Rapidly growing cells typically produce an average of one to two proteins per ribosome per minute (Princiotta et al., 2003). Both the amount and quality of proteins produced from a cell’s transcriptome are crucial for cellular homeostasis (Labbadia and Morimoto, 2015). Thus, cells have evolved numerous mechanisms for regulating translation output (Roux and Topisirovic, 2018) and monitoring protein quality during and after translation (Wolff et al., 2014). One of the earliest monitors of protein quality detects ribosomes that stall during translation (Brandman and Hegde, 2016; Joazeiro, 2019). A ribosome that slows excessively can be an indicator of several potential problems including a damaged or incorrectly processed mRNA (Shoemaker and Green, 2012), amino acid or energy insufficiency (Harding et al., 2019; Inglis et al., 2019), or certain types of cellular stress (Liu et al., 2013; Shalgi et al., 2013). Inability to respond appropriately to these problems leads to disrupted protein homeostasis and disease (Kapur and Ackerman, 2018).

Because a common cause of ribosome stalling is a defective mRNA, stalled ribosome detection is linked to mRNA decay (Shoemaker and Green, 2012) and degradation of the partially synthesised protein (Bengtson and Joazeiro, 2010; Joazeiro, 2019). There are at least two mechanisms that cells use to detect a stalled ribosome. In the specialised case of an mRNA truncated within the coding region, ribosomes stall at the 3’ end with an empty A site that cannot be engaged by either elongation or termination complexes. This allows efficient engagement by the Pelota-Hbs1 complex to initiate a ribosome splitting reaction that separates the peptidyl-tRNA-60S complex from the mRNA-40S complex (Becker et al., 2011; Pisareva et al., 2011; Shao et al., 2016; Shoemaker et al., 2010). The 60S complex is engaged by ribosome-associated quality control (RQC) factors to trigger polypeptide degradation (Shao et al., 2015), while the 40S complex engages mRNA decay pathways by still poorly-understood mechanisms (Schmidt et al., 2016).

When a ribosome slows excessively within the coding region, its detection relies on ZNF598 (Hel2 in yeast), an E3 ubiquitin ligase whose ubiquitination of a 40S ribosomal protein (eS10 in mammals or uS10 in yeast) initiates downstream mRNA quality control (Ikeuchi et al., 2019) and protein quality control (Garzia et al., 2017; Juszkiewicz and Hegde, 2017; Matsuo et al., 2017; Sundaramoorthy et al., 2017). The cue used by ZNF598 to recognise a slow ribosome is the collision that occurs when a trailing ribosome catches up (Juszkiewicz et al., 2018). The collided ribosome shows a distinctive 40S-40S interface where ZNF598/Hel2 is proposed to bind and ubiquitinate nearby target sites (Ikeuchi et al., 2019; Juszkiewicz et al., 2018).

The steps downstream of ZNF598 are incompletely understood and appear to involve nuclease (D'Orazio et al., 2019), a helicase complex (Hashimoto et al., 2020; Matsuo et al., 2017; Sitron et al., 2017), and possibly ribosome rescue factors (Tsuboi et al., 2012). Although the roles and order of action of these factors is only partially resolved (Juszkiewicz et al., 2020; Matsuo et al., 2020), the requirement for ZNF598-mediated 40S ubiquitination (Garzia et al., 2017; Juszkiewicz and Hegde, 2017; Matsuo et al., 2017; Sundaramoorthy et al., 2017) together with the specificity of ZNF598 for ribosome collisions (Juszkiewicz et al., 2018) strongly argues that the collided ribosome is a key proxy of aberrant translation. Hence, collisions are an upstream event in both mRNA decay (Simms et al., 2017) and nascent protein quality control (Ikeuchi et al., 2019; Juszkiewicz et al., 2018).

Nearly all insight into ribosome stalling and the downstream consequences comes from the analysis of model substrates where specific stalling sequences have been introduced. However, the set of substrates that ordinarily engage RQC are poorly understood. The analysis of RNAs that crosslink to ZNF598 identified mRNAs, tRNAs, and rRNAs, consistent with its engagement of translating ribosome (Garzia et al., 2017). Strikingly, the tRNA that was strongly enriched decodes the AAA codon while no specific mRNAs were enriched above others. This observation can be rationalised because ribosomes are well established to stall during translation of poly(A) sequences (Chandrasekaran et al., 2019) and polyadenylation is known to occur inappropriately at near-cognate sites within the coding region (Guydosh and Green, 2017; Ozsolak et al., 2010; Pelechano et al., 2013). Thus, prematurely polyadenylated mRNAs are likely to be one physiologically relevant source of ribosome stalling, collisions, and ZNF598 engagement.

While the quality control events downstream of collisions have evolved to eliminate aberrant proteins and mRNAs, promiscuous triggering of these reactions from normal translation complexes would be detrimental. This might be why reticulocytes, which have exceptionally high ribosome densities on highly translated haemoglobin mRNAs, downregulate the collision sensor ZNF598 (Juszkiewicz and Hegde, 2017; Mills et al., 2016). Indeed, a detectable proportion of ribosomes in native reticulocyte polysomes are collided as judged by their ability to engage exogenously added ZNF598 and resistance to separation by exogenously added nuclease (Juszkiewicz and Hegde, 2017; Juszkiewicz et al., 2018).

Unlike the terminally differentiating reticulocyte with a minimal transcriptome and limited life-span, most cells cannot forgo translation quality control. How cells either minimise incidental collisions or avoid promiscuous activation of quality control from them is not clear. Although the precise frequency of incidental collisions is not well established, they are probably not insignificant considering the wide variation in elongation rates (Boersma et al., 2019; Ingolia et al., 2011; Riba et al., 2019; Yan et al., 2016), combined with closely spaced ribosomes on heavily translated mRNAs (Palade, 1955). Indeed, recent ribosome profiling experiments examining collided di-ribosomes suggest widespread collisions at numerous (potentially physiologic) pause sites throughout the transcriptome (Arpat et al., 2019; Han et al., 2020; Zhao et al., 2019). Here, we have taken an unbiased proteomic approach to identify cellular factors that selectively engage ribosome collisions. In addition to ZNF598, we describe a previously unappreciated response to ribosome collisions that complements the RQC pathway by dynamically adjusting translation initiation.

Cells exploit ribosome collision as an indirect indicator of an aberrant mRNA that cannot be successfully translated (Ikeuchi et al., 2019; Juszkiewicz et al., 2018; Simms et al., 2017). Accordingly, the recruitment of ZNF598 to such collisions can initiate mRNA and protein quality control pathways. However, collisions can also indicate the more innocuous situation of high ribosome density caused by overly frequent initiation. Our discovery of a collision-dependent cis-acting mechanism of inhibiting translation initiation suggests that cells monitor ribosome density at a per-mRNA level and dynamically tune initiation rates accordingly. Thus, ribosome collisions emerge as an important parameter that cells use to adjust the quality and quantity of translational output.

At an average density of one ribosome per ~60 codons and a highly variable translation rate that averages ~6 codons per second, collisions are inevitable (Reuveni et al., 2011; Tuller et al., 2010). This is because considerations of optimal resource utilisation dictate that translation efficiency and ribosome density should be at slightly sub-saturation levels (Reuveni et al., 2011). Consistent with this, electron microscopy images have long shown polysomes contain ribosomes that are typically separated by ~8–15 nm (Palade, 1955). In HEK293 cells growing under laboratory conditions, we estimate that ~2–5% of ribosomes are collided at any given moment based on the small but discernible disome peak seen on sucrose gradients after nuclease digestion (Juszkiewicz et al., 2018). These considerations suggest that a general detector of collisions must be within ~20 fold abundance of ribosomes.

Quantification in Hela cells indicates that EDF1 is present in ~1 million copies per cell (Kulak et al., 2014), roughly one-fourth to one-sixth the concentration of ribosomes as measured in the same study. By contrast, GIGYF2 and ZNF598 are estimated to be around 50-fold lower (Itzhak et al., 2016), although this may vary by cell type. We therefore posit that EDF1 is a general collision sensor that arrives first at any collision due to its very high abundance. If the collision is simply due to stochastic variations in ribosome spacing along an mRNA, elongation would continue, the ribosome spacing would quickly change, and EDF1 would dissociate without any action being taken. Because each round of elongation only takes ~200 milliseconds [i.e., an average of ~5 amino acids per second (Sokabe and Fraser, 2019)], these events would all be complete within a second or so.

If the collision is due to a genuine slowdown, EDF1 would be retained on collided ribosomes for longer, thereby stabilising the ZNF598-GIGYF2-4EHP complex at the collision (Figure 7). While GIGYF2 co-fractionation with collided ribosomes is diminished without EDF1, it is not lost entirely. Similarly, ZNF598 binding to collided ribosomes is substantially less stable but not eliminated in the absence of either EDF1 or GIGYF2. In each case, the ribosome interaction is collision-specific. This is consistent with a model in which the disome-EDF1 complex provides an optimal platform for ZNF598-GIGYF2-4EHP binding. This could occur in one of two ways: (i) a direct model where EDF1 provides part of a binding site for the ZNF598-GIGYF2-4EHP complex on collided ribosomes; (ii) an indirect model where EDF1 engagement of collided ribosomes changes the ribosome complex in a way that favours the bound state of the ZNF598-GIGYF2-4EHP complex. Future structural analysis of the complex engaged on a collided disome is required to address these details.

Figure 7

Download asset Open asset

Once recruited, GIGYF2 can repress translation initiation. As established previously, GIGYF2 forms a complex with 4EHP (Morita et al., 2012), which competes with the translation factor eIF4E for the 5’ cap (Cho et al., 2005). 4EHP is ordinarily not able to outcompete eIF4E (Zuberek et al., 2007) unless its local concentration in the vicinity of an mRNA is high. Such a model would explain how collisions can selectively inhibit translation initiation of the mRNA on which the collision occurred. Consistent with this idea, artificially recruiting the GIGYF2-4EHP complex to an mRNA is sufficient to inhibit translation initiation in cis (Hickey et al., 2019).

