Yeast Ded1 promotes 48S translation pre-initiation complex assembly in an mRNA-specific and eIF4F-dependent manner

DEAD-box RNA helicase Ded1 is thought to resolve secondary structures in mRNA 5'-untranslated regions (5'-UTRs) that impede 48S preinitiation complex (PIC) formation at the initiation codon. We reconstituted Ded1 acceleration of 48S PIC assembly on native mRNAs in a pure system, and recapitulated increased Ded1-dependence of mRNAs that are Ded1-hyperdependent in vivo. Stem-loop (SL) structures in 5'-UTRs of native and synthetic mRNAs increased the Ded1 requirement to overcome their intrinsically low rates of 48S PIC recruitment. Ded1 acceleration of 48S assembly was greater in the presence of eIF4F, and domains mediating one or more Ded1 interactions with eIF4G or helicase eIF4A were required for efficient recruitment of all mRNAs; however, the relative importance of particular Ded1 and eIF4G domains were distinct for each mRNA. Our results account for the Ded1 hyper-dependence of mRNAs with structure-prone 5'-UTRs, and implicate an eIF4E·eIF4G·eIF4A·Ded1 complex in accelerating 48S PIC assembly on native mRNAs.


Introduction
In canonical translation initiation in eukaryotes, a ternary complex (TC), consisting of eukaryotic initiation factor 2 (eIF2), Met-tRNA i Met , and GTP, along with eIF1, eIF1A, eIF5, eIF4B, and eIF3, binds to the small (40S) ribosomal subunit to form a 43S pre-initiation complex (PIC). The 43S PIC binds to the 5'-end of mRNA and scans the 5'-untranslated region (UTR) to identify the start codon, resulting in the formation of the 48S PIC. eIF4F complex, comprised of eIF4E (a cap binding protein), eIF4G (a scaffolding protein), and eIF4A (a DEAD-box RNA helicase), interacts with the mRNA m 7 G cap and aids in recruitment of the 43S PIC to the 5'-end of the mRNA (reviewed in Dever et al., 2016;Hinnebusch, 2014). eIF4A promotes 48S PIC formation in vitro and translation in vivo of virtually all mRNAs regardless of their structural complexity (Pestova and Kolupaeva, 2002;Sen et al., 2015;Yourik et al., 2017). Yeast ribosome profiling studies show that the majority of cellular mRNAs have strong and similar dependence on eIF4A for their proper translation in cells (Sen et al., 2015). Additionally, eIF4A is an essential protein, and small decreases in eIF4A cellular concentrations reduce bulk translation in vivo, further emphasizing the critical role of eIF4A in translation of most mRNAs (Firczuk et al., 2013). Yeast eIF4A is a weak helicase when unwinding RNA duplexes in vitro (Rajagopal et al., 2012;Rogers et al., 1999). Recent evidence suggests that mammalian eIF4A modulates the structure of the 40S subunit to enhance PIC attachment (Sokabe and Fraser, 2017).
Ded1 is a yeast DEAD-box RNA helicase that promotes translation in vivo of reporter mRNAs with longer or structured 5'-UTRs (Berthelot et al., 2004;Chiu et al., 2010;Sen et al., 2015). Like eIF4A, Ded1 is essential for yeast growth and stimulates bulk translation initiation in vivo (Chuang et al., 1997;de la Cruz et al., 1997). However, ribosome profiling of conditional ded1 mutants revealed that native mRNAs with 5'-UTRs that are longer and more structured than average yeast 5'-UTRs exhibit a greater than average reduction in translational efficiency (TE) relative to all other mRNAs on Ded1 inactivation (Ded1-hyperdependent mRNAs); whereas mRNAs with shorter and less structured 5'-UTRs exhibit increased relative TEs in ded1 cells (Ded1-hypodependent mRNAs) (Sen et al., 2015).
Yeast Ded1 can unwind model RNA duplexes and act as an RNA chaperone or RNA-protein complex remodeler in vitro (Bowers et al., 2006;Iost et al., 1999;Yang and Jankowsky, 2006). Translation stimulation by Ded1 requires its ATPase activity (Hilliker et al., 2011;Iost et al., 1999). Mutations in the Ded1 ATPase domain (Ded1-E307A and Ded1-R489A) impair ATP binding and hydrolysis, and these mutants have dominant-negative effects on translation both in cell extracts and in vivo, evoking stress granule formation in the latter (Hilliker et al., 2011).
Ded1 can physically interact individually with purified eIF4E, eIF4G, eIF4A, or Pab1 (poly(A) binding protein), and can bind simultaneously to eIF4AÁeIF4EÁeIF4G (eIF4F) or eIF4EÁeIF4G, in an RNAindependent manner (Gao et al., 2016;Hilliker et al., 2011;Senissar et al., 2014). Ded1 also interacts with the eIF4F complex in yeast extracts, supporting the physiological relevance of these interactions (Hilliker et al., 2011;Senissar et al., 2014). According to a proposed model, Ded1-eIF4G-mRNA interaction is thought to repress translation by promoting accumulation of mRNPs in stress granules, whereas ATP hydrolysis by Ded1 moves the repressed mRNPs back into the translation cycle (Hilliker et al., 2011). However, the role of Ded1-eIF4F interactions in the stimulatory function of Ded1 in translation initiation remains to be elucidated. Ded1 interacts with eIF4A through its N-terminal domain (Ded1-NTD) and this interaction is required for eIF4A stimulation of Ded1's RNAduplex unwinding activity (Gao et al., 2016). Deletion of the Ded1-NTD confers a cold-sensitive growth phenotype in cells, consistent with a role for eIF4A stimulation of Ded1 function in vivo (Banroques et al., 2011;Gao et al., 2016). Yeast eIF4G contains three RNA binding domains, N-terminal RNA1, central RNA2, and C-terminal RNA3 (Berset et al., 2003); and while none of the three is essential, simultaneous deletion of RNA2 and RNA3 is lethal (Park et al., 2011). In vitro, eIF4G variants lacking any of the three RNA binding domains exhibit similar affinities for eIF4A, support similar rates of ATP-hydrolysis by the eIF4F complex (albeit with higher K m values for ATP), but lack the preference of WT eIF4F for 5'-overhang substrates during unwinding (Rajagopal et al., 2012). The Ded1 C-terminal domain (Ded1-CTD) interacts with the eIF4G-RNA3 domain, and Ded1-eIF4G interaction decreases the rate of RNA unwinding while increasing Ded1 affinity for RNA in vitro (Hilliker et al., 2011;Putnam et al., 2015). A Ded1 variant lacking the CTD conferred reduced reporter mRNA translation compared to WT Ded1 in cell extracts, supporting a stimulatory role for the Ded1-CTD interaction with eIF4G in translation initiation (Hilliker et al., 2011). Ded1 also interacts with the RNA2 domain of eIF4G and with eIF4E (Senissar et al., 2014), but the physiological relevance of these interactions is unknown.
