Characterization of cDNAs Encoding the p44 and p35 Subunits of Human Translation Initiation Factor eIF3*

Eukaryotic translation initiation factor 3 (eIF3) is a large multisubunit complex that plays a central role in the initiation of translation. It binds to 40 S ribosomal subunits resulting in dissociation of 80 S ribosomes, stabilizes initiator methionyl-tRNA binding to 40 S subunits, and is required for mRNA binding. eIF3 has an aggregate molecular mass of ∼600 kDa and comprises at least 10 subunits. The cDNAs encoding eight of the subunits have been cloned previously (p170, p116, p110, p66, p48, p47, p40, and p36). Here we report the cloning and characterization of human cDNAs encoding two more subunits of human eIF3, namely eIF3-p44 and eIF3-p35. These proteins are immunoprecipitated by affinity-purified anti-eIF3-p170 antibodies, indicating they are components of the eIF3 complex. Far Western analysis shows that eIF3-p44 interacts strongly and specifically with the eIF3-p170 subunit, and weakly with p116/p110, p66, p40, and itself. eIF3-p44 contains an RNA recognition motif near its C terminus. Northwestern blotting shows that eIF3-p44 binds 18 S rRNA and β-globin mRNA. Possession of cloned cDNAs encoding all 10 subunits of eIF3 provides the tools necessary to elucidate the functions of the individual subunits and the structure of the eIF3 complex.


From the Department of Biological Chemistry, School of Medicine, University of California, Davis, California 95616
Eukaryotic translation initiation factor 3 (eIF3) is a large multisubunit complex that plays a central role in the initiation of translation. It binds to 40 S ribosomal subunits resulting in dissociation of 80 S ribosomes, stabilizes initiator methionyl-tRNA binding to 40 S subunits, and is required for mRNA binding. eIF3 has an aggregate molecular mass of ϳ600 kDa and comprises at least 10 subunits. The cDNAs encoding eight of the subunits have been cloned previously (p170, p116, p110, p66, p48, p47, p40, and p36). Here we report the cloning and characterization of human cDNAs encoding two more subunits of human eIF3, namely eIF3-p44 and eIF3-p35. These proteins are immunoprecipitated by affinity-purified anti-eIF3-p170 antibodies, indicating they are components of the eIF3 complex. Far Western analysis shows that eIF3-p44 interacts strongly and specifically with the eIF3-p170 subunit, and weakly with p116/p110, p66, p40, and itself. eIF3-p44 contains an RNA recognition motif near its C terminus. Northwestern blotting shows that eIF3-p44 binds 18 S rRNA and ␤-globin mRNA. Possession of cloned cDNAs encoding all 10 subunits of eIF3 provides the tools necessary to elucidate the functions of the individual subunits and the structure of the eIF3 complex.
Translational control plays an important role in the regulation of gene expression in eukaryotes. The initiation phase of translation is one of the points at which changes in the rate of protein synthesis occur (1). Initiation of translation begins with dissociation of 80 S ribosomes into 40 and 60 S subunits. The 40 S subunit then binds a ternary complex consisting of eukaryotic initiation factor 2 (eIF2), 1 GTP, and methionyl-tRNA i (Met-tRNA i ). This 40 S preinitiation complex recognizes the m 7 G-capped 5Ј end of a mRNA, binds to the mRNA, and scans toward the 3Ј end until it forms a stable complex at the first AUG initiation codon. Subsequently, the 60 S subunit joins to form the 80 S initiation complex.
The various reactions in the initiation pathway are promoted by 11 or more soluble proteins called eIFs (2,3). One of these, eIF3, plays a central role in the process. It binds to 40 S ribosomal subunits, thereby preventing 60 S association and promoting dissociation of 80 S ribosomes (4). It stabilizes Met-tRNA i binding to 40 S subunits and contributes to mRNA binding (5,6) through its interaction with the eIF4G subunit of the mRNA m 7 G-cap binding protein complex, eIF4F (7) and with eIF4B (8). eIF3 also may be involved in the recognition of the initiation codon (9) and the GTPase activity of eIF2 activated by eIF5 (10). Therefore, elucidating the structure and function of eIF3 is essential for understanding the pathway and regulation of initiation.
