Introduction

RNA degradation is an essential process for maintaining cellular homeostasis, regulating both the quantity and quality of RNAs in the eukaryotic cell. RNA degradation typically occurs by exonuclease attack from either end of the transcript, but in some cases is initiated by endonuclease cleavage at internal sites. Endonuclease cleavage of certain RNAs modulates important cellular processes including differentiation and stress response. As such, mutations in endonucleases have been implicated in many human diseases. However, eukaryotic endonucleases remain largely uncharacterized, and in most cases of disease, specific RNA transcripts that fail to be cleaved have not been identified, highlighting the need for systematic identification of endonuclease targets.

Indeed, recent transcriptome-wide studies have uncovered non-classical RNA targets for well- studied endonucleases. One notable example is the human RNase III enzyme, Drosha. Until recently, Drosha’s only known function has been the maturation of micro-(mi)RNAs (Y. Lee et al., 2003). It is now clear, however, that Drosha directly cleaves other classes of RNAs, including mRNAs (Han et al., 2009; Karginov et al., 2010; Knuckles et al., 2012), long non-coding (lnc)RNAs (Dhir et al., 2015), and exogenous viral RNAs (Lin & Sullivan, 2011; Shapiro et al., 2012, 2014).

Like its distant metazoan homolog Drosha, the Saccharomyces cerevisiae RNase III enzyme, Rnt1, has also been shown to process ncRNAs, including small nuclear (sn)RNA and small nucleolar (sno)RNA precursors. snRNAs are essential components of the spliceosome, and snoRNAs guide the chemical modification of rRNA and snRNAs. The most abundant target of Rnt1, however, is rRNA. Although Rnt1 is not essential, a rnt1Δ strain is extremely slow-growing (Chanfreau et al., 1997; Chanfreau, Rotondo, et al., 1998), and this is presumed to be due to the requirement of Rnt1 in rRNA processing.

Rnt1 carries out 3’ end maturation of pre-rRNA by cleaving at a specific site downstream of the mature 25S 3’ end in the 3’ external transcribed spacer (ETS) (Elela et al., 1996). The U3 pre- snoRNA and the U1, U2, U4, and U5 pre-snRNAs are also cleaved to remove a 3’ extension downstream of their mature ends. At both rRNA and snRNA genes, co-transcriptional 3’ cleavage generates an entry site for the nuclear 5’-3’ exonuclease Rat1. Rat1-mediated digestion of the nascent RNA then triggers transcription termination (Allmang et al., 1999; Allmang & Tollervey, 1998; El Hage et al., 2008). Many pre-snoRNAs, on the other hand, are 5’ end processed by Rnt1 (Chanfreau, Rotondo, et al., 1998). While some snoRNAs contain a trimethylated G (TMG) cap and do not require 5’ processing (Chanfreau, Legrain, et al., 1998), others rely on Rnt1 to either remove a m⁷G-capped extension or to both remove the 5’ extension and liberate them from polycistronic gene arrays (Chanfreau, Legrain, et al., 1998; Chanfreau, Rotondo, et al., 1998; Qu et al., 1999). Failure to remove pre-snoRNA m⁷G caps results in snoRNA mislocalization to the cytoplasm and also affects the processing of their 3’ ends (Grzechnik et al., 2018). Additionally, some intron-encoded pre-snoRNAs require Rnt1 cleavage for release from their host introns (Ghazal et al., 2005; Giorgi et al., 2001). While cleavage of pre-snRNAs and pre-snoRNAs is followed by Rat1 exonucleolytic digestion, the consequences differ; Rat1 processing produces mature 5’ ends of snoRNAs, but causes transcription termination of pre-snRNAs.

Rnt1 recognizes double-stranded (ds)RNA stem loops with a terminal single-stranded tetranucleotide loop containing the consensus sequence AGNN (Catala et al., 2004; Elela et al., 1996; Nagel & Ares, 2000). This tetraloop is the major determinant of Rnt1 substrate recognition as it dictates the three-dimensional conformation of the substrate and facilitates docking of Rnt1 (Chanfreau et al., 2000; Lebars et al., 2001; Wu et al., 2001). Specifically, the conserved A and G nucleotides in the first two positions of the tetraloop allow Rnt1 interaction with the loop (Lebars et al., 2001; Wu et al., 2001). Rnt1 also binds UGNN tetraloops, but with lower affinity, and can cleave 3- and 5-nt terminal loops with diverse sequences in vitro (Gagnon et al., 2015; Lebars et al., 2001). Rnt1 cleaves 14 nts upstream and/or 16 nts downstream of the tetraloop using its RNase III nuclease domain (Chanfreau et al., 2000). The sequences surrounding the tetraloop can impact the stability of the stem and affect Rnt1 binding, while sequences close to the cleavage site can affect the ability of Rnt1 to cleave (Lamontagne et al., 2003, 2004). This results in Rnt1 binding and cleavage efficiency that is difficult to predict systematically (Lamontagne et al., 2003).

Cleavage results in 3’ fragments with a free 5’ monophosphate (Court et al., 2013; Lamontagne et al., 2004) that are substrates for Rat1 and/or the primarily cytoplasmic 5’-3’ exonuclease Xrn1 (Kenna et al., 1993; Stevens, 1980). Upon Rnt1 cleavage, Rat1 processes ncRNA precursor 5’ ends in the nucleus (Chanfreau et al., 1997; Henry et al., 1994), but when a Rnt1 site is artificially inserted into mRNAs, the resulting cleavage products are degraded from their 5’ ends predominantly by Xrn1 in the cytoplasm (Meaux et al., 2011). Rnt1-cleaved 3’ fragments, therefore, accumulate in a rat1-ts xrn1Δ strain background (Geerlings et al., 2000; Petfalski et al., 1998).

While the role of Rnt1 in ncRNA processing is well understood, much less is known about the enzyme’s function in mRNA degradation. Throughout the literature, a total of 102 Rnt1 cleavage sites have been implicated in ncRNA maturation, and many of these have been confirmed by multiple studies. In contrast, the 36 published Rnt1 cleavage sites in mRNAs have largely been identified in isolation and have not been confirmed by other studies. These studies have used in vitro cleavage to identify Rnt1 mRNA targets, but only a limited subset of these mRNAs have been shown to be cleaved in vivo. Additionally, some studies have reported changes in overall mRNA levels, but these include indirect effects. Genome-wide studies have identified a large number of mRNAs that are upregulated when Rnt1 is deleted but are not cleaved by the enzyme in vitro, while many other RNAs that are cleaved by the purified enzyme in vitro are not affected by the deletion of the enzyme in vivo. This raises questions about whether certain Rnt1 substrates might only be recognized in vivo. Here, we take an in vivo approach to identify bona fide Rnt1 cleavage sites throughout the yeast transcriptome, focusing our analysis on protein-coding regions. We find that Rnt1 directly cleaves several mRNAs with features resembling known targets and that cleaved mRNAs are subsequently degraded by Xrn1. We also show that several mRNAs greatly rely on Rnt1 for turnover and that Rnt1’s regulation of one highly cleaved target is essential for maintaining homeostasis.

