Stress-induced Translation of ATF5 mRNA Is Regulated by the 5′-Untranslated Region*

Activating transcription factor (ATF) 5 is a transcription factor belonging to the ATF/cAMP-response element-binding protein gene family. We previously reported that ATF5 mRNA expression increased in response to amino acid limitation. The ATF5 gene allows transcription of mRNAs with at least two alternative 5′-untranslated regions (5′-UTRs), 5′-UTRα and 5′-UTRβ, derived from exon1α and exon1β. 5′-UTRα contains highly conserved sequences, in which the upstream open reading frames (uORFs) uORF1 and uORF2 are found in many species. This study was designed to investigate the potential role of 5′-UTRs in translational control. These 5′-UTRs differentially determined translation efficiency from mRNA. The presence of 5′-UTRα or 5′-UTRβ represses translation from the downstream ATF5 ORF. Moreover, 5′-UTRα-regulated translational repression is released by amino acid limitation or NaAsO2 exposure. This release was not seen for 5′-UTRβ. Mutation of uAUG2 in the uORF2 of 5′-UTRα restored the basal expression and abolished the positive regulation by amino acid limitation or arsenite exposure. We demonstrated that phosphorylation of eukaryotic initiation factor 2α was required for amino acid limitation-induced translational regulation of ATF5. Furthermore, arsenite exposure activated the exogenously expressed heme-regulated inhibitor kinase and induced the phosphorylation of eukaryotic initiation factor 2α in nonerythroid cells. These results suggest that translation of ATF5 is regulated by the alternative 5′-UTR region of its mRNA, and ATF5 may play a role in protecting cells from amino acid limitation or arsenite-induced oxidative stress.

element-binding protein (CREB)/ATF family that was first identified as a protein that binds to the lipopolysaccharide-response element (GPE-1) on the granulocyte colony-stimulating factor (CSF3) gene along with C/EBP␥ (1). It contains a DNAbinding and dimerization domain (bZIP domain) and regulates processes that are involved in cellular differentiation (2,3), the cell cycle (4), and apoptosis (5,6). ATF5 represses cAMP-induced transcription in cultured cells (4) and is shown to inhibit apoptosis (6). Angelastro et al. (2) demonstrated that ATF5 inhibits CRE-mediated expression of neural genes and neural differentiation. Cdc34 is the G 2 checkpoint gene, and ATF5 is a target of Cdc34-dependent ubiquitin-mediated proteolysis (4), expression of which is affected by the cell cycle. Recently, Monaco et al. (7) showed that ATF5 is widely expressed in carcinomas, and interference with its function caused apoptotic cell death of neoplastic breast cell lines. This suggests that ATF5 may be a target for cancer therapy and that studies of the mechanism by which ATF5 expression is regulated might be important in the investigation of treatments for cancer.
Mammalian cells have the ability to alter their gene expression to adapt to a variety of environmental stresses, including nutrient limitation, oxidative stress, and hypoxia, although the exact molecular events controlling the stress response have not been fully elucidated. Recently, we discovered that ATF5 is a stress response transcription factor that responds to amino acid limitation (8). We studied the effect of amino acid limitation on ATF5 mRNA levels in a mammalian cell line. Limitation of a single amino acid (glutamine, methionine, or leucine) increased ATF5 mRNA levels in HeLaS3 cells. This resulted at least in part from increased half-life of the ATF5 mRNA. Cycloheximide, an inhibitor of translation, suppresses the glutamine limitation-induced increase in ATF5 mRNA, indicating that this increase depends on de novo protein synthesis. Relatively few studies have been performed to characterize ATF5 and its biological functions. The amino acid sequence of ATF5 is closely related to that of ATF4, another member of the CREB/ATF transcription factor family. ATF4 expression is also affected by amino acid availability and regulates target genes in an amino acid-dependent manner (9 -12).
In mammals, four eukaryotic initiation factor (eIF) 2␣ kinases recognize distinct stress signals, phosphorylate eIF2␣, and inhibit protein synthesis via translational control (13). These eIF2␣ kinases are general control nonderepressible-2 (GCN2), which is activated by nutritional limitation, RNA-activated protein kinase-related ER kinase/pancreatic EIF2␣ kinase (PERK/PEK), which is activated by protein malfolding because of ER stress, double strand RNA-activated protein kinase, which is activated in response to viral infection, and heme-regulated inhibitor (HRI), which is activated by heme deficiency in the erythroid lineage. HRI is also activated in response to oxidative stresses in Schizosaccharomyces pombe (14) and MEF cells (15). Phosphorylation of eIF2␣ reduces the levels of eIF2-GTP available for initiation of translation and contributes to lowered global protein synthesis coincident with induced translational expression of genes that function in the stress response.
