A Role for the Poly(A)-binding Protein Pab1p in PUF Protein-mediated Repression*

PUF proteins regulate translation and mRNA stability throughout eukaryotes. Using a cell-free translation assay, we examined the mechanisms of translational repression of PUF proteins in the budding yeast Saccharomyces cerevisiae. We demonstrate that the poly(A)-binding protein Pab1p is required for PUF-mediated translational repression for two distantly related PUF proteins: S. cerevisiae Puf5p and Caenorhabditis elegans FBF-2. Pab1p interacts with oligo(A) tracts in the HO 3′-UTR, a target of Puf5p, to dramatically enhance the efficiency of Puf5p repression. Both the Pab1p ability to activate translation and interact with eukaryotic initiation factor 4G (eIF4G) were required to observe maximal repression by Puf5p. Repression was also more efficient when Pab1p was bound in close proximity to Puf5p. Puf5p may disrupt translation initiation by interfering with the interaction between Pab1p and eIF4G. Finally, we demonstrate two separable mechanisms of translational repression employed by Puf5p: a Pab1p-dependent mechanism and a Pab1p-independent mechanism.

PUF proteins regulate translation and mRNA stability throughout eukaryotes. Using a cell-free translation assay, we examined the mechanisms of translational repression of PUF proteins in the budding yeast Saccharomyces cerevisiae. We demonstrate that the poly(A)-binding protein Pab1p is required for PUF-mediated translational repression for two distantly related PUF proteins: S. cerevisiae Puf5p and Caenorhabditis elegans FBF-2. Pab1p interacts with oligo(A) tracts in the HO 3-UTR, a target of Puf5p, to dramatically enhance the efficiency of Puf5p repression. Both the Pab1p ability to activate translation and interact with eukaryotic initiation factor 4G (eIF4G) were required to observe maximal repression by Puf5p. Repression was also more efficient when Pab1p was bound in close proximity to Puf5p. Puf5p may disrupt translation initiation by interfering with the interaction between Pab1p and eIF4G. Finally, we demonstrate two separable mechanisms of translational repression employed by Puf5p: a Pab1p-dependent mechanism and a Pab1p-independent mechanism.
Regulation of messenger RNA (mRNA) is an important part of gene expression. Each mRNA contains instructions that determine where and when it will be expressed. Many of the instructions are contained in the untranslated regions (UTRs) 2 of the mRNA (1,2). UTRs encode instructions for translation activation, repression, mRNA decay, or localization (3). The instructions take the form of primary sequences and structural elements that can be recognized by RNA-binding proteins (RBPs) and microRNAs (miRNAs). The combination of these factors assembled on each transcript determines the fate of each mRNA.
PUF proteins are a family of RBPs that regulate the translation and stability of mRNAs to which they bind, typically by repressing their expression (4 -9). PUFs recognize elements that contain a conserved UGU sequence located in the 3Ј-UTR of target mRNAs (8,10). PUF proteins often act as part of a larger regulatory complex, providing RNA binding specificity while other factors accomplish regulation (7,(11)(12)(13)(14).
PUF proteins utilize two mechanisms to repress mRNA expression: destabilization and translational repression (8). Shortening of the poly(A) tail is the first step of mRNA decay in eukaryotes (15). In yeast, Puf5p interacts with Pop2p, a member of the Ccr4/Not deadenylase complex, facilitating deadenylation and decay of target mRNAs (11,16). PUF-mediated recruitment of the Ccr4/Not complex has been demonstrated in Caenorhabditis elegans, Drosphila, and humans, suggesting that it is a conserved regulatory mechanism (11,12).
PUF proteins employ several different mechanisms to disrupt translation initiation. During the first step of initiation, the mRNA is circularized by the eIF4F complex, which then facilitates ribosome recruitment (17,18). Circularization is achieved by eIF4G binding to both the cap-binding protein eIF4E and the poly(A)-binding protein PABP, bringing both ends of the mRNA in proximity (19). Drosophila Pumilio disrupts circularization by recruiting two different proteins: Nanos, which recruits the eIF4E-binding protein Cup, and Brat, which recruits the cap-binding protein, 4E-HP (14,20,21). Xenopus Pum2 binds directly to the 7-methyl-G cap structure, and may compete with eIF4E although it has a relatively low affinity for the cap (22). Xenopus Pum1 may also recruit the eIF4E-binding protein 4E-T via CPEB (13,23,24). In yeast, the non-canonical PUF protein Puf6p is able to block 40 S ribosomal subunit recruitment, possibly through its interaction with eIF5B (25). Although the mechanisms vary, disrupting translation initiation is a conserved method of PUF-mediated repression.
