The essential ATP-binding cassette protein RLI1 functions in translation by promoting preinitiation complex assembly.

RLI1 is an essential yeast protein closely related in sequence to two soluble members of the ATP-binding cassette family of proteins that interact with ribosomes and function in translation elongation (YEF3) or translational control (GCN20). We show that affinity-tagged RLI1 co-purifies with eukaryotic translation initiation factor 3 (eIF3), eIF5, and eIF2, but not with other translation initiation factors or with translation elongation or termination factors. RLI1 is associated with 40 S ribosomal subunits in vivo, but it can interact with eIF3 and -5 independently of ribosomes. Depletion of RLI1 in vivo leads to cessation of growth, a lower polysome content, and decreased average polysome size. There was also a marked reduction in 40 S-bound eIF2 and eIF1, consistent with an important role for RLI1 in assembly of 43 S preinitiation complexes in vivo. Mutations of conserved residues in RLI1 expected to function in ATP hydrolysis were lethal. A mutation in the second ATP-binding cassette domain of RLI1 had a dominant negative phenotype, decreasing the rate of translation initiation in vivo, and the mutant protein inhibited translation of a luciferase mRNA reporter in wild-type cell extracts. These findings are consistent with a direct role for the ATP-binding cassettes of RLI1 in translation initiation. RLI1-depleted cells exhibit a deficit in free 60 S ribosomal subunits, and RLI1-green fluorescent protein was found in both the nucleus and cytoplasm of living cells. Thus, RLI1 may have dual functions in translation initiation and ribosome biogenesis.

RLI1 is an essential yeast protein closely related in sequence to two soluble members of the ATP-binding cassette family of proteins that interact with ribosomes and function in translation elongation (YEF3) or translational control (GCN20). We show that affinity-tagged RLI1 co-purifies with eukaryotic translation initiation factor 3 (eIF3), eIF5, and eIF2, but not with other translation initiation factors or with translation elongation or termination factors. RLI1 is associated with 40 S ribosomal subunits in vivo, but it can interact with eIF3 and -5 independently of ribosomes. Depletion of RLI1 in vivo leads to cessation of growth, a lower polysome content, and decreased average polysome size. There was also a marked reduction in 40 S-bound eIF2 and eIF1, consistent with an important role for RLI1 in assembly of 43 S preinitiation complexes in vivo. Mutations of conserved residues in RLI1 expected to function in ATP hydrolysis were lethal. A mutation in the second ATPbinding cassette domain of RLI1 had a dominant negative phenotype, decreasing the rate of translation initiation in vivo, and the mutant protein inhibited translation of a luciferase mRNA reporter in wild-type cell extracts. These findings are consistent with a direct role for the ATP-binding cassettes of RLI1 in translation initiation. RLI1-depleted cells exhibit a deficit in free 60 S ribosomal subunits, and RLI1-green fluorescent protein was found in both the nucleus and cytoplasm of living cells. Thus, RLI1 may have dual functions in translation initiation and ribosome biogenesis.
Most of the proteins belonging to the ATP-binding cassette (ABC) 1 superfamily of proteins are membrane transporters that use ATP hydrolysis to transport solute molecules against a concentration gradient (1). Typically, they contain two ABCs and a transmembrane domain. The ABC contains a nucleotide binding domain with Walker A and B motifs and an ␣-helical domain bearing the "LSGGQ" signature motif that, along with several other conserved features, distinguishes ABC proteins from other ATPases. ABCs bind ATP as a dimer, with the LSGGQ motif of one cassette capping the ATP molecule bound to the Walker motifs of the second cassette in the dimer. This produces an "ATP sandwich" with two ATP molecules bound to two hybrid active sites formed by the dimerized cassettes. ATP hydrolysis destabilizes the dimer, and it was proposed that a cycle of dimerization driven by ATP binding and hydrolysis can perform mechanical work. For ABC transporters, this would entail opening and closing a solute channel in the membrane through conformational changes in the transmembrane domains (2)(3)(4)(5)(6).
The yeast proteins GCN20 and YEF3 are soluble ABC proteins whose functions are connected with the binding of tRNAs to ribosomes. YEF3 (eEF3) is an essential translation elongation factor that stimulates release of deacylated tRNA from the ribosomal E-site concurrent with the recruitment of aminoacylated tRNA to the ribosomal A-site (7). GCN20 is a positive regulator of GCN2, a protein kinase that regulates protein synthesis initiation in amino acid starved cells by phosphorylation of eukaryotic translation initiation factor 2 (eIF2). GCN20 resides in a complex with GCN1, a protein that interacts with both GCN2 and ribosomes, and this complex is believed to mediate the activation of GCN2 kinase function by uncharged tRNAs bound to the ribosomal A-site. The N-terminal domain of GCN20 promotes the regulatory function of GCN1 and the ABC domains in GCN20 seem to modulate ribosome association of the GCN1-GCN20 complex (8 -10). Yeast encodes three additional ABC proteins closely related to GCN20 and YEF3, known as NEW1, YER036C, and YNL014W, and a more distantly related nontransporter ABC protein called RLI1 (11). RLI1 is an essential protein (12), and in a systematic analysis of protein complexes in yeast (13) it copurified with TAP affinity-tagged forms of the largest subunit of translation initiation factor 3 (eIF3a/TIF32) and initiation factor 5 (eIF5). These findings raised the possibility that RLI1 functions in the initiation of protein synthesis.
The process of translation initiation in eukaryotes is stimulated by an array of soluble initiation factors (eIFs), beginning with the recruitment of initiator methionyl tRNA (Met-tRNA i Met ) in a ternary complex (TC) with eIF2 and GTP to the 40 S subunit to form the 43 S preinitiation complex. This reaction is stimulated in vitro by eIF1, eIF1A, and eIF3 (14 -16). The eIF3 complex in yeast consists of five essential subunits (eIF3a/TIF32, eIF3b/PRT1, eIF3c/NIP1, eIF3i/TIF34, and eIF3g/TIF35) and one nonstoichiometric component, eIF3j/ HCR1. The recruitment of mRNA to the 43 S complex is stimulated by the m 7 G cap-binding complex eIF4F and poly(A)binding protein, to produce the 48 S complex. This assembly scans the mRNA, and the recognition of an AUG start codon triggers GTP hydrolysis by the TC, dependent on eIF5. In the last step of the pathway, the 60 S subunit joins with the 40 S complex in a reaction stimulated by eIF5B, and translation elongation then begins (14,17).
