eIF4A is stimulated by the pre-initiation complex and enhances recruitment of mRNAs regardless of structural complexity

eIF4A is a DEAD-box RNA-dependent ATPase thought to unwind RNA secondary structure in the 5′-untranslated regions (UTRs) of mRNAs to promote their recruitment to the eukaryotic translation pre-initiation complex (PIC). We show that eIF4A’s ATPase activity is markedly stimulated in the presence of the PIC, independently of eIF4E•eIF4G, but dependent on subunits i and g of the heteromeric eIF3 complex. Surprisingly, eIF4A accelerated the rate of recruitment of all mRNAs tested, regardless of their degree of structural complexity. Structures in the 5′-UTR and 3′ of the start codon synergistically inhibit mRNA recruitment in a manner relieved by eIF4A, indicating that the factor does not act solely to melt hairpins in 5′-UTRs. Our findings that eIF4A functionally interacts with the PIC and plays important roles beyond unwinding 5’-UTR structure is consistent with a recent proposal that eIF4A modulates the conformation of the 40S ribosomal subunit to promote mRNA recruitment.


24"
The goal of translation initiation is to assemble the ribosomal initiation complex containing the methionyl initiator 25" tRNA (Met-tRNAi) at the translation start site on an mRNA. The process begins when the small (40S) subunit of the 26" ribosome binds eIF1, eIF1A, eIF2, GTP, Met-tRNAi, eIF3, and eIF5, to assemble the 43S PIC (Dever et al., 2016). eIF1

27"
and eIF1A bind near the P and A sites of the 40S subunit, respectively, and promote loading of the ternary complex

16"
provided by RNA or eIF4G•4E, but is dependent on the 3g and 3i subunits of eIF3. eIF4A increases the rate of 17" recruitment for all mRNAs tested, ranging from the natural RPL41A mRNA to short unstructured messages. Structures

18"
in the 5'-UTR and on the 3' side of the start codon synergistically inhibit mRNA recruitment in a manner relieved by

19"
eIF4A. Our data indicate that eIF4A can relieve inhibition of mRNA recruitment arising from structure created by 20" elements throughout the length of the mRNA rather than only resolving secondary structures in the 5'-UTR. Overall,

21"
these results are consistent with a recent model suggesting that eIF4A may modulate the conformation of the 40S

25"
ATP hydrolysis by eIF4A promotes recruitment of the natural mRNA RPL41A as well as a short, unstructured 26" model mRNA

27"
To better understand how eIF4A-catalyzed ATP hydrolysis is related to the removal of RNA structure and 28" mRNA recruitment, we compared the kinetics of recruitment of the natural mRNA RPL41A (possessing structural 29" 1A-B; see RNAs 1 and 10 in Supplementary Methods). As for most natural mRNAs, RPL41A is thought to have 3" numerous base-pairing interactions throughout its length while the CAA 50-mer is expected to have little, if any,

5"
mRNA recruitment experiments were performed as described previously, using an in vitro-reconstituted S.

8"
and 4G (see mRNA Recruitment Assay in Methods). Reactions were initiated by simultaneous addition of ATP and an 9" mRNA labeled with a [ 32 P]-7-methylguanosine (m 7 G) cap, enabling mRNA recruitment to the PICs and formation of 10" 48S complexes. Reaction timepoints were acquired by mixing an aliquot with a 25-fold excess of a non-radioactive

11"
("cold") capped mRNA identical to the labeled one in the reaction, effectively stopping further recruitment of 12" radiolabeled mRNA. The rate of dissociation of recruited mRNAs from the PIC in the presence of the cold chase 13" mRNA was negligible for all mRNAs in the study (data not shown). Free mRNA and 48S complexes were resolved via

15"
We first compared the kinetics of recruitment for RPL41A with CAA 50-mer in the presence and absence of 16" ATP ( Figure 1). In the presence of saturating ATP, the rate of recruitment of RPL41A was 0.74 ± 0.01 min -1 with an 17" endpoint in excess of 90%. In contrast, in the presence of ADP, and in reactions lacking either nucleotide or eIF4A, less 18" than 20% of RPL41A mRNA was recruited after 6 hours, indicating a dramatically lower rate that could not be 19" measured accurately due to the low reaction endpoint ( Figure 1A,C). The CAA 50-mer was recruited in the absence of 20" eIF4A and ATP at rates of about 0.90 min -1 , likely due to lack of significant structure, reaching endpoints around 80%.

21"
Surprisingly, the addition of eIF4A and ATP stimulated recruitment of the CAA 50-mer to a rate that could not be 22" measured accurately by manually quenching the reaction; however, we estimate that the increase in rate was at least 4-

24"
To determine whether ATP hydrolysis is required for the stimulation of the rate of mRNA recruitment that we 25" observed, we next measured the rate of recruitment with the non-hydrolyzable ATP analogs ADPCP and ADPNP, as

27"
supported stimulation of the recruitment of either mRNA by eIF4A, producing rates that were comparable to the 28" observed rates measured in the absence of nucleotide or eIF4A (Figure 1; compare grey and blue curves to red and 29" 6" " purple). In the presence of ATP-γ-S, recruitment of RPL41A and CAA 50-mer was 39-fold (0.019 ± 0.001 min -1 ) and 1" nearly 2-fold (2.3 ± 0.2 min -1 ) slower, respectively, than in the presence of ATP; however, both mRNAs achieved 2" endpoints of approximately 80%, consistent with previous observations that eIF4A is capable of utilizing ATP-γ-S (Peck 3" and Herschlag, 2003). Taken together, these results suggest that ATP hydrolysis by eIF4A stimulates the recruitment of 4" both a natural mRNA harboring structure throughout its sequence and the short unstructured CAA 50-mer.

