Characterization of multiple mRNAs that encode mammalian translation initiation factor 5 (eIF-5).

Eukaryotic translation initiation factor 5 (eIF-5) interacts with the 40 S initiation complex (40S·mRNA·MettRNAf·eIF-2·GTP) to promote the hydrolysis of bound GTP with the concomitant joining of the 60 S ribosomal subunit to the 40 S initiation complex to form a functional 80 S initiation complex. In this paper, the multiple mRNAs that encode mammalian eIF-5 have been characterized. In rat tissues, three major eIF-5 mRNAs of 3.5, 2.8, and 2.2 kilobases in length are detected. All major eIF-5 mRNAs are initiated from a single transcription initiation site, contain identical 5′-untranslated and coding regions, but differ from one another only in the length of their 3′-untranslated regions. The different lengths of the 3′-untranslated region of eIF-5 mRNAs are generated by the use of alternative polyadenylation signals. Additionally, we demonstrate tissue-specific variations in eIF-5 mRNA expression as well as preference for polyadenylation sites. These results should lead to increased understanding of the regulation of eIF-5 gene expression.

Eukaryotic translation initiation factor 5 (eIF-5) interacts with the 40 S initiation complex (40S⅐mRNA⅐Met-tRNA f ⅐ eIF-2⅐GTP) to promote the hydrolysis of bound GTP with the concomitant joining of the 60 S ribosomal subunit to the 40 S initiation complex to form a functional 80 S initiation complex. In this paper, the multiple mRNAs that encode mammalian eIF-5 have been characterized. In rat tissues, three major eIF-5 mRNAs of 3.5, 2.8, and 2.2 kilobases in length are detected. All major eIF-5 mRNAs are initiated from a single transcription initiation site, contain identical 5-untranslated and coding regions, but differ from one another only in the length of their 3-untranslated regions. The different lengths of the 3-untranslated region of eIF-5 mRNAs are generated by the use of alternative polyadenylation signals. Additionally, we demonstrate tissue-specific variations in eIF-5 mRNA expression as well as preference for polyadenylation sites. These results should lead to increased understanding of the regulation of eIF-5 gene expression.
Eukaryotic translation initiation factor 5 (eIF-5) 1 plays a key role in initiation of protein synthesis in eukaryotic cells (for reviews, see Refs. [1][2][3][4]. During the initiation process, a 40 S preinitiation complex, consisting of a 40 S ribosomal subunit to which the initiator Met-tRNA f is bound as a Met-tRNA f ⅐ eIF-2⅐GTP ternary complex, scans along the mRNA until it recognizes the initiator AUG codon to form the 40 S initiation complex (40 S⅐mRNA⅐Met-tRNA f ⅐ eIF-2⅐GTP). The initiation factor eIF-5, a monomeric phosphoprotein of about 50 kDa (5)(6)(7), then interacts with the 40 S initiation complex to promote the hydrolysis of ribosome-bound GTP. Hydrolysis of GTP causes the release of eIF-2⅐GDP (and P i ) from the 40 S ribosomal subunit which is essential for the subsequent joining of the 60 S ribosomal subunit to the 40 S complex to form a functional 80 S initiation complex (80S⅐mRNA⅐Met-tRNA f ) that is active in peptidyl transfer (8 -11).
To increase our understanding of the structure and function of eIF-5 protein and regulation of its activity, we have recently cloned, sequenced, and expressed both a rat cDNA and the Saccharomyces cerevisiae gene encoding functional eIF-5 of calculated M r ϭ 48,926, and 45,346, respectively (12)(13)(14). Although the derived amino acid sequences of the yeast and rat eIF-5 protein show considerable sequence homology and identity, analysis of the cDNA-deduced structure of rat eIF-5 mRNA shows several interesting features that are absent in yeast eIF-5 mRNA. First, the 5Ј-UTR of rat eIF-5 mRNA contains two small ORFs upstream of the eIF-5 coding region (12). Such a feature of the 5Ј-UTR is characteristic of mRNAs of many regulatory genes, e.g. genes for growth factors, transcription factors, oncogenes, and signal transduction components (4,15). Second, while Northern analysis of yeast poly(A) ϩ RNA showed only a single size class of eIF-5 mRNA of 1.75 kb (14), multiple mRNAs were found to encode rat eIF-5 (12). The mechanism of formation of these different forms of rat eIF-5 mRNAs was unclear.