The levels of the transcription factor ATF4 are increased in response to phosphorylation of eIF2␣ during amino acid limitation, ER stress, or hypoxia. Induced ATF4 expression occurs predominantly via translational control (9, 11, 16 -18). Up-regulated ATF4 induces the expression of genes that are important for nutrient availability, the stress response, and apoptosis (9,11,16). The 5Ј-untranslated region (5Ј-UTR) of ATF4 mRNA contains two upstream open reading frames (uORFs) that are conserved among species and regulate translation. Under nonstressed conditions, low level eIF2␣ phosphorylation favors reinitiation of translation at uAUG2 by scanning ribosomes, which precludes translation of the ATF4 ORF. Under stressed conditions, highly phosphorylated eIF2␣ decreases ribosome assembly and favors re-initiation of translation at the ATF4 ORF located downstream of uORF2, resulting in ATF4 protein production.
The human ATF5 gene has at least two exon1s, exon1␣ and exon1␤ (19), and the gene is transcribed into at least two mRNAs encoding the same single 30-kDa protein. We designated them ATF5-R1 and ATF5-R2. These exon1s encode alternative 5Ј-UTRs of ATF5 mRNA, designated ATF5-5Ј-UTR␣ and ATF5-5Ј-UTR␤. Whether these alternative 5Ј-UTRs have different properties for the expression of the ATF5 protein remains to be elucidated. Alignment of the 5Ј-UTR␣ of human, mouse, and rat ATF5 mRNA reveals high identity and the presence of two putative translation start sites (uAUG1 and uAUG2). It also reveals mRNA with a similar two-uORF configuration (Fig. 1C). These observations led us to investigate the potential role of 5Ј-UTR in the translational control of ATF5 expression. In this study, we report that a uORF located in the ATF5-R1 mRNA inhibited the translation of ATF5 mRNA. This repression was released by stresses, including nutrient limitation and exposure to arsenite, via eIF2␣ phosphorylation. The 5Ј-UTR of ATF5-R2 mRNA inhibited translation of ATF5 mRNA, but amino acid limitation did not release the translational repression. These findings suggest that translation of ATF5 mRNA is differentially regulated by the alternative 5Ј-UTRs, and ATF5 induction via 5Ј-UTR␣ may be a protective response of cells that are deprived of amino acids or exposed to oxidative stress.
Human HRI cDNA was amplified by RT-PCR using total RNA from HepG2 cells as a template. The primers used were hHRI-F, 5Ј-CGCGAATTCTATGCAGGGGGGCAACTCC-3Ј, and hHRI-R, 5Ј-CCACTGGCGGCCGCTCATCCCACGC-CCCCATC-3Ј. The PCR product was digested with EcoRI and NotI. The resultant human HRI coding region of the cDNA was inserted into pcDNA3.1-FLAG vector (8), which was digested with EcoRI and NotI to produce pcDNA3.1-FLAG-hHRI. Expression plasmids for eIF2␣ S51A and eIF2␣ S51D were kindly provided by D. Ron (New York University, School of Medicine, New York).
Preparation of RNA and Quantification of Transcripts-Total RNA was isolated from cells using a GenElute Mammalian Total RNA Miniprep kit (Sigma). Five g of total RNA was treated with 3 units of DNase I (Nippon Gene, Toyama, Japan). Reverse transcription using 0.5 g of DNase I-treated total RNA and (dT) [12][13][14][15][16][17][18] primers (GE Healthcare) was performed with ReverTra Ace (Toyobo, Osaka, Japan) according to the manufacturer's instructions. The reverse-transcribed first strand cDNA was quantified with a real time quantitative PCR system, the ABI PRISM 7700 sequence detection system, or ABI PRISM 7000 sequence detection system, using a QuantiTect SYBR Green PCR kit (Qiagen, Valencia, CA) or SYBR Premix Ex Taq (Takara Bio, Otsu, Japan). The following oligonucleotides were used for the amplification of 80-, 147-, or 103-bp fragments of the cDNA corresponding to beetle luciferase, Renilla luciferase, or ␤-galactosidase: beetle luciferase primers, 5Ј-ACAAGGATGGATGGCTAC-3Ј and 5Ј-CTTCAGGCG-GTCAACGAT-3Ј; Renilla luciferase primers, 5Ј-ATGGGAT-GAATGGCCTGATA-3Ј and 5Ј-GCTGCAAATTCTTCTG-GTTCT-3Ј; and ␤-galactosidase primers, 5Ј-GCTGCATAAA-CCGACTACACAAA-3Ј and 5Ј-GCCGCACATCTGAACTT-CAG-3Ј. For the ABI PRISM 7700 sequence detection system, the samples were first heated at 50°C for 2 min and then at 95°C for 10 min and amplified during 40 cycles at 95°C for 15 s and 60°C for 1 min. For the ABI PRISM 7000 sequence detection system, the samples were first heated at 50°C for 2 min and then at 97°C for 10 min and amplified during 40 cycles at 95°C for 15 s and 60°C for 1 min. The relative luciferase mRNA level was obtained as the ratio of beetle luciferase mRNA to Renilla luciferase mRNA or beetle luciferase mRNA to ␤-galactosidase mRNA. All values are the mean Ϯ S.E. calculated from the results of three independent experiments.
