The UV-inducible RNA-binding protein A18 (A18 hnRNP) plays a protective role in the genotoxic stress response.

We have previously shown that specific RNA-binding proteins (RBP) are activated by genotoxic stress. The role and function of these stress-activated RBP are, however, poorly understood. The data presented here indicate that the RBP A18 heterogeneous ribonucleoprotein (hnRNP) is induced and translocated from the nuclei to the cytoplasm after exposure to UV radiation. Using a new in vitro system we identified potential cellular targets for A18 hnRNP. Forty-six mRNA transcripts were identified, most of which are stress- or UV-responsive genes. Two important stress-responsive transcripts, the replication protein A (RPA2) and thioredoxin, were studied in more detail. Northwestern analyses indicate that A18 hnRNP binds specifically to the 3'-untranslated region of RPA2 transcript independently of its poly(A) tail, whereas the poly(A) tail of thioredoxin mRNA reinforces binding. Overexpression of A18 hnRNP increases the mRNAs stability and consequently enhances translation in a dose-dependent manner. Moreover, cell lines expressing reduced levels of A18 hnRNP are more sensitive to UV radiation. These data suggest that A18 hnRNP plays a protective role against genotoxic stresses by translocating to the cytosol and stabilizing specific transcripts involved in cell survival.

The cellular response to genotoxic and non-genotoxic stresses is complex. It includes multiple regulatory mechanisms that are generally thought to have protective roles. Cells respond to stress in a limited number of ways by adjusting regulatory components of basic processes such as replication, transcription, and/or translation. Much emphasis has been put on stress responses involving replication or the activation of specific genes in response to DNA damage (1); however, the regulation of post-transcriptional and translational events in response to stress has not been studied extensively. Post-transcriptional regulation can be mediated through interaction of regulatory proteins with an mRNA 3Ј end (2). This mechanism, which occurs in several organisms, is not fully understood. Most regulations of this type have been observed during early development of different organisms from Caenorhabditis elegans to mammals (3). One possible mechanism by which regulation of translation initiation can be mediated through the 3Ј end of an mRNA transcript has suggested that specific proteins bound to this region could contact the basal translation apparatus and influence translational activation or repression (2). A recent review (4) described the possibility for RNA-binding proteins to shuttle between cellular compartments either constitutively or in response to stress and regulate the localization, translation, or turnover of mRNAs. Post-transcriptional regulation can also occur through mRNAs stabilization. Recent studies describe the stabilization of mRNAs by specific RBPs 1 in response to hypoxia (5) or extracellular signals (6).
In a previous study (7) we have shown that RNA binding activity of specific proteins can be induced by DNA-damaging agents. Induction of an RBP at the mRNA levels was first reported for the A18 hnRNP after UV radiation (8). The hnRNPs are a sub-group of ribonucleoproteins (RNPs) found in the nucleus and involved in RNA processing (9). The A18 hnRNP was originally cloned by hybridization subtraction on the basis of rapid induction in UV-irradiated Chinese hamster ovary cells (8). Since then, the human A18 hnRNP was cloned and characterized (10). The human A18 hnRNP is a rather unique RNP. In addition to containing a conserved RNA binding domain, it also contains several repeats of an RGG box in its auxiliary domain (10). The RGG boxes were first identified as single-stranded nucleic acid binding motifs in an hnRNP that does not contain the conserved RNA binding domain (11). The auxiliary domains of RNPs are associated with proteinprotein interaction (9); therefore, the presence of singlestranded nucleotide binding domains in this location is unusual.
In this report we show that the nuclear A18 hnRNP is not only induced but also translocated to the cytoplasm in response to UV radiation. In addition, we established an in vitro system to isolate and identify A18 hnRNP most probable mRNA targets. Forty-six mRNAs transcripts have been identified, a large proportion of which are UV-or stress-responsive. Our data indicate that A18 hnRNP binds specifically to the 3Ј-UTR of the replication protein A (RPA2) and the human thioredoxin mRNAs. Co-transfection of A18 hnRNP with a CAT expression vector resulted in increased message stability and CAT activity. Moreover, cells expressing reduced levels of A18 hnRNP are more sensitive to UV radiation. Taken together these data suggest that stabilization of stress-responsive transcripts by A18 hnRNP may protect the cells against genotoxic insults.
