A threonyl-tRNA synthetase-mediated translation initiation machinery

A fundamental question in biology is how vertebrates evolved and differ from invertebrates, and little is known about differences in the regulation of translation in the two systems. Herein, we identify a threonyl-tRNA synthetase (TRS)-mediated translation initiation machinery that specifically interacts with eIF4E homologous protein, and forms machinery that is structurally analogous to the eIF4F-mediated translation initiation machinery via the recruitment of other translation initiation components. Biochemical and RNA immunoprecipitation analyses coupled to sequencing suggest that this machinery emerged as a gain-of-function event in the vertebrate lineage, and it positively regulates the translation of mRNAs required for vertebrate development. Collectively, our findings demonstrate that TRS evolved to regulate vertebrate translation initiation via its dual role as a scaffold for the assembly of initiation components and as a selector of target mRNAs. This work highlights the functional significance of aminoacyl-tRNA synthetases in the emergence and control of higher order organisms.

• The authors need to tone down their claim about invertebrate-to-vertebrate transition. I would suggest that this point can be emphasized in 'Discussion' and not in 'Abstract'.
• Page 6, para 3: What is meant by "dorsal and lateral surfaces of eIF4E"? It is better to explain these terms (preferably by a supplementary fig, if possible) for the convenience of the general reader.
• Pages 6/7, paras 3/1: "However, TRS does not appear…immediately after Asp78 (Fig. 1f,g )." This point is not clear! It is better to elaborate on this point a bit. How does TGS domain beginning immediately after Asp78 not lead to TRS not utilizing the non-canonical motifs and auxiliary sequences? The reference of Fig 1f,g does not seem to support/clarify this point at all. Moreover, it is not clear whether TRS possesses non-canonical motifs and auxiliary sequences.
• Page 9, para 2: "Remarkably, most genes…development of nervous, skeletal, and circulation systems…" Nervous and circulatory systems are not unique to vertebrates! Thus, the sentence needs to be modified.
• Page 11, para 2: "Silencing of TRS…by oxygen tension ( Supplementary Fig. 4e)." What is the relevance of this aspect in the context of the study?
• Page 12, para 1: "Tube formation and…TRS (M60K)-expressing cells ( Fig. 5d and Supplementary Fig.  5d)." Enhancement in the case of M60K is seen for both tube length and cell migration, but to a lesser extent than wild-type. Hence, the statement is incorrect and needs to be modified accordingly.
• Page 16, para 2: "During evolution,…complexity of higher organisms." Whether increasing complexity resulted in the acquisition of additional functionalities or the latter resulted in increased complexity? This is a contentious and debatable statement, and therefore needs to be modified accordingly.
• Page 17, para 4, Supplementary fig 9: Since this is a dataset generated in the present work, it should be reported in the 'Results' section, and not merely in the 'Discussion' section.
• Crystallographic table: round off the unit cell dimensions to 2 decimal places; mention the number of unique reflections in the highest resolution shell; mention the resolution range for the highest resolution shell used in refinement; mention the number of reflections used in refinement; mention avg B-factor for protein, ligand and water separately; change "Geometry (%)" to "Ramachandran statistics (%)"; mention what % of reflections were used for Rfree calculation during refinement. The authors present a study that proposes a new function of an aminoacyl-tRNA synthetase in stimulating cap-dependent translation of VEGF. Overall, this is an interesting study and has the potential to transform our understanding of how selective translation of VEGF can be regulated by 4EHP. A strength of the study is the structural model that reveals how 4EHP binds to TRS. The validation of this interaction using structure guided mutants is impressive (although mutations in zebrafish TRS did not clearly result in loss of binding (Fig. 3d)).
Several weaknesses of the study lowers enthusiasm regarding the validity of the proposed model. Basically, more rigorous experiments are needed to strengthen the findings.
There is a considerable emphasis on interaction models being generated from immunoprecipitation data from the over-expression of epitope tagged proteins in cells. The possibility of indirect binding partners being involved for interactions is not rigorously explored. This is particularly evident in the case of eIF4A and eIF3 binding to TRS. The ability of TRS to bind eIF4A and eIF3 needs to be more rigorously tested using biochemical and biophysical approaches to substantiate the proposed model whereby TRS functions in place of eIF4G.
The VEGF translation reporter assay requires the over-expression of 4EHP and TRS. This raises some concern about how this interaction promotes translation in an endogenous system when components are not over-expressed.