6"
The ATPase activity of eIF4A•4G•4E is increased by the PIC

7"
Because of the importance of ATP hydrolysis for the ability of eIF4A to stimulate recruitment of both 8" structured and unstructured mRNAs, we next investigated how the mRNA and PIC influence the ATPase activity of 9" eIF4A and the eIF4A•4G•4E heterotrimer. Single turnover conditions in which the concentration of enzyme is 10" saturating and greater than the concentration of the substrate, similar to those employed in the mRNA recruitment 11" experiments above (Figure 1), would be ideal to study the ATPase activity for comparison. However, such an approach

12"
was technically not feasible because the low affinity of yeast eIF4A for ATP (Rajagopal et al., 2012) made it impossible

13"
to create conditions where eIF4A is saturating for ATP binding and stoichiometric to or in excess of ATP. Using

14"
saturating ATP under experimentally accessible concentrations of PIC and eIF4A led to a situation in which the first few

15"
turnovers of ATP hydrolysis produced by any possible eIF4A•PIC complexes would be below the limit of detection

16"
using either radioisotope or spectrophotometric ATPase assays. In addition, based on the high (~5 µM) concentration of

17"
eIF4A required to achieve maximal rates of mRNA recruitment, we also could not achieve a situation in which PICs

18"
were saturating over eIF4A because such concentrations of PICs are not experimentally achievable. Thus, we were not

19"
able to perform pre-steady state kinetic ATPase experiments. Instead, to inquire whether the ATPase activity of eIF4A is 20" affected by the presence of the PIC, we used multiple turnover conditions in which [eIF4A] << [ATP] in the in vitro 21" reconstituted translation initiation system. Although steady-state kinetics do not allow direct comparison to single-

22"
turnover mRNA recruitment assays, the approach still enables detection of the effects of other components of the

23"
system on repeated cycles of ATP hydrolysis by eIF4A and eIF4A•4G•4E. Thus, if the PIC in the absence of mRNA 24" (i.e., prior to mRNA recruitment) or the presence of mRNA (i.e., during and after mRNA recruitment) promotes a state

25"
of eIF4A with altered ATPase activity, it should be possible to detect it using this approach.

26"
ATPase was monitored with an enzyme-coupled assay in which pyruvate kinase and lactate dehydrogenase are

27"
used to regenerate ATP from ADP and, in the process, oxidize NADH to NAD + , producing a change in absorbance at

7" "
Methods). Reactions with varying concentrations of mRNA were assembled in a 384-well plate and initiated by addition 1" of saturating ATP (5 mM). NADH absorbance at 340 nm was recorded every 20 seconds using a microplate reader. By

2"
titrating mRNA, we determined the first-order rate constant (kcat) of ATP hydrolysis at saturating mRNA and ATP

11"
Addition of the PIC to eIF4A•4G•4E in the absence of mRNA increased the kcat 24-fold over the value in the

12"
absence of the PIC ( Figure 2B; compare closed circles and closed squares, respectively, at 0 µM mRNA). This

13"
enhancement could be due to the presence of rRNA and tRNA in the PIC components. However, at saturating

14"
concentrations of mRNA, the PIC components still enhance the kcat for ATP hydrolysis by eIF4A•4G•4E by 3.4-fold

15"
( Figure 2B, compare closed circles to closed squares), indicating that there is an enhancement of ATPase activity not due

16"
to non-specific stimulation by RNA but instead caused by one or more components of the PIC. Leaving out 40S

17"
subunits from the PIC components resulted in a 2-fold lower kcat at saturating mRNA compared to the value observed

18"
in the presence of a complete PIC ( Figure

20"
for the observed stimulation by the PIC or mRNA.

21"
In order to determine which components are responsible for this stimulation of the ATPase activity of

23"
components alone or in combination. We also varied the ATP concentration to determine both the kcat ( Figure 2C) and

26"
tRNAi) -which alone are not sufficient to form the PIC -did not have any additional effect. However, adding the

29"
8" " 1C). Also, omission of eIF4A from an otherwise Complete PIC resulted in a ≥ 67-fold decrease in the rate of ATPase as 1" compared to the Complete PIC (to the limit of detection of the assay), ruling out any significant ATPase contamination 2" in any of the PIC components ( Figure 2C, "-4A").

3"
As with the omission of 40S ribosomal subunits, leaving out eIF2 decreased the kcat by ~2-fold ( Figure 2C;

12"
In contrast to the core 43S PIC components and eIF3, leaving out eIF4B or eIF5 had no effect on the 13" stimulation of the ATPase activity of eIF4A•4G•4E. Neither factor is required for formation of a stable PIC or binding

14"
of eIF3 to the complex, consistent with the proposal that it is the 43S PIC bound to eIF3 that stimulates the ATPase 15" activity.

16"
17" eIF3 subunits g and i, critical for mRNA recruitment, are necessary for ATPase stimulation

18"
S. cerevisiae eIF3 is comprised of 5 core subunits and is involved in numerous steps of translation initiation,

19"
including mRNA recruitment. Previous studies showed that yeast eIF3 is essential for mRNA recruitment both in vitro 20" (Mitchell et al., 2010) and in vivo (Jivotovskaya et al., 2006), and that it stabilizes TC binding and promotes PIC

21"
interactions with the mRNA at both the mRNA entry and exit channels of the 40S subunit (Aitken et al., 2016). In

22"
particular, the eIF3 subunits 3g and 3i have been implicated in scanning and AUG recognition in vivo (Cuchalová et al.,

23"
2010) and are required for recruitment of RPL41A mRNA in vitro (Aitken et al., 2016;Valásek, 2012). Both subunits are 24" thought to be located near the path of the mRNA on the ribosome, at either the solvent or intersubunit face of the 40S

27"
the 3g and 3i subunits -the kcat was 3.9 ± 0.3 min -1 , which is similar to the rate constant observed in the absence of the

2"
activity. One possible scenario is that eIF3g and eIF3i interact with eIF4A near the mRNA entry channel of the 40S

3"
subunit, either directly or indirectly, and this interaction promotes ATP hydrolysis and mRNA recruitment.