XhoI sites of pcDNA3.1 (Invitrogen, Carlsbad, CA). The A18 hnRNP-GFP fusion expression vector was constructed by amplifying the coding region of GFP from pEGFP (Invitrogen) and cloning it downstream of A18 hnRNP into pcDNA3.1 (Invitrogen). For constitutive expression of the CAT protein, the ORF of CAT was cloned into the NheI/BamHI sites of pcDNA3.1 (Invitrogen) to generate the plasmid pcDNA3.1-CAT (CAT). The 3Ј-UTR of RPA2 was amplified by PCR from the full-length cDNA of RPA2 and cloned into the BamHI/XhoI sites of pcDNA3.1-CAT to generate the plasmid pcDNA3.1-CAT/RPA2-3Ј-UTR (CAT-UTR). To generate the antisense A18 hnRNP vector, the ORF of A18 hnRNP was cloned in antisense orientation into the HindIII/XhoI sites of pcDNA3.1.
Cell Lines and Treatments-The human colorectal carcinoma RKO cells were maintained in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum. For transient transfection, cells were cultured to 80% confluency, and the indicated plasmids were transfected with Fugene 6 lipid mix (Roche Molecular Biochemicals). A total amount of 11 g of plasmids DNA was maintained by supplementing with pcDNA3.1. Each dish received 1 g of CAT or CAT-UTR expression vector and up to 10 g of A18 hnRNP. Forty-eight hours after transfection, cells were harvested and lysed for further analyses. Analyses were performed in triplicate. For stable transfection, hnRNP A18-GFP was transfected into 50% confluent RKO cells. Selection was performed with 400 g/ml hygromycin B for 2 weeks.
Clonogenic survival after UV irradiation was performed by colony formation assay. The RKO antisense cell line was established by stable transfection of the pCDNA3.1 expression vector containing the A18 hnRNP ORF in the antisense orientation into RKO cells. Single colonies were expanded, and expression levels of A18 hnRNP were verified by Western blot. About 200 cells were seeded/100-mm dish for overnight growth, then the medium was removed, and the cells were irradiated with the indicated dose of UV. Treatments were performed in triplicate for each dose. Two weeks later, colonies containing more than 50 cells were counted.
CAT Assay-The CAT activity was measured essentially as described before (12). Cellular extracts containing either 50 or 25 g of protein were incubated at 37°C for 1-3 h and separated on TLC plates. The conversion of chloramphenicol was quantified on a PhosphorImager (Molecular Dynamics STORM) with the ImageQuant software.
Fluorescence-The A18 hnRNP-GFP-overexpressing cells were grown on coverslips and treated with UV radiation at a dose of 20 J/m 2 . Cells were put back into culture for 3 h, and fluorescence was observed with a fluorescence microscope (Zeiss, Axioskop, objective 10ϫ, HB100-W mercury lamp).
Expression and Purification of Recombinant hnRNP A18 -The hnRNP A18 cDNA-coding region (516 base pairs) (10) was amplified by PCR and cloned into the NdeI and XhoI sites of pET21a (Novagen, Madison, WI). Expression was achieved in Escherichia coli BL21(DE3) as recommended by the manufacturer (Novagen) except that the bacteria were grown at 30°C. Soluble proteins were loaded on a nickel nitrilotriacetic acid column (Novagen) and stepwise eluted with 100 -350 mM imidazole. The imidazole was removed by gel filtration on a D-Salt Excellulose desalting column (Pierce). The recombinant protein was finally dialyzed against 50 mM sodium phosphate buffer, pH 7.5.
