Rae1/YacP, a new endoribonuclease involved in ribosome‐dependent mRNA decay in Bacillus subtilis

Abstract

The PIN domain plays a central role in cellular RNA biology and is involved in processes as diverse as rRNA maturation, mRNA decay and telomerase function. Here, we solve the crystal structure of the Rae1 (YacP) protein of Bacillus subtilis, a founding member of the NYN (Nedd4‐BP1/YacP nuclease) subfamily of PIN domain proteins, and identify potential substrates in vivo. Unexpectedly, degradation of a characterised target mRNA was completely dependent on both its translation and reading frame. We provide evidence that Rae1 associates with the B. subtilis ribosome and cleaves between specific codons of this mRNA in vivo. Critically, we also demonstrate translation‐dependent Rae1 cleavage of this substrate in a purified translation assay in vitro. Multiple lines of evidence converge to suggest that Rae1 is an A‐site endoribonuclease. We present a docking model of Rae1 bound to the B. subtilis ribosomal A‐site that is consistent with this hypothesis and show that Rae1 cleaves optimally immediately upstream of a lysine codon (AAA or AAG) in vivo.

Introduction

The Gram‐positive model organism Bacillus subtilis currently has 19 known ribonucleases, consisting of eight exoribonucleases and 11 enzymes that cleave RNA endonucleolytically (Condon, 2014). Some of these have relatively specialised functions, such as RNase M5 in 5S rRNA maturation (Condon et al, 2001), and RNase P and RNase Z in tRNA maturation (Hartmann et al, 2009), while others have both stable RNA and mRNA substrates. One of the best known enzymes of this latter class is the 5′‐3′ exoribonuclease RNase J1 that matures the 5′ end of 16S rRNA (Britton et al, 2007; Mathy et al, 2007) and is involved in the degradation of many mRNAs and mRNA fragments (Durand et al, 2012). RNase Y is also involved in both mRNA degradation and stable RNA processing, notably the scRNA (4.5S RNA) and the RNA subunit of RNase P (Lehnik‐Habrink et al, 2011; Durand et al, 2012; Laalami et al, 2013; Gilet et al, 2015). Despite the large number of RNases identified in B. subtilis to date, there remain both orphan substrates, i.e. processed RNAs for which the enzymes are not yet known, and orphan enzymes, i.e. predicted RNases encoded by the genome of B. subtilis whose substrates have not been identified, that are ripe for further study.

In this paper, we describe the characterisation of an orphan enzyme of B. subtilis called YacP that we have renamed Rae1 (ribosome‐associated endonuclease 1). This enzyme was predicted to be a ribonuclease related to the PIN family (PilT N‐terminal) of RNases (Anantharaman & Aravind, 2006). An extensive bioinformatic analysis showed that members of this so‐called NYN (Nedd4‐BP1/YacP nuclease) subfamily of PIN domain enzymes are found in all three kingdoms of life and suggested that they are an ancient class of RNA processing enzymes that can be traced all the way back to the last universal common ancestor (LUCA). The rae1 gene in B. subtilis is the last cistron in a large operon encoding a number of proteins involved in translation, including the glutamyl‐ and cysteinyl‐aminoacyl‐tRNA synthetases (gltX and cysES), the 23S rRNA processing enzyme Mini‐RNase III (mrnC) and the 23S rRNA methyltransferase (rlmB). This organisation is conserved in the Firmicutes (Fig EV1) and suggested that Rae1 may have a role in a translation‐related process in these organisms.

Figure EV1. Synteny of rae1 locus in bacteria.

1. Structure of rae1 locus in Firmicutes. The rae1 gene is shown in red. Genes are labelled for Bacillus subtilis. Conserved genes have the same colour. The per cent co‐occurrence of genes immediately surrounding rae1 is given for 268 Firmicutes.
2. Structure of rae1 locus in Cyanobacteria. The rae1 gene is shown in red. Genes are labelled for Synechocystis elongatus. Conserved genes have the same colour. The per cent co‐occurrence of genes immediately surrounding rae1 is given for 35 Cyanobacteria.

The NYN/PIN domain is found to be fused to a large variety of other domains throughout biology. Some of these are basic RNA binding domains, such as the Zn‐finger CCCH domain, the K‐homology (KH) domain, the RNA recognition motif (RRM) domain or pentatricopeptide repeats (PPR) (Anantharaman & Aravind, 2006). Much larger proteins related to the TetM/TetO family of tetracycline resistance proteins are also found with C‐terminal NYN domains. Although the larger PIN family has been widely characterised (Arcus et al, 2011), only a few NYN subfamily proteins have been studied in any detail. MCPIP1/Zc3h12a/ Regnase‐1 has been shown to be a Zn‐finger containing RNase involved in dampening the mammalian immune response by cleaving the IL‐6 and IL‐12β mRNAs (Matsushita et al, 2009) and to play a role in antagonising micro‐RNA (miRNA) biogenesis by cleaving pre‐miRNAs in their terminal loops (Suzuki et al, 2011). It also has potent broad‐spectrum antiviral activity through degradation of viral RNAs (Lin et al, 2013, 2014). The PPR‐containing NYN protein PRORP is a protein‐only form of the well‐known tRNA 5′ processing ribozyme RNase P and is found in mammalian and plant organelles (Holzmann et al, 2008; Gutmann et al, 2012).

The bacterial Rae1 enzyme consists of an N‐terminal catalytic (PIN‐like) domain and a short highly positively charged helical domain that resembles a number of ribosomal RNA binding proteins. Rae1 is conserved in the Firmicutes, the Cyanobacteria, algae and higher plants, where it is predicted to be localised principally to chloroplasts (Emanuelsson et al, 2007). Despite the identification of this putative novel B. subtilis ribonuclease as a founder member of the NYN/PIN family of RNases more than 10 years ago, nothing is yet known about its function in this organism. To get a handle on its biological role, we employed parallel approaches of crystal structure resolution and RNA sequencing (RNAseq) of a ∆rae1 mutant and a complemented strain. The crystal structure shows that Rae1 has an N‐terminal NYN/PIN domain and a highly positively charged flexible C‐terminal domain, comprised of two α‐helices that likely serve in RNA binding. The RNAseq data show that Rae1 is involved in the degradation of a subset of B. subtilis mRNAs in rich medium. We further demonstrate that RNA cleavage by Rae1 is dependent on both the translation and reading frame of one of these substrates, making it a good candidate for a ribosome acceptor (A)‐site ribonuclease.

Results

Crystal structure of Bacillus subtilis Rae1 (YacP)

The wild‐type (WT) Rae1 protein from B. subtilis and a fortuitously isolated mutant (W164L) were overexpressed in E. coli as C‐terminal His‐tagged fusion proteins. Both proteins behaved as monomers in solution with an apparent molecular weight of 25 kDa (see Appendix Fig S1A for the WT protein). The W164L derivative was first to crystallise and its structure was solved at 3.2 Å resolution using sodium iodide‐soaked crystals and single isomorphous replacement with anomalous scattering (SIRAS) for phase determination. A different set of crystallisation conditions were subsequently obtained for the WT Rae1 protein and its structure was solved at 2.25 Å resolution by molecular replacement using the structure of the W164L mutant as a model. The refinement statistics for both proteins is given in Appendix Table S1.

The N‐terminal 120 amino acids of Rae1 form a PIN‐related fold consisting of a central five‐stranded parallel β‐sheet associated with five helices, two behind and three in front (Fig 1A and Appendix Fig S1B). The first two helices of Rae1 (α1 and “α2”) are orthogonal to the plane of the β‐sheet, while the remaining three are in antiparallel orientation.