In yeast, the TC is associated with eIF3, eIF5, and eIF1 in a multifactor complex (MFC) that can exist free of 40 S ribosomes (18). In the MFC, the ␤-subunit of eIF2 associates indirectly with eIF3c/NIP1 in a manner bridged by eIF5, and eIF2␤ also binds directly to the C-terminal domain of eIF3a/TIF32 (18 -20). A mutation in eIF5 (tif-7A) that disrupts the indirect contact between eIF2 and eIF3, and the overexpression of a dominant negative mutant of eIF3a/TIF32 (TIF32-⌬6), which lacks the binding domain for eIF2␤, confer growth defects that are partially suppressed by overexpressing the TC. When combined in the same strain, these mutations produce an additive defect in translation initiation (20,21) and reduce the binding of TC and eIF3 subunits to 40 S ribosomes (22). These findings plus the fact that tif5-7A impairs Met-tRNA i Met binding to 40 S subunits in vitro (18) led us to propose that the two independent contacts between eIF2 and eIF3 in the MFC have additive stimulatory effects on recruitment of these factors to the 40 S ribosome to form the 43 S complex.
The co-purification of TAP-tagged RLI1 with eIF3a/TIF32 and eIF5 raised the possibility that RLI1 is an additional component of the MFC with a role in translation initiation. We show here that RLI1 is specifically associated with MFC components and 40 S ribosomes and that it promotes the assembly or stability of 43 S preinitiation complexes in vivo.

MATERIALS AND METHODS
Plasmid Constructions-The plasmids employed in this study are listed in Table I. Plasmid pDH177 was constructed as follows. The RLI1 gene, including 662 bp upstream of the ATG and 400 bp downstream of the stop codon, was amplified by PCR using the two primers 5Ј-CCA GAT AAA TGC TTC CAA GAA CCT TTA and 5Ј-AAC TGC AGT GCA GCA AAT GCT TAT CAT GAC GA and total genomic DNA from yeast strain BY4741 as template. The PCR product was cut with SpeI and PstI, and the resulting fragment was inserted between the SpeI and PstI sites of plasmid YCplac33. The DNA sequence of the entire ORF was verified, and it was shown that this plasmid complements an rli1⌬ mutant.
Plasmid pDH178 was constructed in several steps. First, a SalI site was introduced immediately upstream of the RLI1 stop codon by sitedirected mutagenesis using the QuikChange XL kit from Stratagene (catalog no. 200516) to create a transition plasmid, pDH177-SalI. Second, we generated a 0.43-kb SalI-PstI fragment, comprising the SalI site, FLAG coding sequence, stop codon, and sequences downstream of RLI1 extending to the naturally occurring PstI site, by PCR using the two primers 5Ј-GCG TCG ACT ACA AGG ACG ACG ATG ACA AAA TTT AAG CAT CTT GGG ATT CGG AGA AC and 5Ј-AAC TGC AGT GCA GCA AAT GCT TAT CAT GAC GA and plasmid pDH177 as template. This 0.43-kb fragment was cloned into pBLUESCRIPT between the SalI and PstI sites to generate another transition plasmid, pBS0.43kb. The DNA sequence of the entire 0.43-kb fragment was verified. Third, the 0.43-kb SalI-PstI fragment in pDH177-SalI was replaced by the 0.43-kb fragment from pBS0.43kb. It was shown that the final plasmid, pDH178, can complement an rli1⌬ mutant and direct the production of FLAG-tagged RLI1 of the predicted molecular weight in yeast as judged by Western analysis of whole cell extracts (WCEs) using antibodies against the FLAG epitope.
The plasmids pDH181 and pDH183 were constructed in two steps. First, the DNA fragment of the entire RLI1 ORF, except for the start codon, plus 0.4 kb downstream of the stop codon, was PCR-amplified using the primers 5Ј-GCG TCG ACG CGT AGT GAT AAA AAC AGT CGT ATC GCT A and 5Ј-AAC TGC AGC AAA TGC TTA TCA TGA CGA, introducing SalI and MluI sites at the 5Ј-end and a PstI site at the 3Ј-end of the fragment. Total genomic DNA from yeast strain BY4741 was used as the template. The PCR fragment was cut with SalI and PstI and inserted between the SalI and PstI sites of pBLUESCRIPT, resulting in the transition plasmid pBS-RLI1. The DNA sequence of the entire fragment in pBS-RLI1 was verified. Second, the MluI-PstI fragment was removed from pBS-RLI1 to replace the MluI-PstI fragment in the plasmids pYER-M and pYER-R, respectively, resulting in pDH181 and pDH183. (Plasmids pYER-M and pYER-R will be described separately.) Plasmid pDH184 was derived from pDH183 by site-directed mutagenesis via PCR using the two mutagenic primers 5Ј-GAA AAG TTA TCT GAT GAT GAA CTG CAA AGA TTT GCC and 5Ј-GGC AAA TCT TTG CAG TTC ATC ATC AGA TAA CTT TTC and plasmid pDH183 as the template. The plasmid pDH185 was constructed in the same way but using the two primers 5Ј-GTC CAA CAT TTG TCT GAT GAT GAA TTA CAA AGA GTC GCC and 5Ј-GGC GAC TCT TTG TAA TTC ATC ATC AGA CAA ATG TTG GAC. The DNA sequence of the entire ORF was determined to verify the presence of the intended point mutations.
