Amino Acid Starvation Induces the SNAT2 Neutral Amino Acid Transporter by a Mechanism That Involves Eukaryotic Initiation Factor 2α Phosphorylation and cap-independent Translation*

Nutritional stress caused by amino acid starvation involves a coordinated cellular response that includes the global decrease of protein synthesis and the increased production of cell defense proteins. Part of this response is the induction of transport system A for neutral amino acids that leads to the recovery of cell volume and amino acid levels once extracellular amino acid availability is restored. Hypertonic stress also increases system A activity as a mechanism to promote a rapid recovery of cell volume. Both a starvation-dependent and a hypertonic increase of system A transport activity are due to the induction of SNAT2, the ubiquitous member of SLC38 family. The molecular mechanisms underlying SNAT2 induction were investigated in tissue culture cells. We show that the increase in system A transport activity and SNAT2 mRNA levels upon amino acid starvation were blunted in cells with a mutant eIF2α that cannot be phosphorylated. In contrast, the induction of system A activity and SNAT2 mRNA levels by hypertonic stress were independent of eIF2α phosphorylation. The translational control of the SNAT2 mRNA during amino acid starvation was also investigated. It is shown that the 5′-untranslated region contains an internal ribosome entry site that is constitutively active in amino acid-fed and -deficient cells and in a cell-free system. We also show that amino acid starvation caused a 2.5-fold increase in mRNA and protein expression from a reporter construct containing both the SNAT2 intronic amino acid response element and the SNAT2-untranslated region. We conclude that the adaptive response of system A activity to amino acid starvation requires eukaryotic initiation factor 2α phosphorylation, increased gene transcription, and internal ribosome entry site-mediated translation. In contrast, the response to hypertonic stress does not involve eukaryotic initiation factor 2α phosphorylation, suggesting that SNAT2 expression can be modulated by specific signaling pathways in response to different stresses.

Nutritional stress caused by amino acid starvation involves a coordinated cellular response that includes the global decrease of protein synthesis and the increased production of cell defense proteins. Part of this response is the induction of transport system A for neutral amino acids that leads to the recovery of cell volume and amino acid levels once extracellular amino acid availability is restored. Hypertonic stress also increases system A activity as a mechanism to promote a rapid recovery of cell volume. Both a starvation-dependent and a hypertonic increase of system A transport activity are due to the induction of SNAT2, the ubiquitous member of SLC38 family. The molecular mechanisms underlying SNAT2 induction were investigated in tissue culture cells. We show that the increase in system A transport activity and SNAT2 mRNA levels upon amino acid starvation were blunted in cells with a mutant eIF2␣ that cannot be phosphorylated. In contrast, the induction of system A activity and SNAT2 mRNA levels by hypertonic stress were independent of eIF2␣ phosphorylation. The translational control of the SNAT2 mRNA during amino acid starvation was also investigated. It is shown that the 5-untranslated region contains an internal ribosome entry site that is constitutively active in amino acid-fed and -deficient cells and in a cell-free system. We also show that amino acid starvation caused a 2.5-fold increase in mRNA and protein expression from a reporter construct containing both the SNAT2 intronic amino acid response element and the SNAT2-untranslated region. We conclude that the adaptive response of system A activity to amino acid starvation requires eukaryotic initiation factor 2␣ phosphorylation, increased gene transcription, and internal ribosome entry site-mediated translation. In contrast, the response to hypertonic stress does not involve eukaryotic initiation factor 2␣ phosphorylation, suggesting that SNAT2 expression can be modulated by specific signaling pathways in response to different stresses.
The severe nutritional stress caused by amino acid starvation triggers several adaptive changes. One of these is the stimulation of transport sys-tem A for neutral amino acids. Adaptive regulation of system A has been described in most amino acid-starved animal cells (1)(2)(3). Work from several laboratories (4 -6) has demonstrated that this regulatory mechanism is associated with an increase in the mRNA for SNAT2, the ubiquitously expressed member of the sodium-coupled neutral amino acid transporter family (7). The induction of SNAT2 transcription leads to the increased synthesis of SNAT2 carriers, which are found in greater amounts on the membranes of starved cells (5,8). Recently, an amino acid-responsive element has been identified in the first intron of the mouse and human SNAT2 genes (9), supporting the hypothesis that transcriptional activation contributes to the adaptive regulation of system A.
Unlike transcriptional activation, the translation of SNAT2 during amino acid starvation has received little attention thus far. However, it is known that amino acid deprivation profoundly affects overall protein synthesis. Indeed, in amino acid-starved cells uncharged transfer RNAs activate the GCN2 pathway, a master regulator of the cellular response to nutritional stress from yeast to mammals (10), leading to the phosphorylation of the ␣ subunit of the eIF2 2 translation initiation factor (eIF2␣) on Ser-51. eIF2␣ phosphorylation causes reduced levels of ternary complexes (11), thus resulting in the inefficient translation of the majority of cellular mRNAs (12). In contrast, the translation of mRNAs that encode proteins that are essential for the cellular stress response is promoted (13,14). Among these proteins are the transcription factors activating transcription factor 4 (ATF4) and CCAAT/enhancer-binding protein ␤ (C/EBP␤), which are important in stress-induced gene expression (15)(16)(17)(18). Moreover, amino acid starvation causes the dephosphorylation of 4E-BP1 and eIF4E, thus lowering the activity of eIF4F (20). Because the vast majority of eukaryotic mRNAs are translated via the scanning mechanism, which requires the recognition of the 5Ј-end of the mRNA and its m-7G-cap by eIF4F (19), the dephosphorylation of 4E-BP1 and eIF4E is associated with a decreased translation efficiency of most mRNAs in amino acid-depleted cells (20). However, a selected pool of mRNAs, which encode either cell-defense proteins or proteins involved in apoptosis, are endowed with internal ribosome entry sites (IRES) and, thus, are efficiently translated via a cap-independent mechanism during amino acid starvation (21). This is the case for the mRNA that encodes the cationic amino acid transporter cat-1 (SLC7A1), which has an IRES in its 5Ј-UTR that mediates efficient translation of the cat-1 mRNA upon amino acid starvation (22). Interestingly, the preferential translation of cat-1 mRNA in amino acid-starved cells requires eIF2␣ phosphorylation (23). No information is yet available about the presence of IRES sequences in SNAT2 mRNA or the relationships between SNAT2 induction and eIF2␣ phosphorylation.
