|Home | About | Journals | Submit | Contact Us | Français|
The RNA-binding protein TIA-1 (T-cell intracellular antigen 1) functions as a posttranscriptional regulator of gene expression and aggregates to form stress granules following cellular damage. TIA-1 was previously shown to bind mRNAs encoding tumor necrosis factor alpha (TNF-α) and cyclooxygenase 2 (COX-2), but TIA-1 target mRNAs have not been systematically identified. Here, immunoprecipitation (IP) of TIA-1-RNA complexes, followed by microarray-based identification and computational analysis of bound transcripts, was used to elucidate a common motif present among TIA-1 target mRNAs. The predicted TIA-1 motif was a U-rich, 30- to 37-nucleotide (nt)-long bipartite element forming loops of variable size and a bent stem. The TIA-1 motif was found in the TNF-α and COX-2 mRNAs and in 3,019 additional UniGene transcripts (~3% of the UniGene database), localizing preferentially to the 3′ untranslated region. The interactions between TIA-1 and target transcripts were validated by IP of endogenous mRNAs, followed by reverse transcription and PCR-mediated detection, and by pulldown of biotinylated RNAs, followed by Western blotting. Further studies using RNA interference revealed that TIA-1 repressed the translation of bound mRNAs. In summary, we report a signature motif present in mRNAs that associate with TIA-1 and provide support to the notion that TIA-1 represses the translation of target transcripts.
Posttranscriptional mechanisms controlling pre-mRNA splicing and maturation, as well as mRNA transport, turnover, and translation, critically influence gene expression programs in mammalian cells. Central to the posttranscriptional regulatory events is the interaction of RNAs with RNA-binding proteins (RBPs) that influence their splicing, localization, stability, and association with the translation machinery (13, 16, 23, 30, 41). Many ribonucleoprotein (RNP) complexes that govern mRNA stability and translation in response to various stimuli (e.g., developmental, stress-inducing, immune, and proliferative) are comprised of transcripts that bear uridine- or adenine/uridine-rich elements (collectively termed AREs) and the proteins that bind the AREs (ARE-RBPs) (5, 42). ARE-bearing mRNAs have received a great deal of attention, since many of them encode proteins that regulate the cell division cycle, apoptosis, proliferation, immune response, oncogenesis, and inflammation (9). Likewise, many ARE-RBPs have been described that modulate the stability of target mRNAs, their translation, or sometimes both processes: AU-binding factor 1 (AUF1), tristetraprolin (TTP), K homology splicing-regulatory protein (KSRP), butyrate response factor 1 (BRF1), the Hu proteins (HuR, HuB, HuC, and HuD), T-cell-restricted intracellular antigen 1 (TIA-1), and the TIA-1-related protein TIAR (4, 6, 7, 22, 31, 34, 36, 43).
TIA-1 has been reported to participate in the regulation of alternative pre-mRNA splicing of bound mRNAs (18, 19). However, TIA-1 has been best characterized as a suppressor of translation, as shown for the target ARE-bearing mRNAs encoding tumor necrosis factor alpha (TNF-α) and cyclooxygenase 2 (COX-2) (15, 35). Following stimulation with bacterial lipopolysaccharide, macrophages derived from either wild-type or TIA-1−/− mice expressed the same levels of TNF-α mRNA, but TIA-1−/− cells expressed much more TNF-α protein than cells expressing TIA-1. In TIA-1−/− macrophages, the levels of TNF-α mRNA found in polysomes were significantly higher, lending further support to the notion that TIA-1 functions as a translational silencer (35). Similarly, the steady-state levels of COX-2 mRNA were the same in TIA-1-expressing and -deficient fibroblasts, but cells lacking TIA-1 had significantly higher levels of COX-2 mRNA in polysomes and expressed elevated levels of COX-2 protein (15).
The mechanisms whereby TIA-1 represses translation have been investigated most extensively in cells responding to environmental stress agents. Stress-triggered translational inhibition is characterized by the activation of one or more protein kinases (PKR, PERK, GCN2, and HRI) that phosphorylate the α subunit of eukaryotic initiation factor 2 (eIF-2α), a constituent of the ternary complex (eIF-2-GTP-) that loads initiator onto the small ribosomal subunit to initiate protein translation (14, 27). Phosphorylated eIF-2α inhibits translation by reducing the availability of the active ternary complex; under these conditions, TIA-1 has been proposed to interact with the translational machinery on the 5′ region of the mRNA and to promote the assembly of noncanonical, translationally incompetent initiation complexes (3). In situations of stress, when many transcripts are simultaneously subject to such translational silencing, the self-aggregating properties of TIA-1 promote the formation of cytoplasmic foci known as stress granules (SGs), which are generally believed to represent sites of translational inhibition (26). Nonetheless, the underlying translational control mechanisms mediated by TIA-1 are likely to be similar in stressed and unstressed cells (1, 2). In the case of mRNAs bearing 3′ untranslated region (3′UTR) AREs that form complexes with TIA-1 (TNF-α and COX-2), the likelihood will be greater that translationally silent preinitiation complexes assemble on the 5′ region of the transcript, and therefore ARE-bearing mRNAs would be preferentially subject to translational repression (3).
