Role of sequences within the first intron in the regulation of expression of eukaryotic initiation factor 2 alpha.

Resting human peripheral blood T cells synthesize proteins at very low rates and contain very low levels of eukaryotic initiation factor (eIF) 2 alpha mRNA. During mitogenic activation, the level of eIF-2 alpha mRNA increases at least 50-fold, an effect thought to be due primarily to intranuclear stabilization of the primary transcript (Cohen, R. B., Boal, T. R., and Safer, B. (1990) EMBO J. 9, 3831-3837). Analysis of sequences within the first intron revealed a region with homology to the "initiator" (Inr) sequence first described by Smale and Baltimore (Smale, S. T., and Baltimore, D. (1989) Cell 57, 103-113). This Inr element is positioned 450 bases downstream of the eIF-2 alpha promoter and is oriented to generate an overlapping antisense transcript. Deletion or mutation of the Inr element results in a reproducible 5-8-fold increase in the activity of an eIF-2 alpha promoter-driven CAT reporter gene and a corresponding 2.5-fold decrease in activity of an antisense driven luciferase reporter gene in vivo in 293 cells. In vitro transcription analysis also reveals antisense transcripts which depend on an intact Inr element and whose 5' ends map to sequences surrounding the Inr consensus sequence. A potential role for double-stranded RNA generated by these overlapping divergent transcription units in the regulation of eIF-2 alpha gene expression in T cells is suggested.

antisense transcripts which depend on an intact Inr element and whose 5' ends map to sequences surrounding the Inr consensus sequence. A potential role for double-stranded RNA generated by these overlapping divergent transcription units in the regulation of eIF- 2a

gene expression in T cells is suggested.
Resting (Go) T cells are metabolically quiescent and have undetectable levels of DNA synthesis and low levels of RNA and protein production (reviewed in Refs. 1 and 2). A rapid increase in protein synthesis is an early event during activation of T cells following antigenic stimulation or by mitogenic lectins such as phytohemagglutinin (2). Single ribosomes are recruited onto mRNA to form polyribosomes, and increased translational activity is seen within the first few hours of mitogenic activation (3,4). The rate of protein synthesis increases during the first 24-72 h of activation and results in levels of protein synthesis that are at least 10-fold greater than the rate in Go T lymphocytes (5, 6). This rapid increase in protein synthesis following mitogenic stimulation of quiescent human T cells occurs at the level of translation initiation * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$To whom correspondence should be addressed FDA, CBER, Division of Hematology, Laboratory of Hemostasis and Thrombosis, Bldg. 29, Rm. 308, Bethesda, MD 20892. Tel.: 301-496-4833; Fax: 301-402-2780. (4,7).' Although eIF-2a2 mRNA is very well translated in both quiescent and activated T cells (8,9), human Go lymphocytes contain very low levels of eIF-2a message (10). During the first 24 h of activation, however, eIF-2a mRNA increases more than 50-fold. Neither changes in the rate of transcription of the eIF-2a gene nor changes in the half-life of the message appear to account for all of this rapid and large increase in eIF-2a mRNA. We have therefore hypothesized that stabilization of the nuclear precursor sufficient to allow processing and transport to the cytoplasm might account for the large increase seen with mitogenic stimulation (10).
Previous work in this laboratory has revealed a DNase Ihypersensitive site in chromatin mapping to the 5' portion of the first intron of the gene encoding eIF-2a between +220 and +300 (11). Analysis of the sequence of this region revealed an element with perfect homology to the conserved sequence of the initiator (Inr) element first described by Smale and Baltimore (12,13) as well as two TATA-like sequences. The Inr sequence (+448, see Fig. 3) and one of the TATA-like elements (TACAAT, +353) are oriented opposite the direction of transcription for eIF-2a. The second TATA-like element (TACAATAT, +361) is oriented in the same direction as eIF-2a. The presence of an Inr element oriented to generate a transcript opposite in direction from the eIF-2a transcript suggested that regulation of eIF-2a expression by an overlapping antisense transcript may account for the apparent change in stability of the sense primary transcript during mitogenic activation of T cells.
The specific aim of this work was to characterize in a preliminary fashion the effect of mutations to this region in the first intron of the eIF-2a gene. We find that elimination or mutation of the consensus Inr element results in a reproducible 5-&fold increase in the activity of a reporter gene in uiuo, both enzymatically and by Northern analysis, and a corresponding 2.5-fold decrease in activity of an antisense driven luciferase reporter gene. We also identify by in vitro transcription and primer extension analysis antisense transcripts whose start sites map to the Inr region in the eIF-2a gene and overlap the eIF-2a sense transcript for at least 450 bases.

