Heterogeneous Nuclear Ribonucleoprotein A3, a Novel RNA Trafficking Response Element-binding Protein*

The cis-acting response element, A2RE, which is sufficient for cytoplasmic mRNA trafficking in oligodendrocytes, binds a small group of rat brain proteins. Predominant among these is heterogeneous nuclear ribonucleoprotein (hnRNP) A2, a trans-acting factor for cytoplasmic trafficking of RNAs bearing A2RE-like sequences. We have now identified the other A2RE-binding proteins as hnRNP A1/A1, hnRNP B1, and four isoforms of hnRNP A3. The rat and human hnRNP A3 cDNAs have been sequenced, revealing the existence of alternatively spliced mRNAs. In Western blotting, 38-, 39-, 41-, and 41.5-kDa components were all recognized by antibodies against a peptide in the glycine-rich region of hnRNP A3, but only the 41and 41.5-kDa bands bound antibodies to a 15-residue N-terminal peptide encoded by an alternatively spliced part of exon 1. The identities of these four proteins were verified by Edman sequencing and mass spectral analysis of tryptic fragments generated from electrophoretically separated bands. Sequence-specific binding of bacterially expressed hnRNP A3 to A2RE has been demonstrated by biosensor and UV cross-linking electrophoretic mobility shift assays. Mutational analysis and confocal microscopy data support the hypothesis that the hnRNP A3 isoforms have a role in cytoplasmic trafficking of RNA.

Establishment of asymmetry in cells requires selective localization of proteins. This may be accomplished by directed protein transport, a well established pathway for plasma membrane and secreted proteins, or by trafficking and subsequent localization of mRNA. Localization of RNA has been intensively studied in Drosophila and Xenopus oocytes (for reviews see Refs. [1][2][3][4][5][6] and more recently in mammalian somatic cells (7)(8)(9)(10)(11)(12).
In 1982, Colman et al. (13) discovered that myelin basic protein (MBP) 1 mRNA is concentrated in the myelin membrane fraction isolated from brain by density gradient centrifugation. Subsequent experiments demonstrated that MBP mRNA is translated close to myelin and the protein rapidly incorporated into the nascent membrane (14 -16) and lead to a model in which MBP mRNA is recruited into RNA transport granules in the perikaryon and then transported, by indirect attachment to the microtubule-bound motor protein kinesin, to the myelin compartment at the cell periphery (10,(17)(18)(19)(20)(21). The granules are localized in the myelin compartment, and the RNA cargo is translated, with the MBP being incorporated into the myelin membrane. Deletion studies led to the conclusion that a small element in the 3Ј-untranslated region of the MBP mRNA, the RNA transport sequence (RTS), is sufficient and necessary for this cytoplasmic RNA transport in oligodendrocytes (17).
Cytoplasmic trafficking of RNA encoding ␤-actin is also dependent on inclusion in transport granules that are attached to the cytoskeleton. In fibroblasts ␤-actin mRNA transport is microfilament-dependent (9,22), whereas microtubules are implicated in transport of this mRNA in neurons (23,24).
trans-Acting factors have been isolated in pull-down experiments with RTS-labeled magnetic particles. The predominant RTS-binding protein from a number of rat tissues is heterogeneous nuclear ribonucleoprotein (hnRNP) A2 (25), a constituent of nuclear "core particles" that bind to nascent hnRNA and participate in various aspects of RNA processing. Recent mutational analyses have shown a close correlation between oligoribonucleotide binding to hnRNP A2 and the ability to support cytoplasmic RNA trafficking in oligodendrocytes (26). Antisense oligonucleotide experiments have added support to the proposition that hnRNP A2 is involved in this RTS-dependent trafficking, and the RTS has consequently been renamed the hnRNP A2 response element (A2RE). hnRNP A2 also enhances cap-dependent translation of mRNA containing the A2RE (27).
Other, less abundant, rat brain proteins that are also reproducibly isolated on immobilized A2RE have not previously been identified. Direct Edman sequencing of proteins extracted from SDS/polyacrylamide gel slices was unsuccessful, suggesting that all are N-terminally blocked. By using Edman microsequencing and mass spectrometric fingerprinting of tryptic peptides, Western blotting with antibodies raised against peptides unique to individual members of the hnRNP A/B family, and cloning and sequencing of rat and human cDNAs, we have now identified four of the A2RE-binding proteins as isoforms of hnRNP A3 and a fifth component as hnRNP A1. hnRNP A3 has been expressed in Escherichia coli, purified, and shown in biosensor and UV cross-linking electrophoretic mobility shift experiments to recognize A2RE directly, suggesting that hnRNPA3 is not bound to RNA through interaction with hnRNP A2. We have also found that hnRNPs A2 and A3 and microinjected A2RE-containing RNA are localized in cytoplasmic transport granules in cultured neurons, suggesting that hnRNP A3, like hnRNP A2, participates in the trafficking of A2RE-containing RNA. These are the first reported experiments on hnRNP A3, which has recently been identified as a component of 40 S hnRNP complexes (28) but has otherwise been described only at the cDNA level.