Given that TIA-1 is implicated in critical cellular events, including the response to stress agents, apoptotic stimuli, and inflammatory factors, we thought it would be highly valuable to systematically identify the collection of TIA-1 target mRNAs. Here, we describe efforts undertaken to elucidate such TIA-1-associated transcripts in human colorectal cancer cells using en masse methodologies. The analysis was carried out by immunoprecipitating TIA-1-RNA complexes from stressed cells and identifying the bound transcripts using microarrays. Computational analysis of the target transcripts led to the elucidation of a shared, U-rich motif present in TIA-1 target mRNAs. The data revealed that novel TIA-1 target mRNAs can be successfully identified using this motif and that mRNAs associating with TIA-1 are translationally repressed.
Human colorectal carcinoma RKO cells were cultured in minimum essential medium (Invitrogen), supplemented with 10% fetal bovine serum and antibiotics. Cells were subjected to heat shock (HS) by incubating at 45°C for 1 h. Sodium arsenite and carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP) were from Sigma. Four small interfering RNA (siRNA) molecules targeting TIA were assessed: CTG GGCTAACAGAACAACTAA (T1, employed in subsequent experiments), AACGATTTGGGAGGTAGTGAA (T2), CACAACAAATTGGCCAGTATA (T3), and CGGAAGATAATGGGTAAGGAA (T4); the control siRNA was AATTCTCCGAACGTGTCACGT. Small interfering RNAs (100 nM, QIAGEN) were transfected with Oligofectamine (Invitrogen), and cells were harvested 2 to 4 days after transfection, as indicated.
Immunoprecipitation (IP) of TIA-1-mRNA complexes from RKO cell lysates was used to assess the association of endogenous TIA-1 with endogenous target mRNAs. The IP assay was performed essentially as described previously (32, 37), except 100 million cells were used as starting material and lysate supernatants were precleared for 30 min at 4°C using 15 μg of immunoglobulin G (IgG) (Santa Cruz Biotechnology) and 50 μl of protein A-Sepharose beads (Sigma) that had been previously swollen in NT2 buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 1 mM MgCl2, and 0.05% Nonidet P-40 [NP-40]) supplemented with 5% bovine serum albumin. Beads (100 μl) were incubated (16 h, 4°C) with 30 μg of antibody (either goat IgG [Santa Cruz Biotechnology] or goat anti-TIA-1 [Santa Cruz Biotechnology]) and then for 1 h at 4°C with 3 mg of cell lysate. After extensive washes and digestion of proteins in the IP material (37), the RNA was extracted and used either for hybridization of cDNA arrays (below) or for verification of TIA-1 target transcripts. For the latter analysis, RNA in the IP material was used to perform reverse transcription-PCR (RT-PCR) to detect the presence of specific target mRNAs using gene-specific primer pairs (available upon request). PCR products were visualized after electrophoresis in 1% agarose gels stained with ethidium bromide. To assess the proteins present in the IP material, the above procedure was followed, except proteins were not digested and were instead extracted from the beads using Laemmli buffer and detected by Western blot analysis.
Where indicated, purified recombinant proteins (either glutathione S-transferase [GST] or GST-TIA-1 at a concentration of 500 nM) were incubated with the precleared cell lysates for an additional 30 min at 45°C, before adding beads that had been precoated with anti-GST antibody. All subsequent steps were as described above, including IP, washes, RT, and PCR amplification.
RNA in the material obtained after IP reactions using either an anti-TIA-1 antibody or IgG was reverse transcribed in the presence of [α-33P]dCTP (MP Biomedicals), and the radiolabeled product was used to hybridize cDNA arrays (http://www.grc.nia.nih.gov/branches/rrb/dna/index/dnapubs.htm#2,MGC arrays containing 9,600 genes), employing previously reported methodologies (32, 37, 38). All of the data were analyzed using the Array Pro software (Media Cybernetics, Inc.), then normalized by Z-score transformation (8) and used to calculate differences in signal intensities. Significant values were tested using a two-tailed Z-test and a P of ≤0.01. The data were calculated from three independent experiments. The complete cDNA array data are available from the authors.
