Individual RNA recognition motifs of TIA-1 and TIAR have different RNA binding specificities.

TIA-1 and TIAR are two closely related RNA recognition motif (RRM) proteins which possess three RRM-type RNA binding domains (RRMs 1, 2, and 3). Although both proteins have been implicated as effectors of apoptotic cell death, the specific functions of TIA-1 and TIAR are not known. We have performed in vitro selection/amplification from pools of random RNA sequences to identify RNAs to which TIA-1 and TIAR bind with high affinity. Both proteins selected RNAs containing one or several short stretches of uridylate residues suggesting that the two proteins have similar RNA binding specificities. Replacement of the uridylate stretch with an equal number of cytidine residues eliminates the protein-RNA interaction. Mutational analysis indicates that, for both TIA-1 and TIAR, it is the second RNA binding domain (RRM 2) which mediates the specific binding to uridylate-rich RNAs. Although RRM 2 is both necessary and sufficient for this interaction, the affinity for the selected RNA (as determined by filter binding assays) does increase when the second domain of TIAR is expressed together with the first and third domains (K = 2 × 10M) rather than alone (K = 5 × 10M). Although RRM 3 (of either TIA-1 or TIAR) does not interact with the uridylate-rich sequences selected by the full-length proteins, it is a bona fide RNA binding domain capable of affinity-precipitating a population of cellular RNAs ranging in size from 0.5 to 5 kilobases. In contrast, RRM 1 does not affinity-precipitate cellular RNA. The inability of RRM 1 to interact with RNA may be due to the presence of negatively charged amino acids within the RNP 1 octamer.

TIA-1 and TIAR are two closely related RNA recognition motif (RRM) proteins which possess three RRMtype RNA binding domains (RRMs 1, 2, and 3). Although both proteins have been implicated as effectors of apoptotic cell death, the specific functions of TIA-1 and TIAR are not known. We have performed in vitro selection/amplification from pools of random RNA sequences to identify RNAs to which TIA-1 and TIAR bind with high affinity. Both proteins selected RNAs containing one or several short stretches of uridylate residues suggesting that the two proteins have similar RNA binding specificities. Replacement of the uridylate stretch with an equal number of cytidine residues eliminates the protein-RNA interaction. Mutational analysis indicates that, for both TIA-1 and TIAR, it is the second RNA binding domain (RRM 2) which mediates the specific binding to uridylate-rich RNAs. Although RRM 2 is both necessary and sufficient for this interaction, the affinity for the selected RNA (as determined by filter binding assays) does increase when the second domain of TIAR is expressed together with the first and third domains (K d ‫؍‬ 2 ؋ 10 ؊8 M) rather than alone (K d ‫؍‬ 5 ؋ 10 ؊8 M). Although RRM 3 (of either TIA-1 or TIAR) does not interact with the uridylate-rich sequences selected by the full-length proteins, it is a bona fide RNA binding domain capable of affinity-precipitating a population of cellular RNAs ranging in size from 0.5 to 5 kilobases. In contrast, RRM 1 does not affinity-precipitate cellular RNA. The inability of RRM 1 to interact with RNA may be due to the presence of negatively charged amino acids within the RNP 1 octamer.
RNA-binding proteins are involved in a variety of fundamental cellular processes including RNA splicing, polyadenylation, RNA transport, and translation. Specific RNA sequences with which these proteins interact have been identified in some cases, but for the majority of RNA-binding proteins, the RNA targets are unknown. TIA-1 and TIAR are two closely related members of the RNA recognition motif (RRM) 1 family of RNA-binding proteins (1,2). The RRM (also known as the RNP motif, the RNP consensus sequence, the RNP-80, and the consensus sequence RNA-binding domain) consists of 80 -90 amino acids containing two stretches of 8 and 6 highly conserved residues called RNP 1 and RNP 2, respectively (3)(4)(5). TIA-1 and TIAR both possess three amino-terminal RRM domains and a glutamine-rich carboxyl terminus. The RRM domains of TIA-1 and TIAR are very similar with 79% amino acid identity between the first domains, 89% amino acid identity between the second domains, and 91% amino acid identity between the third domains. The carboxyl termini of the two proteins, in contrast, are only 51% identical in amino acid sequence (2).