Amino acid starvation is not the sole stress that modifies SNAT2 transcription because the mRNA is also transiently induced by hypertonic treatment (24), leading to an enhanced expression of SNAT2 carriers (8). The resulting increase in the sodium-dependent transport of neutral amino acids is essential for the fast recovery of cell volume (25). The relationship between the response of system A activity to amino acid starvation and hypertonic stress has been widely debated and is not yet completely clarified (2,26). Our previous results have demonstrated that amino acid starvation causes significant cell shrinkage and induces osmo-responsive genes, suggesting that this experimental condition is also associated with hypertonic stress (27). However, the transduction pathways involved in the response to amino acid starvation and hypertonic stress of system A transport activity appear to differ significantly (27)(28)(29).
In this study we report on the mechanism of SNAT2 induction during stress conditions. We demonstrate that eIF2␣ phosphorylation is required for the increase of SNAT2 expression and the maintenance of cell viability during amino acid starvation but not hypertonic stress. We also provide evidence that the 5Ј-UTR of the SNAT2 mRNA, which has the potential to form a stable secondary structure, contains an IRES that has similar activity in amino acid-fed and -depleted cells. Thus, the expression of the system A transporter SNAT2 during amino acid starvation is the result of both transcriptional activation and IRES-mediated translation, which allows the efficient production of new carriers under conditions where protein synthesis is inhibited.

EXPERIMENTAL PROCEDURES
Cells, Cell Culture, and Cell Viability-HeLa and C6 cells were cultured in high glucose Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM glutamine, 100 units/ml penicillin, and 100 g/ml streptomycin. ATF4 Ϫ/Ϫ , eIF2␣-S/S (S/S), and eIF2␣-A/A (A/A) mouse embryonic fibroblasts were cultured in the same medium supplemented with both essential amino acid solution 50ϫ, Sigma and nonessential amino acids (MEM nonessential amino acid solution 100ϫ, Sigma). S/S and A/A cells, generous gifts from R. Kaufman (12), had been obtained from embryos expressing normal (S/S) or S51A (A/A) eIF2␣ at the homozygous state. The S51A mutation eliminates the phosphorylation site of eIF2␣.
Cells were starved for amino acids by culturing in Krebs-Ringer bicarbonate buffer with 10% dialyzed fetal bovine serum. Hypertonic treatment consisted of the incubation in high glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 100 mM sucrose for a final osmolality of 400 mosmol/kg (hypertonic DMEM).
Cell viability was assessed with resazurin (30). S/S and A/A cells were seeded in complete growth medium in 24-well plates at a density of 5 ϫ 10 4 cells/well and grown for 24 h before treatment. Growth medium was then substituted with Krebs-Ringer bicarbonate buffer, hypertonic DMEM, or complete medium with 1 mM dithiothreitol. After 24 h, the medium was replaced with fresh medium supplemented with 44 M resazurin (Sigma) for 2 h, and fluorescence was measured as described.
Western Blot Analysis-Western blot analysis of eIF2␣ and phospho-eIF2␣ were carried out using a mouse monoclonal antibody (Quality Controlled Biochemicals, Inc.) and rabbit polyclonal antibody (BIOSOURCE), respectively. 4E-BP1 was detected using an anti-PHAS-1 peptide antibody (Zymed Laboratories Inc.), whereas anti-ATF4 antibody was from Santa Cruz Biotechnology Inc. ␣-Tubulin antibody was from Sigma.
Real-time PCR was performed with iCycler (Bio-Rad). Primers for the endogenous mSNAT2 and GAPDH were the same as primers used in RT-PCR; exogenous m/SNAT2-2 and m/SNAT2-2in were 5Ј-GTCGACG-TATTTGACTGAGAGAAACCCACTG-3Ј (sense) at the 5Ј-end of the SNAT2 5Ј-UTR and 5Ј-GGGAGGTAGATGAGATGTGACG-3Ј (antisense) annealing 77 nt downstream of the LUC ATG. Each gene was assayed in triplicate with Sybr Green (Applied Biosystems); PCR reactions (25 l) contained 1.4 l of 1:60-diluted cDNA. GAPDH expression levels were used to normalize gene expression in each sample.