5"
The presence of the 43S PIC and eIF3 increases eIF4A ATPase activity in the absence of eIF4G•4E

6"
We next asked if the PIC could stimulate the ATPase activity of eIF4A in the absence of eIF4G•4E. eIF4A on

7"
its own, in the presence of saturating RPL41A mRNA but in the absence of any PIC components or eIF4G•4E, had a 8" kcat of 0.58 ± 0.08 min -1 in these experiments (titrating the concentration of ATP). Addition of the PIC, without 9" eIF4G•4E, resulted in a 6-fold increase in kcat over eIF4A alone ( Figure 2D and

12"
PIC components markedly enhance the activity of eIF4A even when it is not associated with the eIF4G•4E complex.

13"
We next asked which PIC components are critical for this eIF4G•4E-independent stimulatory mechanism.

16"
compare dark blue bar to light blue cross-hatched bars). Congruent with our earlier results, these data suggest that

17"
stimulation of the ATPase activity of eIF4A, in the absence of eIF4G•4E, also requires the complete 43S PIC and eIF3,

18"
in particular subunits 3i and 3g. This result suggests the possibility that eIF4A can interact directly with one or more

19"
components of the PIC -for example, eIF 3i and 3g -independently of interactions that might be mediated by eIF4G;

20"
and its interactions with the PIC or eIF4G•4E confer comparable stimulation of eIF4A's ATPase activity (6.0-fold vs.

23"
Km values suggest distinct mechanisms of eIF4A activation by the PIC components and eIF4G•4E

29"
10" " only eIF4G•4E reduces the Km for ATP, suggesting different mechanisms of stimulation. The effect of eIF4G•4E is 1" consistent with the proposal that eIF4G acts as a "soft clamp" to juxtapose the two eIF4A RecA-like domains to 2" enhance ATP binding and catalysis (Hilbert et al., 2011;Oberer et al., 2005;Schütz et al., 2008). By contrast, the PIC

3"
components apparently act in a manner that enhances the rate-limiting step of ATP hydrolysis without affecting ATP 4" binding.

5"
6" eIF4A relieves inhibition of recruitment produced by structures throughout the length of mRNAs

7"
Having established that ATP hydrolysis by eIF4A accelerates the rate of recruitment of both the natural mRNA

8"
RPL41A and an unstructured 50-mer made up of CAA repeats ( Figure 1) and that intact PICs activate steady-state ATP 9" hydrolysis by eIF4A (Figure 2), we next set out to probe the effects of mRNA structure on the recruitment process and 10" action of eIF4A. To this end, we created a library of in vitro transcribed and individually purified mRNAs spanning a

11"
range of structures and lengths. This library contains model mRNAs comprised almost entirely of CAA repeats,

12"
containing or lacking a 9 base pair (bp) hairpin in the 5'-UTR, and/or a natural RPL41A mRNA sequence downstream

13"
of the AUG in place of CAA repeats ( Figure 3A; Supplementary Methods). The RPL41A sequence is expected to have

14"
structure throughout its length (Figure 1 -figure supplement 1C). We measured the recruitment kinetics for each 15" mRNA in the absence of eIF4A and as a function of eIF4A concentration (see "mRNA Recruitment Assay" in

16"
Methods) and determined the maximal rate (kmax) and concentration of eIF4A required to achieve the half-maximal rate 17"

18"
Using an eIF4A concentration determined to be saturating for all ten mRNAs (Figure 3 -figure supplement 1),

19"
we observed recruitment endpoints between 85%-95% with all mRNAs tested ( Figure 3B, black bars). In the absence of 20" eIF4A, however, the extent of recruitment varied widely among the mRNAs. Less than 10% of RPL41A mRNA was

21"
recruited in reactions lacking eIF4A ( Figure 3B, RNA 10, red bar), consistent with the low levels of RPL41A

22"
recruitment we observed in the absence of ATP ( Figure 1A). Varying the concentration of eIF4A yielded a kmax of 1.3 ±

24"
1B,E). In the absence of eIF4A, time courses with RPL41A mRNA could not be accurately fit with a single-exponential 25" kinetic model due to low reaction endpoints. However, comparison of estimated initial rates (no eIF4A, 0.22 ± 0.07 min -

27"
magnitude more rapidly in the presence of saturating levels of eIF4A versus in the absence of eIF4A ( Figure 3D, RNA

11" "
Consistent with our observation that the unstructured CAA 50-mer mRNA is efficiently recruited even in the 1" absence of ATP, we observed 74 ± 1 % recruitment of this mRNA in the absence of eIF4A ( Figure 3A-B, RNA 1, red 2" bar). Nonetheless, the addition of saturating eIF4A increased the extent of recruitment to 87 ± 1 % ( Figure 3A-B, RNA

3"
1, black bar) consistent with our observation that the addition of ATP and eIF4A slightly elevates the extent of 4" recruitment of this mRNA above the levels observed in the absence of ATP or eIF4A ( Figure 1B). Beyond this modest

9"
To compare the CAA 50-mer with a longer mRNA, we increased the total mRNA length to 250 nucleotides 10" (250-mer) by adding 200 nucleotides of CAA repeats downstream of the AUG ( Figure 3A, RNA 2). In the absence of

12"
comparable to the results seen for the CAA 50-mer in the absence of eIF4A ( Figure 3B, red vs. black bars, RNA 1 vs. 2).