Luciferase Assay-Cells were lysed with passive lysis buffer (Promega, Madison, WI). The cell lysates were used to determine luciferase activity with a dual-luciferase reporter assay system (Promega, Madison, WI) and Lumat LB 9501 (EG and G Berthold, Badwildbad, Germany). Luciferase activities were normalized with sea pansy luciferase activity. All values are the mean Ϯ S.E. calculated from the results of three independent experiments.
Western Blot Analysis-Cell extracts were prepared with lysis buffer containing 100 mM Tris-HCl, pH 6.8, 70 mM urea, 2% SDS, and 1ϫ Complete TM protease inhibitor mixture (Roche Applied Science). Protein concentration was determined using a BCA protein assay kit (Pierce). Proteins were separated by SDS-PAGE (7.5% for in-frame fusion proteins and 10% for the others) and electrophoretically transferred onto a Trans-Blot transfer membrane or Immun-Blot polyvinylidene difluoride membrane (Bio-Rad) using a wet transfer system or a semi-dry transfer system. Membranes were blocked for 1 h at room temperature with a 5% nonfat milk solution in TTBS buffer (25 mM Tris-HCl, pH 7.3, 150 mM NaCl, 0.1% Tween 20). The blots were then incubated with various antisera, including an antiluciferase antibody (1:1000), an anti-␤-galactosidase antibody (1:5000) (Promega, Madison, WI), an anti-FLAG antibody (1:5000) (Sigma), an anti-eIF2␣ antibody (1:1000), and an antiphospho-eIF2␣ antibody (1:1000) (Cell Signaling, Beverly, MA) in TTBS for 40 min at room temperature, then washed three times in TTBS, 2.5% nonfat milk, and incubated with an antigoat secondary antibody (1:3000), an anti-rabbit secondary antibody (1:3000), or an anti-mouse secondary antibody (1:3000) in TTBS, 2.5% nonfat milk for 20 min at room temperature. Horseradish peroxidase-coupled secondary antibodies were detected by ECL Western blotting detection regent (GE Healthcare) according to the manufacturer's instructions. Fig. 1B and Fig.  6A show the 81 and 65% identity between human cDNA and mouse cDNA for ATF5 exon1␣ and exon1␤ (19). The alignment of the 5Ј-UTR sequences from human, mouse, and rat ATF5-R1 transcripts reveals a high degree of similarity, and the presence of two additional translation start sites, uAUG1 and uAUG2. The 5Ј-leader sequence for human ATF5-R1 is 319 nucleotides in length and contains two uORFs (uORF1 and uORF2) preceding the ATF5 coding region. uORF1 and uORF2 have a Kozak match of A/G at Ϫ3 and G at ϩ4 of A(ϩ1)UG, and these uAUGs, as well as the ATF5 initiation codon, are found within a sequence context that allows initiation of translation by ribosomes (20). The alignment of the amino acid sequences for the uORFs in these species also shows a high degree of similarity (Fig. 1D). The second uAUG delineates an ORF overlap-ping 53 nucleotides with the ATF5 ORF. In contrast, the relative positions of the upstream AUG and uORFs from exon1␤ of human and mouse ATF5 showed no significant similarities. The conservation of these 5Ј-UTR␣ uORFs among species suggests that they are functionally important for the regulation of ATF5 expression. These observations led us to focus on the effects of the uAUGs in modulating the translation of the downstream coding sequence.

ATF5 5Ј-UTR␣ Contains Conserved uORFs-
ATF5 5Ј-UTR␣ Represses Translation of the Downstream Coding Sequence-We investigated whether ATF5 5Ј-UTR␣ can regulate translation of the downstream ORF. 5Ј-UTR␣ was cloned upstream of the luciferase reading frame, driven by the CMV promoter/enhancer in an expression vector (pCMV-hATF5-5Ј-UTR␣-LUC) ( Fig. 2A). The second uAUG delineates an ORF overlapping 80 nucleotides with the luciferase ORF. The construct was transfected into HeLaS3 cells. Fig. 2B shows that the luciferase activity was repressed when 5Ј-UTR␣ was present, whereas the luciferase mRNA level was not repressed. To test the generality of this phenomenon, we analyzed the effect of the 5Ј-UTR on the luciferase activity of COS7 cells. Translational suppression was also seen in COS7 cells. These results suggest that ATF5 translation is strongly repressed by the 5Ј-UTR␣ of the ATF5 tran- script, and this translational regulation may be a general phenomenon in different cell lines.