Because incidental collisions may be a pervasive feature across the transcriptome (Arpat et al., 2019; Han et al., 2020), the GIGYF2-4EHP axis is expected to impact the translation of a wide range of mRNAs. Earlier experiments in cells lacking GIGYF2 or 4EHP showed ~30% higher rates of translation globally (Morita et al., 2012). Strikingly, the native size of polysomes in 4EHP knockout tissue was increased by the size of 1–2 ribosomes, indicating that ribosome density on mRNAs is increased across the transcriptome.

Reduction of initiation rates via the GIGYF2-4EHP axis provides time for resolution of the collision without incurring further collisions. Resolution can occur in one of two ways. If the collision is incidental due to a physiologic pause in translation, elongation would soon resume, at which point EDF1 and the ZNF598-GIGYF2-4EHP complex would dissociate to allow initiation to begin again (Figure 7, left). Given the abundance and speed of elongation factor action, resumption of translation would typically occur in under a second. The broadest estimates of normal elongation rates span at most a ~20 fold range (Rodnina, 2016; Sokabe and Fraser, 2019). Thus, even in particularly unusual instances where a ribosome lingered for 20-fold longer than a normal elongation cycle, it would still re-start in ~4 s.

The longer a collision persists, the greater the likelihood that ZNF598 will ubiquitinate 40S proteins including eS10 (Figure 7, right). Depending on the ligase and other parameters, ubiquitination reactions can occur on the timescale of milliseconds to minutes (Pierce et al., 2009; Melvin et al., 2013). A time course of ZNF598 ubiquitination of ribosomes in vitro at close to physiologic concentrations indicate that it takes at least ~2–5 min to accumulate appreciable 40S ubiquitination (Juszkiewicz and Hegde, 2017). Because there are likely to be deubiquitinase(s) that can act on the ribosome (Garshott et al., 2020), 40S ubiquitination is not the commitment step to trigger RQC. Instead, ubiquitinated eS10 is needed for the ASC-1 complex (ASCC) to act (Juszkiewicz et al., 2020; Matsuo et al., 2020).

ASCC is a conserved complex containing the ASCC3 helicase (Slh1 in yeast) involved in resolving collided ribosomes (D'Orazio et al., 2019; Hashimoto et al., 2020; Ikeuchi et al., 2019; Juszkiewicz et al., 2020; Matsuo et al., 2017; Matsuo et al., 2020). Recent experiments suggest that by directly dissociating the lead ribosome into subunits, ASCC is the factor that irreversibly aborts translation and initiates RQC on the stalled ribosome (Juszkiewicz et al., 2020; Matsuo et al., 2020). Having resolved the troublesome ribosome, the downstream ribosomes are then free to elongate (Juszkiewicz et al., 2020).

Because the commitment to RQC is two steps removed from ZNF598 recruitment and involves the trans-acting ASCC, this limb of the response is likely to take between one to many minutes. By contrast, the simple binding reaction represented by 4EHP interacting with the 5’ cap in cis would occur at millisecond to low-second time scales given the constrained local concentrations of the interaction partners. Although we have not made all of these measurements, their estimates are based on well-established biochemical principles. Thus, we propose that collisions initiate a tiered response that begins with the conservative strategy of inhibiting initiation (the 1° response) followed later by the destructive strategy of initiating quality control (the 2° response). If the lead ribosome elongates at any point prior to ASCC action, the distinctive architecture of the collided disome would be lost, causing dissociation of the EDF1 and the ZNF598-GIGYF2-4EHP complex.

Even though ZNF598 is more stable on collisions containing EDF1 and GIGYF2, ZNF598-mediated eS10 ubiquitination does not require either protein. Hence, RQC can be effectively triggered without EDF1 or GIGYF2. Similarly, GIGYF2 does not require ZNF598 for either its stability on collisions or its effect on translation initiation. While future structural studies are needed to understand the physical relationship of these factors on the collided ribosome, our biochemical and functional analyses suggest that they do not compete for a common site.

Nevertheless, GIGYF2 and ZNF598 do functionally compete for the outcome downstream of ribosome collisions. This can be rationalised because each pathway reduces collisions needed by the other pathway’s output. The ZNF598 limb reduces collisions by ultimately splitting the lead ribosome. Thus, increased output through this limb means fewer collided ribosomes from which the GIGYF2-4EHP complex can inhibit initiation. On the other hand, increased output through the GIGYF2-4EHP limb reduces the initiation rate, reducing collisions on which ZNF598 can act.

The implication of this tiered model is that the ratio of GIGYF2 to ZNF598 effectively sets the persistence threshold at which collisions trigger quality control. This key ratio may vary across cell types as evidenced by the lack of ZNF598 in reticulocytes (Juszkiewicz et al., 2018; Mills et al., 2016) and its particularly high expression in oocytes (Grennan, 2006; www.genevisible.com). Reticulocytes may forego quality control in order to translate at maximal levels, while other cell types might show particularly sensitive quality control to minimise aberrant proteins. The regulatory aspect of how cells respond to collisions remains to be explored.

Although relatively little was previously known about EDF1, the yeast homolog Mbf1 was identified in a screen for genes that suppress frameshifting at ribosome stalling sequences (Wang et al., 2018). Given the recent finding that suggests ribosome collisions can cause frameshifting (Simms et al., 2019), our results provide a plausible explanation for the increased frameshifting observed in ∆Mbf1 yeast strains. By removing a factor that helps to suppress initiation in response to collisions, the collision rate would increase, resulting in increased frameshifting. Consistent with this idea, yeast has two GIGYF2 homologs (Smy2 and Syh1) whose overexpression suppresses translation of a stall-containing reporter (D'Orazio et al., 2019). Although yeast does not have an obvious 4EHP homolog, Smy2 interacts with Eap1, an inhibitor of eIF4E function (Sezen et al., 2009). Thus, it is likely that the cis-acting collision-mediated feedback loop described in our study is widely conserved.

If each collision is accompanied by a slight increase in the probability of frameshifting, minimising their occurrence would be important for avoiding the production of aberrant proteins. Whether the accumulation of aberrant frameshifted proteins is the basis for the neurodegenerative-like phenotype of partial GIGYF2 deficiency (Giovannone et al., 2009) remains to be explored. It is particularly intriguing that this phenotype resembles in some ways the partial deficiency of Listerin (Chu et al., 2009), which acts downstream of ZNF598 (Garzia et al., 2017; Juszkiewicz and Hegde, 2017; Sundaramoorthy et al., 2017). Thus, by acting to minimise (via the GIGYF2 limb) and resolve (via the ZNF598 limb) ribosome collisions, both pathways seem to protect the proteome from detrimental translation products that may contribute to neurodegenerative disease.

All data generated or analyzed during this study are included in the manuscript and supporting files. Source data have been provided for Figure 1. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD020239.

1. 1. Juszkiewicz S
   2. Slodkowicz G
   3. Lin Z
   4. Freire-Pritchett P
   5. Peak-Chew SY
   6. Hegde RS

   (2020) PRIDE

   ID PXD020239. Ribosome collisions trigger cis-acting feedback inhibition of translation initiation.

   https://www.ebi.ac.uk/pride/archive/projects/PXD020239

1. Preprint

   1. Arpat AB
   2. Liechti A
   3. Matos MD
   4. Dreos R
   5. Janich P
   6. Gatfield D

   (2019) Transcriptome-wide sites of collided ribosomes reveal principles of translational pausing

   bioRxiv.

   https://doi.org/10.1101/710061

   • Google Scholar
2. 1. Arthur L
   2. Pavlovic-Djuranovic S
   3. Smith-Koutmou K
   4. Green R
   5. Szczesny P
   6. Djuranovic S

   (2015) Translational control by lysine-encoding A-rich sequences

   Science Advances 1:e1500154.

   https://doi.org/10.1126/sciadv.1500154

   • PubMed
   • Google Scholar
3. 1. Becker T
   2. Armache J-P
   3. Jarasch A
   4. Anger AM
   5. Villa E
   6. Sieber H
   7. Motaal BA
   8. Mielke T
   9. Berninghausen O
   10. Beckmann R

   (2011) Structure of the no-go mRNA decay complex Dom34–Hbs1 bound to a stalled 80S ribosome

   Nature Structural & Molecular Biology 18:715–720.

   https://doi.org/10.1038/nsmb.2057

   • Google Scholar
4. 1. Bengtson MH
   2. Joazeiro CAP

   (2010) Role of a ribosome-associated E3 ubiquitin ligase in protein quality control

   Nature 467:470–473.

   https://doi.org/10.1038/nature09371

   • Google Scholar
5. 1. Boersma S
   2. Khuperkar D
   3. Verhagen BMP
   4. Sonneveld S
   5. Grimm JB
   6. Lavis LD
   7. Tanenbaum ME

   (2019) Multi-Color Single-Molecule Imaging Uncovers Extensive Heterogeneity in mRNA Decoding

   Cell 178:458–472.

   https://doi.org/10.1016/j.cell.2019.05.001

   • Google Scholar
6. 1. Brandman O
   2. Hegde RS

   (2016) Ribosome-associated protein quality control

   Nature Structural & Molecular Biology 23:7–15.

   https://doi.org/10.1038/nsmb.3147

   • Google Scholar
7. 1. Brown A
   2. Shao S
   3. Murray J
   4. Hegde RS
   5. Ramakrishnan V

   (2015) Structural basis for stop codon recognition in eukaryotes

   Nature 524:493–496.