15"
the rate of recruitment of this mRNA, in this case by 13-fold ( Figure 3D, RNA 2). Similar results were obtained for two

16"
additional CAA-repeat 250-mers with the AUG situated 67 or 150 nucleotides from the 5'-end ( Figure 3A, RNAs 3 and 17" 4). These mRNAs had extents of recruitment of 60-70% in the absence of eIF4A, which increased to >90% in the

21"
RNAs 1-4. The reason for these differences is not clear, but does not seem to correlate with overall mRNA length or

22"
number of nucleotides 5' or 3' to the AUG. In summary, all four unstructured CAA-repeat mRNAs that we studied can

23"
be recruited by the PIC at appreciable levels independently of eIF4A, but eIF4A still stimulates their rates of recruitment

24"
by roughly an order of magnitude.

25"
To probe the effects of defined, stable secondary structures on the functioning of eIF4A in mRNA 26" recruitment, we examined 250-mer mRNAs comprising CAA repeats throughout the sequence except for a single 21-nt

29"
12" " insertions of this 21-nt sequence into the 5'-UTR of a luciferase reporter conferred strong inhibition of reporter mRNA 1" translation in yeast cells (Sen et al., 2016). We confirmed the presence and location of the single hairpin in RNAs 5 and 6

2"
and absence of significant secondary structure in RNA 4 by incubating RNAs 4-6 at 26˚C with a 3'-5' RNA exonuclease,

5"
To our surprise, in the absence of eIF4A, neither the cap-proximal nor the cap-distal hairpins significantly

6"
influenced the extent of recruitment, achieving endpoints between 70% and 80%, comparable to the unstructured CAA

11"
both cap-proximal and cap-distal hairpin mRNAs but, surprisingly, to a lesser degree than in the absence of the hairpin:

13"
RNA 4 vs. 5-6). Thus, at odds with the expectation that stable structures in the 5'-UTR would impose strong obstacles

14"
to mRNA recruitment to the PIC, we found that addition of a cap-proximal or cap-distal hairpin in the 5'-UTR of an

15"
otherwise unstructured mRNA confers little or no inhibition of the extent of recruitment and only a modest reduction in

16"
the rate, in the presence or absence of eIF4A. Moreover, the observation that these hairpins in the 5'-UTR actually

17"
decrease the enhancement of the rate of mRNA recruitment provided by eIF4A relative to what we observed with the

18"
unstructured mRNA ( Figure 3D, RNAs 5-6 vs. 4) is not readily consistent with the idea that the factor's predominant

19"
function is to unwind stable secondary structures in the 5'-UTRs of mRNAs to facilitate PIC attachment. If this were the 20" case, one might have expected larger rate enhancements for mRNAs with stable structures in their 5'-UTRs than for 21" mRNAs containing little inherent structure, the opposite of what we actually observe. It is possible that eIF4A is not

22"
efficient at unwinding such stable structures, which results in a lower kmax and a lower degree of stimulation.

23"
To probe further the effects of RNA structural complexity on mRNA recruitment and eIF4A function, we

24"
examined a chimeric mRNA comprising CAA-repeats in the 5'-UTR and the natural sequence (with associated structural 25" complexity) from RPL41A 3' of the AUG start codon ( Figure 3A, RNA 7). In the absence of eIF4A, this mRNA was

26"
recruited to the PIC with an observed rate of 0.06 ± 0.01 min -1 , which is significantly slower than for RNAs 1-6 (~0.2-

27"
0.9 min -1 ), but faster than the rate for full-length RPL41A mRNA, which as noted above could not be determined due

1"
fold of the values for RNAs 1 and 4. Thus, addition of saturating eIF4A conferred a 60-fold increase in the rate of 2" recruitment for RNA 7, a greater degree of stimulation than observed with RNAs 1-6 ( Figure 3D). These results indicate 3" that eIF4A can efficiently resolve inhibition of mRNA recruitment mediated by RNA sequences on the 3' side of the 4" start codon, which are not predicted to form stable secondary structures with the 5'-UTR (Figure 1 -figure supplement

7"
supplement 1E), suggesting that structure on the 3' side of the start codon increases the concentration of eIF4A required 8" to maximally stimulate recruitment.

9"
In contrast to the modest effects of the hairpins when present in the otherwise unstructured RNAs 5-6, both

10"
the cap-proximal and cap-distal hairpins were strongly inhibitory to mRNA recruitment in the absence of eIF4A when

11"
inserted into the unstructured 5'-UTR of the chimeric mRNA harboring RPL41A sequence 3' of the AUG codon,

16"
recruitment and marked dependence on eIF4A in the context of the chimeric mRNA containing native RPL41A

17"
sequences, but not in an otherwise unstructured mRNA. We note however that stimulation of the recruitment rate by

18"
eIF4A was considerably less than the 60-fold and ~190-fold increases observed for the chimeric mRNA lacking a

20"
stimulated maximal rates for mRNAs 8-9 also remain well below those observed for mRNA 7 and RPL41A mRNA 21" ( Figure 3C, RNAs 8-9 vs. 7 and 10). These differences might be explained by the inability of eIF4A to efficiently resolve

22"
these stable secondary structures, as already suggested above.

23"
Comparing kmax values for RNA 4 to RNAs 5-9 shows that the combined effects of structures in the 5'-UTR

26"
codon (RNA 7) on their own have 2-3-fold effects on kmax relative to the unstructured RNA 4, combining them (RNAs

27"
8 and 9) produces a ≥30-fold effect. These data indicate that structures in both regions synergistically inhibit the rate of 28" mRNA recruitment. This synergy is also reflected in the dramatically reduced recruitment endpoints of RNAs 8-9 seen

29"
14" " in the absence of eIF4A ( Figure 3A,B, red bars; RNAs 4-7 vs. 8-9). It is possible that, in the absence of significant RNA 1" structure on the 3' side of the AUG codon, PICs are able to load directly in the unstructured region containing the AUG 2" codon located downstream of the stem loops in RNAs 5 and 6, in a manner only slightly encumbered by the stable

3"
hairpins (Agalarov et al., 2014). When the structured RPL41A mRNA is present beyond the start codon in RNAs 8 and 4" 9, it may make this direct loading impossible, perhaps by sterically occluding or otherwise blocking access to the internal 5" unstructured region, thereby forcing the PIC to attach near the 5'-cap and scan through the structured 5'-UTR to reach 6" the AUG codon. This latter situation likely reflects the state of most (if not all) natural mRNAs, in which PICs do not

7"
have access to large segments of unobstructed, unstructured internal RNA on which to directly load.