EXPERIMENTAL PROCEDURES
Primers-Primers designed against the 5Ј and 3Ј ends of frame shift-corrected human fetal brain ribonucleoprotein (FBRNP; NCBI GI:1710627) DNA coding sequence (nucleotides 31-1140) were purchased from Invitrogen (Mt. Waverly, Australia). NcoI (forward primer) and SacI (reverse primer) restriction sites were added to the 5Ј ends of the primers to facilitate insertion of the PCR product into an expression vector. The primers were forward (hA3F) 5Ј-GTACCATGGAGGTA-AAACCGCCG-3Ј and reverse (hA3R) 5Ј-AGAGAGCTCAGAACCTTCT-GCTACCATATCCAC-3Ј (coding sequences are underlined, and restriction sites are in italics).
RNA Extraction and RT-PCR-21-day-old Wistar rat brain was snap frozen. Human cerebellar tissue was from a 65-year-old female who died of heart failure. Total RNA was isolated from 100 mg of tissue using TRIzol reagent (Invitrogen). The dried RNA-containing pellet was resuspended in 20 l of diethyl pyrocarbonate-treated water. RNA purity and concentration were determined spectrophotometrically and by electrophoresis on 2% agarose gels. RNA-dependent DNA synthesis was performed using avian myeloblastosis virus reverse transcriptase (AMV-RT; Promega, Annandale, Australia, for rat RNA) or SuperScript II (Invitrogen, for human RNA). 2 g of RNA was used in each RT reaction and primed with oligo(dT) 15 . Subsequent PCR reactions were performed using the proofreading enzymes platinum Taq high fidelity DNA polymerase (Invitrogen, for rat) and ELongase (Invitrogen, for human) with 40 pmol of each primer in each 50-l reaction.
3Ј-RACE was performed using an internal gene-specific primer for reverse transcription. Terminal transferase (Invitrogen) was used for homopolymeric 3Ј C-tailing of the first DNA strand. Nested PCR was performed with RACE anchor and adapter primers (5Ј AP and AUAP; Invitrogen). A modified oligo(dT) (3Ј AP; Invitrogen) and internal primers were used for the RT step in 3Ј-RACE.
DNA Sequencing-Purified RT-PCR products were ligated into pGemT-Easy (Promega) and electroporated into E. coli DH5␣ cells. Cells bearing plasmids containing an appropriately sized insert were grown overnight, the plasmids were isolated, and the inserts were amplified using M13 universal primers and hA3F and hA3R for sequencing in both forward and reverse directions with partial overlap. DNA from three individual clones was sequenced. The sequences were analyzed using CLUSTAL W version 1.8 (29) to generate consensus sequence and protein alignment, and the programs at ExPASy (/au. expasy.org/) were used for conceptual translation of the DNA sequences (30).
Protein Sequencing and Mass Fingerprinting-Attempts at direct Edman sequencing of the A2RE-binding proteins were unsuccessful, suggesting that they were N-terminally blocked. Peptides were therefore generated by excision of bands from SDS/polyacrylamide gels and in-gel digestion with 1.5 times the gel volume of 0.02 mg/ml trypsin (Promega) in 40 mM ammonium bicarbonate, 10% (v/v) acetonitrile, pH 8.1 (31). The resultant peptides were purified by reverse-phase HPLC on a microbore C18 column and running a gradient of 10 -40% acetonitrile over 60 min in 1% trifluoroacetic acid at 30 l/min and sequenced on a Procise cLC sequencer (Applied Biosystems, Foster City, CA). Mass spectral analysis was performed on an ABI QSTAR Pulsar i spectrometer (Applied Biosystems) with an electrospray ion source interfaced to a microbore HPLC. The peptides were separated on a C3 reverse-phase column, the output of which was split between the HPLC detector and the mass spectrometer. The proteins were identified by comparison of the observed tryptic peptide masses with those predicted from the gene sequences. For each of the hnRNP A3 isoforms, fragments with masses within Ϯ 0.2 Da of the theoretical value covering at least 33% of the primary sequence were found.
Cloning, Expression, and Isolation of hnRNP A3-Full-length rat hnRNP A3 transcript was amplified and ligated into pGemT. The sequence was verified by sequencing before ligation in-frame into pET30aϩ (Novagen, Madison, WI), which had been mutated to generate a second thrombin cleavage site in place of an enterokinase site between the hexahistidine tag and hnRNP A3. This vector was used to transform E. coli BL21(DE3) cells that were grown to an A 600 of 0.6 before induction with isopropyl-1-thio-␤-D-galactopyranoside. The cell lysates were centrifuged, and the supernatant was passed through a metal ion affinity column (Talon IMAC resin, CLONTECH, Palo Alto, CA) equilibrated with 50 mM NaH 2 PO 4 , 700 mM NaCl, 5 mM imidazole, 1 mM phenylmethylsulfonyl fluoride, pH 7, or 20 mM Tris, 100 mM NaCl, and 6 M guanidinium hydrochloride, pH 7. The hexahistidine fusion protein was released from the resin by increasing the column buffer imidazole concentration to 200 mM and purified by reverse-phase HPLC. Two proteins with the tag removed and an additional Ser-Gly at the N terminus were generated by cleavage with thrombin; they were identified by electrospray/time-of-flight mass spectrometry performed on an Applied Biosystems QSTAR Pulsar i.
