Discrete binding determinants for eIF4A and eIF4E within the yeast Ded1 protein were unknown, which has made it difficult to assess the physiological importance of each interaction in vivo. To identify amino acids important for each interaction, we assayed binding of bacterially expressed glutathione-S-transferase (GST) fusions made to full-length yeast eIF4A1 or eIF4E to in vitro translated [³⁵S]-labeled Ded1 polypeptides, beginning with full-length (FL) Ded1 (amino acids 1–604) and truncated Ded1 variants lacking NTD residues 2–92 (ΔN_2-92), CTD residues 562–604 (ΔC_562-604), or both domains (ΔNC) (Figure 1A). The labeled Ded1 polypeptides recovered with GST-eIF4A, GST-eIF4E, or GST alone on glutathione-agarose beads were resolved by SDS-PAGE, with the result that the FL Ded1 polypeptide bound to both GST-eIF4A and GST-eIF4E, but not GST alone (Figure 1B, lanes 9 and 13 vs. 5), confirming specific, separate interactions of Ded1 with both eIF4A and eIF4E. These and all subsequent pull-down assays included RNAase treatment to insure that the interactions were not bridged by RNA.

The amounts of bound Ded1 polypeptides were quantified and normalized to the input amounts for each reaction, and mean percentages of the input amounts that bound to GST-eIF4A or GST-eIF4E in replicate pull-down experiments were compared between FL Ded1 and the three truncated Ded1 variants. For GST-eIF4A, binding of the ΔC_562-604 and FL Ded1 polypeptides were comparable, whereas drastically reduced amounts of the ΔN_2-92 and ΔNC variants were bound (Figure 1C). These results support previous findings that eIF4A interacts specifically with the Ded1 NTD (Gao et al., 2016). Eliminating the NTD reduced Ded1 binding to GST-eIF4E as well, although binding was also reduced to a lesser extent by deleting the Ded1 CTD (Figure 1D). Importantly, however, comparing the results for ΔNC and ΔN_2-92 revealed that the low-level binding conferred by ΔN_2-92 was not significantly diminished by ΔC_562-604, whereas ΔN_2-92 exacerbated the moderate binding defect of ΔC_562-604 (Figure 1D). These results suggest that the Ded1 NTD is considerably more important than the CTD for eIF4E binding, and provide the first evidence that eIF4E binds directly to the yeast Ded1 NTD.

Ded1 has also been reported to interact through its C-terminus with eIF4G (Gao et al., 2016; Hilliker et al., 2011; Senissar et al., 2014). We verified this interaction under the conditions of our binding assay by determining that the ΔC_562-604 truncation was sufficient to abolish binding by [³⁵S]-labeled Ded1 to a GST-eIF4G fusion expressed in bacteria (Figure 1—figure supplement 1, lanes 10 and 12). Thus, the Ded1 CTD is essential for binding to eIF4G but is either dispensable or relatively unimportant for interactions with eIF4A or eIF4E, respectively.

To identify specific Ded1 residues critical for binding eIF4A and eIF4E in vitro, we generated [³⁵S]-labeled variants of the FL Ded1 polypeptide containing clustered alanine substitutions, or in one case a deletion, of ten blocks of conserved amino acids located throughout the NTD (Figure 2A). While the main consideration for sequence conservation was similarity among other Saccharomycetaceae species (Figure 2—figure supplement 1A), most of these clusters include residues also conserved in higher eukaryotes (Figure 2—figure supplement 1B). Using the same GST pull-down assay as above, we observed that substituting Ded1 NTD residues 21–27, 29–35 and 51–57 conferred the greatest reductions in binding to GST-eIF4A (Figure 2B, cols. 5, 6, 8, bars in green hues), whereas residues 59–65 and 83–89 appear to be most important for binding to GST-eIF4E (Figure 2C, cols. 9,11, orange hues). These findings suggested that the critical binding determinants for eIF4A and eIF4E are located in adjacent, non-overlapping segments of the Ded1 NTD, located within residues 21–57 for eIF4A and within residues 59–89 for eIF4E.

Figure 2 with 4 supplements see all

Download asset Open asset

To determine whether the non-contiguous binding determinants within each of these two intervals make additive contributions to binding eIF4A or eIF4E, we examined additional variants harboring combined substitutions of two different clusters. Whereas combining the contiguous substitutions 21–27 and 29–35 produced no further reduction in binding to GST-eIF4A compared to the single substitutions (Figure 2B, cols. 13 vs. 5 and 6), an additive reduction in binding was observed on combining the non-contiguous substitutions 21–27 and 51–57 (Figure 2B, cols. 14 vs. 5 and eight and Figure 2—figure supplement 2, cols. 5 vs. 1 and 2), which was comparable in its effect to deleting the entire (ΔNTD_2-92) on binding to GST-eIF4A (Figure 2B, cols. 2 and 14, dark green hues). In contrast, combining the non-contiguous substitutions that individually impaired eIF4E binding produced reductions in binding by the 59-65/83-89 variant indistinguishable from those generated by the single substitutions (Figure 2C, cols. 16 vs. 9 and 11 and Figure 2—figure supplement 2, cols. 12 vs. 9 and 10). It is noteworthy that the two double substitutions that impair binding to GST-eIF4A (21-27/29-35 and 21-27/51-57) had little or no effect on binding to GST-eIF4E (Figure 2B–C, cols. 13–14), and that the double substitution 59-65/83-89 strongly reduced binding to GST-eIF4E with no significant effect on binding to GST-eIF4A (Figure 2B–C, col. 16). These last findings support the notion that the binding determinants for eIF4A and eIF4E are segregated into non-overlapping adjacent segments of the NTD.

