Transcriptome-wide analysis of the function of Ded1 in translation preinitiation complex assembly in a reconstituted in vitro system

  1. Section on the Mechanism and Regulation of Protein Synthesis Eunice Kennedy Shriver National Institute of Child Health and Human Development National Institutes of Health Bethesda, MD USA
  2. Section on Nutrient Control of Gene Expression Eunice Kennedy Shriver National Institute of Child Health and Human Development National Institutes of Health Bethesda, MD USA


  • Reviewing Editor
    Nahum Sonenberg
    McGill University, Montreal, Canada
  • Senior Editor
    David Ron
    University of Cambridge, Cambridge, United Kingdom

Reviewer #1 (Public Review):

The authors have developed and optimized a footprinting assay to monitor the recruitment of mRNAs to a reconstituted translation initiation system. This assay is named Recruitment-Sequencing (Rec-Seq) and enables the analysis of many purified mRNAs in the reconstituted system.

This system possesses the ability to determine how competition occurs between mRNAs for the initiation machinery. This is the first approach using a reconstituted system that enables this important feature, and this is an important advance for the field.

Using purified mRNAs in a fully reconstituted system together with the ability to monitor start site selection is an important advance. The method enables one to observe for the first time how competition between mRNAs is altered in response to the absence or presence of different initiation components or accessory proteins.

Start site fidelity in purified reconstituted systems can be altered in different buffer conditions and by the concentration of various initiation factors involved in start site fidelity. Future experiments will reveal how these variables can regulate start site selection in this powerful system.

Comments on revised version:

The authors have addressed all of my original comments. This is an impressive manuscript.

Reviewer #2 (Public Review):


Zhou et al report development of a new method, Rec-Seq, that allows rigorous quantitation of the efficiency of 48S ribosomal pre-initiation complex (PIC) formation on messenger RNAs at transcriptome scale in vitro. With a next-generation deep-sequencing approach, Rec-Seq allows precisely targeted dissection of the roles of translation initiation factors in PIC assembly. This level of molecular precision is important to understanding mechanisms of translational control, making Rec-Seq a significant methodological advance. The authors leverage Rec-Seq to investigate the relative roles of two key helicase enzymes, Ded1p and eIF4A. While past work has pointed to differing roles for Ded1p and eIF4A helicase activity in PIC assembly, unambiguous interpretation of prior in-vivo data has been hindered by technical requirements for performing the experiments in cells. Rec-Seq circumvents these challenges, providing robust mechanistic insights. The authors find that Ded1p stimulates PIC formation selectively on mRNAs with long, structured leaders in the Rec-Seq system, while eIF4A provides much more general stimulation across mRNAs. The findings substantiate the past in-vivo results, along with adding new insights. They contrast with evidence that Ded1p promotes translation by suppressing inhibitory upstream initiation through structural remodeling, or through formation of intracellular, phase-separated granules. The conclusions of the study are well-supported by the data, and are likely to be of broad interest.


The quantitative nature of Rec-Seq, which uses an internal standard to measure absolute recruitment efficiencies, is an important strength.

The methodology decisively overcomes past experimental limitations, allowing the authors to make clear conclusions with regard to the relative roles of Ded1p and eIF4A in PIC formation. An important and useful addition to the toolbox for studying translation and translational control mechanisms, Rec-Seq substantially expands the throughput and scope of mechanistic analyses for translation initiation.

One significant finding to emerge is that the in-vitro reconstituted system used here recapitulates effects of in-vivo perturbations of translation initiation. Despite the lack of a cellular environment and its components, PIC formation appears to operate much as it does in the cell. Importantly, this highlights an inherent "modularity" to the system that is especially of interest in the context of how regulatory machinery beyond the PIC may control translation.


The study finds that Ded1p stimulates accumulation of PICs at internal AUG codons, i.e., within mRNA coding sequences, at an incidence of up to ~50% - thus, bypassing "canonical" translation start sites. Understanding the physiological significance of this activity will require further study. The authors address this in the text.