eIF3 is the largest of the eukaryotic translation initiation factors, with an apparent mass of about 600 kDa and a shape resembling a flat triangular prism (11). Initial characterizations of mammalian eIF3 were based on biochemical methods applied to purified preparations from rabbit reticulocytes and HeLa cells (12)(13)(14). eIF3 also has been purified from the budding yeast, Saccharomyces cerevisiae (15,16), as well as from numerous other species. The human factor has 10 or more different subunits named according to their apparent masses as determined by SDS-PAGE: p170, p116, p110, p66, p48, p47, p44, p40, p36, and p35. More recently, the cDNAs encoding eight of the subunits of human eIF3 have been cloned and sequenced: p170 (17), p116 (18), p110 and p36 (19) p48 (20), and p66, p47, and p40 (21). Similarly, the genes for the eight subunits of yeast eIF3 have been identified (22). The cloning of the cDNAs or genes encoding eIF3 subunits provides structural information about the factor and enables researchers to develop tools to better characterize its structure/function. In this paper, we describe the cloning and characterization of the cDNAs encoding eIF3-p44 and eIF3-p35, provide evidence that these subunits are part of the eIF3 complex, and show that the p44 subunit, which contains an RNA recognition motif (RRM), binds RNA. This report completes descriptions of the cloning of cDNAs encoding the 10 eIF3 subunits, thereby providing a firm base for further structural work on this important initiation factor.

EXPERIMENTAL PROCEDURES
Cloning of p44 and p35-eIF3 was purified from HeLa S3 cells essentially as described (14) and was subjected to SDS-PAGE. Bands corresponding to the p44 and p35 subunits were excised and digested with Lys-C protease in the gel. Following high performance liquid chromatography fractionation, N-terminal sequences of internal peptides were obtained by automated Edman degradation in the Protein Structure Laboratory (University of California, Davis). The sequences (shown by black overlines in Figs. 1 and 4) were used to search for matching sequences in the National Center for Biotechnology Information EST (expressed sequence tag) data base using the TBLASTN program (23).
Mouse ESTs matching p44 peptide sequences were found and used to identify overlapping EST clones (GenBank accession numbers * This work was supported in part by National Institutes of Health Grant GM22135 from the United States Public Health Service. 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 AA109090, AA270800, and W18370), and a hypothetical cDNA for mouse eIF3-p44 was assembled. The deduced amino acid sequence was used in a TBLASTN search to identify matching human EST clones. The longest clone (identification no. 293583) was purchased from ATCC (Rockville, MD) as a 1,237-bp insert in pT7T3D-PAC and sequenced. The length of the open reading frame and comparison to the closely related mouse cDNA sequence suggested that the clone lacked about 100 codons from the N-terminal coding region. To obtain the missing coding sequences, 5Ј-rapid amplification of cDNA ends (5Ј-RACE) was used with a Marathon Ready cDNA library from pancreas (CLON-TECH) as template and primers AP1 (CLONTECH) and 1 (Table I). A PCR product of ϳ650 bp containing the amplified 5Ј end was ligated into pNoTA (5 Prime 3 3 Prime, Inc., Boulder, CO) and sequenced. To obtain full-length cDNA for eIF3-p44, the insert in pNoTA was cut with ApaI and HindIII and the ApaI/HindIII fragment from pT7T3D-PAC was inserted into the pNoTA vector to yield pNo44 -7. The sequence was submitted to GenBank as human eIF3-p44 cDNA (accession number U96074).
For eIF3-p35, searching the EST data base with the C-proximal peptide (Fig. 4) identified a human clone with 12 of 15 amino acids matching, namely Z25214 (clone identification no. HSBA6A022). By searching with the cDNA sequence, a human clone, W05832 (clone identification no. 299811), was identified and purchased from ATCC as an insert in pT7T3D-PAC. The 2,223-bp insert was sequenced on both strands. To obtain a full-length coding region, 5Ј-RACE was employed as described above with primers AP1 and 5 (Table I) to yield DNA encoding the 5Ј region. PCR was used with primers AP2 (CLONTECH) and 5, and with primers 6 and M13 univ, to amplify 530 and 2,088 bp overlapping DNAs, which together extend from the 5Ј end of the RACE product to the 3Ј-terminus of the insert in pT7T3D-PAC. Amplification of the two PCR products (100 -200 ng each) resulted in a 2,443-bp PCR product that was ligated into pNoTA to yield pNo35F2. The sequence was submitted to GenBank as human eIF3-p35 cDNA (accession number U97670).
