Genomic Cloning and Characterization of the Human Eukaryotic Initiation Factor-2β Promoter

Abstract The translation initiation factor eIF2 consists of three subunits that are present in equal molar amounts. The genomic DNA containing the gene for eIF2β and its promoter were cloned and sequenced to characterize further the mechanism of their regulated synthesis. Whereas Southern blot analysis indicated that a number of copies of the gene may exist, only one full-length intron-containing copy was identified. Similar to the eIF2α promoter, the eIF2β promoter is TATA-less, CAAT-less, and GC-rich and contains an α-Pal binding motif. Mutation of the α-Pal binding sequence resulted in an 8-fold decrease in activity when assayed by the luciferase reporter gene constructs. The data suggest a common mechanism of transcriptional control for the two cloned subunits of eIF2.

The translation initiation factor eIF2 catalyzes the first regulated step of protein biosynthesis, the binding of the initiator Met-tRNA i to the 40 S ribosomal subunit. Binding occurs as a ternary complex of Met-tRNA i -eIF2-GTP. All three subunits of eIF2 (␣, ␤, ␥) are required for the catalytic utilization of eIF2 during protein synthesis initiation (1,2). The ␣ subunit of eIF2 is a 36.2-kDa polypeptide whose phosphorylation state regulates activity of the heterotrimer (3). The ␤ subunit appears to bind GTP or GDP and is a 51.9-kDa polypeptide (4). The 38.3-kDa ␥ subunit may be directly involved in binding of the ternary complex to mRNA (5,6). None of the three subunits appears to exist as a monomer outside the eIF2 heterotrimer. Initial experiments demonstrated that balanced synthesis of the ␣ and ␤ subunits is predominantly the result of different rates of ribosomal elongation (7). Subsequent experiments suggested that eIF2␣ expression could be regulated by antisense transcripts that form double-stranded RNA during T-cell activation (8). Regulation of sense transcription is under the control of a transcription factor designated ␣-Pal, which binds to an unusual direct repeat element. ␣-Pal is homologous to the developmental transcription factors P3A2 and ewg (9). Potential target genes of ␣-Pal can be broadly classified as encoding growth-responsive factors (9). To see if the eIF2␤ subunit was regulated via similar mechanisms, the eIF2␤ gene and its promoter region were cloned and characterized.

MATERIALS AND METHODS
Library Cloning Screening-The cDNA for eIF2␤ was a gift from J. W. B. Hershey (GenBank number M29536). The nucleotide sequence of the promoter region has been reported to GenBank (number AF076927). A human lung fibroblast genomic library was constructed in the phage vector. Lambda Fix was purchased from Stratagene (La Jolla, CA). Subgenomic libraries were constructed by digesting high molecular weight DNA from K562 cells (ATCC, Manassas, VA) to completion with the indicated restriction enzymes followed by separation on a 0.5% Sea Kem GTG-agarose gel (FMC, Rockland, ME) and staining with ethidium bromide, and was the region containing the fragment of interest excised. The DNA was then purified and ligated into prepared phage arms following the manufacturer's instructions (Stratagene). To ensure the proper region of the gel was excised, the gel (minus the excised band) was Southern blotted and probed with the fragment of interest. Screening of these libraries was carried out essentially as described previously (10). Northern blots were prepared following standard procedures (10). Southern blots of genomic DNA were prepared following the instructions for use of the Zeta probe blotting membranes (Bio-Rad). All hybridizations were done in Church-Gilbert buffer (0.5 M NaPO 4 (pH 7.0), 1% bovine serum albumin, 7% sodium dodecyl sulfate, 10 mM EDTA) for 18 h at the indicated hybridization temperature and washed with a final stringency of 0.15 M NaCl for 1 h at the indicated hybridization temperature.
Primer Extension Analysis-Primer extension analysis was performed using MLV reverse transcriptase as described by the manufacturer (Life Technologies, Inc.). To determine the size of the products, four sequencing reactions were performed and run on the same gel.
Ribonuclease Protection Assays-Ribonuclease protection assays (RPAs) 1 were performed as described by the manufacturer (Ambion, Austin, TX). Riboprobes were generated using the Maxiscript in vitro transcription kit according to the manufacturer's instruction (Ambion). The size of the protected fragments was determined by comparison with control RNA and by comparison with a dideoxy sequencing ladder.