A PUF protein can participate in multiple regulatory complexes, each with different components and outcomes. For example, the C. elegans PUF protein FBF promotes opposite fates for the same mRNA at different stages of development; The gld-1 mRNA can be repressed or activated by FBF in different regions of the germ line (7). Even PUF complexes that serve the same function can vary; Pumilio forms different repression complexes in Drosophila depending on the target mRNA. The cyclin B mRNA is regulated by only Pumilio and Nanos, whereas repression of hunchback requires a complex of Pumilio, Nanos, and Brat (20). The combinatorial interactions between each of the components in a regulatory complex control the outcome of regulation.
We sought to characterize PUF regulation in yeast. Using the HO 3Ј-UTR as a model target, we first identified several sequence elements in the 3Ј-UTR, outside the PUF binding site that enhanced PUF repression. We show that Pab1p binds to these elements to promote optimal Puf5p repression and that Pab1p must be able to activate translation in order to facilitate repression. Pab1p may be part of a general mechanism of translational control by PUF proteins.
Yeast Extracts and Cell-free Translation Assays-Yeast extracts were prepared as described (26). Briefly, each yeast strain was grown to saturation (OD660 Ն 2.0) in 1-4 liters of YPAD media at 30°C. The yeast were washed and lysed in liquid nitrogen by grinding with a mortar and pestle. The soluble fraction of the lysate was filtered through a G25 Sephadex column and subsequently used in translation reactions.
Cell-free translation assays were performed exactly as described (26). Briefly, translation reactions were assembled using 60 g of yeast extract, 10 ng of firefly reporter mRNA, 30 ng of Renilla reporter mRNA, 2.5 l of 6ϫ translation buffer, 1 l of 4 mg/ml creatine kinase, and 0.1 l of RNasin (Promega) in a 15 l reaction. Reactions were incubated for 1 h at 30°C before luciferase levels were measured using the Dual-Luciferase Reporter Assay System from Promega.
To obtain the ratio of firefly to Renilla luciferase measurements, the firefly reading was divided by the Renilla reading from each reaction. Then the basal translation level (with no added protein) was set to 1, and all ratios for the same RNA were scaled to that basal translation level. The standard deviation was calculated from 3 replicates for each sample for every experiment.
Purified recombinant Puf5p was added to extracts to a final concentration between 200 -350 nM. Half-maximal repression required 45 nM protein (26). The apparent K d of the protein for its binding site in the HO 3Ј-UTR is ϳ100 nM under these conditions (27). When noted, purified recombinant Pab1p was added to a final concentration of 300 -600 nM.
Unless indicated otherwise, the yeast strains used to derive translation extracts possessed a wild-type PUF5 gene. We directly compared these extracts to puf5 deletion extracts by titrating in purified Puf5p into both. Puf5p activity was identical in both strains (data not shown), including the maximal level of repression observed, indicating endogenous Puf5p is either inactive at stationary phase or inactivated by the extraction process.
In Vitro Transcribed Reporter mRNAs-All reporter mRNAs were in vitro transcribed from PCR products using T3/T7/SP6 MEGAscript kits from Ambion with 20% of the suggested GTP and 6 mM m 7 G cap analog (New England Biolabs). PCR products were purified by phenol:chloroform extraction and ethanol precipitation prior to use as templates for transcription. The control Renilla reporters were derived from the pSP65-Ren plasmid (26). The Can1-Pgk1 firefly reporter in Fig. 1B is the same as the Renilla reporter from pSP65-Ren except that it contains the firefly luciferase ORF instead of the Renilla luciferase ORF. For the 1xBS reporter, the Puf5p binding site from the Cin8 3Ј-UTR was inserted into the Pgk1 3Ј-UTR: AGTTG-TAATATTAAATAGCT. For the 3xBS reporter, the Puf5p site from the HO 3Ј-UTR was inserted in triplicate: AGTTGTATGTAAT.