Mutations in Upstream AUGs
Abolish the Repressor Effect of the 5Ј-UTR␣-To determine the importance of the uAUGs of the 5Ј-UTR␣ in regulating ATF5 translation, we constructed a series of plasmids (derived from pCMV-hATF5-5Ј-UTR␣-LUC) containing a point mutation for each uAUG or both. After transfection of these constructions into both HeLaS3 and COS7 cells, cellular extracts were prepared and assayed for luciferase activity and luciferase mRNA level (Fig. 2B). Luciferase activity was up-regulated by mutation of the AUG within the second uAUG (5Ј-UTR␣mt2), whereas mutation of the first uAUG (5Ј-UTR␣mt1) did not result in up-regulation of luciferase activity. Compared with the wild-type leader sequence, mutation of the second uAUG led to increases by 17.5-fold (HeLaS3) or 23.5-fold (COS7) in luciferase activity. When both the first and the second uAUG were mutated (5Ј-UTR␣mt3), the repressor effect was also absent, because the luciferase activity was over the control level. These results show that the repression of luciferase activity because of the presence of the ATF5 5Ј-UTR␣ occurs at the translational level and that uAUG2 is strongly involved in this repression. This repression of luciferase expression was also demonstrated by Western blotting (Fig. 2C). Luciferase protein expression was down-regulated by the ATF5-5Ј-UTR␣, whereas the mRNA levels were similar among all constructs.
Translation Initiation at the Upstream ORF-The results described above suggest that the ATF5 uORF sequence is translated after initiation of the uAUG2 codon. To test directly the possibility that the uORF can be translated, we generated a construct in which the ATF5 uORF2 was fused in-frame to the luciferase ORF (pCMV-hATF5-5Ј-UTR␣-LUC-IF) (Fig. 3A) and transiently transfected into cultured cells. Transfected cell extracts were purified and analyzed for luciferase expression by Western blotting using a specific antibody that recognizes luciferase. As a control ␤-galactosidase expression plasmid, pSV-␤- The pellet was dissolved in passive lysis buffer for the measurement of luciferase activity, and 800 l was used for RNA extraction and quantification of luciferase mRNA. Relative luciferase activity and relative mRNA level were determined as the ratio of beetle luciferase to Renilla luciferase. Each value represents the mean Ϯ S.E. of three independent experiments. Results are given as fold induction defined as the ratio of the luciferase activity divided by the mRNA level of 5Ј-UTR⌬ relative to the others. C, luciferase (LUC) protein expression is down-regulated by ATF5 5Ј-UTR␣. One day before transfection, COS7 cells were plated on 60-mm dishes; 2 g of luciferase constructs were then transiently transfected. Two g of ␤-galactosidase (␤-gal) expression plasmid was co-transfected as an internal control, and its protein content was also detected by immunoblotting. Forty eight h after transfection, cells were harvested in 1 ml of phosphate-buffered saline-(Ϫ), then half was used for protein analysis by immunoblotting, and half was used for RNA extraction and quantification of luciferase mRNA. Relative luciferase mRNA levels were determined as the ratio of beetle luciferase to ␤-galactosidase. Each value represents the mean Ϯ S.E. of three independent experiments.

Stress-induced Translational Regulation by ATF5 5-UTR
FEBRUARY 1, 2008 • VOLUME 283 • NUMBER 5 gal was co-transfected, and ␤-galactosidase protein was also detected using a specific antibody. As shown in Fig. 3B (lane 2), uORF2-LUC fusion protein was expressed at a similar level to the luciferase protein expressed from the correct AUG. As expected, this protein has a molecular weight slightly higher than that of the luciferase protein. These data demonstrate that uORF2 is efficiently translated. We next addressed the possibility that the uORFs encode peptides that inhibit ATF5 translation of the downstream ORF in trans. The reporter plasmid pCMV-LUC or pCMV-hATF5-5Ј-UTR␣-LUC was transfected together with increasing amounts of an ATF5 uORF expression plasmid encoding the putative uORF peptides (pCMV-uORF) (data not shown). A control experiment was performed by co-transfecting the reporter vector with the expression plasmid pCMV-uORF-AUG(Ϫ) in which both uAUG1 and uAUG2 were mutated. Transfection of increasing amounts of the uORF expression vectors did not diminish the activity of luciferase encoded by the reporter plasmid. Although we could not directly assess the synthesis of the uORF-encoded peptide in the cells, the expression of the uORF-luciferase fusion construct (Fig. 3B) shows that translation is initiated at uAUG2. These results suggest that the uORF inhibits ATF5 expression via a cis-acting mechanism.

Stress-induced eIF2␣ Phosphorylation Is Required for ATF5 Translational Regulation-Harding et al.