Biosensor Measurements-An IAsys resonant mirror biosensor (Affinity Sensors, Cambridge, UK) was used to determine the equilibrium affinities for the interactions of hnRNP A3 with A2RE, with NS1 (CAAGCACCGAACCCGCAACUG) being used as a control to distinguish specific binding from nonspecific binding. Recombinant hnRNP A3 was covalently attached to the carboxymethyldextran-coated sensing surface of a biosensor cuvette using a standard procedure (32). After washing the biosensor cuvette with phosphate-buffered saline (PBS) containing 0.1% Tween 20, hnRNP A3 was added to the cuvette, and the equilibrium response was recorded (ϳ20 min). Unbound protein was removed by washing with PBS containing 0.05% Tween 20. Binding of oligonucleotides was monitored at 0.4-s intervals at 25°C until equilibrium had been attained (ϳ15 min). The cuvette was washed with 5 mM NaOH to remove the oligonucleotide and with PBS containing 0.05% Tween 20 to restore the base line between experiments.
UV Cross-linking Electrophoretic Mobility Shift Assays-Recombinant proteins were obtained as described above. Oligoribonucleotide 32 P end-labeled with T4 polynucleotide kinase (New England Biolabs, Beverly, MA) and protein were mixed in binding buffer (10 mM Tris-HCl, pH 7.5, containing 1 mM EDTA, 4% (w/v) glycerol, 0.1% (w/v) Triton X-100, and 1 mM dithiothreitol) and incubated for 20 -30 min on ice. The reaction mixtures were then irradiated with 250 mJ of 254-nm light in a Bio-Rad GS Genelinker UV chamber. The samples were run on 15 ϫ 15-cm SDS/12% or 15% polyacrylamide gels.
Affinity Isolation on Magnetic Particles-A2RE-binding rat brain proteins were isolated using affinity isolation on superparamagnetic particles bearing immobilized A2RE or NS1, as described previously (25). Proteins bound to the particles were eluted by heating for 10 min at 65°C in 0.1% SDS, 1 mM dithiothreitol, or 30% (v/v) acetonitrile in 0.1% trifluoroacetic acid and analyzed on SDS/polyacrylamide gels or used for mass spectrometry.
Cell Culture-Hippocampus dissected from embryonic day 18 Wistar rats was digested with trypsin in Hanks' balanced salt solution, washed twice in Hanks' balanced salt solution, placed in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) fetal calf serum, and mechanically dissociated by trituration. The cells were plated at 600 cells/mm 2 on poly(L-lysine)-coated glass-bottomed microwells (MatTek, Ashland, MA). After 2 h of incubation to allow cell attachment, medium containing Neurobasal, N2 supplements (1:100 dilution), B27 supplements (1:50 dilution), and 5% (v/v) gentamicin was added, and the cells were incubated at 37°C in 95% air, 5% CO 2 . The medium was replaced every other day. All reagents were from Invitrogen except where noted otherwise.
Confocal Laser Scanning Microscopy-Cultured rat brain neurons were washed in PBS then fixed for 20 min in 3.7% paraformaldehyde (Sigma) in PBS. After further washing with PBS, the cells were permeabilized by incubation for 2 min in 0.1% Nonidet P-40 (Sigma) diluted in the same buffer, washed, and then incubated in 5% goat serum in PBS for 10 min. For visualization of hnRNPs A2 and A3, the cells were incubated for 30 min at room temperature in the primary antibody (rabbit antibody against hnRNP A3 and mouse antibody to hnRNP A2), washed extensively, blocked with 5% goat serum in PBS for 10 min, incubated in secondary antibody (fluorescein isothiocyanate-conjugated goat antimouse IgG and Alexa-598-conjugated goat anti-rabbit; Jackson ImmunoResearch Laboratories, West Grove, PA) for 30 min, and washed with PBS. Finally, 70% glycerol containing an anti-fading agent (Dabco) was added to the cells before they were imaged on either Zeiss FIG. 1. A2RE-binding proteins. Rat brain proteins were added, in the presence of heparin, to A2RE immobilized on magnetic particles. A, the bound proteins were eluted with 1% SDS, separated on an SDS/ polyacrylamide gel, and stained with Coomassie Blue. The 36-kDa hnRNP A2 (arrow) was identified earlier (25), and hnRNP A1 is barely visible below this band. The other four bands, which are the focus of this study, are indicated by arrowheads with their apparent molecular masses. B, Western blots of the A2RE-binding proteins. The proteins separated on an SDS/12% polyacrylamide gel were electrophoretically transferred onto nitrocellulose, and the strips were incubated in rabbit primary antibodies raised against peptides of hnRNPs A1, A2/B1, B1, and A3 (N-terminal peptide). The proteins were visualized using an anti-rabbit IgG alkaline phosphatase-conjugated secondary antibody. The positions of molecular mass standards are marked on the left. The two hnRNP A3 bands recognized by the hnRNP A3 antibody, the 41and 41.5-kDa proteins in A, are not resolved on this short gel.
FIG. 2. hnRNP A3 amino acid sequences. The human hnRNP A3 sequence deduced from the cDNA sequence determined in the present work and those of human hnRNPs A1 and A2. The rat hnRNP A3 differs from the human protein only by insertion in the latter of one additional Gly residue in the Gly-rich region (arrowhead above the hnRNP A3 sequence). Residues in hnRNPs A1 and A2 are identified only where they differ from the hnRNP A3 sequence. The residues conserved between hnRNPs A1 and A3 but not A2 and A3 are boxed, and those conserved between hnRNPs A2 and A3 but not A1 and A3 are shaded. Gaps in the sequences are indicated with dashes. The minimal M9 nuclear import/export sequence of A1 (50) and the equivalent segments of hnRNPs A2 and A3 are underlined in the C-terminal Gly-rich region. Exon 1 residues deleted in the shorter isoform of hnRNP A3 are in bold, and the sequence within this segment used to generate the N-terminal hnRNP A3 antibody is underlined. The arrow marks the point at which recombinant hnRNP A3 is cleaved by thrombin.