Determining the RNA-Protein Interaction in Nunc-Immuno Tubes-Recombinant A18 hnRNP and nucleolin were coated in Nunc-Immuno tubes (immunotubes) (Nalgen Nunc International, Rochester, NY) at a concentration of 40 g/ml in carbonate buffer, pH 9.6, overnight at 4°C. As a negative control, bovine serum albumin (BSA) with the same concentration or buffer alone was used. The tubes were washed twice with PBS containing 0.1% Tween 20 and washed two more times with PBS. Nonspecific sites were blocked with 2% milk in PBS containing 2 g/ml yeast tRNA for 1 h at room temperature. The tubes were washed 3 times with PBS and 3 times with RNA binding buffer (RBB, 20 mM Tris-HCl, pH 7.5, 60 mM KCl, 1 mM MgCl 2 , 0.2 mM EDTA, 10% glycerol). Labeled TAR RNA (10 6 cpm) was incubated in the immunotubes with 1 ml of RBB containing 100 units of RNase inhibitor and 2 g of yeast tRNA. After incubation for 2 h at room temperature, the tubes were washed 10 times with 1 ml of RBB and 5 times with 1 ml of PBS. Bound RNA was then eluted stepwise with 1 ml of RBB containing either 500 mM, 1 M, or 2 M NaCl. Aliquots (100 l) of each elution fraction were counted, and binding specificity was obtained by subtracting the BSAcoated tubes counts from the A18 hnRNP-coated tubes counts.
Preparation and Selection of Cellular mRNA-Human colon carcinoma (RKO) cells were irradiated with UV at a dose of 20 J/m 2 . Two hours later the cells were harvested, and the mRNA was prepared by acid phenol extraction as described previously (13).
Ten g of mRNA were incubated in the immunotube coated with A18 hnRNP at room temperature and mixed gently by inversion for 2 h. The tube was then washed 10 times with 1 ml of RBB and 5 times with 1 ml of PBS to remove the unbound mRNA. After the last wash, the bound mRNA was eluted stepwise with 1 ml of RBB containing either 500 mM, 1 M, or 2 M NaCl. The eluted mRNA was precipitated with ethanol and used for cDNA synthesis.
cDNA Synthesis and Library Construction-The first strand cDNA was synthesized using the SmartII cDNA synthesis oligo (CLONTECH, Palo Alto, CA) and superscript reverse transcriptase (Life Technologies, Inc.). The reaction was performed in 20 l at 42°C for 1 h. Five l of the reaction mixture was used to amplify the double strand cDNA by long distance PCR using SmartII cDNA PCR amplification oligo (CLON-TECH). PCR was performed for 30 cycles at 95°C for 1 min, 65°C for 1 min, and 68°C for 6 min. The amplified cDNA was purified, ligated to pGEM-T vector (Promega, Madison, WI), and transformed into E. coli DH5␣. Positives colonies were picked for further analysis by PCR fingerprinting and sequencing. Sequencing was performed on an Applied Biosystems 373 (ABI, Foster City, CA) automated sequencer.
Northwestern Blot Assay-Northwestern blots were performed essentially as previously described (7).
Ribonuclease Protection Assay-RNase protection was performed using the ribonuclease protection kit RPAII (Ambion). Briefly, total RNA was prepared from the cells co-transfected with A18 hnRNP and the CAT reporter gene constructs. A 100-base pair fragment upstream of the stop codon of the CAT open reading frame was in vitro transcribed into the complement RNA probe and purified. For each reaction, 20 g of total RNA was co-precipitated with the RNA probe (2 ϫ 10 4 cpm for 1 h) and hybridized overnight at 42°C. For control, 20 g of yeast RNA was co-precipitated with the RNA probes. The hybridized RNA probe was digested with RNase A/T and precipitated. Protected RNA was separated on a 6% polyacrylamide gel.

RESULTS
A18 hnRNP Translocates to the Cytoplasm after Exposure to UV Radiation-The A18 hnRNP was originally cloned by hybridization subtraction on the basis of rapid induction after UV radiation (8). Since then, the human gene has been cloned and shown also to be UV-inducible (10). Because some hnRNPs are known to shuttle between cell compartments either constitutively (14) or in response to stress (15), we evaluated the cellular location of A18 hnRNP after UV radiation. To follow A18 hnRNP protein in the RKO cells, we fused the A18 hnRNP protein to GFP and established a stable cell line with the recombinant expression vector. Our data ( Fig. 1) indicate that most cells have incorporated the fused protein and distinctly show that A18 hnRNP is a nuclear protein (panel B). Overexpression of A18 hnRNP in the absence of stress does not affect cellular growth or induce apoptosis (data not shown). Three hours after exposure to UV radiation (20 J/m 2 ), the protein becomes clearly visible in the cytoplasm of many cells (panel D). Cytoplasmic mRNAs are thus likely to be targeted by the RNA-binding protein A18 hnRNP after UV exposure.