Northern Blotting-Total RNA was isolated from exponentially growing HeLa S3 cells using the RNeasy Midi total RNA isolation kit (Qiagen, Santa Clarita, CA). RNA was subjected to 1.2% agarose gel electrophoresis (24) and blotted onto a nitrocellulose membrane using a 1-h downward capillary blot procedure (25). Probes corresponding to the coding regions of eIF3-p44 (bp 420 -829) and eIF3-p35 (bp 307-824) were obtained by PCR amplification and were labeled with [␣-32 P]dATP (NEN Life Science Products) using the Multiprime DNA labeling kit (Amersham Pharmacia Biotech). The blots were pre-hybridized 4 h at 42°C in hybridization buffer (6ϫ SSC, 50% formamide, 5ϫ Denhardt's, 0.5% SDS), hybridized overnight (hybridization buffer plus 10 g/ml salmon sperm DNA and ϳ1 ϫ 10 7 cpm probe), and washed once in 1ϫ SSC and 0.1% SDS at room temperature for 15 min, twice in 1ϫ SSC and 0.1% SDS at room temperature for 15 min, and twice in 0.25ϫ SSC and 0.1% SDS at 42°C for 15 min. RNA binding was visualized by autoradiography. The 0.24 -9.5-kb RNA ladder was purchased from Life Technologies, Inc.
Overexpression of p44 and p35 in Escherichia coli-For expression of His-tagged p44 and p35 in E. coli, the initiation codon regions were modified via PCR to contain the restriction site NdeI at the 5Ј end (p44, primer 2; p35, primer 7; see Table I) and XhoI (p44, primer 4) or HindIII (p35, primer 9) restriction sites downstream of the stop codon, respectively. PCR-amplified full-length DNAs were digested (NdeI/XhoI for p44 and NdeI/HindIII for p35) and subcloned into pET28c (Novagen, Madison, WI) to generate pET-NHp44 and pET-NHp35. E. coli strain BL21(DE3) was transformed with each of the plasmids, and synthesis of the tagged protein was induced with 1 mM isopropyl-␤-D-thiogalactopyranoside. N(His) 6 -p44 and N(His) 6 -p35 were purified from the corresponding lysates using His-bind resin (Novagen) under native conditions as recommended by the manufacturer.
In Vitro Transcription/Translation of cDNAs-The 5Ј ends of the inserts in pNo44 -7 and pNo35F2 were modified via PCR to contain NcoI sites at their initiation codons (p44, primer 3; p35, primer 8) and XhoI and HindIII sites, respectively, at the 3Ј ends (as described above). When subcloned into the NcoI and XhoI or HindIII sites of pET28c, no (His) 6 tag is added, but the upstream T7 RNA polymerase promoter allows for efficient transcription with T7 RNA polymerase. One microgram each of the resulting plasmids, pET-T7p44 and pET-T7p35, was transcribed and translated in vitro at 30°C for 60 min with the T7 TNT coupled reticulocyte lysate system (Promega, Madison, WI) containing translation grade [ 35 S]methionine (NEN Life Science Products). Radiolabeled products (1-2-l aliquots) were analyzed by 10% SDS-PAGE, followed by blotting onto polyvinylidene fluoride (PVDF) membranes (Immobilon P, Millipore, Bedford, MA) and exposure to X-Omat film for 16 -24 h.
Far Western Analysis-The untagged or N(His) 6 -p44 and N(His) 6 -p35 proteins were radiolabeled with the in vitro transcription and translation system described above. HeLa lysate, purified human eIF3, and individual recombinant eIF3 subunits were fractionated by 10% SDS-PAGE and transferred to PVDF membranes. The membranes were washed repeatedly with Protein Binding Buffer (PBB ϭ 20 mM HEPES/KOH (pH 7.4), 50 or 150 mM NaCl, 5 mM MgCl 2 , 1 mM EDTA, 0.1 mM ZnCl 2 , 1 mM dithiothreitol, 10% glycerol), incubated for 3 h with PBB containing 5% bovine serum albumin at 4°C, and probed overnight at 4°C with the radiolabeled reticulocyte lysates in 10 ml of PBB containing 3% bovine serum albumin. Blots were washed eight times for 10 min each with 10 ml of 0.1% Nonidet P-40 in PBB (28), dried, and exposed to film.