To elucidate the mechanism by which the translation of ATF5 is induced, we investigated the translation start site shift from uORF to ATF5 ORF using an ATF5 5Ј-UTR␣-LUC inframe fusion construct (Fig. 3A). Fig. 5D shows that the translation initiation site was shifted from uAUG2 to the AUG of the luciferase ORF in response to NaAsO 2 exposure in COS7 cells (lanes 5 and 6). As shown in Fig. 5E, we confirmed that, when HRI protein was exogenously expressed in COS7 cells, NaAsO 2 exposures induced phosphorylation of HRI. From these findings, we conclude that eIF2␣ phosphorylation regulates the translation of ATF5 mRNA via its 5Ј-UTR in response to cellular stresses, and we suggest that in cells exposed to oxidative stresses, HRI is activated and induces eIF2␣ phosphorylation to up-regulate the translation of ATF5.
Alternative 5Ј-UTRs of Human ATF5 mRNA Have Different Translational Regulation Properties-The ATF5 gene has two exon1s, exon1␣ and exon1␤, which are transcribed into ATF5-R1 mRNA and ATF5-R2 mRNA, respectively. The sequences of exon1␣ and exon1␤ did not show any similarity. The identity between human and mouse was 81% for exon1␣ and 65% for exon1␤ (Fig. 6A). Human ATF5-R2 mRNA harbors one initiation codon (GGGatgG) that is in a good context for  translation, according to the Kozak rule. A putative uORF encoding 101 amino acids derived from the human ATF5-R2 transcript overlaps with the ATF5 ORF. The human ␤-specific 5Ј-UTR harbors two other uAUGs that are in a poor context for translation and are followed by a stop codon upstream of the initiation codon for ATF5. To test the regulatory function of the alternative 5Ј-UTRs, we analyzed the translational properties of 5Ј-UTR␤ in a transient transfection assay (Fig. 6, C and D). Human ATF5 5Ј-UTR␤ was cloned upstream of the luciferase reading frame, driven by the CMV in an expression vector (pCMV-hATF5␤-5Ј-UTR-LUC). The constructs pCMV-LUC, pCMV-hATF5-5Ј-UTR␣-LUC, and pCMV-hATF5-5Ј-UTR␤-LUC were transfected into HeLaS3 cells. The luciferase activity from the luciferase transcript containing 5Ј-UTR␤ reached only about 13% of the activity obtained from the transcript containing no UTR, which was comparable with that obtained with 5Ј-UTR␣.
To assess whether ATF5 5Ј-UTR␣ and -␤ are translationally regulated in a cell-specific manner, pCMV-hATF5-5Ј-UTR␣-LUC and pCMV-hATF5-5Ј-UTR␤-LUC were transfected into HeLaS3, HepG2, and U2OS cells, and luciferase mRNA levels and luciferase activities were determined. The luciferase activities from the luciferase expression plasmids containing no 5Ј-UTR, 5Ј-UTR␣, or 5Ј-UTR␤ were the same among these cell lines. We did not observe any cell-specific translational regulation, at least among these cell lines for 5Ј-UTR␣ and 5Ј-UTR␤.
Next we investigated whether 5Ј-UTR␤ can modulate translation in respond to amino acid limitation (Fig. 6D). HeLaS3 cells were transfected with the above three plasmids. The effect of amino acid limitation on luciferase activity and luciferase mRNA level was analyzed. As shown in Fig. 6D, although LUC activity from pCMV-hATF5-5Ј-UTR␤-LUC did not significantly respond to amino acid limitation (comparable with the control construct), luciferase activity from the 5Ј-UTR␣-LUC was increased 3.8fold by glutamine limitation and 11.4-fold by leucine limitation. These results show that ATF5 5Ј-UTR␤ has a repressor property for translation as does 5Ј-UTR␣, but this did not respond significantly to amino acid limitation. This suggests that these alternative 5Ј-UTRs have different roles in translational regulation induced by cellular stress.