In this paper, we have cloned and sequenced a human eIF-5 cDNA. Sequence analysis indicates that like the rat eIF-5 mRNA, the deduced structure of human eIF-5 mRNA also contains two small ORFs preceding the coding region. Additionally, we have studied the pattern of eIF-5 expression in different mammalian tissues. Finally, we show that multiple mRNAs encoding mammalian (rat) eIF-5 are generated by alternative use of polyadenylation signals in the 3Ј-noncoding region.

EXPERIMENTAL PROCEDURES
cDNA Cloning and Sequencing-Specific DNA restriction fragments isolated from previously cloned partial 1.1-kb rabbit and complete 3.55-kb rat cDNA clones (12) were labeled with 32 P by random-priming using [␣-32 P]dCTP (Boehringer Mannheim DNA labeling kit) and used as hybridization probes to screen a HeLa cell cDNA library in phage ZAPII (Stratagene) as described by Sambrook et al. (16). Positive clones were plaque-purified to homogeneity, and the cDNA inserts present in recombinant phages were isolated by in vivo excision as a subclone in the plasmid vector pBlueScript SK(ϩ) (Stratagene). The inserts present in the recombinant plasmids were sequenced from both ends by the dideoxy chain termination method (17) using U. S. Biochemical sequencing kit, and a series of appropriate 17-mer deoxyoligonucleotide primers.
Primer Extension Mapping of the Transcription Initiation Site(s) of eIF-5 mRNAs-The transcription initiation sites were mapped using the oligonucleotide-directed primer extension method of Sambrook et al. (16) as follows. An 18-mer deoxyoligonucleotide 5Ј-GGTATCTTCT-GTCTGGAG-3Ј which is complementary to nucleotides Ϫ57 to Ϫ74 of the eIF-5 cDNA (see Fig. 1) was labeled at the 5Ј end with 32 P using [␥-32 P]ATP (3000 Ci/mmol) and phage T4 polynucleotide kinase. The 32 P-labeled primer (about 100 ng) was mixed with about 1 g of HeLa cell poly(A) ϩ RNA and the mixture was precipitated by adding 70% ethanol. The washed pellet was dissolved in 40 mM Pipes-HCl, pH 6.4, containing 1 mM EDTA, 0.4 M NaCl, and 80% formamide, and allowed to hybridize at 30°C for about 16 h. The DNA⅐RNA hybrids formed were then precipitated by adding 70% ethanol. The washed pellet was dissolved in the reverse-transcription buffer (Promega) containing 40 mM NaPP i and 40 units of RNasin (Promega) and reverse-transcribed at 42°C using 20 units of avian myeloblastosis virus reverse transcriptase. The 32 P-labeled DNA⅐RNA hybrids formed were isolated, * This research was supported by Grant GM15399 from the National Institutes of Health and by Cancer Core Support Grant P30CA13330 from the National Cancer Institute. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) U49436.
treated with DNase-free RNase, and the labeled single-stranded DNA was isolated by phenol-chloroform extraction followed by analysis on a 6% polyacrylamide, 8 M urea-sequencing gel.
Northern Analysis-The multiple tissue-Northern blot used in this study was purchased from Clontech. Each lane contained approximately 2 g of poly(A) ϩ RNA isolated from each rat tissue as indicated in the text. Following prehybridization as described (16), the blot was hybridized in the same solution containing 10% dextran sulfate at 42°C for 20 h with an appropriate 32 P-labeled DNA fragment derived from the complete 3.55-kb rat cDNA (12), labeled with [␣-32 P]dCTP (ϳ6000 Ci/mmol) by random-priming using the Boehringer Mannheim kit. The blot was then washed three times in 0.1 ϫ SSC, 0.1% SDS (1 ϫ SSC is 0.15 M NaCl, 15 mM sodium citrate, pH 7.0) at 60°C for 1 h, air-dried, and then autoradiographed. Following completion of one Northern analysis, the hybridized probe was removed by immersing the blot for 2 min in 0.5% SDS at 100°C. The temperature of the mixture was then brought down to about 20 -22°C by shaking the mixture at room temperature. The blot was then reused for another Northern analysis.