FIG. 3. Antibodies to hnRNP A3 recognize the A2RE-binding proteins. A2RE-binding rat brain proteins from a magnetic particle pull-down experiment were separated on an SDS/polyacrylamide gel and electroblotted onto polyvinylidene difluoride membrane. Strips of the electroblot were incubated with antibodies to hnRNP A2 (A2), to the N-terminal alternatively spliced segment (A3(N)), and to the peptide from the Gly-rich region (A3(C)). The proteins were visualized using an anti-rabbit IgG alkaline phosphatase-conjugated secondary antibody and development with nitroblue tetrazolium chloride/5-bromo-4-chloro-3Ј-indolyl phosphate, p-toluidine salt one-step solution (Pierce). The positions of marker proteins are shown at left, and the hnRNP A3 bands are indicated by arrowheads.
RNA Microinjection and Visualization-After 7-14 days in culture, differentiated hippocampal neurons were microinjected with RNA labeled with Alexa-488-UTP (Molecular Probes, Eugene, OR) and containing or lacking the A2RE11 (GCCAAGGAGCC) sequence inserted in the 3Ј-untranslated region between the green fluorescent protein open reading frame and the segment encoding the poly(A) (26). The cells were injected using a Compic Inject (Cellbiology Trading, Hamburg, Germany) micromanipulator attached to a Zeiss Axiophot inverted microscope. After injection, the cells were incubated at 37°C for 30 min to allow transport to occur. To visualize neurites, the cells were incubated for 30 min in mouse anti-MAP2 (Sigma), washed, and incubated for 30 min in Texas Red-conjugated goat anti-mouse IgG (Sigma). (25,26) showed that immobilized A2RE binds at least six polypeptides from rat brain in an RNA sequence-selective manner (Fig. 1A). The most abundant of these proteins was shown by Edman sequencing of tryptic peptides to be hnRNP A2, but the other polypeptides were less abundant and appeared to be N-terminally blocked. Edman protein sequencing of peptides generated from the 38-and 39-kDa bands had indicated that they were closely related to hnRNP A2 but were not alternatively spliced forms of this 36-kDa protein. The peptide sequences matched those deduced from the cDNA sequences of human hnRNP A3 (identified as FBRNP, as corrected in the Swiss Protein Database, accession number P51991) but were insufficient in number and length to unequivocally identify the protein.

Western Blotting of A2RE-binding Proteins-Previous experiments
We have now used immunoblotting to identify the other rat brain A2RE-binding proteins that were eluted from magnetic particles, separated on SDS/polyacrylamide gels, and transferred to nitrocellulose for immunodetection with polyclonal antibodies raised against peptide antigens. The peptide antigens were selected using published amino acid sequences for the human hnRNP A1, A2, and B1 proteins, and for A3 a peptide was deduced from our DNA sequence (see below). The proteins detected (Fig. 1B) were hnRNPs A1, A1 B (above A1 on the A1 track, ϳ40 kDa), A2, B1 (just above A2 on the A2 track, and separately on the B1 track; the topmost band in the A2 track is unidentified), and A3. As shown below, the antibody to hnRNP A3 detects two proteins, but they are not resolved on the short gel used in this experiment. The predominant A2REbinding proteins are thus the previously identified hnRNP A2 and hnRNP A3. As judged by the staining on polyacrylamide gels, hnRNPs A1, A1 B , and B1 are minor components.
DNA Sequencing-As a basis for further studies we determined the cDNA sequences of rat and human brain hnRNP A3, initially using for both RT-PCR with primers based on the 5Ј and 3Ј coding sequences of FBRNP to amplify hnRNP A3-like mRNA. The high degree of amino acid sequence conservation between humans and rodents observed for other hnRNPs suggested that these primers would be satisfactory for the latter, a proposition subsequently verified by direct sequencing of the rat DNA in these regions (see below).
Two distinct sequences were amplified from multiple clones containing cDNA reverse-transcribed from human brain RNA. One matched and extended the partial murine clone (EMBL accession number Y16641), a truncated expressed sequence tag described as encoding a novel gene product, mBx-3 (28) (Fig. 2). The second differed from that deduced for the human FBRNP (33) only by a substitution of the dipeptide Met 93 -Arg 94 in our sequence for Ile-Gly at the C-terminal end of the second RNA recognition motif of FBRNP. These two cDNAs had 96.5% identity at the nucleotide level and 94.2% (357 of 379 residues) at the amino acid level, with the nucleotide differences between these human sequences spread throughout the DNA, suggesting that the two proteins arise from distinct genes, but subsequent searching of the human genome data bases suggested that the second of these cDNAs (corresponding to FBRNP) resulted from transcription of a pseudogene.
The first human cDNA sequence differed from the single rat sequence obtained from multiple clones in 55 nucleotide substitutions spread throughout the sequence and by the presence of a TGG insert in the latter, but these differences result only in insertion of a single additional glycine residue near the C terminus of the rat protein, suggesting that these proteins are orthologous. No murine cDNA corresponding to FBRNP was detected by multiple approaches including 3Ј-and 5Ј-RACE and direct RT-PCR using total rat brain RNA with an oligo(dT) primer, again suggesting that it represents a pseudogene. However, 5Ј-RACE did result in the identification of a truncated form of rat DNA, which had a similar but longer 5Ј-untranslated region and a 66-nucleotide deletion near the 5Ј end of the coding region (Fig. 2). This corresponds to a mass change of 2566 Da in the protein. The 3Ј-and 5Ј-RACE experiments also confirmed that the human and rat sequences are identical in the regions corresponding to the primers used in the initial amplification of the rat DNA. The sequence of the full and truncated cDNAs are closely related to those reported for A3like proteins from Xenopus, A3a and A3b (34).