Combining substitution 51–57 that selectively impairs binding to GST-eIF4A with the adjacent substitution 59–65 that selectively impairs binding to GST-eIF4E produced a binding defect to GST-eIF4E indistinguishable from that given by 59–65 alone (Figure 2C, cols. 15 vs. 8 and 9), supporting our conclusion that the binding determinants for eIF4E do not extend upstream into the region that binds eIF4A. However, the 59–65 substitution appeared to suppress the eIF4A binding defect of the adjacent upstream substitution 51–57 in the 51-57/59-65 variant (Figure 2B, cols. 15 vs. 8 and 9). One way to explain this ‘context dependence’ of the 51–57 substitution would be to propose that the residues in segment 59–65 that mediate eIF4E binding also antagonize binding of eIF4A to the contiguous segment 51–57 segment, such that their removal lessens the requirement for the eIF4A binding determinants in the 51–57 segment. Consistent with this possibility, we found that the 59–65 substitution did not suppress the stronger binding defect conferred by combining the non-contiguous substitution 21–27 with 51–57 in a 21-27/51-57/59-65 triple mutant (Figure 2—figure supplement 3, col. eight vs. col. 6). Regardless of the explanation, the 21-27/51-57 and 59-65/83-89 double substitutions provide Ded1 variants that selectively impair binding to eIF4A or eIF4E for our subsequent in vivo analysis.

The Ded1 NTD is highly unstructured (Floor et al., 2016) and harbors RGG/RG motifs (Rajyaguru and Parker, 2012). To examine whether these motifs contribute to Ded1 interactions with eIF4A or eIF4E, we examined substitutions R27A, G28A, and R51A within the presumptive eIF4A-binding region identified above. Interestingly, R27A impaired binding to GST-eIF4A to the same extent observed on substituting the entire cluster 21–27 in which R27 resides, whereas G28A and R51A had no effect (Figure 2—figure supplement 4B). This finding is in accordance with the fact that R27 is highly conserved in the family Saccharomycetaceae, as well as in animals (Figure 2—figure supplement 1A–B), whereas, G28 and R51 are not. In contrast, despite its strong sequence conservation, Ala substitution of R62 within the region implicated in eIF4E binding had no significant effect on binding to GST-eIF4E (Figure 2—figure supplement 4B, right graph); although we note that the adjacent residue at position 61 is also an Arg in S. cerevisiae.

We further reasoned that because the Ded1 NTD is intrinsically unstructured, its hydrophobic and/or aromatic residues may be solvent-exposed and available for protein-protein interactions. Accordingly, we examined substitutions of five Phe or Tyr residues located within the NTD blocks of residues implicated above in binding to eIF4A or eIF4E. Strikingly, substituting each of the aromatic residues Y65 and W88 reduced binding to GST-eIF4E to the same extent observed for Ala substitutions of the entire corresponding segments 59–65 and 83–89, respectively (Figure 2—figure supplement 4C). By contrast, the Y21A and F56A/F57A substitutions within the eIF4A binding determinants 21–27 and 51/57, respectively, had no significant effect on binding to GST-eIF4A (Figure 2—figure supplement 4C). Together, these last findings reinforce the identification of segments 59–65 and 83–89 as binding determinants for eIF4E, and further suggest that aromatic residues Y65 and W88 within these segments might mediate key hydrophobic interactions with eIF4E. Although Y65 is not highly conserved, W88 is invariant among the eukaryotic species we examined (Figure 2—figure supplement 1A–B).

Having identified substitutions in the Ded1 NTD that selectively reduce Ded1 binding to GST-eIF4A or GST-eIF4E in vitro, we addressed next whether these substitutions reduce Ded1 function in vivo. We began by asking whether the Ded1 NTD substitutions confer synthetic growth defects when combined with the temperature-sensitive (Ts^-) ded1-952 mutation, which strongly impairs growth at 37°C (Sen et al., 2015) but confers only a moderate slow-growth (Slg^-) phenotype at 34°C (Figure 3B, rows 1–2) when introduced as a plasmid-borne Myc₁₃-tagged allele expressed from the native promoter (henceforth ded1-ts, Figure 3A) in a strain deleted for chromosomal DED1. Compared to ded1-ts alone, the ded1-ts alleles also containing the 21–27 or 51–57 mutations shown above to impair eIF4A binding in vitro conferred stronger Slg^- phenotypes at 34°C, with a slightly greater defect when the two mutations were combined within ded1-ts-21-27/51-57 (Figure 3B, 34°C, rows 6–9). Similar findings were made for the 59–65, 83–89, and 59-65/83-89 mutations that selectively impair eIF4E binding to Ded1 in vitro (Figure 3B, row 6 vs. 10–12), except that the reductions in growth were less pronounced than observed for the 21–27 and 21-27/51-57 mutations that impair eIF4A binding (Figure 3B, cf. rows 10–12 vs. 7 and 9). In contrast, the 14–20 and Δ40–47 mutations that did not affect Ded1 binding to eIF4A or eIF4E in vitro also conferred no Slg^- in combination with ded1-ts in vivo. (The apparent suppression of the ded1-ts phenotype by 14–20 was not reproduced in independent transformants.) Western analysis showed that neither the double substitution 21-27/51-57 nor the quadruple substitution 21-27/51-57,59-65/83-89 reduced the steady-state level of the ded1-ts product (Figure 3D, lanes 5–7), indicating that they impair Ded1 function and not its expression.

Figure 3 with 1 supplement see all

Download asset Open asset

The single amino acid substitutions Y21A and R27A in the eIF4A binding region each conferred a Slg^- phenotype at 34°C in combination with ded1-ts, which for R27A was greater than that observed for Y21A and comparable to that given by the 21–27 clustered substitution that encompasses both single-residue substitutions (Figure 3—figure supplement 1, rows 5–6 vs. 2–3). These phenotypes correlate with the stronger eIF4A binding defect given by R27A versus Y21A, and with the similar binding defects observed for R27A and the 21–27 variant in vitro (Figure 2—figure supplement 4B). Although Y21A had little effect on Ded1 binding to eIF4A in vitro (Figure 2—figure supplement 4B), the residue is highly conserved (Figure 2—figure supplement 1) and, hence, might have a greater impact in vivo. The F56A/F57A substitutions in the 51–57 interval, which had no effect on GST-eIF4A binding in vitro (Figure 2—figure supplement 4B) also had no effect on cell growth in combination with ded1-ts (Figure 3—figure supplement 1, rows 2 and 7). The Y65A and W88A single residue substitutions in the respective 59–65 and 83–89 segments of the eIF4E binding region each conferred moderate Slg^- phenotypes in combination with ded1-ts indistinguishable from those given by the corresponding 59–65 and 83–89 clustered substitutions (Figure 3—figure supplement 1, rows 11–12 vs. 9–10). These last findings are in accordance with the defects in GST-eIF4E binding in vitro shown above for these single and clustered substitutions (Figure 2—figure supplement 4C). Taken together, the effects of single-residue and clustered NTD substitutions on Ded1 binding to GST-eIF4A or GST-eIF4E in vitro are generally well correlated with their effects on cell growth in combination with the ded1-ts allele, with relatively stronger growth defects associated with mutations that reduce Ded1 binding to eIF4A versus those conferring comparable reductions in Ded1 binding to eIF4E.