Co-immunoprecipitation-HeLa S3 cells from six 100-mm tissue culture dishes were lysed in 1 ml of lysis buffer (described above) and clarified. Protein G-Sepharose beads (Amersham Pharmacia Biotech) were pre-incubated at 4°C for 1 h with the appropriate antibody in lysis buffer. After removing unbound antibody, 500 g of the HeLa lysate was incubated with 10 l of the pre-incubated protein G beads for 2 h at 4°C in a total volume of 200 l. Beads were removed from the "supernatant" by centrifugation and washed three times with an equal volume of 300 mM KCl lysis buffer. Bound proteins ("eluate") were obtained by heating at 90°C for 5 min in 20 l of SDS-sample buffer. Proteins in the supernatant were concentrated by precipitation with 10% trichloroacetic acid. The eluate and aliquots of supernatant containing 75 g of protein were subjected to SDS-PAGE and Western immunoblotting as described above.
Northwestern Blot Analysis-Full-length human rDNA was PCRamplified with primers 10 and 11 (Table I) from genomic DNA prepared from BJAB cells (a B cell lymphoma cell line). The PCR product was cloned into pNoTA to yield pNo18Sh, sequenced from both ends, and a 1.8-kb XhoI/Asp718I fragment was subcloned into the corresponding sites of pSP73 (Promega, Madison, WI) to yield pSP18Sh. This plasmid was linearized with Asp718I for in vitro transcription with T7 RNA polymerase (Maxiscript SP6/T7 kit, Ambion, The Woodlands, TX) and [␣-32 P]UTP. In addition to full-length 18 S rRNA, the transcript carries eight additional nucleotides at the 5Ј end encoded by pSP73. Xenopus ␤-globin mRNA was prepared as described previously (29) and transcribed with SP6 as described above. Purified eIF3 (ϳ5 g), rc-eIF3-p44, and rc-eIF3-p66 were subjected to 10% SDS-PAGE and electrotransferred to PVDF. The membrane was prehybridized in binding CCCGGTACCTAATGATCCTTCCGCAGGTTCACC buffer (20 mM HEPES-KOH, 75 mM KOAc, 2 mM Mg(OAc) 2 , 1 mM EDTA, 1 mM dithiothreitol, 0.2% (w/v) CHAPS (pH 7.5)) plus 1 mg/ml yeast tRNA for 20 min. The blot was then hybridized in binding buffer plus yeast tRNA and RNase inhibitor with the 32 P-labeled 18 S rRNA or ␤-globin mRNA (10 7 cpm) for 30 min, washed three times for 5 min each in binding buffer, and subjected to autoradiography, and subsequently to staining or Western immunoblotting.

RESULTS
Cloning and Characterization of Human eIF3-p44 cDNA-To clone the cDNA encoding the p44 subunit of eIF3, a preparation of purified HeLa eIF3 was fractionated by SDS-PAGE and the band corresponding to eIF3-p44 was excised and used to obtain partial peptide sequences. As described in detail under "Experimental Procedures," the peptide sequences were used to search the EST data base at the National Center for Biotechnology Information, and finally, three overlapping human ESTs encoding one of the partial peptides were identified. The longest of these, clone 293583, was sequenced but lacked DNA encoding the N-terminal ϳ100 amino acids, as concluded from comparison to the mouse protein sequence deduced from overlapping ESTs. The missing 5Ј-region of the cDNA was obtained by the 5Ј-RACE procedure with a human pancreas cDNA library to yield DNA that codes for the N-terminal region and the putative initiator codon. DNA encoding the entire eIF3-p44 was constructed in pNoTA to yield pNo44 -7, and the nucleotide sequence was deposited in the GenBank data base.
The cloned cDNA insert in pNo44 -7 comprises 1,115 bp with a 963-bp open reading frame that encodes a putative 35,694 Da protein of 320 amino acids (Fig. 1). The two peptide sequences mentioned above are found within the p44 sequence (identified with black overlines in Fig. 1). The 5Ј-UTR of the cDNA contains 9 bp, whereas the 3Ј-UTR contains 96 bp plus another 47 A residues that are preceded by the sequence AATAA 20 bp upstream. This sequence presumably serves as the polyadenylation signal. Northern blot analysis of HeLa total RNA identifies a major RNA species migrating at ϳ1.4 kb ( Fig. 2A), indicating that the cloned cDNA is not full-length, lacking a substantial portion likely at the 5Ј end. Nevertheless, the putative AUG initiation codon (residues 10 -12) is very likely correctly identified, as the next in-frame AUG codon is far downstream, corresponding to amino acid residue 128. In addition, a nearly identical mouse cDNA (GenBank accession no. AA109090) corresponding to the 5Ј end of eIF3-p44 DNA contains an in-frame termination codon 4 codons upstream from the putative initiator AUG.
