Interaction between hnRNPA1 and IκBα Is Required for Maximal Activation of NF-κB-Dependent Transcription

Abstract

Transcriptional activation of NF-κB is mediated by signal-induced phosphorylation and degradation of its inhibitor, IκBα. NF-κB activation induces a rapid resynthesis of IκBα which is responsible for postinduction repression of transcription. Following resynthesis, IκBα translocates to the nucleus, removes template bound NF-κB, and exports NF-κB to the cytoplasm in a transcriptionally inactive form. Here we demonstrate that IκBα interacts directly with another nucleocytoplasmic shuttling protein, hnRNPA1, both in vivo and in vitro. This interaction requires one of the N-terminal RNA binding domains of hnRNPA1 and the C-terminal region of IκBα. Cells lacking hnRNPA1 are defective in NF-κB-dependent transcriptional activation, but the defect in these cells is complemented by ectopic expression of hnRNPA1. hnRNPA1 expression in these cells increased the amount of IκBα degradation, compared to that of the control cells, in response to activation by Epstein-Barr virus latent membrane protein 1. Thus in addition to regulating mRNA processing and transport, hnRNPA1 also contributes to the control of NF-κB-dependent transcription.

The NF-κB/Rel family of transcription factors is composed of a number of structurally related, interacting proteins that bind DNA and whose activity is regulated by subcellular location. In vertebrates, this family includes p50 and p105, p52 and p100, and p65 Rel A, c-Rel, or Rel B, which bind DNA in a homo- or heterodimeric fashion and are implicated in regulation of a number of cellular genes involved in immune, inflammatory, and antiapoptotic responses (3, 5). Following cellular activation, NF-κB, typically a p50-p65 heterodimer, translocates to the nucleus and activates transcription of NF-κB-responsive genes. NF-κB dimerization, nuclear translocation, and DNA binding are facilitated by a conserved region known as the Rel homology domain. NF-κB transcriptional activity is controlled by the inhibitor IκB proteins, whose association with the NF-κB p50 and p65 subunits occludes their nuclear localization signals (NLSs), thereby leading to cytoplasmic sequestration, but also inhibits NF-κB DNA binding activity (27). Several IκBs have been described, including IκBα (25), IκBβ (63), IκBɛ (68), Bcl 3 (42), and the precursors of p50 (p105) and p52 (p100), which possess inhibitory ankyrin repeat domains that in isolation are known as IκBγ and IκBδ.

Following signal induction, IκBα is phosphorylated on serine 32 and serine 36 (8, 10, 52, 64) by the dimeric IκB kinase (16, 38, 47, 71, 75). Subsequently, IκBα is ubiquitinated on lysine 21 and lysine 22 (4, 51, 55), which targets the protein for degradation by the proteosome 26S complex. Although signal-induced modifications of IκBα are targeted to the N-terminal domain, the carboxyl-terminal domain of IκBα is also required for proteasome-mediated degradation (9, 34). Recognition of phosphorylated IκBα is accomplished by β-TrCP, which is a component of an E3 ubiquitin ligase complex which mediates ubiquitination of IκBα (26, 43, 56, 62, 67, 70, 74). After IκBα degradation, NF-κB translocates to the nucleus, where it induces the transcription of several genes, including that of its inhibitor, IκBα. Following IκBα mRNA translation, newly synthesized IκBα is accumulated in the cytoplasm and also in the nucleus, where it terminates NF-κB transcriptional activity (1). Termination of NF-κB-dependent transcription is achieved by inhibition of the NF-κB–DNA interaction and export of NF-κB back to the cytoplasm (2).

The mechanism by which IκBα localizes to the nucleus has not been precisely defined, but IκBα does not contain a region of basic residues that resembles previously characterized NLSs. However, nuclear entry of IκBα is conferred by a cis-acting nuclear import sequence located in the second ankyrin repeat which can also functionally substitute for the classical NLS in nucleoplasmin (54). Reconstitution of the nuclear import pathway in vitro indicates that IκBα is transported into the nucleus by a “piggy-back” mechanism that involves additional uncharacterized NLS-containing proteins that recognize the ankyrin repeats of IκBα (65). Nuclear export of IκBα is conferred by leucine-rich nuclear export sequences present in the carboxy-terminal (2) and amino-terminal (32) regions of the protein. The nuclear protein CRM1 (exportin 1), which belongs to the karyopherin β family (20), has been identified as the nuclear export sequence receptor (19, 21, 44, 57) and forms a complex with IκBα in the presence of GTP-bound Ran. It has been proposed that this ternary complex is transported through the nuclear pore complex and dissociates in the cytoplasm due to GTP hydrolysis by Ran, induced by Ran GTPase activating protein (19). While nuclear export of the NF-κB–IκBα complex can be demonstrated during the process of postinduction repression, pharmacological inhibition of CRM1 with leptomycin B leads to the nuclear accumulation of NF-κB and IκBα even in the uninduced state. Thus nuclear and cytoplasmic shuttling of IκBα is a highly dynamic process which, in unactivated cells, establishes a steady state where NF-κB is predominantly cytoplasmic (13, 23, 28, 32, 35, 48, 53, 61). Although the precise function of IκBα nuclear export has yet to be defined, the constant surveillance of the nucleus by IκBα results in tight and finely tuned control of NF-κB-dependent transcription.