Figure 1. Structure of Rae1 monomer at 2.25 Å resolution.

1. Ribbon representation of Rae1 superimposed on a transparent surface map. α‐helices and β‐strands of the NYN/PIN domain are shown in red and blue, respectively. The C‐terminal extension is shown in green. Helices α2 and α5 are not recognised as perfect α‐helices and are thus labelled as “α2” and “α5”.
2. Ribbon and stick representation of the Rae1 catalytic site superimposed on that of mammalian MCPIP1 (PDB 3V34). Key aspartic acid residues are shown as orange sticks. MCPIP1 bound to Mg²⁺ is shown in grey. Key residues of Rae1 are labelled in black; equivalent residues of MCPIP1 are labelled in grey.
3. Surface potential map of Rae1, with positively charged residues in blue and negatively charged residues in red. Scale is from −5 to +5 k_b T e_c ⁻¹.
4. Structure and charge distribution of r‐protein S13 from E. coli, with positively charged residues in blue and negatively charged residues in red. Scale is from −5 to +5 k_b T e_c ⁻¹.

The Rae1 catalytic site

Unlike other PIN domain proteins in which the active site is partially covered by a “lid” (Appendix Fig S2), the active site of Rae1 is remarkably exposed at the surface of the protein (Fig 1A and B). This is an unusual configuration for a ribonuclease, leaving open the possibility that another partner binds to this surface of Rae1, creating a bipartite pocket around the active site. Three conserved aspartic acid residues, D7, D81 and D104, occupy positions equivalent to the magnesium (Mg)‐coordinating Asp residues (D141, D226, D244) of MCPIP1 and are located at the C‐terminal ends of strands β1 (D7), β4 (D104) and the N‐terminal end of helix α4 (D81) (Fig 1B and Appendix Fig S1B and C). Mg²⁺ is not visible in the Rae1 structure, suggesting that it is provided by the RNA substrate. In MCPIP1, a fourth key Asp (D225) residue lies adjacent to D226 at the N‐terminus of helix α4 (Xu et al 2012). D226 makes direct contact with the catalytic Mg²⁺, while D141, D225 and D244 bridge to the metal ion via a network of five water molecules. Interestingly, Rae1 differs from MCPIP1 and the other NYN proteins in the PDB by having only one Asp residue at the N‐terminus of helix α4. A fourth Asp (D53), located at the C‐terminus of β2, occupies a position not too distant from MCPIP1 D225 in the catalytic pocket and may serve a similar function (Fig 1B). In the current configuration, all four Asp residues of Rae1 are too far (>3.5 Å) from the predicted position of the Mg²⁺ ion to make direct contact. Asp7 and Asp53 are each hydrogen‐bonded to a water molecule (water 116 in the A‐subunit and 72 in the B‐subunit, respectively) that could allow them to bridge to the Mg²⁺ ion (Fig EV2). Most likely the other Asp residues only move into place or bridge via additional water molecules once the substrate and Mg²⁺ are present, but it is clear that the network is incomplete without these two elements.

Figure EV2. Hydrogen bond networks in the catalytic site of Rae1 crystal structure.

1. Water molecules in the catalytic site of the A‐subunit of the Rae1 crystallographic dimer. Hydrogen bonds are shown as dotted yellow lines with distances are given in Å. Key catalytic residues are labelled. The Mg²⁺ ion (grey) is from MCPIP1 and was placed using the superposition shown in Fig 2.
2. Water molecules in the catalytic site of the B‐subunit of the Rae1 crystallographic dimer. Hydrogen bonds are shown as dotted yellow lines with distances are given in Å. Key catalytic residues are labelled. The Mg²⁺ ion (grey) is from MCPIP1 and was placed using the superposition shown in Fig 2.

One of the distinguishing features of the NYN subfamily of PIN nucleases is a conserved Asn residue (Asn10), three residues downstream of the first catalytic aspartate (the DGYN motif) and a conserved small residue immediately (in this case a serine) upstream of the last catalytic aspartate (Asp104). The hydroxyl group of the Asn10 side chain hydrogen bonds with the backbone amino group of Ser103 to stabilise the relative positions of Asp7 and Asp104 (Figs 1B and EV2). Asn10 further stabilises the relative positions of these two key catalytic Asp residues via a network of water molecules (waters 2, 16 and 116 in the A‐subunit).

Putative RNA binding domain

Crystals of the wild‐type protein contain a dimer of Rae1 per asymmetric unit, with the dimer being formed through the intertwining of the C‐terminal α‐helices (Appendix Fig S1D). Since Rae1 is a monomer in solution, this dimer is not likely to be relevant in vivo. The C‐terminal 50 amino acids of Rae1 form two highly positively charged α‐helices (13 arginine or lysine residues) (Fig 1C), consistent with a role in RNA binding. Indeed, the C‐terminal α‐helical domain bears some overall charge and structural resemblance to a number of ribosomal proteins with positively charged helix‐loop‐helix motifs, most notably S13 (Fig 1D). In the monomeric conformation, the C‐terminal domain of Rae1 is likely to be highly flexible and free to move to interact with RNA.

Identification of potential targets of Rae1

To identify potential substrates of Rae1 in B. subtilis, we performed RNAseq analysis in triplicate on strains either lacking Rae1 (Δrae1) or complemented with a plasmid expressing a C‐terminal Flag‐tagged derivative of Rae1 (Rae1^f) under control of an IPTG‐dependent promoter. All annotated mRNAs, non‐coding RNAs and RNA segments (5′/3′ UTRs) (Nicolas et al, 2012) were included in the analysis. Forty‐six RNAs showed at least a 1.5‐fold increased expression (P < 0.05) in the Δrae1 strain compared to WT, and we consider these to be potential direct targets of Rae1‐mediated RNA degradation (Appendix Table S2). Thirteen RNAs had significantly lower levels than the WT strain and these likely represent indirect effects of the rae1 deletion or potential cases of RNA stabilisation following Rae1 cleavage. Interestingly, 17 of the 46 RNAs (37%) showing increased expression levels were members of the Fur (ferric uptake regulator) regulon, suggesting that the Δrae1 strain is subjected to a basal level of oxidative or iron (Fe³⁺) stress, despite the fact that it has no obvious growth phenotype. Eight of the up‐regulated RNAs are members of the AbrB and/or SigW regulons. AbrB encodes a repressor involved in the regulation of starvation‐induced processes (e.g. sporulation) in B. subtilis, while SigW is an extracellular function (ECF) sigma factor involved in the response to various types of envelope stress in B. subtilis. Globally, these results point to a potential role for Rae1 in stress management. Intriguingly, many of the candidate Rae1 targets (21 in total, both up‐ and down‐regulated) encode proteins destined for the cytoplasmic membrane, periplasm or the cell wall. The vast majority of these cell envelope proteins are non‐overlapping with the members of the Fur regulon.

We also compared the global RNA expression patterns in the plasmid‐complemented Δrae1 strain grown in the presence or absence of IPTG. Eleven RNAs showed decreased levels upon induction of Rae1 expression (Appendix Table S3). Seven of these showed the opposite effect in the rae1 deletion strain and we consider these to be among the best candidates for direct initiation of degradation by Rae1. Ten RNAs showed increased expression upon induction of Rae1 expression, but none of these were down‐regulated in the Δrae1 strain.