We also examined the effects of the NTD mutations on Ded1 function in the absence of the ded1-ts mutation, finding that the 21-27/51-57 double cluster mutation affecting eIF4A binding confers a cold-sensitive Slg^- phenotype, which is exacerbated on combining it with the 59-65/83-89 double cluster mutations that impair eIF4E binding in vitro (Figure 3C, rows 1–4). Interestingly, the latter quadruple mutation conferred a strong Slg^- phenotype comparable to that given by the ded1-Δ2–90 deletion allele lacking nearly the entire NTD, both in otherwise WT DED1 (Figure 3C, rows 4–5) and in the ded1-ts allele, which suggests that binding to eIF4A and eIF4E constitutes the key in vivo function of the Ded1 NTD. Again, Western analysis revealed that the double and quadruple substitutions, as well as the Δ2–90 NTD deletion, had little effect on Ded1 steady-state expression levels (Figure 3D, lanes 1–4).

Finally, we asked whether disrupting binding of the Ded1 NTD to eIF4A or eIF4E would reveal an impact on cell growth of eliminating the Ded1 CTD and its known interaction with eIF4G. Consistent with previous findings (Hilliker et al., 2011), the CTD deletion that removes the C-terminal 43 residues of Ded1 (ded1-ΔC) confers no growth defect when the rest of Ded1 is intact. Importantly, however, the ΔC mutation exacerbates the cold-sensitive Slg^- phenotypes of all of the ded1 mutations examined containing single- or double cluster substitutions in the NTD that impair binding to either eIF4A or eIF4E. While the exacerbation by ΔC is subtle for the 51–57, 59–65, and 83–89 single-cluster mutations, it is pronounced for 21–27 and both double-cluster substitutions (Figure 4A, cf. adjacent rows). The exacerbation of the cold-sensitive growth phenotype of the NTD double-cluster mutations by ΔC occurred without reducing expression relative to the NTD variants with an intact CTD (Figure 4B, lanes 4–6). These findings are consistent with the idea that interaction of the Ded1 CTD with eIF4G is less critical for stabilizing the Ded1-eIF4E-eIF4A-eIF4G quaternary complex, activating Ded1 helicase function, or both, compared to the Ded1 NTD interactions with eIF4E and eIF4A; but that the importance of the CTD is increased when either the Ded1-eIF4E or Ded1-eIF4A interactions are impaired.

Figure 4

Download asset Open asset

Having identified eIF4A and eIF4E binding determinants in the Ded1 NTD in vitro and also demonstrating their contributions to Ded1 function in supporting cell growth, we sought next to demonstrate the importance of these interaction sites for Ded1 association with eIF4A or eIF4E in vivo using immunoprecipitation of myc-tagged Ded1 from yeast lysates. To this end, the myc₁₃-tagged ded1-ts alleles containing the 21-27/51-57 or 59-65/83-89 double-cluster mutations, ded1-ts containing no other mutations, or empty vector were introduced into a DED1 strain lacking the chromosomal genes encoding eIF4G1 and eIF4G2 and expressing HA-tagged eIF4G1 from a plasmid under the eIF4G2 promoter, and whole cell extracts (WCEs) were immunoprecipitated with anti-Myc antibodies. The eIF4A, eIF4E, and eIF4G1-HA all coimmunoprecipitated specifically with the myc₁₃-tagged ded1-ts product, failing to be immunoprecipitated from the extract lacking a myc-tagged protein (Figure 5, rows 5–6), as expected for Ded1 interaction with eIF4F. The presence of mutations 21-27/51-57 in the ded1-ts allele, which impair Ded1 binding to GST-eIF4A in vitro (Figure 2B) reduced coimmunoprecipitation of the myc₁₃-tagged product with native eIF4A, but not eIF4E or eIF4G1-HA from WCEs (Figure 5, rows 6–7). Moreover, mutations 59-65/83-89, which reduced Ded1 binding to GST-eIF4E in vitro (Figure 2C), specifically impaired coimmunoprecipition of eIF4E with the myc₁₃-tagged ded1-ts/59-65/83-89 product (Figure 5, lanes 6 and 8). The results were unaffected by treating the immune complexes with RNAses A and T1 (Figure 5, lanes 9–12 vs. 5–8), suggesting that the interactions are not bridged by RNA. Thus, the binding determinants in the Ded1 NTD for eIF4A and eIF4E defined by in vitro pull-down assays with recombinant GST fusions to eIF4A or eIF4E are also crucial for association of native eIF4A or eIF4E with the ded1-ts product expressed at native levels in yeast cells. Because WT Ded1 is present in these cells, the reductions in eIF4E/eIF4A coimmunoprecipitation with the ded1-ts variants harboring NTD substitutions might have been intensified by competition with WT Ded1 for binding to eIF4F.

Figure 5

Download asset Open asset

Our findings that the NTD substitutions selectively reduced association of the ded1-ts product with only eIF4E or eIF4A versus all three eIF4F subunits might indicate that Ded1 exists predominantly in binary complexes with each of the subunits of eIF4F; however, this would be odds with the expectation that the majority of Ded1 is associated with eIF4F in cells, based on estimates of the binding constant for Ded1 association with eIF4F and the cellular concentrations of these factors (Gao et al., 2016). Rather, it seems plausible that weakening the association of Ded1 with eIF4E leads to a specific reduction in eIF4E coimmunoprecipitation with ded1-ts owing to dissociation of eIF4E from the scaffold subunit eIF4G during the extensive washing steps involved in the experiment, notwithstanding the known stable interaction of eIF4E with eIF4G (Mitchell et al., 2010). The relatively lower affinity of yeast eIF4A for eIF4G (Mitchell et al., 2010; Park et al., 2012) makes this explanation even more likely for the selective loss of eIF4A association with mutated ded1-ts; although the existence of Ded1-eIF4A binary complexes in vivo has also been predicted (Gao et al., 2016).