We previously demonstrated functions of eIF4F, eIF4B, and eIF3 in stimulating the rate and extent of mRNA recruitment by 43S PICs in a fully purified yeast initiation system for native RPL41A mRNA, containing a short and relatively unstructured 5'-UTR (Mitchell et al., 2010). Although Ded1 is essential in vivo, it was dispensable for recruitment of this mRNA in vitro. Considering that RPL41A was judged to be Ded1-hypodependent in vivo by ribosome profiling of ded1 mutants (Sen et al., 2015), we asked whether recruitment of Ded1-hyperdependent mRNAs would require Ded1 in the reconstituted system. We investigated whether the presence of defined stem-loop (SL) structures in native or synthetic mRNAs would confer greater Ded1-dependence for rapid recruitment in vitro. Finally, we examined the role of the RNA2 and RNA3 domains of eIF4G and the NTD and CTD of Ded1 that mediate Ded1 interactions with the eIF4F complex in promoting Ded1's ability to accelerate mRNA recruitment by 43S PICs. Our findings demonstrate that Ded1 accelerates recruitment of native and synthetic mRNAs, overcoming the inhibitory effects of structured leader sequences and conferring relatively greater stimulation for mRNAs hyperdependent on Ded1 in vivo, all in a manner consistent with stimulation of Ded1 function by formation of a Ded1-eIF4F complex.

Ded1 enhances the rate of recruitment of all natural mRNAs tested
We set out to reconstitute the function of Ded1 in 48S PIC assembly in a yeast translation initiation system comprised of purified components (Mitchell et al., 2010;Walker et al., 2013;Yourik et al., 2017). Pre-assembled 43S PICs, containing 40S subunits and factors eIF1, eIF1A, eIF5, eIF2ÁGDPNPÁMet-tRNA i , eIF4GÁ4E, eIF4A, eIF4B, and eIF3, were pre-incubated with or without Ded1, and reactions were initiated by addition of ATP and 32 P-labeled m 7 Gppp-capped mRNA (synthesized in vitro). Formation of 48S complexes was monitored over time using a native gel electrophoretic mobility shift assay (EMSA) to resolve free and 48S-bound mRNAs. An~20 fold excess of unlabeled-capped mRNA was added to reaction aliquots at each time point to quench further recruitment of 32 P-labeled mRNA ('pulse-quench'). By varying the concentration of Ded1, this assay yields the apparent rate constants (k app ) for 48S PIC formation at each Ded1 concentration, the maximal rate at saturating Ded1 (k max ), the Ded1 concentration required for the half-maximal rate of 48S formation (K 1/2 ), and the reaction endpoints (percentage of mRNA recruited) at each Ded1 concentration. The addition of~20 fold excess non-radiolabeled mRNA in the quench was adequate to prevent further recruitment of 32 P-labeled mRNA, and did not dissociate the pre-formed 32 P-labeled 48S complexes on the timescale of the recruitment experiments (  (Iost et al., 1999;Senissar et al., 2014). Moreover, the Ded1 bound a fluorescently labeled single-stranded mRNA in the presence or absence of ADP or ADPNP with K D values consistent with published values (Figure 1-figure supplement 1F) (Banroques et al., 2008;Iost et al., 1999).
We began by analyzing the effect of Ded1 on recruitment of RPL41A mRNA, a short transcript of 310 nucleotides (nt), with 5'-UTR of only 24 nt ( Figure 1A and Figure 1-figure supplement 1G), and a low degree of predicted secondary structure (Mitchell et al., 2010). RPL41A behaved like a Ded1-hypodependent mRNA in ribosome profiling experiments, exhibiting increased relative TE in ded1 mutant versus WT cells (Sen et al., 2015). Consistent with this, recruitment of RPL41A mRNA in vitro was achieved previously without Ded1 (Mitchell et al., 2010). Nevertheless, with our more sensitive pulse-quench approach, we found that addition of saturating Ded1 increased the k max of RPL41A by~2.8 fold from 0.95 ± 0.1 min -1 to 2.7 ± 0.3 min -1 ( Figure 1B 1=2 of 58 ± 8 nM ( Figure 1C). Thus, despite its low Ded1-dependence relative to most other mRNAs in vivo, RPL41A recruitment is appreciably stimulated by Ded1 in vitro.
We next examined Ded1 stimulation of another Ded1-hypodependent mRNA, HOR7, and several Ded1-hyperdependent mRNAs, SFT2, PMA1, OST3, and FET3, which exhibited reduced TEs in ded1 versus WT cells (Sen et al., 2015). SFT2 and PMA1 mRNAs are noteworthy in containing SL structures in their 5'-UTRs detected in vivo (Rouskin et al., 2014). Because these mRNAs exceed the maximum length that can be resolved by EMSA (~400 nt), we generated reporter mRNAs with the 5'-UTR and first 60 nt of coding sequences (CDS) of each mRNA ( Figure 1A and Figure 1-figure supplement 1G). (For brevity, we refer to these reporter mRNAs simply by their gene names.) Without Ded1, the rate of HOR7 recruitment was 1.3 ± 0.16 min -1 , and saturating Ded1 stimulated recruitment by~2 fold (k max = 2.8 ± 0.4 min -1 ) with a K Ded1 1=2 of 115.5 ± 17 nM, results that were similar to those for Ded1-hypodependent RPL41A (Figure 1B-C and Figure 1-figure supplement 1Hblue curve). Strikingly, without Ded1, the four Ded1-hyperdependent mRNAs were recruited at rates 2.5-to 15-fold lower compared to the two Ded1-hypodependent mRNAs ( Figure 1B, blue), and these rates increased by an order of magnitude on addition of Ded1 ( Figure 1B, orange and bottom panel; and Figure 1-figure supplement 1H). The Ded1-hyperdependent mRNAs also required higher Ded1 concentrations to achieve these maximal rates ( Figure 1C). The acceleration of mRNA recruitment by Ded1 required its ATPase activity, as ATPase-deficient Ded1 variant Ded1 E307A (    PMA1 and OST3 also exhibited low endpoints of recruitment without Ded1, 30 ± 5% and 41 ± 4%, respectively, which increased to 91 ± 2% and 86 ± 3%, respectively, on Ded1 addition ( Figure 1D). It was suggested that Ded1 acts as an RNA chaperone to aid transitions between different RNA conformations (Yang et al., 2007). In fact, two or more conformers of OST3 (and other mRNAs) were observed in native gel electrophoresis that likely represent differently folded, stable mRNA conformers (Figure 1-figure supplement 1K). Perhaps only one of these conformers of PMA1 and OST3 is competent for 48S PIC assembly, and Ded1 facilitates isomerization among them.
Interestingly, a linear relationship was observed between the fold-acceleration by Ded1 and the K Ded1 1=2 , such that Ded1-hypodependent and Ded1-hyperdependent mRNAs cluster separately from each other along the line ( Figure 1E). One explanation could be that Ded1-hyperdependent mRNAs have a higher rate-limiting activation energy barrier for 48S PIC assembly in the absence of Ded1 compared to the Ded1-hypodependent mRNAs, consistent with the latter's relatively higher rates of recruitment without Ded1 ( Figure 1B, blue bars, RPL41A and HOR7 versus SFT2, PMA1, OST3, and FET3). Accordingly, the hyperdependent mRNAs require relatively higher Ded1 concentrations to lower this barrier to the point at which a Ded1-independent step becomes rate limiting (SFT2 and PMA1) or where Ded1 cannot lower the Ded1-dependent barrier further (OST3 and FET3) ( Taken together, the data presented above demonstrate that mRNAs with long and structured 5'-UTRs, found to be hyperdependent on Ded1 for translation in vivo, are inherently less capable of PIC recruitment and more dependent on Ded1 for rapid recruitment in vitro than are mRNAs hypodependent on Ded1 in vivo.

