A 110-Kilodalton Subunit of Translation Initiation Factor eIF3 and an Associated 135-kilodalton Protein Are Encoded by theSaccharomyces cerevisiae TIF32 and TIF31Genes*

Translation initiation factor eIF3 is a multisubunit protein complex required for initiation of protein biosynthesis in eukaryotic cells. The complex promotes ribosome dissociation, the binding of the initiator methionyl-tRNA to the 40 S ribosomal subunit, and mRNA recruitment to the ribosome. In the yeast Saccharomyces cerevisiae eIF3 comprises up to 8 subunits. Using partial peptide sequences generated from proteins in purified eIF3, we cloned the TIF31 and TIF32genes encoding 135- (p135) and 110-kDa (p110) proteins. Deletion/disruption of TIF31 results in no change in growth rate, whereas deletion of TIF32 is lethal. Depletion of p110 causes a severe reduction in cell growth and protein synthesis rates as well as runoff of ribosomes from polysomes, indicative of inhibition of the initiation phase. In addition, p110 depletion leads to p90 co-depletion, whereas other eIF3 subunit levels are not affected. Immunoprecipitation or nickel affinity chromatography from strains expressing (His)6-tagged p110 or p33 results in the co-purification of the well characterized p39 and p90 subunits of eIF3 as well as p110 and p33. This establishes p110 as an authentic subunit of eIF3. In similar experiments, p135 and other eIF3 subunits sometimes, but not always, co-purify, making assignment of p135 as an eIF3 subunit uncertain. Far Western blotting and two-hybrid analyses detect a direct interaction of p110 with p90, p135 with p33, and p33 with eIF4B. Our results, together with those from other laboratories, complete the cloning and characterization of all of the yeast eIF3 subunits.

reactions in the initiation pathway. eIF3 alone among the initiation factors binds stably to 40 S ribosomal subunits (3). The factor promotes the dissociation of 80 S ribosomes into 40 S and 60 S subunits, affects the stability of the ternary complex comprising methionyl-tRNA i ⅐eIF2⅐GTP in the absence of ribosomes but in the presence of mRNA, and stabilizes methionyl-tRNA i binding to 40 S subunits (1). It also is absolutely required for mRNA binding to ribosomes, where eIF3 already bound to the 40 S ribosome interacts with a region of eIF4G, a component of the mRNA m 7 G-cap binding complex, eIF4F (4). Thus eIF3 acts as a bridge between the 40 S ribosome and the mRNA⅐eIF4F complex. It is apparent that eIF3 plays a central role in initiation by interacting with numerous other translational components.
To elucidate the function of eIF3 during the initiation phase, we turned our attention to the budding yeast, Saccharomyces cerevisiae, where genetic approaches and gene replacement strategies are well developed. The other initiation factors from mammalian and yeast cells are quite strongly conserved, ranging from 17 to 72% sequence identity (1). Thus information obtained about yeast eIF3 may be applicable to understanding the structure/function of human eIF3. A yeast homolog of mammalian eIF3 was isolated on the basis of its activity in an in vitro methionyl-puromycin synthesis assay constructed with purified mammalian components of initiation (11). The yeast eIF3 preparation contains proteins with masses of 135, 90, 62, 39,33,29,21, and 16 kDa. The genes encoding p90 (PRT1), p62 (GCD10), and p16 (SUI1) had been identified previously, and were subsequently shown to encode eIF3 subunits (11)(12)(13). The genes encoding the p39 (TIF34) and p33 (TIF35) subunits have been cloned and characterized more recently (14,15). The p29 subunit was shown to be a degradation product of p39 (14), whereas a protein of 93 kDa has been identified as an eIF3 subunit (p93) encoded by NIP1 (16,17). Nip1p (p93) is not present in our preparation described above, apparently due to proteolysis during purification. In this report, we describe the cloning of TIF31 encoding p135 and show that the 21-kDa protein is derived from a 110-kDa subunit encoded by TIF32.
Thus, 8 proteins associated with yeast eIF3 have been cloned and characterized: p135, p110, p93, p90, p62, p39, p33, and p16. Interactions of p135, p110, and eIF4B with other eIF3 subunits also are defined. The report thereby completes the detailed characterization of yeast genes encoding known eIF3 components and employs the genes in studies designed to define the subunit composition and structure of eIF3.

MATERIALS AND METHODS
Strains, Cell Growth, and General Procedures-The strains of S. cerevisiae employed in this study are listed in Table I. Yeast cells were grown at 30°C in yeast-peptone (YP) medium or in synthetic complete (SC) medium lacking the indicated amino acids and were supplemented with 2% glucose (D) or galactose (G) (18). Cell growth was monitored as optical density at 600 nm (OD 600 ). Yeast cells were sporulated on 2% agar plates with 0.3% potassium acetate, 0.02% raffinose, and 10 g/ml of each amino acid as sporulation medium, and tetrad analyses were carried out as described (19). Yeast chromosomal DNA and RNA were prepared and transformations performed according to Rose et al. (18). Escherichia coli XL-1 blue was used for all plasmid cloning procedures and the recombinant TRX-p110 fusion protein was expressed in E. coli AD494(DE3) (Novagen). Plasmid preparations, recombinant DNA techniques, and Northern and Southern blotting were performed as described (20,21).
Cloning and Chromosomal Disruption of TIF31 and TIF32-eIF3 was purified as described previously (11) and was fractionated by SDS-PAGE. Internal peptide sequences for the p135 and p21 subunits were determined in the Protein Structure Laboratory (University of California, Davis) as described (22). The 5 peptide sequences from p135 and the two from p21 are reported in Table II and were used to perform a TblastN search of the GenBank data base.