   https://doi.org/10.1038/nature14896

   • Google Scholar
8. 1. Chakrabarti O
   2. Hegde RS

   (2009) Functional Depletion of Mahogunin by Cytosolically Exposed Prion Protein Contributes to Neurodegeneration

   Cell 137:1136–1147.

   https://doi.org/10.1016/j.cell.2009.03.042

   • Google Scholar
9. 1. Chandrasekaran V
   2. Juszkiewicz S
   3. Choi J
   4. Puglisi JD
   5. Brown A
   6. Shao S
   7. Ramakrishnan V
   8. Hegde RS

   (2019) Mechanism of ribosome stalling during translation of a poly(A) tail

   Nature Structural & Molecular Biology 26:1132–1140.

   https://doi.org/10.1038/s41594-019-0331-x

   • Google Scholar
10. 1. Chapat C
    2. Jafarnejad SM
    3. Matta-Camacho E
    4. Hesketh GG
    5. Gelbart IA
    6. Attig J
    7. Gkogkas CG
    8. Alain T
    9. Stern-Ginossar N
    10. Fabian MR
    11. Gingras AC
    12. Duchaine TF
    13. Sonenberg N

    (2017) Cap-binding protein 4ehp effects translation silencing by microRNAs

    PNAS 114:5425–5430.

    https://doi.org/10.1073/pnas.1701488114

    • PubMed
    • Google Scholar
11. 1. Cho PF
    2. Poulin F
    3. Cho-Park YA
    4. Cho-Park IB
    5. Chicoine JD
    6. Lasko P
    7. Sonenberg N

    (2005) A new paradigm for translational control: inhibition via 5'-3' mRNA tethering by bicoid and the eIF4E cognate 4ehp

    Cell 121:411–423.

    https://doi.org/10.1016/j.cell.2005.02.024

    • PubMed
    • Google Scholar
12. 1. Chu J
    2. Hong NA
    3. Masuda CA
    4. Jenkins BV
    5. Nelms KA
    6. Goodnow CC
    7. Glynne RJ
    8. Wu H
    9. Masliah E
    10. Joazeiro CAP
    11. Kay SA

    (2009) A mouse forward genetics screen identifies LISTERIN as an E3 ubiquitin ligase involved in neurodegeneration

    PNAS 106:2097–2103.

    https://doi.org/10.1073/pnas.0812819106

    • Google Scholar
13. 1. D'Orazio KN
    2. Wu CC
    3. Sinha N
    4. Loll-Krippleber R
    5. Brown GW
    6. Green R

    (2019) The endonuclease Cue2 cleaves mRNAs at stalled ribosomes during no go decay

    eLife 8:e49117.

    https://doi.org/10.7554/eLife.49117

    • PubMed
    • Google Scholar
14. 1. Garshott DM
    2. Sundaramoorthy E
    3. Leonard M
    4. Bennett EJ

    (2020) Distinct regulatory ribosomal ubiquitylation events are reversible and hierarchically organized

    eLife 9:e54032.

    https://doi.org/10.7554/eLife.54023

    • Google Scholar
15. 1. Garzia A
    2. Jafarnejad SM
    3. Meyer C
    4. Chapat C
    5. Gogakos T
    6. Morozov P
    7. Amiri M
    8. Shapiro M
    9. Molina H
    10. Tuschl T
    11. Sonenberg N

    (2017) The E3 ubiquitin ligase and RNA-binding protein ZNF598 orchestrates ribosome quality control of premature polyadenylated mRNAs

    Nature Communications 8:16056.

    https://doi.org/10.1038/ncomms16056

    • Google Scholar
16. 1. Giovannone B
    2. Tsiaras WG
    3. de la Monte S
    4. Klysik J
    5. Lautier C
    6. Karashchuk G
    7. Goldwurm S
    8. Smith RJ

    (2009) GIGYF2 gene disruption in mice results in neurodegeneration and altered insulin-like growth factor signaling

    Human Molecular Genetics 18:4629–4639.

    https://doi.org/10.1093/hmg/ddp430

    • Google Scholar
17. 1. Grennan AK

    (2006) Genevestigator. Facilitating web-based gene-expression analysis

    Plant Physiology 141:1164–1166.

    https://doi.org/10.1104/pp.104.900198

    • PubMed
    • Google Scholar
18. 1. Guydosh NR
    2. Green R

    (2017) Translation of poly(A) tails leads to precise mRNA cleavage

    RNA 23:749–761.

    https://doi.org/10.1261/rna.060418.116

    • Google Scholar
19. 1. Han P
    2. Shichino Y
    3. Schneider-Poetsch T
    4. Mito M
    5. Hashimoto S
    6. Udagawa T
    7. Kohno K
    8. Yoshida M
    9. Mishima Y
    10. Inada T
    11. Iwasaki S

    (2020) Genome-wide Survey of Ribosome Collision

    Cell Reports 31:107610.

    https://doi.org/10.1016/j.celrep.2020.107610

    • Google Scholar
20. 1. Harding HP
    2. Ordonez A
    3. Allen F
    4. Parts L
    5. Inglis AJ
    6. Williams RL
    7. Ron D

    (2019) The ribosomal P-stalk couples amino acid starvation to GCN2 activation in mammalian cells

    eLife 8:e50149.

    https://doi.org/10.7554/eLife.50149

    • PubMed
    • Google Scholar
21. 1. Hashimoto S
    2. Sugiyama T
    3. Yamazaki R
    4. Nobuta R
    5. Inada T

    (2020) Identification of a novel trigger complex that facilitates ribosome-associated quality control in mammalian cells

    Scientific Reports 10:3422.

    https://doi.org/10.1038/s41598-020-60241-w

    • Google Scholar
22. Preprint

    1. Hickey KL
    2. Dickson K
    3. Cogan JZ
    4. Replogle JM
    5. Schoof M
    6. D’Orazio KN
    7. Sinha NK
    8. Frost A
    9. Green R
    10. Kostova KK

    (2019) GIGYF2 and 4ehp inhibit translation initiation of defective messenger RNAs to assist Ribosome-Associated quality control

    bioRxiv.

    https://doi.org/10.1101/792994

    • Google Scholar
23. 1. Ikeuchi K
    2. Tesina P
    3. Matsuo Y
    4. Sugiyama T
    5. Cheng J
    6. Saeki Y
    7. Tanaka K
    8. Becker T
    9. Beckmann R
    10. Inada T

    (2019) Collided ribosomes form a unique structural interface to induce Hel2‐driven quality control pathways

    The EMBO Journal 38:e100276.

    https://doi.org/10.15252/embj.2018100276

    • Google Scholar
24. 1. Inglis AJ
    2. Masson GR
    3. Shao S
    4. Perisic O
    5. McLaughlin SH
    6. Hegde RS
    7. Williams RL

    (2019) Activation of GCN2 by the ribosomal P-stalk

    PNAS 116:4946–4954.

    https://doi.org/10.1073/pnas.1813352116

    • PubMed
    • Google Scholar
25. 1. Ingolia NT
    2. Lareau LF
    3. Weissman JS

    (2011) Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of mammalian proteomes

    Cell 147:789–802.

    https://doi.org/10.1016/j.cell.2011.10.002

    • PubMed
    • Google Scholar
26. 1. Itzhak DN
    2. Tyanova S
    3. Cox J
    4. Borner GH

    (2016) Global, quantitative and dynamic mapping of protein subcellular localization

    eLife 5:e16950.

    https://doi.org/10.7554/eLife.16950

    • PubMed
    • Google Scholar
27. 1. Joazeiro CAP

    (2019) Mechanisms and functions of ribosome-associated protein quality control

    Nature Reviews Molecular Cell Biology 20:368–383.

    https://doi.org/10.1038/s41580-019-0118-2

    • PubMed
    • Google Scholar
28. 1. Juszkiewicz S
    2. Chandrasekaran V
    3. Lin Z
    4. Kraatz S
    5. Ramakrishnan V
    6. Hegde RS

    (2018) ZNF598 Is a Quality Control Sensor of Collided Ribosomes

    Molecular Cell 72:469–481.

    https://doi.org/10.1016/j.molcel.2018.08.037

    • Google Scholar
29. 1. Juszkiewicz S
    2. Speldewinde SH
    3. Wan L
    4. Svejstrup JQ
    5. Hegde RS

    (2020) The ASC-1 complex disassembles collided ribosomes

    Molecular Cell S1097-2765:30391–30399.

    https://doi.org/10.1016/j.molcel.2020.06.006

    • Google Scholar
30. 1. Juszkiewicz S
    2. Hegde RS

    (2017) Initiation of quality control during poly(A) Translation requires Site-Specific ribosome ubiquitination

    Molecular Cell 65:743–750.

    https://doi.org/10.1016/j.molcel.2016.11.039

    • PubMed
    • Google Scholar
31. 1. Kapur M
    2. Ackerman SL

    (2018) mRNA Translation Gone Awry: Translation Fidelity and Neurological Disease

    Trends in Genetics 34:218–231.

    https://doi.org/10.1016/j.tig.2017.12.007

    • Google Scholar
32. 1. Koutmou KS
    2. Schuller AP
    3. Brunelle JL
    4. Radhakrishnan A
    5. Djuranovic S
    6. Green R

    (2015) Ribosomes slide on lysine-encoding homopolymeric A stretches

    eLife 4:e05534.

    https://doi.org/10.7554/eLife.05534

    • Google Scholar
33. 1. Kulak NA
    2. Pichler G
    3. Paron I
    4. Nagaraj N
    5. Mann M

    (2014) Minimal, encapsulated proteomic-sample processing applied to copy-number estimation in eukaryotic cells

    Nature Methods 11:319–324.