Secondary structures in the 5'-UTR increase Ded1-dependence in 48S PIC assembly
Given that mRNAs with heightened Ded1-dependence in vivo have a greater than average potential to adopt secondary structures involving the 5'-UTR (Sen et al., 2015), we investigated if the stable SL structures previously detected in the 5'-UTRs of SFT2 and PMA1 were responsible for their elevated Ded1 dependence (Rouskin et al., 2014). To this end, we introduced mutations to eliminate the SL in each mRNA, or (for SFT2) to strengthen the SL (Figure 2A and Figure 2-figure supplement 1A-B). Without Ded1, the SL-disrupted version of PMA1, PMA1-M, showed~2 fold higher endpoints of recruitment than WT PMA1 (PMA1-M = 62 ± 2.5%, PMA1 = 30 ± 5%, Figure 2B, gold). PMA1-M was also recruited at rates~4 fold higher than PMA1 (k max = 1.2 ± 0.1 min -1 (PMA1-M) vs. 0.33 ± 0.06 min -1 (PMA1); Figure 2C, blue), consistent with the idea that the SL inhibits mRNA recruitment. Ded1 increased the rate of PMA1-M recruitment to yield a k max similar to that for PMA1 ( Figure 2C, orange), but at a much lower Ded1 concentration for PMA1-M ðK Ded1 1=2 = 40 ± 6 nM) versus PMA1 (~280 nM) ( Figure 2D and  These results strongly suggest that Ded1 accelerates the recruitment of WT SFT2 and PMA1 mRNAs, and increases the endpoint of PMA1 recruitment, in part by melting the SL structures in their 5'-UTRs. As the PMA1 SL is~120 nt from the 5'cap, it is likely that Ded1 resolves the SL to accelerate scanning of the PIC through the 5'-UTR.

Evidence that Ded1 stimulates the PIC attachment and scanning steps of initiation
Although Ded1-hypodependent mRNAs RPL41A and HOR7 lack any strong, defined SLs, Ded1 still stimulated their recruitment ( Figure 1B). Similarly, even after removal of SLs, the PMA1-M and SFT2-M mutant mRNAs were still stimulated by Ded1 ( Figure 2C). In addition to defined, stable SLs in 5'-UTRs, natural mRNAs likely form large ensembles of weaker structures involving interactions of nucleotides within the 5'-UTR or between 5'-UTR nucleotides and nucleotides in the CDS or 3'-UTR (Kertesz et al., 2010;Yourik et al., 2017), which might also contribute to Ded1-dependence during PIC attachment or scanning. To test this hypothesis, we examined recruitment of chimeric mRNAs with synthetic 5'-UTRs attached to the CDS and 3'-UTR of native RPL41A . A synthetic mRNA dubbed '-SL' (for 'minus stem-loop') contained a 67 nt 5'-UTR comprised of CAA repeats that is devoid of stable secondary structure ( Figure 2A and  Figure 1C. (E) Plot of K Ded1 1=2 from (D) versus fold-change in k þDed1 max =k ÀDed1 max from (C, lower) for the indicated mRNAs. Gray, blue, and green points indicate natural, mutated, and synthetic mRNAs, respectively, with error bars indicating 1 SD; line produced by linear regression analysis (Y = 25.29*X + 15.28, R 2 = 0.89, p-value<0.001). (F) Apparent rate of mRNA recruitment of synthetic mRNAs with different strengths SLs in the cap-distal region in the absence (blue) and presence of Ded1 (orange). DG˚: CD-10.5 = À10.5 kcal/mol, CD-8.1 = À8.1 kcal/mol, and CD-3.7 = À3.7 kcal/mol. Note that Y-axis is discontinuous. (B-D, F) Bars indicate mean values calculated from the three independent experiments represented by the data points. (B-D) WT PMA1 amd WT SFT2 data is added from Figure 1  green, -SL). The finding that Ded1 significantly stimulates 48S PIC assembly on -SL, containing a 5'-UTR incapable of forming stable structures on its own, suggests that Ded1 has a second role in mRNA recruitment, in addition to unwinding stable structures in 5'-UTRs.
To analyze whether Ded1 can stimulate the PIC attachment step of 48S assembly, we examined the synthetic CP-8.1 mRNA with a stable SL (predicted DG˚= À8.1 kcal/mol) inserted in a cap-proximal location 5 nt from the 5'-end of -SL mRNA. We also analyzed a third synthetic mRNA, CD-8.1, containing the same SL inserted 45 nt downstream from the cap of -SL, reasoning that this cap-distal SL might impede scanning without interfering with PIC attachment at the cap (Figure 2A and Figure 2-figure supplement 1A-B). As expected, both synthetic mRNAs with SLs had order-of-magnitude lower rates of recruitment compared to -SL in the absence of Ded1 ( Figure 2C, blue, inset), indicating that SLs in either location strongly inhibited 48S PIC formation. Ded1 increased the maximal rate of CP-8.1 recruitment by~5 fold, such that the maximal rate for CP-8.1 was still~6 fold below that of -SL ( Figure 2C and Figure 2-figure supplement 1F, blue vs. red), and~2 -3 fold below that of RPL41A, HOR7, SFT2, OST3, and PMA1 ( Figure 2C and Figure 1B). Moreover, CP-8.1 exhibited an~7.6 fold higher K Ded1 1=2 versus -SL ( Figure 2D). As a result, CP-8.1 lies between the Ded1-hyperdependent and Ded1-hypodependent mRNAs, and deviates from the line in the plot of 1=2 versus fold-increase in k max ( Figure 2E, green, CP-8.1). This deviation is due to the fact that CP-8.1, like the Ded1-hyperdependent mRNAs, is inefficiently recruited in the absence of Ded1; but unlike the Ded1-hyperdependent mRNAs, its recruitment is accelerated by Ded1 only~5 fold. These results suggest that Ded1 is not very effective at reducing the inhibitory effect of a stable cap-proximal SL on PIC attachment, even at saturating Ded1 concentrations. In contrast, Ded1 conferred an~17 fold acceleration of recruitment for CD-8.1 (Figure 2C), and a high Ded1 concentration was required to achieve the half-maximal rate ( Figure 2D and  Figure 2E, it appears that Ded1 is better at resolving the inhibitory effect of the synthetic SL in a cap-distal versus cap-proximal location in the 5'-UTR. Interestingly, the same conclusion was reached previously from analyzing the relative effects of a cold-sensitive ded1 mutation on expression of reporter mRNAs containing cap-proximal or cap-distal SLs (Sen et al., 2015). The ATPase deficient mutant Ded1 E307A did not increase the recruitment rates for these mRNAs above the levels seen in the absence of Ded1, even in case of -SL mRNA which lacks stable secondary structures in the 5'-UTR ( Figure 2-figure supplement 1G).