Amino Acid Transport Analysis-Cells were seeded in 24-well trays (Falcon) at a density of 5 ϫ 10 4 cells/cm 2 for S/S and A/A cells and cultured for 2 days. The activity of transport system A was evaluated by measuring the uptake of [ 14 C]methylaminoisobutyric acid (MeAIB, 0.1 mM, 2 Ci/ml) for 1 min at 37°C in 0.25 ml of Krebs-Ringer bicarbonate buffer. The incubation was terminated with two rapid washes in 3 ml of ice-cold 300 mM urea. Ethanol extracts of cells were added to 0.6 ml of scintillation fluid (Hisafe, PerkinElmer Life Sciences) and counted for radioactivity in a Wallac Microbeta Trilux counter (PerkinElmer). Cell monolayers were then dissolved with 0.5% sodium deoxycholate in 1 N NaOH, and protein content was determined using a modified Lowry procedure (27).
Expression Vectors-The 5Ј-end of the SNAT2 mRNA was amplified by polymerase chain reaction using 5Ј-SalI (5Ј-AAAGTCGACCGACGC-CGCCGCCTTAGAACGCCT-3Ј) and 3Ј-NcoI (5Ј-TTTCCATGGTA-AGCACTGGGAGGAATCGGGTGCAG or 5Ј-TTTCCATGGCGGC-CTTCTTCATGCTAAGCACTGGGAG) primers. Because the SNAT2 mRNA has two in-frame potential initiation ATGs with the second being in better consensus flanking sequences, we isolated both cDNAs and named them SNAT2-1 and SNAT2-2. These cDNAs were cloned into the SalI/NcoI site of the pSVCAT/BiP/LUC plasmid by replacing the IRES from the BiP mRNA. The resulting vectors were named b/SNAT2-1 and b/SNAT2-2, respectively. The plasmids b/hpSNAT2-1 and b/hpSNAT2-2 were constructed by replacing the BiP IRES within the pSVhpCAT/BiP/LUC vector at the NcoI/SalI sites. The pSVCAT/ ICS/LUC (b/ICS) vector contains 400 nt of irrelevant sequence and was used as a negative control for IRES-mediated translation. The b/SNAT2-2mut contained the same sequence as b/SNAT2-2, with the upstream open reading frame (uORF) ATG changed to TTG using the polymerase chain reaction-based mutagenesis.
The m/SNAT2-1 and m/SNAT2-2 expression vectors were generated by cloning the chimeric LUC cDNAs into the pUHD10 -3 expression vector (33) cleaved at the XbaI site 3Ј to the promoter region and the EcoRI site 5Ј to the polyadenylation signal. This vector contains a minimal cytomegalovirus promoter and the SV40 polyadenylation signal. SNAT2-1 (342 bp) and SNAT2-2 (360 bp) cDNAs were amplified by the polymerase chain reaction using 5Ј-EcoRI (5Ј-AAAGAATTCCGACGCCGCCGCCTTA-GAACGCCT) and 3Ј-NcoI (5Ј-TTTCCATGGCGGCCTTCTTCAT-GCTAAGCACTGGGAG-3Ј and 5Ј-TTTCCATGGTAAGCACTGG-GAGGAATCGGGTGCAG-3Ј, respectively) primers and were cloned into the EcoRI/NcoI site of the pUHD10 -3/LUC vector. We obtained the m/hpSNAT2-1 and m/hpSNAT2-2 vectors by inserting a hairpin in the SacII site. The SacII site is 70 nt downstream of the transcription start site. The hairpin was generated by annealing phosphorylated primers 5Ј-GGAAGCTTATCGATTTCGAACCCGGGGTACCG and 5Ј-AATTCGGTACCCCGGGTTCGAAATCGATAAGCTTCCGC. The m/SNAT2-2in vector was obtained by PCR amplification using adult human liver genomic DNA. The GC-rich PCR System (Roche Applied Science) with 1 M resolution solution was used. The PCR primers were EcoRI (5Ј-AAAGAATTCCGACGCCGCCGCCTTAGAACGCCT-3Ј) and NcoI (5Ј-TTTCCATGGCGGCCTTCTTCATGCTAAGCACT-GGGAG-3Ј). The PCR product was ligated into the m/SNAT2-2 vector at the EcoRI/NcoI site by replacing the SNAT2 5Ј-UTR. The final vector was sequenced to confirm the presence of the intronic sequence.
The LUC-P98 expression vector was obtained from M. W. Hentze (34). Monocistronic and bicistronic DNA vectors for in vitro transcription were generated by introducing the appropriate fragments into the corresponding sites of Bluescript KSϪ (Stratagene) and ligating it to LUC P98, which was obtained from the LUC-P98 vector. The LUC-P98 vector contains 98 A residues cloned at the 3Ј-end of the LUC ORF and upstream of a NotI site that was used to linearize the plasmids and generate RNA transcripts.
Cell Transfections and Analytical Procedures-Cells were seeded in 6-well plates (Falcon) at a density of 20 ϫ 10 4 cells/plate for HeLa, 50 ϫ 10 4 cells/plate for mouse embryonic fibroblast cells (MEFs), and 100 ϫ 10 4 cells/plate for C6 cells. The following day cells were transfected with 500 ng/well of bicistronic vectors or cotransfected with 100 ng/well each of monocistronic vectors and ␤-galactosidase expression vector. Cells were subjected to treatment and/or collected for enzymatic assays 48 h after transfection. LUC, chloramphenicol acetyltransferase (CAT), and ␤-galactosidase activities were measured as described previously (22). The activities were normalized to the protein content of the cell extracts, which was measured using the Bio-Rad D C assay.