    https://doi.org/10.1038/nmeth.2834

    • PubMed
    • Google Scholar
34. 1. Labbadia J
    2. Morimoto RI

    (2015) The Biology of Proteostasis in Aging and Disease

    Annual Review of Biochemistry 84:435–464.

    https://doi.org/10.1146/annurev-biochem-060614-033955

    • Google Scholar
35. 1. Liu B
    2. Han Y
    3. Qian S-B

    (2013) Cotranslational Response to Proteotoxic Stress by Elongation Pausing of Ribosomes

    Molecular Cell 49:453–463.

    https://doi.org/10.1016/j.molcel.2012.12.001

    • Google Scholar
36. 1. Matsuo Y
    2. Ikeuchi K
    3. Saeki Y
    4. Iwasaki S
    5. Schmidt C
    6. Udagawa T
    7. Sato F
    8. Tsuchiya H
    9. Becker T
    10. Tanaka K
    11. Ingolia NT
    12. Beckmann R
    13. Inada T

    (2017) Ubiquitination of stalled ribosome triggers ribosome-associated quality control

    Nature Communications 8:159.

    https://doi.org/10.1038/s41467-017-00188-1

    • Google Scholar
37. 1. Matsuo Y
    2. Tesina P
    3. Nakajima S
    4. Mizuno M
    5. Endo A
    6. Buschauer R
    7. Cheng J
    8. Shounai O
    9. Ikeuchi K
    10. Saeki Y
    11. Becker T
    12. Beckmann R
    13. Inada T

    (2020) RQT complex dissociates ribosomes collided on endogenous RQC substrate SDD1

    Nature Structural & Molecular Biology 27:323–332.

    https://doi.org/10.1038/s41594-020-0393-9

    • Google Scholar
38. 1. Melvin AT
    2. Woss GS
    3. Park JH
    4. Dumberger LD
    5. Waters ML
    6. Allbritton NL

    (2013) A comparative analysis of the ubiquitination kinetics of multiple degrons to identify an ideal targeting sequence for a proteasome reporter

    PLOS ONE 8:e78082.

    https://doi.org/10.1371/journal.pone.0078082

    • PubMed
    • Google Scholar
39. 1. Mills EW
    2. Wangen J
    3. Green R
    4. Ingolia NT

    (2016) Dynamic Regulation of a Ribosome Rescue Pathway in Erythroid Cells and Platelets

    Cell Reports 17:1–10.

    https://doi.org/10.1016/j.celrep.2016.08.088

    • Google Scholar
40. 1. Morita M
    2. Ler LW
    3. Fabian MR
    4. Siddiqui N
    5. Mullin M
    6. Henderson VC
    7. Alain T
    8. Fonseca BD
    9. Karashchuk G
    10. Bennett CF
    11. Kabuta T
    12. Higashi S
    13. Larsson O
    14. Topisirovic I
    15. Smith RJ
    16. Gingras AC
    17. Sonenberg N

    (2012) A novel 4EHP-GIGYF2 translational repressor complex is essential for mammalian development

    Molecular and Cellular Biology 32:3585–3593.

    https://doi.org/10.1128/MCB.00455-12

    • PubMed
    • Google Scholar
41. 1. Ozsolak F
    2. Kapranov P
    3. Foissac S
    4. Kim SW
    5. Fishilevich E
    6. Monaghan AP
    7. John B
    8. Milos PM

    (2010) Comprehensive polyadenylation site maps in yeast and human reveal pervasive alternative polyadenylation

    Cell 143:1018–1029.

    https://doi.org/10.1016/j.cell.2010.11.020

    • PubMed
    • Google Scholar
42. 1. Palade GE

    (1955) A small particulate component of the cytoplasm

    The Journal of Biophysical and Biochemical Cytology 1:59–68.

    https://doi.org/10.1083/jcb.1.1.59

    • PubMed
    • Google Scholar
43. 1. Pelechano V
    2. Wei W
    3. Steinmetz LM

    (2013) Extensive transcriptional heterogeneity revealed by isoform profiling

    Nature 497:127–131.

    https://doi.org/10.1038/nature12121

    • PubMed
    • Google Scholar
44. 1. Peter D
    2. Weber R
    3. Sandmeir F
    4. Wohlbold L
    5. Helms S
    6. Bawankar P
    7. Valkov E
    8. Igreja C
    9. Izaurralde E

    (2017) GIGYF1/2 proteins use auxiliary sequences to selectively bind to 4EHP and repress target mRNA expression

    Genes & Development 31:1147–1161.

    https://doi.org/10.1101/gad.299420.117

    • Google Scholar
45. 1. Pierce NW
    2. Kleiger G
    3. Shan S-ou
    4. Deshaies RJ

    (2009) Detection of sequential polyubiquitylation on a millisecond timescale

    Nature 462:615–619.

    https://doi.org/10.1038/nature08595

    • Google Scholar
46. 1. Pisareva VP
    2. Skabkin MA
    3. Hellen CUT
    4. Pestova TV
    5. Pisarev AV

    (2011) Dissociation by Pelota, Hbs1 and ABCE1 of mammalian vacant 80S ribosomes and stalled elongation complexes

    The EMBO Journal 30:1804–1817.

    https://doi.org/10.1038/emboj.2011.93

    • Google Scholar
47. 1. Princiotta MF
    2. Finzi D
    3. Qian SB
    4. Gibbs J
    5. Schuchmann S
    6. Buttgereit F
    7. Bennink JR
    8. Yewdell JW

    (2003) Quantitating protein synthesis, degradation, and endogenous antigen processing

    Immunity 18:343–354.

    https://doi.org/10.1016/S1074-7613(03)00051-7

    • PubMed
    • Google Scholar
48. 1. Ran FA
    2. Hsu PD
    3. Wright J
    4. Agarwala V
    5. Scott DA
    6. Zhang F

    (2013) Genome engineering using the CRISPR-Cas9 system

    Nature Protocols 8:2281–2308.

    https://doi.org/10.1038/nprot.2013.143

    • PubMed
    • Google Scholar
49. 1. Rauniyar N
    2. Yates JR

    (2014) Isobaric Labeling-Based Relative Quantification in Shotgun Proteomics

    Journal of Proteome Research 13:5293–5309.

    https://doi.org/10.1021/pr500880b

    • Google Scholar
50. 1. Reuveni S
    2. Meilijson I
    3. Kupiec M
    4. Ruppin E
    5. Tuller T

    (2011) Genome-scale analysis of translation elongation with a ribosome flow model

    PLOS Computational Biology 7:e1002127.

    https://doi.org/10.1371/journal.pcbi.1002127

    • PubMed
    • Google Scholar
51. 1. Riba A
    2. Di Nanni N
    3. Mittal N
    4. Arhné E
    5. Schmidt A
    6. Zavolan M

    (2019) Protein synthesis rates and ribosome occupancies reveal determinants of translation elongation rates

    PNAS 116:15023–15032.

    https://doi.org/10.1073/pnas.1817299116

    • PubMed
    • Google Scholar
52. 1. Rodnina MV

    (2016) The ribosome in action: tuning of translational efficiency and protein folding

    Protein Science 25:1390–1406.

    https://doi.org/10.1002/pro.2950

    • PubMed
    • Google Scholar
53. 1. Roux PP
    2. Topisirovic I

    (2018) Signaling pathways involved in the regulation of mRNA translation

    Molecular and Cellular Biology 38:e00070-18.

    https://doi.org/10.1128/MCB.00070-18

    • PubMed
    • Google Scholar
54. 1. Schmidt C
    2. Kowalinski E
    3. Shanmuganathan V
    4. Defenouillère Q
    5. Braunger K
    6. Heuer A
    7. Pech M
    8. Namane A
    9. Berninghausen O
    10. Fromont-Racine M
    11. Jacquier A
    12. Conti E
    13. Becker T
    14. Beckmann R

    (2016) The cryo-EM structure of a ribosome-Ski2-Ski3-Ski8 helicase complex

    Science 354:1431–1433.

    https://doi.org/10.1126/science.aaf7520

    • PubMed
    • Google Scholar
55. 1. Sezen B
    2. Seedorf M
    3. Schiebel E

    (2009) The SESA network links duplication of the yeast centrosome with the protein translation machinery

    Genes & Development 23:1559–1570.

    https://doi.org/10.1101/gad.524209

    • Google Scholar
56. 1. Shalgi R
    2. Hurt JA
    3. Krykbaeva I
    4. Taipale M
    5. Lindquist S
    6. Burge CB

    (2013) Widespread regulation of translation by elongation pausing in heat shock

    Molecular Cell 49:439–452.

    https://doi.org/10.1016/j.molcel.2012.11.028

    • PubMed
    • Google Scholar
57. 1. Shao S
    2. Brown A
    3. Santhanam B
    4. Hegde RS

    (2015) Structure and assembly pathway of the ribosome quality control complex

    Molecular Cell 57:433–444.

    https://doi.org/10.1016/j.molcel.2014.12.015

    • PubMed
    • Google Scholar
58. 1. Shao S
    2. Murray J
    3. Brown A
    4. Taunton J
    5. Ramakrishnan V
    6. Hegde RS

    (2016) Decoding mammalian Ribosome-mRNA states by translational GTPase complexes

    Cell 167:1229–1240.

    https://doi.org/10.1016/j.cell.2016.10.046

    • PubMed
    • Google Scholar
59. 1. Sharma A
    2. Mariappan M
    3. Appathurai S
    4. Hegde RS

    (2010) In vitro dissection of protein translocation into the mammalian endoplasmic reticulum

    Methods in Molecular Biology 619:339–363.

    https://doi.org/10.1007/978-1-60327-412-8_20

    • PubMed
    • Google Scholar
60. 1. Sharma P
    2. Yan F
    3. Doronina VA
    4. Escuin-Ordinas H
    5. Ryan MD
    6. Brown JD