Having observed that the strength of a 5'-UTR SL influences the degree of Ded1-dependence, as observed with WT versus mutant derivatives of SFT2 and PMA1 ( Figure 2C), we went on to analyze synthetic mRNAs containing cap-distal SLs of predicted stabilities either higher (À10.5 kcal/mol) or lower (À3.7 kcal/mol) than that of CD-8.1. Each of these SLs, present in CD-10.5 and CD-3.7, respectively, reduced the recruitment rate in the absence of Ded1 by~20 fold compared to that of -SL, similar to the results for CD-8.1 ( Figure 2F, blue). This result suggests that eIF4EÁeIF4GÁeIF4A alone cannot efficiently mitigate the inhibitory effects of cap-distal SLs of even moderate stability, such as that in CD-3.7, consistent with previous findings . As expected, Ded1 strongly stimulated the recruitment rate of CD-3.7, by~25 fold, slightly more than the~15 fold observed for CD-8.1 ( Figure 2F); however, Ded1 conferred only a modest~2 fold acceleration of CD-10.5 recruitment ( Figure 2F). One possibility to explain these last results would be that the cap-distal SL in CD-10.5 is too stable for efficient unwinding by Ded1, limiting Ded1's ability to accelerate 48S assembly on this mRNA.
In summary, our analysis of synthetic mRNAs supports the notion that Ded1 can accelerate mRNA recruitment by enhancing scanning through cap-distal secondary structures, such as the SLs that occur naturally in SFT2 and PMA1 mRNAs; although if the structure is too stable, Ded1 has a limited ability to unwind it. Additionally, it appears that Ded1 also partially mitigates the inhibitory effects of cap-proximal secondary structures on PIC attachment at the mRNA 5'-end, although not to the same degree that it overcomes cap-distal structures.
Ded1 stimulation is completely dependent on eIF4EÁeIF4G for a subset of mRNAs and all mRNAs require eIF4A in the presence or absence of Ded1 Ded1 has been shown to interact physically with eIF4F components eIF4E, eIF4G and eIF4A in a manner that influences its ability to unwind model RNA substrates in an unwinding assay (Gao et al., 2016;Hilliker et al., 2011;Senissar et al., 2014). Accordingly, we examined whether its interactions with eIF4F influence Ded1's ability to accelerate 48S PIC assembly. As eIF4E is co-purified with eIF4G (Mitchell et al., 2010), and eIF4E is required for full activity of the eIF4F complex, all experiments involving eIF4G utilize the eIF4EÁeIF4G heterodimer. We performed mRNA recruitment assays in the presence and absence of eIF4EÁeIF4G and Ded1 for (i) Ded1-hypodependent mRNAs RPL41A, HOR7, -SL, and SFT2-M; (ii) Ded1-hyperdependent mRNAs SFT2, OST3, and CD-8.1; and (iii) CP-8.1, which exhibits intermediate behavior between groups (i) and (ii) mRNAs in the relationship between

K Ded1
1=2 and k max stimulation ( Figure 2E). The 5'-cap blocks aberrant mRNA recruitment and imposes a requirement for eIF4EÁeIF4G and eIF4A for maximal recruitment rate. 43S PICs can bind to uncapped mRNAs but are unable to either locate or stably associate with their start codons (Mitchell et al., 2010). Accordingly, only 5'-capped mRNAs were used in all experiments to avoid the formation of these aberrant complexes, especially in the absence of eIF4EÁeIF4G or eIF4A.
As described above, in the presence of eIF4EÁeIF4G, all eight mRNAs can be recruited in the absence of Ded1, and Ded1 increased their recruitment rates to different extents ( Figure 3A-H, blue vs. orange). RPL41A, HOR7, and -SL, which contain very short (RPL41A and HOR7) or unstructured (-SL) 5'-UTRs, were recruited at relatively low (but measurable) rates in the absence of both eIF4EÁeIF4G and Ded1 ( Figure 3A,B,F, tan bars); and Ded1 conferred no increase in their apparent rates in the absence of eIF4EÁeIF4G ( Figure 3A,B,F, cf. green vs. tan, orange vs. blue). The complete dependence of Ded1 on eIF4EÁeIF4G to accelerate recruitment for these three mRNAs is consistent with the idea that Ded1 acts exclusively in the context of the eIF4GÁeIF4EÁeIF4AÁDed1 quaternary complex (Gao et al., 2016). The same might be true for CP-8.1, whose observable stimulation by Ded1 also required eIF4EÁeIF4G ( Figure 3H); however, because no CP-8.1 recruitment was observed without eIF4EÁeIF4G, Ded1 might stimulate recruitment of this mRNA on its own at levels below the detection limit of the assay. The finding that CP-8.1 differs from -SL in showing no measurable recruitment in the absence of eIF4EÁeIF4G ( Figure 3H vs 3F), suggests that the cap-proximal SL in CP-8.1 imposes a requirement for eIF4EÁeIF4G.
In contrast to the results described above, Ded1 can accelerate recruitment of SFT2, OST3, SFT2-M, and CD-8.1 mRNAs in the absence of eIF4EÁeIF4G, with comparable apparent rates afforded by eIF4EÁeIF4G alone for SFT2, OST3, and CD-8.1 (Figure 3C,D,G; green vs. blue), but well below the apparent rates observed with eIF4EÁeIF4G for SFT2-M ( Figure 3E, green vs. blue). Thus, Ded1 can stimulate recruitment of these four mRNAs, at least to some extent, acting outside of the eIF4GÁeIF4EÁeIF4AÁDed1 complex. However, because the maximal rates were observed in the presence of both eIF4EÁeIF4G and Ded1 (Figure 3C-E,G, orange bars), Ded1 likely functions within the eIF4GÁeIF4EÁeIF4AÁDed1 complex as well for these four mRNAs.
In contrast to eIF4EÁeIF4G, omitting eIF4A from the reactions essentially eliminated recruitment of all mRNAs tested, yielding endpoints of <10%; with the exception of -SL, which was recruited at the low rate of 0.04 ± 0.01 min -1 with endpoints of 46 ± 3% (Figure 3-figure supplement 1, red). Moreover, in the absence of eIF4A, Ded1 did not rescue recruitment of any mRNAs, nor did it increase the k app or endpoints for -SL mRNA (Figure 3-figure supplement 1, blue). Thus, eIF4A has one or more essential functions in mRNA recruitment that cannot be provided by Ded1, even for an mRNA such as -SL exhibiting appreciable recruitment in the absence of eIF4EÁeIF4G ( Figure 3F, tan). This is consistent with the previous findings that eIF4A is required for robust translation of virtually all yeast mRNAs in vivo and in vitro (Sen et al., 2015;Yourik et al., 2017), and that DHX29 and yeast Ded1 cannot substitute for eIF4A in 48S PIC assembly on native b-globin mRNA in a mammalian reconstituted system (Abaeva et al., 2011;Pisareva et al., 2008). Moreover, the fact that Ded1 does not accelerate recruitment of -SL mRNA in the absence of eIF4A (Figure 3-figure supplement 1H, blue vs. red) indicates that, at least for this mRNA with an unstructured leader, Ded1 can only promote mRNA recruitment in the presence of eIF4A.
Interaction between the RNA3 domain of eIF4G and Ded1-CTD stimulates mRNA recruitment The C-terminal RNA3 domain of eIF4G was shown to interact physically with the Ded1 CTD (Hilliker et al., 2011) and to influence effects of eIF4G on Ded1 unwinding of a model RNA duplex in vitro (Gao et al., 2016;Putnam et al., 2015). Hence, we sought to determine whether this physical interaction between the CTDs of Ded1 and eIF4G is functionally relevant in 48S PIC assembly by performing mRNA recruitment assays with a truncated eIF4G variant lacking RNA3 (eIF4EÁeIF4G-D RNA3) or a Ded1 variant lacking the CTD (Ded1-DCTD) (Figure 4-figure supplement 1A).