8"
The fact that eIF4A restores recruitment of RNAs 8 and 9 ( Figure 3B, RNAs 8-9, black bars), albeit at 9" relatively low recruitment rates, indicates that eIF4A can eventually resolve the synergistic inhibition produced by 10" structures in the 5' and 3' segments of these messages. It is noteworthy, however, that eIF4A gives considerably larger

11"
rate enhancements for the mRNAs harboring only native RPL41A sequences (RNAs 7 and 10) than those burdened

12"
with synthetic hairpins (RNAs 5-6 and 8-9), suggesting that eIF4A is better able to resolve the complex array of relatively 13" less stable structures in RPL41A compared to a highly stable local structure in the 5'-UTR. 14"

16"
To further understand the interplay between eIF4A•4G•4E, the PIC and the mRNA during the recruitment

17"
process, we inquired how the 5'-cap -which binds the heterotrimer via eIF4E -influences the requirement for eIF4A in 18" recruitment of various RNAs. We have previously shown that the 5'-cap enforces the requirement for several eIFs,

19"
including eIF4A, in mRNA recruitment (Mitchell et al., 2010) and more recent work in the mammalian system provided 20" evidence that the 5'-cap-eIF4E-eIF4G-eIF3-40S network of interactions is required to promote mRNA recruitment via

21"
threading of the 5'-end into the 40S entry channel (Kumar et al., 2016). As before, we monitored the kinetics of mRNA

22"
recruitment at various concentrations of eIF4A with capped or uncapped versions of mRNAs described above,

24"
of the AUG; Figure 4 and Figure 4 -figure supplement 1). As summarized in Figure 4B, the kmax observed with

25"
saturating eIF4A was comparable with or without the 5'-cap for RNAs 1 and 7, and was 1.5-fold lower with the cap than

26"
without it for RNA 4. In contrast, in the absence of eIF4A, the rates of recruitment for uncapped versions of the 27" unstructured model mRNAs 1 and 4 were 3.7-and 2.5-fold higher, respectively, than the rates of the corresponding 5'-

29"
15" " effect was even more pronounced for RNA 7, containing natural mRNA sequence 3' of the AUG, which was recruited 1" 15-fold faster when uncapped versus capped in the absence of eIF4A. It is also noteworthy that the rate enhancement 2" provided by eIF4A ( Figure 4B, kmax/k !"# !"!!"#$% ) is larger in all cases for the capped mRNAs than the uncapped mRNAs,

3"
reaching an order of magnitude difference for RNA 7, and this effect is due almost entirely to the reduced rate in the

4"
absence of eIF4A for the capped versus uncapped mRNAs. Taken together, our data indicate that even for short

5"
mRNAs with low structural complexity, the 5'-cap inhibits recruitment in the absence of eIF4A, consistent with our

6"
previous observations with a natural mRNA and our proposal that the cap serves, in part, to enforce use of the canonical

8"
directing the PIC to load at the 5'-end of the mRNA and impeding it from binding directly to downstream, unstructured

14"
In the prevailing model of mRNA recruitment, eIF4F (eIF4A•4G•4E) is localized to the 5'-end of the mRNA

16"
The unwound 5' end of this "activated" mRNA is then accessible for binding by the PIC, which subsequently scans

17"
down the mRNA in search of the start codon. And yet, given the natural propensity of an mRNA to form structure it is

18"
difficult to envision how an mRNA could be unwound by eIF4F and eIF4B, released into the cytoplasm and then bound

19"
by the PIC without the mRNA reforming its structure. Moreover, recent work has demonstrated that yeast eIF4B binds

20"
the ribosome itself  and previous evidence suggested that mammalian eIF4B interacts with rRNA 21" (Methot et al., 1996), thus blurring the lines between the PIC and the activated mRNP composed of mRNA, eIF4F and

22"
eIF4B. In another proposed model, eIF4F and eIF4B could interact with the PIC, forming a "holo-PIC" (Aitken and

23"
Lorsch, 2012) that relaxes the mRNA and attaches to it synchronously. Some support for a model in which eIF4F

24"
interacts with the PIC to promote mRNA recruitment came from hydroxyl radical footprinting experiments that indicate

25"
eIF4G binds to the 40S subunit near the eukaryotic expansion segment 6 of the 18S rRNA (Yu et al., 2011). In addition,

26"
in mammals (but not S. cerevisiae) eIF4G interacts with eIF3, which could serve to bring the eIF4F complex onto the PIC

27"
(des Georges et al., 2015). Our observation that the PIC stimulates the ATPase activity of eIF4A, both in the context of

28"
the eIF4F complex and in the absence of eIF4G•4E, indicates that eIF4A and eIF4F functionally interact with the PIC,

29"
16" " and is consistent with the holo-PIC model. Although our ATPase experiments monitor steady-state kinetics and we are 1" not able to access the pre-steady state regime that would allow us to detect any rapid burst of ATP hydrolysis that might 2" take place during mRNA recruitment, our results still show that association with a complete PIC -in the presence or

3"
absence of mRNA -accelerates ATP hydrolysis by eIF4A, on its own and as part of the eIF4A•4G•4E complex.

4"
Our studies of eIF4A ATPase activity indicated that a 43S PIC bound to eIF3 is necessary for full acceleration

5"
of ATP hydrolysis. The absence of eIFs 2 or 3, or the 40S subunit from the components required for a complete PIC

6"
significantly decreased the rate of ATP hydrolysis, whereas the presence of just the 40S subunit and TC, which alone are 7" insufficient to form a PIC, or eIF3 alone gave no stimulation of ATP hydrolysis beyond that afforded by eIF4A•4G•4E.