    (2012) 2A peptides provide distinct solutions to driving stop-carry on translational recoding

    Nucleic Acids Research 40:3143–3151.

    https://doi.org/10.1093/nar/gkr1176

    • Google Scholar
61. 1. Shoemaker CJ
    2. Eyler DE
    3. Green R

    (2010) Dom34:hbs1 promotes subunit dissociation and peptidyl-tRNA drop-off to initiate no-go decay

    Science 330:369–372.

    https://doi.org/10.1126/science.1192430

    • PubMed
    • Google Scholar
62. 1. Shoemaker CJ
    2. Green R

    (2012) Translation drives mRNA quality control

    Nature Structural & Molecular Biology 19:594–601.

    https://doi.org/10.1038/nsmb.2301

    • PubMed
    • Google Scholar
63. 1. Simms CL
    2. Yan LL
    3. Zaher HS

    (2017) Ribosome collision is critical for quality control during No-Go decay

    Molecular Cell 68:361–373.

    https://doi.org/10.1016/j.molcel.2017.08.019

    • PubMed
    • Google Scholar
64. 1. Simms CL
    2. Yan LL
    3. Qiu JK
    4. Zaher HS

    (2019) Ribosome collisions result in +1 Frameshifting in the Absence of No-Go Decay

    Cell Reports 28:1679–1689.

    https://doi.org/10.1016/j.celrep.2019.07.046

    • PubMed
    • Google Scholar
65. 1. Sitron CS
    2. Park JH
    3. Brandman O

    (2017) Asc1, Hel2, and Slh1 couple translation arrest to nascent chain degradation

    RNA 23:798–810.

    https://doi.org/10.1261/rna.060897.117

    • PubMed
    • Google Scholar
66. 1. Sokabe M
    2. Fraser CS

    (2019) Toward a kinetic understanding of eukaryotic translation

    Cold Spring Harbor Perspectives in Biology 11:a032706.

    https://doi.org/10.1101/cshperspect.a032706

    • PubMed
    • Google Scholar
67. 1. Sundaramoorthy E
    2. Leonard M
    3. Mak R
    4. Liao J
    5. Fulzele A
    6. Bennett EJ

    (2017) ZNF598 and RACK1 Regulate Mammalian Ribosome-Associated Quality Control Function by Mediating Regulatory 40S Ribosomal Ubiquitylation

    Molecular Cell 65:751–760.

    https://doi.org/10.1016/j.molcel.2016.12.026

    • Google Scholar
68. 1. Thompson A
    2. Schäfer J
    3. Kuhn K
    4. Kienle S
    5. Schwarz J
    6. Schmidt G
    7. Neumann T
    8. Johnstone R
    9. Mohammed AK
    10. Hamon C

    (2003) Tandem mass tags: a novel quantification strategy for comparative analysis of complex protein mixtures by MS/MS

    Analytical Chemistry 75:1895–1904.

    https://doi.org/10.1021/ac0262560

    • PubMed
    • Google Scholar
69. 1. Tsuboi T
    2. Kuroha K
    3. Kudo K
    4. Makino S
    5. Inoue E
    6. Kashima I
    7. Inada T

    (2012) Dom34:hbs1 plays a general role in quality-control systems by dissociation of a stalled ribosome at the 3' end of aberrant mRNA

    Molecular Cell 46:518–529.

    https://doi.org/10.1016/j.molcel.2012.03.013

    • PubMed
    • Google Scholar
70. 1. Tuller T
    2. Carmi A
    3. Vestsigian K
    4. Navon S
    5. Dorfan Y
    6. Zaborske J
    7. Pan T
    8. Dahan O
    9. Furman I
    10. Pilpel Y

    (2010) An evolutionarily conserved mechanism for controlling the efficiency of protein translation

    Cell 141:344–354.

    https://doi.org/10.1016/j.cell.2010.03.031

    • PubMed
    • Google Scholar
71. 1. Wang J
    2. Zhou J
    3. Yang Q
    4. Grayhack EJ

    (2018) Multi-protein bridging factor 1(Mbf1), Rps3 and Asc1 prevent stalled ribosomes from frameshifting

    eLife 7:e39637.

    https://doi.org/10.7554/eLife.39637

    • PubMed
    • Google Scholar
72. 1. Wolff S
    2. Weissman JS
    3. Dillin A

    (2014) Differential scales of protein quality control

    Cell 157:52–64.

    https://doi.org/10.1016/j.cell.2014.03.007

    • PubMed
    • Google Scholar
73. 1. Yan X
    2. Hoek TA
    3. Vale RD
    4. Tanenbaum ME

    (2016) Dynamics of translation of single mRNA molecules in Vivo

    Cell 165:976–989.

    https://doi.org/10.1016/j.cell.2016.04.034

    • PubMed
    • Google Scholar
74. 1. Yanagitani K
    2. Kimata Y
    3. Kadokura H
    4. Kohno K

    (2011) Translational Pausing Ensures Membrane Targeting and Cytoplasmic Splicing of XBP1u mRNA

    Science 331:586–589.

    https://doi.org/10.1126/science.1197142

    • Google Scholar
75. Preprint

    1. Zhao T
    2. Chen Y
    3. Wang J
    4. Chen S
    5. Qian W

    (2019) Disome-seq reveals sequence-mediated coupling of translational pauses and protein structures

    bioRxiv.

    https://doi.org/10.1101/746875

    • Google Scholar
76. 1. Zuberek J
    2. Kubacka D
    3. Jablonowska A
    4. Jemielity J
    5. Stepinski J
    6. Sonenberg N
    7. Darzynkiewicz E

    (2007) Weak binding affinity of human 4EHP for mRNA cap analogs

    RNA 13:691–697.

    https://doi.org/10.1261/rna.453107

    • Google Scholar

Acceptance summary:

This article suggests a hitherto elusive mechanism implicated in resolving ribosome collisions that appears to be distinct from the previously described quality control mechanism wherein aborted translation is sensed by ZNF598. To this end, the authors provide evidence that recruitment of EDF1 to collided ribosomes triggers inhibition of translation initiation in cooperation with GIGYF2 and 4EHP. Future studies are warranted to establish the potential physiological role(s) of this EDF1-driven response to ribosome stalling.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Ribosome collisions trigger cis-acting feedback inhibition of translation initiation" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Ivan Topisirovic as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife at this stage.

There was general enthusiasm pertinent to the potential role of the EDF1-GIGYF2-4EHP axis in sensing and resolving ribosome collisions. Unfortunately, however, the reviewers felt that the ribosome profiling studies suffered from insufficient statistical rigor due to the lack of biological replicates. In addition, it was found that the mechanistic data corroborating the latter model were somewhat preliminary and/or open to alternative explanations. Specifically, it was thought that the analysis of ribosome profiling data should be improved by including independent biological replicates and statistically rigorous methodology. This should be complemented by orthogonal validation of ribosome profiling data and ZNF598-dependent stalls. In addition, experiments corroborating the proposed model would also benefit from sufficient replication. Important controls outlined in each of the reviewer's comments should also be included. Addressing issues raised regarding eS10 ubiqutination as well as the potential competition between GIGYF2-4EHP binding to EDF1 vs. ZNF598 would also be appreciated. The latter experiments should be carried out using more direct approaches (e.g. co-IP). Finally, excluding and/or considering alternative explanations pertinent to studies employing emetine was also thought to be warranted.

We realize that this news will be disappointing. If you plan to address these issues and resubmit this study to eLife as a new submission, we will try to expedite the review process by sending the manuscript to the same reviewers that reviewed the initial submission and include your responses to the reviewer comments from this round of evaluation. We hope you will find these comments constructive in planning your next steps.

Reviewer #1

In this study, Juszkiewicz et al. provide evidence suggesting a previously unappreciated mechanism of resolving ribosome collisions centred on EDF1. This mechanism appears to be distinct form the quality control mechanism wherein aborted translation is sensed by ZNF598. To this end, the authors provide evidence that recruitment of EDF1 to collided ribosomes may inhibit translation initiation via recruitment of GIGYF2 and 4EHP. Based on these data, it is proposed that EDF1 examines mRNAs for increased ribosome density, followed by the recruitment of GIGYF2 to block initiation and facilitate resolving of ribosome collisions. Failure of this mechanism is suggested to be followed by induction of ZNF598-directed mechanism. Overall, it was found that this study is of broad potential interest as it describes a hitherto underexplored mechanism of resolution of ribosomal stalling and collisions. Notwithstanding general enthusiasm for the model, it was found that this study falls short when it comes to including a number of important controls as well as biological replicates to support authors' conclusions. Specific comments are outlined below:

Major comments:

– Several major issues were observed concerning inadequate scientific rigor, including insufficient number of replicates. It is stated in the legend of Figure 1—figure supplement 1 that ribosome profiling experiments were done in 2 technical replicates, which, if not a typo, suggest a single biological replicate. This precludes any meaningful statistical analysis. It is also expected that high reproducibility will be obtained from technical replicates, and it is thus not clear why technical replicates were used en lieu of biological replicates.

– A number of additional issues were noted pertinent to ribosome profiling analysis:

"To determine whether a subset of mRNAs are more dependent on ZNF598 than others, we tabulated the KO/WT ratio of CFr for each ORF. This value was plotted relative to that ORF's total dRPFs as an indicator of sampling depth, and hence reliability of the measurement (Figure 1C). 339 ORFs were higher (defined as ZNF598-dependent ORFs) and 159 ORFs lower than a significance interval defined by two standard deviations flanking a rolling mean".

This is lacking any statistical rigor as there is no adjustment for multiple testing. It would be more appropriate to perform biological replicates and to statistics on these.

"When normalized to the number of background rRNA reads as an internal control..."