Without Ded1, the DRNA3 truncation of eIF4G had little or no effect on k max for all seven mRNAs tested ( Figure 4A, compare blue bars to superimposed line/whiskers, the latter indicating results for WT eIF4EÁeIF4G re-plotted from Figure 3A-H for comparison, where the horizontal line shows the mean and the whiskers one SD from the mean). In the presence of Ded1, by contrast, DRNA3 conferred~2 fold reductions in k max for three mRNAs, RPL41A (DRNA3 -1.5 ± 0.2 min -1 ; WT -3.9 ± 0.5 min -1 ), HOR7 (DRNA3 -1.8 ± 0.1 min -1 ; WT -3.9 ± 0.3 min-1) and CP-8. - -  Figure 4A). Thus, as summarized in Figure 4B, deleting RNA3 nearly eliminated the stimulatory effect of Ded1 on k max values for RPL41A, HOR7, and CP-8.1 (black vs. grey bars vs. dashed red line, the latter indicating no stimulation by Ded1).
Importantly, when Ded1 was replaced with Ded1-DCTD in reactions containing WT eIF4EÁeIF4G, we observed effects on k max values and rate enhancements very similar to those described above for DRNA3 ( Figure 4D-E), which is consistent with a functional interaction between RNA3 and Ded1-CTD. (We verified that saturating concentrations of eIF4EÁeIF4G-DRNA3 and Ded1-DCTD were used in these experiments by showing that rates of recruitment for SFT2, RPL41A or HOR7 mRNAs were not elevated even at much higher concentrations of the variants (Figure 4-figure supplement 1B-D, and data not shown). We also confirmed that Ded1-DCTD has RNA-dependent ATPase activity similar to that of the WT Ded1 (Figure 1-figure supplement 1D-E)).
To further verify that RNA3 of eIF4G and Ded1-CTD interact in mRNA recruitment, we combined the eIF4EÁeIF4G-DRNA3 and Ded1-DCTD variants in the same assays (at the saturating concentrations determined for each alone). We examined recruitment of RPL41A, for which eIF4G-RNA3 and Ded1-CTD were each essential for rate enhancement by Ded1; CD-8.1, whose k max was not significantly reduced by eIF4EÁeIF4G-DRNA3 or Ded1-DCTD compared to the corresponding WT proteins; and CP-8.1, for which each domain deletion had an intermediate effect on rate enhancement by Ded1. The effects on the k max values for these three mRNAs on combining the eIF4EÁeIF4G-DRNA3 and Ded1-DCTD mutants were similar to what we observed with each deletion individually (Figure 4-figure supplement 1E). These data strongly suggest that the changes in k max values conferred by eliminating either eIF4G-RNA3 or Ded1-CTD arise from loss of interaction between these two domains, because once the interaction is disrupted by removing one domain no further defect results from also removing the other.
We next examined whether eliminating eIF4G-RNA3 alters the concentration of eIF4EÁeIF4G required to achieve half-maximal rate acceleration, that is, the K eIF4EÁeIF4G

1=2
. To this end, we first deter-   Figures 1 and 2), and error bars represent 1 SD from the mean (this representation will be referred to as line/whisker plot). See accurately because of its endpoint defects at lower eIF4EÁeIF4G concentrations.) Thus, the maximum stimulation of recruitment of the synthetic mRNAs in the absence of Ded1 can be achieved at relatively low eIF4EÁeIF4G concentrations, but higher eIF4EÁeIF4G concentrations are required to support the additional stimulation of recruitment conferred by Ded1. (See Figure 4-figure supplement 2 legend for additional comments.) We then proceeded to determine the effect of eliminating RNA3 on the concentration of eIF4E-ÁeIF4G required for the maximum recruitment rate in the absence of Ded1. For mRNAs RPL41A, HOR7, SFT2, SFT2-M, and -SL, the K 1/2 values for eIF4EÁeIF4G-DRNA3 did not differ substantially from those of WT eIF4EÁeIF4G, although it was~2 fold higher for CP-8.1 ( Figure 4C, blue bars vs. line/whiskers summarizing results for WT eIF4EÁeIF4G taken from Figure 4-figure supplement 2A).
could not be accurately measured for CD-8.1 using eIF4EÁeIF4G-DRNA3 due to endpoint defects.) In reactions containing Ded1, by contrast, the K eIF4EÁeIF4G 1=2 values for eIF4EÁeIF4G -D RNA3 were increased by 2-to 5-fold relative to the values determined for WT eIF4EÁeIF4G for all mRNAs tested ( Figure 4C, red bars versus superimposed line/whiskers results for WT eIF4EÁeIF4G taken from Figure 4-figure supplement 2A). Thus, on removal of RNA3, relatively higher concentrations of eIF4EÁeIF4G are required to achieve maximal rate stimulation by Ded1, supporting a functionally important interaction between RNA3 and Ded1. We also determined the effects of eliminating the Ded1 CTD on K 1/2 values for Ded1. Similar to the results obtained for eIF4EÁeIF4G-D RNA3, higher K 1/2 values were observed for Ded1-DCTD versus WT Ded1 for four of the five mRNAs that exhibit appreciable rate stimulation by Ded1-DCTD (which excludes RPL41A and HOR7) ( Figure 4F, green bars versus superimposed line/whiskers results for WT Ded1 from Figures 1C and  2D).
In summary, our results indicate that there is an interaction between eIF4G-RNA3 and Ded1-CTD that facilitates Ded1 function in mRNA recruitment to the PIC. The increased K 1/2 values evoked by eliminating either domain suggests that their interaction enhances assembly of the eIF4GÁeIF4EÁeIF4AÁDed1 tetrameric complex (Gao et al., 2016). The finding that increased concentrations of eIF4EÁeIF4G-DRNA3 or Ded1-DCTD can rescue the rate (k max ) defects caused by the domain deletions for some mRNAs but not for others suggests that in certain mRNA contexts this interaction plays a role in addition to simply promoting interaction between eIF4G and Ded1.
The RNA2 domain of eIF4G functions in Ded1-dependent mRNA recruitment Ded1 also interacts with the RNA2 domain of eIF4G (Senissar et al., 2014), but the importance of this interaction for Ded1 function is unknown. Comparing the eIF4EÁeIF4G-DRNA2 variant (with an internal deletion of RNA2) to WT eIF4EÁeIF4G, we found that, as for RNA3, deletion of RNA2 influenced recruitment of the mRNAs to different extents. In reactions lacking Ded1, we observed no significant differences in k max for any of the seven mRNAs examined ( Figure 5A, blue bars vs. line/ whiskers for WT eIF4EÁeIF4G data from Figure 3, blue). By contrast, removal of RNA2 increased the K 1/2 for eIF4EÁeIF4G-DRNA2 versus WT eIF4EÁeIF4G by 3 -5-fold for SFT2-M, -SL, and CP-8.1 mRNAs in reactions lacking Ded1 ( Figure 5C, blue bars vs. line/whiskers from Figure 4-figure supplement 2A). DRNA2 also conferred an endpoint defect for SFT2 at lower concentrations, precluding determination of its effects on the K 1/2 for eIF4EÁeIF4G. (Because of endpoint defects for CD-8.1 even with WT eIF4EÁeIF4G, the importance of RNA2 cannot be evaluated for this mRNA.) Considering that DRNA3 increased the K 1/2 for eIF4EÁeIF4G only for CP-8.1 in reactions lacking Ded1, it appears that RNA2 has relatively more important Ded1-independent functions than RNA3 in recruitment of particular mRNAs.