DISCUSSION
UTRs have several roles in gene expression, including mRNA stability, mRNA localization, and translation efficiency (21). The turnover of an mRNA is mostly regulated by cis-acting elements located in the 3Ј-UTR, such as the AU-rich elements, which regulate mRNA decay in response to a variety of specific  . Inhibition of translation through 5-UTR␣ is released by cellular stress. A, inhibition of translation through 5Ј-UTR␣ is released by amino acid limitation. HeLaS3 cells were transiently transfected with ATF5 5Ј-UTR␣-luciferase constructs as shown in Fig. 2. A Renilla luciferase reporter construct was used as an internal control. Twenty-four h after transfection, cells were incubated in DMEM/F-12 medium containing 10% dialyzed FBS as control (C), or in medium lacking glutamine (GlnϪ) or leucine (LeuϪ) for 6 h. Relative luciferase activity and relative mRNA level were determined as described in Fig. 2B. Results are given as fold induction, defined as the ratio of the relative luciferase activity divided by the mRNA level of deprived cells to that of nondeprived cells. Each value represents the mean Ϯ S.E. of three independent experiments. B, inhibition of translation through 5Ј-UTR␣ is released by arsenite exposure. HeLaS3 or COS7 cells were transiently transfected with ATF5 5Ј-UTR-luciferase constructs. Twenty-four h after transfection, cells were incubated in serum-free DMEM as a control (C) or in serum-free medium containing 50 M NaAsO 2 (As) for 8 h. Luciferase assay and mRNA quantification were performed as in A. Each value represents the mean Ϯ S.E. of three independent experiments. Fold inductions are shown as the ratio of luciferase activity to mRNA level. FEBRUARY 1, 2008 • VOLUME 283 • NUMBER 5 FIGURE 5. Stress-induced eIF2␣ phosphorylation is required for ATF5 translational regulation. A, stress induces eIF2␣ phosphorylation. Upper, amino acid limitation induces eIF2␣ phosphorylation. HeLaS3 or COS7 cells were treated with amino acid limitation for 6 h. Total proteins were separated by SDS-PAGE. The extent of eIF2␣ phosphorylation was detected with anti-eIF2␣P antibody. Total eIF2␣ was determined with anti-eIF2␣ antibody. Lower, NaAsO 2 or CdCl 2 treatment induces eIF2␣ phosphorylation. HeLaS3 or COS7 cells were treated with NaAsO 2 or CdCl 2 at the indicated concentrations for 4 h. The extent of eIF2␣ phosphorylation was analyzed as described above. Cont, control. B, activation of ATF5 5Ј-UTR␣-mediated translation is regulated by eIF2␣ phosphorylation. One day before transfection, COS7 cells were plated on 60-mm dishes. The cells were then transiently co-transfected with 1 g of pCMV-5Ј-UTR␣-LUC alone or with expression plasmid for eIF2␣ S51A, or eIF2a S51D (1 g). 0.2 g of Renilla luciferase reporter construct was used as an internal control. Twenty-four h after transfection, relative luciferase activity and relative mRNA level were determined as described in Fig. 2B. Each value represents the mean Ϯ S.E. of three independent experiments. Results are given as fold induction, defined as the ratio of the relative luciferase activity divided by the mRNA level of cells transfected with empty vector to that of the others. C, phosphorylation of eIF2␣ is required for ATF5 translational regulation in amino acid-deprived cells. Left, wild-type or GCN2 Ϫ/Ϫ MEF cells were treated with leucine limitation for 6 h and the extent of eIF2␣ phosphorylation was detected with anti-eIF2␣P antibody. Total eIF2␣ was determined with anti-eIF2␣ antibody. Right, 1 day before transfection, wild-type or GCN2 Ϫ/Ϫ mutant MEF cells were plated on 60-mm dishes. The cells were then transiently transfected with 1 g of pCMV-5Ј-UTR␣-LUC, with 0.2 g of Renilla luciferase reporter construct as an internal control. Four h after transfection, cells were treated with leucine limitation for 8 h. Relative luciferase activity and relative mRNA level were determined as described in Fig. 2B. Each value represents the mean Ϯ S.E. of three independent experiments. Results are given as fold induction, defined as the ratio of the relative luciferase activity divided by the mRNA level of deprived cells to that of nondeprived wild-type MEF cells. D, translation initiation site is shifted from uAUG2 to the AUG of the luciferase (LUC) ORF. COS7 cells were transiently transfected with ATF5 5Ј-UTR␣-luciferase constructs shown in Fig. 3A (5Ј-UTR⌬, 5Ј-UTR␣, and 5Ј-UTR␣ in-frame fusion). Forty-eight h after transfection, cells were incubated in serum-free DMEM containing 50 M NaAsO 2 for 8 h. Whole-cell lysates were then prepared and analyzed for luciferase protein by immunoblotting as described in Fig. 3B. E, HRI is activated by NaAsO 2 treatment of COS7 cells. One day before transfection, COS7 cells were plated on 60-mm dishes. The cells were then transiently transfected with 2 g of FLAG-tagged HRI expression plasmid. Twenty-four h after transfection, cells were incubated in serum-free DMEM as a control or in serum-free DMEM containing 50 M NaAsO 2 for 4 h. Whole-cell lysates were then prepared and analyzed for HRI protein by immunoblotting. Two g of ␤-galactosidase expression plasmid was co-transfected as an internal control, and its protein was also detected by immunoblotting. signals, including cellular stress (22). UTRs also have a significant role in the localization of some mRNA by the mechanisms, active direct transport, local stabilization, or local entrapment of transcripts.