We sought next to provide evidence that eliminating the binding determinants for eIF4A or eIF4E in the Ded1 NTD confer growth defects owing to loss of the specific interaction with eIF4A or eIF4E, respectively. We reasoned that if a Ded1 NTD substitution confers a growth defect, but can still partially interact with its binding partner, then overexpression of the partner should reinstate the interaction by mass action and rescue normal cell growth. If however the Ded1 mutant cannot bind to the partner at all, then overexpressing the latter might have little effect on cell growth (Figure 6A). To test this prediction, we overexpressed eIF4A1 from a high-copy (hc) TIF1 plasmid in the panel of ded1-ts mutants with single or double cluster substitutions in the eIF4A binding region and measured growth rates by cell-spotting assays. eIF4A overexpression mitigated the Slg^- phenotypes at 34°C of the 21–27, 29–35, and 51–57 single-cluster mutations, and also the 21-27/29-35 double cluster mutations (Figure 6B, rows 3–10, cf. adjacent rows), all of which conferred partial reductions in Ded1 binding to GST-eIF4A in vitro (Figure 2B), consistent with restored interaction of Ded1-eIF4A association by mass action. Importantly however, eIF4A overexpression did not mitigate the stronger Slg^- phenotype conferred by the ded1-ts,21-27/51-57 double-cluster mutation (Figure 6B, rows 11–12), which reduced binding to GST-eIF4A to the low level conferred by the ΔN_2-92 deletion that removes the entire eIF4A binding domain (Figure 2B). Nor did eIF4A overexpression mitigate the Slg^- phenotype conferred by the 59-65/83-89 mutations that selectively impair binding to eIF4E (Figure 6—figure supplement 1), supporting the interpretation that suppression of the other NTD mutations that partially impair eIF4A binding results from restoration of Ded1-eIF4A association by mass action (Figure 6A). Western analysis verified that eIF4A was overexpressed similarly in all of the ded1 mutants (Figure 6C). These findings support the idea that Ded1 substitutions in the eIF4A-binding region of the NTD confer growth defects in vivo owing to impaired association with eIF4A in cells.

Figure 6 with 1 supplement see all

Download asset Open asset

When we subjected the ded1-ts alleles containing single- or double cluster mutations in the eIF4E binding region to a similar analysis by overexpressing eIF4E from a hc CDC33 plasmid, we saw a uniform, modest exacerbation of growth defects for all strains (not shown), which may indicate that eIF4E overexpression is toxic for yeast. Instead, we exploited the fact that overexpressing WT DED1 can mitigate the growth defects conferred by the chromosomal eIF4E mutant allele cdc33-1 (de la Cruz et al., 1997). In agreement with this, we found that overexpressing either WT DED1 or the ded1-21-27/51-57 allele defective for eIF4A binding from a hc plasmid increased the growth rate of cdc33-1 cells (Figure 7A, rows 3–4 vs. 1. Importantly, however, the hc ded1-59-65/83-89 allele, defective for eIF4E binding in vitro, had no effect on growth of the mutant cells (Figure 7A, row 5 vs. 1). Western analysis verified that the mutant and WT ded1 alleles were overexpressed similarly in the cdc33-1 mutant (Figure 7B). These findings have two important implications. First, they provide strong evidence that the ded1-59-65/83-89 mutation impairs both a physical and functional interaction between Ded1 and eIF4E in cells. Second, they imply that overexpressing Ded1 partially suppresses the cdc33-1 mutation by enhancing interaction of the cdc33-1 product with Ded1 by mass action, presumably increasing the concentration of the eIF4F·Ded1 complex in which Ded1 functions most efficiently, rather than indirectly compensating for a reduction in eIF4F function (that should occur for all ded1 alleles).

Figure 7

Download asset Open asset

We next investigated changes in bulk translation initiation upon disruption of Ded1-eIF4A or Ded1-eIF4E interactions by Ded1 NTD mutations, examining first the effects on total polysome assembly. Cells were treated with cycloheximide just prior to harvesting to prevent polysome run-off during isolation, and polysomes were resolved from 80S monosomes, free 40S and 60S subunits, and free mRNPs by sedimentation through sucrose density gradients. Both the ded1-ts,21-27/51-57 and ded1-ts,59-65/83-89 mutations, impairing Ded1 interactions with eIF4A or eIF4E, respectively, reduced the ratio of polysomes to monosomes (P/M), indicating a reduction bulk translation initiation, which was not significantly greater for one mutation versus the other (Figure 8A).

Figure 8 with 5 supplements see all

Download asset Open asset

We further analyzed the functional defects in eIF4A- and eIF4E binding mutants by measuring expression of three luciferase (FLUC) reporter mRNAs harboring a Ded1-hypodependent synthetic unstructured 5’UTR harboring no stem-loop (no SL), or either cap-distal or cap-proximal SLs shown previously to confer Ded1-hyperdepence both in vivo (Sen et al., 2015) and in vitro (Gupta et al., 2018; Figure 8B). Relative to the ded1-ts parental mutant, introducing single- or double-cluster NTD mutations that impair binding to eIF4A or eIF4E, or control NTD mutations 14–20 and 40–47 with no effect on eIF4A/eIF4E binding, had little or no effect on expression of the reporter lacking a 5’UTR SL (Figure 8C, dark gray bars). The reporter with a cap-distal SL showed reduced expression relative to the no-SL construct in the ded1-ts parental strain (Figure 8—figure supplement 1, cols. 1–2), which was unaffected by the control NTD mutations; but was driven even lower by all of the mutations affecting eIF4A or eIF4E binding, with the greatest reductions seen for the two double substitutions ded1-ts,21-27/51-57 and ded1-ts,59-65/83-89 that eliminate, respectively, eIF4A or eIF4E binding to Ded1 (Figure 8C, light gray bars). Expression of the cap-proximal SL reporter was also reduced somewhat in the ded1-ts parental strain (Figure 8—figure supplement 1, col. 3 vs. 1), and was not further diminished by the control NTD mutations or the single cluster mutants defective for eIF4E binding (59–65 and 83–89), whereas the single cluster mutants defective for eIF4A binding (21–27 and 51–57) and both double-cluster mutants impairing eIF4A or eIF4E binding showed substantially reduced expression of this reporter, with the greatest reductions seen for the two double-cluster mutants (Figure 8C, white bars). Interestingly, overexpression of eIF4A conferred increased expression of the no SL and cap-distal reporters in the ded1-ts/21–27 mutant, predicted to exhibit partially impaired binding of eIF4A to Ded1, but not in the ded1-ts/21-27/51-57 or ded1-ts/59-65/83-89 strains expected to be fully defective for eIF4A binding, or for eIF4E binding, respectively (Figure 8—figure supplement 5). This supports the notion that the defect in reporter expression conferred by ded1-ts/21–27 involved impaired Ded1 association with eIF4A.

