Ubiquitination of stalled ribosomes enables mRNA decay via HBS-1 and NONU-1 in vivo

As ribosomes translate the genetic code, they can encounter a variety of obstacles that hinder their progress. If ribosomes stall for prolonged times, cells suffer due to the loss of translating ribosomes and the accumulation of aberrant protein products. Thus to protect cells, stalled ribosomes experience a series of reactions to relieve the stall and degrade the offending mRNA, a process known as No-Go mRNA Decay (NGD). While much of the machinery for NGD is known, the precise ordering of events and factors along this pathway has not been tested. Here, we deploy C. elegans to unravel the coordinated events comprising NGD. Utilizing a novel reporter and forward and reverse genetics, we identify the machinery required for NGD. Our subsequent molecular analyses define a functional requirement for ubiquitination on at least two ribosomal proteins (eS10 and uS10), and we show that ribosomes lacking ubiquitination sites on eS10 and uS10 fail to perform NGD in vivo. We show that the nuclease NONU-1 acts after the ubiquitin ligase ZNF-598, and discover a novel requirement for the ribosome rescue factors HBS-1/PELO-1 in mRNA decay via NONU-1. Taken together, our work demonstrates mechanisms by which ribosomes signal to effectors of mRNA repression, and we delineate links between repressive factors working toward a well-defined NGD pathway.

1. While the genetic screens form the basis for the paper and one of its strengths, their description is brief and incomplete. The screen selects for suppressors of the unc-54(srf1004) allele, but the exact sequence of this allele is not given, and no detail are provided on how it was created. The methods describe two screens with the initial screen only identifying znf-598 mutants. This is not explained in the text. Is this simply because the nonu-1 loss has a smaller effect size? The text describes that for the mutants "In each case we observed de-repression of the reporter, manifest as increased fluorescence and an improvement in animal movement and egg laying." This is key data for the paper and should be shown.
Thank you for this suggestion. We have included a more detailed description of the unc-54(rareArg) screen in the main text, including an explanation that the non-znf-598 mutants had less pronounced phenotypes. See paragraph starting with "Using unc-54(rareArg), we performed two genetic screens…" 2.The authors mutate Lys residues in rps-10 and rps-20 to Arg and ascribe any effects on the absence of ubiquitination. The authors do not show that these mutant rps proteins are expressed at their normal level and that the proteins fail to be ubiquitinated. Simply adding a rps10∷HA(K125R) lane to figure 3B would greatly strengthen the data, as would addition of a similar blot for rps-20. The authors should add whether these mutations cause any organismal phenotype that is different from znf-598 to address whether the point mutant ribosomes are otherwise functional.
We agree that it is important to show expression of our mutant ribosomal proteins. We have included these key data, and shown that ribosomal protein expression is not grossly perturbed by the lysine>arginine substitutions. We incorporated these data into the revised manuscript (Fig 3C), updated the text, including commenting on the fact that these animals are healthy. Our original interpretation of the role of ribosomal ubiquitination remains unchanged and has indeed been strengthened by this suggestion.
3. The manuscript uses a catalytic mutation in nonu-1 [nonu-1(AxA)] for several key experiments. Supposedly this protein would still be able to bind stalled ribosomes, and might be dominant negative for some aspects. How do the effects of the nonu-1(AxA) allele compare to a nonu-1 null allele?
We shared this concern. We repeated several of our key findings with a nonu-1 null allele (M1I) and observed identical effects to the catalytic (AxA) mutant ( Fig S3). Our original conclusions remain unchanged. Thank you for this feedback. We have removed the SKI complex figures from this manuscript. Figure 3H, the authors indicate that the znf-598 data are significantly different from wt. I assume the same is true for nonu-1(AxA) but that is not indicated. More importantly, is the znf-598 nonu-1 double mutant different from either single? The authors imply in the text that they are ("The fact that nonu-1 and znf-598 together exhibit additive effects").