The HO firefly reporters were derived from the pYC2-HO plasmid as described (26). HO mutants 1-6 contained 3 nt transversion mutations as shown in Fig. 1A. Each mutation was also optimized to avoid creating new secondary structures in the 3Ј-UTR. The 2x(A) mutant is equivalent to mutant 2 from The E1 and 185 spacer mutations in the HO reporter have been described previously (26). The HO mut reporter that lacks PUF binding sites in Fig. 6B contains mutations in the Puf4p and Puf5p binding sites such that all three UGU sequences are mutated to ACA.
The other 8 reporters tested were created by ligation of the firefly luciferase ORF and the desired 3Ј-UTR from each gene and then PCR-selecting for the correct product. The lengths of the 3Ј-UTRs were based on data from a tiling array on the yeast transcriptome (28) and are as follows: HO, 67 nt; Ase1, 318 nt; Cin8, 253 nt; Dhh1, 167 nt; Lrg1, 213 nt; Rax2, 89 nt; Ypp1, 112 nt; and Pgk1, 76 nt. The PCR products were cut and inserted into the pRS416-TEF vector using the HindIII and BamHI sites. The PCR template for in vitro transcription was amplified from the reporter plasmid with the T3 promoter sequence and contained the start of the luciferase ORF to directly downstream of the HindIII site. For adenylated transcripts, the reverse primer encoded an additional 50 nt poly(A) tail. For the Cin8 (A) del mutant, 13 of the 15 As downstream of the Puf5p site were deleted. The remaining 2 As were left to avoid disrupting the Puf5p binding site. The FBF target reporters, 0 FBE and 3 FBE, have previously been described (26).
EMSA Assays-Electrophoretic mobility shift assays were performed exactly as described (31) with the exception of the binding reaction buffer, which was comprised of 10 mM HEPES (pH 7.4), 1 mM EDTA, 50 mM KCl, 2 mM DTT, 0.1 mg/ml BSA, 0.02% Tween 20, and 60 ng/l competitor RNA. The competi-tor RNA was yeast tRNA from Ambion. The K D values and standard errors for each RNA/protein interaction were derived from an average of at least three independent experiments.

RESULTS
Oligo(A) Tracts Are Important for Puf5p-mediated Repression-We previously developed an in vitro translation system that supports PUF-mediated repression ( Fig. 1A) (26). The system employs two different luciferase reporter mRNAs: a firefly reporter that contains the 3Ј-UTR of a PUF target (such as HO) and a Renilla reporter that acts as an internal control for variations in translation. The reporter mRNAs are combined with cytoplasmic yeast extracts. After 1 h, the levels of luciferase are measured. To account for variations in translation between samples, firefly readings are normalized to the Renilla readings for each sample and the basal level of translation, absent any PUF protein, for each mRNA.
Regulation by Puf5p requires the presence of PUF binding sites in the HO 3Ј-UTR (26). We tested whether a PUF binding A, diagram of the cell-free translation assay. Two different reporter mRNAs encoding firefly and Renilla luciferase were added to each translation reaction. The Renilla reporter served as a control while the firefly reporter typically contained the 3Ј-UTR of a PUF target mRNA. The reporters were combined with purified yeast cytoplasmic extracts and purified recombinant proteins, where noted, and allowed to translate for 1 h at 30°C. Next, the levels of luciferase activity were measured by quantifying luminescence. B, we measured the translation of several reporter mRNAs in response to Puf5p. The light gray bars in each graph indicate the basal level of translation for each mRNA while dark gray bars indicate translation in presence of Puf5p. The HO mRNA is a target of both Puf4p and Puf5p and thus contains two PUF binding sites in its 3Ј-UTR (27). However, the Puf4p site does not contribute substantially to Puf5p regulation (26). A firefly reporter with the Can1 5Ј-UTR and the Pgk1 3Ј-UTR was used as a template for the insertion of Puf5p binding sites. The Pgk1, 1xBS and 3xBS reporters contain 0, 1 or 3 Puf5p binding sites in their 3Ј-UTR, respectively. C, to isolate sequences important in Puf5p repression, we made mutant reporters that contained 3 nt transversions across selected regions in the the HO 3Ј-UTR. Each mutant 3Ј-UTR sequence was analyzed using RNAfold to avoid creating stable secondary structures. D, each of the 3 oligo(A) tracts highlighted in part C was mutated in a series of reporters. To eliminate complications from close proximity of the poly(A) tail, the poly(A) tail was removed from these reporters. E, same reporters as in part D were assayed for Puf5p repression with poly(A) tails.
site alone was sufficient for regulation by placing a Puf5p binding site in the Pgk1 3Ј-UTR. Pgk1 is not a target of Puf5p and has no Puf5p binding sites. The chimeric 3Ј-UTR did not confer repression (Fig. 1B). With insertion of three Puf5p binding sites, we observed only modest repression relative to HO, which contains a single Puf5p binding site. We infer that sequences in the HO 3Ј-UTR outside the PUF binding site enhance repression.