In this study we demonstrate that IκBα interacts directly and specifically, in vitro and in vivo, with another nucleocytoplasmic shuttling protein, heterogeneous nuclear ribonucleoprotein A1 (hnRNPA1). This interaction is mediated by the C-terminal region of IκBα and one of the N-terminal RNA binding domains of hnRNPA1. In cells lacking hnRNPA1, NF-κB activation is defective, but reintroduction of hnRNPA1 into these cells restores an efficient NF-κB response to signal induction. In the absence of hnRNPA1, IκBα does not undergo signal-induced degradation, but IκBα degradation in response to Epstein-Barr virus latent membrane protein 1 (EBV LMP-1) (60) is restored by ectopic expression of hnRNPA1. Thus in addition to regulating splicing, polyadenylation, and mRNA transport (17), hnRNPA1 also contributes to the control of NF-κB-dependent transcription.

MATERIAL AND METHODS

Plasmid construction.

pV44ER.LexA and pKV701.VP16 (31) were received from Colin Goding (Marie Curie Research Institute, Oxted, United Kingdom), and pACT (18) was obtained from Stephen Elledge (Baylor College of Medicine, Houston, Tex.). pLexA-hnRNPA1 and pLexA-IκBα N-T, the DNA binding domain plasmids in this study, were obtained by subcloning the genes from BamHI/EcoRI sites of pCDNA3 constructs into BamHI/EcoRI-digested pV44ER.LexA. The activation domain plasmids pVP16-IκBα, pACT-lysRS (58), and pACT-Ubch9 (15) were obtained from Lesley Stark and Jill Thomson (University of St. Andrews). The IκBα wild type and IκBα S32A S36A SV5 tag were a gift from Manuel Rodriguez (University of St. Andrews). For in vitro protein-protein interactions, the hnRNPA1 wild type and derivatives, IκBα wild type (49) and derivatives, and β-galactosidase (β-Gal) chimeras (34) were expressed as [³⁵S]methionine-labeled proteins (Promega in vitro transcription-translation kit). All hnRNPA1 cDNAs and derivatives and IκBα truncations were amplified by PCR oligonucleotides containing BamHI and an EcoRI site, digested with both enzymes, and cloned into BamHI- and EcoRI-cut pCDNA3. The IκBα and hnRNPA1 wild types and truncated forms of the proteins were expressed as glutathione S-transferase (GST) fusion proteins. All cDNAs used were amplified by PCR oligonucleotides with a BamHI site and an EcoRI site, digested, and cloned into BamHI- and EcoRI-cut pGEX 2T. To allow expression in vivo, the IκBα and hnRNPA1 wild-type and truncated forms were amplified by PCR oligonucleotides with a BamHI site and an EcoRI site, digested, and cloned into BamHI- and EcoRI-cut pcDNA3 SV5 (14). The DNA sequences of the inserts in all new constructions were determined by automated DNA sequencing (Alex Houston, St. Andrews DNA sequencing service).

Western blot analysis antibodies.

Proteins were separated by electrophoresis in 10% polyacrylamide gels containing sodium dodecyl sulfate (SDS) transferred to a polyvinylidene difluoride membrane and subjected to enhanced chemiluminescence (ECL) Western blotting as described previously (58). Primary antibodies to p50 (residues 35 to 381), p65 (residues 12 to 317), and IκBα (residues 1 to 317) were raised in sheep (Scottish Antibody Production Unit, Carluke, Scotland) and were antigen affinity purified. The primary polyclonal antibody to IκBα, C-21, was raised in rabbit and purchased from Santa Cruz Biotechnology Inc. Mouse monoclonal antibody 10B, which recognizes IκBα (30), was obtained from Ellis Jaffray (University of St. Andrews), mouse monoclonal antibody which recognizes the SV5 tag (22) was obtained from Dan Young (University of St. Andrews), mouse monoclonal antibody which recognizes myc-tagged proteins was obtained from Bernie Precious (University of St. Andrews), and monoclonal antibody 4B10, which recognizes hnRNPA1, was a kind gift from G. Dreyfuss (University of Pennsylvania, Philadelphia). The secondary antibodies used to detect immobilized antibody-antigen complexes were anti-sheep horseradish peroxidase (HRP) (DAKO) and anti-mouse HRP (Amersham).

Affinity purification of protein complexes containing IκBα.