8"
An eIF3 subcomplex comprised of subunits 3a, 3b, and 3c but lacking subunits 3g and 3i, resulted in a similar rate 9" decrease as when eIF3 was omitted entirely from the reaction. The 3g and 3i subunits of eIF3 have been implicated in 10" mRNA recruitment and scanning (Aitken et al., 2016;Cuchalová et al., 2010;Valásek, 2012), and structural data suggest 11" that they are located near the mRNA entry channel of the 40S subunit, on either the solvent or intersubunit face (Aylett 12" et al., 2015;des Georges et al., 2015;Llacer et al., 2015). The observation that these eIF3 subunits appear at distinct

13"
locations near the mRNA entry channel in complexes either containing or lacking mRNA has led to the speculation that

14"
they might participate in a large-scale rearrangement of the PIC important for either initial attachment to the mRNA or

23"
slower than necessary to support estimated rates of translation initiation in the range of 10 min -1 in vivo (Palmiter, 1975; 24" Shah et al., 2013;Siwiak and Zielenkiewicz, 2010). The factor is required in reconstituted translation systems for 48S PIC

17" "
have shown that the vast majority of mRNAs in yeast display a similar, strong dependence on eIF4A for efficient 1" translation, whereas mRNAs with long, structured 5'-UTRs generally exhibit a special dependence on the helicase Ded1

2"
in addition to their general requirement for eIF4A (Sen et al., 2015). A recent study using a fluorescence-based

3"
equilibrium binding assay showed that mammalian eIF4A promotes recruitment of both an unstructured model mRNA

4"
and natural globin mRNA, although ATP hydrolysis was only required with the globin message (Sokabe and Fraser,

5"
2017). Thus, there was evidence that eIF4A has a general function in PIC attachment to mRNAs while playing an 6" ancillary role for particular mRNAs burdened with 5'UTR structures. We employed the reconstituted yeast system to test 7" this hypothesis.

8"
We observed that ATP hydrolysis by yeast eIF4A accelerates recruitment of structured as well as unstructured 9" mRNAs, and this acceleration does not correlate with the amount of secondary structure in the 5'-UTR. Taken together

10"
with our observation that the PIC accelerates ATP hydrolysis by eIF4A, one explanation for the ability of eIF4A to

11"
stimulate mRNA recruitment regardless of 5'-UTR structure would be that one function of eIF4A is to modulate the 12"

19"
residues responsible for this activity are preserved in the eukaryotic ribosome. In fact, we recently demonstrated that the 20" equivalent residues in 40S protein uS3 (yeast Rps3) -which is near the 40S latch and interacts with eIF3 -stabilize the

21"
PIC-mRNA interaction (Dong et al., 2017). It is thus possible that ATP hydrolysis by eIF4A, stimulated at least in part 22" by eIF3 subunits, might be employed to remodel this region of the 40S subunit to promote mRNA recruitment by the 23" PIC ( Figure 5A).

26"
We observed that addition of native RPL41A mRNA sequence 3' of the start codon inhibits recruitment of

27"
mRNA to the PIC, even if the 5'-UTR is made up of CAA repeats, and this inhibitory effect is ameliorated by eIF4A.

28"
This result is surprising because in the canonical model of initiation eIF4A functions to resolve structures in the 5'-UTR.

18" "
We also found that, in combination, isolated hairpins in the 5'-UTR and structural complexity 3' of the AUG codon 1" synergistically inhibit mRNA recruitment. Thus, structure created by elements beyond the 5'-UTR strongly influences 2" the rate of mRNA recruitment to the PIC and eIF4A is able to alleviate these inhibitory effects. We envision that the

3"
RPL41A sequences 3' of the AUG introduce an ensemble of relatively weak secondary or tertiary interactions that 4" occlude the 5'-UTR and start codon and thus impede direct PIC attachment and AUG recognition, and that eIF4A can

5"
efficiently resolve these structural impediments. The RPL41A sequences might interact directly with the 5'-UTR or 6" simply create structures that envelop it, sterically occluding access to the 5'-UTR and the start codon. In addition, RNAs 7" longer than their persistence length inherently fold back on themselves (Chen et al., 2012), which could further serve to 8" create a steric block to the 5'-UTR. By binding to single-stranded segments, eIF4A may serve to increase their 9" persistence length, helping to untangle the message. It is intriguing that eIF4A was 15-to 20-fold more effective in

10"
overcoming the inhibitory effect of the RPL41A sequences versus the 5'-UTR hairpins in the RNAs containing one or

11"
the other inhibitory element (RNA 7 vs. RNAs 5 and 6). This could be explained by proposing that eIF4A is more 12" efficient at resolving the relatively weak interactions that contribute to the overall structure of an mRNA compared to

13"
highly-stable local structures. Indeed, only a handful of native mRNAs are known to possess stable hairpin structures in

14"
yeast cells (Rouskin et al., 2014), yet eIF4A is essential for translation of virtually all mRNAs in vivo. Moreover, it was

15"
shown that the DEAD-box RNA helicase Ded1 rather than eIF4A is required to overcome the inhibitory effects of

16"
hairpin insertions on reporter mRNA translation in vivo (Sen et al., 2015). Combined with these previous findings, our

17"
results support a model in which eIF4A acts to disrupt moderately stable, transient interactions throughout an mRNA 18" that sequester the mRNA 5'-UTR or start codon within the overall mRNA structure, whereas Ded1 is more effective in

19"
resolving stable, local secondary structures in the 5'-UTR or the initiation region of an mRNA.