Luciferase Assay-Luciferase assays were performed as described previously (11). Briefly, NIH 3T3 cells were plated in 6-well plates at a density of 1 ϫ 10 5 cells/well and transfected using LipofectAMINE (Life Technologies, Inc.). After 48 h, cell lysates were prepared, and 5-l aliquots were added to 100 l of luciferase reagent and assayed in a Monolight™ 2010 luminometer for 10 s. Final results were given in units of luciferase activity/g of protein.

RESULTS
A cDNA clone for eIF2␤ was isolated from human liver mRNA, and the encoded polypeptide was shown to interact with the eIF2␣ and ␥ subunits (6). Southern blots of DNA isolated from K562 cells (ATCC) probed with the eIF2␤ cDNA produced a large number of fragments, indicating the gene for eIF2␤ is either very large or there are multiple copies (Fig. 1A). Hybridization of duplicate Southern blots with either an oligonucleotide probe from the 3Ј-UTR (bases 1273-1298) or a random-primed fragment from the 5Ј-UTR (1-103) generated a simplified pattern of only 4 or 5 bands (Fig. 1, B and C, respectively). These results suggest the existence of multiple genomic fragments with a high degree of homology to the eIF2␤ cDNA. Because of the high degree of specificity of oligonucleotide probes compared with random-primed fragments, the 3Ј-UTR oligonucleotide probe was initially used to screen a human fibroblast genomic library. * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Bldg. 10 Genomic clones corresponding to three of the four bands detected using the 3Ј-UTR oligonucleotide probe were isolated after screening multiple libraries (Fig. 2). Although all three exhibited a high degree of homology to the cDNA, only one was nearly identical and contained introns. This clone is referred to as clone 1A (Fig. 2). Additional sequencing and restriction mapping showed that clone 1A contained an insert 13 kilobases in length and 550 bases of the cDNA divided into 5 exons. A full-length clone was obtained by using polymerase chain reaction to amplify the DNA between the last exon in clone 1A and upstream exons. The promoter region and initiation codon were cloned by creating subgenomic libraries of the five DNA fragments that hybridized to the 5Ј 103 base pairs of the cDNA as shown in Fig. 1C. From the subgenomic library constructed from the 2.3-kilobase fragments, one clone was isolated that contained the missing cDNA sequence and could be linked to clone 1A. This produced a full-length cDNA and is referred to as clone 1 (Fig. 2). In addition, another pseudogene was isolated from this library, which contained the whole open reading frame with a number of point mutations but lacked introns. This clone is referred to as clone 4 (Fig. 2).
The sequence of clone 1 differed from the published cDNA in two respects. The 5Ј most 18 bases reported to be in the cDNA were not present in the 2.3-kilobase promoter fragment. Clone 1 also differed from the cDNA at position 1000. The cDNA contained a T at this position and the genomic clone contained a C. This difference could be the result of a mutation introduced by reverse transcriptase during the cloning of the eIF2␤ cDNA or it could indicate that more than one copy of the ␤ gene exists. To distinguish between these two possibilities, three identical Northern blots were prepared and probed with complementary oligonucleotide probes that contained the genomic version of this sequence based on clone 1 (5Ј-AGAGTCTGTCCTTC), the cDNA version of the sequence (5Ј-AGAGTCGTATGTCCTTC), or a unique sequence found in one of the pseudogenes, clone 2 (5Ј-TCACATGAAACTATTAAGTAAGC).
Under stringent washing conditions each oligonucleotide probe hybridized to the plasmid from which its sequence was obtained (Fig. 3). Although the cDNA and clone 1 oligonucleotide probes differed by a single nucleotide, no cross-hybridization was observed in the other plasmid. When these same hybridization conditions were used to probe Northern blots, only the oligonucleotide from clone 1 hybridized to a message on the Northern blots (Fig. 4B). The mRNAs detected by the clone 1 oligonucleotide probe are of similar mobility to those detected by hybridization with the full-length cDNA (Fig. 4D). Therefore clone 1A would appear to contain the correct sequence. Because the oligonucleotide probe based on the published cDNA sequence did not hybridize to the Northern blot (Fig. 4A), the T at position 1000 may be an artifact introduced during the cloning process of the cDNA.