A frozen cell pellet from a 200-liter culture of B cells (Namalwa) was resuspended in lysis buffer (50 mM sodium fluoride, 5 mM tetra-sodium pyrophoshate, 1 mM sodium orthovanadate, 10 mM β-glycerophosphate, 2 mM EDTA, 20 mM sodium phosphate buffer [pH 7.5], 0.5% NP-40) containing a cocktail of protease inhibitors (100 μM Pefablock, 1 mM Nα-p-tosyl-l-lysine chloromethyl ketone, 50 μg of Bestatin/ml, 1 μM pepstatin, and 1 μM Leupeptin). Following cellular disruption by sonication, the extract was clarified by centrifugation (20,000 × g) for 30 min at 4°C. The supernatant was removed and recentrifuged (100,000 × g) for 60 min at 4°C to remove any particulate material. The supernatant was passed through a 1-ml protein A Sepharose column and then through a 1-ml preimmune column (preimmune sheep immunoglobulin G [IgG] covalently cross-linked to protein A Sepharose), and finally through a 1-ml anti-IκBα column (antigen affinity-purified IgG from sheep covalently cross-linked to protein A Sepharose). Immunoaffinity matrices were prepared as described previously (24). Anti-IκBα and preimmune columns were washed with 150 ml of lysis buffer followed by 50 ml of 10 mM triethylamine (pH 8.0). Bound proteins were eluted using 15 ml of 100 mM acetic acid. Proteins eluted from each column were freeze-dried overnight, resuspended in H₂O, and trichloroacetic acid precipitated prior to fractionation by electrophoresis in a 10% polyacrylamide gel containing SDS. Polypeptides were visualized by staining with Coomassie R450.

Protein sequencing.

Coomassie-stained polypeptides were subjected to in-gel trypsin digestion, microbore high-pressure liquid chromatography fractionation, and protein sequencing as described previously (14).

Immunoprecipitation.

Preimmune sheep IgG, sheep antigen affinity-purified anti-p65, sheep antigen affinity-purified anti-IκBα, monoclonal 4B10 (anti-hnRNPA1), and monoclonal 336 (anti-SV5 tag) were covalently cross-linked to protein A beads (24). Antibody-linked beads (10 μl) were incubated with cell extract (1 mg of protein) in incubation buffer (100 mM potassium acetate, 1 mM dithiothreitol, 20 mM Tris-acetate, 10 mg of bovine serum albumin [BSA]/ml, and 0.05% NP-40 [pH 7.5]), washed three times with the same buffer and once in distilled water. Immobilized complexes were resuspended in loading buffer (1.25% SDS and 0.35 M 2-mercaptoethanol) and resolved by electrophoresis in 10% polyacrylamide gels containing SDS. Following Western blotting, IκBα- and hnRNPA1-containing complexes were detected using monoclonal antibodies 10B and 4B10, respectively. Immobilized antibody-antigen complexes were visualized with anti-mouse HRP and ECL.

In vitro binding studies.

Glutathione beads (10 μl) containing 10 μg of the respective fusion protein were blocked for 1 h prior to use in phospate-buffered saline containing 10 mg of BSA/ml. After blocking, beads were washed once with the incubation buffer, resuspended in the same buffer with the appropriate volume of the in vitro transcription and translation product, and incubated for 1 h at 4°C. The beads were washed three times with incubation buffer and once in distilled water. Immobilized complexes were resuspended in loading buffer and separated by electrophoresis in a 10% polyacrylamide gel containing SDS, and radioactive species were detected using a phosphorimager (Fuji Bas 1500).

In vitro transcription and translation.

To generate ³⁵S-labeled IκBα and hnRNPA1 proteins, pCDNA and the appropriate linearized cDNA constructs were used as the template in the TNT coupled-wheat germ extract system (Promega). Proteins were translated in a final volume of 50 μl in the presence of 20 μCi of ³⁵S-labeled methionine (Amersham). Proteins expressed were detected and standardized using a phosphorimager (Fuji Bas 1500).

Yeast II hybrid analysis.

The Saccharomyces cerevisiae L40 reporter strain was used in the yeast II hybrid interaction assay, and transformations were carried out as described previously (58). Cotransformants were grown on Sabouraud's dextrose plates with differing levels of 3-amino triazole (3AT) (0 to 30 mM), and β-Gal activity was measured qualitatively using filter lifts.

Purification of GST fusion proteins and recombinant preparation.

GST fusion proteins were purified from isopropyl-β-d-thiogalactopyranoside-induced Escherichia coli by binding to glutathione agarose, essentially as described previously (30). Fusion proteins bound to beads were washed with lysis buffer, and a small fraction of the beads was resuspended in gel loading buffer, and eluted protein was resolved by SDS-polyacrylamide gel electrophoresis (PAGE). Coomassie blue staining using BSA as a standard was used to quantitate GST fusion proteins bound to the glutathione agarose beads. Fusion proteins were eluted from the beads in a solution containing 10 mM glutathione, 0.5 M NaCl, and 50 mM Tris-HCl (pH 8.0) and cleaved with thrombin. Phenylmethylsulfonyl fluoride (1 mM) was added to stop the reaction, and the resulting solution was dialyzed overnight (against phosphate-buffered saline containing 0.5 M NaCl and 2 mM dithiothreitol) to remove excess glutathione. GST and incompletely cleaved GST fusion proteins were removed from the protein preparations by passage over a glutathione agarose column. Protein purity was determined by SDS-PAGE and Coomassie blue staining.

Electroporation, reporter assays, and Western blotting.