21"
A holistic model for eIF4A function in mRNA recruitment

22"
A possible model for eIF4A function that seems consistent with previous studies and the results presented here

23"
is shown in Figure 5B. This model is not mutually exclusive with that depicted in Figure 5A in which eIF4A modulates

24"
the structure of the 40S ribosomal subunit. In the holistic model, eIF4A -which is present in large excess of ribosomes

25"
in vivo (Firczuk et al., 2013) -binds to an mRNA throughout its length (Lindqvist et al., 2008) and mediates the 26" relaxation of local structures, helping to expose the 5'-cap. The eIF4F complex associates with the cap, the 5'-end of the

27"
mRNA and the PIC. Direct loading of the PIC onto the 5'-UTR near the start codon is not possible because the start

28"
codon is still occluded within the overall structural ensemble of the mRNA and because the cap-eIF4E interaction

29"
19" " directs the PIC to the 5'-end. Interaction of the eIF4F complex with the PIC leads to an acceleration of ATP hydrolysis 1" by the bound eIF4A. This hydrolysis event could mediate opening of the mRNA binding channel, as suggested in Figure   2" 5A, and loading of the 5'-end of the mRNA into it. It would also result in a low affinity, ADP-bound state of eIF4A that

3"
would dissociate from the mRNA, eIF4G and the PIC. During loading of the mRNA, the eIF4E-eIF4G association may

4"
be disrupted, as proposed by Pestova and colleagues (Kumar et al., 2016). Movement of the PIC on the mRNA would 5" then lead to its encountering a new eIF4A molecule, which could then associate with eIF4G. Association of the mRNA-

6"
bound eIF4A with eIF4G and the PIC would increase its ATP hydrolysis activity by 14-fold, which would again

7"
promote mRNA loading into the 40S subunit and dissociation of eIF4A•ADP. This cycle could repeat itself until the 8" start codon is located. The successive cycles of PIC-eIF4G-eIF4A interaction at the entry channel followed by eIF4A

10"
In addition, the cycling of eIF4A molecules into and out of the complex at each step is consistent with previous

11"
observations that an ATPase deficient mutant of eIF4A acts in a dominant-negative fashion to inhibit translation

12"
initiation in vitro (Pause et al., 1994), as the mutation would block dissociation of eIF4A molecules when they encounter 13" eIF4G in the scanning PIC, and thereby impede progression through the 5'-UTR. Although speculative, this model

14"
brings together previous results in the field with our observations that the PIC stimulates ATP hydrolysis by eIF4A and 15" eIF4F, that eIF4A accelerates recruitment of mRNAs regardless of their degree of structure, and that structure 16" throughout the length of mRNAs inhibits recruitment in a manner relieved by eIF4A. It builds upon previous models

18"
Regardless of the model, the events we observe in our mRNA recruitment assay reflect only the first 19" engagement of a PIC with an mRNA. Once multiple ribosomes have been loaded onto a message to form a polysome,

20"
the structure of the mRNA presumably changes dramatically, and structures on the 3' side of the start codon would play

21"
a different role than they do in early rounds of initiation. Future studies of initiation events on polysomal mRNA might

22"
reveal interesting differences from the starting phase of translation on a message.

12"
described previously . tRNAi was charged with methionine as described previously (Walker and 13" Fredrick, 2008). Following charging, Met-tRNAi was separated from contaminating ATP and other nucleotides (left over

14"
from the charging reaction) on a 5 mL General Electric (GE) desalting column equilibrated in 30 mM Sodium Acetate

15"
(NaOAc), pH 5.5. This step was essential in order to measure the ATP dependence of mRNA recruitment and to

16"
accurately control the concentration of ATP in experiments. The Met-tRNAi and free nucleotide peaks were confirmed

17"
with individual standards prepared identically to a charging reaction. Eluted Met-tRNAi was precipitated with 3 volumes

24"
with all other conditions identical. Reactions were incubated at 37 °C for 90 minutes and purified using the RNeasy

6"
was then added to the reaction and incubated an additional 7 minutes. The remainder of the components, except mRNA

7"
and ATP, were added to the TC and incubated for an additional 10 minutes to allow complex formation. Reactions were 8" initiated with a mix containing final concentrations of 15 nM mRNA and ATP•Mg 2+ . Experiments varying the 9" concentration of eIF4A were carried out in the presence of 5 mM ATP. Experiments varying ATP were carried out in

11"
blue and xylene cyanol dye in 40% sucrose containing a final concentration of a 25-fold excess of unlabeled mRNA

12"
(cold chase), identical to the labeled mRNA for that reaction. Two µl of the chased reaction were immediately loaded

13"
and resolved on a native 4% (37.5:1) polyacrylamide gel using a Hoefer SE260 Mighty Small II Deluxe Mini Vertical

14"
Electrophoresis Unit at a potential of 200 volts for 50 minutes. The electrophoresis unit was cooled to 22°C by a

16"
MgCl2 ("THEM"). Gels were exposed to a phosphor plate overnight at -20°C, the plates visualized on a GE Typhoon

17"
9500 FLA, and the fraction of recruited mRNA bands (48S complex) versus the total signal in the lane was quantified

18"
using ImageQuant software. Data were plotted and fit using KaleidaGraph 4.5 software. Recruitment time courses were

19"
fit to a single exponential rate equation: y = A*(1-exp(-kobs*t)), where t is time, A is amplitude, and kobs is the observed 20" rate constant. Observed rates were plotted against the concentration of the titrant and fit to a hyperbolic equation: y = 21" b+((kmax*x)/(K1/2+x)) where x is the concentration of the titrant, kmax is the maximal observed rate of mRNA

22"
recruitment when the reaction is saturated by the factor titrated (e.g. eIF4A), K1/2 is the concentration of the factor 23" required to achieve ½Vmax, and b is the rate in the absence of the titrated factor (i.e., the y-intercept).

