Saccharomyces cerevisiae protein Pci8p and human protein eIF3e/Int-6 interact with the eIF3 core complex by binding to cognate eIF3b subunits.

Mammalian, plant, and Schizosaccharomyces pombe eukaryotic initiation factor-3 (eIF3) contains a protein homologous to the product of int-6 (eIF3e), a frequent integration site of mouse mammary tumor viruses. By contrast, Saccharomyces cerevisiae does not encode a protein closely related to eIF3e/Int-6. Here, we characterize a novel S. cerevisiae protein (Pci8p, Yil071cp) that contains a PCI (proteasome-COP9 signalosome-eIF3) domain conserved in eIF3e/Int-6. We show that both Pci8p and human eIF3e/Int-6 expressed in budding yeast interact with the yeast eIF3 complex in vivo and in vitro by binding to a discrete segment of its eIF3b subunit Prt1p and that human eIF3e/Int-6 interacts with the human eIF3b segment homologous to the Pci8p-binding site of yeast Prt1p. These results refine our understanding of subunit interactions in the eIF3 complex and suggest structural similarity between human eIF3e/Int-6 and yeast Pci8p. However, deletion of PCI8 had no discernible effect on cell growth or translation initiation as judged by polysome analysis, suggesting that Pci8p is not required for the essential function of eIF3 in translation initiation. Motivated by the involvement of Int-6 in transcriptional control, we investigated the effects of deleting PCI8 on the total mRNA expression profile by oligonucleotide microarray analysis and found reduced mRNA levels for a subset of heat shock proteins in the pci8Delta mutant. We discuss possible dual functions of Pci8p and Int-6 in transcriptional and translational control.

Translation initiation in eukaryotes is a complex series of reactions leading to the formation of an 80 S ribosomal complex containing Met-tRNA i Met base-paired with the initiation codon in the mRNA. The largest of eukaryotic initiation factors (eIFs), 1 eIF3 serves as a scaffold for assembly of other eIFs, thereby promoting ribosome binding of a ternary complex consisting of Met-tRNA i Met , eIF2, and GTP and of mRNA in asso-ciation with the cap-binding complex eIF4F (for review, see Refs. 1 and 2). Mammalian eIF3 contains 11 non-identical subunits (3) and is very similar in subunit composition to plant eIF3, as both contain homologs of subunits eIF3a-i and eIF3k and only one additional subunit specific to each organism (4 -6). The fission yeast Schizosaccharomyces pombe encodes orthologs of nine of these conserved subunits (all except for eIF3k) (7,8), and eIF3a-c, eIF3e-g, and eIF3i have been identified in a putative eIF3 complex purified from this yeast (9). eIF3 from the budding yeast Saccharomyces cerevisiae is atypical in containing only five subunits present in stoichiometric amounts (eIF3a/ Tif32p, eIF3b/Prt1p, eIF3c/Nip1p, eIF3g/Tif35p, and eIF3i/ Tif34p) (10,11). Hcr1p, the budding yeast ortholog of human eIF3j, appears to be a non-stoichiometric peripheral component of yeast eIF3 (12). Importantly, the five-subunit complex purified from budding yeast was sufficient to restore Met-tRNA i Met binding to 40 S ribosomes in a heat-inactivated extract from a prt1 mutant and thus possesses a key activity ascribed to mammalian eIF3 (11). All five eIF3 subunits are essential in S. cerevisiae (2). For these reasons, it was proposed that the five-subunit complex isolated from budding yeast represents a catalytically active "core" complex of the critical eIF3 subunits (11). This hypothesis is consistent with the finding that, in fission yeast, two of the five core subunits are essential (eIF3g and eIF3i), whereas two of the "non-core" eIF3 subunits are dispensable (eIF3d and eIF3e) (7)(8)(9)13).
Three of the eIF3 subunits in mammals, plants, and S. pombe contain a conserved region (the PCI (26 S proteasome-COP9 signalosome-initiation factor-3) or PINT (proteasome-Int-6-NIP1-TRIP-15) domain) that also occurs in five subunits of the 19 S lid subcomplex of the 26 S proteasome and in six subunits of the COP9 signalosome (14,15). Two of the eIF3 subunits containing the PCI/PINT domain, eIF3a and eIF3c, may be regarded as core subunits (as they are present in the five-subunit budding yeast eIF3), whereas the third, eIF3e, appears to be a peripheral eIF3 subunit. The gene encoding eIF3e in S. pombe, int6 ϩ , is not essential, but strains lacking this gene grow more slowly than the wild-type strain and display moderate reductions in the rate of translation initiation (7,9). Thus, eIF3e/Int6 in S. pombe may play an accessory role in the general function of eIF3 or may be required only for translation of specific mRNAs.
In addition to its proposed role in translation as an eIF3 subunit, human eIF3e and the mouse homolog (Int-6) have been implicated in the control of cell growth and tumorigenesis as well as transcriptional regulation (16 -18). The int-6 coding sequence was found to be a frequent integration site of mouse mammary tumor virus (19), and expression of Int-6 is decreased in a consistent portion of human breast and lung carcinomas (20). Human Int-6 interacts with the Tax transactivator of human T cell leukemia virus type 1 (21). Mammalian Int-6 is localized in the promyelocytic leukemia nuclear body (21,22), whose major constituent, PML protein, is a tumor suppressor and transcriptional regulator (23). In fission yeast, eIF3e/Int6 was identified as an inducer of multidrug resistance genes dependent on AP-1 transcription factor Pap1p (8), and int6⌬ mutants display both caffeine sensitivity (7) and a defect in nuclear partitioning (9,24). Furthermore, plant eIF3e is also reported to co-purify with the COP9 signalosome (25), a multifunctional nuclear complex that may be involved in transcriptional regulation (26). Thus, there is accumulating evidence that Int-6 functions in the nucleus in the transcriptional control of gene expression, although the nuclear localization of Int-6 could still be interpreted as a means of sequestering it from eIF3 (27). In contrast, little is known about the molecular interactions of human eIF3e/Int-6 with the eIF3 core complex or its possible effect on eIF3 activity in translation.