To identify these sequences, we generated a series of 3 nt mutations across the 67 nt HO 3Ј-UTR, with the PUF binding sites intact (Fig. 1C). Mutants 2 and 4 manifested a small but significant disruption of repression (Fig. 1C). Both mutations eliminate "AAA" trinucleotides close to the Puf5p binding site. To further characterize the importance of these oligo(A) tracts, we mutated all three oligo(A) tracts in the HO 3Ј-UTR (underlined in Fig. 1C) in various combinations. Because the presence of the poly(A) tail could complicate analysis of the oligo(A) tracts, we first tested reporters lacking poly(A) tails. As the number of oligo(A) tracts in the HO 3Ј-UTR decreased, Puf5p repression was dramatically relieved (Fig. 1D). The same trend was observed for oligo(A) mutations in reporters that contained poly(A) tails, although the magnitude of the effect was reduced (Fig. 1E).
To determine whether the oligo(A) tract enhancers in the HO 3Ј-UTR were a general feature of Puf5p targets, we examined six other Puf5p targets: Ase1, Cin8, Dhh1, Lrg1, Rax2, and Ypp1. All six targets were significantly repressed by Puf5p, though to varying extents ( Fig. 2A). Because efficient repression of HO by Puf5p requires either oligo(A) tracts or a poly(A) tail, we assayed repression of the six reporters without a poly(A) tail.
Four of the seven targets tested, Ase1, Cin8, Rax2, and HO, remained repressed in the absence of a poly(A) tail (Fig. 2B). Targets that were repressed by Puf5p without a poly(A) tail contained oligo(A) tracts (of 4 or more As) within 15 nt of the Puf5p binding site (Fig. 2C).
The Cin8 3Ј-UTR contains a 15 nt oligo(A) tract directly downstream of the Puf5p binding site. We tested whether the oligo(A) tract was important for Puf5p regulation by excising the oligo(A) tract from the Cin8 3Ј-UTR. Deletion of the oligo(A) tract slightly decreased repression of both poly(A) plus and minus mRNAs (Fig. 2D). These results mirror those of the HO reporter, where oligo(A) tracts are required for Puf5-mediated repression, especially in the absence of a poly(A) tail.
Pab1p Is Required for PUF-mediated Repression-The importance of oligo(A) elements suggested that a protein with high affinity for oligo(A) sequences might be important for Puf5p-mediated repression. Because Pab1p binds A-rich RNA sequences (32), we tested whether it was required for Puf5p repression. We assayed Puf5p repression in yeast extracts derived from a pab1⌬/pbp1⌬ double deletion strain; Deletion of Pbp1 was necessary to suppress the lethality of the Pap1 deletion (33). In extracts from pab1⌬/pbp1⌬ strain Puf5p repression was severely compromised compared with repression in a comparable wild-type strain (Fig. 3A). To test whether this effect was due to the absence of Pab1p, we added back recombinant yeast Pab1p purified from E. coli. Addition of Pab1p activated translation (supplemental Fig. S1a) and partially restored repression by Puf5p (Fig. 3A). Similarly, an HO reporter lacking a poly(A) tail was not repressed in Pab1p deficient extracts but addition of Pab1p restored repression (Fig.  3B). Addition of Pab1p also enhanced translation of the HO reporter without a poly(A) tail but to a lesser extent (supplemental Fig. S1b).