Comparison of the human eIF3-p44 amino acid sequence with translated overlapping mouse EST sequences, using the GAP program (GCG, Madison, WI), indicates that the mouse and human proteins share 97% amino acid sequence identity (Fig. 1). The human eIF3-p44 amino acid sequence is also homologous to the eIF3-p33 subunit from the yeast S. cerevisiae, sharing 33% sequence identity and 42% similarity. As seen from Fig. 1, eIF3-p44 shares a fair amount of sequence The amino acid sequence of human eIF3-p44 is aligned with its yeast homolog, yeIF3-p33 (GenBank accession no. AF004913), its mouse homolog (accession nos. AA109090, AA270800, and W18370), the homolog in S. pombe (accession no. AB011823), and the homolog in C. elegans (accession no. Z50044). The alignment was conducted with the Clustal X program and modified using Boxshade. Residues identical in three or more of the five species are shown with a black background, whereas three or more similar residues are shown with a gray background. The black bars over the sequence identify the peptides sequenced; the light gray bars correspond to the RNP1 and RNP2 motifs. identity and similarity with its homologs in Schizosaccharomyces pombe and Caenorhabditis elegans, where highly conserved regions are potentially important for function or structure. A Prosite search identifies several potential phosphorylation sites and a consensus RRM (30) in the C-terminal quarter of the protein. The RNP1 and RNP2 motifs are located between residues 280 -287 and 241-246, respectively, and are identified with gray lines above them in Fig. 1.
Two experimental approaches were used to demonstrate that the cloned cDNA encodes the p44 subunit of eIF3. The coding region of pNo44 -7 was inserted into pET28c (an E. coli expression vector) such that the final product contains an N-terminal 6-histidine tag. As described under "Experimental Procedures," N(His) 6 -p44 was expressed and the tagged protein was purified by Ni 2ϩ affinity chromatography. The resulting recombinant tagged p44 subunit was used to affinity-purify anti-p44 antibodies present in crude goat serum containing anti-eIF3 antibodies. After SDS-PAGE and immunoblotting of purified eIF3 and HeLa lysate, the affinity-purified anti-p44 antibodies generate a band corresponding to the p44 subunit in purified eIF3 (Fig. 3, A and B, lane 2) and a weak band with the same mobility in a HeLa lysate (data not shown). Furthermore, coupled in vitro transcription of pET-T7p44 (no His 6 tag) with T7 RNA polymerase and translation in a rabbit reticulocyte lysate with [ 35 S]methionine, results in a radiolabeled protein of 44 kDa that comigrates with eIF3-p44 (Fig. 3C, lane 1). Together with the finding that eIF3-p44 is homologous to yeast eIF3-p33, these results provide strong evidence that the cloned cDNA encodes eIF3-p44.
Cloning and Characterization of a cDNA Encoding Human eIF3-p35-When eIF3 was first purified from rabbit reticulocytes in the mid-1970s, the p35 subunit was recognized as a protein only weakly bound to the complex (12). To determine whether or not p35 is a true subunit of eIF3, we sought to clone its cDNA by reverse genetics as described in detail under "Experimental Procedures." A 2.4-kb DNA was obtained that contains a long open reading frame and putative AUG initiation codon. A 32 P-labeled 517-bp probe derived from the coding region of the DNA sequence (bp 367-898) was used in Northern blot analysis. A major RNA species migrating at ϳ2.4 kb hybridizes to the probe (Fig. 2B). This suggests that the constructed cDNA encodes essentially the entire mRNA. The cDNA was inserted into pNoTA to yield pNo35F2. pNo35F2 contains a 2,443-bp insert with a 774-bp open reading frame that encodes a putative 28,989-Da protein of 258 amino acids (Fig. 4). All three sequenced peptides (indicated by black bars in Fig. 4) are found within the amino acid sequence. The 5Ј-UTR of the cDNA insert contains 60 bp, whereas the 3Ј-UTR is unusually long, containing 1,584 bp. The insert ends with a string of 21 A residues preceded by two overlapping potential polyadenylation signals (AATAA) 22-25 bp upstream. The nucleotide sequence was deposited in the GenBank data base as human eIF3-p35 cDNA.