Finally, we verified that the decreased expression of the cap-proximal FLUC reporter in the Ded1 mutants results from impaired translation by examining the effects of the mutations on the polysome size distributions of the reporter mRNA, assayed by qRT-PCR of total mRNA isolated from each fraction of the density gradient used to resolve total ribosomal species, noting that a shift in mRNA abundance from larger to small polysomes, or from polysomes/monosomes to free mRNPs, indicates a reduction in the rate of translation initiation. In the parental ded1-ts mutant, the reporter mRNA is almost equally distributed among all gradient fractions (Figure 8D (ii), gray points). In the eIF4E-binding mutant ded1-ts,59-65/83-89, the proportion of reporter mRNA in the heavy polysome fractions is reduced, and the proportion in the smallest polysomes and monosomes is increased (Figure 8D (ii), orange points) relative to the mRNA distribution in the parental strain (Figure 8D (ii), grey points). Moreover, in the eIF4A binding mutant ded1-ts,21-27/51-57, the proportion of reporter mRNA in heavy polysomes is reduced and the proportion in free mRNP is elevated (Figure 8D (ii), green points) compared to that seen in the ded1-ts strain (Figure 8D (ii), grey points). The more extensive shift from polysomes to free mRNP observed for the eIF4A-binding mutant compared to the shift from larger to smaller polysomes/monosomes conferred by the eIF4E-binding mutant (Figure 8D (ii), green vs. orange) is consistent with the relatively greater reductions in cap-proximal SL reporter expression given by the set of three eIF4A-binding mutants versus the three eIF4E-binding mutants (Figure 8C, eIF4A(-) vs. eIF4E(-) mutants). This shift of the reporter mRNA from polysomes to free mRNP in the eIF4A-binding mutant or to smaller polysomes in the eIF4E-binding mutant is statistically significant (Figure 8—figure supplement 4). In contrast to the behavior of the reporter mRNA, the polysome distribution of native ACT1 mRNA was not substantially altered by either of the Ded1 NTD mutations (Figure 8D (i)), suggesting that Ded1’s interactions with eIF4A and eIF4E are relatively more important for the reporter mRNA harboring a stable SL structure compared to ACT1 mRNA. ACT1 mRNA was not found to be hyperdependent on Ded1 by ribosome profiling of a ded1 mutant (Sen et al., 2015). These results confirm that loss of eIF4A and eIF4E interactions by the Ded1 NTD lead to reduced translation initiation of a Ded1-hyperdependent reporter mRNA, as well as to decreased bulk translation initiation in vivo.

Assaying expression of the same LUC reporters, we found that the single amino acid mutations in the eIF4A binding region Y21A and R27A reduced expression of only the cap-distal SL (Y21A) or both SL reporters (R27A), and that the Y65A and W88A single-residue mutations in the eIF4E binding region reduced expression of both SL reporters, compared to the parental ded1-ts strain (Figure 8—figure supplement 1), thus supporting the importance of these individual residues in Ded1 NTD interactions with eIF4A and eIF4E in vivo.

We also provided evidence that eliminating both eIF4A and eIF4E binding simultaneously essentially inactivates the Ded1 NTD in promoting bulk translation as well as translation of Ded1-hyperdependent reporter mRNAs. As shown in Figure 8—figure supplement 2, the 21-27/51-57/59-65/83-89 quadruple-cluster mutation and the Δ2–90 deletion of the NTD introduced into otherwise WT DED1 conferred nearly indistinguishable reductions in bulk polysome assembly (panel A) and expression of both cap-proximal SL and cap-distal SL FLUC reporters (panel B). In both assays, the defects were quantitatively smaller than shown above for the double-cluster mutations in Figure 8A and 5C, which we attribute to the absence of the ded1-ts mutation in the parental and mutant alleles being compared in the current assays.

Finally, we obtained evidence that the effect of NTD mutations in exacerbating the effect of deleting the Ded1 CTD shown above in cell growth assays (Figure 4) could also be observed in the functional assays for polysome assembly and reporter mRNA expression. As shown in Figure 8—figure supplement 3A, the ded1-ΔC mutation analyzed above has little or no effect on polysome assembly on its own (cf. panels (i)-(ii)). However, ΔC clearly exacerbates the effects of both the 21-27/51-57 and 59-65/83-89 double-cluster mutations in the Ded1 NTD in reducing P/M ratios, with a relatively greater effect for the 21-27/51-57 mutation that eliminates eIF4A binding (cf. (iii)-(iv) and (v)-(vi)). Similarly, ded1-ΔC alone has only small effects on expression of the SL-containing FLUC reporters, but ΔC exacerbates the reductions in reporter expression conferred by 21-27/51-57 (Figure 8—figure supplement 3B, set four vs. sets 2 and 3) and by the 59-65/83-89 mutation (Figure 8—figure supplement 3B, set six vs. sets 2 and 5). These findings support our conclusion that eliminating eIF4G association with the Ded1 CTD imposes a relatively greater requirement for the Ded1 NTD interactions with eIF4A and eIF4E, either in forming or stabilizing the eIF4F·Ded1 complex or in stimulating Ded1 unwinding activity.