Hela or CB3 cells (5 × 10⁶) were incubated in 50 μl of 200 mM NaCl containing 10 μg of plasmid DNAs, 30 μg of salmon testes DNA (Sigma) and electroporated (Easyject plus; EquiBio) at 240 V and 1,200 mA for 40 ms. Following electroporation, cells were incubated in growth medium for 16 h and processed for luciferase and β-Gal activity as described previously (50) or ECL Western blotting (58).

Oligonucleotide primers.

Oligonucleotide sequences are shown in sense orientation with restriction sites underlined. Forward primers were as follows: hnRNPA1 1–320, GTCGGATCCATGTCTAAGTCAGAGTCTCCT; hnRNPA1 196–320, AGAGGATCCATGAGTGGTTCTGGAAACTTTGGT; hnRNPA165–320, ATTATATGGATCCGTGGAGGAGGTGGATGCAGCT; hnRNPA1 75–320, ATAGGATCCATGAGGCCACACAAGGTGGAT; hnRNPA1 85–320, TAGGATCCATGGAACCAAAGAGAGCTGTCTCC; hnRNPA1 90–320, CGGATCCATGGTCTCCAGAGAAGATTCT; hnRNPA1 95–320, ATAGGATCCATGTCTCAAAGACCAGGTGCC; hnRNPA1 105–320, GCGGGCGGATCCATGAAAAAGATATTTGTTGGTGGC; hnRNPA1 142–320, AGTCGGATCCATGAGTGGCAAGAAAAGGGGCTTT; hnRNPA1 162–320, GCGCCGCGGGATCCATGATTGTCATTCAGAAATACCAT; hnRNPA1 182–320, CGGGATCCATGTCAAAGCAAGAGATGGCTAGT; and IκBα 1-317, GTACTAGGATCCATGTTCCAGGCGGCCGAG.

Reverse primers were as follows: hnRNPA1 320–1, GCCGCGAATTCTTAAAATCTTCTGCCACTGCC; hnRNPA1 320–1(-SC), CGGAATTCAAATCTTCTGCCACTGCC; hnRNPA1 267–1, TATATTAAATTGAATTCGTTGTAATTCCCAAAATCATT; hnRNPA1 247–1, CGCGGGAATTCAAATCCATTATAGCCATCCCC; hnRNPA1 227–1, AGAATTCGCCACCACGACCACTGAAGTT; hnRNPA1 207–1, CGAATTCTCCACGACCACCACCAAAGTT; hnRNPA1 196–1, GGAATTCTCGACCTCTTTGGCTGGA; IκBα 296-1, CGCGAATTCTGACGTGAACTCTGACTCTGT; IκBα 303–1, CGAATTCTGACAGCTCGTCCTCTGTGAA; and IκBα 313-1, CGAATTCTGACTGGCCTCCAAACACACA.

RESULTS

hnRNPA1 interacts with IκBα.

To identify proteins involved in the activity of the transcription factor NF-κB, proteins bound to the IκBα inhibitor were isolated by immunoaffinity chromatography. An extract from Namalwa cells was first passed over a column of protein A agarose. The flow-through from this column was then passed over a second column of protein A agarose to which preimmune IgG was linked. Finally the extract was passed over a protein A column to which anti-IκBα was linked. Both IgG columns were extensively washed, and bound proteins were eluted from the separated columns with acetic acid. Western blot analysis of the load and flow-through fractions with an IκBα antibody indicated that IκBα was present in the cell extract but was absent from the flow-through (Fig. 1A). Western blot analysis of the acetic acid eluates indicated that IκBα along with NF-κB p50 and p65 were bound to and eluted from the anti-IκBα column but were not present in the eluate from the preimmune column (Fig. 1B, C, and D). These data indicate that the immunoaffinity purification procedure was functioning efficiently. Eluted proteins were fractionated by electrophoresis in a polyacrylamide gel containing SDS and stained with Coomassie brilliant blue. In addition to the NF-κB and IκBα proteins, a prominent species migrating as a doublet of 34 and 38 kDa was identified (Fig. 1E). After in-gel digestion with trypsin and high-pressure liquid chromatography fractionation, a number of peptides were analyzed by Edman degradation. Sequence analysis indicated that peptides from both species were derived from hnRNPA1 (Fig. 1F), which shuttles between the nucleus and the cytoplasm and is involved in export of mRNA from the nucleus to the cytoplasm. The 34-kDa species is the most abundant form of A1, while the 38-kDa form, termed A1B, is a differentially spliced form of A1 containing an additional exon (11).

FIG. 1.

Affinity purification of IκBα complexes from a B-cell extract. A protein extract from Namalwa cells was first passed through a column of protein A Sepharose and then through a column of preimmune IgG linked to protein A Sepharose (PI) and finally through a column of anti-IκBα IgG linked to protein A Sepharose (α-IκBα). The cell extract prior to passage over the affinity columns (Load) and after passage over the affinity columns (FT) was analyzed by Western blotting (WB) with an IκBα antibody (A). Proteins bound to the immunoaffinity matrices were eluted with acetic acid (HAc), and each of the fractions (lanes 1 to 4) was analyzed by Western blotting with antibodies to IκBα (B), NF-κB p65 (C), NF-κB p50 (D), or Coomassie blue staining of the polyacrylamide gel (E). Stained polypeptides were subjected to in-gel trypsin digestion, and peptides were sequenced by Edman degradation. The sequence output and identification of the peptides are indicated (F).