Identification of the initiation codon indicated in Fig. 4 is based on the fact that the next in-frame AUG occurs 148 codons downstream, whereas no in-frame AUG is found upstream in the 5Ј-UTR. Furthermore, as shown below, translation originating at this AUG generates a protein that co-migrates with authentic eIF3-p35 during SDS-PAGE. A data base search at the EBI (European Bioinformatics Institute, EMBL Outstation, Hinxton, United Kingdom), using the BLITZ program and eIF3-p35 as query, identifies a yeast protein of 265 amino acids (SwissProt identification no. Q05775), which shows 26.7% identity and 42% similarity to p35 when analyzed with the GCG GAP program. The uncharacterized yeast protein is similar to GVPD (gas vesicle protein D) from Halobacterium halobium. A Prosite search of eIF3-p35 identifies several potential phosphorylation sites (not shown).
To prove that the cloned eIF3-p35 cDNA encodes the p35 subunit, the coding region was inserted into the pET28c vector, expressed in E. coli, and the (His) 6 -tagged protein was Ni 2ϩ affinity-purified. The recombinant protein was then used to affinity-purify anti-p35 antibodies present in the crude goat serum containing anti-eIF3 antibodies. Immunoblot analysis of both purified eIF3 and HeLa lysate (Fig. 3B, lanes 3 and 4,  respectively) shows a single reactive band that comigrates with the p35 subunit of eIF3. To demonstrate that the cloned cDNA contains the full-length coding region for p35, in vitro transcription/translation of the cDNA was carried out as described under "Experimental Procedures." Expression of pET-T7p35 yields a single radiolabeled protein that precisely comigrates with eIF3-p35 (Fig. 3C, lane 2). The results show that the cloned p35 cDNA encodes the p35 subunit of eIF3.
eIF3-p44 and -p35 Are Part of the eIF3 Complex-Even though the eIF3-p44 and -p35 subunits co-purify with other eIF3 subunits, it is necessary to show that they are indeed components of the eIF3 complex as opposed to contaminants that co-purify. Toward this end, we have used co-immunoprecipitation and Far Western analyses to examine whether or not p44 and p35 are part of the complex or can bind other eIF3 subunits in vitro. eIF3 complexes were immunoprecipitated from a HeLa cell lysate with anti-eIF3 serum and with affinity-purified anti-p170 antibodies. The high specificity of the affinity-purified anti-p170 antibodies is shown in Fig. 5 (lanes 8 and 9, which contain eIF3 and HeLa lysate, respectively). As shown in Fig.  5A (lane 3), the affinity-purified anti-p170 antibodies clearly immunoprecipitate p170, p116, p110, p47, p40, and maybe p35. Because the titer of antibodies to p44 is low in the crude anti-eIF3 serum, detection of p44 was problematic. Therefore, aliquots of the immunoprecipitated sample were subjected to Western blotting and incubation with affinity-purified antibodies to p44 and p35. As shown in Fig. 5, B (lane 4) and C (lane 2), affinity-purified p170 antibodies co-immunoprecipitate p44 and p35, respectively, from HeLa lysates. These results indicate that p44 and p35 are part of the eIF3 complex. The presence of p35 in both the eluate and supernatant fractions suggests that p35 partially dissociates from the eIF3 complex under the stringent assay conditions.
A demonstration that a putative subunit interacts directly with an authentic eIF3 subunit supports the view that it may be a part of the eIF3 complex. Far Western analysis was performed as described under "Experimental Procedures" to look for possible direct protein-protein interactions of p44 and p35 with eIF3 subunits. Radiolabeled p44 and p35 probes were prepared by in vitro translation and were incubated with blots containing eIF3 subunits fractionated by SDS-PAGE. As shown in Fig. 6A (lane 2), recombinant 35 S-labeled p44 binds most strongly to the p170 subunit, less so to p116/110 and p66, and most weakly to itself (p44) and p40. Comparable analysis of a HeLa lysate shows that 35 S-labeled p44 binds specifically to a 170-kDa protein (lane 3). The other bands in lane 3 also are seen when the lysate is probed with a radioactive reticulocyte lysate lacking mRNA template (lane 7) and therefore are considered artifacts. To confirm interactions of eIF3-p44 with some of the other eIF3 subunits, 35 S-labeled p44 was incubated with a membrane containing purified, recombinant, His-tagged p40 and p44 (Fig. 6B, lanes 3 and 4). The probe binds to both recombinant proteins, confirming that p44 binds both to itself and to p40 under these conditions. To check for specificity, green fluorescent protein was used as a probe against purified eIF3 (Fig. 6C, lane 1) and yielded no interactions. In addition, Novex markers were run as a negative control. For all experiments, except those using p35 as a probe (see paragraph below), no interactions are detected, as shown in Fig. 6C (lane 2).