Molecular Cloning and Characterization of Human eIF-5
cDNA-To clone human eIF-5 cDNA, a ZAPII HeLa cell cDNA library (Stratagene) was screened using a 32 P-labeled 515-bp PvuII-HpaI DNA restriction fragment isolated from the 3Ј-end of the previously cloned partial rabbit liver cDNA (12) as a hybridization probe. Five positive clones were obtained in this screening procedure. An additional three independent clones were isolated when a 544-bp HindIII-SphI fragment isolated from the 5Ј-end of the complete rat cDNA clone (12) was used as the probe. Analysis of the DNA insert size in each homogenous positive phage clone followed by partial sequencing at their ends indicated that the cDNA inserts present in two of these clones encompassed the complete coding sequence as well as the 5Ј-UTR of the human cDNA. These two cDNA inserts were FIG. 1. Comparison of the nucleotide sequence of human and rat eIF-5 cDNAs. The human cDNA sequence was obtained as described in the text while the rat sequence is from Ref. 12. The ATG assigned as the translation start codon (numbered ϩ1) is boxed. The human cDNA contains the complete 5Ј-UTR and the unique transcription start site, mapped by primer extension analysis, is marked by a downward arrow (2) at nucleotide position Ϫ318. There are two other ATG codons in the 5Ј-untranslated regions, both of which generate short ORFs represented by broken lines. The shaded nucleotide residues in the human sequence show that these residues are different from those in the rat sequence. sequenced in their entirety and were found to contain the nucleotide sequence Ϫ335 to ϩ457 and the sequence ϩ439 to ϩ1507, respectively, of the human eIF-5 cDNA. Fig. 1 shows the nucleotide sequence of human eIF-5 cDNA and is compared with the previously determined (12) rat eIF-5 cDNA. There is a high degree of sequence conservation in the 5Ј-UTR as well as in the N-terminal coding region between the human and rat eIF-5 cDNA. Furthermore, a characteristic feature of both the rat and human cDNAs is the presence of two short open reading frames preceding the ATG codon from which the translation of eIF-5 mRNA is known to be initiated (Fig. 1). Although we have not cloned the complete 3Ј-UTR of human eIF-5 cDNA, it appears that the rat and the human sequences diverge considerably in the 3Ј-UTR. Fig. 2 shows the alignment of the amino acid sequence of human, rat (12), and the yeast (14) eIF-5 proteins. eIF-5 shows a high degree of sequence conservation at the amino-terminal half. The yeast protein is, however, shorter than the mammalian protein at the carboxyl-terminal end. Of particular interest is the observation that the derived amino acid sequence of eIF-5 from all three sources contains sequence motifs that are similar to the consensus GTP-binding domains that are characteristic of proteins of the GTPase superfamily (18) ( Table I). Members of the GTPase superfamily have sequence motifs in four distinct domains, designated G 1 through G 4 that have been implicated in mediating guanine nucleotide binding and GTP hydrolysis catalyzed by these proteins. In mammalian and yeast eIF-5, the G 1 domain and its surrounding sequences are highly conserved and is present in human and rat eIF-5 as 27 GKGNGIKT 34 and in yeast eIF-5 as 27 GRGNGIKT 34 (Table   I). The role of these sequence motifs in the hydrolysis of GTP mediated by eIF-5 is unclear at present.