To define the sequence at the 5Ј end of the message, total K562 RNA was hybridized with radiolabeled antisense riboprobes derived from either the 5Ј end of the cDNA or clone 1 and analyzed in a RPA. As a size control, the antisense riboprobes were also hybridized with a non-radiolabeled sense fragment transcribed from the original eIF2␤ cDNA. Fig. 5A (lane  3) shows that incubation of the radiolabeled cDNA probe with the cDNA sense control gave a band that was the correct length for a fully protected cDNA fragment. As a point of reference the length of this band is defined as 1 (Fig. 5, A and B). Incubation of the cDNA sense control message with the clone 1 riboprobe produced a band 18 bases shorter that corresponds to the point of divergence for the cDNA sequence and the genomic sequence. The length of this band is correspondingly Ϫ18 (Fig. 5,  A, lane 11, and B). Incubation of the cDNA probe with total K562 RNA also produced bands of approximately 18 bases shorter than the cDNA/cDNA hybrid, indicating that the 5Јterminal 18 bases contained in the cDNA are not present in the pool of K562 RNA (Fig. 5A, lanes 4 -6). Hybridization of the clone 1 riboprobe with total K562 RNA generated bands approximately 21 bases longer than the cDNA, indicating that the genomic sequence of clone 1 was again correct and that the 5Ј-UTR was longer than previously reported (Fig. 5A, lanes  13-15). Because the cDNA was isolated from a library derived from liver mRNA (6), this difference could be tissue-specific. However, hybridization with liver RNA with each probe gener-FIG. 5. A ribonuclease protection assay. Total RNA from either K562 cells or mouse liver was hybridized with antisense riboprobes from the 5Ј-UTR of either the cDNA or clone 1 and digested with RNase at a 1:50 (lanes 2-4, 7, 11-13, 16) and 1:25 dilution (lanes 5, 6,8,9,14,15,17,18) with either 50 g of total RNA (lanes 4, 5, 7, 8, 13, 14, 16, 17) or 100 g of total RNA (lanes 6, 9, 15, 18). As controls the two riboprobes were each hybridized with yeast tRNA (lanes 2, 12) or in vitro transcribed sense RNA from the cDNA (lanes 3, 11). ated bands of similar size to the K562 RNA (Fig. 5A, lanes 7-9,  16 -18).
To more accurately map the start site of the mRNA, total RNA from K562 cells was hybridized with a radiolabeled antisense primer binding 87 bases from the 5Ј end of the cDNA. The primer was extended with reverse transcriptase, and a series of bands was detected by autoradiography (Fig. 6, lane 1). This pattern was simplified by the addition of actinomycin D (Fig. 6,  lane 2), and a single prominent band of 108 bases was produced, indicating the 5Ј end of the message was 21 bases upstream from the 5Ј end of the cDNA clone. This length is in agreement with the potential start site identified by the RPA analysis. Therefore, of the four possible eIF2␤ genes that were identified by Southern blotting only one contained the entire cDNA and introns. The organization of this gene is shown in Fig. 7A. The gene for eIF2␤ spans 28 kilobases and is divided into 9 exons ranging in size from 50 to 480 bases. The message appears to be transcribed from a single start site 138 bases upstream of the initiation codon and contains a 3Ј-UTR of 280 bases. The number of nucleotides contained in each exon is summarized in Fig. 7B, columns 1 and 2. Further sequencing of clone 4 showed it contains several point mutations and an in frame deletion of 18 bases (amino acids 138 -143) (Fig. 7B, column 3). In vitro transcription and translation of clone 4 generated a protein product very similar in size to that translated from the cDNA (data not shown). Several attempts were made to identify a transcript for this gene in vivo using Northern blots, RPA of K562, and activated lymphocyte RNA, but none could be detected (data not shown).
The promoter region of eIF2␤ (Ϫ1000 to ϩ1321) was searched for transcriptional elements and was found to contain a number of potential transcription factor binding sites (Fig. 8). The eIF2␤ promoter contains a consensus sequence for the ␣-Pal transcription factor at Ϫ25. This element is also present in the promoter of the ␣ subunit of eIF2 (9). Positioned over the cap site is a potential Sp1 site with potential E2F, C/EBP, c-myc, and two SIF sites located downstream at ϩ15, ϩ22, ϩ253, ϩ93, and ϩ106, respectively. Upstream of the cap site, potential p53, Ap2, myoD, F-Act1, and CAAT box motifs exist.
Previous research has demonstrated the importance of the ␣-Pal site in the regulation of the eIF2 promoter (1). To assess the importance of the ␣-Pal element in the expression of the eIF2␤ gene, luciferase reporter constructs were made that contained mutations in the ␣-Pal region of the promoter. The ␣-Pal sequence was mutated from TGCGCAGGCGCA to TG-CAAATTGGAT, which had previously been shown not to bind the ␣-Pal transcription factor (7). This mutation decreased promoter function in the reporter constructs 8-fold compared with the wild type promoter (Fig. 9, A and B, column 2 versus  3). A similar change in activity was detected in transfection of both 3T3 and 293 cells. To detect possible antisense activity, the promoter region was cloned into the luciferase plasmid in the antisense orientation and transfected into 293 cells or 3T3 cells. No luciferase activity was detected using a region corresponding to the first intron (ϩ125 to ϩ1300) alone in the sense or antisense orientation (Fig. 9, columns 4 and 5, respectively). Furthermore, no luciferase activity was detected using the entire promoter region (Ϫ1000 to ϩ1300) in the antisense orientation (Fig. 9, column 6). Experiments using reverse transcriptase-polymerase chain reaction of RNA isolated from G o and activated T-cells also failed to detect any antisense RNA  transcription within the first intron region of eIF2, which has been described for eIF2 (8) (data not shown).