The five p135 peptide sequences match 51 out of 55 residues with a putative protein encoded by cosmid clone c8270 (American Type Cell Culture). The 7.2-kb ApaI and XhoI fragment (GenBank accession number AF004911) was ligated into the same sites of pBluescript KSII (Stratagene) to yield pKSCos-TIF31. For disruption of the open reading frame (ORF), tentatively called TIF31 (for translation initiation factor 3, 1st subunit), the gene was PCR-amplified from total yeast genomic DNA to generate a 4.6-kb DNA fragment that includes 519-and 257-bp flanking the 5Ј and 3Ј ends of the ORF. The PCR product was cloned into the SmaI site of pBluescript KSII to yield pKS-TIF31. pKS-TIF31 was digested with BstBI and HpaI to remove 98% of the TIF31 coding region and a 1.7-kb BamHI DNA fragment containing the HIS3 gene was inserted (see Fig. 1). The resulting plasmid, named pKS-TIF31::HIS3, was digested with SspI and AseI to release a 2.5-kb DNA fragment containing the tif35::HIS3 allele. The fragment was transformed into the diploid yeast strain W303 to create a one-step gene deletion/disruption (23), confirmed by Southern blot analysis, to yield PH135D-8. PH135D-8 was sporulated, tetrads were dissected, and a His ϩ haploid spore colony containing tif31::HIS3 was selected and named PH135H-B.
For the cloning of the p110 gene (TIF32), the two p21 peptide sequences (Table II) matched two regions of a putative 110-kDa protein derived from the yeast genome data base (ORF YBR079c). To obtain TIF32 DNA, yeast genomic DNA (200 g) was digested with NheI and PstI, subcloned into pBluescript KS ϩ (Stratagene) and transformed into E. coli XL1-blue. Colonies containing the desired plasmid were identified by hybridization with a labeled 3.9-kb PCR product amplified from genomic DNA. One of the plasmids carrying TIF32 was named pHV110-1 and the sequence of the 4565-bp insert was determined and submitted to GenBank under accession number AF004912. A SstI/ Asp718I fragment from pHV110-1 carrying the putative TIF32 gene was subcloned into the same sites of the centromeric/URA3 plasmid pRS316 (24) to generate pHV110-2.
To disrupt a chromosomal TIF32 gene by a one-step gene replacement, the 3.9-kb PCR product described above, which includes 557 and  This work a Strain PH1-39 was constructed by transforming strain TND1 with plasmid pRD53pTIF34Myc (29), followed by tetrad dissection and selection for His ϩ Ura ϩ haploids. 462 bp of DNA flanking the 5Ј and 3Ј ends of the TIF32 coding region, respectively, was inserted into the SmaI site of pBluescript KS ϩ to generate pKSp110. The TIF32 coding region was deleted by digestion with StyI and BspMI (see Fig. 1) and the 1.7-kb BamHI fragment carrying the HIS3 gene was ligated to form pHV110-3. The tif32::HIS3 allele was excised by digestion of pHV110-3 with AseI and PacI and the resulting 2.3-kb DNA fragment was transformed into the diploid yeast strain W303. One of the His ϩ transformants was selected and named HV110-D20/D. The correct disruption of one of the TIF32 genes was confirmed by Southern blot analysis (not shown).
Construction of Plasmids and Strains Expressing TIF32-To express TIF32 under a regulatable promoter, plasmids directing expression of wild type p110 and N-or C-terminal (His) 6 -tagged p110 were constructed as follows. The 5Ј and 3Ј halves of TIF32 were PCR amplified from pHV110-1 with suitable primers to generate DNAs encoding either wild type or (His) 6 -tagged fragments. These were combined to encode full-length p110 proteins, then subcloned into p414Gal1 to yield pHPp110EX1 (wild type), pHPp110EX2 (N-terminal tag), and pHPp110EX3 (C-terminal tag) under control of the GAL1 promoter. The wild type construct also was cloned into p414GalS (25) which contains a weakened GAL1 promoter to give pHPp110EX1-S. All four constructs were verified by sequencing. They were transformed into the yeast strain HV110-D20/D, cells were sporulated, and tetrads were dissected. The resulting His ϩ Leu ϩ haploid strains HV110-24, HV110-29, HV110-33, and HV110-42 express, respectively, an untagged p110 from the GAL1 promoter, an untagged p110 from the GALS promoter, an Nterminal His-tagged p110, and a C-terminal His-tagged p110 as the sole source of eIF3-p110.
Antibodies-(His) 6 -p135⌬C was purified with HIS-Bind TM resin (Novagen) from E. coli BL21 (DE3) transformed with pET-NH135⌬C, and used to raise antibodies against p135 in rabbits (BAbCo). Antibodies specific for p110 were obtained from rabbit anti-yeast eIF3 antiserum (11) by affinity purification (26) against a urea-purified His-tagged C-terminal 607-amino acid fragment of p110 fused to the TRX protein expressed in E. coli. Affinity-purified antibodies to p135 were prepared similarly. Rabbit antisera against p33, p39, and Prt1p (a gift from Alan Hinnebusch (National Institutes of Health)) have been described previously (15).
Coimmunoprecipitations-Yeast cells expressing various (His) 6 -or Myc-tagged subunits of eIF3 were grown in YPG to an OD 600 of 1.0. A 50-ml aliquot of each culture was harvested and cells were resuspended in 300 l of buffer C 50 ϩ (50 mM Tris/HCl, pH 7.5, 50 mM KCl, 0.1% (v/v) Triton X-100, 10% (v/v) glycerol, and 1 ϫ Complete TM (Roche Molecular Biochemicals) protease inhibitors). Cells were lysed with glass beads, lysates were clarified by centrifugation, and the protein concentration was determined (27) and adjusted with buffer C 350 ϩ (buffer C 50 ϩ , but containing 350 mM KCl) to 2 mg/ml. Lysates (500 l) precleared with Protein A/G-agarose beads (Santa Cruz Biotechnology) were incubated for 90 min at 4°C with 20 l of Protein A/G-agarose beads loaded with anti-(His) 6 antibody (CLONTECH) or anti-c-Myc antibody 9E10 (Santa Cruz Biotechnology). Beads were washed 3 times with 500 l of buffer C 350 ϩ and resuspended in SDS-PAGE sample buffer. The bead-extracted proteins were separated by SDS-PAGE, blotted onto a polyvinylidine difluoride membrane, and probed with the antibodies described in the figure legends.