Eight PCI proteins have been recognized in the proteome of budding yeast (14,15). These include five 26 S proteasomal subunits 2 ; the eIF3 core subunits eIF3a/Tif32p and eIF3c/ Nip1p; and a hypothetical protein encoded by YIL071c or YIH1 (15), which we designate Pci8p. The absence of a third PCI subunit in S. cerevisiae eIF3 led us to consider that Pci8p could be a distantly related ortholog of the non-core subunit eIF3e that might still interact physically or functionally with the eIF3 core complex in this yeast. Hence, we set out to analyze biochemically and genetically this last uncharacterized PCI protein in budding yeast. By comparing Pci8p with eIF3e/Int-6 in different binding reactions, we discovered that eIF3b/Prt1p is a common binding partner in the eIF3 core complex for both Pci8p and human eIF3e/Int-6. Furthermore, the minimal binding site for Pci8p in yeast eIF3b is homologous to that which we identified for human eIF3e/Int-6 in human eIF3b. This conservation of binding domains suggests a common ancestry and functional similarity between Int-6 and Pci8p.
To investigate the cellular role of Pci8p in budding yeast, we created a pci8⌬ deletion mutant and found that Pci8p is not essential and does not have any detectable effect on general translation initiation. Given its ability to interact physically with eIF3, it is still possible that Pci8p regulates translation of specific mRNAs or under restricted growth conditions. Based on the role of Int-6 in transcriptional control, we used oligonucleotide microarray analysis to compare the gene expression profiles of pci8⌬ and wild-type cells and detected changes in mRNA expression for several heat shock genes. Thus, Pci8p may be involved in transcriptional control.

MATERIALS AND METHODS
Plasmids-The oligodeoxyribonucleotides and plasmids used in this study are listed in Tables I and II, respectively. pAV1427 encodes N-terminally FLAG-and polyhistidine-tagged Gcd6p under the control of the GAL promoter (28) with a unique MluI site located between the coding regions for the N-terminal tags and the first codon of Gcd6p. The DNA segments containing the human int-6 ORF were synthesized from pTZp48 (16) by PCR using primers 1 and 2 and primers 1 and 3 (Table  I), of which primer 3 introduces a FLAG epitope coding sequence fused in-frame to the 3Ј-end of the int-6 ORF. After digestion with MluI and BamHI, the fragments were inserted into the 5Ј-MluI and 3Ј-BamHI sites of pAV1427 to produce pEMBL-Int-6-5ЈFL and pEMBL-Int-6-2ϫFL, respectively. Similarly, the DNA segments encoding yeast Pci8p were synthesized from yeast chromosomal DNA by PCR using primers 4 and 5 and primers 4 and 6, digested with MluI and BamHI, and subcloned into pAV1427 to produce pEMBL-PCI8-5ЈFL and pEMBL-PCI8-2ϫFL, respectively. The N-terminal tag coding sequences in pEMBL-Int-6-2ϫFL and pEMBL-PCI8-2ϫFL were removed by SacI and MluI digestion, fill-in with Klenow enzyme, and self-ligation, generating pEMBL-Int-6-3ЈFL and pEMBL-PCI8-3ЈFL, respectively. Each of these six plasmids was introduced into yeast strain H2557 (see Table  III) and examined for expression of the encoded tagged Int-6 or Pci8p protein following induction on galactose-containing medium. As pEMBL-Int-6-5ЈFL and pEMBL-PCI8-3ЈFL produced the largest amounts of FLAG-tagged Int-6 and Pci8p, respectively (data not shown), they were selected for further analyses. The 2.3-kb DNA segment carrying the PCI8 ORF and its 5Ј-and 3Ј-UTRs (0.5 kb each) was amplified from the yeast genome using primers 7 and 9 (Table I), digested with BglII and BamHI, and subcloned into the BamHI site of YCplac111 (29) to generate YCpPCI8. YCpPCI8-FL, encoding FLAG-tagged Pci8p under the control of its own promoter, was constructed by replacing the 1.0-kb NsiI-BamHI segment of YCpPCI8 with a 0.5-kb NsiI-BamHI segment of pEMBL-PCI8-3ЈFL containing the 3Ј-half of the FLAG-tagged PCI8 ORF. The deletion plasmid pPCI8⌬-URA3 was constructed in two steps. First, the 3Ј-UTR segment of PCI8 was synthesized by PCR using primers 10 and 9 and digested with BclI and BamHI, which cleave at sites immediately following the PCI8 stop codon and at the end of primer 9, respectively. The resulting 0.5-kb fragment was subcloned into the BamHI site of pNKY51 (30) 3Ј of the hisG::URA3::hisG cassette. Second, the 5Ј-UTR of Pci8p and eIF3e/Int-6 Interact with eIF3 Core Complex PCI8 was amplified with primers 7 and 8, digested with BglII and BamHI, and subcloned into the BglII site of the pNKY51 derivative produced in the first step 5Ј of the hisG::URA3::hisG cassette.