Plasmid pDH98 was constructed using the primers 5Ј-CCA GAT AAA TGC TTC CAA GAA and 5ЈGAA GTG GCG CGA ATT CAC TAG TG to PCR-amplify a fragment from total genomic DNA of yeast strain YDH353. The resulting fragment was cut with SpeI and inserted into the SpeI site of plasmid pRS315, resulting in pDH98. The DNA sequence of the entire RLI1 ORF and the presence of the coding sequences for the myc 5 tag at the 3Ј-end of the ORF were both verified. It was shown that this plasmid can complement an rli1⌬ mutant. To introduce the E493Q mutation into pDH98, plasmid pDH202 was constructed by site-directed mutagenesis as described above, using plasmid pDH98 as the template and the mutagenic primers 5Ј-GAC ATA TAC TTG ATT GAT CAG CCA TCT GCC TAC TTA G and 5Ј-CTA AGT AGG CAG ATG GCT GAT CAA TCA AGT ATA TGT C. The DNA sequence of the entire RLI1 ORF and the presence of the mutation were both verified.
Plasmid pDH201 was constructed using the primers 5Ј-C GAG CTC ATG AGT GAT AAA AAC AGT CGT ATC and 5Ј-AAC TGC AGC AAA TGC TTA TCA TGA CGA and plasmid pDH178 as the template. The  Table II. To construct YDH312, the coding sequences for GFP were appended to the 3Ј-end of the chromosomal NEW1 allele immediately upstream of the stop codon by one-step homologous recombination (23). We used two primers, 5Ј-CGT TAC ATT GAG TGG TTG TCA TCG CCA AAA GGT ACA CCA AAA CCA GTT GAT ACT GAC GAT GAA GAA GAT CGG ATC CCC GGG TTA ATT AA and 5Ј-TGA ATT AAA GAA AAA AGA TGG GAT GAA ACA AAC GAA GTT AGC GAA GAT AAA ACA CTA GCC AGT AGG CTT GAA TTC GAG CTC GTT TAA AC and plasmid pFA6a-GFP(S65T)-HIS3MX as the template to PCR-amplify the appropriate DNA fragment and used it to transform yeast strain BY4741 to His ϩ , producing YDH312. The presence of NEW1-GFP was verified by Western analysis of WCEs using monoclonal antibodies against GFP. Growth of this strain on YPD medium was indistinguishable from that of BY4741.
To construct YDH353, the coding sequences for myc 13 were appended to the 3Ј-end of the chromosomal RLI1 allele immediately upstream of the stop codon by one-step homologous recombination, using the two primers 5Ј-AAG CTA GAT TCC CAA ATG GAT AAA GAA CAA AAA TCA TCA GGA AAC TAC TTT TTC TTG GAT AAC ACC GGT ATT CGG ATC CCC GGG TTA ATT AA and 5Ј-TAA TTA TTT CTC CAC ATT TCT AAA CCC AAA AAT AAA ACA ATC GTC CTC TTG GTT CTT CGA ATC CCA AGA TGC GAA TTC GCG CTC GTT TAA AC and plasmid pFA6a-13myc-HIS3MX as template, to PCR-amplify the appropriate DNA fragment and transform yeast strain BY4741 to His ϩ . The presence of RLI11-myc 13 in YDH353 was verified by Western analysis using monoclonal antibodies against c-myc epitope. Growth of this strain on YPD medium was indistinguishable from that of BY4741.
To produce YDH363, pDH177 was introduced into the RLI1/rli1⌬ diploid strain 24026, purchased from Research Genetics. The Ura ϩ transformants were sporulated and subjected to tetrad analysis. An ascospore clone deleted for chromosomal RLI1 but carrying plasmid pDH177 was identified. To generate YDH364, plasmid pDH178 was introduced into YDH369, and a Leu ϩ Ura ϩ transformant was grown in SC-Ura to eliminate plasmid pDH181. Strains YDH367 and YDH368 were produced by transforming YDH363 with pDH183 or YCplac111, respectively.
To produce YDH369, pDH181 was introduced into YDH363, and the transformants were selected on SC Gal -Leu-Ura. A Ura ϩ Leu ϩ transformant was replica-printed on SC Gal medium containing 5Ј-fluoro-orotic acid (5-FOA) to evict pDH177. A similar procedure was used to generate YDH371 by introducing pDH183 into YDH363 and then evicting pDH177. Strains YDH373 and YDH374 were produced by transforming YDH363 with pDH184 or pDH185, respectively.
To construct YDH391, the coding sequences for GFP were appended to the 3Ј-end of the chromosomal RLI11 allele immediately upstream of the stop codon by one-step homologous recombination, as described above for YDH312. We used primers 5Ј-AAG CTA GAT TCC CAA ATG GAT AAA GAA CAA AAA TCA TCA GGA AAC TAC TTT TTC TTG GAT AAC ACC GGT ATT CGG ATC CCC GGG TTA ATT AA and 5Ј-TAA TTA TTT CTC CAC ATT TCT AAA CCC AAA AAT AAA ACA ATC GTC CTC TTG GTT CTT CGA ATC CCA AGA TGC GAA TTC GCG CTC GTT TAA AC and plasmid pFA6a-GFP(S65T)-HIS3MX as the template to produce the appropriate PCR-amplified fragment for transformation of BY4741. The presence of RLI1-GFP was verified by Western analysis using antibodies against GFP. Growth of this strain on YPD medium was indistinguishable from that of BY4741.
Biochemical and Imaging Techniques-Co-immunoprecipitation analysis was performed as described previously (13,24), and Ni 2ϩ chelation chromatography of eIF3 complexes containing His-tagged NIP1 (20) were conducted as described previously. HCHO cross-linking of yeast cells and separation of extracts by velocity sedimentation through sucrose gradients was conducted according to Ref. 22. Polysome analysis using extracts from cells treated cycloheximide was conducted as described previously (25). Purification of RLI1 was carried out essentially as described previously (26) except that 2% instead of 10% galactose was the carbon source and 3-AT was omitted from the galactose induction medium. In vitro translation of LUC mRNA in whole cell yeast extracts was conducted as previously described (27). Imaging of GFP-tagged fusion proteins in living yeast cells was conducted as described previously (28).