DISCUSSION
Cell growth and differentiation require the regulated expression of a large number of proteins. One mechanism for achieving coordinate regulation is through the use of a common regulatory transcription factor. Examples of this theme are the CREB family of responsive genes and the NF-B family. The presence of a functional ␣-Pal site in the two subunits of eIF2 that have been cloned indicates that ␣-Pal may also act as a coordinating factor.
Obtaining the genomic clone for eIF2␤ was much more complicated than for eIF2␣. Multiple pseudogenes were identified for the eIF2␤ subunit as well as an intronless minigene. In contrast, eIF2␣ was encoded by a single gene (12). Furthermore, several differences were identified in the genomic clone compared with the cDNA. Through the use of high stringency Northern blots and RPAs, the genomic clone was shown to be the correct sequence, and no mRNA corresponding to the cDNA sequences could be detected. The differences in the 5Ј-UTR and those at base 1000 in the published cDNA compared with the genomic sequence are most likely the result of ligation artifacts produced during the creation of the libraries and poor fidelity of reverse transcriptase, respectively.
The minigene identified during the cloning may be the result of a retrotransposition event. Although in vitro translation of this open reading frame generated a protein that was very similar in size to that generated from the cDNA, no transcript for this gene could be detected in vivo by Northern blot analysis, RPA in K562, or T-cells. Other bands are present on Northern blots probed with the eIF2␤ cDNA, and these mRNAs are most likely the result of multiple polyadenylation sites as has been reported for other translation factors (12). Although it would appear that only one functional gene for eIF2␤ exists, multiple functional copies of other translation factors have been identified (13). Recently two genomic clones were isolated for eIF4E; one contained introns and the other was intronless (14).
Whereas eIF2␣ and eIF2␤ are regulated at the transcriptional level by ␣-Pal, the cis elements in the two promoters are distinct. The eIF2␣ subunit promoter has two adjacent sites, which bind ␣-Pal with different affinities. Mutation of the high FIG. 8. Promoter sequence. The sequence of the eIF2␤ promoter region (Ϫ1000 to ϩ200) is presented. An arrow indicates the start site of transcription as mapped by both primer extension and ribonuclease protection experiments, and the initiation codon is italic and boxed. The ␣-Pal transcription factor binding site is underlined. Other potential transcription factor binding sites are identified by boxes. The 5Ј-UTR sequence identified in clone 1, which was missing from the published cDNA, is indicated in bold. affinity site decreases expression 12-fold, whereas mutation of the lower affinity site inhibits expression 3-fold. 2 In contrast, eIF2␤ has only a single site whose deletion results in an 8-fold reduction in transcriptional activity. However, both genes are TATA-less and have the ␣-Pal sites positioned in the traditional Ϫ30 TATA box site. Although potential ␣-Pal binding sites have been identified in a number of other genes (9), their importance has yet to be demonstrated. Characterization of ␣-Pal binding sites identified in these other growth-responsive genes is essential to defining the role of ␣-Pal as a coordinating transcription factor involved in the regulation of expression of growth response genes. FIG. 9. Luciferase reporter analysis. Several regions of the eIF2␤ gene were tested for transcriptional activity using a luciferase reporter gene assay in either 3T3 or 293 cells (A and B). Transcriptional activity was measured in both the sense and antisense direction. Column 1, luciferase gene lacking a promoter (basic); column 2, eIF2␤ promoter bases Ϫ1000 to ϩ125 (eIF-2beta); column 3, eIF2 promoter with a mutated ␣-Pal binding site (mut aPal); column 4, eIF2␤ promoter sequence ϩ125 to ϩ1300 in the sense orientation (intron sense); column 5, eIF2␤ promoter ϩ125 to ϩ1300 in the antisense orientation (intron antisense); column 6, eIF2␤ promoter Ϫ1000 to ϩ1300 in the antisense orientation (eIF-2beta antisense). RLU, relative light units.