Human UniGene records were first identified from the most strongly enriched TIA-1 targets derived from the array analysis; the top 185 transcripts from which the 3′UTRs were available (available upon request) served as the experimental data set for the identification of the TIA-1 motif. The 185 3′UTR sequences were first scanned with RepeatMasker (www.repeatmasker.org) to remove repetitive sequences. The remaining sequences were divided into 100-base-long subsequences with 50-base overlap between consecutive sequences and were organized into 50 data sets. Common RNA motifs were elucidated from each of the 50 random data sets. The top 10 candidate motifs from each random data set were selected and were used to build the stochastic context-free grammar (SCFG) model. The SCFG model of each candidate motif was used to search against the experimental 3′UTR data set as well as the entire human UniGene 3′UTR data set to obtain the number of hits for each motif. The motif with the highest enrichment in the experimental data set over the entire UniGene data set was considered to be the best TIA-1 candidate motif. The enrichment was examined by Fisher's exact test. The identification of the RNA motif in unaligned sequences was conducted using FOLDALIGN software (21), and the identified motif was modeled by the SCFG algorithm and searched against the transcript data set using the COVE and COVELS software packages (17). The motif logo was constructed using WebLogo (http://weblogo.berkeley.edu/). RNAplot was used to depict the secondary structure of the representative RNA motifs. The computation was performed using the NIH Biowulf computer farm. Both UniGene and Refseq data sets were downloaded from NCBI.
For in vitro synthesis of biotinylated transcripts, reverse-transcribed total RNA was used as the template for PCRs using 5′ oligonucleotides that contained the T7 RNA polymerase promoter sequence. All oligonucleotide pairs used to synthesize DNA templates for the production of biotinylated transcripts are available upon request. The following genes (with the amplified regions indicated in parentheses) were assayed for biotin pulldown: ACTG1 (1201 to 1904), PFN1 (556 to 733), ACTB (1219 to 1724), APH-1A (864 to 1910), DEK (1312 to 2077), MTA1 (2018 to 2609), GAPDH (981 to 1283), CALM2 (522 to 1071), SNRPF (345 to 445), CDK9 (1229 to 1732), APEX1 (1314 to 1541), and PTMA (546 to 1202). The PCR-amplified products were resolved on agarose gels, and transcripts were purified and used as templates for the synthesis of the corresponding biotinylated RNAs using T7 RNA polymerase and biotin-CTP (39).
Biotin pull-down assays (39) were carried out by incubating 500 nM of either recombinant GST or GST-TIA-1 proteins (prepared in Escherichia coli using an expression vector kindly provided by J. Varcarcel and P. Anderson ) with 0.2 μg of biotinylated transcripts for 30 min at 45°C. Complexes were isolated using streptavidin-conjugated Dynabeads (Dynal), and bound proteins in the pull-down material were analyzed by Western blotting using antibodies recognizing GST (below).
RKO cells cultured on dishes containing coverslips were fixed in 4% formaldehyde (15 min) and permeabilized in cold 0.2% Triton X-100 in phosphate-buffered saline (PBS) (15 min). After incubation in blocking buffer (5% horse serum in PBS) for 1 h at 37°C, coverslips were incubated with either goat anti-TIA-1 or goat anti-TIAR (Santa Cruz Biotechnology) in blocking buffer (1 h at 37°C, 1:200 dilution), washed with PBS containing 0.1% Tween 20, and further incubated with Alexa Fluor 568-labeled donkey anti-goat IgG (heavy plus light chains) (Molecular Probes; 1 h at 37°C, 1:500 dilution). After washes with PBS containing 0.1% Tween 20, coverslips were mounted in Vectashield (Vector Laboratories) and visualized with a Zeiss LSM410 confocal microscope. Representative photographs from three independent experiments are shown. Negative-control incubations were performed without primary antibody.
Newly translated CALM2, SNRPF, CASP8, and (control) glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were assessed by incubating 4 × 106 cells with 1.5 mCi l-[35S]methionine and l-[35S]cysteine (Easy Tag EXPRESS; NEN/Perkin Elmer) per 100-mm plate for 20 min, whereupon cells were lysed using TSD lysis buffer (50 mM Tris [pH 7.5], 1% sodium dodecyl sulfate [SDS], and 5 mM dithiothreitol). IP reactions were carried out as previously described (33) for 1 h at 4°C using appropriate antibodies and IgG as a control. Following extensive washes in TNN buffer (50 mM Tris [pH 7.5], 5 mM EDTA, 0.5% NP-40, 250 mM NaCl), the IP material was resolved by either 15% (for CALM2 and SNRPF) or 10% (for CASP8) SDS-polyacrylamide gel electrophoresis, transferred onto polyvinylidene difluoride filters, and visualized using a PhosphorImager (Molecular Dynamics).