Using rRNA to normalize is highly inadvisable as rRNA typically varies for a number of reasons. It also seems that the way rRNA reads are used to normalize the data, does not affect dRPFs to mRPFs ratio.

Figure 1—figure supplement 2: My opinion is that the correct approach here is to establish whether there are more dRPFs when TE adjusted for mRNA length. To this end, Y axis should represent dRPFs, while X axis should represent translational efficiencies calculated as residuals from a regression between mRPFs and mRNA levels (reads only from the coding region should be used). This will adjust for both length of the mRNA and the expression level.

"One explanation for the absence of a correlation is if collisions on highly translated mRNAs are efficiently resolved by ZNF598-triggered disassembly. However, a matched ribosome profiling dataset from ZNF598 knockout (KO) cells showed a similar range of CFr and no correlation with TE (Figure 1—figure supplement 2A, orange dots). This indicates that highly translated mRNAs (as defined by high TE) neither show a higher than average basal collision frequency nor are they preferentially affected in ZNF598 KO cells"

This seems as over-interpretation given how these effects were calculated, and the inherent noisiness of the assay. To make this claims RUST analysis should also be performed.

"…would be more likely to occur on mRNAs with high ribosome density."

It seems equally plausible that local features of the CDS may explain this phenomenon.

"Due to normalization, the above analysis cannot inform about any uniform and wholesale differences in collision frequency between WT versus KO cell"

The authors should corroborate this statement as their data indicate largely uniform effects.

"As exemplified by HNRNPU and RPS4X, some ORFs that did not score as hits when averaging CFr across the ORF nevertheless showed individual collision site(s) that are strongly dependent on ZNF598 (Figure 2C)."

How were these identified? Do they occur more than by chance?

"Such examples indicate that disome profiling is capable of identifying highly specific and strongly ZNF598-dependent stall sites with codon precision".

Seems overstated without any systematic analysis and validation that these indeed represent true events.

"Third, the collision frequency is remarkably similar across a wide range of translation efficiencies"

Data to support this statement were found to be insufficient.

Several other weaknesses were observed:

– "In addition to ZNF598 as the most collision-specific interaction partner, we identified eight other candidate collision-specific proteins."

Can authors explain how were these proteins identified as candidates?

– "Only two of the candidates, EDF1 and GIGYF2, were collision-specific interactors in both WT and ZNF598 KO cells"

How was this calculated/determined?

– Figure 3D – Western blot indicating the extent of ZNF598 depletion is missing.

Reviewer #2

Ribosome stalling and collisions, as well ribosome-associated quality control have drawn considerable attention over the last several years. It was postulated that major substrate for this ribosome rescue pathway are ribosomes stalled on damaged and prematurely polyadenylated mRNAs. The stalls are resolved by no-go decay factors, Pelota-Hbs1, and ZNF598(Hel 2), however there are still gaps in the mechanism of this process. In this manuscript, Juszkiewicz and colleagues, demonstrate an interesting observation that the ribosome collisions may initiate translation initiation inhibition in cis independently of ZNF598 pathway.

Unfortunately, despite the clear support that EDF1 is involved in ribosome collisions and resolution of this process, most of the claims are not clearly or fully supported by the presented data.

1) The complete analyses of ribosome profiling data and dRPF/mRPF ratio has multiple limitations. It is not clear whether these data are analyzed for the sequencing bias and normalization by dividing both sets with same rRNA background reads is problematic. The three conclusions drawn from these data are as such not supported.

2) Authors never report that experiments shown in Figure 3C and Figure 5B are from the same set of data, but they clearly use the same blots in these two figures. Have authors did replicates of these experiments?

3) It is not completely clear how eS10 gets ubiquitinated by ZNF598 in cells missing GIGYF2 (Figure 6). The supplement Figure 6 shows no recruitment of ZNF598 to poly-ribosomes in those same conditions with a rather massive recruitment in case of the WT cells. Authors argue that this could be due to transient interaction of ZNF598 in cells missing either EDF1 or GIGYF2. Is es10 ubiquitinated in HEK293 cells missing ZNF598, or Caco-2 and MCF-7 cells?

4) Original report by Morita et al., 2012. has 4EHP-GIGYF2-ZNF598 complex as translational repressor with direct co-IPs between all components and especially between 4EHP and ZNF598 as well as GIGYF2 and ZNF598. Is EDF1 binding to GIGYF2-4EHP and competing with ZNF598? There is no co-IP studies shown on any components so the reader is left to believe on interactions through the polysome profiles or functional assays.

Reviewer #3

The authors present a study on the dynamics of ribosome collisions and how modulations of the RQC pathway aim to resolve such aberrant events. By using a combination of disome-seq to identify collisions, clever reporter constructs and unbiased proteomics they shed light on different branching of the RQC pathway. Despite some important shortcomings pointed out in the following comments, the methods and conclusions are sound, the paper is well written, and it represent a strong contribution to the field.

1) Steady-state RNA abundance in the different tested conditions has not been measured, making it impossible to appreciate possible changes in translation efficiency for the transcripts displaying collision events. The authors should show that such concern is not valid (the plot in Figure 2B does not represent a sufficient assessment), by experimental methods or analysis of other datasets. If not possible, the Authors should acknowledge this significant shortcoming, and revisit some of their conclusions where necessary.

2) The authors should compare their findings with recent work outlining the crosstalk between translation initiation and the RQC: https://www.biorxiv.org/content/10.1101/792994v1

3) The authors might want to explore the potential sequence/structure features underlying pausing collisions events and their potential in discriminating between ZNF598-dependent and other pausing events. Results from recent work on detecting pausing events should be considered (e.g. https://www.biorxiv.org/content/10.1101/746875v2 , https://www.biorxiv.org/content/10.1101/710061v1 , https://www.biorxiv.org/content/10.1101/710491v1).

4) In the emetine treated samples, the percentage of dRPF reads mapping to the ORF greatly increases. Do the authors observe robust pausing/collisions outside coding regions in the different conditions, or do they represent possible contaminants?

5) Figure 3C shows how both EDF1 and GIGYF2 migrate to the polysome fraction under emetine treatment. However, this might represent a more general phenomenon where many more protein complexes migrate to the polysome fraction. Can the author comment on this possibility?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Ribosome collisions trigger cis-acting feedback inhibition of translation initiation" for further consideration by eLife. Your revised article has been evaluated by Suzanne Pfeffer (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining textual issues that need to be addressed before acceptance, as outlined below:

1) It was thought that the authors should either tone down some statements or discuss alternative explanations of observed phenomena. In particular, the interpretation that EDF1 recruits GIGYF2 complex at collided ribosomes was found to be insufficiently supported by data. To this end, it was found that although the IP experiments support the role of EDF1 in stabilization of GIGYF2 complex binding to collided ribosomes, their role in its recruitment remains unclear. This should be clearly stated throughout the manuscript and depicted in the model shown in Figure 5. It was also thought that Figure 7 should be removed as this model would require further kinetic studies that, as we agree, are out of the scope of this study.

2) Although highly appreciated, several issues were raised regarding frame-shifting studies. For instance, it appears that EDF1 depletion favours -1 frameshifting which cannot be fully explained by the collision model. Moreover, the authors cite studies that indicate that frameshifting should be in +1 direction. Discussion of these apparent discrepancies appears to be warranted. Potential effects of frameshifting on NMD activation should also be discussed. Finally, RFP signal and RFP/GFP ratios should be presented along with GFP for EDF1 knockdown and KO cells (Figure 4 and Figure 4—figure supplement 1) to further facilitate interpretation of results in Figure 6 and Figure 6—figure supplement 3.

3) It was also thought that the authors should specify number of replicates and precise criteria for some statistical tests. For example, 1.6 fold threshold was found to be quite arbitrary, and the authors should clearly elaborate on criteria to use such cutoff. MS data should be made publicly available and identified factors that are enriched/depleted on collided ribosomes should be at least briefly discussed. The authors are are also encouraged to add all appropriate controls (e.g. showing that ZNF598 is indeed knocked out in the experiment in question rather than referring to previous publications where these lines were established)

4) At times, interpretation of data was thought to require some clarification. Particularly, Figure 2—figure supplement 1A shows apparent increase in mRNA levels, albeit in the text the authors state that there is a lack of the potential mRNA abundance changes. To this end, it was thought that appropriate significance testing is required to corroborate the authors' conclusions, and if it turns out that the changes in mRNA abundance are significant, this should be commented on. Moreover, it is argued that 4EHP, but not GIGYF2 depletion increases translation. However, considering the scale in Figures 2C vs. 2D, it appears that the changes are of comparable amplitude but in the opposite directions. Based on this, significance testing appears warranted.

5) The authors are advised to verify some of the references. For instance in Mills et al., 2016 study does not appear to demonstrate ZNF598 loss during maturation of reticulocytes.

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

We appreciate the overall enthusiasm for our discovery of an EDF1-GIGYF2-4EHP mechanism of collision-specific translation repression. A large majority of the concerns related to inadequate analysis of a ribosome profiling experiment that aimed to provide a survey of collisions across the transcriptome. Because this experiment is tangential to the key findings of a new collision dependent response, we and the handling editor agreed that the study would be better if the profiling was omitted and the rest of the manuscript strengthened. In the response below, we address the non-profiling comments; the comments relating to profiling are no longer relevant.

Key issues identified by the Editor:

1) Is EDF1 binding to GIGYF2-4EHP and competing with ZNF598? Moreover, what is the status of GIGYF2 complexes (with ZNF598 vs. EDF1) in emetine treated vs. control cells? Is ZNF598GIGYF2-4EHP still preserved in EDF1 knockout cells? Is interaction between ZNF598 and 4EHP (Morita et al., 2012, MCB) affected in GIGYF2 knockout cells? IP experiments were thought to be an appropriate approach here.