Characterization of eIF-5 mRNAs-We have previously used 32 P-labeled DNA restriction fragments derived from the coding region of the rat eIF-5-cDNA to carry out Northern analysis of total poly(A) ϩ RNA isolated from rat liver and HeLa cells (12). Three distinct transcripts of 3.5, 2.8, and 2.2 kb were detected in rat liver, whereas HeLa cells contained a 3.5-and a 2.2-kb transcript (12). The mode of formation of multiple eIF-5 transcripts in mammalian cells was unknown. Since the rat cDNA corresponding to the longest 3.5-kb eIF-5 mRNA contained multiple potential polyadenylation signals (12), the possibility exists that some or all of the eIF-5 transcripts may arise by alternative processing at the 3Ј-noncoding region of the eIF-5 mRNAs. Alternatively, since the 5Ј-noncoding region of the 3.55-kb rat cDNA contained an excellent 3Ј-splice site, TCCCT-TCTTCTCCAG, preceding the initiation ATG codon (12) the possibility also exists that some of the shorter eIF-5 transcripts are derived by efficient splicing of eIF-5 mRNAs at this site. It is to be noted that many vertebrate cDNA sequences that have upstream ATG codons contain an unspliced 5Ј-intron (19). Finally, the multiple eIF-5 transcripts in mammalian cells may arise from multiple transcription initiation sites.
To distinguish between the above possibilities, we first carried out a primer extension analysis to map the transcription initiation site(s) of eIF-5 mRNAs. Fig. 3 shows that when a 18-nucleotide deoxyoligonucleotide primer complementary to nucleotides Ϫ57 to Ϫ74 of eIF-5 cDNA was labeled with 32 P at the 5Ј-end and used for primer extension analysis using HeLa cell poly(A) ϩ RNA as a template, a single 262-nucleotide long cDNA was found to be the only major reaction product, indicating that all eIF-5 mRNAs were initiated at nucleotide position Ϫ318 and thus have identical 5Ј-UTR. Next, we investigated whether multiple eIF-5 mRNAs were generated by the use of alternative polyadenylation signals in the 3Ј-noncoding regions. Since the rat cDNA corresponding to the longest 3.5-kb eIF-5 mRNA contained multiple potential polyadenylation signals (12), we carried out Northern blot analysis of total poly(A) ϩ RNA isolated from various rat tissues using 32 P-labeled DNA probes that correspond to different regions of the longest (3.55 kb) rat cDNA species (Fig. 4). When a 0.8-kb DNA restriction fragment that encompasses only the eIF-5 coding region was used as a hybridization probe (probe a, Fig. 4, panel B), three distinct transcripts of 3.5, 2.8, and 2.2 kb were detected (Fig. 4, panel a). In contrast, when the hybridization probe used in Northern analysis was derived from the region encompassing the first and the second putative polyadenylation signals (probe b), the 2.2-kb mRNA was no longer detectable (Fig. 4, panel b). Finally, analysis was carried out with a DNA probe (probe c) that spans the region between the second and the third putative polyadenylation signals. If multiple eIF-5 mRNAs are indeed generated by differential polyadenylation, probe c was expected to hybridize only to the 3.5-kb mRNA. As shown in Fig. 4, panel c, only the 3.5-kb mRNA was detected, whereas the other two shorter transcripts of 2.8 and 2.2 kb were undetectable. These results show that the three major eIF-5 transcripts observed in rat tissues differ only in the length of their 3Ј-untranslated region and in all probability, arise by alternative use of polyadenylation signals in the 3Ј-noncoding region.