The preparation of whole-cell, cytoplasmic, and nuclear lysates was previously described (28, 40). Protein lysates (5 to 20 μg) were resolved by SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose membranes. Antibodies were used to detect α-tubulin, hnRNP C1/C2, S6, GST, TIAR, or TIA-1 (Santa Cruz Biotechnology), CASP8 (BD Pharmingen), CALM2 (Upstate Cell Signaling Solutions), or SNRPF (a gift from D. Ingelfinger). Following secondary antibody incubations, signals were visualized by enhanced chemiluminescence.
The subcellular localization of TIA-1 was investigated in colorectal carcinoma RKO cells exposed to a variety of stress agents, including heat (45°C, 1 h) (HS), sodium arsenite (0.5 μM, 45 min), or the mitochondrial uncoupler FCCP (1 μM, 90 min) (Fig. (Fig.1A).1A). Each of these treatments triggered the accumulation of cytoplasmic TIA-1 into SGs (Fig. (Fig.1A),1A), in keeping with previous reports (24, 25); HS yielded the most robust effects and was chosen for further study. By Western blot analysis, the relative levels of TIA-1 in the nuclear and cytoplasmic compartments remained largely unchanged in response to HS (Fig. (Fig.1B),1B), supporting the notion that SGs were formed by dynamic recruitment of TIA-1 (24) through mechanisms that did not significantly deplete the pool of nuclear TIA-1.
In order to identify the collection of mRNAs that were TIA-1 targets, we first sought to perform immunoprecipitation assays under conditions that preserved the pools of mRNAs bound to TIA-1. As assessed by Western blot analysis, TIA-1 was undetectable in the IgG IP material but was specifically immunoprecipitated by the anti-TIA-1 antibody; in the latter IP group, TIA-1 levels were again found to remain unchanged in response to HS (Fig. (Fig.1C).1C). Under these conditions, no significant TIAR was detected in the TIA-1 IP material, although after we used larger amounts of IP material and longer exposure times to detect Western blotting signals, we did observe low levels of TIAR in the TIA-1 IP samples (unpublished data), consistent with the ability of TIAR and TIA-1 to form protein-protein associations. Given that TIA-1 is an integral constituent of SGs, we tested whether SGs might have been immunoprecipitated along with TIA-1 by monitoring the presence of the small ribosomal subunit S6 in the IP material. As shown, S6 was undetectable in all of the IP lanes, suggesting that intact SGs indeed failed to immunoprecipitate by this procedure. The RNA associated with TIA-1 in the IP material was then extracted (Fig. (Fig.2A)2A) and used to prepare reverse-transcribed products that were subsequently hybridized to human cDNA arrays (http://www.grc.nia.nih.gov/branches/rrb/dna/dna.htm#, MGC arrays, 9,600 genes). The array hybridization patterns obtained from untreated RKO cultures were similar to those obtained from HS cultures, but overall signal intensities were higher in the HS groups (not shown). The latter set of array data was therefore used for the identification of TIA-1 target mRNAs. The association between TIA-1 and target mRNAs was deemed specific on the basis of evidence that the anti-TIA-1 antibody exhibited no appreciable cross-reactivity with other cellular proteins (particularly TIAR, as indicated below). The specificity of the interactions was further ensured through the elimination of nontarget mRNAs associating with either the antibody or the beads (detected in the IgG arrays). Three hundred array spots (~3% of the total spots on the array) had Z scores of >1.00 in a comparison of the signals in TIA-1 IP arrays with those in IgG IP arrays and were thus deemed to represent specific TIA-1-associated transcripts. Of the specific TIA-1-associated transcripts, the 185 transcripts for which full-length mRNAs were available (the experimental data set [available upon request]) were selected for further analysis.