EXPERIMENTAL PROCEDURES
Construction of CAT Expression Vectors and Templates for in Vitro Transcription-The plasmid vector pCAP2CAT and the sequence of the eIF-Pa gene promoter have been described (11,14). The BamHI-EcoRI (-806 to +478) fragment of the eIF-2a promoter was cloned T. R. Boal, manuscript in preparation. The abbreviations used are: eIF, eukaryotic initiation factor; Inr, initiator; CAT, chloramphenicol acetyltransferase; HEPES, 4-(2-hy-droxyethy1)-1-piperazineethanesulfonic acid; MOPS, 4-morpholinepropanesulfonic acid. into the HindIII-SmaI sites of pCAP2CAT to generate pCAP2eIF-2a806CAT. The plasmid pCAP2eIF-2a806AInr was constructed by site-directed mutagenesis of the four core bases of the consensus initiator element ((13) see Fig. 3). The pCAP2eIF-Sa -806/+264 CAT construct was made by cloning the BamHI-NaeI (-806 to +264) portion of the eIF-2a promoter into the HindIII-SmuI sites of pCAP2CAT. The plasmid pCAP2eIF-2a806AFP construct was made by cloning complementary oligonucleotides containing a three-base mutation in a region downstream (+465 to +467) of the Inr sequence into pCAP2eIF-2a-806/+478CAT digested with EspI and EcoRI. Plasmid pa(+1093/+179)Luc was generated by cloning the HindIII-AvaI (+lo93 to +179) fragment of the eIF-2a gene modified with HindIII linkers into the HindIII site of a luciferase cassette kindly provided by Dr. T. Shimada (NHLBI, NIH). Plasmid pa(+1093/+179 del 1nr)Luc was generated by digesting pa(+1093/+179)Luc with EspI and mung bean nuclease to remove three bases in the core of the Inr element. All constructs were confirmed by dideoxy sequencing. The luciferase negative control plasmid was kindly provided by Dr. K. Yoshimura (NHLBI, NIH). The templates for in vitro transcription were generated by removing the first 250 bases of CAT coding sequence to the EcoRI site. The resulting plasmids, containing unique HindIII and EcoRI sites, were digested with HindIII and EcoRI to release 1.3-kb templates capable of supporting in vitro transcription in both the sense and antisense orientations.
Transient Transfections of CAT Constructs into 293 Cells-293 cells, an adenovirus 5 transformed human embryonic kidney cell line that expresses the Ad5 E1A protein, were used for all transient transfections. DNA-mediated gene transfer was performed using the calcium-phosphate coprecipitation method (15,16). Cells were seeded at 5 X 105/plate. For transient transfections, a mixture of the test CAT hybrid gene (5-pg equivalent, 1.6 pmol), the transfection control plasmid pRSV luciferase (0.1 pg), and carrier DNA (PUC 8, 5 pg) was precipitated in HEPES-buffered saline, pH 7.05, and added to the plates. After 6 h, a 3-min glycerol shock was performed followed by two washes with ice-cold phosphate-buffered saline. Fresh medium was then added, and the incubation continued for another 42 h. Equimolar amounts of test plasmid were used. DNA concentrations were measured by fluorometry (Hoefer/Hoechst) and concentrations confirmed by gel electrophoresis. The calcium-phosphate method was also used for transfections in which luciferase-containing constructs FIG. 1. Relative CAT activities of eIF-2aCAT hybrid constructs. Plasmids used for transient transfection assays are shown. p designates the transcription start site. A total of 10 pg of DNA (CAT constructs plus carrier PUC 8 and transfection control pRSV-luciferase) was used. For all constructs, 1.6 pmol (5-pg equivalent) of test DNA was used. 100 ng of the transfection control plasmid, pRSV-luciferase, was included in all transfections. DNA-mediated gene transfer, CAT assays, and luciferase assays were performed as described (15)(16)(17)(18)(19)(20). CAT data are shown normalized to the value for pCAP2eIF-2a-806/ +478CAT which was assigned a value of 1.0. H, HindIII; E, EcoRI; B, BamHI; P, PstI; K, KpnI; A P , ampicillin resistance gene; Cm', chloramphenicol acetyltransferase coding sequence. CONSTRUCT PLASMID were tested. Cells were seeded at 5 X 106/plate and transfected with a total of 15 pg of DNA including 10 pg of salmon sperm DNA as carrier and equimolar amounts of the test plasmids. Six hours after transfection, fresh medium was added, and the incubation continued for another 34 h. Cells were harvested at 48 h, and CAT and luciferase activities were analyzed as described previously (17)(18)(19)(20).
Northern Blot Analysis-Total cellular RNA was prepared from transiently transfected 293 cells using a modification of the Chirgwin method (RNAzol, Cinna Biotecz). A total of 10-15 pg of RNA from each sample was heated to 55 "C in 50% formamide, 20% formaldehyde in 0.02 M MOPS buffer and electrophoresed in a 0.8% agarose gel containing 0.02 M MOPS. RNA was capillary-blotted to Nytran membrane. Similarity of RNA amounts loaded and efficiency of transfer were assessed by ethidium bromide staining of the gel. A single-stranded DNA probe was generated by primer extension from an M13mp19 construct containing the first 250 bases of the CAT coding sequence. RNA blots were hybridized as described by Church and Gilbert (21,22) for 36 h and washed as described.
In Vitro Transcription and Primer Extension Analysis-In vitro transcription analysis of the eIF-2a region was performed in a total of 25 pl containing 60 mM KCl, 20 mM (NH4)2S04, 6 mM MgC12,O.l mM EDTA, 1 mM dithiothreitol, 5 mM creatine phosphate, 10% glycerol, and 12 pl of K562 nuclear extract (50 pg). The DNA template concentration was 0.1 pmo1/25 pl of reaction mixture, and ATP, GTP, and UTP were 100 p~. 5 pCi of [cx-~'P]GTP (800 Ci/mmol) were added per reaction, and incubations were for 30 min at 28 "C. Processing of the transcription products was performed as described previously (23).
Primer extension analysis was performed as previously described (14).