Use of Nunc-Immuno Tubes for RNA-Protein Interactions-To identify A18 hnRNP potential mRNAs targets, we developed an affinity binding technique with immunotubes. We first optimized the RNA binding conditions for A18 hnRNP in immunotubes by using a labeled TAR RNA probe. This probe was selected based on its high affinity for stress-activated RNA-binding proteins (7). To evaluate the extent of nonspecific binding, we treated four tubes in parallel. For controls, one tube was left uncoated to estimate the nonspecific binding of RNA to the tube walls, and one tube was coated with BSA to determine the specificity of the RNA-protein interaction. The last two tubes were coated with either the recombinant A18 hnRNP protein or nucleolin, another RNA-binding protein known for its specific binding to stem loop RNA structure (16). The data presented in Fig. 2 indicate that, under all the elution conditions used, the amount of RNA bound nonspecifically to BSA or the tube walls was less than 20% of the RNA bound to either A18 hnRNP or nucleolin. To make sure that we had eluted all the protein-bound RNA, we treated the tubes with proteinase K. As shown in Fig. 2 (lanes 17-20), proteinase K treatment released some RNA in all the tubes but in much smaller amounts than the salt treatment. These data indicate that under the conditions used most RNAs that are specifically bound to an RNA-binding protein can be eluted. The immunotube technique is thus a suitable and selective technique to identify the potential targets of a given RNA-binding protein.
Identifying A18 hnRNP mRNAs Targets-We then repeated this technique with a newly coated A18 hnRNP immunotube and 10 g of mRNAs isolated from UV-irradiated RKO cells. Because the RKO cells contain several RBP that can be activated by genotoxic stress (7), we reasoned that UV-irradiated RKO cells could be a good source of mRNA targets for A18 hnRNP. After incubation of the UV-treated mRNAs with the A18 hnRNP-coated tube, the specifically bound mRNAs were precipitated, reverse-transcribed, and amplified by PCR. The resulting amplicons were cloned and digested with the fourbase cutter HaeIII (New England Biolabs, Beverly, MA) to evaluate redundancy. Based on this analysis we have determined that several genes were present more than once in the library. Sequencing was performed randomly on non-redundant clones by an automated sequencer. The partial sequences were compared with GenBank TM and EST data bases through the BLAST search engine. As shown in Table I, three main classes of transcripts representing 46 different clones have been identified. A large proportion (ϳ40%) of these genes are UV-or stress-responsive. This indicates that the technique is highly selective.
Four of these genes, RPA2, thioredoxin, ferritin, and nucleophosmin, are either UV-inducible or UV-responsive (17). We also verified the specificity of the mRNAs selected by attempting to amplify by reverse transcription-PCR other transcripts known to be UV-inducible. Transcripts for the growth arrest and DNA damage-inducible gene GADD45 and the cyclin-dependent kinase inhibitor p21 (1) were not found in the pool of mRNAs selected by A18 hnRNP under these conditions.
Another major class of transcripts bound by A18 hnRNP encodes ribosomal proteins. This large number of ribosomal proteins is in good agreement with the increasing body of evidence indicating that translation is an important component of the cellular stress response (18). Several type of stress, including heat shock stress and several chemicals, can induce the synthesis of stress proteins while inhibiting the rate of protein synthesis (Ref. 19 and references within). Stress-induced changes in the stoichiometry of ribosomal proteins may trigger an adaptive response by allowing the translation of a specific set of proteins (20). Three of the ribosomal proteins identified in our screen, L10, S13, and L19, have been associated with the oxidative stress response (21,22). L5 is involved in cellular resistance to general stress stimulus (20), and S6 is induced by cold shock (23). L13 has been suggested as a mediator of the stress induction of the sigma (B) factor in bacteria (24). These data support the idea that our technique is highly selective and that A18 hnRNP specifically targets transcripts involved in the stress response. We do not know whether the other ribosomal proteins identified in our screen play a direct or indirect role in the stress response. Nevertheless, A18 hnRNP may contribute to modulate an adaptive response to cellular stress by targeting specific sets of ribosomal protein transcripts.