The RNA sequences of the experimental data set were used in computational analyses to identify and characterize TIA-1 motifs on the basis of both primary RNA sequences and secondary structures. Of the 100 possible candidate motifs initially identified from the experimental data set, one motif comprising 30 to 37 nt was identified as showing the highest relative number of hits in the experimental data set over the entire UniGene data set. The sequence alignment and motif logo (graphic representation of the relative frequency of nucleotides at each position), as well as examples of the secondary structures of this putative TIA-1, are shown (Fig. 2B and C). The motif was found to be highly U-rich in its 5′ segment and AU-rich in its 3′ segment (Fig. (Fig.2B);2B); Fig. Fig.2C2C depicts 10 examples of the TIA-1 motif, with the corresponding mRNAs indicated below. Each hit of this motif was assigned a score, a value that reflects the degree to which each particular motif matches the motif model (Fig. (Fig.2B).2B). A partial list (after removal of hypothetical proteins) of the experimental data set mRNAs bearing the TIA-1 motif is shown (Table (Table1),1), and the positions of individual TIA-1 motif hits and the score of each motif hit are indicated in Table Table1.1. Similar to what was observed for mRNAs bearing HuR motifs (32), there were multiple occurrences of the TIA-1 motif within many of the transcripts examined. It is important to note that only 42 putative targets (~23% of the experimental data set) had hits for the TIA-1 motif. The fact that the remaining 143 putative TIA-1 target transcripts from the experimental data set did not appear to contain the 30- to 37-base motif suggested that additional TIA-1 motifs may also exist that were not identified in this analysis. In this regard, it should also be explained that other motifs which had a greater number of hits in the experimental data set were identified, but the hit frequency was lower relative to the hit frequency within the UniGene database and were thus considered to be less strong TIA-1 motifs. Importantly, the two mRNAs that have been shown to be targets of TIA-1 (those encoding TNF-α and COX-2 [10, 35]) were found to contain at least one motif hit (Table (Table22).
Table Table33 lists the relative frequency with which the motif was found in the UniGene database; sequences comprising 5′UTR, coding region (CR), and 3′UTR were assessed separately. The number of motif hits for each data set was calculated with respect to the relative sequence length of the data set (presented as hits per kb). Within the UniGene database, the frequency of the TIA-1 motif in the 3′UTR is 0.203 per kb, a markedly higher frequency than that seen for the 5′UTR (0.056 per kb) or the CR (0.029 per kb). Comparable differences were seen in the experimental data set (Table (Table3),3), wherein the TIA-1 motif hits were almost exclusively found in the 3′UTRs of target mRNAs. As anticipated, the frequency ofthe motif occurrence (0.469 per kb in the 3′UTR) was higher than that calculated for the general UniGene transcript set. The complete list of human UniGene transcripts containing the TIA-1 motif is available from the authors. Taken together, the TIA-1 motif identified from the array-derived experimental data set displays features that were previously predicted, including its U-richness, its existence on published TIA-1 target mRNAs, and its preferential location on the 3′UTRs of target mRNAs (12, 29).
To test the usefulness of the motif in predicting TIA-1 target mRNAs, we carried out two types of validation efforts. First, we monitored the ability of endogenous TIA-1 to immunoprecipitate TIA-1-containing mRNA-protein (mRNP) complexes and assessed the presence of bound target mRNAs of interest by RT-PCR (Fig. (Fig.3A).3A). The mRNAs tested by this method included both array-identified mRNAs (array targets) as well as transcripts predicted to be TIA-1 targets after a search of the UniGene database for motif-bearing mRNAs (database targets). Following IP of mRNP complexes present in RKO cell lysates using an anti-TIA-1 antibody (or IgG in control IP reactions), RT-PCR assays were performed to detect the presence of the endogenous mRNAs of interest in the IP material (Fig. (Fig.3A).3A). Of note, nontarget GAPDH and SDHA mRNAs (encoding housekeeping genes) could also be amplified, albeit inefficiently and to the same extent in both IP groups; these findings revealed the presence of low levels of contaminating, unspecific mRNAs in all IP samples, verified the use of equal amounts of input material, and demonstrated that TIA-1 mRNA targets were amplified from TIA-1 IP material much more readily than from IgG IPs.
Second, we assessed the interaction of TIA-1 with target transcripts in vitro. Biotinylated transcripts encompassing motif-containing 3′UTR sequences (details in Materials and Methods and unpublished data) from the mRNAs indicated (Fig. 3B and C) were incubated with recombinant TIA-1 protein (GST-TIA-1). The formation of complexes was then assessed by pulling down the biotinylated RNA using streptavidin-coated beads followed by Western blot analysis of TIA-1 levels in the pull-down material. Included in this validation were several TIA-1 target transcripts identified in the cDNA array: CALM2, ACTB, APH-1A, SNRPF, CDK9, and APEX1. As shown in Fig. Fig.3B,3B, all of the biotinylated transcripts were capable of pulling down TIA-1 but not a control protein (GST); in addition, one biotinylated transcript, corresponding to SLAMF1 3′UTR, failed to show binding (not shown). In further control incubations, a biotinylated GAPDH transcript that lacked a TIA-1 motif failed to show any binding to TIA-1 (Fig. (Fig.3B).3B). The validation efforts were extended to UniGene database genes predicted to encode targets of TIA-1 on the basis of the detection of TIA-1 motif hits in the corresponding mRNAs. In all cases, the biotinylated transcripts tested (MTA1, ACTG1, PTMA, PFN1, and DEK) showed binding to TIA-1, while control incubations with nontarget RNA (GAPDH) or with GST did not (Fig. (Fig.3C).3C). Moreover, in vitro supplementation of recombinant purified protein (either GST-TIA-1 or GST) to cell lysates that were prepared as described above (Fig. (Fig.3A)3A) recapitulated the specific binding of TIA-1 to the target mRNAs tested (Fig. (Fig.3D),3D), indicating that exogenously added TIA-1 retains the target specificity of the endogenous protein (Fig. (Fig.3A3A).