A limitation of the methodology is that, as an endpoint assay, Rec-Seq does not readily decouple effects of Ded1p on PIC-mRNA loading from those on the subsequent scanning step where the PIC locates the start codon. Considering that Ded1p activity may influence each of these initiation steps through distinct mechanisms - i.e., binding to the mRNA cap-recognition factor eIF4F, or direct mRNA interaction outside eIF4F - additional studies will be needed to gain deeper mechanistic insights. The authors discuss this in the text.

Comments on revised version:

In revising their manuscript, the authors have responded very thoughtfully and insightfully to the initial review. The final manuscript is an important contribution to the field, and I am sure it will be of broad interest.

Author Response

The following is the authors’ response to the original reviews.

Reviewer #1


Start site fidelity in purified recons5tuted systems can be drama5cally altered in different buffer condi5ons. Interpreta5on of the observed changes to start site selec5on in mRNAs in the absence or presence of Ded1 using only the one buffer condi5on used is therefore limited.

This is an excellent point and is something we could explore in future studies using the Rec-Seq system. We have added this caveat to the Discussion on lines 797-809. We have previously studied the fidelity of start codon recogni>on in the recons>tuted system (Kolitz et al., [2009] RNA, 15:138-152) and found that under our standard buffer condi>ons the codon specificity generally reflects what we observed in vivo using a dual-luciferase reporter assay, with the most stable 48S complexes forming on AUG codons, followed by first posi>on mismatches (GUG, UUG, CUG), with second and third posi>on mismatches leading to significantly less stable complexes. However, as the reviewer notes, there are some devia>ons: ACG and AUA are poor codons in the in vitro system under the buffer condi>ons used but allowed rela>vely strong expression in our in vivo reporter assay. It should also be noted that the hierarchy of nearcognate start codon usage in vivo in yeast differs according to the study and the reporter used, making it difficult to establish a “ground truth” for start codon fidelity.

I have some specific comments to strengthen the manuscript and address some minor issues.

It is not clear to me whether the authors refold the purified mRNA aEer phenol/chloroform extrac5on? Have the authors observed different results if the mRNA is refolded or not? This is appropriate since the authors compare their Rec-Seq data to PARS scores that were generated from refolded mRNAs. One assumes that the total mRNA used is refolded in the same way as the PARS score study, but this is not clearly stated. The authors should make this point clear in the text and methods.

This is an excellent point. We did not use the final refolding protocol that Kertesz et al. used when they developed their PARS scores and now clarify this in the Methods sec>on (lines 962967). It is possible that we would have seen stronger correla>ons in the analyses using PARS scores had we followed the renatura>on protocol, although the fact that we observed significant correla>ons (e.g., Fig. 3E-H) suggests the structures in the Kertesz et al. mRNAs were similar to those in our mRNAs.

It is not clear how the authors determine the concentra5on of total mRNA that is used in the assay - reported as 60 nM? Are the authors assuming a molecular weight of an average mRNA to determine the concentra5on? The authors should provide more detail for how they quan5fy their mRNA concentra5on and its stoichiometry compared to 43S PICs.

We thank the reviewer for poin>ng out this oversight and have now included this informa>on on lines 849-855 of the Methods sec>on.

Comments regarding start site fidelity in the recons5tuted system:

The authors use in vitro transcribed tRNAi-Met. Since tRNA modifica5ons may play a role in start site fidelity, the authors should perhaps men5on that this will need to be inves5gated in a future study in the discussion.

This is a good point and we now note it as a caveat in the Discussion on lines 806-809.

The authors state that Ded1 promotes leaky scanning regardless of the mAUG start site context (page 24; lines 533-534). The authors then state on page 25 that the level of iAUG ini5a5on rela5ve to mAUG ini5a5on does depend on the mAUG context (lines 545-546). This seems contradictory unless I am not understanding this correctly? It would certainly be surprising that mAUG context didn't regulate leaky scanning in the recons5tuted system given the fact that ini5a5on codon context regulates selec5on in cells (when Ded1 is present).