Other genes encoding a variety of transcripts were also identified. Among these, two encode new sequences, one of unknown function and one that encodes a hypothetical protein.
Binding of A18 hnRNP to the 18 S rRNA is also of interest since this rRNA is part of the 40 S ribosomal small subunit. Binding to this rRNA may thus indicate a potential role for A18 hnRNP in translation. Based on sequence analysis we have determined that the 18 S rRNA sequence bound to A18 hnRNP corresponds to nucleotides 706 -1385 of the 18 S rRNA (data not shown). This sequence is located in the central domain of the 18 S rRNA and is thought to be involved in eIF3 interaction, which prevents premature association of the large and small ribosomal subunits (25). A potential role in translation regulation is also supported by the binding of A18 hnRNP to the transcripts of two translational elongation factors (1␣ and 1␤2).
A18 hnRNP Binds Specifically to RPA2 and Thioredoxin 3Ј-UTRs-Our immunotube technique revealed that A18 hnRNP binds to a large number of stress-responsive transcripts (Table I). To determine the binding specificity of A18 hnRNP, we performed Northwestern analyses with RPA2 and thioredoxin, two stress-responsive transcripts identified in our screen (Table I). RPA2 and thioredoxin were selected for this assay based on their UV responsiveness and their well established role in the stress response. The RPA2 protein is important for DNA replication and nucleotide excision repair and is specifically phosphorylated after exposure to UV radiation (17). RPA2 is also implicated in UV-induced replication arrest (17). Thioredoxin is a UV-inducible protein involved in transcriptional processes such as induction of AP-1 activity and the inhibition of NF-B activation (26). Thioredoxin has also been associated with tumor growth and is now becoming a new target for anti-cancer drugs (26).
We first constructed four overlapping probes with the RPA2 transcript (Fig. 3A) and hybridized them to increasing amounts of recombinant A18 hnRNP protein (Fig. 3B). Our data indicate that A18 hnRNP does not bind to RPA2 ORF or the 5Ј-UTR. On the other hand, A18 hnRNP binds very strongly to RPA2 3Ј-UTR irrespective of the presence of the poly(A) tail. Binding is detectable with as little as 200 ng of recombinant protein (Fig.  3B, lanes 2). We also repeated a similar experiment with the thioredoxin transcript (Fig. 4). Six different overlapping probes were generated (Fig. 4A). Our data (Fig. 4B) indicate that again A18 hnRNP does not bind to the ORF of the transcript and binds only weakly to the 5Ј-UTR. However, in this case, binding is reinforced by the presence of the transcript poly(A) tail.
Interestingly, binding appears to be more sensitive in the presence of the ORF even though A18 hnRNP does not bind to it. The presence of the ORF may affect the overall structure of the RNA and increase binding. These data confirm the binding specificity of A18 hnRNP to two transcripts selected by our immunotube technique.