Cell-free Translation Assays-DNAs were linearized by digesting with NotI, and capped RNAs were synthesized using the mMessage mMachine T3 kit (Ambion). The purity and integrity of the RNA molecules was checked on urea-polyacrylamide gels. In vitro translations were performed using the Promega nuclease-treated rabbit reticulocyte lysate (RRL) cell-free system, and the LUC and CAT enzymatic activities were determined. HeLa (S3) cell extracts were prepared as previously described (34). HeLa cell extracts were treated with 0.01 unit of micrococcal nuclease for 6 min just before the translation reactions. EGTA was added to a final concentration of 2 mM, and after a brief centrifugation at 14,000 ϫ g, the supernatant was used immediately for in vitro translation. Each in vitro translation reaction contained 6 -8 g/l extract, and the reaction was as previously described (34). Trans-lation reactions in the presence of [ 35 S]Met (15 mCi/ml) and analysis of the products on SDS-PAGE gels were as described previously (35).

Differential Requirement for eIF2␣ Phosphorylation of SNAT2 Induction by Amino Acid Starvation and Hypertonic Stress-
The system A family of amino acid transporters has three members: SNAT1, SNAT2, and SNAT4 (7). It has been shown that system A transport activity is enhanced by amino acid starvation via an increased abundance of SNAT2 mRNA (4). We hypothesized that induction of SNAT2 mRNA may be part of the global response of cells to nutritional stress and, therefore, involves eIF2␣ phosphorylation. To test this hypothesis we studied the accumulation of SNAT2 mRNA in MEF homozygous for a Ser-51 to Ala mutation in eIF2␣ (A/A) and wild type MEFs (S/S) during amino acid starvation and hypertonic stress. We found that amino acid starvation (6 h) increased SNAT2 mRNA levels by 5-fold in S/S cells and 3-fold in A/A cells, suggesting that the induction involves eIF2␣ phosphorylation (Fig. 1A). This conclusion was confirmed using real-time RT/PCR analysis of SNAT2 mRNA levels in fed-and amino acid-depleted cells (data not shown). To determine the specificity of SNAT2 behavior, we also tested the accumulation of asparagine synthase (AS) mRNA. Induction of AS mRNA levels requires the transcription factor ATF4, which is induced during amino acid starvation in an eIF2␣ phosphorylation-dependent manner (36). In agreement with those previous results, we observed the induction of AS mRNA levels in S/S cells but not in A/A cells (Fig. 1A). Moreover, AS induction was also completely suppressed in ATF4 Ϫ/Ϫ cells that, on the contrary, exhibit a significant induction of SNAT2 under amino acid free conditions (Fig. 1B, 3.9-and 3.5-fold induction in S/S and ATF4 Ϫ/Ϫ cells, respectively).
SNAT2 synthesis is also induced during hypertonic stress to promote recovery of cell volume by increasing the sodium-dependent transport of neutral amino acids (8,24). In contrast to what was observed for amino acid starvation, hypertonic stress caused a comparable induction of SNAT2 mRNA levels in S/S and A/A cells (Fig. 1C), suggesting that eIF2␣ phosphorylation is not required for this regulation. Hypertonic stress had no effect on AS mRNA levels (Fig. 1C). As a control for the two stress conditions, we tested induction of SNAT2 mRNA levels in HeLa cells, which showed the expected induction by both amino acid starvation and hypertonic stress (Fig. 1, A and C).
To investigate the relationship between the accumulation of SNAT2 mRNA and system A transport activity, we measured the uptake of MeAIB, a specific substrate for system A, in S/S and A/A cells. After 8 h of amino acid starvation, MeAIB uptake showed a 5-fold induction in S/S cells but only a 2-fold stimulation in A/A cells ( Fig. 2A). On the contrary, MeAIB transport was stimulated to a similar extent by hypertonic stress in S/S and A/A cells (Fig. 2B). These data suggest that the increased activity of system A upon amino acid starvation depends at least in part on eIF2␣ phosphorylation, whereas the hypertonic stimulation of the system is an eIF2␣-independent regulation.
To understand the reasons for the differences in system A amino acid transport upon hypertonic and starvation stresses, we tested if hypertonic stress induces eIF2␣ phosphorylation in MEF cells. The results ( Fig. 2C) indicate that, as expected, amino acid starvation caused an increase in eIF2␣ phosphorylation and promoted the expression of the eIF2␣ target, ATF4, in S/S but not in A/A cells. In contrast, hypertonic stress did not increase eIF2␣ phosphorylation in S/S cells. These data are, therefore, consistent with the conclusion that eIF2␣ phosphorylation is not required for the induction of SNAT2 mRNA by this stress. We also examined the phosphorylation of 4E-BP1 by the amino acidsensitive mammalian target of rapamycin (mTOR) kinase, a mechanism which inhibits the binding of this protein to eIF4E, thereby stimulating translation (37). Amino acid starvation caused the increase of underphosphorylated forms of 4E-BP1 only in S/S cells (note the transient increase in the lower under-phosphorylated bands). On the contrary, A/A cells depleted of amino acids did not show a dephosphorylation of 4E-BP1, suggesting a potential link of eIF2␣ phosphorylation and the mTOR pathway. A similar observation was previously reported in yeast (38). Hypertonic treatment caused a transient decrease in 4E-BP1 phosphorylation in both S/S and A/A cells, indicating that osmotic stress decreased mTOR activity (Fig. 2C), as described previously (39).
