Enzymatic activity necessary to restore the lethality due to Escherichia coli RNase E deficiency is distributed among bacteria lacking RNase E homologues

Escherichia coli RNase E (Eco-RNase E), encoded by rne (Eco-rne), is considered the global RNA decay initiator. Although Eco-RNase E is an essential gene product in E. coli, some bacterial species, such as Bacillus subtilis, do not possess Eco-RNase E sequence homologues. B. subtilis instead possesses RNase J1/J2 (Bsu-RNase J1/J2) and RNase Y (Bsu-RNase Y) to execute RNA decay. Here we found that E. coli lacking the Eco-rne gene (Δrne E. coli) was viable conditional on M9 minimal media by introducing Bsu-RNase J1/J2 or Bsu-RNase Y. We also cloned an extremely short Eco-RNase E homologue (Wpi-RNase E) and a canonical sized Bsu-RNase J1/J2 homologue (Wpi-RNase J) from Wolbachia pipientis, an α-proteobacterial endosymbiont of arthropods. We found that Wpi-RNase J restored the colony-forming ability (CFA) of Δrne E. coli, whereas Wpi-RNase E did not. Unexpectedly, Wpi-RNase E restored defective CFA due to lack of Eco-RNase G, a paralogue of Eco-RNase E. Our results indicate that bacterial species that lack Eco-RNase E homologues or bacterial species that possess Eco-RNase E homologues which lack Eco-RNase E-like activities have a modest Eco-RNase E-like function using RNase J and/or RNase Y. These results suggest that Eco-RNase E-like activities might distribute among a wide array of bacteria and that functions of RNases may have changed dynamically during evolutionary divergence of bacterial lineages.


Introduction
Studies of RNA decay began with the discovery of unstable mRNA in Escherichia coli [1,2]. RNA decay investigations over many decades have demonstrated its important role in the post-transcriptional regulation of gene expression to adapt to environmental change. E. coli RNase E (Eco-RNase E) encoded by the essential gene rne (Eco-rne) was initially discovered as an endoribonuclease that processes 9S rRNA into mature 5S rRNA [3]. A sequence homologous to RNase E is conserved in various bacterial species [4,5] (see [6]  various carbon sources to understand the similarities between RNase E and RNase J or RNase Y. In addition to B. subtilis, which lacks Eco-RNase E sequence homologues in its genome [37], we used the maternally transmitted endosymbiont Wolbachia pipientis, an α-proteobacteria member with an extremely short Eco-RNase E homologue, to perform E. coli Δrne complementation experiments. Alpha-proteobacteria typically possess Eco-RNase E and Bsu-RNase J1/J2 homologues in their genomes [6], and expression of these genes in W. pipientis has been confirmed by RNA-seq [53], thereby suggesting that these enzymes are physiologically functional in W. pipientis cells. W. pipientis has one of the shortest RNase E/G family proteins among α-proteobacteria with a length of approximately 600 aa, even lacking the typical GWW motif for the common PNPase binding motif, which is generally conserved in the Cterminus among α-proteobacteria [54] (also see S1 Table). These observations led us to investigate the functional relationship between W. pipientis RNase E/G (Wpi-RNase E/G) and RNase J (Wpi-RNase J) homologues with Eco-RNase E.
In this study, we showed that both RNase J and RNase Y have a common enzymatic activity that phenotypically restored the lethality due to Eco-RNase E deficiency allowing Δrne E. coli to grow (form colonies), thereby suggesting that the distribution of Eco-RNase E-like ribonucleolytic activity occurs in a wider range of bacterial species that harbor RNase J and/or RNase Y than previously considered.

Materials and methods
Isolation, maintenance, and preparation of W. pipientis On May 19, 2014, a female adult Eurema mandarina (Insecta; Lepidoptera; Pieridae), formerly known as Eurema hecabe mandarina or E. hecabe yellow-type, was collected on the roadside of a prefectural road on Tanegashima Island, Kagoshima, Japan, where no specific permissions are required for sampling non-endangered insect species. This female was diagnosed by PCR as singly-infected with a cytoplasmic-incompatibility (CI)-inducing W. pipientis strain, wCI [55,56]. This particular W. pipientis wCI strain, designated DK101, was transinfected into a silkworm Bombyx mori cell line, BmN4 (BmN) [57], and maintained as previously described [58], since W. pipientis cannot be cultivated in artificial media. W. pipientis genomic DNA was extracted as a mixture with B. mori chromosomal DNA and used as a DNA template for PCR, as previously described [58].