RLI1 Is Associated with Specific Translation Initiation Factors in
Vivo -To confirm the reported association of RLI1 with eIF5 and a subunit of eIF3, we introduced the coding sequences for 13 tandem myc epitopes at the C terminus of chromosomal RLI1 and found that the resulting RLI1-myc strain grew indistinguishably from the parental RLI1 strain (data not shown). WCEs from isogenic RLI1-myc and RLI1 strains were immunoprecipitated with anti-myc antibodies, and the immune complexes were probed with antibodies against different translation factors. The results in Fig. 1A showed that RLI1-myc This study This study This study specifically coimmunoprecipitated with eIF5 and subunits of eIF2 or eIF3 but not with eIF5B or with a subunit of the 40 S ribosome (RPS22), translation elongation factor eEF1␤, or translation termination factor eRF3. These findings suggest that RLI1 is specifically associated with the MFC in vivo. In an effort to confirm that RLI1 can associate with these factors free of the 40 S ribosome, we repeated the coimmunoprecipitation analysis using a postribosomal supernatant prepared from a WCE. We found that RLI1-myc coimmunoprecipitated with eIF3 and eIF5, but not with eIF2, from the postribosomal supernatant (Fig. 1B). From these last results, it seems clear that RLI1 associates with eIF3 and eIF5 free of the 40 S ribosome. It is possible that the association of eIF2 with nonribosomal complexes containing RLI1, eIF3, and eIF5 is labile and does not persist in the postribosomal supernatant, which contains a lower concentration of most initiation factors compared with the WCE. Indeed, eIFs are generally purified from the ribosomal pellet obtained in preparing the postribosomal supernatant (27). We asked next whether RLI1 is a component of 43 or 48 S preinitiation complexes. To address this question, we used a recently devised technique in which living yeast cells are treated with formaldehyde to cross-link native preinitiation complexes in vivo. The composition of these complexes is then examined after resolving WCEs prepared from the cross-linked cells by sedimentation through sucrose density gradients (22). For this experiment, we employed a strain lacking chromosomal RLI1 and harboring the RLI1-F allele on a single copy plasmid. The latter encodes a fully functional form of RLI1 tagged at its C terminus with the FLAG epitope under the control of its own promoter. We observed the expected cosedimentation of a proportion of eIF2, -3, -5, -1, and -1A with the 40 S subunit, indicative of 43 or 48 S preinitiation complexes ( Fig.  2A, fractions 9 -11). Importantly, RLI1-F showed a clear peak in the same fractions, suggesting that RLI1 is associated with preinitiation complexes in vivo. When the same extract was resolved on a sucrose gradient of higher density to visualize 60 S subunits, 80 S ribosomes, and polysomes in addition to 40 S subunits, we observed that RLI1-F co-sedimented at much higher levels with 40 S subunits than with 60 S subunits (Fig.  2C). It also appeared that RLI1-F is associated at significant levels with 80 S ribosomes and polysomes. A qualitatively similar sedimentation profile was observed for GCD11, the ␥-subunit of eIF2 (Fig. 2C). We presume that the association of RLI1 and GCD11 with 80 S ribosomes and polysomes reflects the occurrence of a small proportion of "halfmer" polysomes in which a 43 S preinitiation complex is bound to the leaders of mRNAs containing one or more translating 80 S ribosomes.
RLI1 Is Required for Efficient Translation Initiation in Vivo -We investigated next the effect of eliminating RLI1 from yeast cells on the rate of translation initiation in vivo. Because it is essential for cell viability, we constructed a strain in which RLI1 is rendered unstable by attaching the coding sequences for ubiquitin, followed by an arginine codon and tandem FLAG and His 6 affinity tags (FH), at the N terminus of the RLI1 ORF. The ubiquitin is co-translationally cleaved, leading to rapid proteosomal degradation of the remaining R-FH-RLI1 moiety through the N-end rule pathway (34). Transcription of this engineered allele, designated P GAL -UBI-R-FH-RLI1, is regulated by the GAL promoter and thus can be repressed by growing cells with glucose as carbon source. As a control, we constructed a similar P GAL -UBI-M-FH-RLI1 allele in which the destabilizing Arg codon was replaced by a stabilizing Met codon. Western analysis of extracts prepared from the resulting strains using anti-FLAG antibodies showed that the stable M-FH-RLI1 protein was expressed at relatively high levels in cells grown with galactose as carbon source and at much lower levels by 6 h after shifting from galactose to glucose. By comparison, R-FH-RLI1 was expressed at much lower levels in galactose and was undetectable after culturing for only 2 h in glucose (Fig.  3A). (The upper bands in the Western blot are unknown yeast proteins that cross-react with FLAG antibodies.) As expected from the Western analysis, cells expressing R-FH-RLI1 could not grow on medium with glucose as the carbon source (Fig. 3B). These cells double only once following the shift to glucose medium before they cease growth about 8 -10 h later. They also grow more slowly than wild type even with galactose as the carbon source, displaying a doubling time of ϳ3 h compared with the ϳ2-h doubling time for wild type on galactose medium (Fig. 3C and data not shown). Western analysis  2. Depletion of RLI1 reduces 40 S-bound eIF1, eIF2, and eIF5 in vivo. A, RLI1-F strain YDH364 was grown in SC medium at 30°C to saturation and then diluted to an A 600 of ϳ0.02 in SC medium and grown to an A 600 of ϳ1.2. B, P GAL -UBI-R-FH-RLI1 strain YDH369 was grown in SC Gal medium to saturation, diluted to an A 600 of 0.4 -0.5 in SC medium, and grown for 12 h to an A 600 of ϳ1.2. Cells in A and B were cross-linked with HCHO, and WCEs were prepared and resolved by velocity sedimentation through 7.5-30% sucrose gradients. Fractions were collected while scanning continuously for A 254 and subjected to Western analysis using antibodies against the proteins indicated between A and B. The A 254 measurements are depicted at the top, with the locations of the 40 S subunits indicated. C, the same RLI1-F extract described in A was separated on a 7-47% sucrose gradient and analyzed exactly as described above except that a different set of antibodies was used to probe the gradient fractions for the proteins indicated to the right of the Western blot panels.
showed that the level of the unstable R-FH-RLI1 in galactose medium was substantially lower than that of RLI-F expressed from the native RLI1 promoter (data not shown). Thus, the slow growth phenotype of the P GAL -UBI-R-FH-RLI1 strain on galactose medium probably reflects an insufficient amount of protein produced from this construct even when its transcription is fully induced by galactose. By comparison, the P GAL -UBI-M-FH-RLI1 strain doubled at nearly the wild-type rate on galactose medium but had a somewhat longer doubling time (3 h) on glucose medium (Fig. 3C) (data not shown). The latter can be explained by the fact that the repressed level of M-FH-RLI1 in glucose-grown cells is somewhat lower than that of RLI1-F expressed from the native promoter (data not shown).