Detailed comments:
As the authors point out, the translation of VEGF can be regulated by hypoxia. It is shown that knockdown of 4EHP still inhibits VEGF translation in reduced oxygen conditions (FIG S4). It should be noted that the authors need to add error bars to all the data presented in Fig. S4 to show that the data are statistically relevant.
The translation of VEGF has been shown to be sensitive to both eIF4E and eIF4G levels, in both normal and hypoxic conditions (Int J Cancer J Int Du Cancer 1996; 65:785-90; PMID:8631593; Mol Cell. 2007;28:501-512). The authors need to confirm that eIF4E and eIF4G knockdown do in fact reduce VEGF translation in hypoxic conditions in their system.
The claim that eIF4A binds directly to TRS is most unexpected. This finding should be more rigorously investigated to substantiate the proposed model. Is the translation of VEGF sensitive to eIF4A inhibition by PDCD4 and hippuristanol? The in vitro binding assays (ITC and pull-downs using purified recombinant full-length factors) should be repeated with the addition of eIF4A. A mutant TRS that cannot bind to eIF4A should also be generated and tested in the TRS over-expression VEGF reporter cell based assay.
The VEGF mRNA precipitation results are rather confusing. It appears that the input amounts of each segment differ substantially: is it by chance that the positive IP is the most abundant? The input of some segments are already lower than the amount of positive segment pulled down, making it hard to judge whether the TRS binding was specific to this segment. Should TRS binding to anticodon motif be important for VEGF translation, the overexpression of tRNA(Thr)-anticodon stem or threonine depletion should counteract this activity. Has this been tested?
Reviewer #4 (Remarks to the Author): The authors have conducted two morpholino-based knockdowns and analysed the resultant vascular phenotypes. Overall, the methods, controls and rescues are all appropriate however I have a few suggestions for improvement: 1) How did the authors determine which zebrafish gene was the ortholog of TRS and 4EHP? While an obvious eif4e2 gene is annotated in the zebrafish genome (http://zfin.org/ZDB-GENE-050327-59), currently, no zebrafish gene is annotated with "trs". The authors should describe how they identified the zebrafish ortholog, include either the ZFIN or ENSEMBLE number for the gene investigated and use the appropriate zebrafish nomenclature if available.
2) The numbers for each morpholino experiment are very low. I would expect the number of animals analysed to be in the 20-30 range not the 5-10 range in this study. This is especially concerning in the TRS morpholino study ( Figure 5E,F) which shows only a very minor phenotype. Presumably, with numbers this low these experiments have only been done once and should be repeated to determine consistency.
3) The statistical methods for determining significance in Figure 5F, H and Supplementary Figure 6D are not stated. 4) While the 4EHP MO appears to be working well, the same cannot be said for the TRS MO. As the TRS morphant phenotype is, at best, subtle, I would suggest removing the TRS data from figure 5 ( Fig  5E-F) and presenting the 4 EHP MO ISV phenotype (Sup Fig 6d) in the main figure as this is more convincing data.
5) The authors should consider trying another TRS splice-blocking MO that may be more effective. Figure 5F, H and Supplementary Figure 6D -standard deviation would be a more appropriate error bar to present than SEM. 7) In the methods, the authors should include the dose of MO injected rather than the concentration of the injection solution. 8) While not essential, whole mount in situ hybridisation using probes against the TRS and 4EHP genes would determine whether the spatiotemporal expression supports the phenotypes observed. Looking at published eif4e2 data (http://zfin.org/action/figure/all-figure-view/ZDB-PUB-040907-1?probeZdbID=ZDB-EST-041111-328) it appears to be strongly expressed in the somatic muscle and head during ISV and Ct vessel formation, supporting the authors hypothesis. Based on the reviewer's comments regarding the formation of the eIF4F-like complex at the protein level, we first purified full-length TRS with a Strep tag at the C-terminal end following expression in HEK293T cells. We would like to note that full-length TRS was difficult to purify to homogeneity due to its flexible N-terminal region, including UNE-T. 4EHP was purified without a tag as described in the manuscript. We then evaluated the interaction between purified TRS, which forms a dimer via the catalytic domain 1 , and 4EHP using in vitro Strep-Tactin pull-down assay. The results showed that unlike the isolated UNE-T, fulllength TRS interacts weakly with 4EHP (see the figure below, and Supplementary Fig. 1d in the revised manuscript), suggesting that the N-terminal UNE-T may not be fully exposed in the dimeric form of TRS in vitro. It is possible that formation of the cellular TRS-4EHP complex might involve a conformational change of the full-length TRS to fully expose UNE-T.

Description of changes in the revised manuscript
We have described these results in the revised manuscript.
In vitro reconstitution of a stable complex of full-length TRS, 4EHP, and eIF4A was difficult due to the intrinsic biophysical properties of TRS as mentioned above. To enhance the stability of the complex, we performed cross-linking. Reaction mixtures containing 20 μg of purified TRS-Strep, eIF4A, and 4EHP in a total volume of 100 μL were incubated with 0.0025% (v/v) glutaraldehyde in conjugation buffer (100 mM HEPES, pH 7.5, and 150 mM NaCl) for 5 min at 37°C. The reaction was terminated by addition of 10 μL of 1 M TRIS-HCl, pH 8.0. The cross-linked proteins were loaded onto Strep- Tactin resin for 30 min at 4°C,  washed thoroughly five times with ice-cold wash buffer (50 mM TRIS-HCl, pH 7.5, 300 mM   NaCl, 1 mM EDTA, and 2% NP-40), and eluted with Strep elution buffer (100 mM HEPES,   pH 8.0, 150 mM NaCl, and 1 mM EDTA supplemented with 2.5 mM desthiobiotin). The eluted sample was analyzed using denaturing SDS-PAGE and immunoblotting. As shown in the figure below, a cross-linked band corresponding to the covalent TRS-4EHP-eIF4A complex was clearly visible in the monomeric state (stoichiometry 1:1:1) and dimeric state (stoichiometry 2:2:2) following denaturing SDS-PAGE, indicating that they do indeed form a complex. Although we demonstrated the direct interaction between full-length TRS and 4EHP, and trimeric complex formation including eIF4A in vitro, we suspect that other factors might be additionally required for the formation of a stable TRS-4EHP-eIF4A complex in vivo.
Since the main focus of this work was to report the novel functional connection between a protein synthesis enzyme (TRS) and a specific form of an initiation factor (4EHP), we believe that the exact number and identity of additional factors should be systematically investigated in-depth in future studies. For this reason, we have not included the cross-linking results in the revised manuscript, and instead include them in this response letter.
Regarding the enzymatic activity of eIF4A, we evaluated the effect of TRS on the ATPase activity of eIF4A. The results showed that the ATPase activity of eIF4A was enhanced by TRS, but not by LRS (see the results below). Thus, these results further suggest that TRS may play a similar role to eIF4G in the eIF4F complex, as deduced in a previously reported study showing that the ATPase activity of eIF4A is augmented by eIF4G 2 .