RESULTS AND DISCUSSION
We evaluated the effect of the putative Inr element on the relative promoter strength of the eIF-2a gene by in vivo reporter gene assays and by in vitro transcription. We first demonstrated that the transcription initiation sites for the CAT constructs correlated well with the initiation sites mapped for the endogenous gene (data not shown). Initial studies using a construct truncated at +264 showed that ACGCTGAGCACT r -806 A C G C &~€ C A C T

B, "P-labeled run-off transcripts of templates A-E were analyzed by
denaturing polyacrylamide gel electrophoresis and autoradiography. Arrowheads identify the major in vitro transcription products (lanes A-E). Transcription of the adenovirus 2 major late promoter linearized by PstI, included as a positive control, generates a 536-base runoff transcript. A BstNI digest of SV40 was used to calibrate transcript size. C, in vitro transcription analysis was performed as in B in the presence of a-amanitin, 5 ng/ml. bp, base pair(s). removal of the 200 bases between +264 and +478 resulted in a reproducible 4-10-fold increase in eIF-2a promoter activity over constructs containing the downstream element (Fig. 1). Because the CAT constructs contain exon 1 and portions of intron 1 in the untranslated leader of the CAT message, we also analyzed the relative promoter activities of both the wild type and deletion constructs by Northern blot hybridization. By Northern analysis, a single-stranded antisense probe complementary to the first 250 bases of the CAT coding sequence hybridized to an appropriately sized 1.5-kilobase transcript. Deletion of sequences between +264 and +478 resulted in a shortened transcript with approximately 5-fold greater hybridization signal than the wild type construct, confirming that the change in CAT activity was not due to changes in translatability of the CAT message (data not shown).
We next looked at the contribution of the downstream sequence to the relative strength of the eIF-2a promoter by specifically altering the four central bases of the consensus sequence. Fig. 1 is representative of several experiments performed and shows that alteration of the core four bases resulted in an approximately 7-fold increase in relative CAT activity. The mutation of the Inr element reproducibly resulted in greater chloramphenicol conversion than did the deletion of 200 bases between +264 and +478. Alteration of bases outside the consensus sequence but within an in vitro DNase I footprint extending from +459 to +4743 resulted in a reproducible 2-fold decrease in CAT activity. The data indicate that a downstream Inr element can modulate the expression of eIF-Pa.
M. Noguchi, manuscript in preparation.  analyzed on a 6% sequencing gel calibrated using a dideoxy sequencing ladder of human eIF-2a. The 1.5-kilobase SstI-EcoRI genomic fragment containing the first intron-exon junction and the eIF-2a promoter region was used with the same primer. B, six major transcription products indicated by arrows in panel A were mapped to the region of the Inr. The Inr element is identified by dots below the conserved bases. 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2   FIG. 4. I n vitro transcriptional analysis of wild type and mutant eIF-Sa templates. In vitro transcription analysis of the wild type ( w t ) and mutant templates was performed as described for Fig. 2. Plasmids containing the eIF-2a promoter and sequences to +478 in intron 1 were digested with EcoRI and Hind111 to release the 1.3-kilobase templates extending from -806 to +478. The 800-base pair XhoI-PstI fragment of the adenovirus 2 major late promoter was transcribed as a positive control. Lanes 2 and 6 contained a-amanitin (5 ng/ml). The size of the major run-off transcript from the wild type template is approximately 1250 nucleotides. A total of 20 pg of DNA (luciferase constructs plus carrier salmon sperm DNA) was used. For all constructs except pLuc0, 1.2 pmol of test DNA was used. DNA-mediated gene transfer and luciferase assays were performed as described (15,16,(18)(19)(20). Luciferase data are shown normalized to the value for pa(+1093/+179)Luc which was assigned a value of 1.0. The sequence surrounding the Inr is shown. Conserved bases of the Inr promoter are indicated by dots, and the internal deletion of bases within the Inr is indicated by dashes.  Fig. 2. Transcription of the eIF-2a gene truncated at +478 was expected to yield a 478-nucleotide transcript whose abundance might reflect progressive deletion of promoter/enhancer/silencer elements. However, the major transcript from each template varied in size rather than in intensity. The size of the major transcript from each template was always 40-60 nucleotides less than that of the template and was shortened in proportion to the extent of the 5' deletion. This transcript overlapped the exon l/intron 1 boundary of eIF-2a and was a-amanitin-sensitive (Fig. 2, panel C ) . These results strongly suggested the presence of an antisense transcription unit in the 3' portion of the in vitro eIF-2a transcription unit.