Several observations suggest that TIA-1 and TIAR are involved in signaling apoptotic cell death. The introduction of purified TIA-1 or TIAR into the cytoplasm of thymocytes permeabilized with digitonin results in fragmentation of genomic DNA into nucleosome-sized oligomers (1,2). TIAR is translocated from the nucleus to the cytoplasm in response to exogenous triggers of apoptosis (6), and TIA-1 is phosphorylated by a serine/threonine kinase that is activated during Fas-mediated apoptosis (7).
Although these findings implicate TIA-1 and TIAR in the apoptotic process, the specific functions of these proteins have not been determined. The existence of Drosophila (8,9) and Caenorhabditis elegans (10) homologs, each of which shares 46% amino acid identity with human TIA-1 and TIAR, indicates that TIA-1 and TIAR are proteins which have been evolutionarily conserved. In order to understand the functions of these proteins, both in cells undergoing apoptosis and in healthy cells, it is necessary to identify RNAs with which they interact. As a first step toward identifying their RNA targets, we have determined the RNA binding specificities of both TIA-1 and TIAR by selection/amplification from pools of random RNA sequences (11). We have found that both of these proteins interact with RNAs containing short stretches of uridylates, and that for both proteins it is the second RNA-binding domain which mediates this sequence specificity.

EXPERIMENTAL PROCEDURES
In Vitro Selection from Pools of Random RNA Sequences-RNA selection was based on the SELEX method (11). The pool of RNA sequences was in vitro transcribed from an oligonucleotide library provided by Dr. Jack Keene (Duke University). The oligonucleotides consisted of 68 random bases flanked at the 5Ј end by the T7 RNA polymerase promoter sequence and at the 3Ј end by 27 bases of common sequence. The RNA transcribed from the library was incubated with Escherichia coli-derived rTIA-1 or rTIAR which had been immobilized on cyanogen bromide-activated Sepharose 4B (Pharmacia Biotech Inc.) using monoclonal antibodies reactive with TIA-1 (2G9) or TIAR (1H10). The RNA was added to the immobilized protein in 300 l of binding buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 20 mM KCl, 1 mM MgCl 2 , 1 mM EGTA, 1 mM dithiothreitol, 0.05% Nonidet P-40, 0.5 mg/ml tRNA, 0.125 mg/ml bovine serum albumin, 40 units/ml RNasin). After incubation for 20 min at room temperature, the beads were washed five times with 1 ml of NT2 buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 0.05% Nonidet P-40, 1 mM dithiothreitol, 0.5 M urea) The immunoprecipitated material was phenol-extracted, and the RNA was precipitated with 2.5 M NH 4 OAc and 3 volumes of ethanol. The selected RNA was reverse-transcribed with Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) using primers which could anneal to the 5Ј (T7 universal primer) and 3Ј (reverse universal primer) sequences flanking the randomized region. The cDNA products were amplified by 30 cycles of polymerase chain reaction (PCR) using Taq polymerase (Perkin Elmer) under the following conditions: 1 min at 94°C, 1 min at 50°C, and 2 min at 72°C. The PCR products were in vitro transcribed, and the selection process was repeated for a total of 5 rounds. After the final round of selection, the PCR products were cloned into the TA cloning vector (Invitrogen) and sequenced by the dideoxy chain termination method using Sequenase (U. S. Biochemical Corp. coli (BL21/DE3) and purified from bacterial sonicates using glutathione-Sepharose 4B (Pharmacia Biotech Inc.) according to the manufacturer's instructions. Protein purity was assessed by Coomassie staining, and quantification was by the bicinchoninic acid protein assay (Pierce).