Results

PARE identifies known Rnt1 cleavage sites and substrates

To identify in vivo Rnt1 cleavage sites throughout the yeast transcriptome, we used an RNA sequencing approach called parallel analysis of RNA ends (PARE). We have previously shown that PARE is effective for determining cleavage sites of tRNA splicing endonuclease (TSEN), while RNAseq identifies hundreds of changes that do not include direct cleavage targets (Hurtig et al., 2021; Hurtig & van Hoof, 2023). Using PARE, we successfully identified mRNA targets of TSEN (Hurtig et al., 2021) and uncovered Dxo1 as a distributive exoribonuclease required for the final step in 25S rRNA maturation (Hurtig & van Hoof, 2022).

To perform PARE, total RNA is isolated from yeast cells, and an adapter is added onto free 5’ monophosphates using T4 RNA ligase (Figure 1A). The adaptor is then used to build a sequencing library. Exposed 5’ phosphates result from decapping or endonuclease cleavage. Thus, RNAs can be sequenced from their 5’ monophosphate ends, and the precise positions of decapping or cleavage can be identified by accumulated reads starting at those specific nucleotide positions. Analysis of our previously published PARE data (Hurtig et al., 2021) revealed a peak of 5’ monophosphate ends at the known Rnt1 cleavage site in BDF2, suggesting that PARE could be used to identify additional Rnt1 sites. Further, comparison of PARE data using polyA+ and polyA- RNA from strains lacking either the primarily cytoplasmic Xrn1 (xrn1Δ) (Hurtig et al., 2021; Hurtig & van Hoof, 2022) or both Xrn1 and nuclear Rat1 (rat1-ts xrn1Δ; unpublished) revealed that known Rnt1 sites can be most prominently detected in the polyA+ fraction of a rat1- ts xrn1Δ strain. PARE on the polyA- fraction predominantly detected the mature 5’ monophosphate ends of snoRNAs, and PARE on the polyA+ fraction from an xrn1Δ-only strain weakly identified known Rnt1 sites. Thus, to identify Rnt1 cleavage sites, we used PARE to detect 5’ monophosphate ends that are present in a RNT1 rat1-ts xrn1Δ strain but absent from a rnt1Δ rat1-ts xrn1Δ strain (hereafter “RNT1” and “rnt1Δ”, respectively, unless otherwise noted).

Figure 1:

To distinguish bona fide Rnt1 cleavage events, we performed two replicates of PARE and focused on sites with ≥1 cpm (counts per million) in the RNT1 strain in each replicate (Figure 1B; Supplemental Figure 1). This cutoff eliminates very rare RNA 5’ ends. Additionally, we required a log2 fold change (log2(FC)) ≥4 (i.e., a 16-fold difference in the number of reads) between RNT1 and rnt1Δ in each replicate for a site to be considered a legitimate Rnt1 cleavage site (Figure 1B; Supplemental Figure 1). The log2(FC) ≥4 was set empirically because it captured almost all known Rnt1 sites, and the vast majority of transcriptomic sites had a log2(FC) between -4 and 4. Very few sites had a log2(FC) <-4 (i.e., showed increased cleavage in the rnt1Δ strain), suggesting that this was an appropriate cutoff to eliminate noise. These two cutoffs identified 496 Rnt1 sites throughout the yeast transcriptome (Supplemental Table 1). In many cases, two sites were separated by approximately 34 nts (Supplemental Figure 2), as expected for cleavage 14 nts upstream and 16 nts downstream of a 4-nt loop. Other sites were separated by 1 to 5 nts (Supplemental Figure 2), which can be explained by imprecise cleavage of Rnt1 or by slight degradation of the primary cleavage product by a 5’ exoribonuclease such as Dxo1 and/or Rat1 not fully inactivated by the temperature sensitive mutation. Overall, the 496 sites clustered into a total of 166 distinct RNA substrates (Figure 1C; Supplemental Table 1).

Of the 166 putative substrates, 44 were known Rnt1 targets, most of which were ncRNAs, including pre-snRNAs, pre-snoRNAs, and the pre-rRNA 3’ ETS (Figures 1C, D; Supplemental Figure 3A). This validates PARE and our cutoffs as a suitable approach for identifying Rnt1 sites and substrates. These sites could be visualized with Integrative Genomics Viewer (IGV), where the height of the peak at each position reflects the number of sequencing reads beginning at that position. For example, Rnt1 has recently been shown to cleave the snoRNA precursor pre-SNR83 upstream of its mature 5’ end (Grzechnik et al., 2018). This site was detected in our PARE data (Figure 1E) where cleavage at that position can be observed in the RNT1 strain, but not in rnt1Δ. Interestingly, we identified an additional Rnt1 site downstream of the known site. These two sites map on either side of an AGUU tetraloop, consistent with Rnt1’s cleavage pattern. Northern blot using a probe that hybridized downstream of the sites in pre-SNR83 confirmed the Rnt1 cleavage detected by PARE. We observed processing to the mature snoRNA in the RNT1 strain, as well as in a rnt1Δ strain containing RNT1 on a plasmid, while the pre-snoRNA transcript accumulates in the rnt1Δ strain containing empty vector (Figure 1E) Thus, PARE confirmed pre-SNR83 as a Rnt1 substrate but also identified a second cleavage site.

Similarly, PARE identified previously unknown 5’ sites in the known substrates pre-SNR39B, pre- SNR85, pre-SNR87, and pre-SNR81 (Supplemental Figure 3B) and an unknown 3’ site in SNR17B (Supplemental Figure 3C). Additionally, although SNR84 was thought to be capped (Grzechnik et al., 2018), we identified a peak of 5’ monophosphate ends at its mature 5’ end, suggesting that SNR84 is uncapped (Supplemental Figure 3D). We also detected additional sites upstream of the 5’ end, and the sequence between these sites can fold into a stem loop structure with a bifurcated stem and an AGUA loop. All three sites disappeared in the rnt1Δ strain, suggesting that Rnt1 cleaves a 5’ extended pre-SNR84 at -26 and -74 nts to generate the uncapped snoRNA. The processed uncapped SNR84 we detect could, however, co-exist with a capped SNR84. Although we detected these novel ncRNA sites by PARE, most of the other ncRNA sites we detected precisely correspond to those in the published literature, demonstrating the reliability of PARE.

Additionally, PARE detected 3 previously published mRNA targets of Rnt1: BDF2, HSP60, and ARN2 (Figures 1C, D). Although cleavage peaks were detected in an additional 6 known targets (MIG2, EST1, YTA6, RGT1, MTH1, and HSL1), these PARE peaks had low RNT1 reads (<1 cpm) and/or did not meet our log2(FC) cutoff of ≥4 (Figure 1D). Thus, these 6 mRNAs may be targets of Rnt1 but are not the most prominent targets. BDF2 is one of the most well-studied Rnt1 mRNA targets, and it was also one of the top targets in our dataset. PARE confirmed that BDF2 was cleaved at the known sites that have previously been described (Figure 2A). This mRNA adopts a secondary structure that facilitates the formation of a double-stranded stem loop within its ORF. The stem is capped by a UGAU tetraloop, instead of the more common AGNN, and Rnt1 sites are 14 nts away from the loop on the 5’ side and 16 nts away on the 3’ side (Figure 2A). In vivo cleavage was confirmed by northern blot, which showed the presence of the cleavage product (Figure 2E). ARN2, another Rnt1 mRNA target that has been previously described (A. Lee et al., 2005), was also shown to be cleaved in our PARE data (Figure 1D, Supplemental Figure 4A). Identification of these known mRNA targets further validate the use of PARE for determining novel Rnt1 cleavage sites in protein-coding regions.