Ad2-
The 5' end of the in vitro antisense transcript was mapped by primer extension. Six sites were identified within a 40base region, and two of the start sites mapped to the consensus Inr element (Fig. 3). Mutation of the four core bases in the Inr element eliminated the in vitro transcript seen with the wild type construct (Fig. 4). In contrast to the in vivo CAT data, however, we were unable to see a sense transcript under the conditions used here even though the in vitro antisense transcript had been eliminated. In addition, mutation of three bases immediately downstream of the Inr element resulted in an approximately %fold decrease in the in vitro antisense transcript, a result which differs from our in vivo CAT data.
We next evaluated the ability of the region spanning the Inr element to drive a luciferase reporter gene when oriented in the antisense direction. Fig. 5 demonstrates that the sequences between +lo93 and +179 are capable of generating moderate luciferase activity when compared with the promoterless vector. Deletion of the core three bases of the Inr element resulted in a reproducible 2.5-fold decrease in luciferase activity.
We have previously demonstrated that the primary transcript of the gene encoding eIF-2a is apparently destroyed in the nucleus when it is not needed and that during mitogenic stimulation of peripheral blood T cells, this transcript is stabilized long enough to be processed and transported to the cytoplasm (10). Regulation of the primary sense transcript by an overlapping antisense transcript is one mechanism by which such control might occur and has been proposed for a number of eukaryotic RNA polymerase I1 transcripts including c-erbAa, Xenopus basic fibroblast growth factor, and N- myc (24)(25)(26). Indeed, preliminary ribonuclease protection experiments in this laboratory have demonstrated the presence of a stable but weak antisense transcript mapping to this portion of the first intron in vivo in both quiescent and activated T cells: Although no role in regulating sense transcripts has conclusively been proved for any of these antisense RNAs, alterations in RNA splicing or induction of a double-stranded RNA unwinding and modifying activity have been suggested as mechanisms by which these endogenous antisense RNAs might regulate expression of the sense transcript (24,27). In the case of eIF-Ba, the presence of such an antisense RNA in the first intron of the primary transcript might result in premature termination of transcription or transcriptional attenuation in Go T cells. On the other hand, an antisense transcript that overlaps the splice site at the exon l/intron 1 boundary might result in alternative splicing of the eIF-2a transcript and the formation of different eIF-2a mRNAs. In this regard, it is interesting to note that there are two hybridizing bands by Northern blot hybridization using both singleand double-stranded probes containing eIF-2a exonic sequences (10): The function of the downstream in vitro DNase I footprint3 and the role played by an antisense RNA transcript in the regulation of eIF-2a expression as well as the role played by the two TATA-like elements and the Inr sequence in regulating expression of the antisense transcript are currently under investigation.