Figure.
Enhancement of the ATPase activity of eIF4A by TRS. The effect of TRS on the ATPase activity of eIF4A was determined using a malachite green assay (R&D Systems).
The specific activity was determined using a phosphate standard curve, and all experiments were performed in triplicate with similar results.

Much of the authors' data reflect various pulldown experiments followed by Western blotting.
While specificity may be seen in that eIF4A2 is pulled down ( Figure 1A)  Response: We assume that eIF4A2 mentioned here by the Reviewer is eIF4E2, which is 4EHP. As described in the original manuscript, we performed affinity purification of Streptagged TRS-interacting proteins (see the figure below) and analyzed them by LC-MS. From the analysis, we identified 434 proteins as potential TRS-associating proteins, directly or indirectly, including proteins involved in post-transcriptional regulation, mRNA metabolic processes, and translation ( Supplementary Fig. 1a, b). We also carried out yeast LexA-B42 two-hybrid screening and identified four proteins as potential TRS-interacting proteins, including TRS itself ( Supplementary Fig. 1c). Based on these results, we decided to focus to the interaction of TRS with 4EHP because both analyses predicted the potential interaction of the two proteins, and because both are commonly involved in translation processes 3, 4, 5, 6 .
We further examined the potential interaction of many other ARSs with 4EHP and found that only TRS engaged in specific binding to 4EHP (Fig. 1b). The specificity of the cellular interaction between TRS and 4EHP was further demonstrated by BiFC experiments (Fig. 1e).
These new data have been added to the revised manuscript.
Regarding the concerns about reliability of the data raised by the Reviewer, all experiments were repeated independently at least three times, and data in the revised manuscript are representative of experiments with similar results, shown as means ± SEM.

Response:
We acknowledge the Reviewer's comments. As emphasized in the main text, this work revealed a novel translational initiation complex involving TRS (an aminoacyl-tRNA synthetase) and 4EHP (an isoform of eIF4E). We do not yet know how many other unknown factors may be required for a fully functional complex. In this particular work, we focused on determining the structure of the binary complex of TRS and 4EHP, and its functional implications in translational initiation. We believe other factors required for full activity and their interactions with the rest of the components are subjects for future research.
Nonetheless, based on the reviewer's suggestion, we examined the direct interactions of TRS with eIF4A, and also with PABP, using in vitro pull-down assays. Purified eIF4A and PABP proteins (without a tag) were sticky and bound non-specifically to Ni-NTA agarose resin and glutathione Sepharose resin. To avoid these problems, a reaction mixture containing purified full-length TRS-His and eIF4A was incubated with eIF4A antibody, and pulled down with Protein A/G PLUS-agarose beads, and co-purification of the two proteins as a complex was determined by immunoblotting analysis. The results confirmed the direct interaction of the two factors ( Supplementary Fig. 10a).
To test the interaction between TRS and PABP, we purified full-length TRS-Strep and His-GST-PABP, loaded the reaction mixture onto Strep-Tactin resin, eluted with Strep elution buffer, and co-purification of the two factors as a complex was determined by immunoblotting analysis. The results also confirmed the direct interaction of TRS and PABP ( Supplementary   Fig. 12b). These novel insights have been added and described in the revised manuscript.

Q3. The authors often describe pull down results as indicating "a specific interaction" if one
or two other proteins are not seen. This is insufficient (see above).

Response:
We understand that "a specific interaction" can be better supported by using a greater number of control proteins for comparison. Thus, we further validated the specificity of the TRS-4EHP interaction using various aminoacyl-tRNA synthetases (ARSs) in the revised manuscript. Among the 13 different ARSs tested, only TRS engaged in an interaction with 4EHP (Fig. 1b). In the case of eIF4E homologues in humans, only eIF4E2 (4EHP) interacted with TRS (Fig. 1a). We also confirmed the specific interaction between TRS and 4EHP using bimolecular fluorescence complementation (BiFC) assays ( Fig. 1e) in the revised manuscript.

Q4. The TRS-4EHP interaction is cited as specific for vertebrates, yet there is only a single
non-vertebrate tested (fly). It would be useful to extend this to at least two other negative species.
Response: Following the Reviewer's comment, we further evaluated the interaction between TRS and 4EHP from yeast and nematode, and observed no interaction in these two lower eukaryotic organisms. These new results have been added in the revised manuscript ( Fig. 3a, b, f, g). Response: Following the Reviewer's comments, we repeated the mRNA enrichment experiments not only with anti-TRS antibody, but also with anti-PRS and IRS antibodies, and compared the enriched mRNAs. Similar to the results obtained using AlaRS as a control, the number of TRS-enriched mRNAs was reduced from 7,708 (IgG control) to 2,694 and 2,866 after subtraction of PRS-and IRS-enriched mRNAs, respectively. Furthermore, functional annotation clustering analysis revealed that a large proportion of the enriched GO terms were related to vertebrate system development. These results have been described in the revised manuscript ( Supplementary Fig. 3). down experiments, we showed that TRS and 4EHP form a distinct complex separate from the eIF4G-eIF4E complex, and both complexes contain eIF4A (Fig. 6d, e). Although we acknowledge the Reviewer's comments, given the order of our experiments and discoveries in our study, we believe it is better to keep the current dataset. However, to strengthen the original results, we have also checked AlaRS and KRS (as negative controls) and eIF4E (as a positive control) in the cap-dependent complex consisting of 4EHP and TRS, and added the new data to the revised manuscript ( Supplementary Fig. 9). Figure 7F is supposed to show an interaction with eIF3 and TRS. However, only a few eIF3 subunits appear visible. Why are not all of the eIF3 subunits pull down?