Yeast Strains-The yeast strains used are listed in Table III. To generate the pci8⌬ deletion strain KAY76, the BglII-BamHI fragment from plasmid pPCI8⌬-URA3 containing the entire pci8⌬::hisG::URA3:: hisG disruption allele was introduced into a transformant of H2557 carrying YCpPCI8 (PCI8 LEU2). After excluding Leu ϩ Ura ϩ strains with the deletion construct integrated into YCpPCI8, deletion of PCI8 in the candidate strains was confirmed by amplifying the 5Ј and 3Ј ends of the deletion construct, which had been integrated into the chromosome, by PCR using primers 16 and 19, and 17 and 18 (Table I), respectively. Primers 18 and 19 correspond to yeast chromosomal sequences outside of the region covered by the deletion construct, whereas primers 16 and 17 correspond to sequences inside the deletion construct in the hisG repeats flanking the URA3 marker. We confirmed that the amplified DNA fragments were generated from the expected sites by restriction digest of the 5Ј-and 3Ј-fragments. YCpPCI8 was allowed to segregate from the pci8⌬ deletion strain by culturing in rich medium (yeast extract/peptone/dextrose (YPD)), and the progeny cells were streaked out for purification. One of the Leu Ϫ Ura ϩ segregants, designated KAY76 (pci8⌬), was selected for further analysis. The isogenic URA3 ϩ wild-type strain H2557U was constructed by transforming H2557 with YIplac211 (URA3 ϩ without yeast replicon) (29) that had been linearized with EcoRV to direct integration to the ura3-52 locus.
Microarray Analysis-Strain KAY76 (pci8⌬::URA3) and isogenic URA3 ϩ wild-type H2557U were grown in 250 ml of YPD medium at 30°C for 6 h, reaching an A 600 of 1.0 during exponential growth. Cells were harvested by centrifugation, washed, mixed with an equal volume of acid-washed and autoclaved glass beads (425-600 m; Sigma) and 7.5 ml of Trizol reagent (Life Technologies, Inc.), and vortexed eight times for 30 s each with 30-s intervals at room temperature. Total RNA was extracted according to the remaining instructions provided by the manufacturer. Poly(A) ϩ mRNA was prepared from total RNA using the RNeasy kit (QIAGEN Inc.) following the manufacturer's instructions. The buffer was removed, and the sample was concentrated by ethanol precipitation. Further sample preparation, cDNA synthesis, in vitro transcription, and labeling and fragmentation to produce the oligonucleotide probes were performed as instructed by the GeneChip manufacturer (Affymetrix). The probes were first hybridized to a test array (Affymetrix) and then to the GeneChip yeast genome S98 (Affymetrix), both performed at 45°C overnight in a rotisserie box. The chips were washed in a GeneChip Fluidics Station 400 (Affymetrix), and the results were visualized and analyzed with a Gene Array scanner (Hewlett-Packard Co.) using the Affymetrix software.

RESULTS
Human eIF3e/Int-6 and Yeast Pci8p Bind to Yeast eIF3 in Vivo-The eIF3e/Int-6 protein (445 amino acids, 52,187 Da) is a component of mammalian eIF3 (16) and contains a conserved domain termed PCI (14) or PINT (15). Budding yeast homologs have been identified for five of the mammalian eIF3 subunits, but not for Int-6. As the hypothetical S. cerevisiae Pci8p protein (444 amino acids, 51,254 Da; GenBank TM /EBI Data Bank accession number P40512) contains a PCI/PINT domain (14,15) and shows no similarity to any other proteins, 3 we considered the possibility that Pci8p might be a distantly related, functional ortholog of Int-6. To investigate whether Pci8p and its possible human counterpart (Int-6) could bind to yeast eIF3 in vivo, we constructed plasmids pEMBL-Int-6-5ЈFL and pEMBL-PCI8-3ЈFL for expressing FLAG epitope-tagged versions of these proteins under the control of a galactose-inducible (GAL) promoter. The plasmids were introduced into S. cerevisiae H2557, and the resulting transformants were grown on galactose-containing medium. Whole cell extracts (WCEs) were prepared from the transformants and subjected to Western analysis using antibodies against the FLAG epitope to characterize the expression levels of FLAG-Int-6 and FLAG-Pci8p. The WCE from strain KAY35 encoding FLAG-eIF5 as the sole source of this initiation factor (Table III) was analyzed in parallel for comparison. FLAG-Int-6 migrated at 52 kDa and was expressed at a level approximately one-fourth that of FLAG-eIF5. FLAG-Pci8p also migrated at 52 kDa and was expressed at a level ϳ10-fold higher than that of FLAG-eIF5 (data not shown).
To determine whether FLAG-Pci8p and FLAG-Int-6 can bind to yeast eIF3 in vivo, WCEs from the H2557 transformants just described or from a control transformant bearing an empty vector were immunoprecipitated with anti-FLAG affinity resin, and the immune complexes were subjected to Western analysis using antibodies against the FLAG epitope, yeast eIF3b subunit Prt1p, and yeast eIF2␣ subunit Sui2p. The results shown in Fig. 1A indicate that nearly all of FLAG-Int-6 and ϳ20% of FLAG-Pci8p were immunoprecipitated with anti-FLAG resin (indicated by arrows in the first row, lanes 5 and 8). Interestingly, ϳ20% of Prt1p (second row, lanes 5 and 8), but little or no eIF2␣ (third row, lanes 5 and 8), co-immunoprecipitated with both FLAG-Int-6 and FLAG-Pci8p. Under these conditions, nearly all of Prt1p in the WCE was incorporated into yeast eIF3, as indicated by co-immunoprecipitation experiments with antibodies against eIF3a subunit Tif32p (data not shown). 3 According to the YPD TM Database at www.proteome.com/. Single-copy LEU2 vector (29)  YCpPCI8 Single-copy LEU2 PCI8 plasmid YCpPCI8-FL Single-copy LEU2 PCI8-FLAG plasmid pPCI8⌬-URA3 PCI8 disruption plasmid with pci8⌬ϻhisGϻURA3ϻhisG pGEX-Int- 6 GST-Int-6 fusion plasmid pGEX-PCI8 GST-Pci8p fusion plasmid pGEX-p110 GST-human eIF3c fusion plasmid (4) pGEX-p116⌬ GST-human eIF3b⌬ fusion plasmid pT7-PRT1 PRT1 ORF cloned under T7 promoter (31) pT7-⌬-PRT1 series Different parts of PRT1 ORF cloned under T7 promoter (12) a References are given in parentheses for plasmids constructed previously.