Ultraviolet (UV)-cross-linking Assays Using Selected RNA Sequences-The RNA for UV-cross-linking assays was in vitro transcribed using SP6 RNA polymerase (Promega). The DNA templates for transcription were individual reverse transcription-PCR products from the final round of in vitro selection/amplification (denoted "selected sequences") or individual PCR products obtained by amplifying the oligonucleotide library prior to selection for TIA-1 or TIAR binding (denoted "random sequences"). Both the selected sequences and random sequences were cloned into the TA vector (Invitrogen) immediately following PCR. Transcripts were internally labeled by including [␣-32 P]GTP (DuPont NEN) in the in vitro transcription reactions according to the protocol provided by Promega. UV-cross-linking was performed by incubating recombinant protein (5 pmol) with 1 ϫ 10 5 cpm of 32 P-labeled RNA in 20 l total volume of binding buffer (125 mM NaCl, 25 mM KCl, 5 mM Hepes pH 7.6, 2 mM MgCl 2 , 3.8% glycerol) at 30°C for 15 min. The binding reactions were placed on ice and irradiated with UV light for 10 min using a Stratalinker 2400 (Stratagene). Following the UV irradiation, the reactions were treated with 1 g/l RNase A (Sigma) at 37°C for 20 min and analyzed by SDS-PAGE. The gels were dried and exposed to film (Kodak X-Omat AR) with an intensifying screen for 6 -12 h.
Truncated and truncated/mutated selected RNA sequences were in vitro transcribed as described above using DNA templates encoding either the first 50 nucleotides of a selected RNA sequence ("truncated selected sequence") or the first 50 nucleotides of the selected RNA sequence with the uridylate residues of the uridylate stretch replaced by cytidines ("truncated/mutated selected sequence"). The DNA templates were constructed by cloning annealed complementary oligonucleotides encoding the RNA sequences into the EcoRI and BamHI sites of the pSP65 vector.
Nitrocellulose Filter Binding Assays-Nitrocellulose filter binding assays were based on previously described methods (12,13). Increasing amounts of recombinant protein were incubated with in vitro transcribed 32 P-labeled RNA using the same conditions as described for the UV-cross-linking experiments except that the reaction contained bovine serum albumin (20 g/ml), the reaction volume was 30 l, and the incubation at 30°C was for 10 min. Using a dot blot apparatus (Bio-Rad) and gentle vacuum, the samples were applied to nitrocellulose which had been pretreated with binding buffer. Each well was washed with 600 l of ice cold binding buffer. The amount of RNA retained on the filter was determined by scintillation counting, corrected for RNA binding to nitrocellulose which occurred in the absence of protein, and normalized to the percent of RNA bound at saturation. The fraction of each protein active in binding (determined by an RNA excess assay at saturating protein concentration) ranged from 60 -86%. The amount of bound RNA (normalized) was plotted against protein concentration (corrected for the active fraction). Binding curves generated from three independent experiments were fitted using Sigma Plot 4.16 (Jandel Scientific). The apparent dissociation constant (K d ) is the concentration of protein at which 50% of RNA bound at saturation is retained on the filter. 32 P-Labeling of Cellular RNA-Jurkat cells were cultured for 12-14 h in phosphate-free RPMI (Life Technologies, Inc.), 5% fetal calf serum, 4 mM glutamine, and 10 Ci/ml 32 PO 4 (New England Nuclear) at a density of 2-4 ϫ 10 5 cells/ml. The cells were lysed, and total RNA was extracted as described (14). Purified recombinant TIA-1 or TIAR GST fusion protein (or RRM truncation GST fusion protein) was incubated with the 32 P-labeled RNA in 100 l of binding buffer for 15 min at room temperature. 30 l of 50% (v/v) glutathione-Sepharose 4B (Pharmacia Biotech Inc.) was added, and incubation was continued for 30 min with resuspension of beads every 3 min. After washing the beads four times with binding buffer plus 0.2% Triton X-100, the RNA was extracted with phenol/chloroform and precipitated with 2.5 M NH 4 OAc and 3 volumes of ethanol. The RNA was separated by formaldehyde agarose gel electrophoresis and transferred to nitrocellulose as described (14).

RESULTS
The TIA-1 and TIAR Proteins Select Uridine-rich RNAs-Recombinant TIA-1 and TIAR were each immobilized on Protein A-Sepharose with a monoclonal antibody (2G9 for TIA-1 and 1H10 for TIAR) that reacts with a carboxyl-terminal epitope of the protein. The immobilized protein was incubated with a large molar excess of in vitro transcribed RNA containing a region of 68 bases of random sequence. RNAs which bound to the protein were purified and amplified by reverse transcription-polymerase chain reaction. The DNA templates encoding the selected RNAs were transcribed using T7 RNA polymerase, and the process was repeated for a total of 5 cycles. After the final cycle, the DNA products were cloned into the TA cloning vector for sequencing. Twenty clones selected by TIA-1 and twenty clones selected by TIAR were sequenced (Fig. 1). Eighteen of the twenty TIA-1-selected sequences, and nineteen of the twenty TIAR-selected sequences contain one or, in most cases, several short stretches of uridylate residues ranging in length from three to eleven nucleotides.