Figure 2:

PARE identifies novel Rnt1 mRNA targets

In addition to the 44 known Rnt1 substrates, we also detected sites in 122 putative substrates. Although these included interesting novel targets such as pre-mRNA introns lacking snoRNAs (Supplemental Figure 3E) and ncRNA transcripts derived from regions annotated as intergenic (Supplemental Figures 3F-H), approximately half of the newly identified targets were mRNAs (i.e., protein-coding genes cleaved within their coding sequence (CDS)) (Figure 1C). The CDS site with the largest log2(FC) was in a novel substrate, CAF4 (Figure 2B). CAF4 encodes the Ccr4- associated factor 4 (Caf4) whose function is poorly understood. We found that CAF4 mRNA was cleaved on both sides of a stem loop containing the canonical AGNN tetraloop sequence (AGGA), with an average 1000-fold difference (10 log2(FC)) in cleavage between the RNT1 and rnt1Δ strains at the upstream site, and a 2500-fold difference (11 log2(FC)) at the downstream site. The cleavage sites map to 15 nts upstream and 16 nts downstream of the tetraloop. Considering that the 15 nts upstream includes one bulged nucleotide, this resembles the typical Rnt1 cleavage pattern that results in a 2-nt overhang on the 3’ side of the stem loop (Court et al., 2013; Lamontagne et al., 2004). We also validated in vivo cleavage of this target by northern blot (Figure 2E).

Other interesting mRNAs among our top hits were YDR514C (third highest log2(FC); Figure 2C) and MTM1 (tenth highest log2(FC); Figure 2D). These two mRNAs do not have canonical stem loops but instead are predicted to have more complex structures that included an AGNN loop. YDR514C had an AGGU tetraloop on a bulged stem (Figure 2C). Remarkably, there is still a 3’ 2- nt overhang along the double-stranded region of the YDR514C stem loop, consistent with cleavage by Rnt1. Unlike many mRNA substrates, we detected only a single site in MTM1 (Figure 2D). MTM1 has an AGGU stem loop with two other stem loops on either side, all radiating from a central bulge. While it is possible that MTM1 may fold differently in vivo compared to the predicted RNA structure, one explanation for the single Rnt1 cleavage site could be that the two double- stranded regions on either side of the AGGU stem dictate that only one site can be cleaved. This single site still resembles the typical Rnt1 cleavage pattern by being 16 nts 3’ of the tetraloop. Thus, some Rnt1-cleaved mRNAs have variant stem structures that may, in part, be why they have not been previously identified.

Other highly cleaved, novel mRNA substrates with canonical AGNN loops included TCB1, YER145C-A, PAN6, AVT1, and YPL277C (Supplemental Figures 4A-C). Interestingly, although YPL277C and YOR389W are recent evolutionary duplicates, there was no PARE signal detected in YOR389W (Supplemental Figure 4C). These mRNAs are 96% identical, with only a single- nucleotide difference in the 90-nucleotide stem loop. This one nucleotide difference disrupts a base pair 4 nts downstream of the cleavage site and destabilizes the YOR389W stem loop (ΔG of -15.2 for YOR389W, compared to ΔG of -17.8 for YPL277C). The hundreds of reads that mapped to YPL277C all had the expected G in the fourth position, unambiguously identifying them as YPL277C reads. The specificity of YPL277C over YOR389W confirms previous mutational analysis that sequences in the stem can modulate cleavage (Lamontagne et al., 2003, 2004).

Because it has been previously suggested that Rnt1 displays a preference for mRNAs involved in carbohydrate metabolism and respiration (Gagnon et al., 2015), we investigated whether the Rnt1 mRNA targets identified by PARE function in related pathways. Gene ontology analysis, however, showed no significant enrichment for common cellular processes, molecular function, or cellular localization. Nonetheless, examination of individual targets revealed that 25% of the mRNAs encode transporters or proteins involved in regulating protein targeting to organelles (Supplemental Figure 5).

We reasoned that Rnt1 likely selects its mRNA targets by recognizing the same signals used to select ncRNA substrates. Thus, to determine the conservation of the AGNN tetraloop in these mRNA targets, we used sequences surrounding the detected Rnt1 cleavage sites in these ORFs to generate mFold predictions of their secondary structures. Of the 63 ORF sequences, 42 were predicted to fold into double-stranded stem loop structures with A/UGNN tetraloops (Supplemental Figure 6). Alignment of these tetraloop sequences revealed a strong preference for A in the first position, with an occasional U (Figure 2F; Supplemental Figure 6). All aligned sequences possessed a G in the second position of the tetraloop. Although the 2 nts before and after the tetraloop could always form two base pairs, there was no sequence consensus (Figure 2F). Similarly, there was no consensus for the nucleotides surrounding the cleavage sites. Taken together, these results indicate that Rnt1 cleavage of mRNAs relies on the presence of a stem loop with an A/UGNN tetraloop sequence (as observed with all known Rnt1 targets) and is influenced by the mRNA secondary structure but not the function or cellular localization of the encoded protein.

Rnt1 directly and independently cleaves mRNAs

A catalytic mutant of Rnt1 (rnt1-D245R) has previously been shown to prevent cleavage in vitro and to be unable to restore growth in a rnt1Δ background. This mutant, however, does complement the cell cycle and cell morphology defects of rnt1Δ, indicating that Rnt1 has both catalytic and non-catalytic functions (Catala et al., 2004). Thus, we used in vivo and in vitro experimental approaches to rule out the possibility that the mRNA cleavage products we detected reflect an indirect effect and/or a non-catalytic function of Rnt1. In vivo, we tested whether the rnt1-D245R mutant affected cleavage of mRNAs. The catalytic mutant was indeed slow-growing and was unable to cleave pre-SNR83 (Figure 3A) like the complete deletion of RNT1, confirming previous reports (Catala et al., 2004). PARE confirmed that the catalytic mutant resembled rnt1Δ in its inability to cleave mRNA targets (Figure 3B), consistent with Rnt1 directly cleaving these mRNAs. To complement this analysis and to eliminate the possibility that some cleavage events we detected might be indirect effects, we performed an in vitro cleavage assay in which total RNA isolated from a rnt1Δ strain was incubated with varying amounts (0, 4, or 8 picomole (pmol)) of recombinant Rnt1 purified from E. coli (Figure 3C). This experimental approach has previously been used to detect Rnt1 cleavage of individual cellular RNAs by northern blot and uses a Rnt1 concentration lower than its in vivo nuclear concentration (see Materials and Methods). Instead of detecting individually cleaved mRNAs by blotting, we combined this approach with PARE (Figure 3C). We found that BDF2, CAF4, and YDR514C are cleaved by Rnt1 in vitro at the same sites that they are cleaved in vivo (Figure 3D), supporting the conclusion that Rnt1 directly and independently cleaves these mRNAs.