For radiolabeling of cell proteins for subsequent immunoprecipitation, cells were grown in 10 ml of SC/galactose medium lacking methionine to an OD 600 of 0.8, labeled with 0.5 mCi of [ 35 S]methionine and [ 35 S]cysteine (Tran 35 S-label, ICN) for 2 h at 30°C and lysed in 200 l of buffer C 350 ϩ as described above. The lysate volume was adjusted with C 350 ϩ to 500 l and a lysate from the protease-deficient strain YAS538 was added to a final protein concentration of 2 mg/ml. Coimmunoprecipitation was as described above.
Far Western Blot Analysis-To generate the 35 S-labeled proteins used as probes, the full-length open reading frames of the following genes were PCR amplified and cloned into pET28c to allow coupled in vitro transcription/translation in a rabbit reticulocyte lysate (Promega): PRT1 (p90), TIF32 (p110), TIF31 (p135), and TIF3 (eIF4B). eIF3 proteins obtained from the gel filtration step described above or purified recombinant full-length p33 or p33⌬C (15) were separated by SDS-PAGE and blotted onto polyvinylidine difluoride membranes. Membranes were washed 4 times for 10 min in buffer D (20 mM HEPES, pH 7.4, 10% (v/v) glycerol, 1 mM dithiothreitol, 150 mM KCl, 5 mM MgCl 2 ) and blocked by incubation for 1 h in buffer D containing 5% (w/v) bovine serum albumin. Membranes were incubated for 5 h at the indicated temperatures with the in vitro-translation extracts containing a radiolabeled protein supplemented with 3% (w/v) bovine serum albumin in buffer D, unbound protein was removed by washing the membranes 2 times for 10 min in buffer D containing 0.1% (v/v) Nonidet P-40 and the membranes were subjected to autoradiography.

Identification and Cloning of TIF31 and TIF32
Encoding p135 and p110 -To complete the identification of genes encoding possible subunits of eIF3, we report the cloning and characterization of TIF31 and TIF32. Each of these genes was cloned by obtaining partial peptide sequences derived from the 135-and 21-kDa proteins present in our preparations of eIF3. The sequences reported in Table II were used to search yeast data bases, an open reading frame was identified that encodes each set of peptides, and the genes were cloned as described under "Materials and Methods." pKSCos-TIF31 carries an insert with a 3834-bp coding region (assuming the first AUG serves as the initiator codon), with 1752 and 1638 bp of DNA flanking the 5Ј and 3Ј regions, respectively. An in-frame stop codon exists 21 nucleotides upstream of the first AUG, ruling out initiation sites upstream. However, AUGs found 62 and 88 codons downstream of the first AUG possibly may serve as initiator codons. The absence of a strict consensus sequence surrounding the initiator AUG in yeast (28) does not allow a firm prediction of which AUG is actually used. The hypothetical p135 product of TIF31 translated from the first AUG contains 1277 amino acid residues with a calculated mass of 145,165 Da and a pI of 6.0. A search of data bases indicates that p135 has no mammalian homolog (but see "Discussion" below).
For p110, plasmid pHV110-1 was constructed as described under "Materials and Methods" which contains a 2892 bp ORF with 920 and 748 bp of 5Ј-and 3Ј-flanking DNA, respectively. The first AUG is preceded by an in-frame stop signal 16 codons upstream, whereas the second in-frame AUG is 169 codons downstream. TIF32 mRNA translated from the first AUG generates a protein with 964 amino acid residues, a mass of 110,329 Da and a calculated isoelectric point of 6.1. The two peptide sequences from the 21-kDa protein are derived from a stable C-terminal fragment of p110. Apparently, the p110 subunit is readily degraded in yeast strains derived from W303, since little or no p110 is detected in eIF3 prepared by classical protein fractionation techniques (11). Northern blot hybridizations (not shown) with probes derived from either the N-or C-terminal coding regions of TIF32 show a single 3.2-kb band indicative of only one size class of mRNAs. Further evidence that TIF31 and TIF32 encode the p135 and p110 proteins in eIF3 is provided below and is summarized under "Discussion." The p110 amino acid sequence has a high content of charged residues (33.5%) evenly distributed over the entire protein.
Data base searches identify homologous proteins from human, mouse, Caenorhabditis elegans, and Nicotina tabacum with amino acid sequence identity/similarity values around 26/38%. The mammalian homologs encode the p170 subunit of eIF3. An alignment of these sequences has been published elsewhere (5).
The p110 Subunit Is Essential for Cell Growth, but p135 Is Not-To explore the physiological role of p135 and p110 in eIF3, we attempted to construct haploid strains lacking these proteins. Individually, most of the coding region of one of the TIF31 or TIF32 genes in the diploid strain W303 was deleted and disrupted by insertion of HIS3 as described under "Materials and Methods" and depicted in Fig. 1. The presence of both the wild type and disrupted allele was confirmed in each case by Southern blotting (not shown). Following sporulation and tetrad dissection, four viable spore colonies were obtained from the TIF31-disrupted diploid PH135D-8 (results not shown). The four spore colonies are of equal size, indicating that p135 is not required for rapid growth of yeast. The presence of only the disrupted tif31::HIS3 allele and the absence of p135 protein in the two His ϩ spore colonies of one of the tetrads were confirmed by Southern blotting (not shown) and by Western immunoblotting of cell lysates (Fig. 2).