Far Western analysis showed that 35 S-labeled p35 binds to nearly all the subunits of eIF3 (Fig. 6A, lane 4), and to a vast number of proteins in the HeLa lysate (lane 5). Analyses under more stringent conditions, such as higher salt concentrations or increasing the amount of detergent, do not result in fewer bands (data not shown). The results indicate that eIF3-p35 is a very "sticky" protein, making the Far Western blotting approach unfeasible for this subunit. Nevertheless, it is possibly significant that especially strong binding to p170 and p66 occurs, whereas no binding to p48, p47, or p36 is detected.
eIF3-p44 Binds RNA-Since eIF3-p44 contains a consensus RRM and other eIF3 subunits and initiation factors have been shown to bind RNA, we wished to determine whether or not p44 binds RNA in vitro. To accomplish this aim, we employed Northwestern blot analysis. rc-eIF3-p66, rc-eIF3-p44, and purified eIF3 were subjected to SDS-PAGE, transferred to a PVDF membrane, and incubated with 32 P-labeled 18 S rRNA or ␤-globin mRNA. eIF3-p66 had been shown previously to bind 18 S rRNA (31) and was therefore regarded as a positive control. Proteins in purified eIF3 corresponding to p170, p66, and p44 bind to both 18 S rRNA and ␤-globin mRNA (Figs. 7, A and  B, lane 1). Recombinant p66 and p44 also bind to either of the radiolabeled probes (lanes 2 and 3). The identities of the three RNA-binding subunits seen with purified eIF3 are established by immunoblotting the membrane used in panel A with affinity-purified antibodies to eIF3-p170, -p66, and -p44 (Fig. 7C). Proteins in the gel that bind RNA also react with the appropriate antibodies. Recombinant p44 appears to bind the 18 S ribosomal RNA ϳ4 times more strongly than recombinant p66 (as calculated using a PhosphorImager (STORM 860; Molecular Dynamics, Sunnyvale, CA)), whereas rc-p44 and rc-p66 bind ␤-globin mRNA with equal intensity, comparable to what is seen when eIF3 subunits are probed (lane 1). It is noteworthy that p170 binds both RNAs, since the p170 sequence contains no discernible RNA-binding motif. The RNA binding activities appear to be specific because other proteins on the membrane do not bind to the RNA probe. However, one cannot determine absolute affinities with the Northwestern blotting procedure because this method may yield similar extents of binding even when binding affinities differ substantially. DISCUSSION We have cloned and characterized cDNAs encoding human eIF3-p44 and eIF3-p35. The following facts indicate that the cloned cDNAs indeed encode the two subunits of human eIF3. In vitro transcription and translation of the cDNAs in reticulocyte lysates result in radiolabeled protein products that comigrate in SDS-PAGE with the corresponding subunits in a purified preparation of eIF3. Antibodies from a crude anti-eIF3 goat antiserum that were affinity-purified with recombinant p44 and p35 overexpressed in E. coli specifically recognize the cognate protein in purified eIF3 and in HeLa lysates (Fig. 3B). In addition, partial peptide sequences for p44 and p35 match regions in the amino acid sequences deduced from the cloned cDNAs.