To control for the effects of pbp1⌬, we assayed repression in a pbp1⌬ extract. Puf5p repression was unaffected (Fig. 3C). To further demonstrate loss of Pab1p disrupted repression, we assayed Puf5p repression in another yeast strain deficient for Pab1p: pab1⌬/rpl39⌬. Puf5p repression of HO was completely abolished in pab1⌬/rpl39⌬ extracts regardless of the presence of a poly(A) tail (Fig. 3D). The requirement for Pab1p was not restricted to regulation of the HO reporter as four other Puf5p targets, Ase1, Cin8, Rax2, and Ypp1, also required Pab1p to be repressed (Fig. 3E). The C. elegans PUF protein FBF-2 was previously shown to repress a target mRNA in yeast extracts (26). To determine whether Pab1p was part of a more general regu-FIGURE 3. Pab1p is required for Puf5p repression. A, to test whether Pab1p was required for repression, we derived yeast extracts from a pab1⌬/pbp1⌬ yeast strain where deletion of pbp1 supresses the lethality of the pab1 deletion. The HO reporter with a poly(A) tail was tested in the Puf5p repression assay using either WT or Pab1p-null extracts. The light gray bars in each graph indicate the basal level of translation for each mRNA while dark gray bars indicate translation in presence of Puf5p. In the second panel, purified GST-Pab1p was added back to the translation reactions lacking Pab1p. In experiments with added Pab1p, the light gray striped bars in each graph indicate the basal level of translation for each mRNA while dark gray striped bars indicate translation in presence of Puf5p. B, similar to part A, Puf5p repression was assayed in WT and pab1⌬/pbp1⌬ extracts using a reporter lacking a poly(A) tail. The requirement for Pab1p was even more obvious in the absence of a poly(A) tail. C, as a control for the absence of Pbp1p in pab1⌬/pbp1⌬ extracts, Puf5p repression was assayed in a Pbp1p-null strain. D, lethality of a pab1⌬ can be suppressed by several mutations. Translation extracts were prepared from a double deletion strain where deletion of rpl39 suppresses the lethality of pab1⌬. Puf5p repression was assayed for the HO reporter with and without a poly(A) tail. E, four other reporters that were repressed by Puf5p in WT extracts ( Fig. 2A) were tested for Puf5p repression in pab1⌬/pbp1⌬ extracts. F, C. elegans PUF protein FBF-2 represses translation in yeast extracts (26). We assayed repression by FBF-2 in WT and pab1⌬/pbp1⌬ extracts using reporters that contained either 0 or 3 FBF binding elements (FBEs). latory mechanism for PUF proteins, we tested whether FBF-2 could repress in the absence of Pab1p. Indeed, repression by FBF-2 was nearly eliminated in the absence of Pab1p (Fig. 3F). We conclude that Pab1p may be a part of a conserved translational control mechanism employed by PUF proteins.

Pab1p Binds to the Oligo(A) Tracts in the HO 3Ј-UTR-
The data thus far suggested that Pab1p might bind the oligo(A) tracts in the 3Ј-UTR to mediate repression. To demonstrate the requirement for both oligo(A) tracts and Pab1p in Puf5p repression, we assayed whether absence of oligo(A) tracts would prevent Pab1p from facilitating Puf5p repression in pab1⌬/pbp1⌬ extracts. Indeed, the ability of Pab1p to facilitate repression was proportional to the number of oligo(A) tracts in the 3Ј-UTR (Fig. 4). An HO reporter lacking all three oligo(A) tracts and a poly(A) tail was no longer repressed by Puf5p. Thus, Pab1p requires oligo(A) tracts or a poly(A) tail to facilitate Puf5p repression.
These findings implied that Pab1p interacts directly with the oligo(A) tracts in the HO 3Ј-UTR. To test this directly, we measured binding of purified recombinant Pab1p to the HO 3Ј-UTR in electrophoretic gel shift assays (EMSAs). Pab1p bound to the HO 3Ј-UTR (lacking a poly(A) tail) with an affinity of ϳ28 nM while mutations that eliminated the oligo(A) tracts (HO no(A)s) reduced binding at least 30-fold (Fig. 5). Pab1p bound to a poly(A) RNA with an affinity of ϳ1 nM (supplemental Fig. S2) demonstrating that Pab1p has a higher affinity for poly(A) versus internal oligo(A) sequences. We conclude that Pab1p binds oligo(A) sequences within the HO 3Ј-UTR in vitro.