In contrast, when the disrupted TIF32 diploid strain, HV110-D20/D, was sporulated and tetrads were dissected, a 2 ϩ :2 Ϫ segregation pattern for cell growth was observed for all dissected spores (results not shown). All viable spores were His Ϫ , demonstrating the absence of the disrupted tif32::HIS3 allele. Microscopic analysis of the nonviable spores showed that the spores are capable of undergoing two to four cell divisions before ceasing to grow. The observed segregation pattern together with the microscopic analysis demonstrate that a functional TIF32 gene is required for cell growth, but not for germination.
To demonstrate that the disruption of TIF32 is the sole cause of defective cell growth, we cloned TIF32 with its flanking regions into the centromeric plasmid pRS316 to generate pHV110-2. The plasmid was transformed into HV110-D20/D and transformants were sporulated and tetrads dissected. For all dissected asci two large colonies were observed with the additional presence of zero to two slower growing colonies (not shown). Marker analysis revealed that all fast growing colonies are His Ϫ and therefore carry a wild type chromosomal allele. All slow growing colonies are His ϩ and Ura ϩ , demonstrating disruption of the chromosomal copy of TIF32 and the presence of the plasmid carrying a wild type TIF32 gene. One His ϩ /Ura ϩ haploid strain was named HV110-13. The reduced growth rate of cells carrying pHV110-2 as sole source of p110 might be due to the lack of TIF32 regulatory elements in the plasmid, but such elements would have to be located more than 920 bp upstream from the translational start codon. TIF32 carried on the plasmid is a recessive allele, because several of the normal growing His Ϫ cells are also Ura ϩ and therefore carry both TIF32 alleles. The slow growth phenotype is not seen in liquid cultures and was not analyzed further.
Depletion of p110 Inhibits Cell Growth and Protein Synthesis-The cellular level of eIF3-p110 was reduced by placing TIF32 under control of glucose-repressible promoters. pHPp110EX1 and pHPp110EX1-S were constructed to express TIF32 from the strong GAL1 promoter and a weaker GALS promoter as described under "Materials and Methods." Each plasmid was transformed into the diploid strain HV110-D20/D, followed by tetrad dissection, to yield haploid strains HV110-24 and HV110-29, respectively, that express TIF32 only from the plasmid. Both haploid strains grow rapidly (100 min doubling times) in YPG liquid medium and no overexpression of p110 is detected by Western immunoblotting (results not shown), even though these expression vectors normally produce higher levels of protein. This suggests that p110 is rapidly degraded when not incorporated into the eIF3 complex. The absence of intact p110 in our purified eIF3 preparations (11) also indicates that p110 is an unstable protein readily susceptible to degradation. Because no difference was detected in either growth rate or p110 expression level for strains HV110-24 and HV110-29 (results not shown), all subsequent experiments were carried out with HV110-29.
To determine the effects of p110 depletion, we compared the growth rates of strains HV110-29 and HV110-13 (expressing TIF32 under control of its own promoter) in both galactose and glucose media (Fig. 3A). Although both strains grow equally well in galactose medium, only HV110-29 shifted to glucose grows progressively more slowly, exhibiting a doubling time of Depletion of an eIF3 subunit is expected to result in reduced protein synthesis. The rate of protein synthesis in strain HV110-29 after the shift to glucose was measured by [ 35 S]methionine pulse labeling as described previously (15). A substantial reduction of the rate of protein synthesis is seen by 4 h, and after 10 h of depletion the rate has dropped to about 20% of the non-depleted level (Fig. 3B, solid circles). Little or no inhibition of protein synthesis is seen in strain HV110-13 in glucose (closed squares) or HV110-29 in galactose (open circles), where p110 depletion does not occur; the slight decrease in protein synthesis rates at later times likely is due to their approach to stationary growth. The continued slow growth and low-level protein synthesis with HV110-29 after 12 h in glucose suggest either that p110 is not absolutely essential, or that low levels are sufficient to provide some eIF3 function. Maintenance of low p110 levels may be due to incomplete glucose suppression of TIF32 expression.
Effects of p110 Depletion on Polysome Profiles and eIF3 Subunit Composition-To determine the phase of protein synthesis that is inhibited upon p110 depletion, we analyzed the distri-bution of ribosomes on mRNAs. Reducing the rate of translation initiation without changing the rate of elongation leads to a reduction of polysome size (number of ribosomes per mRNA) and an increase in free 80 S ribosomes. When HV110-29 cells are shifted to glucose for 2 h, a time when protein synthesis already is inhibited over 2-fold (Fig. 3B), a reduction of polysome size and an increase in 80 S ribosomes are observed compared with cells grown in galactose for the same time (Fig.  4). After 6 h the polysome size is severely reduced and after 12 h polysomes are hardly detectable. When the same cells are grown in galactose, little or no reduction of polysome size is apparent after 6 h. The more pronounced effect on polysome size after 12 h again may be due to exit from the exponential growth phase (see Fig. 3A). The results demonstrate that the reduced protein synthesis rate observed after depletion of p110 (Fig. 3B) is due to decreased efficiency of translation initiation, indicating that eIF3-p110 plays an important role in the initiation pathway.
To demonstrate that p110 actually is depleted from the HV110-29 cells after the shift to glucose, we analyzed the levels of p110 and some other eIF3 subunits in crude cell lysates prepared at various times after the carbon source shift. eIF3 subunit levels in equal amounts of cell lysates were determined by Western blotting with specific antibodies as described under "Materials and Methods." As shown in Fig. 5A, an immunoreactive protein of 110 kDa is indeed reduced over the time course of glucose repression of TIF32 expression. This is seen with blots subjected to treatment with a crude anti-eIF3 antiserum (upper panel), and also with antibodies affinity purified from the anti-eIF3 antiserum with recombinant p110 protein.
Specific detection of a 110-kDa protein in undepleted cells with the affinity-purified antibodies demonstrates that TIF32 indeed encodes a 110-kDa protein.