To further test the role of eIF2␣ phosphorylation in the response to different stress conditions, we tested the viability of S/S and A/A cells subjected to hypertonic stress, amino acid depletion, or endoplasmic reticulum (ER) stress. ER stress was induced by incubation with dithiothreitol, which also promotes eIF2␣ phosphorylation (40). The viability of S/S cells was not affected by any of these treatments, whereas only ϳ50% of the A/A cells survived amino acid starvation or ER stress (Fig.  2D), indicating the importance of eIF2␣ phosphorylation in the response to both stresses. In contrast, the viability of A/A cells was not affected by hypertonic stress, consistent with our finding that eIF2␣ is not involved in the cellular response to this condition.

SNAT2 Is the Only System A Transporter Induced by Amino Acid Starvation and Hypertonic Stress in S/S and A/A Cells-To ensure that
SNAT2 is the only contributor to increased system A amino acid transport activity during the two stress conditions in the S/S and A/A MEFs, we employed RT/PCR to assess the abundance of the mRNAs for the other two system A transporters, SNAT1 and -4. Consistent with the transport induction, SNAT2 mRNA levels were increased by either amino acid starvation or hypertonic stress (Fig.  3B). In contrast, neither SNAT1 no-4 mRNAs were visibly increased in S/S cells by either stress (Fig. 3B). Interestingly, although basal transport activity is comparable in the two cell types (see Fig. 2, A  and B), SNAT1 was expressed at a much higher level in A/A cells than in S/S cells, whereas SNAT4 was undetectable in A/A cells, and SNAT2 expression was comparable in the two cell types. The reasons for these differences are not known. Therefore, these data suggest that SNAT2 is the major regulated system A transporter in MEFs. This is in agreement with previous findings (41).
The SNAT2 5Ј-UTR Contains an Internal Ribosome Entry Site-The human SNAT2 5Ј-UTR has two in-frame ATGs separated by 15 nt that could be used as initiator codons. Use of these two codons will result in mRNAs with 5Ј-UTRs of 342 and 357 nt. These will be referred to as SNAT2-1 and SNAT2-2, respectively, in the studies that follow. Folding the 5Ј-UTR RNA using the Mfold program (42) predicts a very stable structure (⌬G ϭ Ϫ90 kcal/mol), suggesting that cap-dependent translation via the scanning mechanism will be inefficient. We, therefore, hypothesized that the SNAT2 5Ј-UTR may contain an IRES that can be used efficiently under conditions of amino acid starvation.
To test this hypothesis we constructed expression vectors that express monocistronic chimeric mRNAs containing the SNAT2 5Ј-UTR and a LUC reporter from a cytomegalovirus minimal promoter (Fig. 4A, m/SNAT2-1 and m/SNAT2-2). To distinguish between scanning-dependent and -independent translation initiation, constructs were generated with a stable RNA hairpin upstream of the SNAT2 UTRs (Fig. 4A,  m/hpSNAT2-1 and m/hpSNAT2-2). These vectors were transfected into cells derived from three species: HeLa (human), wild type MEF S/S (mouse), and C6 (rat). LUC activity was expressed from the constructs containing the SNAT2 UTR in all three cell lines (Fig. 4B). Moreover, the presence of the RNA hairpin upstream of the LUC open reading frame did not affect expression, suggesting that these RNAs are translated via an IRES. To support this conclusion, we analyzed constructs that contained a UTR of 100 nt of irrelevant sequence from the vector between the cytomegalovirus transcription start site and the cloning site (Fig. 4A, m/CON and  m/hpCON). In this case, LUC expression was largely inhibited by the hairpin (Fig. 4B), which is the expected result for cap-dependent translation initiation.
Although these results strongly suggest that the SNAT2 5Ј-UTR contains an IRES element, there are alternative explanations. To rule out the possibility that a cryptic promoter within the UTR allows the expression of an mRNA that does not contain the hairpin, we tested LUC expression from derivatives of the m/SNAT2-1 and m/SNAT2-2 vectors that  had deletions of the promoter. In all three cell lines there was little expression from constructs with the deleted promoters (Fig. 4C), demonstrating that the SNAT2 5Ј-UTR does not contain a cryptic pro-moter. To rule out the possibility that there is unexpected splicing or differences in stability that account for the observed LUC expression in Fig. 4B, the integrity of mRNAs from m/SNAT2-2 and m/hpSNAT2-2 FIGURE 5. The SNAT2 5-UTR functions as an IRES in a bicistronic vector. A, bicistronic constructs used in this study. b/BiP has the BiP 5Ј-UTR inserted between the CAT and LUC cistrons and is used as a positive control. b/ICS has an irrelevant sequence between the CAT and LUC cistrons and is used as a negative control. The hairpin (hp) inserted upstream of the CAT cistron is the same sequence that was used in the monocistronic vectors (Fig. 4). was tested in stably transfected wild-type MEF cells. Northern blot analysis shows that the mRNAs from the two vectors have similar sizes (Fig.  4D). LUC activities in these cells were 25 units/g of protein (for m/hpSNAT2-2) and 55 units/g of protein (for m/SNAT2-2), which is in agreement with the 2-fold difference in mRNA levels (Fig. 4D, lanes 2  and 3). The integrity of the constructs was confirmed by RT/PCR using primers within the 5Ј-end of the RNA transcript and the LUC cistron (Fig. 4E, lanes 3 and 4). We next determined if the RNA that is generated from the m/hpSNAT2-2 vector contains the hairpin. To determine this, we digested the cDNAs amplified from m/SNAT2-2-and m/hpSNAT2-2expressing cells (Fig. 4E, lanes 3 and 4) with KpnI, which cleaves the cDNA within the hairpin. It is shown that only the cDNA from m/hpSNAT2-2 was cleaved by KpnI, producing two fragments of the expected sizes (Fig. 4E, compare lanes 7 and 8). These findings rule out alternative explanations for the similar LUC activity expressed from SNAT2 constructs with or without hairpin and support the existence of an IRES in the SNAT2 5Ј-UTR.