Effects of A18 hnRNP on Translation and mRNAs Stability-Binding of proteins to the 3Ј-UTR of a transcript can affect both the rate of translation initiation and the stability of a transcript. To evaluate the overall effect of A18 hnRNP on translation, we performed transient co-transfection of A18 hnRNP with a constitutive expression vector for the CAT protein. We used two different CAT expression vectors to determine whether A18 hnRNP required the RPA2 3Ј-UTR to mediate an effect on translation. The first construct (CAT) contains the CAT cDNA under the control of the pCMV promoter and a poly(A) tail from the bovine growth hormone transcript (Fig. 5,  A and B). The second construct (CAT-UTR, Fig. 5, C and D) is identical to the CAT construct except that we have inserted the RPA2 3Ј-UTR without its poly(A) tail upstream of the bovine growth hormone poly(A)s. Our data (Fig. 5) indicate that cotransfection of A18 hnRNP with either construct produced a dose-dependent increase of the CAT activity (Fig. 5, A and C). The addition of the RPA2 3Ј-UTR to the CAT construct resulted in a marked (60%) decrease of the basal CAT activity (Fig. 5, A  and C, lanes 0). However, co-transfection of increasing amounts of A18 hnRNP with either construct generated similar levels of CAT activity (Fig. 5, A and C). These data suggest that A18 hnRNP can relieve the repressive effect of the RPA2 3Ј-UTR. FIG. 2. Use of immunotubes for protein-RNA interaction. A18 hnRNP or nucleolin was coated in immunotubes at a concentration of 40 g/ml. BSA and RNA binding buffer alone were used as controls. In vitro transcribed human immunodeficiency virus TAR RNA (1 ϫ 10 6 cpm) was incubated in the protein-coated immunotubes. Labeled RNA was eluted with increasing concentrations of NaCl and counted. Levels of RNA released from each tube are expressed in cpm. Proteinase K was used as the last step of elution. Myosin regulatory light chain (51) Cystein-glycine-rich protein CSRP2 S5 Laminin-binding protein (52) S6 (23) Human G10 homolog edg-2 (53) S8 Translational elongation factor 1␤2 (46) S11 Protein kinase C inhibitor I (54) S12 Putative C-myc-responsive (rcl) (55) S13 (22) ATPase (56) TAXREB107 c (57) a Chaperonin containing T-complex. b Translocase of inner mitochondrial membrane. c DNA-binding protein.
The overall effect of A18 hnRNP on the stimulation of the CAT activity is thus more pronounced when the RPA2 3Ј-UTR is present. This is better illustrated by the fold conversion of chloramphenicol measured with the two constructs (Fig. 5, B  and D). Although a less than 4-fold conversion is obtained with the CAT construct (Fig. 5B, 10 g of A18 hnRNP), the addition of A18 hnRNP (10 g) resulted in a more than 12-fold conversion in the presence of the RPA2 3Ј-UTR (panel D, CAT-UTR).
To determine whether the increased CAT activity was due to stabilization of the CAT mRNAs by A18 hnRNP, we performed RNase protection assays on these transcripts (Fig. 6). Our data indicate that, similarly to the CAT results (Fig. 5, A and C), the basal levels of the CAT-UTR transcript (Fig. 6B, lane 3) was much less than the basal level of the CAT transcript in the absence of the RPA2 3Ј-UTR (Fig. 6A, lane 3). The addition of increasing amount of A18 hnRNP enhanced the stability of the CAT transcript by less than 2-fold (Fig. 6A, lane 5), whereas it enhanced the stability of the CAT-UTR transcript by more than 5-fold (Fig. 6B, lane 5). The data presented in Figs. 5 and 6 indicate that A18 hnRNP can stimulate translation by stabilizing mRNA transcripts. Mechanisms other than mRNAs stabilization are also probably involved in the overall effect of A18 hnRNP on translation since A18 hnRNP can stimulate translation at a greater rate than it stabilizes mRNAs (Fig. 5, B-D,  versus Fig. 6).
A18 hnRNP Has a Protective Role-Our data show that A18 hnRNP binds specifically to the 3Ј-UTRs of RPA2 and thioredoxin (Figs. 3-4) and may influence their translation (Fig. 5). Both of these proteins have been associated with increased survival after stress. RPA protein can reverse the replication arrest induced by UV radiation (27), whereas expression of thioredoxin is associated with UV resistance (28). To evaluate the general effect of A18 hnRNP on cell survival after genotoxic stress, we first established a cell line expressing reduced levels of A18 hnRNP by stably transfecting an antisense vector for A18 hnRNP in RKO cells. After selection, the protein levels of A18 hnRNP were measured by Western blot. The data presented in Fig. 7A indicate that in cells transfected with an empty vector (RKO), A18 hnRNP was induced 4 h (lane 3) after exposure to UV radiation (20 J/m 2 ). The levels remain consistently high, up to 8 h after radiation (lane 4). This is in good agreement with early reports of the induction of A18 hnRNP mRNA by UV radiation (8). However, in the antisense cell line, the basal levels of A18 hnRNP were reduced (lanes 1 and 5), and no induction of the protein was detected even 8 h after radiation (lanes 5-8). Consistency of the protein loading was measured by analyzing the amount of the actin protein in all the samples (lanes 1-8). Similar results were also obtained with another A18 hnRNP antisense cell line (RKO AS-2, data not shown).