Validation of potential Rae1 targets by Northern blot

We divided the RNAs into three categories for validation of the RNAseq results by Northern blot: (I) those showing effects restricted to the comparison between the WT and Δrae1 strains; (II) those showing effects only upon induction of Rae1 expression; and (III) those showing opposite effects in the two sets of strains. A Northern blot probed for the category I ykuNOP operon (Fur regulon) indeed showed increased expression in the rae1 deletion strain compared to WT and confirmed the lack of effect of Rae1 overexpression (Appendix Fig S3A). No effect of ∆rae1 was seen on the fur mRNA itself (Appendix Fig S3B), suggesting that the effect of Rae1 on members of the Fur regulon occurs through changes in Fur protein levels or in Fur activity. Northern analysis also confirmed the increased expression of the category II sigH mRNA only in the condition of Rae1 overexpression (Appendix Fig S3C), further validating the RNAseq data.

Many of the seven candidates showing opposite effects in the Δrae1 and plasmid‐complemented strain (category III) were encoded in the same operons. bmrC and bmrD are part of one operon encoding two subunits of an ABC multidrug exporter, previously known as YheIH (Torres et al, 2009; Galian et al, 2011). We confirmed the increased expression of this operon in the absence of Rae1 and its decrease in the plasmid‐complemented strain by Northern blot (Appendix Fig S3D). The S1024, S1025, S1026 and yrzI (encoding a 49‐amino acid peptide of unknown function) RNAs also form an operon that includes yet another RNA segment S1027 identified in an extensive study of the B. subtilis transcriptome (Nicolas et al, 2012; Fig 2A). Ribosome profiling data suggest that a number of these short RNA segments are translated in rich medium, notably S1027, S1025 and to a lesser extent S1024, encoding potential peptides of 38, 17 or 52 amino acids, respectively (Li et al, 2012). For the rest of this study, we will focus on the effect of Rae1 on expression of the yrzI operon.

Figure 2. Role of Rae1 in yrzI mRNA turnover.

1. Structure of the yrzI operon. Putative promoters are represented by rightward arrows and putative transcription terminators by hairpin structures. Transcripts from this locus that include yrzI are shown as wavy lines.
2. Northern blot showing expression of the yrzI transcript in WT, ∆rae1 and plasmid‐complemented strains grown in the absence (−) and presence (+) of IPTG for 1 h. Number of biological replicates (n) = 3. The blot was probed with an oligonucleotide complementary to yrzI (CC1589) indicated by the black bar in panel (A). The correspondence between bands and transcripts shown in panel (A) is given to the right of the blot. R‐T4 refers to the 521‐nt species extending from 21 nt upstream of the yrzI coding sequence to the main operon terminator (T4). A Western blot showing the overexpression of Rae1 in the presence of IPTG is shown underneath the Northern blot; note that Rae1 is undetectable in WT cell extracts, even upon overexposure (n = 2).
3. Northern blot showing levels of yrzI mRNA at times after rifampicin addition (n = 2).
4. Northern blot showing that Rae1 catalytic activity is required for accumulation of yrzI mRNAs. Plasmid pDG‐Rae1^3fMut expresses the 3×Flag‐tagged (3f) catalytic mutant D7N D81N (n = 2).

Data information: Blots in panels (B–D) were rehybridised with a probe complementary to 16S rRNA (oligo CC058) as a loading control. Molecular weight markers (kb) are shown to the left of autoradiograms.Source data are available online for this figure.

The rae1 deletion increases the stability of the yrzI operon mRNAs

Deletion of the rae1 gene led to an accumulation of three yrzI‐containing transcripts of ~2.4, ~0.8 and ~0.5 kb in size (Fig 2B, lane 2). The two larger transcripts likely correspond to primary transcripts from the P1 and P2 promoters upstream of the yrhF and yrhG genes, respectively, while the 5′ end of the ~0.5‐kb transcript maps to 21 nt upstream of the yrzI coding sequence (Fig 3A, lane 2), with no obvious corresponding promoter. Induction of plasmid‐borne Rae1 expression reduced the level of these transcripts significantly (Fig 2B, lane 4), although not quite to WT levels, suggesting that the Flag tag may interfere slightly with Rae1 activity.

Figure 3. Mapping of Rae1 cleavage site in the yrzI mRNA.

1. Primer extension assay (oligo CC1671) on total RNA isolated from wild‐type and strains lacking Rae1 (∆rae1), RNase J1 (∆rnjA) and both (∆rae1 rnjA) (n = 2). Sequence lanes are labelled as their reverse complement to facilitate direct read‐out. The left and right panels represent different exposures of the same gel. Reverse transcriptase (RT) stops corresponding to the T2 and T3 terminator structures, the predicted P3 promoter and a series of 5′ ends mapping within the yrzI ORF in the ∆rae1 rnjA strain are indicated.
2. Mapping of the Rae1 cleavage site to the secondary structure of the yrzI‐S1024 region predicted by LocaRNA (Smith et al, 2010) using sequences shown in panel (C). Coordinates are given relative to the AUG start codon of S1024. The start and stop codon of the putative 17‐aa S1025 ORF are also boxed.
3. Conservation of S1025 nucleotide and amino acid sequence in Bacilli. Conserved nucleotides are marked with an asterisk. Conserved amino acids at the C‐terminus of the peptide are indicated. The SD sequence is underlined in red and the start codon highlighted in green. The Rae1 cleavage site in Bacillus subtilis is indicated with a vertical red arrow.

Source data are available online for this figure.

We confirmed that the effect of the rae1 deletion on yrzI transcript levels occurred at the level of mRNA stability. In the WT strain, two of the three yrzI transcripts were only barely detectable at the zero time point and all three disappeared rapidly after rifampicin addition (Fig 2C). All three transcripts were clearly stabilised in the Δrae1 strain but to varying degrees. The ~0.5‐kb RNA showed little if any degradation over the time course of the rifampicin experiment, showing that the degradation of this short yrzI transcript is strongly dependent on Rae1.

We performed an additional experiment to confirm that the levels of the yrzI transcripts observed in vivo were dependent on the catalytic activity of Rae1. The crystal structure of Rae1 allowed us to identify the key amino acids likely to be involved in Mg binding (D7, D81 and D104). As a tool to allow us to better co‐immunoprecipitate RNAs bound to Rae1 (to be reported elsewhere), we constructed a second Rae1 complementation strain expressing either a WT or catalytic mutant (D7N, D81N) of Rae1 bearing a 3×Flag tag (Rae1^3f). An equivalent mutation in one of these residues (D141N) was shown to functionally inactivate MCPIP1 (Matsushita et al, 2009). Induction of the expression of the 3×Flag derivative of WT Rae1 complemented the rae1 deletion strain to the same extent as the 1×Flag construct in terms of the yrzI expression profile seen in Northern blots (Fig 2D; compare lanes 4 and 5). However, the Rae1^3f D7N D81N mutant construct failed to complement upon induction of expression (Fig 2D; lane 6), confirming that Rae1 catalytic activity is required for degradation of yrzI mRNAs.