Pci8p and eIF3e/Int-6 Interact with eIF3 Core Complex These results suggest that human Int-6 and Pci8p can interact specifically with yeast eIF3 in vivo. eIF3e/Int-6 and Pci8p Bind Specifically to Purified eIF3 in Vitro-In an effort to confirm these findings in vitro using purified proteins, we constructed GST fusions to Int-6 and Pci8p, purified them from bacteria (see Fig. 1B (upper panel) for Coomassie Blue staining of these proteins), and tested them for interaction with eIF3 purified from yeast. This eIF3 preparation contains polyhistidine-tagged eIF3b/Prt1p, FLAGtagged eIF3g/Tif35p, and hemagglutinin-tagged eIF3i/Tif34p and was purified by nickel chelation chromatography, followed by anti-FLAG affinity chromatography (33). The purified eIF3 was incubated with the GST fusion proteins or GST alone and immobilized on glutathione-Sepharose beads, and the bound proteins were visualized by Western analysis using antibodies against Prt1p or the hemagglutinin or FLAG epitope. As shown in Fig. 1B (lower panels), the His-Prt1p, FLAG-Tif35p, and hemagglutinin-Tif34p subunits of purified eIF3 bound to GST-Pci8p and GST-Int-6, but not to GST alone. These results suggest that both Pci8p and Int-6 can interact directly with yeast eIF3.
eIF3e/Int-6 and Pci8p Bind Specifically to the eIF3b/Prt1p Subunit in Vitro-To identify which yeast eIF3 subunit(s) bind to Int-6 and Pci8p, we performed additional binding experiments with GST fusions to each of the five core subunits of yeast eIF3 (31) and to the peripheral eIF3 subunit eIF3j/Hcr1p (34). These GST fusion proteins (Fig. 2, A and B, upper panels) and GST alone were incubated with 35 S-labeled Int-6 synthesized in rabbit reticulocyte lysates or with FLAG-Pci8p present in the yeast WCE used for co-immunoprecipitation analysis described above. The bound proteins were isolated on glutathione-Sepharose beads and visualized by autoradiography or by Western analysis with anti-FLAG antibodies, respectively. As shown in Fig. 2 (A and B, lower panels), both 35 S-labeled Int-6 and FLAG-Pci8p bound specifically to GST-eIF3b/Prt1p (lanes 4) and, to a lesser extent, to GST-eIF3a/Tif32p (lanes 3). Thus, a References are given in parentheses for strains constructed previously.

FIG. 1. Human eIF3e/Int-6 and S. cerevisiae Pci8p interact with yeast eIF3 in vivo and in vitro.
A, co-immunoprecipitation. The transformants of H2557 carrying pEMBL-Int-6-5ЈFL (FLAG-Int-6), pEMBL-PCI8-3ЈFL (FLAG-Pci8p), or pEMBLyex4 (vector) were inoculated into 50 ml of synthetic complete medium lacking uracil (containing 10% galactose and 2% raffinose) at A 600 ϭ 0.05 and grown overnight to A 600 ϭ 0.5-2. Cells were collected by centrifugation, suspended in buffer A (40), and broken with glass beads (425-600 m), by eight pulses of 30 s in a Braun homogenizer at 4°C with 30 s of cooling between pulses. Homogenized cell extracts were clarified by centrifugation, yielding the supernatants as WCEs, which were used for immunoprecipitation with anti-FLAG affinity resin as described previously (40). The immune complexes were analyzed by SDS-polyacrylamide gel electrophoresis and Western blotting using antibodies listed on the left. I, 20% of the input amount of WCEs used for immunoprecipitations; P, the entire precipitated fractions; S, 10% of the supernatant fractions. The asterisk indicates the position of anti-FLAG immunoglobulin. B, GST pull-down experiment. Extracts prepared from isopropyl-␤-D-thiogalactopyranoside-induced BL21(DE3) transformants bearing pGEX-4T-1, pGEX-Int-6, or pGEX-PCI8 were incubated with 5 l of glutathione-Sepharose beads (Amersham Pharmacia Biotech) according to the manufacturer's instructions. After extensive washing, beads attached to GST fusion proteins were incubated with 10 g of purified eIF3 (33)  it appears that Int-6 and Pci8p both can interact directly with the same two subunits of yeast eIF3. 4 The Minimal Binding Site in Yeast eIF3b/Prt1p for Int-6 and Pci8p Contains the RNA Recognition Motif (RRM)-We proceeded next to identify the minimal binding site in Prt1p required for its interaction with Pci8p or Int-6. Full-length and various truncated versions of 35 S-labeled Prt1p (depicted in Fig. 3A) were synthesized in rabbit reticulocyte lysates and incubated with GST-Pci8p, GST-Int-6, or GST alone. The bound proteins were isolated on glutathione-Sepharose beads and visualized by autoradiography. As reported previously (31), three polypeptides were produced from the construct (pT7-PRT1) that encodes full-length Prt1p (Fig. 3B, first panel, lane  5), but only the largest one corresponding to full-length Prt1p bound specifically to both GST-Pci8p and GST-Int-6 (first panel, lanes 2-4). These results confirm the conclusion reached from Fig. 2 that Pci8p and Int-6 both interact with Prt1p. The 35 S-labeled Prt1p polypeptide