The crude anti-eIF3 antiserum also detects the p33 subunit, whose level does not change following p110 depletion. The p39 and p90 subunits are not readily detected in lysates by the antiserum to eIF3, but analyses with affinity-purified anti-p39 were removed and labeled as described previously (15). The results shown are representative of independent experiments performed three times. Protein synthesis rates are expressed as counts/min ϫ 10 Ϫ4 of incorporated methionine ϫ g Ϫ1 protein min Ϫ1 .
FIG. 4. Effect of p110 depletion on polysome profiles. Strain HV110-29 was shifted into medium containing either galactose (right panels) or glucose (left panels), cells were chilled and harvested with cycloheximide (100 g/ml) at the indicated times, lysed and analyzed by sucrose gradient centrifugation essentially as described (15). Absorption (254 nm) was monitored during upward displacement with a gradient fractionator (ISCO). The figure shows A 254 scans of the gradients, with sedimentation from left to right. The large peak at about fraction 5 corresponds to 80 S ribosomes.
antibodies and an antiserum specific for p90 show that p39 levels remain constant, whereas p90 levels decrease. Quantitation of the Western blot signals (Fig. 5B) reveals that the p110 level drops following the shift to glucose and by 12 h is only 10% of the normal concentration. The p90 level decreases similarly. The quantitations also confirm that p39 and p33 levels, as well as that of eIF5A which is not associated with eIF3, do not significantly change over the time course of the experiment. Whether p110 depletion regulates the synthesis of p90 or makes p90 more accessible to degradation remains to be determined. The observed co-depletion of p90 and p110 suggests an interaction between the two subunits. Since p90 is a proven component of eIF3, the finding supports the view that p110 is associated with the eIF3 complex.
p110 Is a Integral Part of the eIF3 Complex-Further evidence that p110 is in fact a subunit of yeast eIF3 was obtained by immunoprecipitation of the eIF3 complex. Strain HV110-33 was constructed which overexpresses an N-terminal (His) 6 tagged p110 protein as the sole source of p110. Since the growth rate of the strain is normal, we conclude that the (His) 6 tag is not detrimental to the function of the protein. Strains expressing (His) 6 -tagged p33 and p135, and Myc-tagged p39, were subjected to protein labeling with [ 35 S]methionine and immunoprecipitation with (His) 6 -specific or anti-Myc antibodies as described under "Materials and Methods." To control for nonspecific immunoprecipitation, strain W303-1A, which contains no (His) 6 -or Myc-tagged protein, was treated similarly. A number of radiolabeled protein bands are seen in the autoradiogram of the tagged immunoprecipitates that also are seen with W303-1A (Fig. 6A), and these are labeled with the letter u, indicating nonspecific precipitation. The anti-(His) 6 antibodies specifically precipitate proteins in the HV110-33 lysate that have masses of 110 and 105 kDa, the former corresponding to (His) 6 -tagged p110 and the latter probably being its partial degradation product (labeled p110* in Fig. 6A). Additionally, proteins of 90, 39, and 33 kDa are coimmunoprecipitated in the HV110-33 strain but not in strain W303-1A, except that the 39-kDa protein is precipitated in both strains (see "Discussion" below). A similar pattern of labeled immunoprecipitated proteins is seen with strain PHS33 containing (His) 6 -tagged p33 and strain PH1-39 containing Myc-tagged p39 (Fig. 6A). Interestingly, PH1-39 produces an additional weak band of ϳ95 kDa (labeled with a dot) which might correspond to the p93 (Nip1p) subunit of eIF3. The results indicate that p110 is found in complexes with the 90, 39, and 33 kDa proteins. Since all of these proteins correspond in size to subunits of eIF3, the results imply that p110 is a part of the eIF3 complex. It is noteworthy that in Fig. 6A, the bands corresponding to the (His) 6 -and Myc-tagged p33 and p39 subunits are very intense, indicating substantial accumulation of these proteins in the cells, whereas that for (His) 6 -tagged p110 is not intense, consistent with an innate lability of p110.
To confirm the identity of the coimmunoprecipitated proteins detected by autoradiography, nonradiolabeled lysates from strains W303-1A, HV110-33, and PHS33 were subjected to immunoprecipitation with the anti-(His) 6 antibodies as described above. Following SDS-PAGE fractionation of the proteins in the immunoprecipitates, we used anti-eIF3 antiserum (Fig. 6B, upper panel) as well as highly specific antibodies (lower panels) to p110, p90, p39, and p33 to detect these eIF3 subunits in the immunoprecipitates. The p110, p90, and p33 subunits are not detected when the immunoprecipitating antibody is omitted or in the precipitate from W303-1A. However, they are present in the anti-(His) 6 precipitates from strains HV110-33 and PHS33 when analyzed by anti-eIF3 and the specific antibodies. Detection of the prominent 105-kDa band (just below the faint p110 band labeled by dots) from strain HV110-33 with the anti-p110 antibodies indicates that this protein is very likely a (His) 6 -p110 degradation product. In the case of p90, its absence in the W303-1A precipitate is obscured by a defect in the blot with the highly specific antibody; however, it clearly is not present in the anti-eIF3 blot or in the labeled precipitate analyzed in panel A. Unexpected is the apparent precipitation of p39 from W303-1A by the anti-(His) 6 antibodies seen in panel A and confirmed with the anti-p39 antibodies (panel B). Noteworthy is the fact that the p39 subunit from W303-1A also co-purifies by IMAC (see Fig. 8 below). Although this might suggest that p39 contains an oligo-histidine tract, no such tract longer than two residues is present in the protein (14). We are not able to explain the presence of p39 in the W303-1A precipitates. In summary, the coimmunoprecipitation experiments demonstrate that p110 is part of a multisubunit complex containing at least two of the known subunits of eIF3, namely p33 and p90.