Nucleotide sequence determination for W. pipientis rne and rnj
The primers used in this study are listed in Table 1. The oligonucleotide primers 5'-UTR-Wpi-rne and 3'-UTR-Wpi-rne were designed based on conserved regions outside the W. pipientis rne (Wpi-rne) open reading frame (ORF) based on eight W. pipientis strains, i.e., wMel, wRi, wHa, wNo, wPip, wBm, wOo, and wCle, the sequences for which were obtained from the KEGG Genes Database (http://www.genome.jp/kegg/). PCR amplification was performed using the total DNA extracted from DK101-infected BmN4 cells. The nucleotide sequence of the complete Wpi-rne ORF was determined using two independent DNA isolates: (i) by directly sequencing the amplified PCR products using the primers 5'-UTR-Wpirne and 3'-UTR-Wpi-rne, including the 5'-untranslated region (UTR) and 3'-UTR; and (ii) by sequencing pLAC-Wpi-rne, which was constructed using the primer set 5'-NotI-Wpi-rne and 3'-SpeI-Wpi-rne, to confirm that the two sequencing results were consistent (GenBank accession number LC177346 for Wpi-rne). The same strategy was used to determine the nucleotide sequence of the complete W. pipientis rnj (Wpi-rnj) ORF using the oligonucleotide primers 5'-UTR-Wpi-rnj and 3'-UTR-Wpi-rnj (GenBank accession number LC177347 for Wpi-rnj). Amplification of the PCR bands for Wpi-rne and Wpi-rnj was confirmed only by DK101-infected BmN4 cells, which showed that the band was derived from DK101 (S1a and S1b Fig).

Bacterial strains and plasmids
The bacterial strains used in this study are listed in Table 2. As the base strain, we used an E. coli strain, CM2100, where a chromosomal deletion in Eco-rne was complemented by a plasmid-borne Eco-rne gene under the control of an araBAD promoter (the kanamycin-resistant [Km r ] plasmid pBAD-RNE) [28]. Eco-rne is essential for E. coli growth, so the addition of 0.1% L-(+)-arabinose to the CM2100 culture allowed E. coli to grow, whereas RNase E was depleted from the cells in the absence of L-(+)-arabinose, which led to growth cessation in both liquid and solid media (so-called E. coli Δrne lethality). MT567 and MT570 were obtained by deleting the ORF of Eco-rng by introducing a Δrng::Km PCR fragment into MT498 and MT504, respectively, followed by eliminating the Km resistance marker using pCP20, as previously described [59], and by transformation with pBAD-RNE after eliminating pCP20. MT875 Table 1. Primer names and sequences used in this study.

Protein analyses
Cultures of DH5α, MT912 (pLAC-Wpi-rne) or MT949 (pLAC-Wpi-rnj) cells were freshly grown to an approximate optical density at 600 nm [OD 600 ] of 2.0 in LB medium containing the appropriate antibiotics but without isopropyl-β-D-thiogalactopyranoside (IPTG). Cells were diluted to an OD 600 of 0.4 in fresh LB medium containing the appropriate antibiotics and 50 μM IPTG, before further culture for 90 min at 37˚C. They were then harvested and prepared for sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Lysates containing total cellular proteins (0.07 OD 600 units per lane) were electrophoresed on 10% polyacrylamide gel and stained with Coomassie Brilliant Blue (CBB) (Quick-CBB PLUS, Wako).

Plating experiments
Plating experiments were essentially performed as described previously [60]. E. coli strains were streaked from glycerol stock and grown overnight at 37˚C on LB plates containing appropriate supplements and antibiotics. After colonies of adequate size (approximately 1 mm in diameter) had formed, the cells were freshly grown (never grown overnight) from a single colony to the mid-to late-log phase at 37˚C in LB medium containing the indicated supplements and antibiotics. Cells were diluted from 10 −4 to 10 −5 , and 100 μl of the diluted culture was spread on LB or M9 plates containing the indicated carbon sources, supplements, and antibiotics. The plates were incubated at 37˚C for 6 days, as indicated, and then scanned. As previously reported [60], the Δrne Δrng double mutant E. coli grows more slowly than the Δrne single mutant E. coli, and thus the incubation period was extended to 14 days (as indicated) to facilitate visualization of the colonies. The incubation conditions in each experiment are described in the figure legends. In general, the absence of detectable colony formation (with 100% colony-forming efficiency (CFE)) after incubation for 6 days was defined as a colony-forming ability (CFA)-negative phenotype.