These statements are correct as wrihen. As shown in Figure 5O, the frequency of leaky scanning (as measured by rela>ve ribosome occupancy of the internal region of the ORF, not including the main start codon, to the whole ORF, including the main start codon; RRO) decreases as the context score around the start codon gets stronger (green and purple lines). The RRO is increased to the same extent when 500 nM Ded1 is added, regardless of the strength of the start codon context, indica>ng that Ded1 enhances leaky scanning equally (compare slopes of the green line without Ded1 to the purple line with Ded1). Because of this, the effect of Ded1 on RRO (DRR0) is constant across context score bins (orange line). There is no discrepancy between our two conclusions that leaky scanning of the mAUG increases as context score decreases and that Ded1 increases leaky scanning equally for good and bad mAUG contexts, indica>ng that Ded1 does not inspect the mAUG context and simply decreases the dwell >me equally at all contexts.

Further to the start site context ques5on. It is possible that the fidelity of the recons5tuted system (i.e. buffer condi5ons) is not fully reflec5ng in vivo-like start site selec5on. A rigorous characteriza5on of commercially available re5culocyte lysate systems iden5fied buffer condi5ons that provided similar start site fidelity to that observed in live cells (Kozak. Nucleic Acids Res. 1990 May 11;18(9):2828). While I feel that it is beyond the context of the current work to undertake a similar rigorous buffer characteriza5on, one must be careful about interpre5ng the results about leaky scanning and upstream ini5a5on sites in the current work. Perhaps one would observe similar results to Guenther et al. if the fidelity (buffer condi5ons) of the recons5tuted system were different? I appreciate that the authors state that their results only apply to their recons5tuted system and do not necessarily suggest that previous data are incorrect, but with only one buffer condi5on being tested in the current study it may be appropriate to further soEen the interpreta5on of the current results when compared to published data in live cells.

This point is well-taken. As noted above, we have added a caveat about possible effects of buffer condi>ons on start codon fidelity to the Discussion (lines 797-809). In terms of the possibility that upstream ini>a>on is more frequent in vivo than we observe in the in vitro RecSeq system, we previously studied 5’UTR transla>on in vivo using ribosome profiling (Kulkarni et al. [2019] BMC Biol., 17:101). The ra>o of RPFs in 5’UTRs to coding sequences in this study was 0.0027, very similar to the value measured in the in vitro Rec-Seq system in the presence of Ded1 (0.0016-0.0017). Thus, it does not seem that the frequency of upstream ini>a>on is drama>cally higher in vivo than in our in vitro system. We have now made note of this point in the Results (lines 594-598). Guenther et al. employed a ribosome profiling protocol in which they added cycloheximide to their cells prior to lysis, which has been shown to create significant ar>facts, par>cularly in 5’UTR transla>on (e.g., Gerashchenko and Gladyshev [2014] Nucleic Acids Res., 42:e134). Nevertheless, as suggested by the reviewer, we have modified the text in the Results and Discussion to somen the interpreta>on somewhat (lines 582-583; 616-618; 761763).

Reviewer #2


Several findings in this report are quite surprising and may require addi5onal work to fully interpret. Primary among these is the finding that Ded1p s5mulates accumula5on of PICs at internal site in mRNA coding sequences at an incidence of up to ~50%. The physiological relevance of this is unclear.

We agree with the reviewer that understanding the physiological significance, if any, of the apparent leaky scanning of main AUG start codons induced by Ded1 is an unanswered ques>on that will require addi>onal studies. It is possible that rapid 60S subunit joining and forma>on of the 80S ini>a>on complex amer start codon recogni>on on most mRNAs reduces the leaky scanning effect in vivo. We now bring up this possibility in the Discussion sec>on (lines 804809). However, as noted in lines 568-580, mRNAs that display significantly decreased mRPFs at 500 nM Ded1 in the Rec-Seq system also tend to have TEs that are increased in the ded1-cs- mutant rela>ve to WT yeast in in vivo ribosome profiling experiments, sugges>ng that Ded1 ac>vity also diminishes ini>a>on on mAUG codons in these mRNAs in vivo.