We then proceeded to evaluate the antisense cell line (RKO AS-1) capacity to survive UV radiation. Our data (Fig. 7B) indicate that the RKO AS-1 cells are much more sensitive to UV radiation. A difference in sensitivity can be observed at every dose, even with doses as low as 2 and 5 J/m 2 . An increase in sensitivity of almost 6-fold is achieved at 7 J/m 2 in the antisense cell line. We have also performed survival analyses with RKO cells stably transfected with an A18 hnRNP vector in the sense orientation. Our data (not shown) indicate that the cells were slightly more resistant than the parent cells, but the difference was not significant. Taken together these data suggest that the translocation of A18 hnRNP to the cytosol after UV radiation (Fig. 1) stimulates the translation of specific transcripts (Fig. 5) that increase cell survival after stress (Fig.  7). A protective role for A18 hnRNP in the genotoxic stress response seems likely. DISCUSSION To ascertain the role of A18 hnRNP in the genotoxic stress response, we first aimed to identify its potential mRNA targets. Several techniques exist to isolate mRNAs bound to a specific RBP, but most of them have a high degree of nonspecific mRNAs binding. We took advantage of the Nunc Immuno tube inert matrix to design a technique that would allow specific binding of mRNAs to A18 hnRNP. Our data (Fig. 2) indicate that less than 12% of the bound RNAs were nonspecifically bound to the matrix. Moreover, the fact that a large proportion of the isolated mRNAs are stress-activated indicates that the binding was not random. Furthermore, we also used this technique with nucleolin, another RBP, and obtained a totally different set of mRNAs. 2 This technique is thus highly specific and could conceivably be used for studying protein-DNA or protein-protein interactions.
Using Northwestern analyses (Figs. 3-4) we have confirmed the binding specificity of A18 hnRNP to two of the transcripts identified in our screen ( Table I). Regulation of RPA2 at the post-transcriptional or translational level is for the most part unexplored. Recent studies indicate that RPA2 is phosphorylated by the DNA-dependent protein kinase and probably the ataxia telangiectasia-mutated protein kinase in response to DNA damage (17). However, the functional significance of this phosphorylation is still unclear. It is believed that phosphorylation of RPA2 may alter the conformation of RPA heterotrimer and modulate its interaction with other proteins such as p53 (29). A18 hnRNP binds specifically to RPA2 3Ј-UTR (Fig. 3) and stimulates translation (Fig. 5). Translation initiation can be regulated by specific binding of proteins to the 3Ј-UTR of a transcript. It has been suggested that proteins bound to this region could contact the basal translation apparatus and influence translational activation or repression (2). Alternatively, as described in a recent review (4), RNA-binding proteins that shuttle between cellular compartments either constitutively or in response to stress may regulate the localization, translation, or turnover on mRNAs. Our data (Fig. 1) indicate that A18 hnRNP is translocated in response to UV radiation and increases mRNAs stability (Fig. 6). However, we cannot rule out the possibility that A18 hnRNP could also increase translation through direct interaction with the translational machinery. Recent evidence has shown (30) that interaction of the polyadenylate-binding protein with the eukaryotic initiation factor 4G, which interacts with the 5Ј cap binding initiation factor 4E (31), enhances translation. Furthermore, a human homologue of eukaryotic initiation factor 4G, PAIP-1, can also stimulate translation by bridging the polyadenylate-binding protein to the initiation factor 4A (32). These mechanisms explain how 2 C. Yang and F. Carrier, unpublished observation.

FIG. 5. Effect of A18 hnRNP on translation.