To confirm the interactions between endogenous hnRNPA1 and IκBα–NF-κB complexes, HeLa cell extracts were immunoprecipitated with polyclonal antibodies to p65, IκBα, or preimmune IgG. Immunoprecipitated proteins were analyzed by Western blotting with monoclonal antibodies recognizing hnRNPA1 or IκBα. Both forms of hnRNPA1 were immunoprecipitated by antibodies directed against IκBα or NF-κB p65 but not by preimmune IgG (Fig. 2A). As expected, IκBα was immunoprecipitated by antibodies to both p65 and IκBα (Fig. 2A). IκBα was also detected in immunoprecipitates using an hnRNPA1 monoclonal antibody but was not detected when an irrelevant monoclonal antibody was used (Fig. 2B).

FIG. 2.

Interaction between IκBα and hnRNPA1 in HeLa cells. (A) Extracts from HeLa cells were immunoprecipitated with preimmune IgG (PI), antibodies to NF-κB p65 (α-p65), or antibodies to IκBα (α-IκBα), and immunoprecipitates were analyzed by Western blotting with the 4B10 monoclonal antibody to hnRNPA1 (WB α-hnRNPA1) or the 10B monoclonal antibody to IκBα (WB α-IκBα). (B) Extracts from HeLa cells were immunoprecipitated with the 4B10 monoclonal antibody (α-hnRNPA1) or an irrelevant monoclonal antibody (α-SV5) and analyzed by Western blotting with the 4B10 monoclonal antibody to hnRNPA1 (WB α-hnRNPA1) or the 10B monoclonal antibody to IκBα (WB α-IκBα).

The interaction between hnRNPA1 and IκBα was also examined in a heterologous system using the yeast two-hybrid system. When the yeast L40 reporter strain was cotransformed with the LexA hnRNPA1 and IκBα-VP16 expression constructs, an interaction between the two expressed proteins was detected. Thus the yeast could grow on minimal medium containing 3AT but lacking histidine and could activate the LexA-dependent β-Gal reporter (Table 1). Appropriate positive and negative controls confirmed the specificity of this interaction (Table 1).

TABLE 1.

hnRNPA1 interacts with IκBα in vivo^a

Construct	Growth of yeast		Colony color
	(+ His)	(− His + 3AT)
pLexA-hnRNP A1, pVP16.701 vector	+	−	White
pV44ER.LexA vector, pVP16-IκBα wt	+	−	White
pLexA-hnRNP A1, pVP16-IκBα wt	+	+	Blue
pLexA-hnRNPA1, pACT-lysRS	+	−	White
pLexA-IκBα N-T, pACT-Ubch9	+	+	Blue

^aS. cerevisiae reporter strain L40 was cotransformed with the following combinations of plasmid: pLexA-hnRNPA1 and pVP16-IκBα wt (test plasmids); pLexA-hnRNPA1 and pACT-lysRS (negative control); pLexA-hnRNPA1 and pVP16.701 empty vector; pLexA-IκBα N-T and pACT-Ubch9 (positive control); and pV44ER.LexA empty vector and pVP16-IκBα wt. Transformed yeast cells were grown either on medium containing histidine (+ His) or on medium lacking histidine but containing 8 mM 3AT (− His + 3AT). Filter lift assays were used to determine if the yeast expressed β-Gal activity. In the presence of chromogenic substrate, colonies which express β-Gal are blue and colonies which do not express the enzyme are white. 

hnRNPA1 interacts directly with IκBα in vitro.

To determine that the interaction between hnRNPA1 and IκBα was direct and was not mediated by a bridging protein present in the human cell extracts or in the yeast, interactions were studied using bacterially produced recombinant proteins. GST-hnRNPA1 was allowed to interact with IκBα in the presence or absence of NF-κB p50 or p65. Bound proteins were analyzed by Western blotting with an IκBα antibody. GST-hnRNPA1 bound IκBα irrespective of whether it was bound to p50 or p65 (Fig. 3A). As a positive control, GST-p65 was shown to interact with IκBα under all conditions, while no interaction with hnRNPA1 was demonstrable with either GST, GST-IκBγ, or GST-NFIII (Fig. 3A). To determine if hnRNPA1 could interact with p65, GST and GST fusion proteins GST-IκBα and GST-hnRNPA1 were incubated with ³⁵S-labeled in vitro-translated p65. Analysis of the bound proteins indicated that GST and GST-hnRNPA1 did not interact with p65 (Fig. 3B), whereas GST-IκBα bound the ³⁵S-labeled p65. Thus hnRNPA1 interacts directly with IκBα and does not interact with p65.

FIG. 3.

hnRNPA1 interacts directly with IκBα in vitro. (A) Recombinant IκBα (present in all lanes) and NF-κB p50 and NF-κB p65 (as indicated) were incubated with either GST, GST-NFIII, GST-IκBγ, GST-p65, or GST-hnRNPA1 immobilized on glutathione agarose. Bound proteins were eluted and analyzed by Western blotting with the 10B monoclonal antibody to IκBα. (B) In vitro-translated ³⁵S-labeled p65 was incubated with GST-IκBα, GST-hnRNPA1 fusion proteins, and GST immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging.

hnRNPA1 binds to the C terminus of IκBα.