A limita5on of the methodology is that, as an endpoint assay, Rec-Seq does not readily decouple effects of Ded1p on PIC-mRNA loading from those on the subsequent scanning step where the PIC locates the start codon. Considering that Ded1p ac5vity may influence each of these ini5a5on steps through dis5nct mechanisms - i.e., binding to the mRNA cap-recogni5on factor eIF4F, or direct mRNA interac5on outside eIF4F - addi5onal studies may be needed to gain deeper mechanis5c insights.

We agree that this is a limita>on of the Rec-Seq assay and now men>on this point in the Discussion sec>on (lines 810-817). It is possible that future work using cross-linking agents to stabilize 43S complexes bound near the cap and scanning the 5’UTR, similar to the methodology used in 40S ribosome profiling, could enable us or others to disentangle these steps from one another.

As the authors note, the achievable Ded1p concentra5ons in Rec-Seq may mask poten5al effects of Ded1p-based granule forma5on on transla5on ini5a5on. Addi5onal factors present in the cell could poten5ally also promote this mechanism. Consequently, the results do not fully rule out granule forma5on as a poten5al parallel Ded1p-mediated transla5on-inhibitory mechanism in cells.

We agree. As stated in the Discussion sec>on (lines 735-741): “It is possible that at higher concentra>ons of Ded1 than were achievable in these in vitro experiments or in the presence of addi>onal factors that modify Ded1’s ATPase or RNA binding ac>vi>es the factor could directly inhibit a subset of mRNAs, by ac>ng as an mRNA clamp that impedes scanning by the PIC, or by sequestering the mRNAs in insoluble condensates. It might be interes>ng in the future to test candidate factors in Rec-Seq to determine if they switch Ded1 from being a s>mulatory helicase to an inhibitory mRNA clamp that removes transcripts from the soluble phase.”

It is certainly clear why the 15-minute 5mepoint was chosen for these assays. However, I wondered whether data from an earlier 5mepoint would provide useful informa5on. The descrip5on on line 210 of the compiled PDF suggests data from different 5mepoints may be available; if it is, in my view it could be a useful addi5on. More generally, including language about the single-turnover nature of these reac5ons may be helpful for the benefit of a broad audience.

In preliminary experiments, we have used the Rec-Seq system to measure the kine>cs of 48S PIC forma>on transcriptome-wide. As you probably can imagine, this is a challenging experiment and requires addi>onal work before we would feel comfortable publishing it. We very much agree with the reviewer that resolving the kine>cs of these events will provide important addi>onal informa>on. As suggested, we have added caveats about the endpoint and single-turnover nature of the assay to the Discussion (lines 821-828).

I wondered whether it might be useful to present addi5onal informa5on on the mRNAs not found in the assay. For example, are these the least abundant mRNAs, which may not have had 5me to recruit the 43S PIC?

75% of mRNAs (2719 of 3640) not observed in the Rec-Seq analysis had densi>es below the median (2.3 reads per nucleo>de). We now men>on this in the Methods sec>on (lines 855856).

The Rec-Seq recruitment reac5ons were carried out at 22C˚ . Considering that remodeling of RNA structure by helicase enzymes is a focal point of the study, linking the results to the recruitment landscape at a closer-to-physiological temperature may bolster the conclusions.

In the future, it would be interes>ng to test the effects of temperature on 48S PIC forma>on using the Rec-Seq system. As the reviewer suggests, the interplay between temperature and mRNA structure could reveal interes>ng phenomenon. It is worth no>ng, however, that there is no clear “physiological” temperature for S. cerevisiae. For consistency and convenience, lab yeast is usually grown at 30 ˚C, but in the wild yeast live at a wide range of temperatures, which generally change throughout the day. From this standpoint, 22 ˚C seems reasonably physiological.