Mapping of the putative Rae1 cleavage site

Our data suggested that the primary pathway to initiate degradation of the ~0.5‐kb yrzI mRNA is through cleavage by Rae1. We reasoned that mapping of the Rae1 cleavage site would require stabilisation of the downstream cleavage product by inactivating the 5′–3′ exoribonuclease RNase J1. We therefore performed primer extension assays on total RNA isolated from WT, Δrae1, ΔrnjA and Δrae1 rnjA double mutant strains, using an oligonucleotide that hybridised to the coding sequence of S1024. In the WT strain, we observed a weak band corresponding to a 5′ end within the putative S1025 ORF, in the loop of a predicted stem‐loop structure in untranslated forms of the RNA (Fig 3A and B). In the ΔrnjA strain, this 5′ end was significantly more abundant, confirming that RNase J1 indeed degrades the downstream product of Rae1 cleavage. In the double Δrae1 rnjA mutant, this 5′ end was absent as expected and instead a number of new 5′ ends mapping to within the upstream yrzI ORF were visible. These data suggest that another (unknown) enzyme can give RNase J1 access to the yrzI mRNA in the absence of cleavage by Rae1. However, the strong stabilisation of the ~0.5‐kb yrzI‐S1024 seen in the Δrae1 strain indicates that this alternative pathway is not very efficient.

Addition of a Rae1 cleavage site destabilises a heterologous mRNA

The Rae1 cleavage site mapped to the yrzI‐S1024 intergenic region, within the putative 17‐aa ORF encoded by S1025 (Fig 3C). We asked whether insertion of this 156‐nt sequence into the 3′UTR of the highly stable hbs mRNA could destabilise it in a Rae1‐dependent manner. The hbs gene encodes an orthologue of the E. coli HU protein and is essential in B. subtilis. We therefore used an ectopic chromosomally expressed deletion derivative of this mRNA called hbs∆ that we have previously characterised extensively (Fig 4A) (Daou‐Chabo & Condon, 2009; Daou‐Chabo et al, 2009). Like the native hbs mRNA, the hbs∆ construct yields three transcripts corresponding to expression from promoters P3 (P3‐ter) and P1 (P1‐ter) and a highly stable ribosome‐protected species (R‐ter). Using a probe common to all expected hbs transcripts, we compared the stabilities of the three hbs∆ transcripts to those containing the S1025 fragment in their 3′UTR (hbs∆‐S1025; Fig 4A) in both WT and ∆rae1 strains. The rae1 deletion had no effect on the stability of either the native hbs mRNA or the control construct hbs∆ (Fig 4B). In WT cells, the P1 and P3 transcripts of the hbs∆‐S1025 construct were significantly less stable than their native hbs or hbs∆ counterparts. But more remarkably, only traces of the ribosome‐protected species (R‐ter) were detected for hbs∆‐S1025 (expected size 312 nt), suggesting it has been almost fully destabilised by insertion of the Rae1 cleavage site. In the ∆rae1 strain, the P1 and P3 transcripts of hbs∆‐S1025 were significantly stabilised compared to WT and the R‐ter species reappeared and was fully stable. Thus, insertion of the S1025 fragment in hbs∆ 3′UTR destabilises this mRNA in a Rae1‐dependent fashion, confirming that this fragment contains a bona fide and transposable Rae1 cleavage site.

Figure 4. Insertion of the Rae1 cleavage site in the 3′UTR of the hbs mRNA results in its instability.

1. Schematic of various hbs transcripts in the strains tested. The top section shows the structure of the native hbs gene and its three transcripts (wavy lines). The 380‐nt species is depicted with a ribosome at its 5′ end. The middle and lower sections show the equivalent transcripts and their sizes from the hbs∆ and hbs∆‐S1025 constructs, respectively. The insertion sequence is shown in red and Rae1 depicted with a scissors symbol.
2. Northern blot analysis of WT and ∆rae1 strains harbouring either the hbs∆ or hbs∆‐S1025 construct in the amyE locus at times after rifampicin addition (n = 2). The blot was probed with an oligo complementary to the early hbs ORF (CC463; black bar in panel A). The origin and sizes of the different transcripts are shown to the left and right of the autoradiogram. Transcripts from hbs, hbs∆ or hbs∆‐S1025 are labelled in black, blue and red, respectively. The blot was rehybridised with a probe complementary to RNase P RNA (rnpB; oligo CC1006) as a loading control.

Source data are available online for this figure.

Rae1‐dependent destabilisation of the hbs∆‐S1025 mRNA depends on S1025 translation and reading frame

The C‐terminal half of the S1025 reading frame is highly conserved at both the amino acid and nucleotide sequence levels (Fig 3C). This observation, coupled with an inability to cleave the S1025 transcript with purified Rae1 in vitro (see below), led us to ask whether Rae1 activity might be translation‐dependent in vivo. To test this, we made two derivatives of the hbs∆‐S1025 construct in which we mutated the SD sequence (∆SD) of the putative S1025 ORF or changed its reading frame by +1 (F + 1). Since the F + 1 mutation caused a premature stop (UAG) at the third triplet of this new reading frame, we mutated it to Gln (CAG) so that translation could continue over the Rae1 cleavage site on the mRNA. The F + 1, stop3 → Q peptide is 20 amino acids long and a completely different sequence to the wild‐type S1025 peptide (Fig 5A). However, the nucleotide sequence and potential secondary structure encompassing the Rae1 cleavage site (Fig 3B), located about 30 nt downstream of these mutations, are unchanged. Remarkably, both the ∆SD and F + 1, stop3→Q mutations resulted in a complete loss of the destabilising effect of the S1025 sequence on the hbs∆‐S1025 mRNA and the stability of these mRNAs became totally Rae1‐independent (Fig 5B). Thus, Rae1 cleavage within the S1025 ORF is dependent not only on translation of S1025, but also on the peptide reading frame.

Figure 5. Effect of translation on stability of the hbs∆‐S1025 transcript.

1. Nucleotide and amino acid sequence of WT S1025, and the ∆SD and frameshift (F + 1, stop3→Q) derivatives in the context of the hbs∆‐S1025 mRNA. Mutations are shown in red; the cleavage site in green.
2. Northern blot of WT and ∆rae1 strains harbouring the ∆SD or F + 1, stop3→Q derivatives of the hbs∆‐S1025 construct (top) at times after rifampicin addition (n = 2). The blot was rehybridised with a probe complementary to RNase P RNA (rnpB; oligo CC1006) as a loading control.

Source data are available online for this figure.

Rae1 binds to the ribosome

The translation and reading frame dependence of Rae1 cleavage of the S1025 mRNA suggested that Rae1 is likely to bind the ribosome. As endogenous levels of Rae1 in B. subtilis are too low to detect (Fig 2B), we added Rae1 purified from E. coli to a mixture of 30S, 50S and 70S ribosomes isolated from B. subtilis on sucrose gradients. The mixture was then reseparated on a fresh sucrose gradient and fractions assayed for ribosomal RNA (16S and 23S) on agarose gels and probed for YacP by Western blot (Fig 6). Rae1 was detected primarily in fractions near the top of the gradient and in fractions containing the 50S and 70S ribosomes (Fig 6C). The peak 30S fraction (fraction 5) contained the least amount of Rae1. These experiments suggest that the tightest association under the conditions tested (i.e. without tRNA or mRNA) is between Rae1 and the 50S ribosomal subunit. This observation prompted the renaming of the enzyme to ribosome‐associated endoribonuclease 1.

Figure 6. Rae1 associates with Bacillus subtilis ribosomes.