Protein Features Deduced from DNA Sequence-hnRNP A1 possesses a transportin-binding nuclear localization signal (M9) that is thought to be important for shuttling between the nucleus and cytoplasm (35)(36)(37). The minimal 15-residue sequence is underlined in Fig. 2 (38). A similar sequence is present in hnRNP A3; this suggests that hnRNP A3, like hnRNPs A1 and A2 (39), may shuttle in and out of the nucleus.
Comparison of the amino acid sequences of the human hnRNPs A1, A2, and A3 reveals a close relationship between them. Overall, the hnRNP A3 amino acid sequence matches hnRNP A1 more closely than it matches hnRNP A2/B1. Within the tandem RNA recognition motif region there are many, mostly conservative, substitutions in hnRNP A2 where the other two proteins are identical and relatively few residues where hnRNP A2 but not hnRNP A1 matches the hnRNP A3 sequence (Fig. 2). In the C-terminal glycine-rich region, this trend is reversed, suggesting that the hnRNP A3 gene may have arisen from the recombination of the 5Ј RNA recognition motif-encoding segment of a former hnRNP A1 gene with the 3Ј Gly-rich segment of an earlier hnRNP A2 gene.
Protein Identification-Antibodies were raised against two peptide sequences deduced from the FBRNP and hnRNP A3 gene sequences. The first peptide (VKPPPGRPQPDSGRR), which was used in initial experiments (Fig. 1B), is in the N-terminal alternatively spliced region. The second (GYDGY- EESGK PGAHV TVK GGNFG GR a At the underlined positions the low signal levels for the 41.5-kDa band resulted in ambiguity in the identification of the residue. Ser and Met were observed in the first position and Gly and Lys in the second. These ambiguities do not interfere with the exclusion of hnRNPs A1 and A2. The residues in hnRNPs A1 and A2 that differ from those observed are in bold type. b EMBL accession numbers G296650 (hnRNP A1) and G500638 (hnRNP A2). NEGGNF) is in a segment of the glycine-rich region predicted to be common to all splice variants of A3 but not fully conserved in hnRNPs A1 and A2/B1. In Western blots the first antibody recognized just the 41-and 41.5-kDa bands, suggesting that only these proteins contain the 22-residue segment encoded within exon 1 (Fig. 3). By contrast, the second antibody associated with all four 38-, 39-, 41-, and 41.5-kDa bands, identifying them as hnRNP A3-like proteins and not isoforms of hnRNPs A1 or A2.
This identification was confirmed by Edman sequencing and electrospray mass spectrometry. Bands cut from stained SDS/ polyacrylamide gels were pulverized and digested with trypsin, and the reverse-phase HPLC-purified peptides were subjected to Edman degradation. The resultant amino acid sequences, presented in Table I, confirmed the identification of three of these proteins as hnRNP A3. Each of the tryptic digests of the excised bands was also subjected to liquid chromatographymass spectrometry (LC-MS), and the resultant peptide masses were compared with those predicted from the putative protein sequences. For each band, fragments spanning 33% or more of the translated cDNA sequence matched the predicted masses of tryptic fragments of hnRNP A3 (Table II). Where the putative protein sequences differed, the observed masses corresponded to the expected sequence and excluded the alternative FBRNP protein, with the exception of one peptide, which gave a weak signal at a mass corresponding to a peptide predicted for this protein. The two protein sequences differ substantially in the predicted tryptic fragments, and the absence of FBRNP peptides in the mass spectral fingerprinting therefore indicates that this protein is not expressed at levels comparable with hnRNP A3. Thus, all four bands appear to be alternatively spliced forms of hnRNP A3, with two of the forms lacking the majority of exon 1. Alternatively spliced forms of both hnRNPs A1 and A2 are expressed, with the inclusion or exclusion of exon 7bis, resulting in hnRNPs A1 and A1 B (40) and the alternative splicing of exons 2 and 9 giving rise to A2, B1, B0a, and B0b (41,42). By analogy we anticipated that the four forms of A3 would arise from inclusion or exclusion of two alternatively spliced exons.