Figure 3:

Although ∼75% of the in vivo targets were also cleaved in vitro, a subset of targets were not detected by our in vitro approach (Figure 3E; Supplemental Table 1). This might be explained by several possibilities. First, the in vitro conditions (ion concentration, pH, temperature, etc.) may not perfectly reflect those of the cell and could potentially interfere with RNA secondary structure, thereby inhibiting cleavage of some targets in vitro. Second, our in vivo PARE analysis (which was performed in a rat1-ts xrn1Δ background) required incubation of the strains at 37 °C to inactivate rat1-ts, while the in vitro PARE used RNA isolated from a rnt1Δ strain grown at 26 °C. Thus, some mRNAs specifically expressed during heat stress would be included in the in vivo dataset but would not be detectable in the in vitro conditions. Third, considering the different strains used for our in vivo versus in vitro PARE experiments (i.e., rnt1Δ rat1-ts xrn1Δ vs rnt1Δ), there may be mRNAs for which degradation by Rnt1 is redundant with Xrn1- or Rat1-mediated degradation. Fourth, it is possible that Rnt1 may require a cofactor for the cleavage of some specific mRNAs, in which case cleavage would be carried out successfully in vivo, but not in vitro. Indeed, it has previously been shown that the Nop1 protein is required for Rnt1 cleavage of two snoRNAs that lack an obvious AGNN loop (Giorgi et al., 2001). Regardless of these possible limitations, our results show that most of the cleavage sites detected by PARE are direct products of Rnt1 and that Rnt1 does not require a cofactor for cleavage of most mRNA targets.

Interestingly, there were approximately 7 times as many Rnt1 targets cleaved in vitro than were detectable in vivo (Figure 3E). One example is MIG2 (Figure 3F). While our data concur with previous reports that MIG2 is cleaved by Rnt1 in vitro, cleaved MIG2 products did not accumulate to exceed our >1 cpm cutoff in the RNT1 strain (Figure 1D). It is possible that these targets are simply expressed at low levels in the cell, potentially during heat stress, and thus would not be detectable by our in vivo experiment. We hypothesized, however, that the drastic increase in Rnt1 targets in vitro could be due to the increased accessibility of these targets to Rnt1, while in vivo, these same targets are mostly cytoplasmic and therefore inaccessible to the primarily nuclear Rnt1.

Localization is a key determinant in Rnt1 mRNA selection and cleavage

Not surprisingly, colocalization of endonucleases with specific mRNAs is critical in determining endonuclease target selection in vivo. For example, in the Regulated Ire-Dependent Decay (RIDD) and tRNA Endonuclease-Initiated mRNA Decay (TED) pathways, endonucleases on the ER membrane and mitochondrial outer membrane, respectively, degrade mRNAs that are in close proximity due to co-translational protein sorting into those organelles (Hollien & Weissman, 2006; Hurtig et al., 2021; Tsuboi et al., 2015). Rnt1 localization is regulated during the cell cycle, with the enzyme being concentrated in the nucleolus during G1 phase but being more diffusely nuclear during G2/M phases (Catala et al., 2004). Additionally, Rnt1 cleavage of BDF2 is thought to increase during osmotic stress due to BDF2 retention in the nucleus under this condition (Wang et al., 2021). Thus, we hypothesized that some potential in vivo targets might escape Rnt1 cleavage due to their rapid cytoplasmic export.

To examine whether Rnt1 cleavage efficiency of mRNA targets is limited because Rnt1 is sequestered away from the bulk of the mRNA, we performed PARE using Rnt1 mutants that altered its localization within the cell. A rnt1-K45I mutant primarily localizes to the nucleolus, while a Rnt1 mutant lacking its nuclear localization sequence (rnt1-ΔNLS) localizes to the cytoplasm (Figure 4A). Both mutants are fully catalytically active and complement the growth phenotype of rnt1Δ (Catala et al., 2004) (Supplemental Figure 7A). As expected, rnt1-ΔNLS cleaved the BDF2, CAF4, and YDR514C mRNAs even more efficiently than wild-type Rnt1 (Figures 4B, C; Supplemental Figure 7B). The PARE peaks corresponding to the cleaved mRNAs became more pronounced, as did the corresponding bands detected by northern blotting. We did not observe this increase with the more nucleolar rnt1-K45I. Instead, we observed increased cleavage of ncRNA sites with nucleolar Rnt1, compared to RNT1 (Supplemental Figure 7C). Notably, although we observed in vitro cleavage of MIG2, this target was not detectably cleaved by cytoplasmic Rnt1 (Supplemental Figure 7D), indicating that even with increased access to mRNAs, Rnt1 likely does not cleave MIG2 in an in vivo context. Overall, these results suggest that cleavage of at least some mRNAs is regulated by the localization of Rnt1 and by mRNAs being predominantly cytoplasmic. Conversely, we speculate that Rnt1-mediated mRNA cleavage may be increased under conditions that slow mRNA export from the nucleus or slow Rnt1 import into the nucleolus.

Figure 4:

Rnt1-cleaved mRNAs are subsequently degraded by Xrn1

To investigate the fate of Rnt1-cleaved mRNAs, we used two complementary approaches. Three different fates have been described for RNAs cleaved by Rnt1. First, pre-snoRNAs are cleaved by Rnt1 upstream of the mature RNA or in between two snoRNAs in a polycistronic cluster, which allows for 5’ end maturation by Rat1 (Qu et al., 1999). Second, Rnt1 cleaves snRNAs, rRNAs, and some other RNAs co-transcriptionally downstream of their mature RNA ends, which allows for Rat1-mediated degradation of the nascent RNA and transcription termination (Ghazal et al., 2009). Third, we have previously shown that if a Rnt1 site is artificially introduced into a reporter mRNA, the 3’ cleavage product is exported to the cytoplasm and degraded by Xrn1 (Meaux et al., 2011).

Thus, to determine whether mRNAs are degraded upon Rnt1 cleavage, as well as which enzyme is responsible for this downstream degradation, we examined published xrn1Δ PARE data from Hurtig et al., 2021 and Hurtig & van Hoof, 2022. We found that cleaved targets accumulate in the xrn1Δ strain, as shown by clear peaks at the Rnt1 cleavage sites in BDF2, CAF4, and YDR514C (Figure 5A). This suggests that once these targets are cleaved by Rnt1, they are degraded by Xrn1. To confirm that Xrn1 is the enzyme responsible for degrading these mRNAs following Rnt1 cleavage, we used rat1-ts and xrn1Δ single mutants to analyze the accumulation of BDF2 and CAF4 cleaved transcripts by northern blot. We found that a deletion of XRN1 led to the accumulation of cleaved BDF2 and CAF4 mRNAs, while these cleaved transcripts were still degraded when only Rat1 was inactivated (Figure 5B). This indicates that Xrn1 is responsible for the degradation of Rnt1-cleaved mRNA targets and suggests two possibilities: (1) the 3’ cleavage product is exported from the nucleus, allowing for degradation by Xrn1 and avoiding Rat1, or (2) these mRNAs escape the nucleus uncleaved and are cleaved by a low amount of cytoplasmic Rnt1. The latter observation is inconsistent with previous findings that the BDF2 mRNA is cleaved upon nuclear retention (Wang et al., 2021). Furthermore, while Rnt1 is synthesized in the cytoplasm, its cytoplasmic concentration is very low (Figure 4), presumably because of very efficient nuclear import. Thus, we hypothesize that 3’ cleavage products are exported for Xrn1- mediated decay.