To evaluate the effect of depleting RLI1 on the rate of translation initiation, we examined the polysome profiles and quantified the ratio of polysomes to 80 S monosomes (P/M in Fig. 4) after shifting the P GAL -UBI-R-FH-RLI1 strain from galactose medium to glucose medium. We observed a substantial reduction in both the size and abundance of polysomes in the P GAL -UBI-R-FH-RLI1 strain after incubating for 4 h in glucose me-dium compared with that seen in the wild-type strain growing exponentially in the same medium (cf. Fig. 4, A and B). Importantly, the 2-mer polysomes were the most abundant species in the mutant cells at this time point. Both defects became progressively worse with a more prolonged incubation of the P GAL -UBI-R-FH-RLI1 strain in glucose medium for 8 h, with the polysomes/80 S monosomes ratio falling to ϳ5% of that seen in the wild-type strain. Thus, the depletion of R-FH-RLI1 in glucose medium led to a severe impairment of translation initiation. In these experiments, we consistently observed an increase in the abundance of free 40 S subunits relative to free 60 S ribosomal subunits upon depletion of R-FH-RLI1 in glucose medium (Fig. 4). This may indicate a defect in 60 S ribosome biogenesis, leading to a relative excess of 40 S subunits, in cells lacking RLI1. were grown on rich medium containing galactose (YPGal) and replica-plated to rich medium containing glucose (YPGlu) and incubated at 30°C for 2 days. C, the P GAL -UBI-R-FH-RLI1 strain YDH369 was grown in SC Gal to an A 600 of ϳ1.5, and the cells were harvested by centrifugation, washed with SC, and resuspended in SC at an A 600 of ϳ0.17. In parallel, another aliquot of these cells were diluted in SC Gal to an A 600 of ϳ0.17. The P GAL -UBI-M-FH-RLI1 strain YDH371 was grown in SC Gal to an A 600 of ϳ1.5 and diluted in SC Gal to an A 600 of ϳ0.17. Cell growth was measured every 2 h, and the cells were diluted four times with the same medium when the A 600 was between 1.0 and 1.5 to maintain exponential growth. The isogenic RLI1 strain BY4741 was grown in SC and analyzed in parallel. t.d., doubling time.

RLI1 Is Required for Efficient Binding of MFC Components to the 40 S Ribosome-We
of MFC components to 40 S ribosomes. We found that depletion of R-FH-RLI1 by growing the P GAL -UBI-R-FH-RLI1 strain for 12 h in glucose medium did not lead to reduced levels of eIF1, -2, -3, or -5 in cells that were cross-linked with HCHO prior to extract preparation (Fig. 5A). To analyze the integrity of the MFC, we overexpressed a hexahistidine-tagged form of eIF3c/ NIP1 in the P GAL -UBI-R-FH-RLI1 strain and affinity-purified the MFC following depletion of R-FH-RLI1. As shown in Fig.  5B, we observed no significant decrease in the association of subunits of eIF3, eIF2, or eIF5 with NIP1-His compared with that seen in the wild-type RLI1-F strain.
We then asked whether the binding of MFC components to 40 S subunits is diminished in vivo following depletion of R-FH-RLI1. To answer this question, we compared the amounts of initiation factors that co-sedimented with 40 S subunits in extracts of HCHO-cross-linked RLI1-F (wild-type) and P GAL -UBI-R-FH-RLI1 cells grown in glucose to deplete RLI1 from the latter strain. As shown in Fig. 2B, we observed a marked reduction in 40 S-bound eIF2 and eIF1, a lesser reduction in 40 S-associated eIF5, and relatively small decreases in 40 S binding by the eIF3 subunits and eIF1A. This experiment was carried out three times with very similar results. These results show that depletion of RLI1 decreases the association of eIF2 and eIF1 with 40 S subunits, reducing the abundance of fully assembled 43 S preinitiation complexes in vivo.
Conserved Residues in the ABC Cassettes of RLI1 Are Required for Its Essential Function-To determine whether the predicted ATPase activity of RLI1 is essential for its function in vivo, we conducted site-directed mutagenesis of conserved residues in the two ABCs of the protein. We first made mutations to replace individually the pairs of conserved Gly residues in the signature sequences of the ABCs with aspartate residues (G224D,G225D and G470D,G471D) (Fig. 6A). Mutations in the corresponding residues of ABC transporter proteins were shown to inactivate transporter function in vivo and ATPase activity in vitro (35,36). The double Asp substitutions were introduced into the P GAL -UBI-M-FH-RLI1 construct, and the resulting alleles were tested for the ability to rescue the growth of an rli1⌬ strain following eviction of wild-type plasmid-borne RLI1. Neither mutant construct could support the growth of rli1⌬ cells in medium containing glucose or galactose (Fig. 6B). Western analysis of viable strains harboring these mutant alleles and wild-type RLI1 showed that the UBI-M-FH-RLI1 mutant proteins were expressed at levels somewhat below that of the corresponding wild-type protein in galactose medium (Fig. 6C). However, based on the results in Fig. 3A, we can infer that the levels of the mutant UBI-M-FH-RLI1 proteins in galactose would be much higher than that of wild-type UBI-M-FH-RLI1 in glucose, and the latter is adequate to support growth. Hence, the failure of the mutant UBI-M-FH-RLI1 proteins to support growth in galactose medium clearly results from inactivation of RLI1 function rather than inadequate expression of the mutant proteins.