Stronger effects of DRNA2 were observed in reactions containing Ded1, markedly reducing the k max for the two mRNAs containing cap-distal SLs, SFT2 (k max of 1.4 ± 0.1 min -1 vs. 3.4 ± 0.1 min -1 for WT) and CD-8.1 (k max of 0.7 ± 0.03 min -1 vs. 1.9 ± 0.2 min -1 for WT), with smaller reductions for SFT2-M and CP-8.1 ( Figure 5A, orange bars vs. line/whiskers; and Figure 5B, black vs. grey bars vs. dotted red line). By contrast, RNA2 was dispensable for maximal Ded1 acceleration of RPL41A, HOR7, and -SL mRNA recruitment ( Figure 5B, black vs. grey bars). It is noteworthy that RPL41A and HOR7 exhibited the strongest dependence on RNA3 ( Figure 4B), but were insensitive to loss of RNA2 ( Figure 5B) for maximal rate stimulation by Ded1.
All of the mRNAs exhibited increases in K 1/2 for eIF4EÁeIF4G-DRNA2 versus WT eIF4EÁeIF4G (3 -20-fold; Figure 5C, red bars vs. line/whiskers). Thus, eliminating RNA2 increases the concentration of eIF4EÁeIF4G required for maximal Ded1 stimulation for all mRNAs tested, which might reflect its importance in promoting formation of the eIF4GÁeIF4EÁeIF4AÁDed1 complex, in the manner concluded above for RNA3. Elevated concentrations of eIF4EÁeIF4G-DRNA2 enable maximum recruitment rates similar to those achieved with WT eIF4EÁeIF4G for the mRNAs with lower degrees of structure -RPL41A, HOR7, and -SL; whereas the more structured mRNAs display varying reductions in k max at saturating eIF4EÁeIF4G-DRNA2 concentrations, with the two mRNAs harboring capdistal stem loops -SFT2 and CD-8.1 -having the largest rate enhancement defects. This suggests that RNA2 enhances Ded1 function on the structured mRNAs beyond its ability to simply stabilize Ded1-eIF4G interaction.
The N-terminal domain of Ded1 enhances mRNA recruitment It was shown previously that the N-terminal domain (NTD) of Ded1 physically interacts with eIF4A, and is required for eIF4A stimulation of Ded1 unwinding activity in vitro (Gao et al., 2016;Senissar et al., 2014). We tested the effects of Ded1 on K 1/2 of eIF4A and, as observed with WT eIF4EÁeIF4G, Ded1 influenced K elF4A 1=2 differently on these mRNAs, providing evidence that Ded1 has functional interactions with eIF4A during mRNA recruitment (Figure 4-figure supplement 2B). Hence, we examined the effect of eliminating the Ded1-NTD on recruitment of our panel of mRNAs. The k cat and K m values for RNA-dependent ATP hydrolysis were indistinguishable between WT Ded1 and the DNTD variant (Figure 1-figure supplement 1D-E). For the SFT2, CD-8.1, and CP-8.1 mRNAs, which harbor defined SLs in their 5'-UTRs, eliminating the Ded1 NTD decreased k max by 1.5 -2-fold, whereas the k max values for RPL41A, HOR7, SFT2-M, and -SL mRNAs were not significantly altered by DNTD ( Figure 6A, purple bars vs. line/whiskers for WT Ded1; and Figure 6B). However, 1 -2 orders of magnitude higher K 1/2 values for the Ded1-DNTD versus WT Ded1 were observed with all mRNAs except CD-8.1 ( Figure 6C, green bars vs. line/whiskers for WT Ded1). Thus, removing the Ded1 NTD significantly increases the concentrations of Ded1 required to achieve enhancement of recruitment of six out of the seven mRNAs tested. These data are consistent with the idea that interaction of eIF4A and the Ded1 NTD enhances assembly or stability of the eIF4GÁeIF4EÁeIF4AÁDed1 complex, and stimulates Ded1 helicase activity; although the Ded1-NTD might also mediate important interactions with other components of the system. As with deletions of the eIF4G RNA domains and Ded1-CTD, mRNA-specific defects were conferred by deleting the Ded1 NTD.

Discussion
Employing a purified yeast translation initiation system, we reconstituted the function of DEAD-box helicase Ded1 in stimulating the rate of 48S PIC assembly on both native and model mRNAs. This stimulation in vitro recapitulates the Ded1-dependence of translation of mRNAs observed in vivo using ribosome profiling, in which mRNAs having longer and more structured 5'-UTRs display hyperdependence on Ded1 relative to mRNAs with shorter and less structured 5'-UTRs (Sen et al., 2015). We showed that defined SL structures both decrease rates of 48S PIC assembly in the absence of Ded1 and increase the fold-stimulation afforded by Ded1. These results provide direct biochemical evidence supporting the proposition that Ded1 enhances translation initiation in vivo by resolving secondary structures formed by 5'-UTR sequences. Our results also showed that Ded1-accelerated recruitment of several mRNAs depends completely on the presence of eIF4EÁeIF4G, and that domains mediating Ded1 interactions with eIF4G or eIF4A enhance Ded1 stimulation of 48S PIC assembly for all mRNAs tested, consistent with the previous work indicating that Ded1 binding to the eIF4GÁeIF4EÁeIF4A complex enhances its activity in RNA unwinding assays (Gao et al., 2016). However, Ded1 can also stimulate recruitment of some mRNAs in the absence of eIF4EÁeIF4G, indicating that it can act independently of eIF4F as well.

DEAD-box proteins Ded1 and eIF4A have complementary but distinct functions in mRNA recruitment
Inactivation of conditional mutants of either Ded1 or eIF4A in vivo results in strong reduction in bulk polysomes and decreased expression of reporter mRNAs bearing unstructured 5'-UTRs (Chuang et al., 1997;de la Cruz et al., 1997;Sen et al., 2015), indicating that both proteins are important for translation of most mRNAs. These general reductions of translation are masked in ribosome profiling studies because the ribosomal footprint and mRNA read counts must be normalized to the total reads obtained in each sample/strain, such that the TE change for each mRNA is determined relative to the average TE change for all mRNAs examined in each sample/strain. mRNAs judged to hyperdependent or hypodependent on Ded1 in ribosome profiling experiments exhibit larger or smaller than average reductions in relative TE, respectively, but they may all exhibit decreased absolute TEs in ded1 vs. WT cells (Sen et al., 2015). The ribosome profiling analysis of ded1 mutants revealed that~10% of all mRNAs -particularly those with long, structured 5'-UTRsexhibit greater than average TE reductions in the ded1 vs. WT cells, and were designated as Ded1hyperdependent. Consistent with a stimulatory role for Ded1 in translation of most mRNAs, we observed here that Ded1 increases the maximal rate of 48S PIC formation by~2 -3 fold on native Ded1-hypodependent mRNAs or mRNAs with 5'-UTRs of low structural complexity (RPL41A, HOR7, and SFT2-M), as well as on a synthetic mRNA with an unstructured 5'-UTR (-SL) ( Figures 1B and  2C). Importantly, Ded1 conferred much greater acceleration of 48S PIC assembly on all four Ded1hyperdependent mRNAs examined ( Figure 1B).