RRMs 1 and 2 and the C Terminus of Pab1p
Are Required for Repression-Pab1p contains four RNA-recognition motifs (RRM), a linker domain and a C-terminal helical domain known as "PABC" or the "MLLE" domain (34,35). To identify domains necessary to facilitate Puf5p-mediated repression, we constructed a set of Pab1p mutants (Fig. 6A). Using these purified mutant proteins, we supplemented pab1-null extracts with each type of mutant Pab1p protein.
We assayed whether each Pab1p mutant facilitated Puf5p  repression (Fig. 6B) or enhanced translation of the control Renilla reporter (Fig. 6C). Table 1 summarizes the results of these experiments.
Three mutant Pab1p proteins were capable of promoting Puf5p repression: ⌬RRM3-4, ⌬linker, and ⌬MLLE (Fig. 6B) suggesting these domains are dispensable for repression. Simultaneous deletion of the linker and the MLLE domain eliminated repression activity (Fig. 6B, ⌬C-term), indicating that the linker and the MLLE domains have redundant functions in repression. RRMs 1 and 2 were the only domains that were absolutely required for repression.
RNA binding activity was required for Pab1p to mediate Puf5p repression. Pab1p was only able to support Puf5p repression if two or more RRMs were present (Fig. 6B). However, the RRMs were not functionally equivalent; RRMs 1 and 2 were required for Puf5p repression while 3 and 4 were dispensable (Fig. 6B, ⌬RRM1-2 versus ⌬RRM3-4). RRMs 1 and 2 are  However, RRM 2 is also required for the interaction with eIF4G (37). The Pab1p 4G mut cannot bind eIF4G due to a mutation in RRM 2 and therefore also lacks the ability to activate translation (30). The 4G mutant was unable to promote translation (Fig.  6C) or facilitate Puf5p repression (Fig. 6B). The same eIF4G mutant is active in stimulating cap-dependent, poly(A) tail-independent translation (30). These data suggest that the interaction between eIF4G and Pab1p may be required for Puf5p repression.
Two Distinct Repression Mechanisms-Puf5p relies on two distinct mechanisms of translational repression: a Pab1p-dependent mechanism and a Pab1p-independent mechanism. The existence of two mechanisms is evident in Fig. 7; The HO reporter was still repressed, albeit inefficiently, in the absence of Pab1p, demonstrating Puf5p can repress through an additional, Pab1p-independent mechanism. Previous experiments uncovered another mutation in the HO 3Ј-UTR that disrupted repression: the 185 spacer mutation increases the distance between the ORF and the PUF sites (26). In WT extracts, the 185 spacer mutation moderately decreased repression by ϳ20% (26). In contrast, the spacer mutation eliminated repression of HO in pab1⌬/pbp1⌬ extracts (Fig. 7). Thus, we conclude that the effect of oligo(A) tracts and spacing are additive, indicating there are two separable mechanisms contributing to Puf5p-mediated translation repression: a Pab1p-dependent mechanism and Pab1p-independent mechanism.

DISCUSSION
We set out to understand the role of 3Ј-UTR regulatory elements in Puf5p regulation in vitro. We draw three main conclusions. First, oligo(A) elements in the HO 3Ј-UTR enhance Puf5p-mediated repression. Second, the oligo(A) elements recruit Pab1p. Third, Puf5p requires Pab1p to efficiently promote translational repression. We conclude that Pab1p bound to the 3Ј-UTR enhances repression by PUF proteins and Puf5p functions through Pab1p-dependent and -independent mechanisms.
Pab1p-dependent Repression-Pab1p could be required for PUF-mediated repression either because Puf5p interferes with Pab1p activation or because Pab1p can inhibit translation. We favor the former hypothesis because Pab1p translation activation was required to observe Puf5p-mediated repression. The Pab1p 4G mutant, which cannot interact with eIF4G, could not activate translation and was incapable of facilitating Puf5p repression (Table 1, 4G mut).
Puf5p may interfere with the interaction between Pab1p and eIF4G. The interaction between eIF4G and Pab1p was specifically required to observe Puf5p-mediated repression (Fig. 6B,  4G mut). Furthermore, Puf5p repression did not extend beyond Pab1p ability to activate translation (supplemental Fig. S1b). The proximity of Pab1p to Puf5p also affected Puf5p ability to repress. Puf5p efficiently repressed translation when only the oligo(A) tracts were present (Fig. 1D) but repression was less effective if Pab1p was more distantly located on the poly(A) tail (Fig. 1E). In one simple model, Puf5p represses translation by disrupting the interaction between Pab1p and eIF4G, an activity that might be promoted by physical proximity.