TIA-1 and TIAR Interact Specifically by UV-cross-linking with the Selected RNAs-To confirm that TIA-1 and TIAR bind to the RNAs selected by the random RNA selection procedure, recombinant protein was incubated with in vitro transcribed, 32 P-labeled RNA and exposed to ultraviolet (UV) light which will cross-link proteins and nucleic acids that are in direct contact. Following RNase A treatment and SDS-polyacrylamide gel electrophoresis, RNA cross-linked to protein (thereby protected from RNase) was visualized by autoradiography. Fig. 2A shows the cross-linking resulting from incubation of recombinant TIA-1 (left panel) or TIAR (right panel) with four different TIA-1-or TIAR-selected RNAs (lanes 1-4 and 9 -12), with individual RNAs picked randomly from the entire library (lanes 5-7 and 13-15), or with a mixture of 20 RNAs picked randomly from the library (lanes 8 and 16). Both TIA-1 and TIAR cross-link more efficiently to the selected RNAs than to the randomly picked RNAs. As shown in Fig. 2B, the selected RNAs (lanes 2-5 and 11-14), but not the random RNAs (lanes 6 -9 and 15-18), effectively compete with the selected sequence for binding to TIA-1 and TIAR. Addition of yeast tRNA (0.2 mg/ml) as a nonspecific competitor had no effect on the interaction between TIA-1 or TIAR and the selected RNAs (data not shown). Replacement of the uridylate stretch of selected sequence 1-1 (five consecutive uridine residues beginning at nucleotide 3) with an equal number of cytidine residues eliminated the ability of this RNA to compete for binding to TIA-1 (Fig. 2C, lanes 8 -10) confirming the importance of the uridylate stretch for the interaction. Similar results were obtained when this RNA was used as a competitor in TIAR-RNA binding reactions (data not shown).
The Second RNA Binding Domain of Both TIA-1 and TIAR Binds Specifically to the Selected RNA Sequences-TIA-1 and TIAR each contain three RRM-type RNA-binding domains. In order to investigate the contributions of the individual domains to the overall RNA binding activity of the protein, constructs which express various combinations of the RRM domains fused to GST were made by PCR cloning (see Fig. 3). As shown in Fig. 4, RRM 2, when expressed by itself or together with RRM 1 and/or RRM 3, is able to interact by UV-cross-linking with the selected RNA sequence (i.e. the RNA sequence selected by the fulllength protein). The first and third RNA-binding domains of FIG. 2. UV-cross-linking of TIA-1 and TIAR to the selected RNA sequences. A, purified GST-TIA-1 or GST-TIAR was incubated with 32 P-labeled RNA and UV-irradiated as described under "Experimental Procedures." Following RNase A treatment, cross-linked protein was visualized by autoradiography of 10% SDS-polyacrylamide gels. Four TIA-1-selected RNAs (lanes 1-4), four TIARselected RNAs (lanes 9 -12), three randomly picked RNAs (lanes 5-7 and 13-15), or a mixture of 20 randomly picked RNAs (lanes 8 and 16) were used. The numbers above lanes 1-4 and 9 -12 correspond to the RNA sequences depicted in Fig. 1. The sequences of the randomly picked RNAs (A, B, C, and Mix) are not depicted. B, purified GST-TIA-1 or GST-TIAR was incubated with 32 P-labeled selected RNA 1-1 (lanes 1-9) or R-2 (lanes 10 -18) in the presence of increasing amounts of unlabeled competitor RNA (selected RNA 1-1, selected RNA R-2, or the mixture of 20 RNAs picked randomly from the library) as indicated. UV-crosslinking and subsequent analysis was performed as in A. The unlabeled competitor RNA was present in 10-, 20-, 40-, or 60fold molar excess compared with labeled RNA. C, purified GST-TIA-1 was incubated with 32 P-labeled selected RNA 1-1 in the presence of increasing amounts of unlabeled competitor RNA (selected RNA 1-1, truncated selected RNA 1-1, or mutated/truncated selected RNA 1-1) as indicated. The unlabeled competitor RNA was present in 10-, 20-, or 60-fold molar excess compared with labeled RNA.