Statistical analyses
To infer the molecular history of Eco-RNase E homologues derived from W. pipientis, the aa sequences of RNase E/G family proteins derived from 20 bacterial representatives, including W. pipientis and E. coli, were aligned using the online program Clustal Omega [63] and subjected to phylogenetic analyses. Using MEGA version 6.06 [64], the evolutionary history was inferred by the maximum likelihood method based on the JTT matrix-based model [65]. Initial trees for the heuristic search were obtained automatically by applying the neighbor-joining (NJ) and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, before selecting the topology with the best log-likelihood value. A discrete Gamma distribution was used to model the evolutionary rate of differences among sites. The rate variation model allowed some sites to be evolutionarily invariable. All positions containing gaps and missing data were eliminated.
To obtain evolutionary insights into the length variation in RNase E among α-proteobacterial lineages, the nucleotide sequences of the16S rRNA genes from 39 α-proteobacterial lineages with known RNase E lengths as well as from E. coli were aligned using ClustalW, and then subjected to phylogenetic analyses based on the maximum likelihood method using MEGA. Similarly, the aa sequences of RNase E from the 40 lineages were aligned using Clustal Omega and subjected to phylogenetic analyses as described above.
To infer the selective forces operating on RNase E and RNase J in W. pipientis lineages, the nucleotide sequences of the two genes from 11 lineages of W. pipientis were aligned using ClustalW, and the number of nonsynonymous substitutions per nonsynonymous site (Ka) and the number of synonymous substitutions per synonymous site (Ks) [66] were estimated using DnaSP version 5 [67].

Nucleotide sequence accession numbers
The sequences reported in this study have been deposited in the GenBank database (http:// www.ncbi.nlm.nih.gov/genbank/) under accession no. LC177346 for Wpi-rne and no. LC177347 for Wpi-rnj from W. pipientis strain DK101.

Results
The W. pipientis genome contains two ORFs with high similarity to Eco-RNase E and Bsu-RNase J1/J2 B. subtilis RNase J1/J2 encoded by rnjA/rnjB and RNase Y encoded by ymdA are well characterized [38,40,50]. However, genetic analyses have not been reported for the Eco-RNase E homologue or Bsu-RNase J1/J2 homologue from W. pipientis, an endosymbiont α-proteobacterium [6]. Using BLAST search [68], we found that a putative single ORF in W. pipientis in the database included a segment that shared high similarity with the catalytic domain (N-terminal 529 residues) of Eco-RNase E (87% coverage with 39% shared identity), and also that a putative single ORF in W. pipientis included a segment highly similar with the full-length Bsu-RNase J1/J2 (98% coverage with 31% shared identity for J1, and 91% coverage with 31% shared identity for J2). The transcription of these genes has been confirmed by global RNA-seq analysis [53], thereby suggesting that they are functional in W. pipientis cells. No sequences with homology to Bsu-RNase Y were found in the W. pipientis genome.
To construct plasmids that expressed each of these enzymes for use in further analyses, we determined the complete nucleotide sequences of the Eco-rne homologue in W. pipientis (Wpi-rne) and the B. subtilis rnjA/rnjB homologue in W. pipientis (Wpi-rnj). The correspondence between the sequencing results for both the PCR products and the cloned ORFs of Wpirne (pLAC-Wpi-rne) or Wpi-rnj (pLAC-Wpi-rnj) confirmed the presence of an ORF encoding a predicted 591-aa protein (designated as Wpi-RNase E/G) with a calculated molecular weight of 67.1 kDa and an ORF encoding a predicted 544-aa protein (designated as Wpi-RNase J) with a calculated molecular weight of 60.5 kDa (S1 Fig). In silico analyses of Wpi-RNase E/G and Wpi-RNase J Eco-RNase E and Eco-RNase G in the E. coli genome are homologous enzymes, which were probably derived from a single ancestral enzyme via a gene duplication event [69]. In other organisms, the gene sequences homologous to Eco-RNase E and Eco-RNase G are provisionally annotated as "RNase E/G family protein" because their enzymatic activities have not been confirmed, genetically or biochemically. Our phylogenetic analysis indicated that the RNase E/ G family protein members could be separated into two clades, each of which included either Eco-RNase E or Eco-RNase G, and Wpi-RNase E/G was classified as a member of the RNase E group (Fig 1).