The connection between Pab1p and Puf5p activity could be direct or mediated by an as yet unidentified protein. We did not detect an interaction between the two recombinant proteins, despite each protein being able to bind RNA specifically (data not shown).
Pab1p Binding to Short Oligo(A) Tracts-Our data also imply that Pab1p may bind more promiscuously than previously thought. One reporter, Ypp1, lacked obvious Pab1p binding sites yet still required Pab1p to be repressed (Figs. 2C and 3E). It is possible that Pab1p may bind other elements in target 3Ј-UTRs to facilitate repression. Repression of Pab1p translation activation may be a general mechanism of PUF-mediated control; Another PUF protein, Puf3p, was recently isolated in a complex with Pab1p (38). Similarly, repression by the C. elegans PUF protein, FBF-2, was dependent on Pab1p (Fig. 3F).
Based on experiments with Pab1p mutant proteins, only two of the four RRMs (RRMs 1 and 2) were necessary for Puf5p repression ( Table 1). The requirement for only these two RRMs suggests they may be important for Pab1p binding to oligo(A) tracts.
Pab1p-independent Repression-Our data show that there are two distinct mechanisms of translational repression employed by Puf5p: a Pab1p-dependent mechanism and a Pab1p-independent mechanism. The existence of two mechanisms is evident in Fig. 3, A and B; The HO reporter was efficiently repressed in WT extract, partially repressed in pab1⌬/ pbp1⌬ extract, and no longer repressed when HO lacked a poly(A) tail in pab1⌬/pbp1⌬ extract. Thus, the Pab1p-independent mechanism relies on the presence of a poly(A) tail.
The residual repression observed in pab1⌬/pbp1⌬ extract could be due to other poly(A) binding proteins. Nab2p and Sgn1p also interact with poly(A) sequences (39,40) which may explain why the poly(A) tail is necessary to observe repression in the absence of Pab1p. Nab2p and/or Sgn1p may be able to facilitate repression via the poly(A) tail while only Pab1p is able to bind internal oligo(A) sequences. Alternatively, Sgn1p and Nab2p may compete with Pab1p for binding to poly(A) and oligo(A) RNA and thereby inhibit Puf5p repression. In pab1⌬/rpl39⌬ extracts, Puf5p repression of HO is entirely abolished regardless of poly(A) status (Fig. 3D), suggesting Rpl39p may be involved in Puf5p repression either directly or indirectly. Rpl39p is part of the 60 S ribosomal subunit and participates in a gating mechanism that controls release of the polypeptide chain (41,42). In conclusion, the Pab1p-independent translational repression mechanism requires a poly(A) tail, close proximity to the ORF, and possibly Rpl39p. Elucidating the molecular mechanisms involved will require understanding how these elements combine to facilitate translational repression.
Broad Implications; Pab1p in Repression-PABP activates translation by enhancing translation initiation. In a prevalent model, it does so by interacting with eIF4G, which in turn binds the cap-binding protein ( Fig. 8A and see Introduction). Several reports suggest that Pab1p also has roles in translational repression. In the simplest instances, repressors bound to the 3Ј-UTR interfere with the ability of PABP to bind eIF4G. For example, the protein GW182, which is recruited by miRNAs via Argonaute, binds PABP and competes with the binding of eIF4G (Fig. 8B) (34,(43)(44)(45). In this way, the miRNA bound to the 3Ј-UTR prevents PABP's activation function. The Musashi (MSI) repressor protein acts similarly (Fig. 8B) (46). PABP can also promote repression by binding to repressors to form complexes that interfere with initiation at a different step. For example, the protein UNR binds both SXL and PABP, and the SXL-UNR-PABP complex bound to the 3Ј-UTR prevents recruitment of the 43 S pre-initiation complex to the RNA (Fig.   