lacking only the N-terminal 27 residues (construct ⌬0) bound to GST-Pci8p and GST-Int-6 above the background level seen for GST alone (second panel); however, none of the polypeptides with more extensive Nterminal truncations bound specifically to either fusion protein (constructs ⌬1, ⌬2, ⌬3, and ⌬6 in the third through sixth panels). Thus, the N-terminal boundary of the binding domain lies between residues 28 and 111, within the predicted RRM (35) in Prt1p. Deletion from the C terminus up to residue 261 in Prt1p did not reduce its interaction with GST-Pci8p or GST-Int-6 (eighth panel), whereas deletion to residue 136 abolished the interaction with both fusion proteins (seventh panel). Thus, the C-terminal boundary of the binding domain for Pci8p and Int-6 is between residues 136 and 261 of Prt1p. The minimal binding domain defined by this deletion analysis corresponds to the RRM plus ϳ140 residues flanking it on the C-terminal side. The fact that Pci8p and Int-6 both interacted with the same N-terminal segment of eIF3b/Prt1p supports the idea that these proteins might be functional homologs.
Human eIF3e/Int-6 Binds a Segment of Human eIF3b (p116) Related to the Pci8p-binding Domain in Yeast eIF3b/Prt1p-To test whether human eIF3e/Int-6 also binds to the human hom- olog of eIF3b, originally called p116 (4, 32), we attempted to conduct binding studies with a GST fusion to human eIF3b. However, after finding that full-length GST-human eIF3b was not expressed well in bacteria (32), we constructed GST-human eIF3b⌬ containing the portion of human eIF3b corresponding to the minimal binding domain for Int-6 in yeast eIF3b/Prt1p (Fig. 4A). Because it was reported that human eIF3c bound to human eIF3e/Int-6 in yeast two-hybrid assays (22), we also examined a GST fusion to full-length human eIF3c for interaction with Int-6. The two GST fusion proteins or GST alone were purified from bacteria on glutathione-Sepharose beads (Fig.  4B, upper panel) and incubated with the yeast WCE containing FLAG-Int-6 described above for co-immunoprecipitation analysis (Fig. 1A). The bound proteins were visualized by Western analysis using anti-FLAG antibodies. As shown in Fig. 4B (middle panel), FLAG-Int-6 bound to GST-human eIF3b⌬ (lane 3), but not to GST alone (lane 2) or to GST-human eIF3c (lane 4). The unidentified 98-kDa protein present in the WCE that cross-reacts with anti-FLAG antibodies did not bind to any of the GST proteins (Fig. 4B, arrowheads), nor did the yeast eIF3b/Prt1p subunit present in the WCE. This last result suggests that the observed interaction between FLAG-Int-6 and GST-human eIF3b⌬ was not mediated by yeast eIF3. Thus, we conclude that human Int-6 can bind specifically to the RRMcontaining region of human eIF3b. The fact that S. cerevisiae Pci8p and human eIF3e/Int-6 interact with the corresponding RRM-containing regions of the cognate eIF3b subunits further supports the idea that Pci8p is the S. cerevisiae ortholog of eIF3e/Int-6.

PCI8 Is Not Essential, and Its Overexpression or Deletion Does Not Alter the Growth Rate of Wild-type and Mutant eIF3
Strains-To investigate the physiological function of Pci8p, we investigated whether PCI8 is essential for yeast cell growth. Toward this end, wild-type strain H2557 bearing YCpPCI8 (LEU2 PCI8) was transformed with a pci8⌬::URA3 deletion construct, and deletion of PCI8 in the resulting Leu ϩ Ura ϩ transformants was confirmed by PCR analysis (see "Materials and Methods"). By culturing without selection, we found that YCpPCI8 was lost from the pci8⌬ and PCI8 strains with similar frequencies, yielding Leu Ϫ Ura ϩ and Leu Ϫ Ura Ϫ clones, respectively (data not shown). These results indicate that PCI8 is not an essential gene.
The pci8⌬ strain lacking YCpPCI8 (designated KAY76) and its isogenic wild-type strain H2557 were found to have virtually identical doubling times when grown on rich and minimal media at 18, 25, 30, and 37°C (data not shown). Overexpression of Pci8p from pEMBL-PCI8-3ЈFL in a transformant of H2557 grown on galactose medium also did not produce a detectable change in cell doubling time (data not shown). Thus, elimination or overexpression of Pci8p did not affect the rate of yeast cell growth under standard culture conditions.
To investigate whether Pci8p is required for wild-type rates of translation initiation, we compared the polysome profiles of pci8⌬ strain KAY76 and its isogenic PCI8 parent strain H2557 grown to exponential phase on rich medium (YPD) at 30°C. As shown in Fig. 5, the two strains had very similar size distributions and abundance of polysomes, with the latter quantified by the ratio of polysomes to 80 S monosomes. These findings suggest that the rate of protein synthesis initiation is unaffected by the absence of Pci8p under the culture conditions employed in these experiments.