To test this hypothesis, A2RE-binding proteins were isolated using magnetic particle pull-down experiments and subjected to LC-MS. Previous reverse-phase HPLC experiments 2 had shown that the hnRNP A3 isoforms co-elute ahead of hnRNP A2. The LC-MS chromatogram gave a similar profile, but only two proteins were detected in the hnRNP A3 peak, with average masses of 39,863 Ϯ 4 and 37,297 Ϯ 4 Da (Fig. 4); these values are both 211 Da greater than the masses calculated from the protein primary structures predicted from the DNA sequences, indicating that the proteins have undergone posttranslational modification. The difference in mass between these two isoforms, 2566 Da, corresponds to the mass of the segment encoded by the N-terminal insertion (2567 Da), reinforcing the evidence for expression of these two forms and identifying them as the nominally 38-and 41-kDa isoforms ( Fig. 1 and below). The 39-and 41.5-kDa isoforms were not observed in the mass spectra, and the relationship between them and the other two isoforms is not known. Although not detected in the mass fingerprinting experiments, hnRNPs B1 and A1 B , which migrate close to the hnRNP A3 bands on SDS/polyacrylamide gel electrophoresis, did bind the A2RE (Fig. 1B), as did the faster migrating hnRNP A1. hnRNP A3 Interacts with the A2RE Independently of hnRNP A2-Pull-down experiments with A2RE immobilized on magnetic particles have consistently yielded hnRNP A2 and the less abundant hnRNP A3s. Given previous demonstrations that purified A2 binds the A2RE, it was possible that hnRNP A3 was isolated because it interacted with hnRNP A2 rather than through a direct interaction with the oligoribonucleotide, as suggested earlier (25). We therefore used the expressed hnRNP A3 in UV cross-linking electrophoretic mobility shift assays and biosensor experiments. Full-length rat hnRNP A3 (Fig. 2) was expressed in E. coli as a hexahistidine-tagged protein and purified (Fig. 5A). Thrombin cleavage yielded hnRNP A3 with two additional N-terminal residues, Gly-Ser, arising from the thrombin cleavage site (calculated mass, 39,796 Da; measured average mass, 39,793 Ϯ 4 Da). Cleavage of the fusion protein tag was accompanied by cleavage at a 2 T. Munro and R. Smith, unpublished results.  4. Mass spectrometry identifies two hnRNP A3 isoforms. Rat brain A2RE-binding proteins were isolated using A2RE immobilized on magnetic particles. The proteins were eluted from the particles in 30% acetonitrile in 0.1% trifluoroacetic acid, concentrated by vacuum centrifugation, centrifuged to remove any particulate material, and analyzed by C18 reverse-phase liquid chromatography-orthogonal quadrupole/time-of-flight mass spectrometry. In the total ion current chromatogram the peaks labeled A, B, and C yielded mass spectra of hnRNP A2 (36,076), the hnRNP A3 isoform missing part of exon 1 (37,297), and the hnRNP A3 isoform possessing the full exon 1 (39,863), respectively. The other peaks in the chromatogram are derived from smaller proteins that are bound nonspecifically to the magnetic particles. Expansion of the mass spectra for peaks A-C revealed small amounts of sodium adduct but no other components. The tracks show proteins bound to nonspecific oligonucleotide (NS) and detected with the Gly-rich region hnRNP A3 peptide antibody (A3(C)), rat brain A2RE-binding proteins (A2RE) detected with A3(C) and N-terminal peptide (A3(N)) antibodies, respectively, and partly digested recombinant hnRNP A3 detected with A3(N), showing the full-length protein (arrow) migrating with the rat brain 41-kDa isoform, behind the C-terminally truncated hnRNP A3 (arrowhead). polyacrylamide gel (Fig. 5B), suggesting that this rat protein corresponds in sequence to the expressed protein. This conclusion is consistent with the Western blots (Fig. 3) showing that the 41-and 41.5-kDa proteins contain the N-terminal insertion shown in Fig. 2.
Biosensor measurements in which A2RE or nonspecific oligoribonucleotide (NS1) was added to purified, expressed Histagged hnRNP A3 immobilized on the cuvette showed that the oligonucleotide binding to this protein closely parallels the binding to hnRNP A2 (43). A saturating concentration of NS1 (30 M) or A2RE (4 M) was added to the biosensor, resulting in a response with A2RE double that with NS1. After attainment of equilibrium (Fig. 6A, arrow), sufficient A2RE was added to each biosensor to double its concentration. This addition led to little change in the response upon addition to A2RE but to a doubling of the biosensor response upon addition of A2RE to the cuvette previously containing only NS1 (Fig. 6A, top panel). By contrast, doubling of the concentration of NS1 in the cuvette previously equilibrated with this oligoribonucleotide did not further increase the biosensor response (Fig. 6A, bottom panel). The parallel between the oligoribonucleotide binding to hnRNPs A3 and A2 indicated that the former also possesses one site that binds RNA sequence specifically and a second site that manifests no strong sequence specificity. Additional support for this proposition was obtained from studies of the effect of 10 g/liter heparin. This polyanion halved the binding of A2RE to hnRNP A3 (Fig. 6B, top panel) and eliminated NS1 binding (Fig. 6B, bottom panel). The affinity of recombinant hnRNP A3 for both NS1 and A2RE, as reflected in the dissociation constants derived from the binding curves (Fig. 6C), is lower than for human recombinant hnRNP A2. The K d for the specific site is 276 Ϯ 35 nM compared with 44 Ϯ 7 nM for hnRNP A2, and the corresponding values for the nonspecific site are 3.0 Ϯ 0.6 M (A2RE) and 4.8 Ϯ 0.6 M (NS1) for hnRNP A3 compared with 267 Ϯ 41 nM (A2RE) and 246 Ϯ 29 nM (dNS1) for hnRNP A2. UV cross-linking electrophoretic mobility shift experiments showed binding of radiolabeled A2RE to hnRNPs A2 and A3 and lower binding of NS1, in accord with the biosensor experiments (Fig. 7A). Competition assays confirmed the specificity of the RNA-protein interaction (Fig. 7B).