To identify the region of IκBα required for interaction with hnRNPA1, a series of deleted IκBα molecules were fused to GST (Fig. 4A) and tested for their ability to interact with ³⁵S-labeled in vitro-translated hnRNPA1. Fusion proteins which contained sequences from the C terminus of IκBα were capable of binding with wild-type affinity to hnRNP1, while GST fusions which lacked the C-terminal region were unable to interact with hnRNPA1. While the GST fusion containing IκBα residues 265 to 317 bound hnRNPA1, a fusion containing IκBα residues 275 to 317 was unable to bind hnRNPA1 (Fig. 4B). Neither GST-IκBγ, GST-NFIII, nor GST displayed any interaction with hnRNPA1 (Fig. 4B). To confirm the role of the IκBα C terminus in the interaction with hnRNPA1, a previously described series of LacZ molecules linked to either the IκBα C terminus, N terminus, or both the N and C termini were utilized (34). ³⁵S-labeled in vitro-translated forms of these molecules were tested for interaction with either GST or GST-hnRNPA1. While the construct containing only the IκBα N-terminal domain failed to interact with hnRNPA1, both constructs which contained the IκBα C-terminal region were bound by GST-hnRNPA1 (Fig. 4C). To delimit the C-terminal boundary of the IκBα region required for interaction with hnRNPA1, a previously described series of IκBα molecules (49) with deletions in various regions of the C terminus were used. In vitro-translated ³⁵S-labeled IκBα molecules were tested for their abilities to interact with either GST or GST-hnRNPA1. None of the C-terminally deleted IκBα molecules bound GST hnRNPA1 (Fig. 4D). Thus the IκBα region required for interaction with hnRNPA1 has boundaries of between residues 265 and 275 at the N terminus and residues 292 and 317 at the C terminus.

FIG. 4.

hnRNPA1 binds to the C terminus of IκBα. (A) Diagrammatic representation of IκBα and truncated versions of IκBα, with their abilities to bind to hnRNPA1 indicated. (B) In vitro-translated, [³⁵S]methionine-labeled hnRNPA1 was incubated with GST, GST-NFIII, GST-IκBγ, GST-IκBα wild type (WT), and GST-IκBα truncation mutants immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging. The location of ³⁵S-labeled hnRNPA1 is indicated (A1). (C) In vitro-translated ³⁵S-labeled β-Gal or fusions with either the N terminus of IκBα (N-T), the C terminus of IκBα (C-T), or both the N and C termini of IκBα (CT + NT) were incubated with either GST or GST-hnRNPA1 immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging. (D) In vitro-translated, ³⁵S-labeled IκBα wild type (WT) or truncation mutants were incubated with either GST or GST-hnRNPA1 immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging. The input in vitro-translated products are shown in the right-hand panel.

An hnRNPA1 RNA binding domain is required for interaction with IκBα.

To identify the domain in hnRNPA1 required for interaction with IκBα, a strategy analogous to that employed with IκBα was adopted. hnRNPA1 is a multidomain protein that contains two RNA binding domains in the N-terminal half of the protein (72), an RGG box, which also constitutes an RNA binding motif (33), in the central region, and the M9 nuclear import and export sequence in the C-terminal region (29) (Fig. 5A). A series of GST-hnRNPA1 fusions were constructed in which sequences from the N and C termini had been progressively deleted. Equal molar amounts of each bacterially expressed fusion protein were incubated with ³⁵S-labeled in vitro-translated IκBα, and bound proteins were collected on glutathione agarose. Removal of sequences between residues 207 and the C terminus (1 to 207) did not affect binding of IκBα, whereas an hnRNPA1 molecule containing only residues 1 to 196 was unable to bind IκBα (Fig. 5B and C). In addition, GST fusions containing hnRNPA1 sequences between residues 95 and 320 bound IκBα efficiently, whereas a fusion containing hnRNPA1 residues 105 to 320 was unable to bind IκBα (Fig. 5D). These data were confirmed using ³⁵S-labeled IκBα and GST-hnRNPA1 and truncated forms (Fig. 5E). Thus sequences in hnRNPA1 between residues 95 and 207, a region which encompasses a single RNA binding domain, are required for interaction with IκBα.

FIG. 5.

An hnRNPA1 RNA binding domain is required for interaction with IκBα. (A) Diagrammatic representation of the hnRNPA1 molecule with the two RNA binding domains (RBD), the RGG box, and the M9 nuclear transport signal indicated. Truncation mutants used in this study and their abilities to bind IκBα are indicated. (B and C) In vitro-translated, ³⁵S-labeled hnRNPA1 wild type (WT) or truncation mutants were incubated with either GST, GST-IκBα, or GST-IκBα_265–317 immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging. (D) In vitro-translated, ³⁵S-labeled hnRNPA1 wild type (WT) or truncation mutants were incubated with GST or GST-IκBα immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging. (E) In vitro-translated ³⁵S-labeled IκBα was incubated with GST, GST-hnRNPA1 wild type (WT), or GST-hnRNPA1 truncation mutants immobilized on glutathione agarose beads. Bound proteins were analyzed by SDS-PAGE and phosphorimaging.

hnRNPA1 enhances NF-κB-dependent transcriptional activation.