We also examined the effects of substituting the conserved Glu residue 493 in the Walker B motif of the second ABC domain of RLI1 (Fig. 6A). Mutation of this residue in several ABC transporters strongly impaired transport function and ATP hydrolysis (3,4,37,38), and this residue is thought to function as the catalytic carboxylate in the active site (4,39). Hence, we substituted Glu-493 with Gln in a functional RLI1 allele tagged at the C terminus with the coding sequences for five tandem c-myc epitopes that is expressed from the native promoter on a low copy plasmid. The E493Q mutation destroyed the ability of plasmid-borne RLI1-myc to rescue the growth of an rli1⌬ strain following eviction of wild-type RLI1 from the strain (Fig. 6D), but it had little or no effect on myc-RLI1 protein expression (Fig. 6E). Thus, we conclude that conserved residues known to be involved in ATP binding or hydrolysis in other ABC transporters are critical for the essential function of RLI1 in vivo.
Interestingly, overexpression from the GAL promoter of the RLI1-E493Q-F allele (containing the FLAG epitope at the C terminus) produced a dominant negative growth phenotype on galactose medium. Even overexpression of wild-type RLI1-F FIG. 4. Depletion of RLI1 reduces the amount and average size of polysomes in vivo. RLI1 strain BY4741 (A) and P GAL -UBI-R-FH-RLI1 strain YDH369 (B and C) were grown in SC Gal medium overnight to an A 600 of ϳ7. The saturated cultures were harvested, washed with SC medium, and resuspended in SC medium at A 600 ϳ 0.3 (RLI1 strain) or A 600 ϳ 0.5 (P GAL -UBI-R-FH-RLI1 strain). The cells were cultured for the times indicated in each panel (shown below the relevant genotype) and treated with 50 g/ml cycloheximide for 5 min to block translation elongation and freeze the polysomes immediately before harvesting. WCEs were prepared and resolved by velocity sedimentation through 7-47% sucrose gradients. The gradients were collected with continuous scanning at 254 nm. Positions of 40 S, 60 S, 80 S, and polysomes are indicated on the A 254 tracings. P/M, ratio of total polysomes to 80 S monosomes.
reduced the growth of otherwise wild-type cells, but the mutant protein clearly had a more pronounced inhibitory effect on growth (Fig. 6F). Analysis of polysome profiles indicates that overexpression of RLI1-E493Q-F from the GAL promoter reduced the abundance and average size of polysomes (Figs. 7, A  and C). Overexpression of wild-type RLI1-F also reduced the polysome/monosome ratio (Fig. 7B) but to a lesser extent than seen with overproduction of RLI1-E493Q-F. These findings suggest that interaction of the RLI1-E493Q-F protein with the 40 S ribosome or with the MFC interferes with translation initiation in vivo.
RLI1 Affects the Efficiency of Translation in Cell-free Extracts-We wished to determine whether depletion of RLI1 would reduce the efficiency of translation initiation in vitro. WCEs were prepared from the P GAL -UBI-R-FH-RLI1 strain following depletion of the R-FH-RLI1 protein in glucose medium and also from the isogenic WT RLII-F strain, and the two extracts were assayed in parallel for translation of a luciferase (LUC) reporter mRNA. Using the WT extract, we identified a range of LUC mRNA concentrations in which synthesis of luciferase occurred in proportion to the amount of mRNA used to program the extract (data not shown). Under these conditions, we found that the depleted extract supported a level of luciferase synthesis that was only ϳ15% of that given by the WT extract (Fig. 8B, y intercepts). We then tried to rescue translation activity in the depleted extract by adding RLI1-F purified from a strain overexpressing this protein from the P GAL -RLI1-F construct (Fig. 8A), by adding the purified RLI1-F to the extract in amounts equivalent to or severalfold higher than the level of RLI1-F present in extracts from a strain expressing this protein from the native RLI1 promoter. We observed no significant recovery of translational activity by adding purified RLI1-F to the depleted extract in several replicate experiments and no effect of adding RLI1-F to the WT extract (Fig. 7B). Thus, either the purified RLI-F is inactive or there is an additional defect in the depleted extract that cannot be corrected by replenishing RLI1-F.
Interestingly, we found that the addition of the purified mutant protein RLI1-E493Q-F strongly inhibited LUC mRNA translation in the WT extract in a dose-dependent manner (Fig.   FIG. 5. Depletion of RLI1 does not affect abundance of initiation factors or integrity of the MFC. A, RLI1-F strain YDH364 was grown in SC medium, and P GAL -UBI-R-FH-RLI1 strain YDH-369 was grown in SC Gal medium, at 30°C to saturation. The RLI1-F strain was diluted in SC medium to an A 600 of ϳ0.015, and the P GAL -UBI-R-FH-RLI1 strain was washed in SC medium and diluted in the same medium to an A 600 of ϳ0.5 and grown for ϳ12 h to an A 600 of ϳ1.2. Cells were cross-linked with HCHO, and the WCEs were subjected to Western analysis using antibodies against the proteins indicated on the right. B, RLI1-F strain YDH364 bearing high copy plasmid YEp-NIP1-His (p3928) was grown in SC-Ura-Leu medium overnight to saturation and then diluted in SC-Ura-Leu to an A 600 of ϳ0.015 and grown to an A 600 of ϳ1.2. The P GAL -UBI-R-FH-RLI1 strain (YDH369) bearing high copy plasmid YEpNIP1-his (p3924) was grown in SC Gal -Ura-Leu medium overnight to saturation, and the cells were harvested, washed, and diluted in SC-Ura-Leu medium to an A 600 of ϳ0.5 and grown to an A 600 of ϳ1.2. WCEs were prepared (without prior cross-linking of the cells) and incubated with Ni 2ϩ -nitrilotriacetic acid resin for 1 h at 4°C. The bound proteins were eluted with 300 mM imidazole and subjected to Western analysis using antibodies against the proteins indicated on the right. I, one-tenth of the input extract; P (from left to right), 14, 28, or 58% of the entire pellet fraction. S, one-tenth of the supernatant fraction. 7B). This inhibitory effect of RLI1-E493Q-F on translation in vitro is reminiscent of the dominant negative phenotype observed in WT cells overexpressing this mutant protein. These last findings are consistent with the idea that RLI1 has an important role in translation initiation that involves its ATPbinding cassettes.