Figure 5:

Rnt1 cleavage is the major event for initiating the turnover of some mRNAs

Finally, we interrogated the biological relevance of Rnt1 mRNA cleavage. First, we investigated the overall contribution of Rnt1 cleavage to mRNA turnover. The usual methods to measure decay rates depend on the disappearance of the mRNA from a steady-state pool. Because the steady-state pool is predominated by cytoplasmic RNA, these methods are not suitable to assess nuclear degradation. Instead of substrate disappearance, we assessed the abundance of degradation intermediates. The major mRNA degradation pathway is through decapping (Muhlrad & Parker, 1992; Shyu et al., 1991), which leads to 5’ monophosphate ends at transcript start sites that can be detected by PARE (Figure 6A). This allowed us to estimate the contribution of Rnt1 cleavage to the overall mRNA “degradome” by examining the frequency of Rnt1 cleavage events compared to decapping (i.e., comparing number of PARE reads at the transcript start site and the cleavage site) for each mRNA target. Thus, the relative contribution to decay was estimated as the total number of reads in Rnt1-dependent peaks in a CDS, divided by the total number of reads that reflect both Rnt1 cleavage and decapping (Figure 6A). This analysis suggests that for nearly half (28) of our mRNA targets, Rnt1 performs at least 20% of their degradation (Figure 6B). In fact, for approximately one-third (19) of the targets, Rnt1 cleavage products were detected more frequently than decapping products. These highly cleaved targets include our top PARE hits (BDF2, CAF4, and YDR154C), indicating a substantial contribution of Rnt1 to the decay of a subset of mRNAs. We noted that most of these highly cleaved targets form a predicted stem loop (Supplemental Figure 6) and are cleaved both in vivo and in vitro (Supplemental Table 1). The proteins encoded by these targets, however, shared no common biological function or cellular localization.

Figure 6:

To determine whether Rnt1 cleavage of mRNAs affects gene expression levels, we re-analyzed published Rnt1 RNAseq data from Grzechnik et al., 2018. We observed that some mRNA targets such as CAF4, YDR514C, and MTM1 were highly upregulated in the absence of Rnt1, compared to wild type, while others like BDF2 were unaffected (Supplemental Figure 8). This suggests that Rnt1 cleavage may regulate the gene expression of certain mRNAs (like CAF4, YDR514C, and MTM1) but that others such as BDF2 may undergo more complex regulation to maintain their homeostatic levels.

Rnt1 cleavage of YDR514C mRNA regulates cellular homeostasis

As an alternative approach to interrogate the biological relevance of Rnt1 mRNA cleavage, we analyzed two kinds of genetic interactions. In one experiment, we focused on the rRNA processing function of Rnt1, since rRNAs are the most abundant Rnt1 product. If Rnt1 cleavage of the pre- rRNA is the main contributor to the rnt1Δ slow-growth phenotype, then mutating the Rnt1 cleavage site in the pre-rRNA should have the same effect as deleting RNT1. To do this, we used a yeast strain with all copies of the rDNA deleted from the genome (rDNAΔ) and replaced by a plasmid- borne rDNA (Figure 7A). The plasmid-borne gene contained either its wild-type AGNN tetraloop (AGGA) or a mutant ACGA or AUGA tetraloop that impairs Rnt1 cleavage. Fascinatingly, the ACGA- and AUGA-containing strains grew as well as the AGGA control, suggesting that a defect in rRNA cleavage is not the major contributor to the growth phenotype of rnt1Δ.

Figure 7:

In a second experiment, we identified spontaneous suppressors that restored growth to a rnt1Δ strain. To isolate suppressors, we grew 13 cultures of a rnt1Δ strain to saturation, diluted them 1000-fold, and repeated the growth for 10 cycles (Figure 7B). Each of the final cultures grew substantially faster than the starting strain (Supplemental Figure 9A). We then isolated an individual colony from the final culture and sequenced its genome. Most of the mutations we identified were in RRP6, RRP47, and MTR4, which encode cofactors of the nuclear RNA exosome (Supplemental Figure 9A). This suppression probably reflects the fact that Rnt1 and the RNA exosome are both involved in processing rRNA, snRNA, and snoRNA precursors (Allmang et al., 1999; Chanfreau et al., 1997; Elela & Ares, 1998; Seipelt et al., 1999) but is not informative as to which of these RNA processing defects causes the slow growth of rnt1Δ. Our most interesting mutation, however, was in the newly identified Rnt1 target YDR514C (Supplemental Figure 9B). YDR514C encodes a putative nuclease of unknown function, and our spontaneous suppressor strain (evolved strain 13; Figure 7B) contained a G220S mutation in the nuclease domain (Supplemental Figure 9B). Of note, this glycine is 100% conserved in the Saccharomycetaceae (budding yeast) family. While most of the spontaneous suppressors contained a single mutation, the evolved strain 13 contained a second mutation in PUF4. To determine whether the suppression observed in this evolved strain was due to the YDR514C mutation, the PUF4 mutation, or both, we constructed rnt1Δ ydr514cΔ and rnt1Δ puf4Δ double-mutants. Interestingly, we observed suppression of the rnt1Δ growth defect in both mutants (Figure 7C). This suggests that lack of YDR514C cleavage may contribute to the rnt1Δ slow-growth phenotype and that this genetic interaction plays a role in maintaining normal cell growth. We conclude that Rnt1 maintains mRNA homeostasis by regulating the turnover and/or the gene expression of specific mRNAs and that its cleavage of at least one mRNA target contributes to overall cellular homeostasis.

Discussion

Expanding the Rnt1 target repertoire

Here, we greatly expand the known mRNA target repertoire of Rnt1 and highlight the biological importance of its mRNA cleavage function. Using PARE, we identify a novel set of 60 mRNAs that are cleaved by Rnt1 in vivo. As with ncRNA targets, we find that Rnt1 cleaves mRNAs that possess double-stranded stem loops with terminal AGNN tetraloops and that it does not require a cofactor for cleavage of most mRNAs. Instead, most mRNAs cleaved in vivo can also be cleaved in vitro. For the small subset of mRNAs that were not cleaved in vitro, further analysis is needed to determine whether these targets fold differently in vitro; are only expressed (or fold differently) during heat stress; can be degraded by Rat1 or Xrn1 without Rnt1-specific cleavage; and/or require a cofactor for Rnt1 to cleave.

Rnt1-cleaved mRNAs encode proteins with no common biological function or subcellular localization. However, the localization of the mRNA targets to the cytoplasm, in contrast to Rnt1’s nuclear localization, limits cleavage of these targets. This partially explains why Rnt1 cleaves only specific mRNAs in vivo, despite its capacity to cleave a wider set of mRNAs in vitro. One outstanding question is whether mRNAs efficiently cleaved by Rnt1 remain in the nucleus for longer periods, perhaps due to slower export rates, compared to mRNAs that are less efficiently cleaved.

Once these mRNA targets are cleaved by Rnt1, they are degraded by Xrn1, not by Rat1. This likely reflects that Rnt1-cleaved mRNAs are exported to the cytoplasm for further degradation by Xrn1. In support of the alternate possibility that some mRNAs can be cleaved in the cytoplasm, we show that Rnt1 relocalized to the cytoplasm retains its ability to cleave mRNAs. This would also present the most energetically economic scenario for the cell. However, Rnt1 cytoplasmic localization conflicts with the scarcely detectable level of Rnt1 in the cytoplasm and with our current understanding of the localization of this well-studied nuclease. It has also been shown that BDF2 cleavage increases upon nuclear retention (Wang et al., 2021) and that the 5’ cleavage product of BDF2 accumulates in the absence of the nuclear exosome subunit Rrp6 (Roy & Chanfreau, 2014). Neither of these would be expected if BDF2 were cleaved by a pool of cytoplasmic Rnt1. Nuclear cleavage by Rnt1 followed by export of the 3’ cleavage product to the cytoplasm seems most consistent with all the available data.