Stress-induced Translational Regulation by ATF5 5-UTR
It is known that less than 10% of eukaryotic mRNAs contain uAUG codons within the 5Ј-UTR region (23). However, uAUGs are common in certain groups of genes, such as oncogenes and many other genes related to cell growth, differentiation, and stress response. Some of the mechanisms by which UTRs regulate translation are beginning to be understood. CHOP, C/EBP␣, C/EBP␤, S-adenosylmethionine, and ␤ 2 -adrenergic receptor have uORFs, and these repress the basal translation of the ORFs (24 -27). The uORF peptide encoded in the UTR of CHOP and vigilin mRNA participates in repression of translation (24,28). However, so far translational up-regulation of these genes has not been reported. In other cases, several mechanisms can account for translational regulation via the 5Ј-UTR of the transcript: ribosome re-initiation, internal ribosome entry, and leaky scanning.
Stress that induces phosphorylation of eIF2␣ inhibits translation and attenuates global protein synthesis. However, eIF2␣ phosphorylation promotes the translation of mRNA for stressinduced genes. Translational control of the yeast GCN4 gene is the first and best studied example of selective translation by eIF2␣ phosphorylation (23). The GCN4 is a transcription factor that activates gene expression for amino acid biosynthesis. Four uORFs in the 5Ј-UTR have an important role in the translational control of GCN4 mRNA. When the levels of eIF2␣ phosphorylation are low, the ribosomes initiate at uORF1 followed by re-initiation at uORF2, -3, or -4, and the basal GCN4 protein expression is repressed. However, during amino acid limitation, elevated levels of eIF2␣ phosphorylation caused by eIF2␣ kinase, GCN2, allow the scanning ribosomes to bypass the uORF2, -3, or -4, and translate the GCN4 protein. These describe the sequences code for exon1␤, a part of the ATF5 5Ј-UTR␤ RNA sequences. The conserved nucleotide sequences are shown by asterisks. B, schematic representation of the human ATF5 5Ј-UTR␣ and ␤-luciferase constructs. Human ATF5 5Ј-UTR␤ was cloned into the upstream of the luciferase coding region to produce pCMV-hATF5-5Ј-UTR␤-LUC. C, ATF5 5Ј-UTR␤ represses translation of downstream coding sequences. One day before transfection, HeLaS3, HepG2, or U2OS cells were plated on 60-mm dishes and then 1.5 g of the luciferase construct shown in B was transiently transfected, with 125 ng of Renilla luciferase reporter construct as an internal control. Twenty-four h after transfection, cells were harvested and analyzed for luciferase activity and mRNA level as described in Fig. 2B. D, 5Ј-UTR␤ does not modulate translation in response to amino acid limitation. HeLaS3 cells were transiently transfected with ATF5 5Ј-UTR-luciferase constructs as described in Fig. 5C. Twenty-four h after transfection, cells were treated with amino acid deprivation and analyzed for luciferase activity and relative mRNA levels as described in Fig. 2B. Each value represents the mean Ϯ S.E. of three independent experiments.
The adaptive response to nutritional stress involves increased translation of CAT-1 mRNA via internal ribosome entry sites (IRES) within the 5Ј-UTR, and this induction requires the uORF in the 5Ј-UTR and eIF2␣ phosphorylation (29). Translation of the uORF of CAT-1 5Ј-UTR unfolds an inhibitory structure in the 5Ј-UTR. It is proposed that eIF2␣ phosphorylation induces the synthesis of an IRES trans-acting factor, which stabilizes the inducible IRES. This conformational change yields an active IRES that allows ribosomes to initiate translation at the CAT-1 ORF (30).
Ribosome re-initiation involving upstream ORFs also regulates ATF4 mRNA translation (17,31). The 5Ј-UTR of ATF4 mRNA has two conserved uORFs, uORF1 and uORF2. Scanning ribosomes that initiate translation from uAUG1 efficiently re-initiate translation at the downstream AUG, uAUG2. Under control conditions, low levels of eIF2␣ phosphorylation favor re-initiating ribosomes, directing them to initiate from uAUG2, which abrogates the translation of the ATF4 ORF. Under stressed conditions, high levels of eIF2␣ phosphorylation delay ribosome recognition of uAUG2 and favor re-initiation at the ATF4 initiation codon.
ATF5 belongs to the ATF4 subfamily of CREB/ATF transcription factors (32). The 5Ј-UTR regions of the ATF5-R1 mRNAs of human, mouse, and rat show high sequence similarity (Fig. 1) and the presence of two putative translation start sites (uAUG1 and uAUG2). Both the relative position and the sequence context of the uATGs of exon1␣, which is suitable for efficient translation, are well conserved. These findings prompted us to investigate the role of ATF5 5Ј-UTR. We found that the presence of 5Ј-UTR␣ severely inhibits the translation of downstream ORFs. 5Ј-UTR␣-dependent repression of translation is recovered by introducing a mutation in uAUG2 or exposure to stresses, including amino acid limitation and arsenite exposure. Amino acid limitation or arsenite exposure activates GCN2-or HRI-dependent eIF2␣ phosphorylation (13). As shown in Fig. 5, stress-induced eIF2␣ phosphorylation is required for ATF5 translational regulation, although the precise mechanism for the shift of the translation initiation site remains to be elucidated.