eIF4A also enhances the translation of nearly all mRNAs in vivo, although, unlike Ded1 where sizable sets of mRNAs are hyper-or hypo-dependent on its function, most mRNAs are similarly (strongly) dependent on eIF4A for translation (Firczuk et al., 2013;Sen et al., 2015). In line with these in vivo observations, in the reconstituted system eIF4A promotes 48S PIC assembly on all mRNAs tested, increasing the k max for the synthetic mRNA with unstructured 5'-UTR (-SL) by 60-fold and even accelerating recruitment of completely unstructured model mRNAs by !7-fold . However, although Ded1 and eIF4A both facilitate recruitment of most mRNAs, and both are essential in vivo (Chuang et al., 1997;Linder and Slonimski, 1989); their functions are distinct. Ded1 cannot substitute for eIF4A in vitro (Figure 3-figure supplement 1), but it promoted recruitment of all mRNAs tested beyond the level achieved by saturating concentrations of eIF4A and eIF4EÁeIF4G ( Figures 1B and 2C).
We previously proposed that eIF4A stimulates a step of 48S PIC assembly common to all mRNAs, such as disrupting the ensemble of weak RNA-RNA interactions that impede PIC attachment to the 5'-UTR or subsequent scanning . In addition, eIF4A might also directly promote loading of mRNA onto the PIC, for example by modulating conformational changes in the 40S subunit or by threading the 5'-end into the mRNA binding channel (Kumar et al., 2016;Sokabe and Fraser, 2017). In common with eIF4A, Ded1 may enhance recruitment of all mRNAs by disrupting their global structures created by dynamic ensembles of base-pairing throughout their lengths. Unlike eIF4A, however, Ded1 can efficiently resolve more stable structures, including local stemloops-achieving an order-of-magnitude acceleration for mRNAs with the most structured 5'-UTRs. If the proposed Ded1 function in promoting 48S PIC formation by disrupting global mRNA structure requires lower Ded1 concentrations than its role in resolving more stable structures within or involving the 5'-UTR, it would be consistent with our findings that mRNAs with SLs require higher Ded1 concentrations to achieve the much greater fold-stimulation of 48S assembly afforded by Ded1 compared to mRNAs lacking SLs (Figures 1 and 2).
Evidence supporting the functional importance of an eIF4GÁeIF4EÁeIF4AÁDed1 tetrameric complex Ded1 alters the K 1/2 of eIF4EÁeIF4G and eIF4A for most mRNAs (Figure 4-figure supplement 2), and these new K 1/2 values may signify the changes in the concentrations of eIF4A and eIF4EÁeIF4G required for proper assembly of the eIF4GÁeIF4EÁeIF4AÁDed1 tetrameric complex on each mRNA. The deleterious effects of eliminating known interactions between Ded1 and eIF4G or eIF4A further suggests the importance of the tetrameric complex formation for robust Ded1 function. With only two exceptions (CP-8.1 for Ded1-DCTD and CD-8.1 for the Ded1-DNTD), we found that deleting the RNA2 or RNA3 domain of eIF4G, or the CTD or NTD of Ded1, increased the concentrations of the corresponding eIF4G or Ded1 variants required to achieve the half-maximal rate of 48S PIC assembly (ie., their K 1/2 values) on each mRNA examined, as summarized by the heatmap in Figure 7A. This holds for the mRNAs with the shortest or least structured 5'-UTRs (-SL, RPL41A, and HOR7) as well as those with the most highly structured 5'-UTRs (SFT2, CD-8.1, and CP-8.1). Because all of these domain deletions abrogate known interactions linking Ded1 to eIF4G or eIF4A, a plausible way to account for these findings is to propose that, regardless of the amount of secondary structure in the mRNA 5'-UTR, rapid mRNA recruitment depends on Ded1 functioning within the eIF4EÁeIF4GÁeIF4AÁDed1 complex; and that eliminating any interaction between Ded1 and eIF4G or eIF4A necessitates a higher concentration of the mutant variant for efficient complex formation. Judging by the magnitude of the increases in K 1/2 conferred by eliminating different domains ( Figure 7A), it would appear that eIF4G-RNA2 and Ded1-NTD are generally more important than the eIF4G-RNA3/Ded1-CTD duo in promoting assembly or stability of the eIF4EÁeIF4GÁeIF4AÁDed1 complex. However, we cannot rule out the possibility that the domain deletions also impair a different interaction, for instance, with mRNA or another factor such as eIF4B or eIF3 that is crucial for rapid mRNA recruitment (Mitchell et al., 2010).
Ded1 and eIF4G mutants affect the maximal rates for recruitment differently depending on the mRNA With some mRNAs, the maximum rate of recruitment observed with WT Ded1 and WT eIF4EÁeIF4G (k max ) could also be achieved using elevated concentrations of the mutant variant; whereas in other cases, the observed k max was diminished from the WT value even at saturating amounts of the mutant Ded1 or eIF4G variant. We regard such reductions in k max as indicating impairment of a fundamental role of the deleted eIF4G or Ded1 domain in rapid recruitment of the affected mRNA. Accordingly, we summarized these effects for each factor truncation in a heatmap ( Figure 7B) to evaluate the requirements for particular domains or interactions for maximum rate stimulation by Ded1 with each mRNA tested.
It is evident from the heatmap that deletion of the eIF4G-RNA3 or Ded1-CTD domain confers a wide range of k max reductions: 2 -3-fold for RPL41A and HOR7;~1.5 fold for SFT2, SFT2-M, and CP-8.1, and almost no change for -SL and CD-8.1 (Figure 7B, cf. cols. 1 -2, all rows). Importantly, however, in all cases the effects of deleting eIF4G-RNA3 and Ded1-CTD are similar for a given mRNA, supporting the proposition that these effects result from loss of the eIF4G-RNA3/Ded1-CTD interaction. Eliminating this interaction essentially abolishes the rate enhancement provided by Ded1 for recruitment of RPL41A and HOR7 mRNAs ( Figure 4B,E and Figure 4-figure supplement 1E), indicating that the Ded1-CTD/eIF4G-RNA3 interaction is essential for the eIF4EÁeIF4GÁeIF4AÁDed1 complex (i.e., at saturating Ded1 concentration) to accelerate a slow step in 48S PIC assembly on these mRNAs ( Figure 7C(i)). This step is apparently enhanced through different interactions or is less rate-limiting for the other mRNAs tested.