both TIA-1 and TIAR, on the other hand, when expressed by themselves, do not cross-link to the selected RNA sequence, suggesting that the second domain is required for this interac-tion. GST alone does not cross-link to the RNA (data not shown). The finding that the second RNA binding domains of TIA-1 and TIAR interact with the selected RNA sequence but not with a mixture of randomly picked RNAs (Fig. 4) suggests that the second RNA-binding domains of TIA-1 and TIAR bind to RNA with the same sequence specificity as their full-length counterparts.
To assess the strength of the interaction between full-length TIAR or TIAR RRM truncation mutants and the selected RNA sequence, nitrocellulose filter binding assays were performed. Fig. 5 shows the normalized binding curves for the full-length GST-TIAR protein and several of the RRM truncation mutants. Although neither RRM 1 nor RRM 3, when expressed in isolation, bind to the selected RNA sequence in this assay even when present in micromolar concentrations (data not shown), the affinity of RRM 2 for the selected RNA sequence is higher when RRM 2 is expressed together with RRM 1 and RRM 3 (K d ϭ 2 ϫ 10 Ϫ8 M) rather than alone (K d ϭ 5 ϫ 10 Ϫ8 M). Consistent with the UV-cross-linking studies, the TIAR constructs did not bind to the random mixture of in vitro transcribed RNAs (data not shown). The filter binding data, together with the UVcross-linking results, suggest that the second RNA-binding domains of TIA-1 and TIAR contain the determinants of sequence-specific binding. However, at least for TIAR, additional structural domains of the protein appear to contribute to the affinity of the protein-RNA interaction.
Unlike RRMs 2 and 3, RRM 1 Does Not Interact in Vitro with  Fig. 3 was UV-cross-linked to [ 32 P]RNA as described under "Experimental Procedures." The [ 32 P]RNA was either selected RNA (S) or a mixture of 20 RNAs picked randomly from the library (R). The selected RNA used in the crosslinking studies with the TIA-1 mutants was sequence 1-1; the selected RNA used with the TIAR mutants was sequence R-2 (see Fig. 1). The migration of molecular mass markers is shown at the right.

FIG. 5. Kinetics of binding of TIAR RRM 2-containing mutants
to the selected RNA. Nitrocellulose filter binding assays were performed to measure the affinity of TIAR and mutants containing RRM 2 either alone or in combination with RRMs 1 and/or 3 for selected RNA R-3 (see Fig. 1). Each point represents the average of three independent binding reactions. The data for each curve are normalized to the saturation point for the RNA and corrected for the fraction of protein active in RNA binding as described under "Experimental Procedures." The legend to the data points is shown in the upper left corner of the plot. The K d values shown below the plot are equal to the protein concentrations at which 50% of RNA is bound.

FIG. 3. Schematic diagram of TIA-1 and TIAR mutants.