Earlier experiments had shown a correlation between hnRNP A2 binding to A2RE and corresponding oligonucleotides with point mutations and the ability of these sequences to support transport of RNAs (26). Together with antisense oligonucleotide data, these observations suggested a role for hnRNP A2 in cytoplasmic mRNA trafficking. Because the binding to A2RE of hnRNP A3 correlated closely with that of A2 in these earlier experiments, it appears that the former proteins may also play some part in RNA trafficking, but confirmation of this proposal requires a more direct experimental demonstration.
hnRNP A3 Distribution in Rat Tissues Mirrors hnRNP A2-The distribution of hnRNP A3 has not previously been investigated. Equal amounts of protein extracted from rat tissues were separated on multiple lanes of an SDS/polyacrylamide gel and electroblotted onto nitrocellulose for detection with antipeptide antibodies to hnRNPs A2 and the N-terminal peptide of A3. hnRNP A3 was found in several tissues, most prominently in brain, lung, and testis, and its levels in these tissues paralleled those of hnRNP A2 (Fig. 8). Little or no hnRNP A2 or A3 was detected in muscle, kidney, heart, or liver. Lower molecular mass forms of both proteins were reproducibly observed in extracts of spleen.
hnRNP A3 Is Co-localized with A2RE-containing RNA in the Cytoplasm of Neurons-If hnRNP A3 is involved in cytoplasmic RNA trafficking, it might be expected to be localized in cytoplasmic granules and to be co-localized with A2RE-containing RNA. Immunofluorescence microscopy showed hnRNP A3 to be present in the nucleus (not shown) and in cytoplasmic granules in the neurites of cultured hippocampal neurons (Fig. 9A), with hnRNPs A2 and A3 being localized to different populations of granules (Fig. 9). Microinjected fluorescently labeled A2REcontaining RNA was also co-localized with hnRNP A3 in a subset of cytoplasmic granules (Fig. 10). These results, together with earlier data (26) showing that mutations in the A2RE that lower binding to hnRNP A2 and A3 interfere with RNA trafficking, 3 suggest that hnRNPs A2 and A3 both play a role in RNA trafficking in neurites. DISCUSSION The four most abundant of the A2RE-binding rat brain proteins migrating behind hnRNP A2 on SDS/polyacrylamide gels have been identified by mass fingerprinting and Edman sequencing of tryptic peptides as hnRNP A3 isoforms (Tables I  and II). These peptides had sequences consistent with our hnRNP A3 cDNA sequences and excluded the possibility that any of the four proteins were splice variants of hnRNPs A1 or A2. This conclusion was also consistent with the results of Western blotting using two antibodies raised against peptides from hnRNP A3 (Fig. 3).
In the course of identifying these proteins, we completed the  9. hnRNP A3 is present in cytoplasmic granules. A, confocal laser scanning microscopy image of the neurites of a cultured hippocampal neuron using a mouse antibody to hnRNP A2 and a rabbit antibody to an N-terminal peptide unique to hnRNP A3. Both hnRNP A2 (green, arrows) and hnRNP A3 (red, arrowheads) were detected in granules, which had the appearance and distribution of transport granules, in the neurites. Scale bar, 5 m. B, statistical analysis of the fluorescence of individual granules showed that the majority of granules in the neurites contained either hnRNP A2 or hnRNP A3. A small number of granules were yellow, indicating the presence of both proteins. sequences of human and rat hnRNP A3 cDNAs. The hnRNP A3 amino acid sequence is highly conserved between humans and rats, as are those of hnRNPs A1 and A2. Initially two human cDNAs were amplified; both appeared to encode full-length hnRNP A/B-like proteins, one corresponding in protein sequence with the previously described FBRNP expressed sequence tag, and the other corresponding with the reported hnRNP A3 partial sequence (28). However, the FBRNP sequence appears to arise from a processed pseudogene; although it is transcribed and has appropriately located start and stop codons and polyadenylation signal, it corresponds to a gene on human chromosome 10 that possesses a single exon, in contrast to the 10 -12 exons of other human and mouse hnRNP A/B genes (41,44). Although an intronless paralog has been discovered for hnRNP E (45), the potential protein product of FBRNP DNA, which has a predicted mass close to one of the hnRNP A3 isoforms, was not detected in mass spectrometric fingerprinting of peptides derived from any of the four hnRNP A3 bands, indicating either that it is not translated in amounts comparable with the other hnRNP A3 isoforms or that the resultant protein does not bind the A2RE. A search of the DNA data bases revealed several other hnRNP A3 pseudogenes on different human chromosomes.
Both hnRNPs A2 and A3 are expressed as four isoforms. The hnRNP A2 isoforms arise from exclusion or inclusion of exons 2 (36 nucleotides) and 9 (120 nucleotides), generating B0a, B0b (ϩ exon 2), A2 (ϩ exon 9), and B1 (ϩ exons 2 and 9) (44, 46). The two higher molecular mass forms isoforms (41 and 41.5 kDa) of hnRNP A3, but not the two lower molecular mass forms (38 and 39 kDa) contain an N-terminal 22-amino acid insertion (Fig. 2), which was discovered using 5Ј-RACE; only the two higher molecular mass proteins bound antibodies raised against a peptide within this N-terminal insertion. Within these two doublets the apparent mass difference is 1-1.5 kDa, corresponding to 10 -15 amino acid residues, but RT-PCR did not reveal any mRNAs varying by this size and in attempts to determine the masses of all four isoforms by LC-MS only two proteins with masses in the appropriate range were detected. Their masses and the co-migration on SDS/polyacrylamide gels of the recombinant hnRNP A3 with the 41-kDa isoform (Fig. 4) suggest that the two observed masses, 37,297 and 39,863 Da, are those of the nominally 38-and 41-kDa isoforms. Several potential reasons for the nonappearance of the 39-and 41.5-kDa isoforms in mass spectrometry experiments have been eliminated; every chromatographic peak from the LC-MS runs was analyzed, but none contained proteins in the 37-45-kDa mass range, other than the two mentioned above. All four isoforms were present in the sample used for mass spectrometry, and all usually elute with similar retention times on reverse-phase HPLC, as shown by SDS/polyacrylamide gel electrophoresis.