In agreement with the phylogenetic relationships of W. pipientis based on five housekeeping genes [70], W. pipientis strains wCI and wPip (a close relative of wCI) shared the most similar RNase E sequences with each other, where both had a 3-aa deletion ca. 20 aa from the C-terminus and they possessed only two amino acid substitutions ( S2 Fig). Strikingly, a 9-aa insertion (FSVRRCTHI) was observed at ca. 50 aa from the C-terminus only in wCI RNase E (S2 Fig). The 9-aa insertion at this position was found consistently in three isolates of wCI derived from three independently-collected butterflies.
To infer the selective forces that operate on Wpi-RNase E and Wpi-RNase J, we aligned and analyzed the nucleotide sequences of the two wCI genes and 10 other lineages of W. pipientis with published whole genome sequences. The average Ka/Ks ratios were 0.1407 for Wpi-RNase E and 0.1093 for Wpi-RNase J, thereby suggesting these are not under positive selection.
RNase J or RNase Y is sufficient partially to reverse the effects of Eco-RNase E deficiency, thereby restoring CFA in Δrne E. coli Previously reported cleavage assays indicate that Bsu-RNase J1/J2 shares a similar RNA cleavage profile with Eco-RNase E [40,48], while the similarity of interaction partners and Prevalence of enzymatic activity to restore lethality by RNase E deficiency degradosome assembly between Bsu-RNase Y and Eco-RNase E also suggests a functional relationship between these two enzymes [51]. However, there have been no reports of the successful growth restoration of Δrne E. coli using these enzymes. Recently, we reported that a part of the essentiality for Eco-RNase E is nutrient-dependent and that M9 minimal media, instead of LB media, supported the CFA of various Δrne revertants [28,52]. Thus, we were interested in learning whether the similar shared profiles of RNase Y and RNase J when complementing the essentiality of Eco-RNase E depended on different carbon sources, thereby helping to visualize the functional distinction between these ribonucleases. In these experiments, an IPTG-inducible plasmid containing an ORF for Bsu-RNase Y (pLAC-ymdA), Bsu-RNase J1 (pLAC-rnjA), Bsu-RNase J2 (pLAC-rnjB), Wpi-RNase E (pLAC-Wpi-rne), or Wpi-RNase J (pLAC-Wpi-rnj) was introduced via transformation into E. coli strain CM2100 harboring the pBAD-RNE plasmid with the chromosomal rne deleted. The ability to complement E. coli Δrne was tested by observing CFA without L-(+)-arabinose, as described previously [28,52,60]. Appropriate concentration of IPTG for each RNase to restore CFA of Δrne E. coli was determined by preliminary plating experiments (S3 Fig) and the expression of GFPuv or each RNase in the complemented strain was confirmed by SDS-PAGE (S4 and S5 Figs). The CFA was restored for the full-length Eco-RNase E-complemented Δrne E. coli on all media plates tested (Fig 2a).
Colonies with approximately 1 mm in diameter could be observed within 24hr for full-length Eco-RNase E complemented Δrne E. coli strain (MT1282). GFPuv did not restore CFA on any of the media plates tested (Fig 2b), but the restoration of CFA was observed on M9 minimal media plates when Δrne E. coli was complemented with Bsu-RNase Y, Bsu-RNase J1, Bsu-RNase J2, or Wpi-RNase J (Fig 2c-2e and 2g), which suggests that these enzymes have an enzymatic activity similar to Eco-RNase E. Wpi-RNase E did not cause restoration on any of the media plates tested (Fig 2f). Except for the full-length Eco-RNase E, none of the complementation effects restored CFA in Δrne E. coli on LB media plates (Fig 2).