Q9.
Response: Consistent with the Reviewer's comments, we observed that all eIF3 subunits except for eIF3A and I were immunoprecipitated with TRS in pull-down experiments following co-expression of TRS-Strep with each of the FLAG-tagged eIF3 subunits (see the results below). However, we believe that some of these eIF3 subunits could be pulled down with TRS indirectly via direct TRS-binding subunits. To identify potential direct TRS binders, we conducted in vitro pull-down assays with GST-TRS and each of the eIF3 subunits expressed as FLAG-tagged proteins using glutathione-Sepharose. To exclude indirect binders, we thoroughly washed the TRS-binding mixture bound to the beads, and found that only the subunits eIF3B, D, F, and L survived the stringent washing step and co-precipitated with GST-TRS (Fig. 7e). Endogenous TRS was immunoprecipitated from WI-26 cells with a mouse anti-TRS antibody and examined for associated eIF3 subunits using antibodies recognizing each of the subunits (Fig. 7f). We also performed intensive washing steps for these experiments, and again found that subunits B, D, F, L (and E) co-purified with TRS, further confirming the above results.

Figure.
Co-immunoprecipitation assay of TRS-Strep co-expressed with each of the FLAGtagged eIF3 subunits in 293T cells. TRS-Strep was purified using Strep-Tactin resin, and coprecipitated eIF3 subunits were determined by immunoblotting with anti-FLAG antibody.
In addition, we also analyzed the interaction between eIF3 and TRS at the protein level.  Fig. 4b). We have further evaluated the effects of loop length on translation regulation in the revised manuscript, and observed that the length of the loop is also important for recognition by TRS (Supplementary Fig. 4c). We also identified the potential TRS-recognizing loop located in the 5′ UTR of ANG mRNA, and similar results were obtained with a loop incorporated upstream of the luciferase gene, as detailed in the revised manuscript ( Supplementary Fig. 4d, e).

Q2. In vitro pull-down assay needs to be checked to confirm direct interaction between TRS
and PABP because the latter's detection in co-IP with TRS ABD can also be due to interaction of mRNAs with ABD. The results revealed that the two proteins directly interact with each other ( Supplementary   Fig. 12b). These results have been described in the revised manuscript.

Q3. TRS LLL mutant also can be probed for interaction with eIF4A, PABP and eIF3 to rule out any overlap between binding sites of these factors and that of UGU anticodon-like loop.
Response: We evaluated whether the TRS ABD site binding to the anticodon Thr -like loop overlaps regions interacting with eIF4A, PABP, and eIF3. The results clearly showed that mutation of residues in the TRS ABD that is critical for binding to the anticodon Thr -like loop did not disturb the interactions with other translation initiation factors. These results have been added to Supplementary Fig. 10b in the revised manuscript. Minor comments Q1. Page 3, para 2: "Cell condition-specific…eIF4 isomers mediate…sequesters eIF4E." Change "isomers" to "isoforms".

Response:
We have corrected this in the revised manuscript.

Response:
We have corrected this in the revised manuscript. Q7. Page 9, para 2: "Remarkably, most genes…development of nervous, skeletal, and circulation systems…" Nervous and circulatory systems are not unique to vertebrates! Thus, the sentence needs to be modified.

Response:
We have modified this sentence in the revised manuscript. Fig. 4a)."

Explanation? Error bars?
Response: It has been reported that silencing of eIF4G reduces VEGF production more than does silencing of eIF4E 9 . We have added error bars in the revised manuscript.
Q9. Page 11, para 2: "Silencing of TRS…by oxygen tension ( Supplementary Fig. 4e)." What is the relevance of this aspect in the context of the study?
Response: VEGF is generally induced during hypoxia. In addition, as described in the Introduction section, together with oxygen-regulated hypoxia-inducible factor 2α and RNAbinding protein RBM4, 4EHP regulates global hypoxic protein synthesis during hypoxia 4 . Therefore, whether TRS and 4EHP-mediated translation initiation is affected by oxygen availability should be checked in this study. Fig. 5d and Supplementary Fig. 5d)." Enhancement in the case of M60K is seen for both tube length and cell migration, but to a lesser extent than wild-type. Hence, the statement is incorrect and needs to be modified accordingly.

Response:
We thank the Reviewer for pointing out this incorrect statement. We have modified this statement "Tube formation and…TRS (M60K)-expressing cells (Fig. 5d and Supplementary Fig. 5d)." to "Tube formation and cell migration of HUVECs were significantly enhanced by supernatants from TRS-transfected WI-26 cells, but to a lesser extent by supernatants from 4EHP-binding-defective TRS (M60K)-expressing cells ( Fig. 5d and Supplementary Fig. 6f, g)." in the revised manuscript. Q15. Fig 2a: The Fo-Fc map should be an unbiased map generated before modeling of the ligand. Fig. 2a showing the Fo-Fc map was generated using the procedure recommended by the reviewer.  Supplementary Fig. 4g).

Q17. Supplementary fig 4c: Why do eIF4E bands appear so faint? Error bars?
Response: We have repeated these experiments and updated the results in the revised manuscript ( Supplementary Fig. 5c). Supplementary fig 4a,c,e.

Response:
We have added error bars in the revised manuscript ( Supplementary Fig. 5).

Reviewer #3 (Remarks to the Author):
The authors present a study that proposes a new function of an aminoacyl-tRNA synthetase in stimulating cap-dependent translation of VEGF. Overall, this is an interesting study and has the potential to transform our understanding of how selective translation of VEGF can be regulated by 4EHP. A strength of the study is the structural model that reveals how 4EHP binds to TRS. The validation of this interaction using structure guided mutants is impressive (although mutations in zebrafish TRS did not clearly result in loss of binding (Fig. 3d)). and TRS. This raises some concern about how this interaction promotes translation in an endogenous system when components are not over-expressed.