1. Sucrose gradient separation of 30S, 50S and 70S ribosomal particles from B. subtilis incubated with Rae1.
2. Agarose gel analysis of rRNAs present in sucrose gradient fractions. 23S and 16S rRNAs are indicated.
3. Western blot analysis of sucrose gradient fractions using anti‐Rae1 antibody (n = 2). Lane C contains 0.05 μg of purified Rae1. The migration positions of molecular weight markers are shown to the left of the blot.
4. Control showing migration of Rae1 in sucrose gradient without ribosomes.

Source data are available online for this figure.

Rae1 cleavage occurs precisely between a conserved glutamate (GAG) and lysine (AAG) codon in positions 13 and 14 of the 17‐aa S1025 coding sequence (Fig 3C). The fact that Rae1 can read codons in frame makes Rae1 a good candidate for an A‐site RNase. To show this was possible, we manually docked the crystal structure of Rae1 to the A‐site of the B. subtilis ribosome (pdb code 3j9w), positioning the catalytic site appropriately for cleavage of the scissile bond between the A‐site and P‐site codons (for details, see Materials and Methods). In this configuration, Rae1 fits snugly into the contours of the A‐site formed primarily by the 16S rRNA and the P‐site tRNA (Fig 7 and Movie EV1). The only significant clash with 16S rRNA (helix 44) is with residues in the loop between strands β2 and β3 of Rae1. Residues in this loop have high B‐factors (Fig EV3A), indicative of structural flexibility, and that some accommodation of this region may be possible for docking of Rae1. Furthermore, helix 44 of 16S rRNA is known to move during the translocation step (Frank & Agrawal, 2000); thus, both partners have flexibility in the region of the clash. In this conformation, helices 38 and 69 of 23S rRNA lie close enough to Rae1 to potentially account for the interaction with the 50S ribosomal subunit seen in the sucrose gradient experiment (Fig EV3B).

Figure 7. Docking model of Rae1 in the A‐site of the ribosome.

1. Ribbon view of Rae1 modelled with Mg²⁺ and scissile phosphate docked on Bacillus subtilis 30S ribosomal subunit. The P‐site tRNA is shown in blue and mRNA in green. The scissile phosphate is in red. A‐, P‐ and E‐site triplets are indicated.
2. Surface model of Rae1 in A‐site showing neighbouring 16S rRNA helices: h18 (524–544), h25 (791–809), h28 (928–947; 1,383–1,409), h29 (948–953; 1,346–1,358), h30 (954–967; 1,234–1,244), h31 (968–990), h32 (991–1,000; 1,222–1,233), h34 (1,057–1,221), h44 (1,410–1,513), h45 (1,514–1,534) (B. subtilis numbering). The rest of the 30S subunit is shown as a partially transparent surface map. Each panel shows a ˜90° rotation compared to the previous panel. Ribosomal proteins S12, S13 and S9 are indicated in appropriate views. A rotation of the docking model is shown in Movie EV1.

Figure EV3. Regions of Rae1 showing highest levels of flexibility.

1. Putty ribbon representation of Rae1 docked in A‐site of Bacillus subtilis 30S ribosomal subunit, shown in the same orientation as the first panel in Fig 7B. B‐factors are plotted on the backbone structure of Rae1 as a gradient from dark blue (low) to red (high) and increasing thickness of the putty ribbon. The 30S ribosomal subunit is shown in grey and key 16S rRNA helices are labelled. The P‐site tRNA is in blue.
2. Putty ribbon representation of Rae1 docked in A‐site of B. subtilis 30S + 50S ribosomal subunits. Colouring is as in panel (A) with the additional 50S ribosomal subunit shown in wheat.

Rae1 recognises the lysine 14 codon in S1025

If our docking model were correct, we would predict that the Glu13 and Lys14 codons of the S1025 mRNA are in the P‐site and A‐site positions, respectively, and that Rae1 should primarily recognise the AAG lysine codon. To test this idea and identify the codon specificity of Rae1 cleavage, we made a number of mutations around the cleavage site in the S1025 ORF in the chimeric hbs∆‐S1025 mRNA, first changing the Glu13 and Lys14 codons to either synonymous or non‐synonymous variants. Since the ribosome‐protected species (R‐ter) is essentially fully stable in the absence of Rae1 and almost completely destabilised in its presence, the accumulation of this species is a highly sensitive measure of Rae1 inhibition levels. Changing the Glu13 and Lys14 codons, either singly or together, to their synonymous codons (GAA or AAA, respectively) had no major effect on the stability of the hbs∆‐S1025 transcripts compared to the WT construct (Fig 8A). Thus, synonymous Glu13 and Lys14 codons function well for Rae1 cleavage.

Figure 8. Activity of Rae1 on S1025‐containing mRNAs with various codons in positions 13 and 14.

1. Northern blot showing the effect of synonymous mutations in codons Glu13 and Lys14 on the stability of the chimeric hbs∆‐S1025 construct (n = 2). Total RNAs were probed with oligo CC463 at times after rifampicin addition. Transcripts from the native hbs gene and the hbs∆‐S1025 construct are labelled in black and red, respectively.
2. Northern blot showing the effect of non‐synonymous mutations in codons Glu13 and Lys14 on hbs∆‐S1025 mRNA stability (n = 2). For quantification, the R‐ter band from the hbs∆‐S1025 construct was normalised to the equivalent species from the native hbs gene and this ratio was arbitrarily set to 1 for the T0 time point in the strain carrying the Lys14→Gln mutant. The histogram shows the quantification of two independent experiments (light and dark grey) with the average indicated.

Source data are available online for this figure.

Upon changing Glu13 to Gln (CAG), a low level of the R‐ter species accumulated, suggesting a contribution of this codon to Rae1 cleavage efficiency (Fig 8B). Changing the Lys14 codon to Gln (CAG) had a significantly greater effect on the accumulation and stability of the three hbs∆‐S1025 transcripts, and changing both Glu13 and Lys14 to Gln had an additive effect. These experiments suggest that Rae1 primarily recognises the Lys14 codon in S1025 (either AAG or AAA), with a contribution from the preceding Glu13 codon. We will discuss possible ways that Glu13 might influence Rae1 cleavage efficiency that are consistent with the A‐site binding model in the discussion section.

The middle A‐residue of codon 14 is important but not sufficient for optimal Rae1 activity

The results obtained with the different synonymous (AAA) and non‐synonymous (CAG) mutations at codon 14 left open the possibility that Rae1 recognised primarily either the first or second A‐residue (or both) of the lysine codon in this position and could therefore potentially cut at other codons. To further probe the specificity of Rae1, we tested the effect of mutating the Lys14 codon to asparagine (AAU and AAC) or threonine (ACG). A clear hierarchy emerged (Fig 8B). Replacing Lys14 (AAG) by either Asn codon caused a significant accumulation of the R‐ter species (on a par with the change to Gln), while mutation to ACG (Thr) caused R‐ter to accumulate to higher levels still. The accumulation levels of R‐ter in the Lys14→Thr mutant were the same in the presence or absence of Rae1, whereas they were higher in the ∆rae1 strain for both the Gln and Asn mutants (Appendix Fig S4). This result indicates that Rae1 can weakly recognise Gln (CAG) and Asn codons (AAC or AAU), but cannot cleave the Thr codon (ACG). Globally, our data suggest that the middle A‐residue of codon 14 is most important but insufficient for optimal recognition by Rae1; i.e. both the first and third positions also contribute towards cleavage efficiency. The overall hierarchy observed for codon 14 was Lys > Asn = Gln > Thr. The resolution of the docking model does not allow us to say with any certainty which residues of Rae1 might account for this discrimination.