Tissue Distribution of eIF-5 mRNAs-Northern analysis presented in Fig. 4 also shows the tissue-specific distribution of the three eIF-5 transcripts in various rat organs. The concentration of all three eIF-5 transcripts relative to other mRNAs is highest in rat testis. Furthermore, in this tissue, while all three eIF-5 transcripts are present in high abundance, molar concentration of the 2.2-kb species is much higher than that of either the 3.5-kb or the 2.8-kb RNA transcript. In contrast, in all other rat tissues examined, the 3.5-kb transcript is the most abundant form of eIF-5 mRNA while the 2.2-kb transcript is the least abundant species. It should be noted that the blot was also hybridized with the cDNA probe for actin as an internal control (data not shown). Except for heart and skeletal muscle where the expression of actin mRNA was highest, in all other rat tissues examined including testis, the relative abundance of actin mRNA was comparable (data not shown). DISCUSSION Our interest to clone the human eIF-5 cDNA stems from our earlier observation (12) that in rat eIF-5 cDNA, the ATG codon from which translation of eIF-5 protein is initiated is preceded by two short open reading frames. The presence of short ORFs in the 5Ј-UTR is characteristic of mRNAs of many regulatory genes (4,15,19). Results presented in this paper show that the 5Ј-UTR of mammalian eIF-5 cDNAs is remarkably conserved (ϳ98% identical), including the presence of two short ORFs. Furthermore, primer extension analysis clearly shows that this 5Ј-UTR is part of the mature eIF-5 mRNAs and does not contain an unspliced 5Ј-intron. It is interesting to note that in contrast to mRNAs of housekeeping genes and many other Restriction fragments that were labeled with 32 P and used as hybridization probes in Northern blots (probes a-c) are shown. Probe a, 812-bp HpaI-SpeI fragment containing part of the coding region and the 3Јuntranslated region preceding the first putative polyadenylation signal; probe b, 266-bp HindIII-HindIII fragment spanning the region between the first and the second putative polyadenylation signals in the 3Ј-UTR of rat eIF-5 cDNA (12); probe c, 832-bp HindIII-HpaI fragment located between the second and the third putative polyadenylation signal in the 3Ј-UTR of rat eIF-5 cDNA (12). vertebrate genes which have a GC-rich 5Ј-leader sequence (19), the 5Ј-UTR of eIF-5 mRNA is relatively GC-poor (only 40% G ϩ C). Furthermore, the putative initiation ATG codons in these short ORFs are in relatively poor sequence context with respect to translation initiation (pyrimidines at both-3 and ϩ4 positions). Thus a significant fraction of the scanning 40 S preinitiation complex is expected to bypass these upstream ATG codons and initiate translation from the ATG codon at ϩ1 of the known coding sequence of eIF-5. This ATG codon is in good sequence context for initiation having an A at the Ϫ3 position. Further work is clearly necessary to understand the role of these upstream ORFs, if any, in the regulation of translation of eIF-5 mRNAs in mammalian cells.
In earlier studies reported from several laboratories, eIF-5 isolated from rabbit reticulocyte lysates, was described to be a protein of about 150 kDa (for a review, see Refs. 2 and 3). In contrast, later work published from this laboratory on purification and characterization of the protein from mammalian cells (5-7) as well as cloning and expression of its cDNA (12,13) clearly showed that mammalian eIF-5 is a protein of about 50 kDa. However, Northern analysis of total poly(A) ϩ RNA isolated from rat liver and HeLa cells (12) showed that multiple mRNAs encode mammalian eIF-5. The presence of a 3Ј-splice site in the 5Ј-UTR immediately preceding the ATG start codon in the eIF-5-cDNA raises the possibility that one of the eIF-5 mRNAs could be a splice variant that might encode the high molecular weight isoform of eIF-5. Distinct molecular forms of the protein (isozymes) have been observed for wheat germ eIF-4F (20,21), and mammalian and yeast eIF-4A (22,23).
Experiments presented in this paper are consistent with the conclusion that all eIF-5 transcripts originate from a single initiation site, contain an identical 5Ј-UTR and coding region but differ from one another in the length of their 3Ј-UTR. The different lengths of the 3Ј-UTR of eIF-5 mRNAs are most likely generated by the use of alternative polyadenylation signals. It is also clear that the 2.2-kb eIF-5 mRNA, which is a relatively minor species in most rat tissues, is highly expressed in rat testis indicating a tissue-specific polyadenylation preference. The biological significance of alternative polyadenylation in the generation of multiple mRNAs encoding the same protein is not clear. However, a large number of mammalian genes including that for translation initiation factor eIF-4E have been shown to use differential polyadenylation signals to generate multiple mRNAs encoding the same protein (24,25). It is particularly noteworthy that recent experiments in a variety of eukaryotic systems have shown that the 3Ј-UTR and the length of the poly(A) tail are important determinants in the translational regulation of gene expression particularly during embryonic development and differentiation (for a review, see Refs. 26 and 27). Further work will undoubtedly focus on whether the activity of eIF-5 and expression of this initiation factor gene are regulated during development and differentiation.