Response:
We repeated the zebrafish TRS mutation validation experiments and obtained consistent but more convincing results. The new results have been added to the revised manuscript (Fig. 3d).
We agree with the point raised by the Reviewer that the possibility of indirect binding partners being involved in interactions was not rigorously explored in the original manuscript.
We should emphasize that we do not yet know how many other unknown factors may be required for a fully functional TRS-mediated translation initiation machinery. In this particular work, we focused on determining the structure of the binary complex of TRS and 4EHP, and its functional implications in translational initiation. Therefore, we believe that determining how many other factors may be required, directly or indirectly, for full activity, and their interactions with the rest of the components, are subjects for future studies.
Regarding TRS binding to eIF4A, we further evaluated the interaction at the protein level in the revised manuscript. Please see the response to Q3 below for detailed information.
We also analyzed the interaction between eIF3 and TRS at the protein level. We first transfected each of the eIF3 subunits tagged with FLAG, and conducted in vitro pull-down assays with GST-TRS. The results confirmed that TRS makes direct interactions with the eIF3 subunits B, D, F and L (Fig. 7e). We further conducted co-immunoprecipitation experiments between endogenous TRS and the eIF3 subunits in HEK293T cells and observed that TRS can also interact with the eIF3 subunit B, D, F and L (also with E) (Fig. 7f).
We subsequently attempted to purify each of the eIF3 subunits using Strep-Tactin resin and found that most of the eIF3 subunits except for A, G, I, and J were tightly associated with the subunit D (see the results below, Figure a, b). Thus, we used the eIF3 subunits enriched with subunit D to assess the interaction with TRS in vitro. The purified GST-TRS was incubated the eIF3 subcomplex containing the subunits (B, C Regarding the point raised by the Reviewer that the VEGF translation reporter assay requires overexpression of 4EHP and TRS, and that this raises some concern about how this interaction promotes translation in an endogenous system when components are not overexpressed, this is a reasonable concern and generally applicable to any biological experiments in which the overexpression effects of the factor of interest are monitored. Since the majority of TRS molecules would be occupied during catalysis, only a small portion would be used for translation initiation complex formation. In this work, we also showed that 4EHP is strongly induced at embryonic day 10.5, which would affect its availability to form a translation initiation complex (Supplementary Fig. 13). Based on these facts, we predict that rather than other translation factors, TRS and 4EHP could be the limiting components involved in the formation of the translation initiation complex. Similar to the reviewer, we were also surprised to observe an interaction between TRS and eIF4A. Although we tried many different approaches to obtain a clear picture of the interaction, most experiments did not produce desirable results, mainly due to the challenging biochemical properties of full-length TRS (perhaps due to the flexible linker between the catalytic domain and UNE-T) and eIF4A. After numerous attempts, we finally purified full-length TRS-His and eIF4A (without a tag), and the mixture was reacted with anti-eIF4A antibody, mixed with Protein A/G PLUS-agarose beads, washed extensively, and subjected to immunoblotting analysis. The results confirmed the direct interaction of TRS and eIF4A (Supplementary Fig. 10a). These results have been described in the revised manuscript.
We also evaluated the effect of TRS on the ATPase activity of eIF4A. The results showed that the ATPase activity of eIF4A is enhanced by TRS, but not by LRS (see the results below). Thus, these results further indicate that TRS may play a similar role to eIF4G in the eIF4F complex, as previously reported for the ATPase activity of eIF4A, which is augmented by eIF4G 2 .

Figure.
Enhancement of the ATPase activity of eIF4A by TRS. The effect of TRS on the ATPase activity of eIF4A was determined using a malachite green assay (R&D Systems).
The specific activity was determined using a phosphate standard curve generated. All experiments were performed in triplicate with similar results.
We attempted to identify a TRS mutant that does not interact with eIF4A. We should emphasize that finding TRS residues that are critical for interacting with eIF4A is very difficult without structural information on the interaction between the two proteins. Nonetheless, using previously reported structural information for each of the proteins, we screened mutants and found that mutations at positions R699, E703, E706, and R707 (mutant #4) in TRS weakened the interaction with eIF4A (see the results below, Figure a)   We also examined the Reviewer's suggestion to check the effects of threonine depletion and tRNA Thr -anticodon stem overexpression on the translation of VEGF mRNA. Since it is already known that amino acid depletion can enhance the expression of VEGF regardless of amino acid type 10, 11 , we did not expect that VEGF translation would be differentially affected by the depletion of different amino acids. When we compared that VEGF expression levels between the leucine-and threonine-depletion conditions, we did not see apparent difference between the two conditions (see the results below, Figure a). We also compared the effects of threonine and alanine tRNA-anticodon stem overexpression on VEGF translation and observed little effect (see the results below, Figure b). Although the experiments did not provide the results the reviewer suggested, perhaps, the expression level or TRS-binding affinity of the ectopically expressed tRNA-anticodon stem loop have not been sufficiently high enough to take off TRS from the TRS-mediated translation initiation complex in which TRS was multiply associated not only with VEGF mRNA but also with many translation factors (as shown above).

Reviewer #4 (Remarks to the Author):
The  (4) and (5), we found a more effective trs morpholino (trs-i6e7 MO) that worked better than the previous one described in the original manuscript. Using this new trs-i6e7 MO, we have increased the number of animals and repeated each experiment with at least 20 embryos per condition (up to 30 embryos in some cases). The results are included in the revised manuscript (Fig. 5e-h and Supplementary Fig. 7d). Figure 5F, H and Supplementary Figure 6D are not stated.