Results from Rec-seq experiments conducted at 15° C might be more directly comparable to in vivo Ribo-seq data with the ded1-cs mutant. However, already ~90% of the Ded1hyperdependent mRNAs iden>fied by Ribo-seq analysis of that mutant were iden>fied here as Ded1-s>mulated mRNAs in Rec-Seq experiments at 22°C. The Ribo-seq experiments conducted by Guenther et al. were conducted on the ded1-ts mutant at 37°C; thus, any structures that confer Ded1-dependent leaky-scanning through uORFs detected in that study should have been stable in our Rec-Seq experiments.

The introduc5on provides an important, detailed exposi5on of the state of the field with respect to Ded1p ac5vity. Nevertheless, in my view, it is quite lengthy and could be streamlined for clarity. As just one example, the proposed func5on of Ded1p in the nucleus seems like a detail that could be dispensed with for the present work.

We have ahempted to shorten the Introduc>on, as suggested. However, we did not remove the short sec>on describing Ded1’s possible roles in the nucleus and ribosome biogenesis because we felt it was important to emphasize that one of the strengths of the Rec-Seq system is that it allows us to isolate the early steps of transla>on ini>a>on from later steps and from other cellular processes. In addi>on, at the sugges>on of Reviewer #3, we added a brief explana>on of Ded1’s possible role in the subunit joining step of transla>on.

Reviewer #3


The slow nature of the biochemical experiments could bias results.

We agree that the 15-minute >me point used could mask effects that are manifested at a purely kine>c level. It should be noted that we have measured the observed rate constants for 48S forma>on on a variety of mRNAs in the in vitro recons>tuted system in the presence of satura>ng Ded1 (Gupta et al. [2018] eLife, hhps://>cles/38892 ) and found that they are generally in the range of es>mates of rate constants for transla>on ini>a>on in vivo in yeast (~1-10 min-1; e.g., Siwiak and Zielenkiewicz [2010], PLOS Comput. Biol., 6: e100865). In preliminary experiments, we have used the Rec-Seq system to measure the kine>cs of 48S PIC forma>on transcriptome-wide in the absence of Ded1 and find that the mean rate constant observed (~2 min-1) is also within the range of es>mates of the rate of transla>on ini>a>on in vivo in yeast. We hope to publish this analysis in a future manuscript.

It has been suggested that Ded1 and its human homolog DDX3X could play a role in subunit joining postscanning (Wang et al. 2022, Cell and Geissler et al. 2012 Nucleic Acids Res). Could the authors poten5ally inves5gate this by adding GTP, eIF5B and 60S subunits into the reac5on mixture and isola5ng 80S complexes?

This is a very interes>ng sugges>on. One of our plans with the Rec-Seq system is to see if we can also observe 80S forma>on with it and dis>nguish 80S from 48S complexes. Although we haven’t yet tried this and there might be technical obstacles to doing it, if it works we would like to examine the poten>al effects of Ded1, as suggested. We now men>on this possibility in the Discussion sec>on (lines 709-716 and 810-817).

An incuba5on 5me of 15 minutes is quite long on the 5mescale of transla5on ini5a5on. Presumably, the compe55on for 40S among mRNAs is par5ally kine5cally controlled so it would be interes5ng if the authors could do a 5me series on the incuba5on 5me. Does Ded1 increase ini5a5on on more structured UTRs even at shorter incuba5ons or are those only observed with longer incuba5ons?

We agree. See the response to the ques5on about kine5cs above.

Does GDPNP lead to off-pathway events? What happens when GTP is used in the TC? Presumably in the absence of eIF5B the 48S PIC should remain stalled at the start codon.

In previous experiments in the recons>tuted system, we showed that using GTP instead of GDPNP resulted in 48S complexes that were less stable than those stalled prior to GTP hydrolysis (e.g., Algire et al. [2002] RNA 8:382-397). This is presumably because eIF2•GDP and eIF5 release from the complex and the Met-tRNAi can dissociate in the absence of subunit joining. Although we haven’t tried it in the Rec-Seq system, we suspect that the resul>ng PICs would fall apart during sucrose gradient sedimenta>on.