Rae1 cleaves the S1025 mRNA in a translation‐dependent manner in vitro

The experiments presented thus far strongly support the idea that Rae1 is a translation‐dependent endoribonuclease in vivo. To prove that the effects observed were direct, we sought to recapitulate Rae1 cleavage of the S1025 mRNA in a purified in vitro translation system supplemented with 70S ribosomes isolated from B. subtilis. A 157‐nt mRNA corresponding to the S1025 ORF and flanking 5′ and 3′ UTRs was transcribed in vitro and 5′‐labelled. Incubation of Rae1 alone with this in vitro‐transcribed RNA failed to produce any cleavage products (Fig 9, lane 2). However, addition of Rae1 to in vitro translation reactions primed with this mRNA resulted in a cleavage product of the expected size, 73 nt (Fig 9, lanes 4 and 5). Addition of the translation inhibitor tetracycline to these reactions showed a strong inhibition of cleavage (Fig 9, lanes 6 and 7), confirming that translation is necessary for Rae1 activity. Note that although tetracycline binds the A‐site and may compete with Rae1 binding, it is likely to have blocked translation before the ribosome gets as far as the Rae1 cleavage site in these experiments. Lastly, no cleavage was observed with the Rae1 D7N/D81N catalytic mutant (Fig 9, lanes 8 and 9). These experiments show conclusively that Rae1 is translation‐dependent for its cleavage activity and that this activity is directly associated with the catalytic site of the Rae1 NYN domain, rather than an induced cleavage by the ribosome as has been proposed for the RelE family of A‐site toxin RNases (Neubauer et al, 2009).

Figure 9. Rae1 cleaves the S1025 mRNA in a purified in vitro translation system.

5′‐labelled S1025 mRNA was translated using a purified in vitro translation system supplemented with Bacillus subtilis 70S ribosomes (n = 4). Rae1 was added at sixfold molar excess over ribosomes (++) or diluted 10‐fold (+). The full‐length (FL) RNA is 157 nt and the expected cleavage product is 73 nt (indicated with a scissors symbol). Addition of ribosomes causes a visible gel shift of the FL species even under denaturing gel conditions. The migration positions of a DNA size standard (bp) are given to the right of the autoradiogram. Source data are available online for this figure.

Discussion

In this paper, we have identified potential mRNA substrates, the molecular structure and the modus operandi of the orphan enzyme Rae1 (YacP), a founder member of the NYN subfamily of PIN domain ribonucleases found in all three domains of life. The crystal structure of Rae1 confirmed what was already suspected from sequence alignments, i.e. that the NYN subfamily differs from other PIN family members by the absence of a lid over the catalytic site (Anantharaman & Aravind, 2006). Indeed, the catalytic site of Rae1 is completely exposed to the solvent at the surface of the protein, an unusual scenario for an RNase and suggestive that a co‐factor might be necessary to create a pocket for catalysis. Our data would suggest that this pocket is generated when Rae1 binds to the ribosome. However, all NYN proteins are clearly not ribosome‐dependent for cleavage, since both MCPIP1 and PRORP have non‐coding RNA substrates.

Our crystal structure of Rae1 lacked the catalytic metal ion necessary for NYN/PIN domain activity, suggesting that it is provided by the substrate bound to the ribosome. In the different structures of ribosomes bound to mRNA, there are plenty of Mg²⁺ ions in the immediate vicinity that could be “borrowed” for Rae1 activity, most notably a Mg²⁺ ion that is always found at the kink between the P‐site and A‐site codons in the mRNA, for example (Demeshkina et al, 2012). The ability of Rae1 to use this, or any other locally available Mg²⁺ ion, would depend on the relative affinities of the Rae1 coordination site and the binding site on the ribosome. Our data do not rule out the possibility of a second metal ion in the Rae1 active site, similar to the catalytic sites of certain other PIN domain proteins (Hosfield et al, 1998; Miallau et al, 2009). Indeed, a second Mg²⁺ ion would fit our model better as it would be predicted to lie closer to the scissile bond in the ribosome A‐site. In the structure of wild‐type Rae1, a tryptophan residue (Trp109) projects into the active site between residues Asp81 and Asp104 (Appendix Fig S5) that may be inhibitory for enzymatic activity. Interestingly, in the W164L mutant (which lies close to the C‐terminus of the protein), W109 is flipped out of the active site. Neither the wild‐type nor the W164L crystal structure contains Mg²⁺, so the position of W109 alone is not sufficient to explain the absence of Mg²⁺ in the WT structure. Nonetheless, the relative positions of this Trp residue in these two structures may suggest a possible activation mechanism for Rae1 upon ribosome binding that will be the subject of future investigation.

The best candidate substrates for Rae1 cleavage showed opposite effects in the ∆rae1 strain and the complemented strain (category III). However, a number of RNAs (categories I and II) only showed effects in one or the other strain. We confirmed these effects for ykuN and sigH by Northern blot (Appendix Fig S3). For category I mRNAs, which only showed effects in the ∆rae1 strain compared to WT, it is possible that a small amount of leakiness from the Pspac promoter on the complementation plasmid provides sufficient Rae1 for full cleavage of these mRNAs, i.e. that these are relatively high‐affinity substrates. In this case, providing more Rae1 upon addition of IPTG would make little difference to cleavage efficiency. Most category II mRNAs, exemplified by sigH, showed up‐effects upon addition of IPTG to the plasmid‐complemented strain. It is possible that these mRNAs represent very low‐affinity or off‐target RNAs that are subjected to stabilising cleavages by Rae1 only when expressed from a plasmid.

Our data converge on the idea that Rae1 cleaves mRNAs from within the A‐site of the ribosome. Nonetheless, we considered two alternatives to the A‐site where Rae1 could potentially read codons in frame, the E‐site or the entrance to the tunnel through which the translated mRNA feeds into the 30S subunit. We could not dock Rae1 in the E‐site without major steric clashes and an interaction with the mRNA entrance tunnel would be too far away to account for the association with the 50S subunit we observed in sucrose gradients. Thus, while we cannot formally rule out other possible interaction sites, Rae1 binding to the A‐site best fits our current data.

In our docking model, Rae1 is positioned to read the identity of the A‐site codon and has no obvious contacts with the P‐site codon, normally occupied by the peptidyl‐tRNA. Yet the Glu to Gln mutation in position 13 of the S1025 nascent peptide (which would constitute the P‐site codon in our model) has a measurable effect on Rae1 cleavage efficiency, in particular in concert with a similar mutation in position 14 (Lys to Gln). At least two possibilities can be envisaged to explain how the P‐site codon might affect Rae1 cleavage without the enzyme making direct contact with this triplet. In the first, the amino acid sequence of the S1025 nascent peptide may cause the ribosome to pause with the Glu13 in the P‐site and Lys14 in the A‐site. In this scenario, Glu13 would be more important for ribosome pausing than for Rae1 cleavage per se, and it would be the alteration of pausing frequency or pausing time that leads to the effect on cleavage efficiency by giving Rae1 a greater opportunity to enter the A‐site. This would explain why only a few substrates were identified under the rich medium growth conditions tested, rather than every mRNA with consecutive Glu‐Lys codons. In the second scenario, it is possible that Rae1 makes specific contacts with the Glu‐tRNA occupying the P‐site and that these contacts are important for Rae1 recruitment or activity. In this case, both Glu and Lys codons would be key for full cleavage efficiency. Further experiments are required to distinguish between these possibilities.