To determine the functional consequences of the interaction between IκBα and hnRNPA1, we obtained a mouse erythroleukemia cell line (CB3) which lacks endogenous hnRNPA1 (7). It was thus possible to introduce hnRNPA1 into these cells and evaluate its influence on NF-κB-dependent transcriptional activation. CB3 cells were electroporated with an NF-κB-dependent luciferase reporter and expression constructs for wild-type and mutant forms of hnRNPA1 in either the presence or absence of constructs expressing EBV LMP-1. EBV LMP-1 is a potent inducer of signal transduction pathways that lead to NF-κB activation (60), and expression levels of this protein were adjusted to ensure that the NF-κB response was not saturated. In the absence of hnRNPA1, EBV LMP-1 expression results in a 13-fold increase in NF-κB-dependent reporter activity. In the presence of hnRNPA1 and LMP-1, reporter activity was increased to 101-fold over that observed in the absence of hnRNPA1 and LMP-1 (Fig. 6B). hnRNP1 mutants which did not interact with IκBα (105–320 and 1–196 constructs) failed to substantially increase LMP-1-activated NF-κB reporter activity, whereas mutants which were capable of interacting with IκBα (95–320 and 1–207 constructs) increased LMP-1-activated NF-κB reporter activity above that for the pCDNA control (Fig. 6B). Western blotting indicated that the wild-type and mutant forms of hnRNPA1 were expressed at comparable levels and that proteins containing the M9 shuttling domain (40) (wild type and 95–320 and 105–320 constructs) were nuclear, whereas proteins lacking the M9 domain (1–196 and 1–207 constructs) were cytoplasmic (Fig. 6A). Neither LMP-1 nor hnRNPA1 had any influence on the activity of a control luciferase reporter lacking NF-κB binding sites (Fig. 6B) or on a LacZ reporter with an RSV promoter (data not shown) which was employed as an internal control. An AP1-dependent luciferase reporter which was activated fivefold by cotransfected LMP-1 was not further activated by expression of hnRNPA1 (Fig. 6). Thus a lack of hnRNPA1 expression in CB3 cells results in defective NF-κB-dependent transcriptional activation, and this can be rectified by expression of exogenous hnRNPA1.

FIG. 6.

hnRNPA1 enhances NF-κB-dependent transcriptional activation. (A and B) The NF-κB-dependent luciferase reporter 3 enh conA luc and the RSV-lacZ reporter were electroporated with pCDNA3 empty vector or pCDNA3 expression constructs containing either the wild-type (WT) hnRNPA1 cDNA (A1) or the indicated truncation mutants into CB3 cells which do not express hnRNPA1. (A) Expression levels of hnRNPA1 were determined by Western blotting. (B) To provide an NF-κB activation signal, cells were electroporated with an expression construct containing the cDNA for EBV LMP-1 or empty vector. Sixteen hours after electroporation, cells were lysed for determination of luciferase and LacZ activity. The activity of the Rous sarcoma virus LacZ reporter was used as an internal control, and the values indicated represented the ratio of luciferase activity to LacZ activity. Assays were performed in triplicate, and error bars represent 1 standard deviation. (C) The control experiments using ConA Luc and AP1 ConA Luc were performed as described above, in duplicate, and results are quoted in relative light units per milligram of protein.

hnRNPA1 enhances NF-κB-dependent transcriptional activation by potentiating IκBα degradation.

To determine the level at which hnRNPA1 expression potentiates NF-κB activation, plasmids expressing a tagged version of IκBα and hnRNPA1 were introduced into CB3 cells in the presence or absence of cotransfected DNA encoding EBV LMP-1. Western blotting indicated that LMP-1-induced degradation of IκBα is not apparent in the absence of hnRNPA1 but is efficient in the presence of hnRNPA1 (Fig. 7A). Analysis of various deletion forms of hnRNPA1 in this assay revealed that in addition to the wild-type form, versions containing amino acids 1 to 207 and 95 to 320 also allowed IκBα degradation, although this was less efficient with the 95–320 construct. Signal-induced degradation of IκBα was not observed in the presence of versions of hnRNPA1, the 105–320 and 1–196 constructs which did not interact with IκBα (Fig. 7A). Transfection efficiency was controlled by cotransfection of myc-tagged pyruvate kinase, which was detected by Western blotting (Fig. 7A). hnRNPA1 constructs were expressed at comparable levels (Fig. 6A). EBV LMP-1-induced degradation of IκBα in the presence of hnRNPA1 was dependent on phosphorylation of S32 and S36, since an S32A, S36A mutant of IκBα failed to undergo EBV LMP-1-induced degradation (Fig. 7B). To determine the sequences in IκBα that are required for degradation, C-terminally truncated forms of IκBα were tested for their ability to undergo EBV LMP-1-induced degradation in the presence of hnRNPA1. While an IκBα construct containing residues 1 to 303, which interacts with hnRNPA1, was efficiently degraded in the presence of EBV LMP-1 and hnRNPA1, a construct containing residues 1 to 292, which did not interact with hnRNPA1, was not degraded (Fig. 7C).