The yeast reconstituted system provides a unique mechanistic view of translation initiation (Acker et al., 2007; Algire et al., 2002; Gupta et al., 2018; Mitchell et al., 2010), which allowed us previously to reconstitute the function of Ded1 in stimulating the rate of 48S PIC assembly on native mRNAs, and to demonstrate substantially greater rate-enhancement for mRNAs that harbor structured 5’UTRs that confer hyperdependence on Ded1 for efficient translation in vivo, compared to mRNAs with less structured 5’UTRs that are Ded1-hypodependent in cells. Moreover, we showed that deleting the N-terminal 116 residues of Ded1 (ΔNTD_1-116) reduced the maximum rate of PIC assembly achieved at saturating levels of Ded1 (k_max) and increased the concentration of Ded1 required for the half-maximal rate (K_1/2) for several mRNAs containing stable SL structures in the 5’UTRs (Gupta et al., 2018). As eliminating the entire N-terminal region did not distinguish between impairing binding to eIF4A, eIF4E, or eliminating some other stimulatory function of this region of Ded1, we sought to determine whether selectively impairing Ded1 interaction with eIF4A or eIF4E would impair Ded1 acceleration of 48S assembly on a SL-containing mRNA. Accordingly, we purified full-length (FL) WT Ded1, Ded1-ΔNTD_1-116 and the two Ded1 variants harboring the double clustered substitutions that disrupt binding to eIF4A 21-27/51-57, or the single cluster substitution 59–65 that reduces binding to eIF4E, and compared them for acceleration of 48S PIC assembly on an mRNA (dubbed CP-8.1) containing the same cap-proximal SL present in the FLUC reporter mRNA analyzed above appended to the coding sequences and 3’UTR of of the Ded1-hypodependent native RPL41A mRNA. The three purified mutant proteins had ATPase activities similar to that of WT Ded1 (data not shown).

In agreement with previous results (Gupta et al., 2018), FL Ded1 substantially increased the k_max of 48S PIC formation on CP-8.1 mRNA by ~5 fold compared to the absence of Ded1, and deleting the NTD diminished the rate enhancement conferred by addition of Ded1 compared to no Ded1 in the reaction by ≈60% (Figure 9A, cols. 1–3; see Figure 9—figure supplement 1C for data summary). Interestingly, disrupting eIF4A binding to the NTD with the 21-27/51-57 double substitutions was comparable to the ΔNTD_1-116 in reducing k_max, whereas the 59–65 substitution that affects eIF4E binding had no effect on the k_max (Figure 9A, cols. 3–5). The ΔNTD_1-116 truncation also greatly increased the K_1/2 for Ded1 on this mRNA (Figure 9B, cols. 1–2), as observed previously (Gupta et al., 2018); and both the 21-27/51-57 and 59–65 substitutions also increased the K_1/2 for Ded1 to an extent ≈40% of that given by the ΔNTD_1-116 (Figure 9B, cols. 2–4). These findings suggest that binding of eIF4A, but not eIF4E, to the Ded1 NTD is required for maximum acceleration of 48S PIC assembly on CP-8.1 mRNA, but that both eIF4A and eIF4E interactions decrease the amount of Ded1 required to achieve this stimulation, possibly by enhancing formation of the eIF4F·Ded1 complex (Gupta et al., 2018).

Figure 9 with 1 supplement see all

Download asset Open asset

Figure 9—source data 1

CP-8.1 mRNA recruitment source data.

https://cdn.elifesciences.org/articles/58243/elife-58243-fig9-data1-v2.xlsx

Download elife-58243-fig9-data1-v2.xlsx

We showed previously that Ded1 also increases the rate of 48S PIC assembly on Ded1-hypodependent RPL41A mRNA, but to a lesser degree and at a much lower Ded1 concentration compared to Ded1-hyperdependent mRNAs such as CP-8.1 (analyzed above). As observed previously (Gupta et al., 2018), FL Ded1 increased the k_max for RPL41A mRNA by ~3 fold, and eliminating the Ded1 NTD by ΔNTD_1-116 did not impair this stimulation (Figure 9—figure supplement 1A, cols. 2–3); however, ΔNTD_1-116 increased the K_1/2 for Ded1 on RPL41A mRNA by nearly 70-fold (Figure 9—figure supplement 1B, cols. 2–3). As expected, neither the 21-27/51-57 double substitution nor the 59–65 substitution altered the increase in k_max for RPL41A mRNA conferred by FL Ded1 (Figure 9—figure supplement 1A, cols. 4–5 vs. 2). While both substitutions increased the K_1/2 for Ded1 compared to FL-Ded1 (~7 fold for 21-27/51-57 and ~4 fold for the 59–65 substitution), these increases were much smaller than the ~70 fold increase produced by ΔNTD_1-116 (Figure 9—figure supplement 1B, cols. 3–4 vs. 2 cf. col.1;and Figure 9—figure supplement 1D, cols. 4–5 vs. 2 cf. col. 1). These last results suggest that strong interaction of either eIF4A or eIF4E with the Ded1-NTD is sufficient to greatly reduce the amount of Ded1 required for maximum acceleration of 48S PIC assembly on RPL41A mRNA. This contrasts with our findings on CP-8.1 mRNA, where eliminating either of the interactions of the Ded1-NTD with eIF4A and eIF4E conferred an increase in the Ded1 K_1/2 only slightly smaller than that given by deleting the entire NTD. Thus, it appears that both interactions are needed simultaneously to enhance the function of the eIF4F-Ded1 complex on the more structured CP-8.1 mRNA, whereas each interaction can contribute independently and additively on the less structured RPL41A mRNA.