To ascertain whether the presence of the TIA-1 motif was sufficient for TIA-1 to associate with a given RNA, we synthesized two sets of biotinylated transcripts (Fig. (Fig.4A,4A, schematic). The first were three small RNAs comprising either the TIA-1 motif from the TNF-α mRNA (similar results were observed with the TIA-1 motif present in the CASP8 mRNA [data not shown]) plus 20-base flanking RNA from each side of the motif [TIA-1(+)], only the flanking regions [TIA-1(−)], or a mutated TIA-1 motif with flanking regions [TIA-1(mut)]. In the second set, we tested whether the presence of the TIA-1 motif would render a heterologous, nontarget transcript (the GAPDH 3′UTR) capable of forming complexes with TIA-1. As shown in Fig. Fig.4B,4B, only transcripts bearing the TIA-1(+) motif in each transcript set showed specific interaction with GST-TIA-1. No binding was seen when the TIA-1 motif was absent [TIA-1(−) transcripts] or when it was mutated [TIA-1(mut) transcripts], indicating that the presence of an intact TIA-1 motif was required for binding. No complexes were detected in control incubations with GST.
We also tested whether TIA-1 colocalized with target mRNAs by in situ hybridization using labeled antisense transcripts. While this approach posed major technical challenges, preliminary experiments showed that one abundant target mRNA colocalized with TIA-1 in cytoplasmic SGs after HS (unpublished data). Together, these approaches support the notion that TIA-1 associates with motif-bearing mRNAs.
Additional insight into the association of TIA-1 with target mRNAs and the functional consequences of these interactions was sought by RNA interference (RNAi)-mediated reduction of TIA-1 levels. The effects of four different small interfering RNAs targeting TIA-1 are shown (Fig. (Fig.5A;5A; T1 was used in subsequent experiments). Transfection with each of the siRNA molecules caused a dramatic reduction in TIA-1 levels (to less than 5% of the levels seen in control siRNA populations). It was important to assess the levels of TIAR in the TIA-1 siRNA populations, given the extensive sequence homology between the two proteins; as shown, TIAR levels remained unaltered (Fig. (Fig.5A).5A). In keeping with this reduction in TIA-1 levels, RT-PCR amplification of putative target transcripts encoding CALM2, SNRPF, and CASP8 was markedly reduced when using IP material obtained from the TIA-1 siRNA population (Fig. (Fig.5B).5B). Analysis of TIA-1 and TIAR by immunofluorescence confirmed the effects of RNAi on TIA-1 expression and further showed that SGs were readily detectable in the TIA-1 siRNA population (as seen by TIAR fluorescence), suggesting that TIAR alone may be sufficient for the assembly of SGs (Fig. (Fig.66).
To investigate the influence of TIA-1 on the expression of target transcripts, the steady-state levels of mRNAs encoding CALM2, SNRPF, and CASP8 were first examined in cells transfected with either a control siRNA or an siRNA targeting TIA-1. As shown in Fig. Fig.7A,7A, the levels of these mRNAs remained unchanged in all of the treatment groups, as assessed by semiquantitative RT-PCR analysis. To test whether TIA-1 contributed to regulating the levels of CALM2, SNRPF, and CASP8 by influencing their translation, nascent protein synthesis was measured by performing a brief (20-min) incubation in the presence of 35S-labeled amino acids, immediately followed by IP using specific antibodies (20, 33). HS was found to considerably lower the translation of these target mRNAs (Fig. (Fig.7B).7B). Importantly, their translation was enhanced in each TIA-1 siRNA group (Fig. (Fig.7B),7B), in keeping with the role of TIA-1 as a translational suppressor. The levels of the encoded proteins changed less dramatically in each population, although HS reduced their levels in each case, and TIA-1 silencing caused increases in protein abundance, particularly in the HS groups (Fig. (Fig.7C).7C). These observations support the notion that TIA-1 suppressed the levels of expression of the proteins encoded by target mRNAs in both unstimulated and heat-stressed cells (Fig. (Fig.7C7C).