Response:
We have stated the statistical methods (one-way ANOVA for Fig. 5f and 5h, and student t-test for Supplementary Fig. 7d) in the Methods section of the revised manuscript.

Q4.
While the 4EHP MO appears to be working well, the same cannot be said for the TRS MO. As the TRS morphant phenotype is, at best, subtle, I would suggest removing the TRS data from figure 5 (Fig 5E-F) and presenting the 4 EHP MO ISV phenotype (Sup Fig 6d) in the main figure as this is more convincing data.

Response:
Since we now have more convincing data for the TRS morphant phenotype, we have replaced the original data with the new data in the revised manuscript. We are grateful for the Reviewer's suggestion to improve the data quality.

Q5. The authors should consider trying another TRS splice-blocking MO that may be more effective.
Response: Based on the reviewer's comments (Q4) and (Q5), we designed and tested several new candidate trs morpholinos to find a more effective one. We eventually identified a morpholino (trs-i6e7 MO) that worked better than the one described in the original manuscript. Since the knock-down of TRS using the trs-i6e7 MO elicited more overt cerebrovascular phenotypes, we have replaced the original data with the new data obtained using the trs i6e7 MO in the revised manuscript (Fig. 5e, f). Figure 5F, H and Supplementary Figure 6D -standard deviation would be a more appropriate error bar to present than SEM.

Q6.
Response: Following the reviewer's suggestion, we have changed the error bars to standard deviation (SD) for all graphs in the revised manuscript (Fig. 5f, h; Supplementary   Fig. 7d).

Q7.
In the methods, the authors should include the dose of MO injected rather than the concentration of the injection solution.
Response: As suggested, we have described the dose of MO injection in the Methods section of the revised manuscript. Response: Following the reviewer's suggestion, we investigated the expression patterns of trs and 4ehp using whole-mount RNA in situ hybridization during zebrafish embryogenesis at 24 hpf, 48 hpf, 72 hpf, and 4.5 dpf. As the reviewer mentioned, 4ehp was highly expressed in the developing trunk at 24 hpf ( Supplementary Fig. 8e, e'), consistent with the ISV defects upon 4ehp knock-down ( Supplementary Fig. 7d). In addition, the expression of both trs and 4ehp was also detected in the developing hindbrain during the period in which CtAs in the hindbrain were actively formed, further supporting their involvement in cerebrovascular formation. We have described the correlation of the vascular phenotypes (ISVs in the trunk and CtAs in the hindbrain) and their spatiotemporal expression patterns in the revised manuscript ( Supplementary Fig. 8).
The authors have improved their manuscript and they have answered extensively concerns from the previous review. This continues to be a very interesting manuscript and one with content appropriate for publication in Nature. However, this reviewer still has several concerns (in large part owing to the extensive data presented).
Major concerns 1. Line 103 "…434 proteins identified …" An examination of the data in Figure S1 leaves open the possibility for non-direct interactions (i.e. either RNA sensitive or ribosome associated). There is no immediate ranking of these interactions given (i.e. often seen as percentage of the entire protein sequence found by mass spectrometry). At the same time, eIF4A is not identified in the screen see in Figure S1C.
2. Line 130 -If the full length TRS interacted poorly with 4EHP, is the complex that is immunoprecipitated with TRS strep a monomer or a dimer of TRS?
3. Line 222 -The identification of over 2,900 transcripts would seem to suggest a lack of specificity. If the yeast mRNA transcripts were tested, how many might be found? It would also be helpful to know the number of transcripts that were isolated using the ARS (no number is given).

Line 354
-Although there appears to be a TRS-eIF4A interaction seen in Figure 6, panel A, there does not appear to be an equivalent interaction seen in panel D where one would expect to see eIF4A as a subunit of eIF4F. Also, why is the amount of 4EHP reduced in the siTRS lane? Again, in panel E, in the absence of eIF4G which might allow for more TRS-eIF4A complexes, there is essentially no eIF4A found in the sieIF4G lane.

5.
Line 397 -It is not clear why not all of the eIF3 subunits are pulled down together (i.e. as seen in the WCL lane in panel F of Figure 7). Do these represent free or excess subunits not in the entire factor? Or do they represent a subspecies of eIF3?
Minor comments 1. Line 36 -This should be "analogous to the eIF4F-mediated translation" rather than eIF4G.
2. Line 222 -it would be of value to know how many of the 2,928 transcripts contained elements that would be similar to the stem loop identified in the VEGF mRNA, especially in the 5' UTR. Secondly, it would ultimately be important to know the molar amount of interacting components with TRS to have some feeling for which were more likely to be found in cells. As noted in the Discussion (line 459), the timing for the expression of both 4EHP and VEGF seems to coordinated (i.e. for day 10.5).
Reviewer #2 (Remarks to the Author): The revised manuscript by Sunghoon Kim and coworkers is very much improved compared to the original version. While the original manuscript itself had substantial data, the authors have now performed additional experiments in response to the reviewers comments. I must say that I am overall impressed with the quantity and quality of data presented and they have addressed satisfactorily all my comments. I therefore strongly recommend publication of the manuscript.