RLI1 Is Present in Both Nucleus and Cytoplasm-If RLI1
participates directly in both translation and ribosome biogenesis, we would expect to find the protein localized in both cytoplasm and nucleus. This prediction was confirmed by our analysis of a GFP-tagged form of RLI1. The coding sequences for GFP were appended to the 3Ј-end of the RLI1 ORF in the . C, WCEs of the strains described in B were grown in SC Gal -Leu-Ura to an A 600 of ϳ1.2 and subjected to Western analysis with anti-FLAG antibodies. D, patches of rli1⌬ strains containing RLI1 on a URA3 plasmid (pDH177) and additionally containing plasmid-borne RLI1-E493Q-myc 5 (pDH202; row 1), RLI1-myc 5 (pDH98; row 2), or empty vector (p702; row 3) were replica-plated to SC-Leu-Ura (left panel) and 5-FOA medium (right panel). E, WCEs of the strains described in D were grown in SC-Leu-Ura to an A 600 of ϳ1.2 and subjected to Western analysis with anti-myc antibodies. F, serial dilutions of transformants of an rli1⌬ strain containing plasmid-borne RLI1-myc 5 and additionally harboring plasmid-borne P GAL -RLI1-F (YDH422; row 1), empty vector pEMBLyex4 (YDH425; row 2), or P GAL -RLI1-E493Q-F (YDH426; row 3) were spotted on SC-Leu-Ura medium (left panel) or on the same medium containing galactose instead of glucose (right panel). wt, wild type. chromosome, with no effect on the cell growth rate. As shown in Fig. 9, RLI1-GFP was present in both nucleus and cytoplasm. By contrast, a functional NEW1-GFP fusion constructed in parallel was restricted to the cytoplasm, showing nonfluorescent "holes" that coincided with the chromosomal DNA regions visualized by DAPI straining. Under the same conditions, a NOP1-GFP fusion was restricted to the nucleus, in agreement with previous results (40). DISCUSSION In this report, we show that RLI1 is specifically associated in vivo with eIF2, eIF5, and multiple subunits of eIF3 and that its interactions with eIF3 and eIF5 can be observed free of ribosomes. We demonstrated by HCHO cross-linking of live cells that RLI1 is specifically associated in vivo with 40 S subunits and polysomes, suggesting an interaction with 43 and 48 S preinitiation complexes. Depletion of RLI1 in vivo leads to a reduction in the abundance and average size of polyribosomes, indicating a decreased rate of general translation at the initiation step. Importantly, we observed a marked decrease in the amounts of eIF2 and eIF1 associated with 40 S subunits following depletion of RLI1, and a less dramatic decrease in 40 S-bound eIF5 was likewise observed (Fig. 2B). Thus, it appears that RLI1 is required for high level binding of eIF2, eIF1, and eIF5 to 40 S ribosomes and the formation of 43 and 48 S preinitiation complexes. Based on estimates of RLI1 abundance, it appears that there is considerably less RLI1 per cell than the subunits of eIF3, eIF2, or eIF5 (41). Thus, RLI1 may play a catalytic role in the assembly of preinitiation complexes on the 40 S ribosome rather than stabilizing these complexes as a stoichiometric component. Consistent with a catalytic function in preinitiation complex assembly, we showed that mutations in conserved residues in the ABCs predicted to be required for ATP hydrolysis inactivated the essential function of RLI1 in vivo. Furthermore, overexpression of one such mutant (RLI1-E493Q-F) had a dominant negative effect on cell growth and the rate of translation initiation.
Supporting the idea that RLI1 functions in translation initiation, we found that an extract depleted of RLI1 was defective for translation of a luciferase reporter mRNA, but unfortunately, we could not recover translational activity in this extract by adding purified RLI1. Thus, another critical initiation factor may be missing or damaged in the RLI1-depleted extracts, although we cannot rule out the possibility that the purified RLI1 was inactivated during purification. It is noteworthy, however, that the purified mutant protein RLI1-E493Q-F strongly inhibited translational activity of the WT extract when added in amounts comparable with the native level of RLI1. This dominant-interfering effect of RLI1-E493Q-F on LUC mRNA translation may be the in vitro correlate of the dominant negative growth phenotype observed when this mutant protein is overexpressed in otherwise WT cells. However, we have been unable to demonstrate that RLI1-E493Q-F impairs binding of Met-tRNA i Met to 40 S ribosomes in the inhibited extracts using a previously described assay for this partial reaction of the initiation pathway (29) (data not shown). Nor did we observe any effect of nonhydrolyzable ATP FIG. 7. Overexpression of RLI1-E493Q-F inhibits translation initiation in vivo. RLI1-F strain YDH364 (A), P GAL -RLI1-F strain YDH422 (B), and P GAL -RLI1-E493Q-F strain YDH426 (C) were grown in SC medium overnight. The saturated cultures were harvested, washed with SC Gal medium, and resuspended in SC Gal medium at A 600 ϳ 0.1. The cells were cultured to an A 600 of ϳ1.2 and treated with 50 g/ml cycloheximide for 5 min before harvesting. WCEs were prepared and resolved by velocity sedimentation through 7-47% sucrose gradients. The gradients were collected with continuous scanning at 254 nm. Positions of 40 S, 60 S, 80 S, and polysomes are indicated on the A 254 tracings. P/M, ratio of total polysomes to 80 S monosomes.
on Met-tRNA i Met binding to 40 S ribosomes in wild-type extracts (data not shown). It may be difficult to observe a reduction in rate versus a lower end-point of 40 S binding of Met-tRNA i Met binding using the in vitro assay we employed. Nevertheless, it is possible that RLI1 is not rate-limiting for 43 S complex formation in cell extracts. In this event, the inhibition of LUC mRNA translation by RLI1-E493Q-F in vitro could result from elevated binding of the mutant protein to 40 S subunits and interference with a postassembly step in the initiation process, such as scanning, GTP hydrolysis by eIF2 at the start codon, or 60 S subunit joining.
