Much of what is known about NGD comes from studying the effect of individual factors and events, and it is unclear how these steps relate to one another to bring about target mRNA repression. Here, we deployed C. elegans to unravel the series of events during NGD. We show that mutation of ribosomal ubiquitination sites on RPS-10 (eS10) and RPS-20 (uS10) phenocopies knock out of ZNF-598. We present data in support of a model in which ZNF-598 first ubiquitinates ribosomes at stall sites, followed by mRNA degradation via NONU-1. Interestingly, we also recovered a role for HBS-1 and PELO-1 in mRNA decay via NONU-1 cleavage, consistent with early northern data in S. cerevisiae, suggesting that ribosome rescue may be an important step that precedes mRNA cleavage.
100 Nonu Model
(A) Gene diagram showing annotated exons (black rectangles) of unc-54(rareArg). Colored rectangles represent CRISPR/Cas9 insertions at the endogenous unc-54 locus: T2A sequence (gray), FLAG (dark gray), 12 rare arginine codons (blue), and GFP (green). (B) Schematics of rareArg genetic screens. (C) znf-598, uba-1, nonu-1, and hbs-1 alleles with representative image of one allele per gene on the left. Black rectangles represent exons, thicker rectangles are CDS, and thin lines are introns. Mutations made via EMS in the rareArg screen 1 (light blue) or rareArg screen 2 (dark blue), and via CRISPR/Cas9 (black) or CGC (gray) are shown. For HBS-1, multiple sequence alignment shows conserved glycine (G200) in GTPase domain.
To initially validate the unc-54(rareArg) reporter as a target of NGD, we crossed it with alleles of two factors known to be required for NGD in C. elegans (nonu-1(AxA)) and other systems (znf-598(Δ)) [8,9,15,23]. In each case we observed de-repression of the reporter, manifest as increased fluorescence and an improvement in animal movement and egg laying (UNC-54 is required in the vulva muscles for egg laying). The nonu-1 result is consistent with our prior work showing that this stretch of twelve rare arginine codons confers nonu-1-dependent mRNA decay [15]. Notably, the phenotypic effects seen upon nonu-1 knockout differ from that seen in S. cerevisiae: knockout of the homologous CUE2 in S. cerevisiae only confers effects upon simultaneous knockout of additional factors [16]. Given that most of our mechanistic understanding of NGD comes from work in S. cerevisiae [4,5,16], and that genetic screens in human K562 cells failed to identify the NONU-1 homolog [24], we reasoned that a genetic screen in C. elegans would prove insightful and augment information gained from other systems.
To check the functionality of factors expressed from an array, we made overexpression arrays in each of the cognate mutant backgrounds. Overexpression of the wild-type ZNF-598 and NONU-1 protein restored repression of unc-54(rareArg) and rescued the loss-of-function phenotype of znf-598 (Fig 2A) and nonu-1 (Fig 2B), respectively. Thus the overexpressed factors were functional. Arrays can be stochastically lost or silenced in cell lineages of the animal, allowing us to examine rescue on a cell-by-cell level. For each of znf-598 and nonu-1, we observed an inverse relationship between factor expression (monitored via mCherry) and unc-54(rareArg) (monitored via GFP). Thus the rescue was cell-autonomous, as would be expected under current models of ZNF-598 and NONU-1 acting directly on ribosomes and mRNAs.
To quantify the inverse relationship of factor overexpression (mCherry) to NGD (GFP), we calculated an overlap score, based on the brightest red and green pixels across multiple, independent animals (Methods) (S1 Fig). If mCherry and GFP are non-overlapping (as would be expected from rescue), the overlap score would be negative. If mCherry and GFP overlap (as would be expected without rescue), the overlap score would be positive. For znf-598 and nonu-1, we observed values close to -1 (Fig 2C), demonstrating the generality of rescue across animals.
With a functional readout of NGD (unc-54(rareArg)), mutants in factors (znf-598, nonu-1, hbs-1, and pelo-1), and the overexpression/rescue system, we set out to characterize the molecular mechanisms by which factors relate and repress gene expression in response to ribosomal stalling.
Prior work underscored the importance of ribosomal ubiquitination events in NGD [8,9]. However, the precise functional contributions of individual ubiquitination sites has remained unclear, in part due to the difficulty in obtaining viable mutants in the relevant ribosomal proteins. In prior work, overexpression of RPS-10 with K>R substitutions at ubiquitination sites conferred de-repression of a stalling reporter in between that observed in wild-type and ZNF598 knockout human cells [8]. Interpretation of this experiment is complicated by a number of factors, including residual expression of wild-type RPS-10. Would complete removal of RPS-10 ubiquitination sites mimic loss of ZNF598? Does ZNF598 contribute to functional repression outside of its role in ribosomal ubiquitination? Our work thus far supported a NGD mechanism similar to that of human cells, and therefore our system seemed a useful model to explore these questions. In particular, the viability of ribosomal point substitutions at endogenous loci as well as the ease of making double mutants and overexpression constructs provided us with the means to test models of the functional importance of ribosomal ubiquitination and its relationship to ZNF-598.