26"
The NADH-coupled ATPase assay was adapted from previously described methods with some modifications 27" (Bradley and De La Cruz, 2012;Kiianitsa et al., 2003). All ATPase experiments were carried out in 384-well Corning

28"
3544 plates on a Tecan Infinite M1000PRO microplate reader at 26°C. Using a standard curve we determined that a 10 29"

22" "
µL reaction with 1 mM NADH on a Corning 3544 microplate gives an absorbance of 1.23 Optical Density of 340 nm 1" light (OD340) in the microplate reader. OD340 was measured every 20 seconds for 40 minutes, plotted vs. time for

2"
individual reactions, and fit to y = mx + b where m is the slope, x is time in minutes, and b is the y-intercept. Thus, m is

5"
Note that the absolute value of m (|m|) was used because the slope is a negative value due to loss of absorbance over

12"
incubations and experiments were performed at 26°C. PICs were formed at 2x of the final concentration in 1x Recon

13"
buffer, in the absence of mRNA and ATP. PIC formation was initiated by incubating eIF2 and GDPNP for 10 minutes,

15"
4B, 4G•4E, and RiboLock were added. Order of addition of eIFs did not make a difference. The reaction was incubated

16"
for 10 minutes to allow complex formation. Subsequently, the PICs were combined with the "Reporter Mix" containing 17" phosphoenolpyruvate, NADH, pyruvate kinase, and lactate dehydrogenase (added as a 10x stock of the final

18"
concentrations all in 1x Recon buffer) resulting in concentrations in the final reaction of 2.5 mM phosphoenolpyruvate,

20"
1400 units/mL) mix (PK/LDH mix). Reactions were brought up to volume with 1x Recon buffer, such that when they

21"
were initiated by addition of mRNA (added as a 10x stock of the final concentration in 1 x Recon) and ATP•Mg 2+

22"
(added as a 4x stock of the final concentration in 1x Recon buffer) the total reaction volume was 12 µl. 10 µl of the 23" reaction were then immediately transferred to the microplate for analysis by the Tecan plate reader and changes in

24"
absorbance of 340 nm light were monitored over time, taking readings every 20 seconds for 40 minutes. Because the 25" assay measures the slope of a straight line (i.e. multiple turnover conditions), the lag in time between initiating the 26" reaction and start of the measurements by the plate reader has no effect on the results. Initiating the reactions using an

27"
injection, capable of monitoring rapid kinetics did not reveal any differences in results; i.e., there was no evidence of a 28"

23" "
"burst" phase in the initial part of the reaction. Increasing or decreasing the concentration of the PK/LDH mix by 3-

1"
fold did not influence the observed rate of ATP hydrolysis (Figure 2 -figure supplement 1A), indicating that the rate of

2"
NADH oxidation is not limited by PK/LDH activity. Also, when ATP, eIF4A, or PK/LDH was absent from the 3" reaction, there was no change in absorbance at 340 nm over 1 hour (Figure 2 -figure supplement 1A).

6"
In a 20 µl reaction 0.5 pmol/µl of RNA was incubated with 0.75 U/µl of RNase Exonuclease T in 1x NEB buffer 4 at 7" 26°C for 18 hours. RNA (4 pmol total RNA per lane) was loaded and resolved on a Novex 15% Tris Borate EDTA

8"
Urea gel and stained with SYBR Gold nucleic acid gel stain, diluted 1/10,000, for 5 minutes and visualized on a General

12" 13"
The authors would like to acknowledge Tom Dever and Nicholas Guydosh for their thoughtful and critical suggestions.

14"
This work was supported by the Intramural Research Program (JRL and AGH) of the National Institutes of Health

18"
The authors declare no competing interests.
Solid black circles: "Complete PIC," as in Figure 2B (k cat of 8 ± 1 min -1 and ! ! !"# of 120 ± 40 µM). (2) (3) (8) (1) (3) (8) (3) (    Observed rates of mRNA recruitment (min -1 ) when eIF4A was not included in the reaction, in the presence (black bars) or absence (red bars) of the 5'-cap. (See Figure 3A Key for explanation of mRNA diagrams). (B) The observed rates of recruitment in the absence of eIF4A (k !"# !"!!"#$% ), k max , fold enhancement by eIF4A (k max /k !"# !"!!"#$% ), and fold difference in the absence of eIF4A (k !"#!!!"#) !"!!"#$% !/!k !"#!!!"# !"!!"#$% ). The data for the capped mRNAs are reproduced from Figure  in the presence or absence of a 5'-cap, as indicated to the right of each curve. Numbers to the right of the plots correspond to RNAs in Figure 3A. Insets show the observed rates versus eIF4A concentration up to 0.5 µM to focus on the 0 µM eIF4A points. All data presented in the figure are mean values (n ≥ 2) and error bars represent average deviation of the mean. mRNAs have complex, often dynamic, structures due to numerous local and distant interactions (red dotted lines) as well as the inherent tendency of polymers longer than their persistence lengths to fold back on themselves. These complex global structures can occlude the 5'-ends of the mRNAs and the start codon, making it difficult for the PIC to bind and locate them. Individual eIF4A molecules might keep the mRNA in a partially-unwound state so that the 5'-end can be located by eIF4F and the PIC. ATP hydrolysis by the eIF4A molecule bound to eIF4F allows it to load the 5'-end of the mRNA into the entry channel of the 40S subunit. Loading might involve opening of the mRNA entry channel or another conformational change in the PIC and active transfer of the mRNA into the channel. After loading of the 5'-end, eIF4A•ADP dissociates from the complex. This allows the next (3') eIF4A molecule on the mRNA to bind to eIF4G, which activates its ATPase, enabling loading of that mRNA segment and thus movement of the PIC down the message. This cycle repeats until the start codon is located. The two models presented in (A) and (B) are not mutually exclusive and it is possible that eIF4A, given its high cellular concentration, performs multiple tasks.