In an effort to uncover a nonessential stimulatory role for Pci8p in eIF3 function, we introduced the pci8⌬ mutation into strains harboring the temperature-sensitive lethal mutation prt1-1, rpg1-1, or nip1-1, altering the yeast eIF3b, eIF3a, and eIF3c subunits, respectively. The resulting double mutants showed growth rates identical to those of the isogenic prt1-1, rpg1-1, and nip1-1 single mutants at the semipermissive temperature for each eIF3 subunit mutation. Additionally, overexpression of wild-type Pci8p from the pEMBL-PCI8-3ЈFL construct during growth on galactose medium did not increase or decrease the growth rates of the eIF3 mutant strains (data not shown). We conclude that Pci8p does not contribute to the Finally, we investigated whether Pci8p might be involved in translational control of GCN4 mRNA, encoding a transcriptional activator of amino acid biosynthetic genes. High level translation of this mRNA is restricted to amino acid starvation conditions and is triggered by limiting the level of the eIF2 ternary complex via phosphorylation of eIF2 by protein kinase Gcn2p (36). Mutants lacking Gcn2p cannot induce GCN4 translation and thus fail to grow on medium containing an inhibitor of histidine biosynthesis, 3-aminotriazole. Mutations in eIF2 or eIF2B subunits that lower ternary complex levels can rescue growth of gcn2⌬ cells on 3-aminotriazole medium (2). However, we found that isogenic PCI8 gcn2⌬ and pci8⌬ gcn2⌬ strains (H2557 and KAY76, respectively) both failed to grow on medium containing 5 or 30 mM 3-aminotriazole (data not shown). We also explored the possibility that removal of Pci8p could impair the induction of GCN4 translation in GCN2 cells by transforming H2557 and KAY76 with a plasmid containing GCN2 and testing the resulting strains for growth on 3-aminotriazole medium. Again, we found that PCI8 GCN2 and pci8⌬ GCN2 strains grew indistinguishably on medium containing 30 mM 3-aminotriazole. Thus, removal of Pci8p seems to have no impact on GCN4 translational control, hence the cellular level or ribosome binding of the eIF2 ternary complex.
Gene Expression Profiling in Wild-type and pci8⌬ Strains- FIG. 4. Recombinant human eIF3b segment interacts with FLAG-tagged Int-6. A, shown are the primary structures of eIF3b homologs found in S. cerevisiae and human, drawn schematically with empty boxes. Shaded boxes indicate the minimal binding site for Pci8p and Int-6 found in the S. cerevisiae homolog (Fig. 3) or the homologous segment found in the human eIF3b polypeptide. Filled boxes depict the RRM. The horizontal bar below the human eIF3b schematic indicates its segment (human eIF3b⌬ (heIF3b⌬)) fused to GST for the binding studies described for B. Numbers at the ends of each box or bar indicate amino acid positions at the boundary. B, the GST fusion proteins listed at the top were incubated with WCE containing FLAG-Int-6, and the GST fusion protein complexes were analyzed as described in the legend to (PCI8) and KAY76 (pci8⌬) growing exponentially on YPD medium at 30°C were treated with cycloheximide for 5 min prior to harvesting the cells. WCEs were prepared and resolved by velocity sedimentation on 15-40% sucrose gradients as described previously (41). Fractions were collected while scanning continuously at A 254 . The x axis indicates A 254 ϭ 0. The positions of different ribosomal species are indicated. P/M, ratio of A 254 in the combined 2-4-mer fractions to that in the 80 S peak.
Mammalian Int-6 has been implicated in transcriptional regulation in addition to its association with eIF3. To investigate the possibility that Pci8p is involved in transcriptional control in yeast, we compared the total mRNA expression profiles of KAY76 (pci8⌬::URA3) and its isogenic wild-type strain H2557U. We constructed this strain by integrating wild-type URA3 into the ura3-52 locus of H2557 since KAY76 contains URA3 inserted at the pci8⌬ locus. We then prepared polyadenylated mRNA from these two strains grown on YPD medium at 30°C and processed the samples for two-chip oligonucleotide microarray analysis. Since this system allows estimates of the absolute levels of mRNA expression, we first compared the level of PCI8 mRNA with the levels of mRNAs encoding the five core subunits of eIF3 and other eIFs in the wild-type strain. The results showed that PCI8 mRNA was present at levels between 7 and 0.5% of the mRNA encoding eIF3 subunits and eIF5 (Table IV). In fact, it was expressed at very low levels compared with all other known eIFs, except for TIF4632.
To measure the level of Pci8p protein expression, we constructed YCpPCI8-FL, a derivative of YCpPCI8 encoding Pci8p FLAG-tagged at its C terminus. A transformant of pci8⌬ strain KAY76 carrying YCpPCI8-FL was grown to exponential phase on synthetic complete medium lacking leucine at 30°C, and a WCE was prepared and subjected to Western analysis using anti-FLAG antibodies. Strain KAY35 expressing FLAG-tagged eIF5 from a single-copy plasmid under the control of its own promoter was analyzed in parallel. The FLAG-Pci8p protein was undetectable (Ͻ5%) compared with the amount of FLAG-eIF5 observed in the WCE from KAY35 (data not shown). Thus, Pci8p is present at a level well below that of the canonical initiation factor eIF5 under our conditions.