HRI has been shown to regulate the translation of blood cellspecific globin protein (33). Recently, McEwen et al. (15) reported that HRI functions in cells other than blood cells. In this study, an exogenously expressed HRI was activated in COS7 cells. Furthermore, we demonstrated that NaAsO 2 exposure induced eIF2␣ phosphorylation in both HeLaS3 and COS7 cells and caused a shift of translation initiation site from the uAUG to the correct ATF5 AUG site. These findings support the view that HRI may function in other types of cells besides blood cells in response to oxidative stress. ATF4 uAUG1 has a positive function for mRNA translation (31). When the uAUG1 of ATF5 5Ј-UTR␣ was mutated, the basal translational level was down-regulated by more than 50% that for wild-type 5Ј-UTR␣ (Fig. 2B). It is noteworthy that the mt1 construct enhances expression of uORF2-luciferase and diminishes expression of luciferase (Fig. 3B). These suggests that the uAUG1 of the ATF5 5Ј-UTR␣ may have the same positive function as the ATF4 uAUG1. There is no homology between the ATF5 5Ј-UTR␣ and ATF4 5Ј-UTR sequence, but their function in translation has been conserved during evolution.
Although sequence similarity is moderately conserved between human and mouse exon1␤ sequences, neither the relative position nor the sequence context of the uAUGs are conserved. From this, we speculated that exon1␤ does not have any regulatory function and cannot suppress the translation of downstream ORFs. However, surprisingly, 5Ј-UTR derived from exon1␤ also inhibits translation, as does exon1␣. This is supported by the in vitro observation that the presence of 5Ј-UTR derived from mouse exon1␤ hampered the translation of the ATF5 ORF (19).
Although the 5Ј-UTR from both exon1␣ and exon1␤ suppressed the translation of the ATF5 ORF, these two UTRs displayed differential properties in the stress response (Fig. 6). In contrast to 5Ј-UTR␣, amino acid limitation does not release 5Ј-UTR␤-dependent translational repression. In some cases, 5Ј-UTR regulates translation in a cell type-specific manner (34 -36). For example, the 5Ј-UTR of the branched-chain ␣-ketoacid dehydrogenase kinase transcript inhibits its translation in a cell type-specific manner (34). To test the cell type specificity of ATF5 5Ј-UTRs, we investigated 5Ј-UTR␣ and -␤ activity in various types of cells. Among these cells, we did not observe any cell type-specific translational regulation for either UTR. The physiological significance of the two types of 5Ј-UTR and their different properties remain to be elucidated.
We have reported that ATF5 mRNA expression was induced by amino acid limitation (8). In this study, we demonstrated that translational repression by 5Ј-UTR␣ is recovered by stresses, including nutrient limitation and exposure to arsenite. Recently, Sarraji et al. (37) showed that overexpression of ATF5 in HepG2 cells stimulates gene transcription via the nutrientsensing unit of the asparagine synthetase gene. These suggest that ATF5 may be a regulator of gene expression in the response to stresses, including amino acid limitation and oxidative stress.
Eukaryotic cells have evolved several proofreading mechanisms to degrade aberrantly processed or mutant mRNAs. Eukaryotic mRNAs containing premature termination codons are subject to accelerated turnover, known as nonsense-mediated decay (NMD) (38). Inhibitors of protein synthesis have been widely used to demonstrate the role of NMD in the degradation of mRNA transcripts (39). Cycloheximide treatment increases ATF5 mRNA levels (8), indicating that NMD may be involved in the stability of ATF5 mRNA. In this study, we have identified two putative uORFs in the 5Ј-UTR␣ of ATF5 mRNA, and studies are ongoing in our laboratory to determine whether NMD recognizes the premature stop codon derived from upstream AUG and regulates stress-responsive gene expression via ATF5 5Ј-UTR␣. The identification of this pathway provides a new area of research into stress responses in mammalian cells.
The molecular mechanism by which dietary protein intake and oxidative stress affect gene regulation is important for the regulation of physiological functions of mammalian cells. Mammalian cells adapt to amino acid limitation or oxidative stress by regulating numerous genes. Exploring the mechanisms that regulate gene expression under stresses, including amino acid limitation and oxidative stress, will increase our understanding of metabolic regulation and stress responses in mammalian cells.
Because ATF5 is a transcription factor, many questions arise. What are the target genes of ATF5? Does ATF5 regulate the same or a different set of genes as ATF4? Although the physiological role of ATF5 gene regulation in response to amino acid limitation and oxidative stress is not yet understood, ATF5 may be a key transcription factor regulating gene expression with important functions in response to amino acid limitation and oxidative stress in relation to physiological functions and diseases.