The SNAT2 5Ј-UTR Functions as an IRES in a Bicistronic mRNA-A widely accepted approach to test IRES-mediated translation uses bicistronic expression vectors. We used a bicistronic mRNA expression vector that contains CAT as the first reporter and LUC as the second reporter (Fig. 5A). In these bicistronic mRNAs, the CAT cistron is translated in a cap-dependent manner, whereas the LUC cistron is translated only if the intercistronic region contains an IRES (43). The SNAT2 UTRs (b/SNAT2-1 and b/SNA2-2) or the BiP IRES (b/BiP) were placed in the intercistronic region. A bicistronic vector containing an irrelevant sequence was used as a control (b/ICS). Vectors containing a stable hairpin in front of the CAT reporter were also used to test the capindependent translation of the LUC cistron (Fig. 5A, hp). LUC activity was expressed from constructs containing SNAT2 sequences in all three cell lines (Fig. 5, B-D). Similar results were seen in constructs that contain the BiP IRES. In contrast, low levels of LUC expression were seen in the control construct (b/ICS), supporting the conclusion that the SNAT2 5Ј-UTR can function as an IRES. Moreover, the presence of the RNA hairpin decreased CAT expression but had no effect on LUC activity (Fig. 5, B-D), consistent with the cap-dependent translation of the CAT cistron and the cap-independent translation of the LUC cistron in the constructs that contain the SNAT2 UTR. The integrity of the bicistronic SNAT2-2 RNA was confirmed by Northern blot analysis (Fig. 4D,  lane 1) and RT/PCR using primers within the CAT and LUC cistrons (Fig. 5F). These data support our previous conclusion that the SNAT2 UTR contains an IRES. Because the human SNAT2 UTR contains a small uORF (Ϫ150 to Ϫ164) and because uORFs have been associated with regulation of IRES activity (35), we mutagenized the uORF ATG initiation codon within the SNAT2-2 UTR (Fig. 5A, b/SNAT2-2mut) and tested its activity in the bicistronic mRNA expression vector. The results (Fig. 5E) indicate that prevention of uORF translation had no effect on IRES-mediated translation of the LUC reporter, suggesting that the uORF is not needed for IRES activity. This conclusion is also supported by the lack of a similar uORF in the mouse and rat SNAT2 UTRs.
The SNAT2 IRES Is Constitutively Active in Fed and Starved Cells-The increased SNAT2-mediated system A activity during amino acid starvation can be due to increased SNAT2 mRNA levels, to induced IRES-activity, or to both mechanisms. To investigate these possibilities, we tested the effect of amino acid starvation on the LUC activities of the m/SNAT2-2 constructs in MEF S/S cells. LUC activity was not altered after 12 h of amino acid starvation for constructs with or without the RNA hairpin (Fig. 6A). In contrast, LUC activity of the m/CON vector decreased by 70% (Fig. 6B). The levels of the m/SNAT2-2 and m/hpSNAT2-2 mRNAs did not increase during amino acid starvation (data not shown). As a positive control, we analyzed a chimeric monocistronic mRNA under the control of the AS promoter, which is transcriptionally induced by amino acid starvation (44). As expected, amino acid starvation induced LUC expression from cells transfected with this construct (Fig. 6B). These data suggest that the SNAT2 IRES is constitutively active and similar in this regard to the BiP IRES (23). Moreover, these results suggest that the increased SNAT2-mediated system A amino acid transport during amino acid starvation is most likely due to increased SNAT2 mRNA levels combined with IRES-mediated translation. To provide evidence for this hypothesis, a construct endowed with both these features was tested. The SNAT gene contains an amino acid response element within the first intron that contributes to transcriptional induction during amino acid starvation (44). The first intron of the human SNAT2 gene was introduced into the 5Ј-UTR at its natural position (9) to generate the m/SNAT2-2in (Fig. 4A). Our hypothesis was that the amino acid response element within the intron will increase transcription from this vector resulting in increased mRNA levels and increased LUC expression. In agreement with our hypothesis, amino acid starvation caused a 2.7-fold increase in LUC activity from the m/SNAT2-2in but no change in expression from m/SNAT2-2 (Fig. 6C). This increase was due to increased m/SNAT2-2in mRNA levels (data not shown). The presence of the intron also increased LUC activity in the amino acid fed cells, a result that can be explained by the effect of the enhancer element on basal promoter activity (9) or the effect of splicing on gene expression (45). We, therefore, conclude that the increased SNAT2-mediated amino acid transport depends on increased SNAT2 mRNA levels and the SNAT2 IRES. The SNAT2 IRES Functions in a Cell-free System-The SNAT2 5Ј-UTR has a predicted RNA structure that is very stable (⌬G ϭ Ϫ90 kcal/mol) and should inhibit ribosome scanning in a cap-dependent translation system in vitro. To test this hypothesis we compared the translational efficiency of monocistronic-capped and -polyadenylated m/SNAT2/PA98 and m/CON/PA98 RNAs in rabbit reticulocyte and HeLa cell lysates (Fig. 7A). We first measured the dependence of LUC expression on mRNA concentration and determined that 10 ng/l is an appropriate concentration of m/SNAT2-1/PA98 RNA because it does not saturate the in vitro translation reaction (Fig. 7B). Next, we measured the translation of the various mRNAs in the presence of a capped CAT-PA36 RNA as an internal control. The m/SNAT2 RNAs were translated less efficiently than the m/CON RNA, which should be translated efficiently because it contains only 20 nt of 5Ј-UTR sequence ϳ5-fold difference for RRL and HeLa extracts, Fig. 7C). These data indicate that the 5Ј-UTR of the SNAT2 mRNA is inhibitory to ribosome scanning, a finding consistent with the properties of the predicted RNA structure.