FIGURE 8. PABP is involved in multiple repression mechanisms. PABP is involved in several cases of translational repression. Each panel shows one example of translational control involving PABP (shown in dark gray). A, simplified model of translation initiation. mRNA circularization occurs via the interactions between PABP, eIF4G, and eIF4E. eIF4G interacts with eIF3 to recruit the 40 S ribosomal subunit. The 40 S subunit then scans across the 5Ј-UTR for the start codon, whereupon the 60 S subunit joins and translation elongation begins. B, in some cases, translational activation by PABP is blocked as is the case for the miRNA/GW182 complex and Musashi. GW182, a component of RISC, interacts with PABP and blocks the eIF4G/PABP interaction, inhibiting translation initiation (34,(43)(44)(45). In addition, the complex facilitates recruitment of the CAF1/CCR4 deadenylase complex to enhance mRNA decay (43,45). Similar to GW182, Musashi competes with eIF4G for binding to PABP (46). C, in other cases, PABP is a member of the repression complex. For example, the SXL-UNR complex in Drosophila silences the msl-2 mRNA by inhibiting 40 S subunit recruitment (47,54). PABP interacts directly with UNR and is required for translational repression of the msl-2 mRNA by the SXL-UNR complex (47,55). The mechanism of translational repression remains unknown but lies downstream of mRNA circularization and eIF4F recruitment (47). In PABP auto-regulation, PABP auto-regulates its own expression by binding to a conserved element in the 5Ј-UTR of its mRNA termed the ARS, or A-rich auto-regulatory sequence (48 -50). A tripartite complex of RNA-binding proteins assembles on the ARS to promote repression: IMP1, UNR, and PABP (49,56). Presumably the RNP complex sterically blocks 40 S ribosome scanning, but repression requires the MLLE domain of PABP, indicating that protein-protein interactions may also be important (53,57). D, Puf5p disrupts PABP-mediated translational activation, possibly by interfering with the interaction between Pab1p and eIF4G (this work). PABP binding in close proximity to Puf5p in the 3Ј-UTR may enhance the Puf5p ability to disrupt the PABP/eIF4G interaction. In addition, Puf5p recruits the CAF1/CCR4 deadenylase complex through an interaction with CAF1/Pop2 (11). 8C) (47). Similarly, a UNR-IMP1-PABP complex bound to the 5Ј-UTR appears to block scanning of PABP own mRNA (Fig.  8C) (48 -50).
Comparisons of the work reported here to these examples is instructive. Repression by Puf5p is Pab1p-dependent, and blocked by point mutations that disrupt the Pab1p ability to bind eIF4G (Fig. 8D). Thus the PUF protein is likely to interfere with eIF4G binding as in Fig. 8B, though this has not been demonstrated directly. Puf5p also recruits the Caf1p/Pop2p deadenylase complex to promote mRNA deadenylation and instability (11,16). In these respects, repression by PUF proteins directly parallels that by the miRNA/GW182 complex. The duality of mechanisms may ensure repression is complete.
Our work shows that Pab1p binds directly to oligo(A) segments within the 3Ј-UTR of the HO mRNA. These oligo(A) tracts are reminiscent of the oligo(A) segements in the 5Ј-UTR that PABP binds to auto-regulate its own mRNA (48 -50). Similarly, PABP binds to the 3Ј-UTR of the non-polyadenylated Dengue Virus RNA genome, near structured elements (51). Oligo(A) segments containing as few as four contiguous adenosine residues promoted Pab1p binding in our experiments. These data suggest that internally bound oligo(A) segments may be much more prevalent than previously thought. The nature of "non-canonical" PABP/RNA interactions and their specificities are important unsolved problems.
Our work has demonstrated a key role for Pab1p in repression by PUF proteins in vitro. In vivo, the balance between occupancy of the poly(A) tail and the oligo(A) segments in the 3Ј-UTR is not known. The presence of internal oligo(A) segments may anchor Pab1p after a minimal poly(A) length is reached during deadenylation. In that fashion, continued repression would be ensured, and decay enhanced. Indeed, like Puf5p, yeast Puf3p interacts with Pab1p and enhances the rate of decapping of target mRNAs (52).
Each 3Ј-UTR nucleates the assembly of a unique regulatory complex responsible for mRNA expression. In turn, the context of each sequence element can drastically alter the outcome of a regulatory complex; A Puf5p binding site is not sufficient to confer repression in a non-native 3Ј-UTR but in its native context reduces translation 10-fold (Fig. 1A). It appears that regulators such as PUF proteins must employ redundant mechanisms to ensure proper expression. Deciphering the code of 3Ј-UTR regulation will require a better understanding of the combinatorial interactions of RNA regulators and the multiplicity of their outcomes.