To exclude the possibility that the restoration of CFA by Bsu-RNase Y, Bsu-RNase J1/J2, and Wpi-RNase J complementation was due to the leaky expression of Eco-RNase E from the pBAD-RNE plasmid in these cells, we eliminated the pBAD-RNE plasmid from MT1072 (Bsu-RNase Y), MT1070 (Bsu-RNase J1), MT1125 (Bsu-RNase J2), and MT956 (Wpi-RNase J) via displacement with pSC101 [71,72] after a few culture passages by streaking the colonies on M9-glycerol plates in the presence of Tc but without Km. The colony size and growth rate were not increased by the addition of 0.1% L-(+)-arabinose after this replacement, thereby demonstrating the ability of Bsu-RNase Y, Bsu-RNase J1/J2, and Wpi-RNase J to restore the lethality due to Eco-RNase E deficiency to form colonies on solid media. The loss of the rne gene from these strains was confirmed by PCR (data not shown).
More than 4 days were required to form solid colonies (approximately 1.0 mm in diameter) and the resulting Bsu-RNase Y-or RNase J-complemented Δrne E. coli cells exhibited extensive filament formation with unevenly-distributed nucleoids, which are typical of the E. coli Δrne mutation (Fig 3), thereby indicating that the ability of these enzymes to complement Eco-RNase E essentiality was conditional and only partial.
Restoration of CFA in Δrne E. coli by Bsu-RNase Y, Bsu-RNase J1/J2, or Wpi-RNase J has different dependencies on endogenous Eco-RNase G The overproduction of Eco-RNase G and its derivatives is known to restore the growth of Δrne E. coli [13,24,26], and we also found that endogenous Eco-RNase G was necessary for establishing second-site suppression of E. coli Δrne lethality in N3433-based strains [52]. To investigate whether the restoration of CFA in Δrne E. coli by Bsu-RNase Y, Bsu-RNase J1/J2, and Wpi-RNase J required the presence of endogenous Eco-RNase G, we constructed Δrne Δrng double mutant E. coli strains complemented by each of these RNases (MT1177, MT1158, MT1176, and MT1173) (Fig 4). In all four strains, the rate (speed) of colony formation was slower than that in Δrne rng + E. coli (cf. Fig 2) and incubation for 14 days was required to visualize the colonies clearly. Three different profiles of Eco-rng dependence were observed for the restoration of CFA. No CFA in MT1177 (Bsu-RNase Y) and MT1176 (Bsu-RNase J2) were observed either on glycerol or pyruvate media (Fig 4a and 4c), thereby indicating that endogenous Eco-rng is necessary to support basic sugar utilization for CFA even when Bsu-RNase Y or Bsu-RNase J2 is present. CFA in MT1158 (Bsu-RNase J1) was observed in both media conditions (Fig 4b), which indicates that the basic functions of both glycolysis and gluconeogenesis (upstream of pyruvate) were restored independently of endogenous Eco-rng. MT1173 Prevalence of enzymatic activity to restore lethality by RNase E deficiency (Wpi-RNase J) exhibited CFA on glycerol but not on pyruvate media (Fig 4d), thereby indicating that pyruvate utilization is dependent on endogenous Eco-rng but without affecting basic glycolysis. These results suggest that independent genetic factors are required for CFA with each carbon source, and that the combination of endogenous Eco-rng and either one of Bsu-RNase Y, Bsu-RNase J2, or Wpi-RNase J was necessary to restore CFA in all three medium conditions. It should be noted that decreased CFE was observed for MT1176 (Bsu-RNase J2) on LB but not on glycerol in the presence of 0.1% L-(+)-arabinose (Fig 4c), which cannot be explained at present. These results suggest that the requirement of Eco-RNase E for carbon utilization is multifaceted and that E. coli endogenous Eco-rng has essential roles in the utilization , and MT1094 (Wpi-RNase J) (d) grown on M9-glycerol plates (0.6% gellan gum) were suspended in M9 minimal medium and then spread directly onto glass slides. Slides were prepared for microscopic observations, as previously described [25], except that poly-L-lysine was omitted. Cells were stained with DAPI (ProLong Diamond Antifade Mountant with DAPI) and microscopic images were obtained using an Axiovert 200 system (Zeiss). A scale bar measuring 5 μm is shown below each DIC image. https://doi.org/10.1371/journal.pone.0177915.g003 Prevalence of enzymatic activity to restore lethality by RNase E deficiency of certain carbon sources in the absence of Eco-RNase E. Our results also indicate that different RNase J homologues from different bacteria species had distinct activities in the restoration of Eco-RNase E deficiency.