A comparison of the S1025 ORF from a number of Bacilli shows that the amino acid sequence around the S1025 cleavage site is a highly conserved YT(M)EKDQV motif, conserved at even the nucleotide level (Fig 3C). This motif is not found in any of other potential Rae1 substrates or indeed in any other B. subtilis protein (even with two allowed mismatches). Other less‐stringent motifs must therefore also make acceptable Rae1 cleavage sites. No obvious sequence homology or enriched sequence motifs were observed with the different potential Rae1 substrates listed in Appendix Tables S2 and S3.

Rae1 is conserved in Firmicutes to a higher degree than the yrzI operon expressing the S1025 peptide, which is essentially confined to the Bacillales. This suggests that a broader function of Rae1 under more specific growth conditions remains to be identified. One possibility we considered is that Rae1 may gain access to the ribosome A‐site during conditions of starvation for lysine. Intriguingly, lysine codons are highly overrepresented in the second codon position of B. subtilis ORFs, with ~20% of all ORFs having either AAG or AAA immediately downstream of one of the three start codons AUG, GUG or UUG (Rocha et al, 1999; Fig EV4A). We tested three mRNAs (rnjA, rnpA, rnmV) with a lysine codon in position 2 of their coding sequences for Rae1‐dependent effects under conditions of lysine starvation in an auxotrophic (lysA) mutant. While we saw some evidence for an accumulation of degradation intermediates under starvation conditions, presumably the result of cleavage events, this accumulation was not Rae1‐dependent and, in the cases of rnjA and rnpA, clearly did not correspond to cleavage early in the respective open‐reading frames (Fig EV4B). Furthermore, the addition of lysine hydroxamate, known to block cognate tRNA charging, to the in vitro translation system did not effect Rae1 cleavage efficiency of the S1025 mRNA (Fig EV4C). These experiments appear to rule out lysine starvation as the general trigger for Rae1 cleavage in B. subtilis.

Figure EV4. Lysine starvation does not appear to be a general trigger for Rae1‐dependent mRNA cleavage in Bacillus subtilis .

1. Per cent of B. subtilis open‐reading frames with each of the 20 amino acids immediately downstream of AUG, GUG or UUG start codons.
2. Northern blots of total RNA isolated from a B. subtilis ∆lysA strain and a B. subtilis ∆lysA ∆yacP strain probed with oligos specific for the rnjA (CC1978), rnpA (CC1005) and rnmV (CC1979) mRNAs (n = 1). Cells were grown in minimal medium in the presence of all 20 amino acids to OD₆₀₀ = 0.3, then centrifuged and resuspended in the same medium with or without lysine for 30 min. Starvation‐specific degradation intermediates are indicated by red arrows.
3. Rae1 cleavage of the S1025 mRNA in a purified in vitro translation system in the presence of lysine (KHX) or serine (SHX) hydroxamate (n = 2). 1 pmol of 5′‐labelled S1025 mRNA was translated using a purified in vitro translation system supplemented with 6.5 pmol B. subtilis 70S ribosomes (see Materials and Methods). Rae1 was added at 38 pmol per reaction. The full‐length (FL) RNA is 157 nt and the expected cleavage product is 73 nt (indicated with a scissors symbol). Lane M is a labelled pBR322/MspI DNA size standard with sizes in base pairs (bp) given to the left of the autoradiogram. KHX and SHX were added at a final concentration of 0.75 mM.

Source data are available online for this figure.

Most known RNases have catalytic activity in isolation, with protein co‐factors acting as guides or providing specificity. Rae1, on the other hand, is completely inactive without the ribosome (Fig 9) and, in this regard, is similar to the RelE family of A‐site toxins. However, while RelE family members induce RNA cleavage by the ribosome (Neubauer et al, 2009), our structural and mutational data show that Rae1 has a bona fide catalytic site and thus represents a new family of ribosome‐dependent RNases. Curiously, some members of the RelE family (HigB and YafQ) also appear to have a preference for AAA triplets. YafQ shows a preference for in‐frame AAA lysine codons, while for HigB reading frame it was less important (Hurley & Woychik, 2009; Prysak et al, 2009). This is a remarkable example of convergent evolution between structurally unrelated RNases from two organisms (E. coli and B. subtilis) that separated more than a billion years ago. In addition to the ribosome‐associated activity and preference for A‐rich triplets shared with the RelE family of toxin RNases, Rae1 is structurally related to the well‐characterised VapC type II toxin. VapC consists of a solitary PIN domain that cleaves tRNAs or 23S rRNA to stop cell growth (Winther et al, 2016). These similarities suggest an intriguing functional overlap between Rae1 and classical toxin–antitoxin (TA) systems. TA systems are involved in mediating bacterial stress responses, with the toxin and antitoxin typically encoded in the same operon. The antitoxin is usually significantly less stable than the toxin, and stress conditions that inhibit expression of the operon rapidly lead to a toxin:antitoxin imbalance, activation of the toxin and an arrest of cell growth (Aakre et al, 2013; Goeders & Van Melderen, 2014). Although there is no obvious antitoxin encoded downstream of the rae1 gene, we have been unable to overproduce large quantities of the enzyme in B. subtilis, suggesting that the protein is unstable. In this regard, the protease(s) responsible for Rae1 degradation could play an analogous role to classical antitoxins.

The rae1 gene is found primarily in Firmicutes, Cyanobacteria and higher plants (Appendix Fig S6), where it is predicted to be localised to chloroplasts. The synteny of the rae1 locus is not conserved between the two bacterial phyla, however. In the Firmicutes, rae1 is found to be surrounded by genes involved in the translation and transcription machineries. In 268 different Firmicutes, it was found to be associated with the rlmB gene 98% of the time (Fig EV1A). RlmB methylates residue G2251 close to the peptidyl transfer centre on 23S rRNA (Lovgren & Wikstrom, 2001). The rae1 gene in Firmicutes is also highly associated with the mrnC gene (91% co‐occurrence), encoding the 23S rRNA maturase Mini‐RNase III and the cysS gene (90% co‐occurrence), encoding cysteinyl tRNA synthetase. This gene organisation with components of the translation machinery is fully coherent with Rae1's role as a translation‐dependent ribonuclease in B. subtilis.

In Cyanobacteria, the rae1 gene is most highly associated with a gene encoding an ABC transporter, a gene encoding a small protein of unknown function and psbDII, encoding photosystem II reaction centre D2 protein (57%, 57% and 49% co‐occurrence, respectively, in 35 different cyanobacterial species; Fig EV1B). Orthologs found in algae and higher plants have predicted chloroplast transit peptides, suggesting that Rae1 is likely to have a conserved function in photosynthesis from the Cyanobacteria to the chloroplasts of plants and algae. The conservation pattern of this gene may reflect a very ancient transfer between B. subtilis and photosynthetic organisms in their shared niche in the soil. Interestingly, the mrnC gene, located in the same operon as rae1 in B. subtilis, has a similar phylogenetic distribution despite the fact that the synteny has been lost outside of the Firmicutes (Redko et al, 2008).

Materials and Methods

Oligonucleotides and bacterial strains used in this study are given in Appendix Tables S4 and S5, respectively. Strain construction is described in Appendix Supplementary Methods.

Protein production and purification

Rae1 expression was induced with 1 mM IPTG for 4 h and purified on NiNTA resin according to previously published protocols (Condon et al, 2008). Peak fractions were dialysed against 20 mM Tris pH 8.0, 150 mM NaCl, 10% glycerol, concentrated using Amicon filters with 3‐kDa cut‐off and applied to a Superdex 26/60 sizing column equilibrated in the same buffer. Peak fractions were reconcentrated for crystallography trials. Crystallography conditions and data collection are described in Appendix Supplementary Methods.

