRNP F protein were added (compare lanes 4 and 1).

While the amounts of protein added are large, addition of similar amounts of control proteins such as DSEF-1 (Fig. 8, lane 5), glutathione S-transferase, or bovine serum albumin did not interfere with the ability of CstF-64 to bind to pre-mRNA (data not shown) (1).

Since recombinant hnRNP F had a negative effect on CstF-64 binding, we wondered if it would have an influence on the cleavage reaction on pre-mRNA. Using nuclear extracts from plasma cells and SVL pre-mRNA we found that addition of increasing amounts of hnRNP F blocked production of the cleavage product (Fig. 9, lanes 5 to 7). When equivalent amounts of recombinant hnRNP H′ (lanes 2 to 4) or a His-tagged recombinant sucrose nonfermenting protein (not shown) were added to the plasma cell extract, neither enhancement nor diminution of activity was seen. When the hnRNP F-to-H′ ratio was increased, the cleavage reaction was also blocked (Fig. 9, lanes 9 to 11). We therefore conclude that high levels of hnRNP F are inhibitory for the cleavage reaction. The regulatory function of hnRNP F may extend beyond blocking CstF-64 from RNA binding since F seems to act in vitro at a lower level to block cleavage than is required to inhibit CstF-64 binding. It is known that hnRNP F blocks other cleavage and/or splicing factors (see Discussion).

Plasma cells stably transfected with hnRNP F show reduced Ig sec-to-mb mRNA ratios.

AxJ myeloma and plasma cells expressing Ig heavy-chain mRNA with a 36:1 ratio of secretory form-to membrane-encoding species were stably transfected with recombinant Flag-tagged hnRNP F driven by the cytomegalovirus promoter. A Western blot of the protein expressed in individual G418-resistant transfectants expressing various levels of the Flag-tagged hnRNP F are shown in Fig. 10. In Fig. 10A, there is a protein in the transfected cells (lanes 2 and 3) which reacts with both anti-Flag and anti-F antibodies. This Flag-F protein is missing in the untransfected AxJ cells. There is a protein that cross-reacts with the anti-Flag in the untransformed and transfected AxJ cells (lanes 1 and 2), but this protein is not related to hnRNP F (lanes 3 and 4). Several clones of AxJ transfected with F were isolated; in some but not all of the transfected cells the overall levels of Flag-F increased significantly (Fig. 10B). The increase in F levels in the transfectants is not as high as the increase in authentic F levels in memory and plasma cells (Fig. 3, compare hnRNP F in lane 3 with that in lane 4 or 5).

To determine if the increased expression of F influenced the poly(A) site choice in the resident Ig heavy-chain gene, the poly(A)⁺ selected RNA from the indicated (Fig. 11) Flag-F transfectants was obtained and probed on a Northern blot for both the Ig secretory form-encoding (1.8-kb) and membrane-encoding (3.6-kb) species (23). A sample of one of the four determinations is shown and all the data obtained are summarized in Fig. 11. In cells expressing endogenous as well as recombinant hnRNP F, the ratio of secretory form- to membrane-encoding species decreased dramatically, while transfection with the Flag vector alone had a more modest effect. In addition, the overall amount of Ig heavy-chain mRNA, relative to that of GAPDH control mRNA, was decreased in the Flag-F transfectants, consistent with what we previously observed in memory B cells (23). We conclude that overexpression of the Flag-tagged hnRNP F significantly alters the regulation of the endogenous Ig heavy-chain gene in plasma cells.