Rajan Sankaranarayanan
Reviewer #3 (Remarks to the Author): Overall, I am satisfied that the authors have strong data to reveal a novel interaction between 4EHP and TRS. However, I do not feel that they have shown what the functional significance of this interaction is. Without this information, we are left with the discovery of an interesting interaction, the function of which will need to be determined in the future. Overall, I am not convinced from the revised manuscript that this is really a translation initiation complex.
Some specific comments: No details are provided for how the ATPase data in the author rebuttal letter was generated. For example, is this RNA-dependent ATPase activity? The lack of rigor in how this data has been generated leaves me unable to assess its quality. From this data, I am not satisfied that the authors have rigorously established that TRS regulates the ATPase activity of eIF4A.
As originally stated in the first review -the interaction data rely on over-expression of proteins to test interactions. The authors state that this is a limitation of any experiments that rely on overexpression. I agree with this statement -which is why one needs to be more rigorous before generating models from overexpression studies! Other approaches should have been used to rigorously test these models -for example, inducible promoters could have be used to limit overexpression levels of proteins (perhaps equal of endogenous). The fact that the study hasn't even tried to overcome this type of limitation is a problem that lowers overall enthusiasm for this study.
The eIF4A interaction data presented in Fig S10a needs to include a control lane where TRS is not included. Without this control it is not possible to interpret the data since eIF4A may bind the resin in the conditions used. I also don't understand what "The beads were washed intensively" means in the figure legend. Please provide details about how the experiment was done -how much resin, what buffer, what volume, how many washes etc. Without this information, it is impossible to assess the rigor of the experiment and it is not possible for anyone else to replicate the data in the future.
Reviewer #4 (Remarks to the Author): The authors have addressed all my concerns.

Responses to Reviewers` Comments
Reviewer #1 (Remarks to the Author): The authors have improved their manuscript and they have answered extensively concerns from the previous review. This continues to be a very interesting manuscript and one with content appropriate for publication in Nature. However, this reviewer still has several concerns (in large part owing to the extensive data presented).

Response:
We deeply appreciate the Reviewer #1 that he/she continuously supports our study.
Major concerns Q1. Line 103 "…434 proteins identified …" An examination of the data in Figure S1 leaves open the possibility for non-direct interactions (i.e. either RNA sensitive or ribosome associated). There is no immediate ranking of these interactions given (i.e. often seen as percentage of the entire protein sequence found by mass spectrometry). At the same time, eIF4A is not identified in the screen see in Figure S1C.
Response: As Reviewer #1 mentioned, we also assume that many proteins identified from the TRS-interactome analysis would be indirectly TRS-associating proteins. As the Reviewer guessed, they might possibly be either RNA sensitive or ribosome associated proteins because human TRS functions as a translation initiation factor that interacts with other translation initiation proteins, which might recruit those proteins. As requested, we have ranked gene list of the TRS-interactome as a percentage of the protein sequence found by mass spectrometry in the revised manuscript (Fig. S1b).
As pointed out by the Reviewer, we did not identify eIF4A as a TRS-interacting protein in the screen using yeast two-hybrid system (Fig. S1c). Although the two-hybrid screening system has identified many previously unknown protein-protein interactions for last decades, it cannot uncover all the potential interactors as most of other screening approaches. There are many reasons for missing the potential interactors. Mostly frequently, the cDNA library prepared for the two-hybrid screening may not cover 100% of the human genome proteins.
Besides, some proteins may not be able to fold correctly within the yeast cells and steric hindrance resulting from the fusion to the yeast proteins may inhibit the interactions between the two interactors. For this limitation, most of the primary screening results are complemented by another approaches (in our case, we did the yeast two-hybrid and affinity purification screening combined with mass spectrometry) and also further validated by manual approaches based on the hypothesis and previous knowledge. Q3. Line 222 -The identification of over 2,900 transcripts would seem to suggest a lack of specificity. If the yeast mRNA transcripts were tested, how many might be found? It would also be helpful to know the number of transcripts that were isolated using the ARS (no number is given).

Response:
We fully understand the Reviewer's concern. It is known that this "omics" type of experiments (such as RIP-seq in this case) usually bear many false (positive and negative) hits and that is why the selected results are subjected to further experimental validation (which we did using cell and in vivo models). To minimize the potential non-specificity in our experiments, the TRS-enriched transcripts were independently subtracted by three negative control sets using AlaRS, IRS and PRS (also following the Reviewer's previous advice). The results consistently showed the transcripts for system development in the range of 30~40% in the TRS-enriched transcripts while they normally exist in 5.8% in the whole transcripts (see the figure below). Thus, these results suggest that the transcripts for system development should be preferentially targeted by TRS. Nevertheless, we actually tried to analyze yeast mRNA transcripts, but, in the process of preparing the experiment, we realized that it is not technically feasible at this moment because no antibody is available specific to yeast TRS that should be used to enrich TRSbound transcripts. We wish the Reviewer's kind understanding of the results mentioned above and current technical limitations.
Lastly, we apologize for somewhat confusing schematic representation of the workflow to identify TRS-targeted mRNAs (shown in Fig. 4a). We have thus provided the workflow with the information on the number of transcripts that were isolated using the ARS in Fig. 4a in the revised manuscript. Accordingly, the workflows shown in Supplementary Fig.   3a and d have been also changed in the revised manuscript.
Q4. Line 354 -Although there appears to be a TRS-eIF4A interaction seen in Figure 6, Like many other cases, we believe that some of the eIF3 subunits could be pulled down with direct TRS-interacting subunits. To distinguish potential direct and indirect TRS binders, we conducted in vitro pull-down assays with GST-TRS and each of the eIF3 subunits expressed as FLAG-tagged proteins using glutathione-Sepharose. To exclude indirect binders, we thoroughly washed the TRS-binding mixture bound to the beads, and found that the subunits eIF3B, D, F, and L survived the stringent washing step and coprecipitated with GST-TRS (Fig. 7e).
Endogenous TRS was immunoprecipitated from WI-26 cells with a mouse anti-TRS antibody and examined for associated eIF3 subunits using antibodies recognizing each of the subunits (Fig. 7f). We also performed intensive washing steps for these experiments, and again found that subunits B, D, F, L (and E) co-purified with TRS, further confirming the above results. For reference, direct eIF4G-interaction eIF3 subunits were assigned to eIF3C,

Q1.
Line 36 -This should be "analogous to the eIF4F-mediated translation" rather than eIF4G.