Fusion proteins between GST and different combinations of the RNA binding domains (RRMs) were constructed for TIA-1 and TIAR as described under "Experimental Procedures." The structural domains of the wildtype proteins include the three RRMs, the RNP 1 and RNP 2 motifs, and the glutamine-rich carboxyl region. The sequence of each RNP 1 and RNP 2 motif is depicted. In each case, the upper sequence is that of TIA-1, the lower sequence is that of TIAR. Cellular RNAs-In an effort to identify cellular RNAs with which TIA-1 and TIAR interact, recombinant TIA-1 and TIAR GST fusion proteins were incubated with total RNA which had been extracted from 32 PO 4 -labeled Jurkat cells. Following affinity precipitation with glutathione-Sepharose, associated RNA was isolated by phenol/chloroform extraction, separated on a formaldehyde-agarose gel, and transferred to nitrocellulose. For both TIA-1 and TIAR, the associated RNA appears by agarose gel electrophoresis to consist of a population of RNAs ranging in size from 0.5 to 5 kilobases (see Fig. 6). RRM 2 and RRM 3 when expressed individually, also interact with a population of RNAs. Surprisingly, however, for both TIA-1 and TIAR, the first RNA-binding domain does not affinity-precipitate RNA. This apparent lack of RNA binding activity is not due to RNase contamination of the RRM 1 protein preparations since in mixing experiments neither TIA-1 nor TIAR RRM 1 inhibits the RNA binding by the wild-type protein or the other truncation mutants (data not shown). Electrophoresis of the affinity-precipitates through 5% polyacrylamide gels (to allow visualization of snRNAs and tRNA) also failed to show RNA associated with RRM 1 (data not shown). DISCUSSION We have employed in vitro selection/amplification from pools of random RNA sequences to identify those RNA sequences for which the TIA-1 and TIAR proteins have the highest affinity. After five rounds of selection, 90% of the RNA sequences selected by both proteins contain short stretches of uridylate residues ranging in length from 3 to 11 nucleotides. The sequences do not appear to have similarity with regard to the distance between the uridylate stretches, the regions flanking the uridylate stretches, or the predicted secondary structure. UV-cross-linking experiments confirm that TIA-1 and TIAR bind the selected sequences, that the interaction is specific, and that the uridylate stretch is required for the interaction (see Fig. 2). The affinity of the interaction (K d of 8 nM as determined by filter binding assays) is similar to that observed for other RNA-binding proteins (such as the hnRNP A1 protein) and their selected RNA sequences (15). Given the 80 -90% amino acid identity between the RNA binding domains of TIA-1 and TIAR, it is not surprising that that the two proteins selected very similar RNA sequences.
Although TIA-1 and TIAR each contain three RRM-type RNA-binding domains, our results suggest that the second domain of each of these proteins mediates the sequence-specific binding. Mutational analysis demonstrates that RRM 2 is both necessary and sufficient for binding to the RNAs selected by the full-length protein. Studies of other RRM-type RNA-binding proteins (hnRNP C, U1A, and U1 70K), have suggested that their RNA binding specificities are dependent on amino acids immediately carboxyl-terminal to the RRM domain (13,16,17). Our results are not inconsistent with this, since the 15 amino acids linking the second and third RRMs are present in each mutant that contains RRM 2. A single RRM of a multi-RRM protein having the same sequence specificity as the full-length protein is not unique to the TIA-1 and TIAR proteins. The U1 snRNP A protein has two RRMs, only one of which appears to be required for binding to U1 RNA (18,19). Similarly, the third RNA binding domain of Hel-N1, like the full-length protein, binds to the 3Ј untranslated region of c-myc mRNA (20). An analysis of the binding specificity of the hnRNP A1 protein, on the other hand, shows that its two RRMs have sequence specificities which differ both from each other and from the fulllength protein (15). Thus, within the RRM family of RNAbinding proteins, there does not appear to be uniformity with regard to the contribution of individual RRMs to overall binding specificity.
Although the third RNA-binding domains of TIA-1 and TIAR do not bind to the uridylate-rich RNAs selected by the fulllength protein, they clearly do have RNA binding activity as illustrated by their ability to affinity-precipitate cellular RNA (see Fig. 6). The ability of RRM 3 to bind to cellular RNA but not to the mixture of random RNAs in vitro transcribed from the oligonucleotide library (see Fig. 4) may be due to differences in the sensitivities of the assays. It is quite possible that RRM 3 interacts with specific RNA sequences (that differ from the uridylate stretches) but that these sequences were not identified using the selection/amplification method because they are of low affinity or because, as has been demonstrated for U2 snRNP-BЉ binding to U2 RNA, the interactions require additional proteins (21)(22)(23).