Using co-IP experiments, we examined the potential interactions between EDF1, GIGYF2, and ZNF598 to complement previous experiments by Morita et al., 2012. The IPs were performed from the ribosome-free cytosolic fraction to avoid co-precipitation due to bridging via polysomes. As shown in the new Figure 5, we find that ZNF598 interacts with GIGYF2, but that EDF1 does not interact with either. The absence of EDF1 interaction was verified with two different tags. The ZNF598-GIGYF2 interaction is retained in EDF1 knockout cells, further arguing that EDF1 is not an integral part of this complex away from the ribosome. Earlier work established that the ZNF598 interaction with 4EHP is via GIGYF2 (Figure 5B in Morita et al., 2012). Here, mutation of the 4EHP interaction motif in GIGYF2 prevented co-IP of 4EHP by ZNF598 despite recovery of GIGYF2. Because the GIGYF2-4EHP interaction is well established and validated with mutants, it was not re-explored in our experiments. Collectively, these findings show that free in the cytosol, ZNF598 interacts with GIGYF2, which interacts with 4EHP. This complex does not appear to include or require EDF1.

Interactions among these proteins on the ribosome cannot be deduced by co-IP because the ribosome and/or mRNA may bridge proteins that are nowhere close to each other. Nevertheless, we can observe by co-fractionation that EDF1, GIGYF2 (Figure 1B), and ZNF598 (Figure 6—figure supplement 2) are all selectively on collided ribosomes. While this finding by itself does not demonstrate they are all together on the same collided ribosome, this is a likely interpretation for the following reasons:

i) GIGYF2 interacts with ZNF598 and 4EHP off the ribosome (Figure 5, as described above), indicating that they are probably recruited together to collisions.

ii) Recruitment of GIGYF2 to collisions is impaired without EDF1 (Figure 4B) arguing that GIGYF2 is recruited to ribosomes that contain EDF1.

iii) Recruitment of ZNF598 to collisions is impaired without GIGYF2 or EDF1 (Figure 6—figure supplement 2) arguing that ZNF598 is recruited to ribosomes containing these factors.

Based on the collection of these data, together with the finding that EDF1 recruitment to collisions does not depend on ZNF598 (Figure 1C) or GIGYF2 (Figure 5—figure supplement 1), we conclude that EDF1 engages collided ribosomes and facilitates the recruitment of pre-formed ZNF598GIGYF2-4EHP complexes (see diagram in Figure 5E). The spatial arrangement of this complex on the ribosome will require structural analysis that is beyond the scope of this study.

2) How is eS10 ubiquitinated by ZNF598 in cells missing GIGYF2 (Figure 6)? The supplement Figure 6 shows that ZNF598 is not recruited to polysomes in the latter cells. It is argued that this may be due to transient interaction of ZNF598. However, it was thought that direct experimental evidence corroborating this tenet should be provided.

Long exposures of the immunoblots show that although weak, ZNF598 does interact selectively with collided ribosomes even in cells lacking GIGYF2 or EDF1 (Figure 6—figure supplement 2). This is consistent with a transient interaction mediating the observed eS10 ubiquitination. To verify this conclusion and exclude the possibility that another ligase ubiquitinates eS10, we analysed eS10 ubiquitination upon ZNF598 depletion in wild type cells and those lacking GIGYF2 (Figure 6C). The complete loss of ubiquitinated eS10 in each case demonstrates that ZNF598 is the sole E3 ligase for eS10, arguing that indeed a weak interaction is sufficient for ubiquitination to proceed effectively. This is now explained in the revised text that accompanies the above new experiments.

3) MbF1 (yeast EDF1 homolog) is thought to prevent frameshifting in yeast cells on stalled ribosomes (Wang et al., 2018). Is there more frameshifting with reporters in EDF1-depleted Hek293 cells relative to controls?

We analyzed this in some detail using four different reporters: (i) a weakly stalling poly(A) sequence; (ii) the Xbp1 stalling sequence; (iii) a strong-stalling variant of Xbp1; (iv) a weakstalling variant of Xbp1. As detailed in Figure 3 and its accompanying two supplements, depletion of EDF1 results in increased frameshifting at the Xbp1 stall and its strong-stalling variant, but not at the weaker stalling Xbp1 mutant or the weakly stalling poly(A) sequence. As discussed in the relevant parts of the text, we conclude that stalling and the ensuing ribosome collisions cause frameshifting (as in yeast), and that EDF1 mitigates this similar to Mbf1 in yeast (although there are clearly differences, as discussed in the text).

Reviewer #1

– "In addition to ZNF598 as the most collision-specific interaction partner, we identified eight other candidate collision-specific proteins."

Can authors explain how were these proteins identified as candidates?

– "Only two of the candidates, EDF1 and GIGYF2, were collision-specific interactors in both WT and ZNF598 KO cells"

How was this calculated/determined?

Sorry for not explaining this. We have now amended the text to explicitly indicate how many total proteins were identified in the mass spectrometry experiment (~600), what criteria were used to identify candidates (1.6-fold enrichment on collided polysomes relative to both untreated and fully stalled polysomes), and what further criteria led us to EDF1 and GIGYF2 (1.6-fold enrichment on collided ribosomes even in ZNF598 knockout cells).

– Figure 3D – Western blot indicating the extent of ZNF598 depletion is missing.

We apologise for the confusion. These are CRISPR-mediated knockout cells that were previously validated to have no detectable ZNF598 (Juszkiewicz and Hegde, 2017).

Reviewer #2

Authors never report that experiments shown in Figure 3C and Figure 5B are from the same set of data, but they clearly use the same blots in these two figures. Have authors did replicates of these experiments?

We apologise for this oversight. In a previous draft of this manuscript, Figure 3C and 5B were part of a single figure from an experiment that was later presented in two parts. That is why the GIGYF2 and RACK1 blots were the same (sorry for not making this explicit in the respective legends). To avoid any concern or confusion, one of the figures (now Figure 4B) has been replaced with a different version of that same experiment.

It is not completely clear how eS10 gets ubiquitinated by ZNF598 in cells missing GIGYF2 (Figure 6). The supplement Figure 6 shows no recruitment of ZNF598 to poly-ribosomes in those same conditions with a rather massive recruitment in case of the WT cells. Authors argue that this could be due to transient interaction of ZNF598 in cells missing either EDF1 or GIGYF2. Is es10 ubiquitinated in HEK293 cells missing ZNF598, or Caco-2 and MCF-7 cells?

We have now verified that the ubiquitination of eS10 is completely dependent on ZNF598 in both wild type cells and in cells lacking GIGYF2 or EDF1 (Figure 6C). Thus, the ubiquitination must be due to transient interactions that do not result in stable co-fractionation through sucrose gradients. Consistent with this conclusion, long exposures of the immunoblots show that although weak, ZNF598 does interact selectively with collided ribosomes even in cells lacking GIGYF2 or EDF1 (Figure 6—figure supplement 2).

Original report by Morita et al., 2012. has 4EHP-GIGYF2-ZNF598 complex as translational repressor with direct co-IPs between all components and especially between 4EHP and ZNF598 as well as GIGYF2 and ZNF598. Is EDF1 binding to GIGYF2-4EHP and competing with ZNF598? There is no co-IP studies shown on any components so the reader is left to believe on interactions through the polysome profiles or functional assays.

We have now added co-IP experiments to our study showing that GIGYF2 interacts with 4EHP and ZNF598, and that EDF1 is not part of this complex away from the ribosome (Figure 5; see detailed reply to the Editor’s comments above).

Reviewer #3

The authors should compare their findings with recent work outlining the crosstalk between translation initiation and the RQC: https://www.biorxiv.org/content/10.1101/792994v1

This is now cited at a few key points in the paper. We avoid extensive comparison or discussion as this study has not undergone peer review and might change in the process.

The authors might want to explore the potential sequence/structure features underlying pausing collisions events and their potential in discriminating between ZNF598-dependent and other pausing events. Results from recent work on detecting pausing events should be considered (e.g. https://www.biorxiv.org/content/10.1101/746875v2 , https://www.biorxiv.org/content/10.1101/710061v1 , https://www.biorxiv.org/content/10.1101/710491v1).

As noted in the overall reply to the Editor, we have omitted our own ribosome profiling analysis. Instead, we cite the above studies as evidence for widespread pausing and collisions.

Figure 3C shows how both EDF1 and GIGYF2 migrate to the polysome fraction under emetine treatment. However, this might represent a more general phenomenon where many more protein complexes migrate to the polysome fraction. Can the author comment on this possibility?

This is now clarified in the text where we explain that extensive proteomic analysis shows that the vast majority of the ~600 detected proteins do not show this behavior. Furthermore, of these ~600 proteins, only EDF1 and GIGYF2 showed this behavior in both wild type and ZNF598 cells, which is why these were pursued in downstream experiments. Sorry for not explaining how uniquely specific these two hits proved to be.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The manuscript has been improved but there are some remaining textual issues that need to be addressed before acceptance, as outlined below:

1) It was thought that the authors should either tone down some statements or discuss alternative explanations of observed phenomena. In particular, the interpretation that EDF1 recruits GIGYF2 complex at collided ribosomes was found to be insufficiently supported by data. To this end, it was found that although the IP experiments support the role of EDF1 in stabilization of GIGYF2 complex binding to collided ribosomes, their role in its recruitment remains unclear. This should be clearly stated throughout the manuscript and depicted in the model shown in Figure 5. It was also thought that Figure 7 should be removed as this model would require further kinetic studies that, as we agree, are out of the scope of this study.