Restoration of E. coli Δrne lethality by RNase J and RNase Y depend on their endonucleolytic activities
Eco-RNase E is an endoribonuclease, so we also investigated whether E. coli Δrne complementation depended on the endonucleolytic activities of RNase J and RNase Y. RNase J has both endonucleolytic and 5 0 -to-3 0 exonucleolytic activities [40,73]. According to structural analyses, RNase J has only one catalytic center and the introduction of double mutations (D78K and H79A) into the catalytic region of Bsu-RNase J1 abolishes both its endo-and exonucleolytic activities [74]. Equivalent mutations in Wpi-RNase J are D77K and H78A according to our sequencing analysis. Bsu-RNase Y exhibits only endonucleolytic activity [50] and the introduction of double mutations (H367A and D368A) abolishes its endonucleolytic activity [49].
Interestingly, the introduction of double mutations (D85K and H86A) into RNase J (Msm-RNase J) from M. smegmatis strongly impaired majority of its exonucleolytic activity and considerable endonucleolytic activity was retained [46]. We consider that Msm-RNase J is a good tool for distinguishing the endonucleolytic and exonucleolytic activities, and thus we prepared plasmids that expressed the wild-type Msm-RNase J enzyme or D85K-H86A mutated enzyme (Msm-mutRNase J) for the E. coli Δrne complementation assay. Disruption of both the endonucleolytic and exonucleolytic activities of Bsu-RNase J1 with D78K-H79A mutations (Bsu-mutRNase J1) eliminated its ability to restore CFA to Δrne E. coli on glycerol or pyruvate (~15% CFE) (Fig 5a). Disruption of the endonucleolytic activity of Bsu-RNase Y with H367A-D368A mutations (Bsu-mutRNase Y) also eliminated its ability to restore CFA in Δrne E. coli on glycerol or pyruvate (~13% CFE) (Fig 5b). In addition to Bsu-RNase J1/J2 and Bsu-RNase Y (see Fig 2), the natural form Msm-RNase J conferred CFA in Δrne E. coli on glycerol or pyruvate after incubation for 6 days (Fig 5c). In contrast to Bsu-RNase J1 or Bsu-RNase Y, D85K-H86A mutations in Msm-RNase J did not abolish the ability to restore CFA, although the rate of colony formation slowed down (Fig 5d). Equivalent amount of RNase expression between wild-type enzyme and mutant enzyme was confirmed by SDS-PAGE analysis (S6 Fig). These results suggest that the endonucleolytic activity of RNase J and RNase Y is sufficient to restore CFA in Δrne E. coli on glycerol or pyruvate, although the exonucleolytic activity is necessary to enhance the speed of colony formation. The D77K-H78A mutated Wpi-RNase J also lost its ability to restore CFA in Δrne E. coli on glycerol or pyruvate (Fig 5e), thereby suggesting that Wpi-RNase J is a ribonuclease with similar enzymatic features to Bsu-RNase J1 and that the ribonucleolytic activity of Wpi-RNase J is necessary to restore the lethality due to Eco-RNase E deficiency.