The authors use assembly of a 48S PIC at the start codon as evidence of scanning but could use more evidence to back this claim up. Does removing the cap structure on the two luciferase mRNA controls disrupt ini5a5on using this approach? That would be direct evidence of 5' end 40S loading and scanning to the start codon.

In previous work using the recons>tuted system, we studied the effect of the 5’-cap on 48S PIC forma>on (Mitchell et al. [2010] Mol. Cell 39:950-962; Yourik et al. [2017] eLife hhps://>cles/31476 ). We found that stable 48S PIC forma>on is strongly dependent on the presence of the 5’-cap. In addi>on, the cap prevents off-pathway events and enforces a requirement for the full set of ini>a>on factors to achieve efficient 48S PIC forma>on. As the reviewer indicates, the cap-dependence of the system supports the conclusion that 5’end loading and scanning take place. We have now added this informa>on and the relevant cita>ons to the Introduc>on (lines 147-153). We thank the reviewer for poin>ng out this oversight. It should also be noted that the cases of mRNAs in which 5’UTR transla>on is increased by addi>on of Ded1 support the conclusion that the factor promotes ahachment of the PIC to the 5’ ends of mRNAs and subsequent 5’ to 3’ scanning, as noted in lines 608-618.

The authors state that "The correla5on between CDS length and RE could be indirect because CDS length also correlates with 5'UTR length". Could the authors bin the transcripts into different 5' UTR length ranges and then probe for CDS length differences on RE for each 5' UTR length bin? This could be useful to truly parse the mechanism by which CDS length is influencing RE.

This was an excellent sugges>on. We now include this analysis in a new supplementary figure, Figure 3S-2. Corresponding text was added in lines 380-387:

“Importantly, correlations between Ded1 stimulation and 5’ UTR lengths are evident for all three groups of mRNAs containing distinct ranges of CDS lengths (Fig. 3-S2A-C). In contrast, a marked correlation between Ded1 stimulation and CDS length was detected only for the group of mRNAs with longest 5’UTRs (Fig. 3-S2D-F), and only the latter group showed a clear correlation between 5’UTR length and CDS length (Fig. 3-S2G-I). Thus, the correlation between Ded1 stimulation and CDS length appears to be indirect, driven by the tendency for the mRNAs with the longest 5’UTRs to also have correspondingly longer CDSs.”

We thank the reviewer for this very useful idea.

In Figure 3I, why does RE dip for the middle bins of CDS length in both 100 nM and 500 nM condi5ons, and then rise back up for the later bins? In other words, why do the shortest and longest CDS have the best RE in the presence of ded1?

We do not know the reason for this dip and now say this in the Results on lines 377-378.

The discussion sec5on would be well served to discuss proposed roles of Ded1 post-scanning and how those fit, if at all, with the data presented throughout the manuscript.

We have now added this to the Discussion (lines 709-716 and 810-817). We thank the reviewer for poin>ng out this oversight.

Minor comments:

  • Define bins on figures rather than using bin number for axis labels. For example, Figure 3A-D x-axis labels indicate the length range of each bin.

Thank you for the sugges>on. We have made this change.

  • Figure 3I: the data seem to indicate that shortest CDSs have a ded1 dependency similar to the longest CDSs. This result seems inconsistent with the given rela5onship between UTR length, structure, CDS length. Please clarify.

See answer to this ques>on above.

  • Replace qualita5ve statements, such as "substan5ally smaller reduc5ons" with percent change, numbers, etc.

We have tried to replace qualita>ve statements with quan>ta>ve ones, where possible.

  1. Howard Hughes Medical Institute
  2. Wellcome Trust
  3. Max-Planck-Gesellschaft
  4. Knut and Alice Wallenberg Foundation