A second approach to obtaining eIF3 complexes rapidly is to employ IMAC. The eIF3 complex was purified on a HIS-Bind TM resin by utilizing the N-terminal (His) 6 -tagged p110 from strain HV110-33, as described under "Materials and Methods." Fractions from each elution step were subjected to SDS-PAGE and Western immunoblotting with the anti-eIF3 antiserum (Fig. 7A). The majority of eIF3 elutes with 250 mM imidazole. Because the 250 mM imidazole fraction contains numerous proteins as detected by Coomassie staining (result not shown), it was subjected to size exclusion chromatography on a Superdex 200 column as described under "Materials and Methods." Aliquots of various fractions were subjected to SDS-PAGE, and eIF3 was detected with the anti-eIF3 antiserum (Fig. 7B, upper  panel) and by Coomassie staining (middle panel). The strongest signals with the antibody are seen in fractions 11-13. Coomassie staining also shows the presence of 33, 39, 90, and 110 kDa proteins in fractions 11-13. A complex of identical subunit composition was obtained from strain HV110-42, which expresses p110 with a C-terminal (His) 6 tag (results not shown). The proteins in fractions 11-13 elute with an apparent molecular mass of 530 kDa (Fig. 7B, lower panel), consistent with the size of the eIF3 complex.
p135 Also May be Present in a Complex with Other eIF3 Subunits-Further evidence that TIF31 encodes a protein associated with eIF3 was obtained by constructing strain PH1-135 which expresses (His) 6 -tagged p135 as the sole source of this protein from a GAL1 promoter. Ribosomal salt wash fractions were prepared from strain PH1-135 and from strain PH1-414 which expresses wild type p135 (untagged) from its own promoter. The preparations were subjected to IMAC and analyzed by SDS-PAGE and Western immunoblotting as described under "Materials and Methods." When the blots are analyzed with antiserum to eIF3, a number of bands corresponding to eIF3 subunits (135, 110, 90, 39, and 33 kDa) are detected with strain PH1-135 which are not present with PH1-414 (results not shown). To demonstrate that these bands correspond to eIF3 subunits, the blot was probed with antibodies affinity purified against recombinant p135, p110, p39, and p33, and with an antiserum specific to p90 (Fig. 8). Each of these proteins is found in the eluate fractions from the (His) 6 -tagged p135 strain, PH1-135, but barely (e.g. p39) or not at all in the eluate fractions from the non-tagged strain PH1-414 (labeled Wt in the figure). The faint p110 immunoreactive band with PH1-135 suggests either partial degradation during purification or dissociation from the eIF3 complex under the conditions used. However, an association of p135 with the eIF3 complex is not always seen, as immunoprecipitation of (His) 6 -p135 (Fig.  6A, lane 5) does not bring down other eIF3 subunits. The relationship of p135 to eIF3 is discussed below.
Interactions of p110 or p135 with Other eIF3 Subunits and Initiation Factors-To obtain further evidence for the presence of p135 and p110 in the eIF3 complex, and to generate information leading to a structural model of eIF3, we attempted to define subunit-subunit interactions by two methods: far Western blotting and two-hybrid analyses. Because depletion of p110 leads to depletion of p90 as well, it seemed likely that these two proteins interact directly in the eIF3 complex. We therefore asked whether p110 might bind to p90 directly, rather than through other subunits in the eIF3 complex. As probes for the far Western blot analyses, p110 and p90 were synthesized in vitro in reticulocyte lysates in the presence of [ 35 S]methionine as described under "Materials and Methods." Full-length p110 and p90, as well as numerous smaller proteins likely due to degradation during the translation incubation, are seen when analyzed by SDS-PAGE and autoradiography (Fig. 9A). The radiolabeled p110 and p90 proteins were used to probe eIF3 subunits fractionated by SDS-PAGE. The eIF3 complexes were purified from strain HV110-33 by IMAC and Superdex chromatography as described above (Fig. 7B). Radiolabeled p110 binds only to p90 (Fig. 9B, lane 2), whereas radiolabeled p90 binds most strongly to p110 (lane 1), confirming the interaction. p90 also binds to another protein of ϳ95 kDa, possibly a degradation product of p110, and less strongly to p39. Interactions of p90 with p39 and p110 also have been detected by two-hybrid analyses and by glutathione S-transferase pull-down experiments (29,30).
Possible interactions of p135 with other eIF3 subunits were sought by far Western blotting. Radiolabeled p135 binds to recombinant p33 (Fig. 9C, lane 3), but not to recombinant p39 or p62 under these conditions (results not shown). To further substantiate this interaction, the two-hybrid system was employed as described in the legend to Table III. A positive interaction is seen between p135 and p33 (Table III), but not with p93 or p39 although the fusion proteins are expressed as detected by Western immunoblotting (not shown). Thus a p135-p33 interaction is confirmed.
Because mammalian eIF4B has been shown by far Western blotting to bind to the p170 subunit of human eIF3 (31), we asked whether or not yeast eIF4B might bind to the yeast homolog of human p170, namely p110. Radiolabeled eIF4B (Fig. 9A) was used to probe the eIF3 complex that contains (His) 6 -p110, but no binding was detected (Fig. 9B, lane 3). Instead, labeled eIF4B binds to recombinant p33 (Fig. 9C, lane  1), but not to recombinant p39, p93, or p110 (results not shown). The binding interaction also is seen in a two-hybrid analysis (Table III). The eIF4B-p33 interaction appears to depend on the C-terminal 71 amino acid residues of p33 which contain the RNA recognition motif (29), as the mutant form lacking this region fails to bind to eIF4B (Fig. 9C, lane 2). The result suggests that eIF4B binds to the C-terminal region of p33. We cannot rule out the possibility, however, that the C-terminal deletion of p33 causes misfolding of other parts of the protein and thereby its loss of eIF4B binding activity.