This study has also expanded the number of known Rnt1 targets in other classes of RNAs, specifically UTRs of mRNAs, pre-mRNA introns lacking snoRNAs, antisense transcripts, and regions annotated as intergenic. These sites may be cleaved to initiate degradation or a processing event, and cleavage may be essential for the regulation of these targets or may be redundant with other pathways. Of the intronic sites, 5 did not contain snoRNAs. Most previously reported Rnt1-cleaved introns encode snoRNAs, with exceptions being the spliced lariat intron of RPL18A, which is degraded upon Rnt1 cleavage, and the first intron of RPS22B, which is cleaved to initiate the degradation of unspliced RPS22B transcripts (Danin-Kreiselman et al., 2003). Thus, these 5 newly identified Rnt1-cleaved introns lacking snoRNAs dramatically expand this class of targets. It remains to be determined whether these introns are cleaved to trigger the degradation of unspliced transcripts or whether cleavage and degradation occur after splicing. Also of note, the intergenic regions targeted by Rnt1 included 3 that contain uncharacterized ncRNAs (Gao et al., 2021; Nagalakshmi et al., 2008), possibly representing a novel class of Rnt1 targets. These sites may be cleaved to liberate the ncRNAs, while other intergenic sites may be cleaved to prevent transcriptional read-through, as in the case of the NPL3-GPI17 dicistronic transcript (Ghazal et al., 2009).

Biological relevance of Rnt1 mRNA cleavage

There are diverse implications of Rnt1 mRNA cleavage. First, Rnt1 appears to recognize the same secondary structures for mRNA degradation and ncRNA processing. We show that compared to decapping, many mRNAs rely heavily on Rnt1 to initiate turnover, and these mRNAs tend to have strong matches to the AGNN stem-loop motif. For 7 targets in particular, almost all of the degradation products we detected reflect Rnt1-initiated degradation. For others, decapping products are more abundant than Rnt1 products, and a number of these primarily decapped mRNAs did not have an obvious, strong predicted stem loop structure. We suspect that they may form weak stem loops and/or may adopt multiple alternative structures in vivo. Specific RNA binding proteins may also affect cleavage by affecting the mRNA structure, analogous to Nop1’s effect on cleavage of the SNR18 and SNR38 snoRNAs. Additionally, other factors including the sequences surrounding the AGNN loop and those around the cleavage site affect Rnt1 cleavage efficiency (Lamontagne et al., 2003, 2004). The example of YPL277C indeed highlights the importance of the stem for target selection, as a single base difference from YOR398W in the stem sequence affects cleavage by Rnt1. Another notable target we identified was MTM1, which did not form the typical extended stem loop recognized by Rnt1. We suspect that its unique structure results in cleavage at a single site. Overall, Rnt1 cleavage appears to be finely tuned, and the targets that heavily depend on Rnt1 for turnover implicate the enzyme in regulating a range of biological pathways including protein targeting and membrane assembly, DNA damage response, and various chemical biosynthesis pathways. It will be interesting to dissect the biological consequences of Rnt1 cleavage of individual mRNA targets.

Second, as we observed with published Rnt1 RNAseq data (Grzechnik et al., 2018), Rnt1 cleavage appears to affect the gene expression levels of several mRNAs, including YDR514C. However, the steady-state level of mRNAs determined by RNAseq includes many secondary effects. The fact that the protein-coding targets reported here are cleaved within their coding regions and subsequently degraded also supports our reasoning that a subset of mRNAs are cleaved as a means of regulating their expression levels. In this case, it follows that cleavage of some targets like YDR514C may be critical for maintaining cellular homeostasis. Others like BDF2 are not upregulated when RNT1 is deleted, indicating that these targets may be regulated by more intricate mechanisms. For example, BDF2 is also regulated by the spliceosome-mediated decay (SMD) pathway, which is redundant with cleavage by Rnt1 (Roy & Chanfreau, 2014).

Third, in relation to the previous point, genetic interaction(s) between RNT1 and YDR514C suggest that Rnt1-mediated cleavage of YDR514C mRNA is important for normal cell growth. The function of Ydr514c is unknown, but we postulate that its expression needs to be tightly controlled. We hypothesize that rnt1Δ results in aberrant YDR514C expression, which is detrimental to growth. This explains why a rnt1Δ strain is slow-growing, but a deletion of YDR514C partially suppresses the slow-growth phenotype. Uncovering the function of this gene would provide better insight into its mechanism of suppression. The lack of complete suppression, however, indicates that this is not the only detrimental effect in a rnt1Δ strain. We identified mutations in PUF4 and genes encoding the nuclear RNA exosome cofactors Rrp6, Rrp47, and Mtr4 as other suppressors of rnt1Δ. Puf4 is an mRNA binding protein that regulates the stability of hundreds of mRNAs that are required for ribosome biogenesis, and Rrp6, Rrp47, and Mtr4 are all required for rRNA maturation. This suggests that Rnt1 indeed plays a role in ribosome biogenesis that affects growth, and we can envision three possible mechanisms for this. First, Rnt1 plays a direct role in pre-rRNA processing. However, mutating the AGNN tetraloop in pre-rRNA had surprisingly little effect on growth, suggesting that this function of Rnt1 is not growth-rate determining. Second, Rnt1 has a function in the maturation of pre-snRNAs that are critical for splicing, and introns are enriched in ribosomal genes. Third, Rnt1 and the RNA exosome cofactors are both required for snoRNA processing, which in turn are required for ribosome biogenesis. Strikingly, the snoRNP subunits Nop1, Nop56, Nop58, Gar1, Nhp2, Nop10, and Cbf5 have all been identified as targets of Puf4 (Lapointe et al., 2017). Deletion of most individual snoRNAs have no effect on growth, but a few are essential (e.g., U3 snoRNA). Thus, defects in processing one of these essential snoRNAs or the simultaneous defect in processing multiple snoRNAs may explain the rnt1Δ growth defect and its suppression by RNA exosome and puf4 mutations.

Although our results suggest that Rnt1 affects growth through cleavage of the YDR514C mRNA under standard lab conditions, the other identified mRNA targets affect a vast range of cellular processes and may affect growth under specific conditions. Furthermore, the broad distribution of Rnt1 cleavage sites throughout the yeast transcriptome in regions other than mRNA CDSs and ncRNAs (i.e., 3’ and 5’ UTRs, intergenic regions, introns, and antisense transcripts) suggests to us that the cleavage function of Rnt1 may exert its effect on multiple cellular processes and pathways not yet discovered for the enzyme. An additional possibility is that Rnt1 is important for the eradication of transcripts that are poorly exported from the nucleus and/or under conditions in which export is blocked. In this role, Rnt1 would act as a quality control on nuclear export competence similar to the role of nonsense-mediated mRNA decay in the quality control of translation. Our finding that most mRNA targets contain sites that can be cleaved in vitro suggests that such a quality control mechanism could survey many mRNAs.