We had previously discovered, from mutational analysis and antisense oligonucleotide experiments, that hnRNP A2 is a trans-acting factor for A2RE-mediated, cytoplasmic RNA trafficking in oligodendrocytes (26) and neurons. 3 In these experiments it was noted that the hnRNP A3 isoforms were also FIG. 10. Microinjected A2RE RNA and hnRNP A3 are co-localized in neuronal neurites. A, hippocampal neurons were microinjected with fluorescently labeled A2RE RNA. The subcellular distributions of the injected RNA (left panels) and hnRNP A3 (right panels) were visualized and analyzed by dual channel confocal microscopy, using rabbit primary antibody and Alexa 598-labeled secondary antibody to locate hnRNP A3 after fixing the cells. In each image, the arrows indicate granules that contain both A2RE RNA and hnRNP A3, and the arrowheads indicate granules with only hnRNP A3 labeling. Granules positive for hnRNP A3 but negative for injected RNA may transport endogenous A2RE-containing RNA. Scale bars, 5 m. B, analysis of hnRNP A3 and A2RE-containing RNA distributions in individual granules, showing a linear correlation between the levels of the two proteins and a population of granules that contain hnRNP A3 but low levels of exogenous RNA. RNA lacking the A2RE is not transported into the processes of oligodendrocytes (26) or neurons. 3 isolated from rat brain protein extracts in pull-down experiments with immobilized A2RE and that their binding to mutated forms of the A2RE paralleled that of hnRNP A2. This left open the possibility that hnRNP A3 bound directly to A2RE or indirectly through association with hnRNP A2. The biosensor and gel mobility shift data presented here support the former interpretation. The biosensor responses with purified recombinant hnRNP A3 (Fig. 6) closely paralleled those recorded for hnRNP A2 (43), indicating that hnRNP A3, like hnRNP A2, possesses two RNA-binding sites; one of them is sequencespecific, binding to A2RE, and the other binds with little discrimination between sequences. Although the RNA recognition motifs of hnRNP A3 are closer in sequence to hnRNP A1 than to hnRNP A2, hnRNP A3 mimics hnRNP A2 more closely than hnRNP A1 in its binding to A2RE, although with dissociation constants for A2RE binding that are severalfold higher than for hnRNP A2.
Although hnRNP A3 was not described originally as a component of the core particles identified in HeLa cell nuclei, it is present in multiple isoforms in these cells (47). We have shown that hnRNP A3 is abundant in several tissues, paralleling the tissue distribution of hnRNP A2. hnRNP A3, like hnRNP A2, is mostly localized in the nuclei of neurons and oligodendrocytes (not shown) and probably has a similar localization in other cell types.
Association of hnRNP A3 with A2RE in vivo is suggested by two of our observations. First, in neurons this protein is localized to granules in neurites that are similar in size and number to those shown previously to participate in trafficking of A2RE-containing RNA. 3 Interestingly, most granules in the neurites were positive for either hnRNP A2 or hnRNP A3, but not both. It has been shown that inclusion in transport granules requires a cooperative interaction between hnRNP A2 and RNA and that there are multiple copies of A2RE-containing RNA and probably multiple copies of hnRNP A2 in each granule. 4 Our observation that few cytoplasmic granules contain both hnRNPs A2 and A3 thus indicates that these two proteins do not interact cooperatively with each other in recruiting RNA to the transport granules. The second observation that implicates hnRNP A3 in the cytoplasmic trafficking of RNA is the co-localization of microinjected A2RE RNA with hnRNP A3 in granules that are distributed along the neurites.
The parallels in A2RE binding, tissue, and subcellular distribution of hnRNPs A2 and A3 beg the question of whether these proteins fulfill the same or similar roles in vivo in A2RE-dependent RNA trafficking. hnRNP A2 has been shown to be involved in cytoplasmic trafficking (25,26) and in the regulation of translation (27), and hnRNP A3 could also be involved in these aspects of RNA metabolism or in others such as nuclear export, RNA tethering at its destination, or mRNA stability. It will be of particular interest to explore the differences in role between the isoforms of each protein. The observation that hnRNP B1 expression is selectively upregulated in oncogenically transformed cells (48,49) is an indication that regulation of splicing of RNA encoding hnRNPs and hence of protein isoform expression may play an important role in cell biology.
In summary, we have shown that the predominant rat brain A2RE-binding proteins are hnRNP A2 and four isoforms of hnRNP A3. Bacterially expressed hnRNP A3 has been found in biosensor and gel mobility shift assays to bind A2RE, showing that this protein can bind A2RE directly and does not necessarily associate with A2RE indirectly by binding to an RNA-hnRNP A2 complex. A role for hnRNP A3 in cytoplasmic RNA trafficking, which may parallel that of hnRNP A2, is suggested by its co-localization with hnRNP A2 in tissues and its subcellular co-localization with A2RE RNA in neuronal transport granules. The sequestration of these hnRNPs into two separate populations of granules in the neuronal neurites raises interesting questions about the mechanism by which they are recruited to the transport granules.