Distinction of the RNase E and RNase G activities using phenotypic defects with E. coli rne rng double mutant bacteria The length of Wpi-RNase E is 591 aa with an approximately 50-aa insertion in the S1 domain, which is typical of α-proteobacteria (see [6] for review). Considering that this insertion does not exist in the Eco-RNase E sequence, the "net" length of Wpi-RNase E is closer to that of Eco-RNase G (489 aa) rather than that of Eco-RNase E (1061 aa), although BLAST search and phylogenetic analysis detected higher similarity in the catalytic region of Eco-RNase E (also see Fig 1). Eco-rng is not an essential gene; thus, the function of RNase G was considered mostly covered by the RNase E function, which made it difficult to distinguish the RNase G function phenotypically from the RNase E function, although the cleavage or degradation of certain transcripts is RNase G-specific [7,13,75,76]. We reported previously that ydfV::Tn10 mutation (second-site suppressor) allowed E. coli to grow in the absence of essential RNase E presumably by repressing the expression of RelB, the anti-toxin of RelBE anti-toxin/toxin system [52]. Experiments conducted to examine the Tn10-inserted Δrne E. coli revertant strain (Δrne ydfV::Tn10 strain) showed that endogenous Eco-rng was essential for growth on pyruvate minimal medium in this strain background (Fig 6b). This growth defect was not restored by adding casamino acid (0.05%) or the aa isoleucine (50 μg/ml) (Tamura et al., unpublished data), which suggests that the CFA-defective phenotype of Δrne ydfV::Tn10 Δrng was not caused simply by the defect in a certain aa production pathway, as shown previously [77]. However, the endogenous Eco-rng, which is not normally sufficient to complement E. coli Δrne lethality, was sufficient to restore the CFA (Fig 6a).
We employed this Eco-rng null defective CFA phenotype to distinguish the RNase G-like activity by complementing various RNases of interest based on multi-copy plasmids. The presence of endogenous Eco-rng was sufficient to restore CFA (Fig 6a) whereas the overproduction of GFPuv, as a negative control, did not confer CFA in Δrne ydfV::Tn10 Δrng bacteria (Fig 6b). The overproduction of Eco-RNase G from the pNRNE4 plasmid-based gene expression system, which is known to restore E. coli Δrne lethality [13,60], restored the defective CFA obtained via Eco-rng null mutations, thereby confirming the validity of this assay system (Fig  6c). Wpi-RNase E, which did not restore the E. coli Δrne lethality (cf . Fig 2f), restored CFA to indicate a functional relationship between Wpi-RNase E and Eco-RNase G (Fig 6d). Wpi-RNase J and Bsu-RNase J1, which restored the E. coli Δrne lethality (cf . Fig 2d and 2g), also restored the CFA, which indicates that these enzymes possess the features of both Eco-RNase E and Eco-RNase G (Fig 6e and 6f). Bsu-RNase Y, which restored the E. coli Δrne lethality (cf .  Fig 2c), did not restore the defective CFA obtained via Eco-rng null mutations, thereby suggesting separate activities for Eco-RNase E and Eco-RNase G in the pyruvate utilization pathway of E. coli (Fig 6g). These results indicate that homologues of RNase E/G, RNase J, and RNase Y exhibit a variety of enzymatic features in terms of their Eco-RNase E-and Eco-RNase G-like activities.
To further evaluate the RNase E/G-like activity in RNase Y and RNase J, accumulation of 16S rRNA precursors were examined by a method described by Wachi et al. [8]. Overproduction of Eco-RNase E or Eco-RNase G restored the cleavage of the precursors to produce mature 16S rRNA (S7b Fig, lane 2 and 3, respectively). Overproduction of Bsu-RNase J1 or Wpi-RNase J also restored the production of 16S rRNA (S7b Fig, lane 4 and 6, respectively). Overproduction of Bsu-RNase Y or Wpi-RNase E accumulated 16.3S precursor band and produced mature 16S rRNA inefficiently (S7b Fig, lane 5 and 7, respectively). These results suggest that Bsu-RNase J1, Bsu-RNase Y, Wpi-RNase J, and Wpi-RNase E have Eco-RNase E/G-like activity sufficient to produce 16S rRNA, although the efficiency of Bsu-RNase Y or Wpi-RNase E cleavage was lower than RNase J.

Discussion
Our results indicate that RNase J and RNase Y have the ability partially to restore the lethality due to Eco-RNase E deficiency. A similar cleavage site to Eco-RNase E was reported for Bsu-RNase J [40] and RNase Y [49,50], and similarity of interaction partners and Eco-RNase Elike degradosome assembly was reported for Bsu-RNase Y [38,51]. There is no published phenotypic evidence, however, for a functional relationship between these enzymes. Our results suggest that bacteria that lack RNase E homologues have a similar enzymatic activity, which is presumably endonucleolytic, to Eco-RNase E via RNase Y or RNase J, thereby indicating that an Eco-RNase E-like activity is distributed among a wider range of bacterial species than previously considered. Various types of genomic and biochemical data suggest that all bacterial species, except for some Candidatus strains, contain at least one RNase E, RNase G, RNase J, or RNase Y homologue [6,50], which indicates that an Eco-RNase E-like enzymatic activity exists in all these bacteria. The exosome is the major RNA decay machinery in eukaryotes [34] but human cells also possess ard-1 with a sequence that is not homologous to Eco-RNase E, although it can phenotypically complement rne ts [78]. Overall, our results suggest that an enzymatic activity that establishes Eco-RNase E essentiality might be distributed widely in different classes of enzymes among living organisms ranging from bacteria to eukaryotes.