Implications for human RNase III

While the details vary, the fundamentals of Rnt1 function resemble those of its human homolog Drosha (Nicholson, 2014). As RNase III family enzymes, they both cleave double-stranded stem loops and leave 2-nt 3’ overhangs (Court et al., 2013; Elela et al., 1996; Y. Lee et al., 2003). Both are also nuclear enzymes with prominent roles in the processing of ncRNAs (Chanfreau et al., 1997; Chanfreau, Rotondo, et al., 1998; Elela et al., 1996; Y. Lee et al., 2003), and both have previously been described to cleave a few select mRNAs (Han et al., 2009; Karginov et al., 2010; A. Lee et al., 2005; Roy & Chanfreau, 2014). Although Rnt1 and Drosha differ in their ncRNA targets (rRNA, snRNA, and snoRNA vs miRNA) and in the way they recognize substrates, each enzyme utilizes the same mechanism to cleave its respective ncRNA and mRNA targets (Nicholson, 2014).

The ability of Drosha and Rnt1 to cleave mRNAs suggests that eukaryotic RNase III family members are multifunctional and active in mRNA degradation, in addition to their earlier discovered ncRNA processing functions. We speculate that the mRNA cleavage role of Rnt1 has been more difficult to detect because only a fraction of each individual mRNA is cleaved, and such nuclear cleavage of a subset of RNAs is not readily detectable by methods that measure mRNA stability by disappearance of a steady-state pool. This limitation would equally apply to Drosha, and therefore it appears likely that Drosha may also play a more prominent role in mRNA degradation than has been previously described. It would be interesting to see whether PARE could detect a more extensive effect of Drosha cleavage on the human protein-coding transcriptome. Performing this in a genetic background devoid of XRN1 and the Rat1 ortholog XRN2 makes such PARE experiments challenging, but our results imply that an XRN1Δ background may be sufficient to identify at least some Drosha targets. A recent study has uncovered two independent missense variants in DROSHA as a cause of neurodevelopmental disease (including severe intellectual disability, white matter atrophy, microcephaly, epilepsy, and dysmorphic features) (Barish et al., 2022), making the need for a full understanding of Drosha function even more important.

Applications and advantages of PARE

As mentioned, PARE can be used for the detection of non-classical targets of other well-known endonucleases. Cellular conditions can also be modulated to determine the specific circumstances under which these enzymes might cleave atypical classes of substrates. Additionally, this technique might be applied to the study of poorly characterized proteins that appear to be endonucleases. This may include proteins that possess a predicted nuclease domain, interact with other RNA processing or decay factors, and/or localize to sites of RNA processing or degradation in the cell.

Combining the current characterization of Rnt1 with our previous analysis of TSEN reveals some informative commonalities. Rnt1 and TSEN both have very well characterized functions in the processing of stable ncRNAs, and inactivating either one of these enzymes causes a large reduction in growth rate and many downstream effects. This makes it challenging to find direct RNA substrates through traditional RNA sequencing. In contrast, by sequencing RNA cleavage products, PARE can capture the primary products of the cleavage events as they have occurred in the cell. In the case of TSEN, we showed that its inactivation induced the general stress response, and the large number of genes affected by the general stress response made the direct TSEN targets undetectable by RNAseq. PARE, however, readily detected TSEN-cleaved mRNAs. Similarly, several previous studies have identified many effects of Rnt1 but failed to detect prominent mRNA cleavage events (other than BDF2). We suspect that there are other RNases well-characterized for other functions that may moonlight as mRNA degradation enzymes.

Materials and methods

Yeast strains and plasmids

All yeast strains and plasmids used in this study are listed in Supplemental Tables 2 and 3. The rnt1Δ::NEO, puf4Δ::NEO, and ydr514cΔ::NEO single mutants were obtained from the MATa haploid yeast gene knockout collection (http://sequence-www.stanford.edu/group/yeast_ deletion_project/deletions3.html). The rat1-ts::URA3 strain was obtained from the Hieter essential yeast gene temperature-sensitive collection (Kofoed et al., 2015). The xrn1Δ::HYG strain was created by Hurtig et al., 2021. The rnt1Δ::HYG strain was created by standard cloning methods. Briefly, the pAG32 vector containing the hygromycin resistance (HYG^R) gene cassette was cut with HindIII-HF (New England BioLabs, cat. R3104S, USA) and SpeI-HF (New England BioLabs, cat. R3133S, USA) restriction enzymes and vector fragments were separated on a 1% agarose gel. The HYG^Rcassette (∼1.75 kb) was purified using the Zymoclean Gel DNA Recovery Kit (Zymo Research, cat. D4008) and transformed into the rnt1Δ::NEO strain using standard LiAc transformation methods. Positive transformants were selected on YPD + Hygromycin plates and counter-selected on YPD + Neomycin plates. All other strains were created by standard genetic crosses and LiAc transformations. Plasmids containing the rDNA sequence with 3’ ETS tetraloop mutated to ACGA or AUGA were created by site-directed mutagenesis using oligos listed in Supplemental Table 4.

Yeast growth conditions

All strains used for in vivo PARE experiments were grown at room temperature, then shifted to 37 °C for 1 hour to inactivate rat1-ts. The rnt1Δ::HIS3 strain used for in vitro PARE, rnt1Δ::NEO strains used for fluorescence microscopy, and all strains used for experimental evolution were grown at 30 °C. For solid-media growth assays, temperature-sensitive strains were grown at 27 °C, and all other strains at 30 °C. Liquid cultures of strains containing plasmids were grown in SC- Leu, and all other strains were grown in YPD. All solid-media growth assays were spotted on YPD.

Solid media growth assays

Yeast cultures were grown to mid-log phase (OD600 of ∼0.6) at room temperature for temperature- sensitive strains or at 30 °C for all other strains. Cells were washed, resuspended to an OD600 of 0.6, then serially diluted and spotted on YPD.

RNA isolation

For in vivo PARE experiments and northern blots, yeast cells were grown in 20-mL cultures to mid-log phase (OD600 of 0.6-0.8) at 27 °C, then shifted to 37 °C for 1 hour to inactivate rat1-ts. Cells were harvested by centrifugation, and RNA was isolated using hot phenol as previously described by He et al., 2008. In brief, cells were resuspended in 500 µL cold RNA buffer A (3M NaOAc, 0.5 M EDTA pH 8.0) and 500 µL phenol heated to 65 °C, then incubated at 65 °C for 4 min, with vortexing for 10 sec per minute. The cell suspension was then centrifuged and the aqueous layer collected and mixed with 500 µL phenol heated to 65 °C. The incubation, vortexing, centrifugation, and aqueous layer collection steps were repeated as described above. The aqueous layer was then mixed with phenol:chloroform (1:1), centrifuged, and removed to a new tube. Total RNA was precipitated overnight in 100% ethanol at -80 °C. For in vitro PARE, the rnt1Δ::HIS3 strain was grown to mid-log phase (OD600 of 0.4-0.6) at 26 °C and a 50 mL culture was harvested for total RNA isolation as described by Catala & Elela, 2019.

In vitro cleavage assay

25 μg total RNA isolated from rnt1Δ::HIS3 was incubated with either 0, 4, or 8 pmol of recombinant Rnt1 purified from E. coli in 100 µL reaction buffer (30 mM Tris pH 7.5, 150 mM KCl, 5 mM spermidine, 0.1 mM DTT, 0.1 mM EDTA) as described by Catala & Elela, 2019. MgCl2 was added to a final concentration of 10 mM to initiate the cleavage reaction, which was allowed to proceed for 20 min at 30 °C. The reaction was stopped by adding 1 volume LETS buffer (100 mM LiCl, 10 mM EDTA, 10 mM Tris, 0.2% SDS pH 7.5), and the cleaved RNA was isolated using phenol:chloroform:isoamyl alcohol (25:24:1). The isolated RNA was then submitted to LC Sciences for PARE.