Response:
We have changed "analogous to the eIF4G-mediated translation" to "analogous to the eIF4F-mediated translation" in the revised manuscript.
elements that would be similar to the stem loop identified in the VEGF mRNA, especially in the 5' UTR. Secondly, it would ultimately be important to know the molar amount of interacting components with TRS to have some feeling for which were more likely to be found in cells. As noted in the Discussion (line 459), the timing for the expression of both 4EHP and VEGF seems to coordinated (i.e. for day 10.5).
Response: Since TRS appears to mainly target the transcripts for system development, we focused to the transcripts (see the

Reviewer #3 (Remarks to the Author):
Overall, I am satisfied that the authors have strong data to reveal a novel interaction between 4EHP and TRS. However, I do not feel that they have shown what the functional significance of this interaction is. Without this information, we are left with the discovery of an interesting interaction, the function of which will need to be determined in the future. Overall, I am not convinced from the revised manuscript that this is really a translation initiation complex.

Response:
The entire data shown in our work are converged to show the functional significance for translation initiation of the interaction between TRS and 4EHP. In addition to comparative structural and biochemical analyses, we proved the function of this complex using molecular and cell-based assays as well as in vivo animal models at the level that current techniques allow.
To our knowledge, a few translation initiation complexes containing 4EHP have been Some specific comments:

Q1.
No details are provided for how the ATPase data in the author rebuttal letter was generated. For example, is this RNA-dependent ATPase activity? The lack of rigor in how this data has been generated leaves me unable to assess its quality. From this data, I am not satisfied that the authors have rigorously established that TRS regulates the ATPase activity of eIF4A.

Response:
We acknowledge that the ATPase data would need more detailed description.
We found that we also made typos in the figure for ATPase data. The amounts of proteins we used for the assays were not 20 μM or 40 μM for each reaction, but 2 μM or 4 μM (see the figure below). We sincerely apologize for our carelessness and are much thankful the Reviewer for pointing this.  Q2. As originally stated in the first review -the interaction data rely on over-expression of proteins to test interactions. The authors state that this is a limitation of any experiments that rely on overexpression. I agree with this statement -which is why one needs to be more rigorous before generating models from overexpression studies! Other approaches should have been used to rigorously test these models -for example, inducible promoters could have be used to limit overexpression levels of proteins (perhaps equal of endogenous). The fact that the study hasn't even tried to overcome this type of limitation is a problem that lowers overall enthusiasm for this study. Briefly, MIA PaCa-2 cells were cultured in DMEM supplemented with 10% FBS and antibiotics and seeded evenly in a 60 mm dish and incubated for 12 h to reach ~90% confluence. When the cells were ready for transfection, 2 μl of Myc-tagged TRS lentiviral particles and 3 μl of 10 mg/ml polybrene were supplemented with 2 ml of complete media and added to the plate. After 16 h of incubation, the culture media was replaced with 3 ml of fresh complete media containing 1 μg/ml puromycin and incubated for an additional 48 h.
The cells were gradually selected by treating with puromycin every 48 h. The efficiency of TRS overexpression was checked using the cells treated with 2.5 μg/ml of doxycycline (Dox) for the indicated times, followed by immunodetection.
MIA PaCa-2 cells that expressed inducible Myc-tagged TRS were cultured in the presence or absence of Dox for 24 to 48 h. Myc-tagged TRS was immunoprecipitated with anti-Myc antibody, and co-precipitation of 4EHP was determined by immunoblotting with 4EHP antibody. Levels of VEGF in the cell culture supernatants were determined by ELISA.
As seen in the figure below, Myc-tagged TRS showed Dox-dependent expression.
Immunoprecipitation using anti-Myc antibody followed by Western blot revealed that the interaction of 4EHP with TRS was increased in a Dox-dependent manner. VEGF levels were also increased in a Dox-dependent manner. Thus, these results further indicate that the TRS interaction with 4EHP positively regulates protein synthesis. Without this information, it is impossible to assess the rigor of the experiment and it is not possible for anyone else to replicate the data in the future.
Response: As requested, we re-performed the interaction experiment and the result has been changed in the revised manuscript. Regarding the experiment method for the interaction, we described the details in "In vitro binding assay" of the "Methods" section to be comparable to other figure legends in the previous manuscript. Fig. 9) was indeed commented by the Reviewer #1 in the first review and addressed in the "Description of Responses-to-Reviewers" in the first revision as following:
Regarding the level of 4EHP in whole cell lysates for WI-26 and VSMC showing much less than in the other cells in Supplementary Fig. 9, we think that it might be depending on cell type. Figure 1d -it is still not clear to this reviewer why so little 4EHP is pulled down in the TRS-Strep/4EHP lane considering how much was present as seen in input.

Q3. Supplementary
Response: As we described in the manuscript, unlike the isolated UNE-T, the N-terminal UNE-T in the dimeric form of full-length TRS may not be fully exposed for the interaction with 4EHP in vitro, resulting in the weak interaction with 4EHP ( Supplementary Fig. 1d). Perhaps, formation of the cellular TRS-4EHP complex might involve a conformation change of the fulllength TRS to fully expose UNE-T.
In sum, this manuscript adds a new pathway to protein synthesis initiation. Although this may easily be a minor pathway such as seen with IRES-mediated initiation or re-initiation, it none the less appears to be important. It is anticipated by this reviewer that additional pathways will emerge as reflects either development or the ability of cells to resist stress and maintain homeostasis.

Reviewer #3 (Remarks to the Author):
I am satisfied that the authors have addressed my original comments and feel that the manuscript is improved and ready for publication.