In contrast, we have been unable to demonstrate any RNA binding activity for the first RNA-binding domain of either TIA-1 or TIAR. The RNP 1 octamer of the first RNA-binding domains of TIA-1, TIAR, and homologous proteins found in Drosophila (8,9) and C. elegans (10) are all somewhat atypical in that they contain aspartic acid in the first position (see Fig.  3 for human TIA-1 and TIAR RNP 1 sequences). The RNP 1 octamers of the second and third RNA-binding domains of the TIA proteins each have a lysine in the first position and thereby resemble the RNP 1 sequences of the majority of RRM-type proteins which contain positively charged amino acids (either arginine or lysine) in that position (5,24). It has been proposed that the four-stranded antiparallel ␤-sheet within the RNAbinding domain of RRM-type proteins provides a general surface for RNA binding, and that the highly conserved basic and aromatic amino acids of RNP 1 and 2 (which form two of the ␤-strands) are involved in direct interactions with RNA (5,13,24,25). The first residue of the RNP 1 octamer of the U1A RRM 1 has been shown to be critical for U1A binding to U1 RNA; replacement of the arginine with the uncharged glutamine completely abolished binding to U1 RNA (26,27). One other RRM-type protein with a negatively charged amino acid in the first position of RNP 1 is ASF/SF2. The ASF/SF2 RRM 2 RNP 1 octamer begins with an aspartic acid (although there appears to be debate as to whether the eight amino acid sequence beginning with aspartic acid, or an eight amino acid sequence three residues downstream and beginning with a glycine, is the actual RNP 1 sequence) (24,28). Additional studies are necessary to determine whether the negatively charged aspartic acid FIG. 6. Interaction between TIA-1/R RRMs and cellular RNA. Total RNA extracted from 32 PO 4 -labeled Jurkat cells was incubated with equal molar quantities of GST-TIA-1, GST-TIAR or truncation mutants of each of these proteins, and affinity-precipitated with glutathione-Sepharose. The precipitated RNA was purified, separated on 1.2% formaldehyde agarose gels, and transferred to nitrocellulose. The identity of each protein is indicated at the top of each lane. The lanes labeled "Total RNA" show a fraction (approximately 1/20) of the RNA that was used in each binding experiment. The migration of DNA size markers is shown at the right.
of TIA-1/R RNP 1 affects the RNA binding activity of RRM 1. It is possible that RRM 1 may preferentially interact with RNA that has been modified in ways which decrease its net negativity (e.g. by capping).
Uridylate stretches, frequently found in regulatory regions of RNAs such as the 3Ј splice site of introns (29,30) and 5Ј and 3Ј untranslated regions (20,31), appear to be common targets of RRM-type RNA-binding proteins. The hnRNP C and Hel-N1 proteins both selected sequences containing short uridine stretches when similar selection/amplification approaches were used with these proteins (13,20), and the Sex-lethal (Sxl) protein of Drosophila has been shown to interact with poly(U) sequences in transformer (tra) and Sxl pre-mRNAs (32)(33)(34)(35)(36). Because cells contain a multitude of RNA-binding proteins, some of which may compete with TIA-1 and TIAR for binding to RNA, it is important to characterize the RNA binding activity of TIA-1 and TIAR not only in vitro but also in cells. The similarity between the uridylate stretch consensus sequences identified for the TIA-1/R and hnRNP C proteins, together with the size distribution of the cellular RNAs interacting in vitro with TIA-1/R, suggested to us that the TIA proteins may associate in cells with pre-mRNA or mRNA. Although we have not been able to demonstrate an in vivo interaction between endogenous TIA-1 or TIAR and RNA, both of these proteins can be isolated by oligo(dT) affinity chromatography of lysates prepared from UV-irradiated TIA-1 or TIAR COS cell transfectants 2 indicating that, at least when overexpressed, the proteins interact in cells with polyadenylated RNA (37,38).
The identification of uridylate stretches as high affinity binding sites for TIA-1 and TIAR, together with the demonstration of in vivo interactions between these proteins and polyadenylated RNA, suggest that the TIA proteins might interact with regulatory regions of RNA transcripts to modulate gene expression. It is not known how the RNA binding activities of these proteins relate to their effector functions in apoptotic cell death. An intriguing possibility is that the TIA proteins regulate the processing or translation of RNA transcripts encoding mediators of apoptosis. Alternative splicing of ICE (39), Ich-1 (40), and bcl-x (41) can produce mRNAs encoding both positive and negative regulators of apoptosis. Studies aimed at determining whether the TIA proteins bind to uridine-rich regions of such RNAs are under way.