FIG. 7.

hnRNPA1 enhances IκBα processing in response to LMP-1 activation. (A, B, and C) IκBα wild type (WT) and the S32A S36A, 1–292, and 1–303 constructs were electroporated with pcDNA 3 empty vector or pcDNA 3 expression constructs containing either hnRNPA1 cDNA or the indicated truncation mutants into CB3 cells which do not express hnRNPA1. To provide an NF-κB activation signal, cells were electroporated with an expression construct containing the cDNA for EBV LMP-1 or empty vector. To control for the level of transfection between the different conditions, the cells were transfected with an expression construct containing the cDNA for pyruvate kinase with a myc tag. Sixteen hours posttransfection the cells were lysed and separated by SDS–10% PAGE. Following separation, cells were transferred to a polyvinylidene difluoride membrane and analyzed using the SV5 monoclonal antibody to SV5-tagged proteins and the myc monoclonal antibody to myc-tagged proteins.

DISCUSSION

The experimental data reported here demonstrate that the IκBα inhibitor of the transcription factor NF-κB interacts directly, both in vivo and in vitro, with hnRNPA1. Cells lacking hnRNPA1 are defective in NF-κB-dependent transcriptional activation, but the defect in these cells is complemented by ectopic expression of hnRNPA1. Cells lacking hnRNPA1 do not induce efficient degradation of IκBα in response to stimuli such as EBV LMP-1. However, an efficient signal-induced degradation of IκBα is fully restored by expression of hnRNPA1 or a derivative containing residues 1 to 207 and is partially restored with a derivative containing residues 95 to 320.

While IκBα and hnRNPA1 both shuttle between the nucleus and the cytoplasm (32, 39, 45, 50), modification of this activity by interactions between IκBα and hnRNPA1 is unlikely to explain the role of hnRNPA1 in NF-κB activation, since an hnRNPA1 molecule lacking the M9 domain, which fails to shuttle, is capable of interacting with IκBα and restoring the NF-κB response in hnRNPA1-deficient CB3 cells (Fig. 6 and 7). Thus hnRNPA1 molecules that are restricted to the cytoplasm can still complement the defect in NF-κB signaling in CB3 cells.

The region in hnRNPA1 which is required for interaction with IκBα is located between residues 95 and 207. Not only is this region required for interaction between hnRNPA1 and IκBα, but it is sufficient to complement the defect in NF-κB activation in CB3 cells. Amino acids 95 to 207 in hnRNPA1 constitute a single RNA recognition motif (RRM) linked to a fragment of the RGG box, which also represents an RNA binding motif (33). The proteins of an hnRNP complex are involved in diverse aspects of pre-mRNA metabolism. There is considerable evidence that suggests a role for some hnRNPs in the export of mRNA from the nucleus to the cytoplasm (39, 40, 45, 66). In addition to its presumed role in pre-mRNA packaging and transport, hnRNPA1 has other activities of biological importance. Both in vitro and in vivo studies demonstrate that hnRNPA1 has the potential to influence 5′ splice site selection in pre-mRNAs that contain multiple 5′ splice sites (12, 36, 73) and promotes the renaturation of complementary single-stranded nucleic acids (41, 46). While both RRMs are required for the alternative splicing activity of hnRNPA1, a single RRM is sufficient for the RNA binding and nucleic acid annealing properties of hnRNPA1 (37). Thus it is unlikely that the mRNA transport and alternative splicing activities of hnRNPA1 are involved in the ability of hnRNPA1 to influence NF-κB-dependent transcription.

On the basis of this information, the possible mechanisms by which hnRNPA1 can participate in NF-κB activation are limited. One possibility is that hnRNPA1 acts as a scaffold-like molecule, bringing IκBα into the optimal environment for signal-induced modification. However, preliminary investigations (data not shown) suggest that hnRNPA1 does not coprecipitate with the dimeric IκB kinase α or β molecules. Gel electrophoresis DNA binding assays (data not shown) also demonstrate that hnRNPA1 does not displace NF-κB from IκBα; thus NF-κB-dependent transcription is not potentiated in this manner. hnRNPA1 interacts with the carboxy-terminal region of IκBα. The carboxy terminus of IκBα contributes to protein destabilization upon cell activation by different stimuli. Moreover, removal of upstream C-terminal sequences, adjacent to the PEST domain, renders the protein highly refractory to signal-induced proteolysis (6, 10, 49, 59, 69). Although we do not know the exact mechanism by which hnRNPA1 influences NF-κB transcription, a potential model would involve hnRNPA1 enhancing IκBα degradation. In this respect, it is worth noting that the C-terminal region of IκBα, required for interaction with hnRNPA1, is also required for proteasome-mediated degradation of IκBα (34).