A, CAT assay. One g of the CAT reporter gene construct was cotransfected with 0, 5, 10 g of A18 hnRNP expression vector. CAT activity was measured with 25 g of cellular extracts for 1 h. The conversion of chloramphenicol was quantified with a PhosphorImager and expressed as either a percentage conversion of the total counts of [ 14 C]chloramphenicol or, in panel B, as a relative fold conversion compared with basal levels of (%) conversion (A, 0 g of A18 hnRNP). C, CAT assay was performed as in A, except that the assays were performed with the CAT-UTR construct, and 50 g of cellular extract were used for 3 h. D, fold conversion calculated as in B. regulation through the 3Ј-UTR of a transcript can influence translation initiation at the 5Ј end. Our data indicate that A18 hnRNP binding to mRNAs is reinforced by poly(A) tails (Figs. [3][4]. Hence, A18 hnRNP could potentially affect translation through similar mechanisms. As mentioned earlier, regulation of protein translation is apparently an important component of the cellular stress response (18). Modification of protein synthesis patterns was among the first phenomena observed after cellular stress (33). Typically, an immediate arrest followed by an increased rate of protein synthesis was observed after UV radiation (33). Downregulation of protein synthesis in response to stress is thought to be an adaptive response triggered to protect the cells and conserve the resources required to survive (34). On the other hand, induction of specific ribosomal proteins in response to stress may indicate the involvement of the translational machinery in the sensoring and response to cellular stress (20). The idea that ribosomes could be sensitive to stress is supported by the finding that bacterial strains lacking the ribosomal protein L11 can not activate the transcription factor sigma B in response to environmental stress (35). Moreover, the association of several ribosomal proteins with the oxidative stress response (36) is additional evidence that translation regulation is a significant component of the cellular stress response. Our data (Table I) indicate that several transcripts targeted by A18 hnRNP code for ribosomal proteins, some of which are stressresponsive. Regulation of ribosomal protein expression could thus represent an alternative mechanism by which A18 hnRNP may regulate the translation of a specific set of proteins in response to stress.
As RPA2, very little is known on the post-transcriptional or translational regulation of thioredoxin in response to stress.
Thioredoxin expression is increased in several tumors, specially gastric tumors (37), and in response to a variety of stress such as x-ray irradiation, UV radiation, and other types of oxidative stress (38). Increased levels of thioredoxin are associated with resistance to chemotherapy and cellular proliferation, probably through inhibition of apoptosis (26). In cultured human retinal epithelial cells, thioredoxin induction by prostaglandin E 1 and H 2 O 2 has been shown to be dependent on cAMP levels (39). However, the mechanism for the increased levels of endogenous thioredoxin in cancer cells is still unknown (26). Thioredoxin is a key regulator of cells signaling events that involve transcriptional processes such as induction of AP-1 activity and the inhibition of NF-B (26). Translocation of thioredoxin to the nucleus in response to UV radiation (40) and other types of stress may contribute to this activity. The cytoprotective effect of thioredoxin against oxidative stress was best illustrated in experiments where the levels of thioredoxin had been reduced with an antisense expression vector (28). These cells became more sensitive to H 2 O 2 , a variety of anticancer drugs, and UV radiation. Our data indicate that cells expressing reduced levels of A18 hnRNP are also more sensitive to UV light (Fig. 7). A18 hnRNP could thus contribute to increase cell survival by stimulating the translation of RPA2, thioredoxin, and other stress-activated transcripts such as ferritin (Table I). As mentioned earlier, ferritin is induced by UV light (41) and protects DNA against UV-induced damage (42).
Other RBP have been shown to translocate to the cytoplasm in response to stress (15). The uniqueness of A18 hnRNP is that it is also induced (Ref. 8 and Fig. 6) by UV radiation and targets stress-activated transcripts (Table I). Moreover, A18 hnRNP can stimulate translation (Fig. 5) and increase survival after genotoxic stress (Fig. 7). In the absence of stress, A18 hnRNP is FIG. 7. Survival of RKO cells after UV treatment. A, Western blot analysis of A18 hnRNP expression after UV radiation (20 J/m 2 ) in RKO cells and in RKO cells stably transfected with an A18 hnRNP antisense vector (RKO-AS1). Proteins were analyzed at different time intervals after UV radiation. Expression of actin was measured as a loading control. B, cells were treated with the indicated UV dose, and the colonies were counted 2 weeks later. Each point represents the average of three experiments. nuclear (Fig. 1) and does not affect cell proliferation (data not shown). The capacity of A18 hnRNP to stimulate translation and increase survival is apparently directly related to its translocation to the cytosol in response to stress. This translocation may allow A18 hnRNP to play a protective role by stabilizing the transcripts of genes involved in cell survival.