In summary, the TIA-1 motif identified using this approach appears to predict with a high degree of confidence if a novel mRNA will associate with TIA-1 and hence become subject to TIA-1-mediated translational repression.
Here, we describe studies aimed at systematically identifying mRNA subsets associated with TIA-1, a pivotal regulator of the cellular response to inflammatory, proapoptotic, and stress-inducing agents. We employed IP assays to isolate TIA-1-containing mRNPs and then used the mRNA in the IP material for hybridization of cDNA arrays. Computational analysis of the target mRNAs led to the identification of a 30- to 37-nt motif present in 23% of the target mRNAs, suggesting that other, as-yet-unknown signature motifs might exist on TIA-1 target mRNAs. The 5′ segment of the 30- to 37-nt RNA motif was remarkably U-rich, in agreement with earlier analyses of TIA-1-bound RNAs (12, 29), while the 3′ segment was predominantly AU-rich. The TIA-1 motif was strikingly more abundant in the 3′UTRs of TIA-1-bound mRNAs on the array, as well as in putative TIA-1 target mRNAs in the UniGene database, and was predicted to adopt a distinct secondary structure consisting of a loop of variable size and a bent stem. Importantly, the motif was found in the TNF-α and COX-2 mRNAs, two reported TIA-1 target transcripts.
To assess the usefulness of the TIA-1 motif for identifying bona fide target mRNAs, several validation approaches were undertaken. First, the mRNAs present in the TIA-1-associated IP material were tested by RT-PCR analysis using sequence-specific primers; among the particular target mRNAs examined were several transcripts that were identified during the initial cDNA array analysis and several that were predicted to be targets after the UniGene database was searched for transcripts containing the TIA-1 motif. By this methodology, all of the predicted targets were validated (Fig. (Fig.3A),3A), except for two array transcripts which did not display appreciable enrichment in the TIA-1 IP material (not shown). In the second set of validation efforts, we sought to determine whether putative TIA-1 target transcripts, synthesized in the presence of biotinylated CTP, formed complexes with recombinant TIA-1. All of the predicted TIA-1-mRNA interactions were also validated by this in vitro assay (Fig. (Fig.3B);3B); only biotinylated SLAMF1 3′UTR transcript did not show the expected binding to TIA-1, despite being enriched by IP plus RT-PCR analysis (not shown), suggesting that perhaps the recombinant transcript did not retain the proper native conformation of the SLAMF1 mRNA. Importantly, when the TIA-1 motif was artificially linked to a heterologous RNA (the nontarget GAPDH 3′UTR), the resulting chimeric transcript was rendered a TIA-1 target; binding was undetectable or at background levels when the motif was absent or mutated (Fig. (Fig.4).4). These experiments further support the validity of the TIA-1 motif elucidated in this study. An additional line of investigation used to substantiate the existence of these interactions assessed the subcellular colocalization of TIA-1 and target SNRPF transcript. The in situ hybridization signals for SNRPF RNA and the immunofluorescent TIA-1 signals (available upon request) overlapped at the sites of SGs. Taken together, the TIA-1 motif reported here successfully identified 23% of TIA-1 target RNAs.
TIA-1 has been reported to participate in the regulation of bound transcripts in at least two distinct processes: pre-mRNA splicing and translational suppression. TIA-1 was previously shown to regulate the splicing of several mRNAs, including the TIA-1 pre-mRNA itself and those encoding fibroblast growth factor receptor 2 and the Fas receptor (11, 18, 29). In this regard, it should be noted that no TIA-1 motif hits were found on either the TIA-1 mRNA or the fibroblast growth factor receptor 2 mRNA (data not shown), although TIA-1 motifs were found on two Fas receptor isoforms (TNFRSF6 and ARTS-1). In this investigation, we have not directly tested whether the mRNAs bearing the motif described here are also targets of TIA-1-regulated splicing. Since the array-based identification of TIA-1 target transcripts was conducted selectively on poly(A)-containing RNA (Materials and Methods), we anticipate that only mature mRNA species were detected on the cDNA arrays. Accordingly, it remains to be directly tested whether a different motif signals the TIA-1-dependent regulation of pre-mRNA splicing.