DISCUSSION
We report here the cloning and analysis of TIF31 and TIF32 encoding p135 and p110, the largest and second largest proteins associated with eIF3. The conclusion that TIF31 encodes the 135-kDa protein found in eIF3 preparations is based on 4 major lines of evidence. 1) The encoded protein contains regions that match 51 out of 55 residues of 5 sequenced peptides derived from the 135-kDa protein in purified eIF3. 2) The calculated molecular mass of the protein encoded by TIF31 (145,165 Da) is consistent with the apparent mass of 135 kDa determined by SDS-PAGE. 3) Antibodies in an antiserum raised against purified eIF3 recognize a polypeptide of 135 kDa in crude extracts prepared from TIF31-expressing cells but not in extracts of cells lacking the TIF31 gene. Affinity-purified antibodies raised against recombinant p135 give the same result (data not shown). 4) When (His) 6 -tagged p135 is expressed in cells where the chromosomal copy of TIF31 is disrupted, the protein that is immunoreactive with anti-p135 antibodies possesses an apparent mass of 135 kDa. We conclude that the cloned TIF31 gene encodes the 135-kDa protein found in our eIF3 preparations.
The role of p135 in yeast cells is not known. TIF31 deletion/ disruption displays no obvious phenotype when cells are grown at temperatures lower or higher than 30°C (data not shown) and thus is not an essential gene. Furthermore, analysis of polysome profiles of strain PH135H-B carrying the disrupted gene reveals no defect in protein synthesis (results not shown). It is noteworthy that p135 is not related to any subunit of mammalian eIF3, but is 27% identical and 50% similar to a 150-kDa protein encoded in the Dictyostelium genome. Recently, Zhu et al. (32) isolated the Dictyostelium gene and named it cluA. Disruption of cluA impairs cytokinesis and results in the clustering of mitochondria near the cell center. While this manuscript was in preparation, Fields et al. (33) cloned the S. cerevisiae homologue of Dictyostelium cluA and named it CLU1. CLU1 and TIF31 are identical. Consistent with the results reported here, deletion of CLU1 from S. cerevisiae does not affect cell viability even when cells are exposed to stress conditions such as heat and osmotic shock. However, disruption does result in the clustering of mitochondria as was observed in Dictyostelium. It is difficult to envision a mechanism whereby eIF3 directly affects cytokinesis and the clustering of mitochondria. The results suggest instead that p135 has a role in cells other than, or in addition to, initiation of protein synthesis. Nevertheless, there is strong evidence that p135 can associate with eIF3: it is present in highly fractionated eIF3 preparations; it co-purifies with other eIF3 subunits by IMAC when either p135 (Fig. 8) or p33 (15) is (His) 6 -tagged; and it binds specifically to p33 as determined by far Western blotting and two-hybrid analyses. In effect, the precise physiological function of p135 and why it is sometimes found associated with the eIF3 complex are unclear.
Evidence that TIF32 encodes the p110 subunit of eIF3 also is strong. 1) The sequence of two internal peptides from p21 match precisely the C-terminal region of the putative product of TIF32. Although no 110-kDa protein is detected in our initial preparations of eIF3 (11), the 21-kDa protein in such preparations is derived from the p110 subunit through partial prote- FIG. 9. Far Western blot analyses with eIF3 subunits and eIF4B. Panel A, analysis of the radiolabeled probes. In vitro transcription/translation reactions were employed to label p90, p110, p135, and eIF4B as described under "Materials and Methods." A 2-l aliquot of each translation reaction was fractionated by SDS-PAGE, proteins were blotted on a polyvinylidine difluoride membrane, and radiolabeled products were detected by autoradiography. Migration positions of molecular mass markers are indicated on the left. Panel B, eIF3 purified by IMAC and Superdex 200 size exclusion chromatography as described in the legend to Fig. 7 was subjected to SDS-PAGE and blotted on a polyvinylidine difluoride membrane. The blot was probed with the 3 different in vitro labeled proteins as described under "Materials and Methods." Migration positions of eIF3 subunits are indicated on the right. The panel shows a scan of the autoradiogram. Panel C, radiolabeled eIF4B was used to probe equal amounts of purified recombinant p33 (lane 1) and p33⌬C (lane 2) and radiolabeled p135 was used to probe recombinant p33 (lane 3) as described above. A scan of the autoradiogram is shown. Coomassie Blue-stained gels with purified recombinant p33 and p33⌬C have been shown elsewhere (29).

TABLE III
Two hybrid interactions within yeast eIF3 subunits An interaction between p135 and p33 was analyzed in vivo by using the MATCHMAKER two-hybrid system 2 (Clontech). pNo-NH33 (15) was digested with NcoI and SalI and the 0.9-kb fragment containing the TIF35 coding sequence was ligated to pAS2-1 to yield pAS-TIF35. pACT-TIF31 was constructed by ligating simultaneously NcoI/NruI and NruI/BamHI fragments of the TIF31 coding sequence into pACT2. pACT-4B was generated by PCR amplication of the TIF3 coding region from genomic DNA and insertion into pACT2. pVA3-1 and pTD1-1 are positive control vectors provided by the manufacturer. All transformations were done in S. cerevisiae strain Y190 (Clontech). Interactions were detected by plating on S.D.-Leu-His-Trp media containing 30 mM 3-aminotriazole (3-AT). The colony-lift protocol was used for qualitative analysis of ␤-galactosidase activity according to the manufacturer's protocol.
olysis, likely during the purification process. 2) TIF32 encodes a putative protein of 110 kDa which is present in eIF3 preparations prepared rapidly by IMAC. 3) Antibodies affinity purified from anti-eIF3 with recombinant p110 expressed from TIF32 recognize a 110-kDa protein in eIF3 rapidly purified by IMAC. Taken together with other experiments discussed below, the data clearly establish that the 110-kDa protein in eIF3 is encoded by TIF32.