What role could an ABC protein play in the formation of translation preinitiation complexes? It can be predicted that binding of ATP would stimulate dimerization of the two ABC domains present in RLI1, with two molecules of ATP sandwiched between the dimerized cassettes. ATP hydrolysis should destabilize the dimer, and the cycle of dimerization driven by ATP binding and hydrolysis could perform mechanical work on the 40 S ribosome. A recent model for the function of the Rad50 ATPase in DNA repair posits that the ATP sandwich formed upon dimerization of the Rad50 ABCs contains two grooves to accommodate the broken ends of DNA destined for repair (42). By analogy with this model, it could be proposed that one of the ABCs in RLI1 interacts with eIF2 or eIF1 in the MFC and the other ABC binds to a site on the 40 S ribosome. In this event, ATP binding and dimerization could promote binding of the factor to the ribosome. Similar mechanisms could also be involved in isomerizing interactions between initiation factors or between a factor and the ribosome during postassembly reactions of the preinitiation complex, such as scanning or AUG selection.
Interestingly, Proud and co-workers (43) have reported that a mammalian nontransporter ABC protein, ABC50, interacts with eIF2 and stimulates TC formation in vitro. In addition, ABC50 co-sedimented with 40 and 60 S subunits in cell extracts in a manner stimulated by ATP. Mammals contain an ortholog of yeast RLI1, originally identified by its ability to inhibit RNase L (RLI stands for RNase L inhibitor) (44); thus, ABC50 and RLI1 are clearly distinct proteins. In addition, yeast RLI1 is more tightly associated with eIF3 and eIF5, whereas human ABC50 co-purified with eIF2. Nevertheless, it is intriguing that ABC50 and RLI1 both participate in different aspects of the recruitment of TC to 40 S subunits. At present, it is unclear how the function of mammalian RLI as an inhibitor of RNase L might be related to the role of yeast RLI1 in translation initiation, except that both functions involve ribonucleoprotein complexes. Since RLI1 is much more widely distributed in evolution than is RNase L (45), the inhibition of RNase L may be an auxiliary function restricted to mammals. It is intriguing that all four nontransporter ABC family members for which functional information is available, eEF3, GCN20, ABC50, and RLI1, interact with tRNA or ribosomes. Perhaps all of the members of this subfamily of ABC proteins are somehow connected with protein synthesis.
We found that a proportion of a functional RLI1-GFP fusion protein was localized in the nuclei of living yeast cells. In FIG. 8. Purified RLI1-E493Q-F inhibits translation of luciferase mRNA in vitro. A, SDS-PAGE analysis of FLAG affinity-purified RLI1-F and RLI1-E493Q-F proteins. Lanes 1 and 2 contain 2 and 4 l of eluate, respectively, from the FLAG resin for RLI1-F; lanes 3 and 4 contain 2 and 4 l of eluate, respectively, for RLI1-E493Q-F. The arrow indicates the RLI1-F and RLI1-E493Q-F proteins. B, depletion of RLI1 reduces the efficiency of LUC mRNA translation, and the addition of purified RLI1-E493Q-F, but not RLI1-F, inhibits translation in vitro. RLI1-F strain YDH364 was grown in SC medium to an A 600 of ϳ1.2. The P GAL -UBI-R-RLI1-FH strain YDH369 was grown in SC GAL medium to A 600 ϳ 7.0, and the cells were harvested and washed with SC, diluted into SC at A 600 of ϳ0.5, and grown for 8 h to an A 600 of ϳ1.2. The extracts were prepared for in vitro translation assays. Aliquots of extracts containing ϳ100 g of protein were programmed with 250 ng of LUC mRNA synthesized in vitro and incubated with the indicated amounts of the purified RLI1-F or RLI1-E493Q-F proteins shown in A. Luciferase synthesis was assayed by measuring luminescence after 20 min of incubation at room temperature, over which time luminescence increases in a linear fashion. The filled and unfilled triangles indicate assays using the extract from RLI1-F cells. The filled and unfilled circles indicate the assays using RLI1-depleted extracts. Each point represents the mean value calculated from three independent assays except that the unfilled circles are mean values from two independent assays, and the error bars indicate the S.E. values. The extract from the RLI1-F strain was about 1.35-fold more concentrated than that prepared from the P GAL -UBI-R-RLI1-FH strain.
FIG. 9. RLI1-GFP is located in both nucleus and cytoplasm. Isogenic strains derived from RLI1 strain BY4741 containing chromosomal RLI1-GFP (YDH391) or NEW1-GFP (YDH312) in place of the corresponding RLI1 and NEW1 alleles, respectively, were grown in YPD medium to an A 600 of ϳ0.5. The transformant of BY4741 containing plasmid-borne NOP1-GFP on a low copy plasmid was grown in SC-Ura to an A 600 of ϳ1.0 and then diluted in YPD at an A 600 of ϳ0.1 and grown to an A 600 of ϳ0.5. The cells were fixed and incubated with DAPI to stain the nuclear DNA. The green fluorescence protein moieties of the GFP fusion proteins were visualized by fluorescence microscopy of living cells. NOMARSKI, phase-contrast imaging of cells.
addition, we observed that depletion of RLI1 led to an excess of free 40 S versus 60 S subunits, suggesting a relative deficiency in 60 S subunits. These findings may indicate that RLI1 has a nuclear function in the biogenesis of 60 S ribosomal subunits. Although the bulk of RLI1 was found associated with 40 S versus 60 S subunits, the small fraction of RLI1 that we detected in the 60 S fraction (Fig. 2C) could be involved in 60 S subunit processing or nuclear export. Interestingly, we came to a similar conclusion for the substoichiometric j subunit of eIF3, HCR1, which has dual functions in 40 S biogenesis and translation initiation in yeast. In addition to being required for strong interactions among components of the MFC that enhance postassembly functions of the preinitiation complex (25), HCR1 is needed for optimal processing of 20 S to 18 S rRNA in the 40 S subunit and wild-type levels of 40 S subunits (46). Recent findings on the mammalian homologue of HCR1 indicate an important role for this protein in promoting interaction of eIF3 with the 40 S ribosome (47). Similarly, eIF3j/HCR may stimulate the binding to 40 S subunits of factors involved in 40 S maturation. It seems likely that additional ribosome-binding proteins will be uncovered that have dual functions in translation and ribosome biogenesis.