All plasmids employed in this study are listed in Table 1. To create plasmids for bacterial expression of GST fusions to eIF4A1 and eIF4E, TIF1 and CDC33 coding sequences were amplified by PCR from the genomic DNA of WT yeast strain BY4741 (using primers listed in Table 2) and inserted between the BamHI and XhoI sites of pGEX4T1. pET-C-Ded1 cut with NdeI and XhoI was used to construct pSG49-pSG51 by inserting fragments encoding Ded1 residues 93–604 (pSG49), residues 1–561 (pSG50), or residues 93–561 (pSG51), PCR-amplified from pET-C-Ded1. Plasmids pSG52-pSG75 were derived from pET-C-Ded1 using the Agilent QuikChange Lightning Mutagenesis kit, according to the vendor’s instructions, as were pSG76-pSG81 by mutagenesis of pET-N-Ded1. pSG1was constructed by inserting a fragment containing the DED1 promoter (beginning 331 nt upstream of the AUG), entire coding sequence, coding sequence for the myc₁₃ epitope, and the ADH1 terminator, PCR-amplified from the genomic DNA of yeast strain H3666, between into the SphI and SacI sites of LEU2/CEN4/ARS1 vector YCplac111. pSG2, containing ded1-ts-myc was derived from pSG1 by mutagenesis with Agilent QuikChange Lightning Mutagenesis kit to introduce the T408I/W253R substitutions. pSG27-pSG43 were derived from pSG1; and pSG3-pSG26 were derived from pSG2, by QuikChange mutagenesis. pSG44-pSG46 were similarly derived from p4504/YEplac195-DED1.

All yeast strains employed in this study are listed in Table 3. Novel strains SGY1-SGY43 harboring plasmid pSG1-pSG43, were constructed by plasmid shuffling using 5-fluoroorotic acid (5-FOA) (Boeke et al., 1987) from strain yRP2799, as follows. Transformants of yRP2799 containing the relevant LEU2 pSGY plasmid were selected on SC-Leu-Ura, patched on medium of the same composition, and replica-plated to SC-Leu containing 1 mg/ml 5-FOA. After 48–72 hr at 30°C, Ura^- segregants were purified from patches able to grow on 5-FOA medium by streaking for single colonies on SC-Leu.

Yeast strains were cultured in SC-Leu to A₆₀₀ of ~1.0, dilutions of 10⁻¹, 10⁻², 10⁻³, and 10⁻⁴ were prepared in sterile water, and 5 μL aliquots of undiluted and diluted yeast cells were spotted on SC-Leu medium and incubated at the temperatures indicated in the figure legends.

Agilent BL21-CodonPlus (DE3)-RIPL competent E. coli were transformed with pGEX4T1 or its derivatives containing the TIF4631, TIF1 or CDC33 coding sequences. Transformants were grown in 500 ml LB+Amp+Cam medium at 37°C to A₆₀₀ of ~0.5 and protein expression induced by adding IPTG to 1 mM and incubating at 18°C overnight with agitation. Harvested cells were washed with 50 ml phosphate-buffered saline (PBS) and lysed by sonication in the presence of 50 μg/mL lysozyme and 10% glycerol. Lysates were treated with 10% Triton X-100 at 4°C and cleared by centrifugation at 10,000xg. 500 μL of GE Healthcare Glutathione Sepharose 4B beads were added per 20 mL of lysate and incubated with rotation overnight at 4°C. The beads were washed twice with 5 ml PBS in 10% glycerol and treated with 5000 gel units of NEB Micrococcal Nuclease in the presence of 2 mM calcium chloride in 1 ml PBS. The reaction was stopped by addition of EDTA to 5 mM, followed by washing the beads with 1 ml of 500 mM sodium chloride in PBS/glycerol. The beads were washed two more times with 1 ml PBS in 10% glycerol and the washed beads were stored at −80°C. SDS-polyacrylamide gel electrophoresis (SDS-PAGE), followed by Coomassie Blue staining, was employed to estimate the amounts of GST-tagged proteins bound to the glutathione-agarose beads by comparison to known amounts of bovine serum albumin.

Promega TNT Quick Coupled Transcription/Translation System was employed for in vitro translation of Ded1 proteins for GST pull-down assays. Perkin Elmer EasyTag L-[³⁵S]-Methionine was used for radioactive labeling following the vendor’s instructions.

GST pull down reactions were conducted as follows. Aliqouts of 10 μL of GST or GST-tagged bait proteins bound to glutathione-agarose beads were combined with 8 μL of pull-down buffer (40 mM Tris, pH 8.0, 0.01% IGEPAL CA630, 2 mM DTT, 50 mM NaCl, 20% glycerol, and 5 mM MgCl₂), 0.5 μL of 5 mg/ml bovine serum albumin, 0.5 μL of ThermoFisher Scientific RNase A/T1 mix and 1 μL of [³⁵S]-labeled Ded1 protein. The reactions were incubated at 4°C with end-over-end rotation overnight. The beads were washed three times with 30 μL of pull-down buffer without glycerol and proteins eluted with 20 μL 15 mM Glutathione [pH 8.0]. The input proteins and eluates were resolved by SDS-PAGE and visualized by fluorography, as follows. Gels were fixed in 50% methanol and 15% acetic acid, stained with Coomassie Blue, treated with Amersham Amplify Fluorographic Reagent, dried under vacuum and exposed to X-ray film for 18–72 hr at −80°C. After developing, the films and gels were scanned, and band intensities analyzed by NIH ImageJ software. Pull down efficiency was calculated as intensity of pulled down band divided by the corresponding input band intensity and converted to percentages. For statistical analyses, three or four replicates were performed. Unpaired student’s t test was employed to determine p-values. GraphPad software was used to graph the data as dot plots.