Some of the figures indicate statistical significance, but this is not consistently shown. For example in
Thank you for this suggestion. We annotated statistical significance throughout the figures (see updated Fig 3B, Fig 4A, Fig 5B, Fig 5C, Fig 5G, Fig 6D, Fig 7B, Fig 7C, Fig 7F, Fig S3, Fig S6).

I could not find how many biological replicates were performed for most of the experiments.
Thank you for the opportunity to clarify; we added n values in the figure legends, and briefly describe them here: for overexpression experiments, we used 3 independent genetic isolates, with at least 4 images of each isolate; for GFP fluorescence experiments, we imaged at least 15 independent animals of the same genotype; for RNA-seq experiments, we performed two biological replicates. 7. In the RNA-seq data of figure 3F, unc-54(rareArg) is clearly affected, but it appears that very few if any other genes are. Can the authors comment on this? Does this mean znf-598 does not have any endogenous targets, or that the 12 Rare Arg residues cause stalling beyond the normal physiological range, or something else?
The short answer is that very few endogenous genes change to the extent of unc-54(rareArg). As the reviewer keenly points out, this could be for several reasons, but it will take some time for us to unpack and distinguish between these possibilities. We look forward to elaborating on this ("the long answer") in a separate manuscript (Monem et al., in prep).
8. The section "Structural, computational, and functional evidence for ubiquitin-binding by HBS-1" is highly speculative and despite the section heading does not contain any "functional evidence" and this section should probably be deleted. Alternatively, it requires substantial experimental evidence to show that the hbs-1 N-terminus indeed binds ubiquitin or ubiquitinated proteins.
Our updated manuscript expands on and clarifies the function of the HBS-1 N-terminus. Despite the HBS-1 N-terminal domain's homology to ubiquitin-binding domains ( Fig 6A, Fig 6B, Fig 6C), we observed no binding to ubiquitin in vitro ( Fig S5). Moreover, we observed that the HBS-1 N-terminus is dispensable for repression from ribosomal stalling (Fig 6D). We've included these findings and updated our manuscript accordingly (see paragraph starting with "To determine whether the N-terminus of HBS-1 is required…" and "While our genetic analyses support a role…"). 9. It is not clear that the last two sections of the results "Deletion of ZNF-598 increases mRNA levels of and ribosomes on a No-Go mRNA Decay substrate" and "Ribo-seq libraries lack accumulations at arginine stalls" add to the manuscript. The main conclusion the authors seem to draw is that "standard monosome and disome Ribo-seq protocols are unable to capture metazoan ribosomes at strong stalls." If the main conclusion is indeed that the experiments in figure 5 and 6 are uninformative for technical reasons they probably shouldn't be a main focus of the main results section. I do understand that there is some value to publishing this information, but by appending it to this paper it distracts from the main point.
We have removed this section from this current manuscript.

Table S1: C. elegans strains and oligos is missing from the version I was provided.
Please see the attached Table S1 in this version. Thank you for the keen eyes. Removed from the manuscript.

Can the authors comment on why they only identify missense mutations in skih-2 in their nonstop screen, and only nonsense mutations in its partner ttc37?
Removed from the manuscript. Figure 5 shows 5 ttc-37 alleles. The text (pages 5 and 7) indicate only 4 were isolated.

13.
Removed from the manuscript.
14. "cerevisiae" and "sapiens" should be spelled out in the text instead of S. cer. and H. sap.

Done.
15 The authors describe C89 of znf-598 as "catalytic", but as far as I know this residues is structural, not catalytic.
Thank you-changed.
16. The results section "Exosome divergence at the Ski7 interface in C. elegans" and figure 2 does not contain any results and at a minimum should be moved to the discussion or supplemental materials. This section only relates to the nonstop mRNA screen and thus is not connected to the main topic of the manuscript.
Removed from the manuscript.