The abundance of Rnt1 has been estimated to be ∼2,512 molecules/cell (Ho et al., 2018). Given that the yeast cell nucleolus is about 0.6 fL, if 2,512 molecules of Rnt1 were uniformly distributed through the nucleolus its concentration would be 7 µM (2,512 molecules/nucleolus in 0.6 fL) (Uchida et al., 2011), or approximately 100 times higher than our highest in vitro concentration (80nM).

Parallel Analysis of RNA Ends (PARE)

PARE was performed as previously described by Hurtig et al., 2021. 10 µg of RNA isolated from each strain, or 10 µg of RNA cleaved by recombinant Rnt1, was submitted to LC Sciences.

LC Sciences PARE workflow

PolyT-conjugated magnetic beads are used for two rounds of polyA-RNA purification, NEBNext Ultra II RNA Library Prep Kit (New England BioLabs, cat. E7770, USA) is used for cDNA library preparation, and T4 RNA ligase is used to ligate 5’ adapters onto 5’-monophosphates. Reverse transcription is performed using a 3′-adapter random primer, and the cDNAs are amplified by PCR (3 min at 95 °C; 15 sec at 98 °C for 15 cycles; 15 sec at 60°C, 30 sec at 72 °C; and 5 min at 72 °C). AMPureXP (Beckman Coulter, ref. A63882, USA) beads are used for size selection (200-400 bp), and 50-bp single-end sequencing is performed with Illumina Hiseq 2500.

Bioinformatic analyses

PARE

FastQ Groomer, FastQC, TopHat, bamCoverage, and bigwigCompare were used to analyze raw PARE sequencing data (Supp. Figure 1). The resulting bed graphs were combined using bedtools Merge BedGraph files, and the combined bedgraphs (forward and reverse strands separate) were filtered to include only positions with reads >1 cpm in the RNT1 strain (Supp. Figure 1). Bedgraphs were exported for further analysis in Microsoft Excel, and bigwig files were analyzed in the Integrative Genomics Viewer (IGV) software. A list of all published Rnt1 cleavage sites was compiled by mining the literature, and a BED file of these sites (which includes the chromosome, strand, and start and end positions of cleavage) was created. This BED file was analyzed alongside PARE data generated in this study.

Whole-genome sequencing

The rnt1Δ::NEO parent strain and evolved strains were submitted to SeqCenter for whole-genome sequencing. The returned raw sequencing reads were trimmed using Trim Galore!, quality control was performed using FastQC, and reads were mapped to the yeast genome using Bowtie2. FreeBayes was used to detect variants. Candidate mutations were visualized in IGV, and VEP (https://useast.ensembl.org/Tools/VEP) was used to annotate the effects of the identified mutations, including affected genes.

Northern blots

Northern blots of SNR83 were performed by separating 10 μg of total RNA from the indicated strains on a 1.3% denaturing agarose formaldehyde gel in 1X MOPS buffer (MOPS, NaOH to pH 7.0, 3M NaAc, 0.5 M EDTA pH 8). RNA samples were resuspended in 10 μL formaldehyde loading dye (Invitrogen, ref. 8552, USA) prior to loading. Northern blots of BDF2 and CAF4 were performed by separating 7 μg of total RNA in formaldehyde loading dye on 1.3% denaturing agarose gels following poly-A RNA enrichment using Poly(A)Purist-MAG Magnetic mRNA Purification Kit (Invitrogen, ref. AM1922, USA). RNA was transferred from the gel onto a charged nylon membrane, UV-crosslinked to the membrane, and hybridized with P³²-labeled gene-specific oligonucleotide probes (Supp. Table 4) overnight at 42 °C. Blots were visualized using the Cytiva Amersham Typhoon phosphor imager. All oligonucleotide sequences used to generate probes for this study are listed in Supplemental Table 4.

RNA structure predictions

Secondary structures of mRNA targets were predicted in MFold (https://bio.tools/mfold) using ∼60-nt sequences that encompassed the Rnt1 cleavage sites detected by PARE.

Sequence alignments

Predicted AGNN tetraloop sequences of mRNA targets (plus 3 nts up- and downstream of the tetraloop) were aligned using WebLogo (https://weblogo.berkeley.edu/logo.cgi).

Gene ontology analysis

Gene ontology analysis was performed for all mRNA targets using YeastEnrichr (https://maayanlab.cloud/YeastEnrichr), PANTHER (https://pantherdb.org/), and Saccharomyces Genome Database Gene Ontology Term Finder (https://www.yeastgenome.org/goTermFinder).

Confocal fluorescence microscopy

Yeast strains were grown in 10-mL cultures to late log phase (OD600 of ∼1.0) at 30 °C, then harvested by centrifugation and fixed overnight in 1% paraformaldehyde (paraformaldehyde, 1 M NaOH, 10X PBS, HCl to pH 7.4) at 4 °C. Cells were then harvested again, permeabilized with 0.1X Triton for 30 min at 27 °C, and stained with NucBlue Fixed Cell Stain ReadyProbes reagent (Invitrogen, ref. R37606, USA) for 1 hr at 27 °C. (Cells were washed three times in 1X PBS prior to fixation, permeabilization, and staining.) Cells were washed and resuspended in 1X PBS, mounted to slides using ProLong Diamond Antifade Mountant (Invitrogen, ref. P36965, USA), and imaged on the Olympus Fluoview FV3000 laser scanning confocal microscope. Images were analyzed in the cellSens Dimension software.

Experimental evolution

Thirteen 10-mL cultures of the rnt1Δ::NEO parent strain were grown to saturation in YPD at 30 °C, then diluted 1:1000. This was repeated 10 times, then cultures allowed to grow to mid-log phase (OD600 of ∼0.6). Strains were streaked on YPD + Neomycin plates, and cells were washed, resuspended to an OD600 of 0.6, serially diluted (1:5), and spotted on YPD. Evolved strains showing a suppression of the rnt1Δ growth defect were analyzed by whole-genome sequencing.

Protein structure modeling

Suppressor mutations in nuclear RNA exosome cofactors were modeled using the published structure of the exosome (Schuller et al., 2018) (RCSB PDB: 6FSZ) in the Chimera software. The Ydr514c protein structure was predicted using the AlphaFold 3 online server (Abramson et al., 2024) (https://alphafoldserver.com/).

Supplemental figures and tables

Supplemental Figure 1:

Supplemental Figure 2:

Supplemental Figure 3:

Supplemental Figure 4:

Supplemental Figure 5:

Supplemental Figure 6:

Supplemental Figure 7:

Supplemental Figure 8:

Supplemental Figure 9:

Table S2:

Table S3:

Table S4:

Acknowledgements

This work was funded by the National Institutes of Health (NIH) grant R35GM141710 to A.v.H. and NIH grant F31GM149143 to L.N.S. We are also very thankful to members of the van Hoof and Abou Elela labs for their thoughtful comments and critiques on this manuscript.

Additional files

Table S1.