RNA sequencing

Cultures were grown in 2×YT in the presence of appropriate antibiotics (mls or kan) with shaking. Strains used were SSB1002, CCB375 and CCB604. The CCB604 culture was grown to OD600 = 0.15, split in two and 1 mM IPTG added to one flask for 1 h before harvesting. Experiments were performed in triplicate. Total RNA was isolated from 20 ml culture (pelleted and frozen) by the glass beads/phenol method described previously (Bechhofer et al, 2008). RNA samples were treated with RQ DNase Promega (37°C for 20 min) to remove potential contaminating chromosomal DNA. Ribosomal RNA was removed from 5 μg total RNA using RiboZero kit from Epicenter, according to the manufacturer's instructions. Ribosomal RNA depletion and overall RNA quality was analysed by Bioanalyser (Agilent). cDNA libraries were prepared using the Smarter Stranded RNA‐Seq Kit (Clontech) with adapters for multiplexing, according to the manufacturer's instructions. RNA concentration and quality were checked by Bioanalyser (Agilent). The 12 samples were normalised to 2 nM, multiplexed and denatured at a concentration of 1 nM using 0.1 N NaOH (5 min at room temperature) before dilution to 10 pM and loading on a HiSeq2000 flow cell. Reads were cleaned of adapter sequences and low‐quality sequences using an in‐house program (https://github.com/baj12/clean_ngs). Quality control checks were performed before and after trimming by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Fifty‐nt reads were mapped by Bowtie (Langmead et al, 2009) and yielded about 10⁷ aligned reads per sample. Details on data analysis are provided in Appendix Supplementary Methods.

Ribosome association of Rae1

Bacillus subtilis 30S, 50S and 70S ribosomal particles were isolated from 200 ml of log phase B. subtilis cells at OD₆₀₀ = 0.5. Cells were centrifuged and resuspended in 2 ml ice‐cold Buffer A (10 mM Tris–HCl pH 7.5, 100 mM NH₄Cl, 10 mM MgCl₂, 6 mM β‐mercaptoethanol) plus 10 μg/ml DNase I and lysed by two passages in a French Press (20,000 psi). The lysate was cleared at 16,100 g for 30 min at 4°C in a bench‐top centrifuge. 1 ml of lysate was loaded on a 10–30% sucrose gradient in Buffer A and centrifuged at 45,300 g for 14 h at 4°C in an SW28 rotor (Beckmann). 1 ml fractions were collected using a Piston Gradient Fractionator (Biocomp). Peak fractions of each ribosome particle were pooled and RNA concentrations measured.

Five hundred picomole Rae1 was incubated with a mixture containing 100 pmol each 30S, 50S and 70S ribosome particles in Buffer A for 10 min on ice, 10 min at 37°C and then 10 further min on ice. The mixture was loaded on a 10–30% sucrose gradient in Buffer A and centrifuged for 16 h at 42,700 g at 4°C in an SW41 (Beckmann) rotor. Fractions were separated as before, and 30 ul of each fraction was assayed by Western blot (ECL; GE) using Rae1‐specific rabbit antibodies.

Manual docking of Rae1 to the ribosome A‐site

A Mg²⁺ ion was first placed in the catalytic site of the Rae1 crystal structure by superposition with the known structure of the Rae1‐homologue MCPIP1 (pdb code 3V34) in PyMol (Fig 2). Rae1 + Mg was then superposed on the PIN domain of T4 RNase H bound to Mg²⁺ (1tfr) and DNA (2ihn) to correctly orient the catalytic site and Mg²⁺ ion relative to the phosphate group of the scissile phosphodiester bond. The Rae1 + Mg + phosphate model was then used for docking. A number of ribosome structures with A‐site‐bound proteins (RelE, HigB, YoeB, EFTu, RF1, RRF, IF2, TetM, RelA, EFG) were tested to find the best fit, using the scissile phosphate between the P‐site and A‐site codons as an anchor. The RF1‐bound structure (5j4d) and the B. subtilis 30S structure with an empty A‐site (3j9w) yielded the fewest steric clashes with 16S rRNA.

Northern blots and primer extension assays

Northern blots were performed on total RNAs isolated either by the glass beads/phenol method described in Bechhofer et al (2008) or by the RNAsnap method described in Stead et al (2012). Northern blots were performed as described previously (Durand et al, 2012). Primer extension assays were performed on glass bead/phenol‐extracted RNAs as described previously (Britton et al, 2007).

In vitro assays of Rae1 activity

70S ribosomes were purified from 1 g of frozen B. subtilis W168 cells as follows. The cell pellet was resuspended in 3 ml Buffer A (10 mM Tris–HCl pH 7.5, 100 mM NH₄Cl, 10 mM MgCl₂, 6 mM β‐mercaptoethanol) plus DNase I (10 μg/ml) and lysed by two passages in a French Press. Cell debris was removed by centrifugation for 30 min at 16,100 g in a bench‐top centrifuge (Eppendorf) at 4°C. Ribosomes were pelleted at 271,700 g for 40 min in a bench‐top ultracentrifuge (Beckmann 100.3 TLA rotor) at 4°C. Ribosomes were resuspended in 0.5 ml Buffer A and layered on a 2.5 ml sucrose 35% cushion. Ribosomes were pelleted a second time at 80,000 rpm for 3.5 h in a bench‐top ultracentrifuge (100.3 TLA rotor Beckmann) at 4°C and resuspended in 0.4 ml Buffer A + 10% glycerol for use in in vitro translation assays.

S1025 mRNA was transcribed by T7 RNA polymerase in vitro using a MegaShortScript kit (Ambion) and a PCR template amplified from B. subtilis chromosomal DNA using oligo pair CC1660/1661. An eightfold excess of guanosine was used relative to GTP to ensure production of an in vitro transcript bearing a 5′OH group. The RNA was purified from unincorporated nucleotides using G50 columns (GE), quality‐controlled by agarose gel analysis and quantified by NanoDrop. 40 pmol of RNA was labelled using γ‐³²P‐ATP and polynucleotide kinase (Biolabs).

In vitro translation assays (PURExpress ∆ ribosomes; New England Biolabs) were typically performed in 5 μl reactions with 1 pmol 5′‐labelled RNA and 6.5 pmol B. subtilis 70S ribosomes. Rae1 was added at 38 or 3.8 pmol per reaction and freshly diluted tetracycline was used at 3 μM final concentration. Reactions were incubated at 37°C for 30 min, stopped with 5 μl RNA loading dye (Biolabs) and run on 5% sequencing gels.

Data availability

The coordinates and structure factors have been deposited in the Brookhaven Protein Data Bank (accession numbers 5MQ8 Rae1 WT and 5MQ9 Rae1 W164L mutant). RNAseq data has been deposited at GEO (access code: GSE93894).

Author contributions

ML, JP, LG, OP, and SF all contributed to the experimental work. The RNAseq experiments and data analysis were performed by CP and JYC. ML, JP and CC designed the experiments and analysed the data. CC wrote the paper.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Movie EV1

Source Data for Expanded View and Appendix

Review Process File

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 8

Source Data for Figure 9

The EMBO Journal (2017) 36: 1167–1181

See also: D Lalaouna & E Massé (May 2017)