To determine whether the SNAT2 UTR can function as an IRES in a cell-free system, we translated bicistronic SNAT2 RNAs in RRL and HeLa extracts. The LUC activities (Fig. 8A) demonstrated that both SNAT2 UTRs function as IRESs in a cell-free system. Moreover, the 5Ј half of the SNAT2 UTR is important in IRES activity because trunca-tions of this region caused a 50% reduction in LUC activity (Fig. 8B). However, the construct that contains residues Ϫ50 to ϩ16 retained IRES activity, suggesting that this sequence may be able to recruit ribosomes.
As a final test of the relative translation efficiencies of monocistronic and bicistronic mRNAs containing the SNAT2 UTR, we measured incorporation of [ 35 S]Met in the LUC and CAT proteins in RRL (Fig.  8C). LUC protein levels were higher with the m/CON than the m/SNAT2 mRNAs (Fig. 8C), in agreement with the results in Fig. 7C. Interestingly, LUC was translated from the monocistronic and bicistronic mRNAs at similar levels (Fig. 8C). These data confirm the conclusion that the SNAT2 UTR functions as an IRES during in vitro translation.

DISCUSSION
Marked stimulation of system A transport activity is observed upon either amino acid starvation or hypertonic stress. Because cell shrinkage occurs under both stress conditions, system A stimulation may be viewed as an adaptive response aimed at restoring cell volume. However, although the biochemical characteristics of system A transport have been extensively studied in a variety of cell types, the molecular mechanisms underlying transport stimulation have not been fully characterized. It was previously shown that the induction of system A amino acid transport by both stresses is associated to the increased abundance of SNAT2 mRNA, the system A transporter encoded by the SLC38A2 gene (4,9). The increased abundance of SNAT2 mRNA would lead to the enhanced synthesis of SNAT2 carriers, the amount of which has been found markedly augmented on the plasma membrane of starved cells (5,8). It was also found that an enhancer element in the first intron of the gene (9) causes transcriptional activation during amino acid starvation. This suggests that expression of the SNAT2 gene is also regulated at the level of transcription during this stress.
In this study we have extended our understanding of the regulation of SNAT2 expression by cellular stress with the following novel findings. (i) The induction of the SNAT2 mRNA and system A amino acid transport by amino acid starvation involves signaling via the phosphorylation of eIF2␣, which is the master regulator of this stress response (13,16). In contrast, the induction of SNAT2 mRNA levels by hypertonic stress does not involve eIF2␣ phosphorylation. (ii) Translation of the SNAT2 mRNA is mediated via an IRES that is located within the 5Ј-UTR of the SNAT2 mRNA. (iii) The SNAT2 IRES is constitutively active in fed and fasted cells. Therefore, the IRES supports translation of the SNAT2 mRNA during fasted conditions when cap-dependent translation decreases. (iv) The SNAT2 IRES can function in a cell-free system, and maximum activity requires sequences at the 5Ј-end of the leader. As a consequence, the increased expression of SNAT2 protein, observed in amino acid-starved cells (5,8), should be considered the result of both the transcriptional activation of SLC38A2 and the cap-independent translation of the mRNA, allowed by the IRES in the 5Ј-UTR.
The signaling pathway induced by eIF2␣ phosphorylation is essential for cell survival during amino acid deprivation (Fig. 2) as well as in other stress conditions (13). Interestingly, hypertonic stress did not induce eIF2␣ phosphorylation, and cell survival was independent of eIF2␣ phosphorylation under this condition. The differential response to the two stresses may indicate that eIF2␣ phosphorylation is associated with conditions that lead to cell death rapidly. On the other hand, hypertonic stress activates signaling pathways that, through the inhibition of mTOR, transiently adjust protein synthesis rates until recovery occurs. Our data are consistent with the idea that two distinct signaling pathways promote system A up-regulation upon osmotic cell shrinkage and amino acid limitation and that maximal stimulation is achieved when both pathways are activated. This hypothesis may explain the 5-fold increase in SNAT2 mRNA levels in amino acid-depleted cells (associated with both nutritional and osmotic stress) and the 3-fold increase caused by hypertonicity alone (associated with osmotic stress only).