Starting on page 15 and continuing on p17 the authors mention a Gly residue in S. cerevisiae Hbs1. Missing is which Gly of S. cerevisiae Hbs1 they refer to and whether this residue is conserved. Is there any evidence that this Gly is required for yeast Hbs1 function?
Removed from the manuscript.

On page 17 "In both structures, the HBS-1 N-terminus is bound to the 40S subunit near uS10 and uS3 (Hilal et al., 2016; Becker et al., 2011)" is offered as support for the idea that the Hbs1 N-terminus binds ubiquitinated ribosomes, but the data in the two references appear to support that this domain binds un-ubiquitinated ribosomes. Is there any density in the structural data that suggests ubiquitination? Have the authors entertained the idea that rps-10 or rps-20 ubiquitination could block binding of the Hbs1 N-terminus instead?
Thank you for this insightful comment. We have considered this idea in the Discussion (see paragraph starting with "Here, we discovered that the N-terminal domain of HBS-1 is dispensable for NGD…".

on page17 the authors wrote "We also noticed that plant HBS-1 homologs contain a conserved C4-type zinc finger homologous to a known Ub-binding domain from rat (Fig 4C) (Alam et al., 2004). Thus, we hypothesize that the Ub-binding function of HBS-1 emerged relatively early in the evolution of this factor." It is not clear to me how the addition of a C4-type zinc finger to plant Hbs1 leads the authors to hypothesize that the Ub-binding function of HBS-1 emerged early. As far as I can tell the C4-type zinc finger shows that is the focus of figure 4C shows no similarity to the CUE/Aha/Hbs1-N domain of figure 4A and B and it is not at all clear how this clarifies function of Hbs1
Thanks for the opportunity to clarify our language and our thinking. Indeed, the result supports no hypothesis on the timing of the evolution of this domain. "These observations show that HBS-1 N-termini from diverse organisms lack sequence or structural homology with each other, and yet many N-termini instead share homology with domains that bind ubiquitin." We have added this sentence to the results section (see paragraph starting with "The HBS-1 protein consists of an N-terminal domain…").
--------Reviewer #2: In the manuscript, Monem et al. utilized their own Nonstop mRNA Decay reporter worms to demonstrate that SKI complex subunits SKIH-2 and TTC-37 are required for efficiently repressing translation of the Nonstop mRNA. The authors also generated a novel No-Go mRNA Decay reporter worms that rely on a rare Arg codon stretch and confirmed that an E3 ubiquitin ligase ZNF-598 and an endonuclease NONU-1, known factors for No-Go mRNA Decay, are required for efficient repression of the reporter. Genetic screening with the No-Go mRNA Decay reporter further identified uba-1 encoding the sole E1 ubiquitin-activating enzyme in C. elegans. hbs-1 encoding a GTPase homologous to eRF3 was identified in both the Nonstop mRNA Decay screen and the No-Go mRNA Decay screen. PELO-1/Dom34 (homologous to eRF1) was also required for efficient repression of both the reporters. Cys89 of ZNF-598 was crucial for ubiquitination of RPS-10, an rps-10 (K125R); rps-20 (K6R+K9R) double mutant phenocopied the znf-598 (null) mutant in de-repression of the No-Go mRNA Decay reporter and an rps-10 (K125R) znf-598 (null) double mutant was comparable to the znf-598 (null) single mutant, indicating that ubiquitination of these ribosomal proteins by ZNF-598 is critical for No-Go mRNA Decay. A catalytic mutation in nonu-1 did not enhance de-repression of the znf-598 (null) mutant implying that they function in the same pathway. hbs-1 (null) mutation did not enhance de-repression of the znf-598 (null) mutant either. RNA-seq and Ribo-seq analysis revealed that the No-Go mRNA Decay reporter mRNA was stabilized in the znf-598 (null) mutant. However, the distributions of monosomes and disomes were not affected upon de-repression from No-Go mRNA Decay.