We also performed additional experiments to clarify the function of ribosomal ubiquitination in relation to znf-598. First, we performed a double mutant analysis. In a model where ribosomal ubiquitination is functionally independent of znf-598, we would expect that a rps-10(K125R); znf-598(Δ) mutant would exhibit a stronger NGD phenotype (more GFP) than either single mutant alone. However, if ZNF-598 and Ub-RPS-10 lie on the same pathway, we would expect the rps-10(K125R); znf-598(Δ) phenotype to resemble one of the single mutants. The rps-10(K125R); znf-598(Δ) phenotype was indistinguishable from the znf-598(Δ) single mutant (Fig 4A), consistent with the latter model. Second, we used our overexpression system. In a model where rps-10(K125R) and rps-20(K6R+K9R) are on a partially redundant pathway to znf-598 (i.e., that znf-598 contains additional repressive functions outside of ubiquitination at these two sites), we would expect that overexpression of ZNF-598 would provide restoration of NGD in the rps-10(K125R); rps-20(K6R+K9R) mutant. However, if ZNF-598 acts through RPS-10 and RPS-20, overexpression of ZNF-598 in a rps-10(K125R); rps-20(K6R+K9R) mutant would yield the same phenotype as rps-10(K125R); rps-20(K6R+K9R). We observed the latter (Fig 4B), again suggesting that ZNF-598 works through ubiquitination of RPS-10 and RPS-20.
Having established a requirement for ubiquitination by ZNF-598 in this system, we decided to investigate the relationship of ubiquitination to mRNA decay. A key effector of NGD is NONU-1, which was identified in our screen (Fig 1C) and also in our prior NSD screen [15]. In our prior work, we noted that NONU-1 and its homologs contain CUE domains (Fig 5A) [15,29], which are known to bind ubiquitin, suggesting a mechanism of recruitment for NONU-1 to sites of ribosome stalling. To test a requirement for the CUE domains in nonu-1 function, we deleted the CUE domains and observed a phenotype indistinguishable from other nonu-1 mutants (Fig 5B). This result is consistent with CUE domains being essential to NONU-1 function. We also attempted to examine expression of NONU-1 by tagging the N-terminus (S1 Table), but tagged alleles were non-functional. We did not tag the C-terminus of NONU-1 as it is conserved. We look forward to future work where we can determine whether the requirement of CUE domains in NONU-1 is merely for NONU-1 expression or for its biochemical functions.
To determine whether NONU-1 acts in the same pathway as ZNF-598, we performed a double mutant analysis. If the two factors function in different pathways, we would expect additive phenotypes on the unc-54(rareArg) reporter in a double mutant. However, if NONU-1 and ZNF-598 work in the same pathway of repression, we would expect the double mutant to resemble one of the single mutants. We combined mutations in znf-598 and ribosomal ubiquitination sites with nonu-1(AxA), a catalytic mutation of nonu-1 that is indistinguishable from a deletion of nonu-1 (S3 Fig) [15]. The nonu-1; znf-598 double mutant mimicked the znf-598 single mutant (Fig 5C), consistent with the two factors functioning in the same pathway in NGD.
We next investigated the ordering of ZNF-598 and NONU-1 relative to one another in NGD. In a first pair of experiments, we overexpressed ZNF-598 in a nonu-1 mutant, and also overexpressed NONU-1 in a znf-598 mutant. According to classical logic when ordering genes in functional pathways, we expect that overexpression of an upstream factor will not compensate for loss of a downstream factor, but that overexpression of a downstream factor will compensate for loss of an upstream factor [30,31,32,33]. We observed little effect of the nonu-1 phenotype by ZNF-598 overexpression (Fig 5D), and we saw rescue of the znf-598 phenotype by NONU-1 overexpression (Fig 5E). These results support a model where NONU-1 acts downstream of ZNF-598. In a second set of experiments, we examined RPS-10 ubiquitination by immunoblot in nonu-1(AxA) and observed it to be unchanged (Fig 5F). We interpret this result to indicate that the defect in a nonu-1 mutant is after ribosomal ubiquitination, i.e., in the commitment of ubiquitinated ribosomes to mRNA decay.
Taken together, our results with the unc-54(rareArg) reporter support a model in which ZNF-598 and NONU-1 function together in NGD. Based on our genetic and molecular analyses, we favor a model in which ribosomal ubiquitination recruits NONU-1 to mRNAs for cleavage.
Several of our NGD suppressor mutants (znf-598, nonu-1, uba-1) encoded factors involved in ubiquitin-dependent processes, so we were initially surprised that this screen also identified the ribosome rescue factor hbs-1. We therefore examined HBS-1 to determine whether it could conceivably function in a ubiquitin-dependent manner. 2ff7e9595c