In contrast to p135, the p110 subunit plays an important role in protein synthesis. The protein is essential for growth of yeast, and depletion results in slowed growth and an inhibition of protein synthesis. Polysome profile analyses of p110-depleted cells indicate that p110 is required for efficient initiation, consistent with its being a subunit of eIF3. Similar effects are observed when other eIF3 subunits are depleted (14,15).
Considerable additional evidence has been generated to demonstrate that p110 is a component of the eIF3 complex. Coimmunoprecipitates from lysates containing epitope-tagged p110 contain p33 and p90 previously characterized as eIF3 subunits. In addition, precipitates from lysates with other tagged eIF3 subunits contain p110. The p110 subunit interacts with p90 as shown by far Western blotting experiments. Depletion of p110 causes degradation of p90 as well, consistent with a direct interaction within the eIF3 complex. Furthermore, the p110 subunit is homologous to the p170 subunit of mammalian eIF3 (5), thus supporting the notion that it is present in yeast eIF3. In work published while this manuscript was in preparation, the p110 subunit was identified by mass spectroscopic analysis of preparations of eIF3 obtained from cells expressing (His) 6 tagged p90 (16) and was shown to interact with p90 by glutathione S-transferase pull-down experiments (30).
The experiments above establish that p110 is an integral part of the eIF3 complex and that it plays an important role in the initiation phase of protein synthesis. A gene identical to TIF32, called RPG1, was cloned independently through a screen for yeast proteins that cross-react with antibodies to mammalian microtubule-associated protein 2 (34). The protein product of RPG1 was not characterized then, but analysis of a temperature-sensitive rpg1-1 allele demonstrated that cells shifted to nonpermissive temperature arrest in the G 1 phase (34). A similar phenotype was observed for mutant alleles of PRT1 (encoding p90) (35) and TIF34 (encoding p39) (29). These data suggest that a properly functioning eIF3, presumably providing efficient translation initiation and protein synthesis, is required for progression through START and the cell cycle (36,37). In a recent publication (38), RPG1 was further characterized by employing the temperature-sensitive mutant to show that initiation of protein synthesis is diminished. Their results and the results reported here are entirely compatible and support and extend each other.
The composition of yeast eIF3 has been controversial, as different preparations appear to contain different subunits with different apparent masses (11,39). (His) 6 -tagging of eIF3 subunits followed by IMAC has shed light on the composition of eIF3. When (His) 6 -tagged p90 is employed, a complex is isolated that contains p33, p39, p90, p93, and p110 as well as eIF5 (16). This may represent a core complex of eIF3, as it possesses activity in vitro. In the IMAC and immunoprecipitation experiments reported here, tagged p33, p39, or p110 lead to copurification of the same 5 subunits, except for p93 which is known to be unstable in strains derived from W303 (17) and presumably is degraded in the lysates examined. It is noteworthy that only the 5 core subunits have homologs in mammalian eIF3 (see below).
The results presented above indicate that p135 also may associate with the eIF3 complex, albeit likely with lower affin-ity than the core subunits. Since p16 and p62 are co-immunoprecipitated with antibodies specific for other eIF3 subunits (12,13), these proteins may be present in eIF3 as well, although p62 is not required for eIF3 activity in vitro (16). We lack suitable antibodies to p16, p62, and p93, and thus are not able to determine unambiguously whether or not they are present in our (His) 6 -tagged preparations. Taking into account all of the evidence available, we believe that yeast eIF3 may contain up to 8 subunits, although the p16, p62, and p135 subunits appear not to be as firmly associated with the complex as the core subunits. The more loosely associated subunits may not play essential roles in all aspects of eIF3 function.
The 5 core subunits in yeast eIF3 are homologous with 5 proteins identified as subunits of mammalian eIF3: yeast p110 is homologous with mammalian p170 (29% identity); p93, with p110 (31%); p90, with p116 (31%); p39, with p36 (46%); and p33 with p44 (33%). Thus yeast and mammalian eIF3 subunit structures are conserved, although more weakly than most of the other initiation factors (1). This conservation is reflected in the ability of yeast eIF3 to substitute for mammalian eIF3 in an in vitro initiation assay (11). On the other hand, there is considerable structural diversity in that the less tightly bound yeast subunits are not related to mammalian eIF3 subunits (although yeast p16 is homologous with mammalian eIF1). Since the overall eIF3 structures appear to be related, interactions of eIF3 with other translational components might be conserved as well. One such interaction concerns mammalian eIF4B, which binds to the p170 subunit of eIF3 (31). The interaction involves the DRYG-rich motif in the central region of eIF4B as demonstrated by far Western blotting. It has been postulated that the C-terminal repeat region of p170 also may be involved in this interaction (5). Our attempts to detect a yeast eIF4B interaction with yeast p110, the homolog of mammalian p170, were not successful. Close examination of the yeast and mammalian eIF4B sequences (40,41) indicates that the two proteins are poorly conserved (17% identity), especially in the DRYG region implicated in the interaction with eIF3. Furthermore, yeast p110 differs considerably from the mammalian p170 subunit. The N-terminal third of yeast p110 shares 35% sequence identity with eIF3-p170, but the C-terminal two-thirds is less well conserved (20% identity) and the C-terminal repeat region of human p170 is lacking entirely in yeast p110. Our two-hybrid and far Western blotting results indicate that yeast eIF4B interacts instead with the p33 subunit of eIF3. Thus the interaction of yeast eIF4B with eIF3 may differ from that occurring in mammalian cells. Along similar lines, mammalian eIF4G binds to eIF3, but no such interaction has been demonstrated in yeast although numerous attempts have been made to do so. Further work is required to determine the three-dimensional structure of eIF3 in both yeast and mammals and to define the subunits responsible for its interaction with other translational components. Until then, caution must be exercised when extrapolating results from one system to the other.