Two different approaches were used to prove that p44 and p35 are part of the eIF3 complex. First, co-immunoprecipitation of eIF3 subunits with affinity-purified antibodies to the p170 subunit was performed. The results show that both p44 and p35 are co-immunoprecipitated with the p170 antibodies. Far Western analyses were then used to demonstrate that p44 and p35 interact with other eIF3 subunits. 35 S-Labeled p44 binds strongly and specifically not only to the p170 subunit in purified eIF3, but also to a 170-kDa protein in a HeLa lysate. The labeled probe also appears to bind unresolved p116/p110, p66, and more weakly to itself and to p40 (Fig. 6A). The weak interactions were confirmed with purified recombinant p44 and p40. Further support for these findings comes from preliminary yeast two-hybrid data, which show human eIF3-p44 interacts FIG. 5. Co-immunoprecipitation of eIF3 subunits. A, 500 g of HeLa lysate was incubated with Gamma Bind G-Sepharose beads that were preincubated with anti-p170 antibodies (lanes 2 and 3), crude anti-eIF3 serum (lanes 4 and 5), or pre-immune serum (lanes 6 and  7). Supernatant and bound fractions were subjected to SDS-PAGE and immunoblotting as described under "Experimental Procedures." Lanes 2, 4, and 6 were loaded with 75 g of trichloroacetic acidprecipitated supernatant protein from the immunoprecipitation. Lanes 3, 5, and 7 were loaded with the antibody-bound proteins removed from the beads by heating in SDS-sample buffer (eluate). Purified eIF3 was run as a standard in lane 1. Samples were then processed as described for Fig. 3 and immunoblotted with crude anti-eIF3 serum. Purified eIF3 (0.5 g) and HeLa lysate (50 g) were examined to demonstrate the specificity of affinity-purified anti-p170 antibodies (lanes 8 and 9, respectively). eIF3 subunit migration positions are indicated on the left. B, purified eIF3 (500 ng, lane 2) and proteins immunoprecipitated with anti-p170 antibodies (lanes 3 and 4) were processed as described in panel A, except that the blots were incubated with anti-p44. Lane 1 contains purified eIF3 (2 g) stained with Amido Black. Lane 3 contains 75 g of supernatant protein, and lane 4 contains the eluate (proteins eluted from protein G beads with SDS sample buffer). The cross-reacting protein in lane 4 with a mobility slightly faster than p35 has not been identified. C, samples immunoprecipitated with anti-p170 antibodies were processed as described for panel B, except that the membrane was incubated with anti-p35 (lane 1, supernatant; lane 2, eluate). Lanes 3 and 4 contain purified eIF3 (500 ng) and HeLa lysate (75 g), respectively, and were immunoblotted with anti-eIF3 serum. with p170, p116, p110, p44, and p40. 2 Unfortunately, Far Western analysis is not a useful tool for studying proteins interacting with p35, because the protein is unusually sticky and binds to nearly everything on the membrane, even at high stringency.
A Prosite search with eIF3-p44 identifies several potential protein kinase C and casein kinase II phosphorylation sites, myristoylation sites, and glycosylation sites (not shown). An RRM consensus sequence (30) close to the C terminus of the protein also is apparent. Northwestern analysis shows that p44 binds 18 S rRNA and ␤-globin mRNA (Fig. 7) and therefore appears to be a nonspecific RNA-binding protein. Our Northwestern analyses show for the first time that eIF3-p170 also binds RNA even though no RNA-binding motif is discernible, and confirms that eIF3-p116, the human homolog of yeast eIF3-p90 (Prt1), does not bind RNA (18) under these conditions. eIF3-p44 is homologous with yeast eIF3-p33, with which it shares 33% sequence identity and 42% similarity, especially in the region containing the RRM (Fig. 1). Yeast eIF3-p33 binds RNA and this activity is destroyed when the RRM is deleted. 3 Although yeast eIF3 functionally replaces human eIF3 in a methionyl-puromycin synthesis assay (15), p44 does not complement a p33 yeast knock-out. 3 Similar failures to complement have been reported for the human p170 (32) and human p36 (33) subunits, suggesting that there is considerable divergence in the overall structures of yeast and human eIF3. Yeast eIF3-p33 has been shown to interact with both the p39 (homolog of human p36) and p90 (homolog of human p116) subunits (34,35), whereas here we show a strong interaction of human p44 with p170 (homolog of yeast p110), and none with p36. 4 The interaction of p44 with p116 seen by both the Far Western and two-hybrid analyses appears to be conserved, however. A precise determination of the composition and subunit stoichiometry of both human and yeast eIF3 remains to be made. The identification of subunit-subunit interactions, some of which are defined above, contributes to elucidating the structure of eIF3. Further such experiments are in progress and exploit the availability of the cloned cDNAs reported here. The blot in panel A was subsequently immunoblotted with affinitypurified anti-p170, -p66, and -p44 to confirm the identities of the RNAbinding subunits (C).