We observed a similar diversity of effects on k max values depending on the mRNA for the eIF4G-RNA2 and Ded1-NTD deletions. DRNA2 had little effect on k max for RPL41A and HOR7, in contrast to the deleterious effects of DRNA3 for these two mRNAs. Similarly, DRNA2 markedly reduced the k max for SFT2 and CD-8.1 by~2.5 -3-fold, whereas DRNA3 had little effect on these mRNAs ( Figure 7B, col. 3 vs. col. 1). Similar to DRNA3 however, DRNA2 decreased the k max values for SFT2-M and CP-8.1 by~1.5 -2-fold, with minimal effect on -SL mRNA ( Figure 7B). In this case, the RNA2 domain appears to be most important for the ability of the eIF4EÁeIF4GÁeIF4AÁDed1 complex to stimulate recruitment of the two mRNAs with cap-distal SLs, SFT2 and CD-8.1; nearly dispensable for the mRNAs with the lowest degrees of structure, -SL, RPL41A and HOR7; and of intermediate importance for CP-8.1 and SFT2-M. These observations suggest that RNA2 facilitates Ded1 function in melting out structures encountered by the PIC during attachment or scanning ( Figure 7C(ii-iv)). (Although SFT2-M lacks the major cap-distal SL in WT SFT2, it contains an additional cap-proximal structure in vitro that might underlie its greater dependence on RNA2 versus -SL, RPL41A, and HOR7.) Whereas deletion of the Ded1-NTD had the largest effect of the four eIF4G or Ded1 truncations on K 1/2 values ( Figure 7A), it had the smallest effects on maximal rates of recruitment, reducing k max values between~1.5 to~2 fold for SFT2, CD-8.1, and CP-8.1, but having little effect on the other mRNAs ( Figure 7B). As all three affected mRNAs have stable SLs and the unaffected mRNAs do not, these data might indicate that the Ded1-NTD, presumably by interacting with eIF4A, modestly enhances the ability of the eIF4EÁeIF4GÁeIF4AÁDed1 complex to unwind stable secondary structures-complementing the function of eIF4G-RNA2 in this reaction ( Figure 7C(ii-iii)).
What is perhaps most striking about the effects of the eIF4G and Ded1 truncations on both the K 1/2 ( Figure 7A) and k max ( Figure 7B) values is that mRNAs have distinct patterns of responses. The geometry' model depicting how different mRNAs exhibiting distinct configurations of the occurrence and location of RNA structures (shown as hairpins or stem-loops) could influence the relative importance of different domain interactions linking Ded1 to eIF4G or eIF4A within the eIF4GÁeIF4EÁeIF4AÁDed1 tetrameric complex. (D) 'Kinetic' model depicting how different mRNAs might differ in the extent to which PIC attachment or scanning are the rate-limiting steps in 48S PIC assembly. Depending on which step is rate-limiting, the requirements for Ded1, either acting alone or within the eIF4GÁeIF4EÁeIF4AÁDed1 complex, could be different on different mRNAs. DOI: https://doi.org/10.7554/eLife.38892.024 various eIF4G-Ded1 domain interactions affect CP-8.1 and CD-8.1 recruitment quite differently even though these mRNAs differ only by the location of the same SL in an unstructured 5'-UTR. This suggests that there is not a single, uniform mechanism through which Ded1 operates on all mRNAs; instead the diversity of structures in mRNAs requires that Ded1 and the eIF4EÁeIF4GÁeIF4AÁDed1 complex can operate in multiple modes. The structural diversity inherent in mRNAs presents a challenge for the translational machinery because, once the eIF4F complex attaches to the 5'-cap, structural elements could be oriented in a variety of locations in three-dimensional space relative to its functional domains ( Figure 7C). This problem could explain why eIF4G is so large and flexible and has multiple RNA-and factor-binding domains, which might confer sufficient plasticity to interact with mRNA structures presented in a variety of orientations and distances. Likewise, the multiple Ded1 binding domains on eIF4G might allow Ded1 to assume different positions relative to the diverse mRNA structures it encounters on different mRNAs. The mRNA specificity of effects of the truncation mutants of Ded1 and eIF4G on both K 1/2 and k max values are consistent with the notion that the eIF4EÁeIF4GÁeIF4AÁDed1 complex can interact with and modulate the structures of mRNAs in different ways, with the mRNA structure dictating the particular interactions of Ded1 with eIF4G or eIF4A that are most crucial for rapid recruitment.
It is likely that the rate-limiting step(s) for 48S PIC formation will also vary depending on the unique structural features of the mRNA. For some mRNAs, PIC attachment to the 5'-UTR might be rate-limiting because of structures proximal to the 5'-end or because the 5'-end is occluded within the global structure of the mRNA ( Figure 7D (i)). For other mRNAs, PIC attachment might be relatively fast, but scanning to the start codon could be impeded by stable structures that require Ded1 in the context of the eIF4EÁeIF4GÁeIF4AÁDed1 complex to resolve ( Figure 7D (ii)). Since the RPL41A and HOR7 mRNAs have short and less-structured 5'-UTRs, it is plausible that PIC attachment could be rate-limiting on these two mRNAs ( Figure 7D (i)), whereas scanning of the 5'-UTR could be ratelimiting on the two cap-distal SL-containing mRNAs, SFT2 and CD-8.1 (Figure 7D (ii)).
Recruitment of some mRNAs was accelerated by Ded1 in the absence of eIF4EÁeIF4G, including SFT2, SFT2-M, OST3, and CD-8.1 (Figure 3C-E,G) indicating that Ded1 is also capable of stimulating one or more aspects of 48S PIC assembly outside of the context of the eIF4EÁeIF4GÁeIF4AÁDed1 complex on certain mRNAs. The three mRNAs that most clearly exhibit complete dependence on eIF4EÁeIF4G for Ded1 stimulation, -SL, RPL41A, and HOR7, have low degrees of structure in their 5'-UTRs. Hence, an intriguing possibility is that for mRNAs lacking strong local secondary structure in the 5'-UTR, Ded1 is only needed to promote eIF4F binding to the cap or initial attachment of the PIC at the 5'-end, and this process requires direct interaction of Ded1 with eIF4F at the mRNA 5'end ( Figure 7D (i)). For other mRNAs harboring strong local structures in the 5'-UTR, in addition to acting in a eIF4F-Ded1 tetrameric complex to facilitate PIC attachment, Ded1 might unwind these structures and promote scanning independently of its association with eIF4F ( Figure 7D (ii)). The apparent inability of Ded1 to accelerate recruitment of CP-8.1 independently of eIF4EÁeIF4G might be explained by noting that the SL in this mRNA is cap-proximal, which could require eIF4F-Ded1 interaction for unwinding; however, because no recruitment of this mRNA was observed in the absence of eIF4EÁeIF4G or Ded1, it is possible that Ded1 can actually accelerate CP-8.1 recruitment on its own but the rate is too low to be detected in the absence of eIF4EÁeIF4G. Additionally, Ded1 can interact with other factors such as eIF4A, the 40S ribosomal subunit, or the mRNA itself, which might aid in the recruitment of these mRNAs without eIF4EÁeIF4G (Gao et al., 2016;Guenther et al., 2018).
The two different models ('mRNA geometry' and 'rate-limiting steps', Figure 7C-D) we are considering to explain the differential effects of the eIF4G and Ded1 domain deletions on different mRNAs are not mutually exclusive. In fact, the proposal that the domains have some specificity for mediating PIC attachment versus scanning probably requires that they localize Ded1 to different parts of the mRNA because the former reaction would occur closer to the 5'-end whereas the latter would occur distal to it. The length and flexibility of eIF4G, coupled with the complex network of interactions possible among eIF4G, eIF4E, eIF4A, Ded1 and mRNA, could have evolved to support the plasticity required to deal with the wide variety of mRNA shapes, sizes and structures that must be loaded onto PICs for translation in eukaryotic cells where transcription and translation are uncoupled.
mRNAs can form long-range interactions between their 5'-UTRs and coding sequences or 3'-UTRs, but because our reporter mRNAs consisted of only 5'-UTRs and the first 60 nucleotides of