We have changed the wording in the text to avoid making any claims about EDF1 directly recruiting the GIGYF2 complex to collided ribosomes. Instead, we are cautious to always state that EDF1 “stabilizes” or “facilitates stable engagement” or “facilitates stable binding” of GIGYF2 with collided ribosomes. These are accurate descriptions based on the data: GIGYF2 is selectively associated with collided ribosomes (Figure 1 and Figure 1—figure supplement 1), and this association is impaired in cells lacking EDF1 (Figure 4B; Figure 5C, lanes 3-6). Importantly EDF1 is itself specific for collided ribosomes and this does not depend on ZNF598 (Figure 1) or GIGYF2 (Figure 5—figure supplement 1). Furthermore, a point mutation in EDF1 that impairs its ribosome binding (Figure 4C) is also impaired in translation repression at stalls (Figure 4D), an effect that is mediated by GIGYF2. These findings support the conclusion that EDF1, and specifically its ribosome binding, is needed for effective function of GIGYF2. In the Discussion, we note the two alternative interpretations of these data: (i) a direct model where EDF1 provides part of a binding site for the GIGYF2 complex on collided ribosomes; (ii) an indirect model where EDF1 engagement of collided ribosomes changes the ribosome complex in a way that favours the bound state of the GIGYF2 complex.

We have edited the diagram in Figure 5 to be agnostic to either model. Here, and in Figure 7, we depict all of the factors associated with collided ribosomes rather than showing a hypothetical order of binding. Figure 5 further depicts that off the ribosome, interactions between ZNF598GIGYF2-4EHP are observed, which is supported by the co-IP data in Figure 5 and earlier cited work. In the text our wording is now more cautious: “Although we cannot know the order of binding to the ribosome from these data, we speculate that EDF1 binding may occur first because it is ~50 fold more abundant than either GIGYF2 or ZNF598 (Itzhak et al., 2016).” We are similarly circumspect in the Discussion section on this issue.

We have opted to retain Figure 7 because it is helpful to illustrate for readers what we think is happening. To alleviate the reviewers’ concern, we have now edited the Discussion to better explain our rationale for proposing each aspect of our model. The main point of speculation is that re-starting elongation after an incidental collision is faster than ubiquitination and helicase-mediated ribosome disassembly. We now provide appropriate references and cite specific time ranges for the appropriate reactions. Elongation reactions take between ~200 ms to 4 seconds, while the sum of ZNF598- and ASCC-mediated reactions is likely to take minutes. We are also careful to state that although we have not measured the rates of all binding and enzymatic reactions, our estimates of these are based on sound and well-established biochemical principles. All other aspects of the model are well supported by data, which we will not enumerate again here.

2) Although highly appreciated, several issues were raised regarding frame-shifting studies. For instance, it appears that EDF1 depletion favours -1 frameshifting which cannot be fully explained by the collision model. Moreover, the authors cite studies that indicate that frameshifting should be in +1 direction. Discussion of these apparent discrepancies appears to be warranted. Potential effects of frameshifting on NMD activation should also be discussed. Finally, RFP signal and RFP/GFP ratios should be presented along with GFP for EDF1 knockdown and KO cells (Figure 4 and Figure 4—figure supplement 1) to further facilitate interpretation of results in Figure 6 and Figure 6—figure supplement 3.

The mechanistic basis of frameshifting at collisions is not known. At this point, it remains a correlative observation, both in our study and in two earlier studies. The earlier studies used different reporters, a different organism (yeast), and different means of stalling the ribosome. The mechanism of stalling is clearly influenced by organism (since di-codons or stem-loops that stall in yeast have little or no effect in mammals), so direct comparisons between yeast and mammals have not been possible. Importantly, the mechanism of stalling seems to influence frameshifting: di-codons seem to prefer +1, whereas a stem-loop permits both -1 and +1 (Wang et al., 2018). Thus, on the basis of earlier studies in yeast, there was no reason to necessarily expect any particular result in mammalian experiments that use qualitatively different mechanisms of stalling. Thus, there are no discrepancies; simply a set of difficult-tocompare observations using different reagents, methods, and organisms to probe a poorly understood phenomenon. We now state that our results “roughly approximate” recent observations in the yeast system, then explicitly highlight the above issues.

We also mention that an appreciable role for NMD can be excluded because the GFP signal does not change with the frameshifting constructs despite clearly showing premature termination (i.e., the RFP signal is markedly reduced). This is consistent with past work showing that in mammals, a downstream exon junction complex is typically needed to trigger NMD, so it would not be expected for an intronless reporter. Finally, the requested individual GFP, RFP, and ratio plots of the experiments in Figure 4 are provided as a new supplement to Figure 6 so they can be compared by any interested reader (Figure 6—figure supplement 4).

3) It was also thought that the authors should specify number of replicates and precise criteria for some statistical tests. For example, 1.6 fold threshold was found to be quite arbitrary, and the authors should clearly elaborate on criteria to use such cutoff. MS data should be made publicly available and identified factors that are enriched/depleted on collided ribosomes should be at least briefly discussed. The authors are are also encouraged to add all appropriate controls (e.g. showing that ZNF598 is indeed knocked out in the experiment in question rather than referring to previous publications where these lines were established)

The number of replicates as well as other important information related to data acquisition and analysis for each experiment is available in the Materials and methods section. The complete quantitative mass spectrometry data (as an annotated Excel file) is provided with the paper, just as it was provided with the initial submission and the revision. The raw files have been uploaded to the PRIDE database by our mass spectrometry facility. Dotted lines indicating 3 standard deviations from the mean have been added to the graphs in Figure 1 to provide a point of reference. EDF1 and GIGYF2 were chosen for further study because they are the two most collision-specific proteins in both wild type and ZNF598 knockout cells. This reason is stated in Results paragraph one. A cut-off or threshold is irrelevant, since our choice to study these was because they were the most enriched, which would be the case regardless of any chosen threshold.

Other proteins that we did not study further are not discussed because we have nothing productive to say about them. We only labelled them on the graph in Figure 1A because they were all provisional candidates until we did the secondary analysis in Figure 1B that highlighted EDF1 and GIGYF2. Because we provide the full dataset, any interested colleague can read about them and pursue them as desired.

The absence of ZNF598 in knockout cells is documented in Figure 1 and in the MS dataset provided, where it is preferentially detected in wild type, but not knockout cells. This should alleviate any concern that the cell line has somehow reverted since our previous publication, which is also clear from its characteristic phenotype (Figure 3C and Figure 6—figure supplement 3). The knockdowns and knockouts of all other components are documented with immunoblots within the paper.

4) At times, interpretation of data was thought to require some clarification. Particularly, Figure 2—figure supplement 1A shows apparent increase in mRNA levels, albeit in the text the authors state that there is a lack of the potential mRNA abundance changes. To this end, it was thought that appropriate significance testing is required to corroborate the authors' conclusions, and if it turns out that the changes in mRNA abundance are significant, this should be commented on. Moreover, it is argued that 4EHP, but not GIGYF2 depletion increases translation. However, considering the scale in Figures 2C vs. 2D, it appears that the changes are of comparable amplitude but in the opposite directions. Based on this, significance testing appears warranted.

No differences are seen in mRNA levels by statistical comparisons, and this is stated in the legend. Regarding the second point, the reviewer appears to have mis-interpreted the data, perhaps by overlooking the fact that Figure 4C, 4D, and 4E are performed in different genetic backgrounds. Both GIGYF2 depletion and 4EHP depletion each increases translation in wild type cells (Figure 4C). Figure 4D documents that 4EHP depletion does not increase translation further in GIGYF2 knockout cells, providing genetic evidence that they act in the same pathway. Figure 2E documents that GIGYF2 depletion and 4EHP depletion each increases translation in ZNF598 knockout cells, illustrating that ZNF598 is not required for this effect.

5) The authors are advised to verify some of the references. For instance in Mills et al., 2016 study does not appear to demonstrate ZNF598 loss during maturation of reticulocytes.

Mills et al. did not study ZNF598 specifically, but they do provide ribosome profiling datasets of reticulocytes. Inspection of this dataset reveals that footprints corresponding to ZNF598 are not detected in reticulocytes. Importantly, footprints of other mRNAs that are similar in abundance to ZNF598 in a pre-reticulocyte cell line are detected in reticulocytes. This supports the idea that ZNF598 mRNA is reduced or absent in reticulocytes, consistent with the absence of ZNF598 protein in reticulocyte lysate as documented in Juszkiewicz et al., 2018, which is cited together with Mills et al., 2016.

Author details

1. Szymon Juszkiewicz

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Conceptualization, Investigation, Writing - original draft, Performed all experiments displayed in this study and co-wrote the manuscript

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-3361-7264

2. Greg Slodkowicz

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Data curation, Formal analysis, Writing - review and editing, Provided analysis of a ribosome profiling experiment that influenced interpretation of this study, but was not included in the final revised manuscript

   Contributed equally with

   Zhewang Lin

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-6918-0386

3. Zhewang Lin

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Conceptualization, Investigation, Writing - review and editing, Prepared samples for a ribosome profiling experiment that influenced interpretation of this study, but was not included in the final revised manuscript

   Contributed equally with

   Greg Slodkowicz

   Competing interests

   No competing interests declared

4. Paula Freire-Pritchett

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Data curation, Formal analysis, Writing - review and editing, Helped to analyze a ribosome profiling experiment that influenced interpretation of this study, but was not included in the final revised manuscript

   Competing interests

   No competing interests declared

5. Sew-Yeu Peak-Chew

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Investigation, Writing - review and editing, Performed and help analyze mass spectrometry experiments

   Competing interests

   No competing interests declared

6. Ramanujan S Hegde

   MRC Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, United Kingdom

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing - original draft, Project administration

   For correspondence

   rhegde@mrc-lmb.cam.ac.uk

   Competing interests

   Reviewing editor, eLife

   "This ORCID iD identifies the author of this article:" 0000-0001-8338-852X

• 7,305

  Page views

• 1,006

  Downloads

• 63

  Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.