1. The Nonstop and No-Go mRNA Decay reporters are unique and excellent for studying mRNA surveillance system in vivo and enabled the authors to characterize the factors required for efficient repression of the problematic mRNAs. It seems to me, however, that the manuscript relies too much on assumption that molecular mechanisms of No-Go mRNA Decay in C. elegans is similar to those in other species such as yeast and human, leading to biased interpretation of the results. The readers would be interested in whether or not this independent genetic study with the excellent reporter system in C. elegans supports the model from other species.
Thank you for the opportunity to clarify. We think there are similarities as well as differences. In C. elegans, our analysis of ZNF-598-dependent ubiquitination sites on ribosomal proteins suggest that C. elegans detects and responds to stalling in a manner similar to human cells. See paragraph starting with "In the NGD screen, we recovered several alleles of znf-598…". In C. elegans, we can clearly see strong phenotypic effects with knockout of nonu-1, whereas in yeast knockout of the homologous CUE2 does not have effects (unless other factors are also knocked out). This is a difference that is not yet understood, and given the lack of studies on the human homolog (N4BP2), we do not know whether humans are similar or different from either organism. See paragraph starting with "To initially validate the unc-54(rareArg) reporter…".
2. In Abstract, the authors claim that "we present data consistent with a model where ubiquitination recruits the endonuclease NONU-1 via CUE domains and the ribosome rescue factor HBS-1 via its poorly characterized N-terminus" (lines 27-29). The provided results are actually consistent with the model, but do not preclude other models. Biochemical analysis and/or genome editing to disrupt critical residues are necessary to claim the model. "Our work (1) presents a model for how ribosomal ubiquitination directly causes mRNA decay and ribosome rescue" (lines 29-30). The evidence for this claim is also weak (see below). It is still unclear whether ribosomes are actually rescued by HBS-1 upon No-Go mRNA Decay in C. elegans.
Thank you for this concern. In the updated manuscript, we added genetic analysis of mutants to test ubiquitin-binding by both NONU-1 (deletion of the CUE domains) and HBS-1 (deletion of the N-terminal domain). Our additional data supports ubiquitin-binding by NONU-1 (Fig 5B), and are consistent with our previous conclusions. As for ubiquitin-binding by HBS-1, we observed the N-terminal domain to be dispensable for NGD in our system (Fig 6D), and its ubiquitin binding was comparable to background in vitro (Fig S5). For more on HBS-1, please see our response to Reviewer #1 point 8. Figure 4H. Gain of function or overexpression of the upstream genes should be suppressed by loss of function of the downstream genes. Gain of function or overexpression of the downstream genes should suppress loss of function of the upstream genes.

The genetic evidence provided in the manuscript do not determine the order of genes in the Model for No-Go mRNA Decay shown in
Thank you for this great suggestion. The revised manuscript considerably expands on this area. We have performed several additional genetic experiments to test the model, specifically, overexpression as the reviewer suggested (Fig 2A, Fig 2B, Fig 5D, Fig 5E, Fig 7D, Fig 7E). Our genetic analyses of nonu-1 places it downstream of znf-598 (Fig 5D, Fig 5E). Therefore, our previous conclusions remain unchanged (see paragraph starting with "We next investigated the ordering of ZNF-598 and NONU-1 relative to one another…"). Our genetic analyses of hbs-1 revealed a role for this factor downstream of znf-598 ( Fig 7D) and upstream of nonu-1 (Fig 7E). Our RNA-seq analyses of hbs-1 and other mutants substantiate a role for HBS-1/PELO-1 in facilitating the RNA cleavage reaction brought about by NONU-1/Cue2. Our updated manuscript contains these findings, as described in the results and discussion (see paragraphs starting with "HBS-1 and PELO-1 are predominantly known…" and "Given prior work in S. cerevisiae revealing a role for Hbs1 in NGD cleavage…"). This significantly expands the scope and novelty of our findings.