However, TIA-1 has been characterized most extensively as a translational suppressor. The findings reported here indeed support this role for TIA-1, as the translation of the target mRNAs studied (CALM2, SNRPF, and CASP8) was enhanced when TIA-1 levels were knocked down by RNA interference (Fig. 7B and C). The TIA-1-mediated translational suppression likely relies on the ability of TIA-1 to promote the formation of noncanonical preinitiation complexes by usurping the position of the ternary complex (eIF-2-GTP-) at the 5′UTR of an mRNA. Whereas the active ternary complex (featuring unphosphorylated eIF-2α) promotes the initiation of translation, TIA-1 instead triggers the aggregation of TIA-1-associated ribonucleoprotein complexes into translationally silent SGs (2). In order to assess the influence of TIA-1 on the translation of target mRNAs, we have employed a methodology for assessing nascent protein biosynthesis which measures the pulse incorporation of 35S-labeled amino acids onto nascent polypeptide chains. This method uniquely provides a measure of new translation but has major limitations, as it can be used only to analyze abundant proteins for which highly sensitive and specific antibodies are available. Approximately one dozen additional antibodies were tested to assess as many additional proteins encoded by putative TIA-1 target mRNAs. Unfortunately, the IP signals in each case were well below the levels of detection of the assay (not shown), and therefore, the nascent translation of the corresponding proteins could not be studied by this approach. A comprehensive assessment of the translation of TIA-1's target mRNAs thus awaits the development of more sensitive methods.
The changes in abundance of the CASP8, CALM2, and SNRPF proteins in cells expressing different levels of TIA-1 determined by Western blotting (Fig. (Fig.7C)7C) mirrored those observed when measuring nascent translation (Fig. (Fig.7B),7B), suggesting that the changes in protein abundance are indeed linked to the changes in protein biosynthesis influenced by TIA-1. Interestingly, our findings support the notion that TIA-1 functions as a translational inhibitor even in the absence of stress, since the nascent translation of CASP8, CALM2, and SNRPF was elevated in TIA-1 knockdown cells that had been left without HS. These results are also in agreement with earlier observations that in unstimulated macrophages derived from TIA-1−/− mice, the levels of TNF-α mRNA found in polysomes were higher than those seen in macrophages from wild-type mice, and the TIA−/− cells expressed elevated levels of the cytokine (35). In our studies, HS did not seem to increase the levels of TIA-1 in the cytoplasm (Fig. (Fig.1B)1B) or promote the binding of TIA-1 to target mRNAs (data not shown). Thus, the mechanism(s) whereby HS silences target mRNA translation in a TIA-1-dependent fashion, including possible TIA-1 posttranslational modification through phosphorylation or its association with other proteins, remains to be formally investigated. After the 1-hour HS treatment, however, it is unlikely that the pronounced decline in the levels of these three proteins is due solely to reduced translation rates (Fig. (Fig.7B);7B); instead, it is likely to be influenced by altered proteolysis or other posttranslational events. In light of these observations, we propose that TIA-1 contributes to altering protein expression by influencing the biosynthesis of encoded proteins. This level of regulation likely functions in juxtaposition with additional processes controlling protein levels, such as subcellular protein transport, proteolysis, and/or protein secretion.
It is noteworthy that in populations in which TIA-1 was knocked down (Fig. (Fig.5),5), SGs were still detected (Fig. (Fig.6);6); similarly, silencing of TIAR failed to block SG formation (K. Mazan-Mamczarz and M. Gorospe, unpublished data). Whereas TIA-1 and TIAR appear to have interchangeable roles regarding SG formation, their relative affinities for target mRNAs have not been compared systematically. En masse efforts to identify TIAR and TIA-1 target mRNAs under way in our laboratory have indicated the existence of both shared and specific target ARE-containing transcripts (I. López de Silanes, K. Mazan-Mamczarz, and M. Gorospe, unpublished data), in keeping with the notion that TIAR and TIA-1 are functionally distinct (42), although they both appear to bind to several common targets, such as the COX-2 and TNF-α mRNAs.
In summary, we have systematically identified many TIA-1 target mRNAs and describe a common RNA motif among them. Using a variety of approaches, the association of TIA-1 with mRNAs that were either detected as microarray targets or identified on the basis of the presence of the TIA-1 motif in the UniGene database transcripts was validated using several approaches. Importantly, TIA-1 was found to repress the translation of target mRNAs. These discoveries provide comprehensive and valuable insight into the ribonucleoprotein complexes that govern gene expression at the posttranscriptional level.
We thank K. G. Becker and the NIA Array Facility for providing cDNA arrays for analysis and A. Lal and T. Kawai for valuable discussions.
This research was supported by the Intramural Research Program of the NIA, NIH.