For coimmunoprecipitation analysis, transformants of H3446 containing the relevant plasmid-borne DED1-myc alleles or empty vector were cultured in 100 mL of SC-Leu-Trp to A₆₀₀ of ~0.8, harvested by centrifugation, washed with coimmunoprecipitation (CoIP) buffer (100 mM Tris-HCl pH 7.5, 100 mM sodium chloride, 0.1% NP-40, 1 mM DTT, 1 mM PMSF, 1 × Complete Protease Inhibitor Mix tablets without EDTA [Roche]) and lysed by vortexing with glass beads. Lysates were cleared by centrifugation and total protein amount determined colorimetrically using BioRad Bradford Reagent. The WCEs extracts were diluted to 5 mg/mL in CoIP buffer and aliquots of 1 mg (200 μL) were immunoprecipitated with 5 μg Roche anti-myc antibody (clone 9E10). After 2 hr of rotation at 4°C, 50 μL ThermoFisher Scientific Dynabeads Protein G were added and rotation continued for 2 hr more. Using a magnetic rack, the beads were washed three times with 150 μL CoIP buffer. The beads were resuspended in 100 μL 2x Laemmli buffer and boiled for 5 min. SDS-PAGE and Western blot were employed to detect the immunoprecipitated proteins, using the following antibodies: mouse monoclonal anti-HA antibody (12C5; Roche Cat# 11 666 606 001), mouse monoclonal anti-c-Myc antibody (9E10; Roche Cat# 11 667 203 001), rabbit polyclonal anti-eIF4A/Tif1 (kindly provided by Patrick Linder), rabbit polyclonal anti-eIF4E/Cdc33 (kindly provided by J.E.G. McCarthy), and rabbit polyclonal anti-Ded1 antibody (kindly provided by Dr. Tien-Hsien Chang). The immune complexes were visualized by enhanced chemiluminescence.

For Western blot analysis of yeast WCEs, trichloroacetic acid (TCA) precipitation was employed to extract total protein (Reid and Schatz, 1982), extracts were resolved by. SDS-PAGE, and Western blots were probed using the antibodies listed above, as well as rabbit polyclonal anti-Hcr1 antibodies (Valásek et al., 2001).

Two-ml cultures of yeast transformants harboring the relevant reporter plasmids were harvested by centrifugation and the cell pellets were washed with 500 μL PBS and lysed by vortexing with glass beads in 200 μL PBS supplemented with 1 × Complete Protease Inhibitor Mix tablets without EDTA [Roche]. Lysates were cleared by centrifugation, diluted 1:100 in PBS supplemented with 1 × Complete Protease Inhibitor Mix tablets without EDTA and luciferase was assayed using Promega Luciferase Assay System and a luminometer. Total protein concentration was measured colorimetrically using Bradford Reagent (BioRad) and known amounts of bovine serum albumin as standards. Luciferase units were normalized by the total protein amounts.

For analyzing polysomes profiles, 200 mL of cells were cultured in the appropriate medium at 30°C and shifted to 36°C for 3 hr or to 15°C for 1 hr, and harvested at A₆₀₀ ≈ 1. Cells were treated with 50 μg/ml cycloheximide for 5 min prior to harvesting. WCEs were prepared by resuspending the cell pellet with an equal volume of breaking buffer (20 mM Tris-HCl, pH 7.5, 50 mM KCl, 10 mM MgCl₂, 1 mM dithiothreitol [DTT], 5 mM NaF, 1 mM phenylmethylsulfonyl fluoride [PMSF], 1 × Complete Protease Inhibitor Mix tablets without EDTA [Roche], 1 U/μl SUPERase-In RNase inhibitor) and vortexing with glass beads, followed by two cycles of centrifugation for 10 min at 15,000 rpm at 4°C. 20 A₂₆₀ units of cleared lysate were resolved on 10–50% (w/w) sucrose gradients by centrifugation at 39,000 rpm for 160 min at 4°C in a Beckman SW41Ti rotor. Gradients were fractionated with a Brandel Fractionation System with continuous monitoring at 254 nm with an ISCO UV detector. Fractions (0.7 mL) were precipitated overnight at −20°C by adding 1.5 volumes of RNA precipitation mix (95% ethanol, 5% sodium acetate pH 5.2), and centrifuged at 15,000 rpm for 30 min. 5 ng of ‘spike-in RNA’ was added to each fraction to control for differences in RNA recovery: a mixture of in vitro transcribed capped Bacillus subtilis mRNAs (Arava et al., 2003): 80 pg/µL lysA (prepared from plasmid pGIBS-LYS; ATCC#87482), 160 pg/µl trpCDEF (prepared from plasmid pGIBS-TRP; ATCC#87485), and 320 pg/µl of pheB (prepared from plasmid pGIBS-PHE; ATCC #87483). This mixture of in vitro transcribed mRNAs was a kind gift from Dr. Neelam Sen (Sen et al., 2019). Pellets were washed in 1 mL cold 80% ethanol, dried in a speed vacuum, resuspended in 100 µL TE buffer (10 mM Tris-HCl pH 8.0, 1 mM EDTA) and treated with 10 Kunitz units of DNase I. Qiagen RNEasy Mini kit was used to extract RNA and resuspended in 25 μL RNase-free water.

For RT-qPCR analysis of mRNA abundance in each gradient fraction, reverse transcription was carried out using SuperScript III First-Strand cDNA Synthesis SuperMix (Invitrogen) and 2 μL RNA purified from each gradient fractions. qPCR was carried out using Brilliant III Ultra-Fast SYBR Green Master Mix (Agilent) in a Mx3000P System (Stratagene) and the oligonucleotide pairs listed in Table 2. Normalization and analyses were performed as described in Chiu et al., 2010 and Sen et al., 2015.

T-Coffee (www.ebi.ac.uk/Tools/msa/tcoffee) was used to align Ded1 amino acid sequences from fungal and animal species. Saccharomyces Genome Database (https://www.yeastgenome.org/) was the source of Ded1 sequences from the genus Saccharomyces. Other sequences were retrieved from NCBI Protein (https://www.ncbi.nlm.nih.gov/protein) and UniProt (https://www.uniprot.org/). The results were reformatted with MView (www.ebi.ac.uk/Tools/msa/mview) and WebLogo (weblogo.berkeley.edu).

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Roy Parker, Angie Hilliker, Alan Sachs, Patrick Linder, and Michael Altmann for gifts of yeast strains or plasmids, Patrick Linder, JEG McCarthy, and Tien-Hsien Chang for gifts of antibodies, and Neelam Sen for spike-in RNAs. We thank Neelam Sen, Fan Zhang, Swati Gaikwad, other members of our laboratories, Nick Guydosh, and Tom Dever for helpful comments and suggestions. This work was supported in part by the Intramural Research Program of the National Institutes of Health.

1. Received: April 24, 2020
2. Accepted: May 28, 2020
3. Accepted Manuscript published: May 29, 2020 (version 1)
4. Version of Record published: July 8, 2020 (version 2)

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.