To compare the mRNA expression profiles of the pci8⌬ and wild-type strains, two independent sets of microarray hybridization experiments were performed using duplicate preparations of poly(A) ϩ mRNA. Comparison of the mRNA levels measured in the two mRNA preparations from wild-type cells showed a high degree of correlation with a coefficient of 0.83. The corresponding correlation coefficient for the two mRNA preparations from the pci8⌬ strain was 0.94. Since the chips contain internal standards, the results from all four experiments could be compared directly with one another, yielding four different comparisons. Table V represents a summary of genes that showed an average difference in expression of Ͼ2-fold. 5 Only genes demonstrating at least a 1.5-fold change between the mutant and wild type in all four comparisons were included.
Interestingly, the levels of mRNAs encoding five heat shock proteins or their relatives (Hsp30p, Ddr2p, Yro2p, Sse2p, and Pir3p) and Ygr138cp, a putative transmembrane protein of the drug/H ϩ antiporter 12 family, were 2-fold or more lower in the deletion mutant versus the wild-type strain (Table V). Expression of HSP30, DDR2, YRO2, and SSE2 has been shown to be induced by heat (37,38). Pir3p is a cell wall membrane protein that is 80% identical to Hsp150p/Pir2p, but its expression is not induced by heat (39). Functional genomic analyses from several laboratories indicate that these genes belong to different groups of genes regulated by a variety of cellular stresses, including heat, osmolarity, and cell-damaging agents. 3 Thus, the genes listed in Table V might compose part of a novel stress-responsive regulon whose expression depends on Pci8p under our experimental conditions. This activity of Pci8p may be akin to the eIF3-independent role of fission yeast int-6 in transcription of multidrug resistance genes (8). DISCUSSION In this report, we showed that one of the eight predicted PCI domain proteins encoded in S. cerevisiae, Pci8p, can interact in vivo and in vitro with the eIF3 complex (Fig. 1). We identified the binding partners of Pci8p as the eIF3b/Prt1p and eIF3a/ Tif32p subunits (Fig. 2B) and localized a Pci8p-binding site in the N-terminal part of eIF3b/Prt1p that contains an RRM (Fig.  3). Interestingly, the human eIF3e/Int-6 protein also bound to the yeast eIF3 complex (Fig. 1) and the isolated eIF3b/Prt1p subunit ( Fig. 2A) via the N-terminal RRM domain (Fig. 3). Furthermore, human Int-6 interacted with the corresponding RRM-containing segment of the human eIF3b protein (p116) (Fig. 4). These results support the notion that yeast Pci8p is a divergent ortholog of human Int-6. They further suggest that the binding site for human Int-6 in the cognate eIF3b subunit is conserved in yeast eIF3b/Prt1p and that Pci8p can interact with this domain in a manner similar to the interaction of Int-6 with eIF3b/p116, as shown schematically in Fig. 6.
Interactions between eIF3e/Int-6 and Other eIF3 Subunits-Results from our binding assays provide new insights into the subunit interactions in mammalian eIF3. The interaction we detected between human eIF3e/Int-6 and eIF3b/p116 (Fig. 4) may be instrumental in tethering Int-6 to the human eIF3 core complex (Fig. 6B). However, there are likely other contributing interactions, as human Int-6 interacts with eIF3c/p110 in the yeast two-hybrid assay (22). Although FLAG-Int-6 did not bind to GST-human eIF3c in our binding assays (Fig. 4), there is evidence that plant eIF3e directly interacts with the cognate plant eIF3c subunit (27). On the other hand, Yen and Chang (24) reported that fission yeast eIF3e/Int6 interacts with eIF3d/ Moe1 in yeast two-hybrid and far-Western assays, that this interaction is conserved with the cognate human homologs, and that it may depend on the PCI domain in human Int-6. Perhaps the PCI domain of eIF3e/Int-6 can interact simultaneously with eIF3d and eIF3c to bridge the interaction between eIF3d and the eIF3 core complex (Fig. 6B). Consistent with this model, disruption of moe1 ϩ reduces the level of Int6 and vice versa (24). The absence of one of the eIF3 non-core subunits e and d may lead to loss of the other from the eIF3 complex, 5 We also conducted the same experiments with KAY76 (pci8⌬::URA3) and isogenic wild-type strain H2557 (PCI8 ϩ ura3-52) and confirmed that expression of the genes listed in Table V is reduced in a pci8⌬::URA3 strain. a Measured as average difference in intensity between perfect match and mismatch over an entire gene probe set. Results represent mean of two independent experiments. TIF1 and TIF2 used the same probe set. Therefore, the sum of expression levels of these genes is given as eIF4A expression level.
Pci8p and eIF3e/Int-6 Interact with eIF3 Core Complex resulting in increased degradation of both proteins. Because mouse mammary tumor virus insertion at int-6 results in truncation of the PCI domain (19), at least a subpopulation of eIF3 complexes found in int-6-disrupted mammary tumors may lack both eIF3e/Int-6 and eIF3d. Finally, we note that human Int-6 and Pci8p interacted with eIF3a/Tif32p (the third PCI protein in eIF3) in our binding experiments (Fig. 2), that eIF3a/Tif32p and eIF3c/Nip1p interact directly (31), and that eIF3a/Tif32p binds to the RRM domain in eIF3b/Prt1p (12). Thus, a network of interactions seems to link the three PCI proteins in eIF3 (subunits a, c, and e) to one another, to eIF3b, and to eIF3d (Fig. 6).