In this regard, regulation of SNAT2 mRNA levels by amino acid starvation is clearly different from regulation of expression of other amino acid sensitive genes. Increased levels of mRNAs encoding AS, CHOP (C/EBP homologous protein), and cat-1 are induced by limitation of any single essential amino acid (44,46), which stimulates eIF2␣ phosphorylation (47). Therefore, transcription of these three genes can be stimulated by the eIF2␣ phosphorylation signaling pathway. In contrast, SNAT2 mRNA levels are not increased by amino acid limitation if any of its substrate amino acids are present at high concentrations (4). This is consistent with the complex regulation of SNAT2 gene expression observed in our study; eIF2␣ phosphorylation is only partially required for SNAT2 mRNA accumulation during amino acid starvation and is not required during hypertonic stress. On the contrary, AS induction is completely suppressed in the absence of eIF2␣ phosphorylation. This is consistent with the observation that AS and SNAT2 transcription also exhibit different dependence on ATF4, whose expression is increased by eIF2␣ phosphorylation. Indeed, although ATF4Ϫ/Ϫ cells are not competent for AS induction upon amino acid starvation, they still exhibit a significant stimulation of SNAT2 expression under the same conditions (Fig. 1B).
We have previously shown that amino acid starvation induces SNAT2 mRNA levels via a mechanism that involves activation of the extracellular signal-regulated protein kinase (ERK) 1/2 pathway (48). It is possible that ERK 1/2 activation causes a post-translational modification of a transcription factor. In fact, C/EBP␤ is a substrate of ERK 1/2 and is activated by this phosphorylation (49). Given the partial dependence of SNAT2 induction on eIF2␣ phosphorylation, these data suggest that there might be cross-talk between eIF2␣ phosphorylation and the ERK 1/2 signaling. eIF2␣ phosphorylation causes increased expression of several transcription factors, including ATF4 and C/EBP␤ (15,17,18), which in turn increases transcription of target genes (16). It is, therefore, possible that eIF2␣ phosphorylation induces expression of transcription factors that are further activated when they are phosphorylated by ERK 1/2. Moreover, induction of other transcription factors has been described under both stress conditions, such as TonEBP/ NFAT5 (27,50). Future studies will determine the role of cross-talk between these signaling pathways in activation of SNAT2 gene transcription.
It is shown here that the SNAT2 mRNA is translated via an IRES element in the 5Ј-UTR. Amino acid starvation causes a decrease in cap-dependent translation initiation, which affects the translation of most mRNAs (51). However, translation initiation of mRNAs that contain IRESs occurs independently of the 5Ј-cap. Prediction of the structure of the SNAT2 mRNA 5Ј-UTR (42) suggests that it forms a very stable RNA structure that would inhibit cap-dependent translation initiation. In agreement with this prediction, we show that mRNAs with the human SNAT2 UTR were translated much less efficiently than the control mRNA in vitro (Fig. 7). We found that the SNAT2 5Ј-UTR contains an IRES that supports cap-independent translation initiation both in cultured cells and in cell-free systems. This conclusion is based on the finding that similar levels of expression were observed from the m/SNAT2-2, m/hpSNAT2-2, and b/SNAT2-2 constructs (Fig. 4D). The LUC activities were very similar when they were normalized to the mRNA levels (data not shown). Interestingly, translation mediated by this IRES is nearly as efficient as cap-dependent translation from control mRNAs (Fig. 4B), suggesting that the SNAT2 IRES competes well with the cellular mRNAs.
The activity of the SNAT2 IRES did not increase during amino acid starvation, suggesting that it is not regulated by this condition. The BiP mRNA also contains a constitutive IRES (22). For both genes cellular stress increases mRNA expression, and the IRES allows mRNA translation and protein expression when cap-dependent translation initiation is inhibited. In contrast, the activity of other cellular IRESs is regulated, whereas viral IRESs are constitutive (for review, see Ref. 21). The fact that the SNAT2 IRES is active in both normal and stressed conditions emphasizes the importance in this protein for normal metabolism and survival during stress.
This study did not determine the RNA structure in the 5Ј-UTR that forms the SNAT2 IRES. However, in vitro translation of deletion mutants demonstrated that the 5Ј-end of the UTR is an important component of the IRES (Fig. 8). We also do not know the protein factors required for initiation from the SNAT2 IRES. Our current knowledge of IRES structure and function suggests that internal ribosome binding involves factors required for cap-dependent translation and novel proteins (21). All these factors work together to form an RNA structure that recruits the initiation codon of the mRNA to the P site of the ribosome (19). A mammalian mRNA that uses the prokaryotic-type mode of ribosome recruitment has also been described (52). The Gtx homeodomain mRNA contains a 9-nucleotide element that is believed to facilitate translation initiation by base pairing with the 18 S ribosomal RNA. The 5Ј-end of the SNAT2 UTR contributes to its activity, suggesting that recruitment of the ribosome could require a structural component.
Although the induction of system A and SNAT2 in response to various stress conditions has been discussed in several reviews (1,2,26,44), the physiological significance of this regulation is not yet fully understood. The results presented here suggest that the increase in system A activity is an integral part of the cell adaptation to amino acid starvation. This stress condition, encountered in severe ischemia as well as upon disruption of placental function, elicits complex adaptation mechanisms that also involve autophagy and proteasomal activity (53,54). Interestingly, low system A activity was associated with reduced fetal growth when placental function is insufficient (55). Therefore, the dissection of the molecular mechanisms involved in the adaptation to nutritional stress may constitute an important step for the understanding of conditions in which these responses are dysregulated.