The S. cerevisiae eIF3 core complex can be divided into two subcomplexes assembled at opposite ends of the eIF3b/Prt1p subunit: eIF3a/Tif32p bridges interaction between eIF3c/Nip1p and the N-terminal domain of eIF3b/Prt1p containing the RRM (N-terminal subcomplex), whereas eIF3g/Tif35p and eIF3i/ Tif34p bind to the extreme C terminus of eIF3b/Prt1p (Fig. 6A)  (12, 31, 33). Interestingly, the interactions described above involving eIF3e/Int-6 would involve specifically the N-terminal eIF3 subcomplex (Fig. 6B). This subcomplex mediates physical interaction of eIF1 and eIF5 with eIF3 via the N terminus of eIF3c, and eIF5 additionally interacts with the Met-tRNA i Metbinding factor eIF2 ( Fig. 6) (11,40,41). eIF1, eIF2, and eIF5 were all implicated in stringent AUG recognition in S. cerevisiae (42). The RRM domain of yeast eIF3b/Prt1p is also required for eIF3 integrity and ribosome binding (12). Assuming that mammalian eIF3 and yeast eIF3 have similar substructures and functions, then association of eIF3e/Int-6 (and possibly eIF3d) with the N-terminal subcomplex might influence ribosome binding or the functions of eIF1, eIF2, and eIF5 in Met-tRNA i Met binding or AUG recognition. There is precedence for an eIF3 non-core subunit interacting with the N-terminal subcomplex of eIF3. Budding yeast eIF3j/Hcr1p is a nonessential protein that can bind to eIF3a and eIF3b simultaneously in vitro (Fig. 6A) and promote formation of the multifactor complex containing eIF3, eIF1, eIF5, and eIF2 in vivo (12).
As yet, there is no evidence that human eIF3e/Int-6 influences the biochemical activity of human eIF3, although the interferon-inducible protein p56 may down-regulate eIF3 ac-  Yeast eIFs implicated in correct AUG selection by genetic approaches (42) and their human homologs are pale blue. The rounded rectangle indicates the 40 S subunit. Double-headed arrows indicate direct interactions between different eIFs and 40 S ribosome. The eIF3 subunits were named after genes encoding them (A) or their sizes in kDa (B), and letters in parentheses indicate the unified eIF3 subunit nomenclature proposed by Burks et al. (6). Primary structures of some eIF3 subunits are drawn with N-and C-terminal ends, evolutionarily conserved domains (ovals), and less conserved charged domains (thick lines). The five eIF3 core subunits found in S. cerevisiae are pale orange, whereas peripheral, non-essential subunits are white. Proteins containing the PCI/PINT domains are shown in red. Brackets refer to N (N-term)-and C (Cterm)-terminal subcomplexes of eIF3. For A, direct contacts indicate direct interactions (12,31). Thick arrows indicate interaction of Pci8p, the possible eIF3e ortholog, found in this study. For B, direct contacts between the five core subunits are deduced from the model in A, except for the eIF3a-eIF3b (RRM) interaction in human (32). Human eIF3 subunits f, h, k, and l are not shown, since no information is available for their interactions. tivity by binding to eIF3e/Int-6 (43). In contrast, the eIF3e/Int6 protein of fission yeast has been implicated more directly in eIF3 function (see Introduction). Thus, it was conceivable that Pci8p also plays a role in translation by binding to the same site in the eIF3 core as human Int-6 does. At odds with this prediction, binding of Pci8p to yeast eIF3 has little or no effect on the general function of this complex in translation initiation under the growth conditions of our experiments (Fig. 5 and other data not shown). It is still possible that binding of Pci8p to eIF3 could alter the translation of specific mRNAs (as suspected for S. pombe Int6), or the protein may have to be induced or undergo a post-translational modification that is restricted to a specific stress condition or developmental state in order to influence eIF3 function. Alternatively, Pci8p may have lost a role in translation and become specialized as a transcriptional regulator, as described below.
Is Pci8p Involved in Transcriptional Control?-Growing evidence has suggested that eIF3e/Int-6 has dual regulatory functions in the nucleus and cytoplasm (see Introduction). Accordingly, we investigated the effect of deleting PCI8 on genome-wide expression of mRNAs using microarray technology. The results showed that pci8⌬ mutants exhibit reduced expression of mRNAs for several heat shock-or stress-inducible proteins (Table V). It remains to be determined whether Pci8p functions directly or indirectly to enhance the expression of these mRNAs, although a recent report of two-hybrid interaction between Pci8p and Tfb1p (the transcription factor IIH subunit of RNA polymerase II holoenzyme) may favor a direct role in transcription (44). 3 The possible role of Pci8p in stressinducible transcription is reminiscent of a recent finding that, in plants, the occurrence of PR500, a free form of the proteasome lid subcomplex, and its localization to nuclear speckles are regulated by heat and canavanine (45). The fact that budding yeast Pci8p can bind to eIF3 and also influence the expression of heat shock mRNAs may indicate that it too performs dual functions in the nucleus and cytoplasm.
It was recently reported that the COP9 signalosome promotes cleavage of a ubiquitin-like protein, NEDD8, from proteins conjugated with it (46). This activity appears to be conserved from budding yeast to humans, even though S. cerevisiae encodes only one COP9 signalosome subunit ortholog termed Rri1p, a Mov34 family protein previously called D0888 (10) or YDL216c (14,15). In S. cerevisiae, deletion of RRI1 is not lethal, 3 but allows accumulation of a NEDD8conjugated protein (46). Since the COP9 signalosome is a hetero-octamer containing six PCI proteins and two Mov34 family proteins (26), Pci8p may be directly involved in this COP9 signalosome activity and activate transcription by regulating NEDD8 conjugation of one or more transcription factors. According to our microarray data, RRI1 expression is very low and comparable to that of PCI8 (Table IV). The similar expression levels and dispensability of PCI8 and RRI1 are at least consistent with this interesting model.