In this study, we cloned and genetically characterized Wpi-RNase E/G and Wpi-RNase J from W. pipientis. Our phylogenetic analyses based on aa sequences suggested that Wpi-RNase E/G should be categorized as an RNase E, but the complementation assays of E. coli Δrne and E. coli Δrng demonstrated that Wpi-RNase E is capable of complementing Eco-RNase G rather than Eco-RNase E. This result might reflect a functional shift from RNase E to RNase G, or vice versa, in an ancestor of W. pipientis (α-proteobacteria) or E. coli (γproteobacteria).
According to the KEGG database, we found a remarkable variety of RNase E/G lengths (344 to 1123 aa) among α-proteobacteria (S1 Table). This is an interesting observation, considering that the average gene length is generally highly conserved in prokaryotes [79]. Thus, we examined whether there was any sign of directional selection in the length of RNase E/G during the diversification of α-proteobacterial lineages. The phylogenetic tree constructed for αproteobacterial lineages based on 16S rRNA gene sequences showed that the subclass Rickettsidae (including both Rickettsiales and Pelagibacterales) consistently had remarkably short RNase E lengths ( 720 aa) (S8 Fig). The phylogenetic tree based on the aa sequences of RNase E had a similar topology to that based on 16S rRNA gene sequences (S9 Fig), thereby excluding the possibility of the horizontal transfer of RNase E. The members of the subclass Caulobacteridae showed longer RNase E lengths with exceptions in two species, R. elongatum and Ca. Liberibacter asiaticus, which had short RNase E length (608 and 723 aa, respectively). It is considered more parsimonious to speculate that the short RNase E is a derived trait rather than an ancestral trait. However, detailed functional and structural analyses remain to be made to draw any conclusions about the evolutionary history of the length polymorphism of RNase E.
Our finding that mutations in conserved amino acids necessary to establish a ribonucleolytic activity for Bsu-RNase J1 [74] or Msm-RNase J [46] also abolished the ability of Wpi-RNase J to restore the lethality due to Eco-RNase E deficiency strongly suggests that Wpi-RNase J shares similar enzymatic characteristics with RNase J. Using mutated Msm-RNase J, we also found that the endonucleolytic activity of the enzyme was necessary to restore CFA in Δrne E. coli, which is unsurprising considering that Eco-RNase E is an endoribonuclease. Previous studies have not reported a common cleavage site for RNase J and RNase Y. The common feature of conferring CFA in Δrne E. coli implies that they share similar characteristics as endoribonucleases, but our results cannot exclude the possibility that bulk RNA decay rather than the specific cleavage of a target RNA might be the mechanism for restoring CFA. Although recent studies revealed the role of Hfq on processing or degradation of various RNAs by RNase E [20,80], involvement of Hfq on RNase E essentiality is improbable, since hfq is not an essential gene in E. coli and RNase E lacking the scaffold region necessary to interact with Hfq still is able to restore CFA of Δrne E. coli.
We developed an assay system to evaluate Eco-RNase G based on the defective growth of the rne ydfV::Tn10 rng strain on pyruvate minimal medium. Eco-RNase G is not an essential enzyme for E. coli growth and this feature made it difficult to evaluate its activity by genetic screening. Eco-RNase G is recognized as a homologue of Eco-RNase E, but the influence on certain RNAs is specific to Eco-RNase G [13,[75][76][77]81]. This feature may have caused the conditional synthetic lethality only on pyruvate minimal medium in the absence of both Ecorne and Eco-rng. Thus, using this assay, we could demonstrate that Wpi-RNase E has an Eco-RNase G-like activity, and we phenotypically distinguished the ability to complement Δrne or Δrng with Bsu-RNase Y for Eco-RNase G and RNase J.
We found that some medium conditions allowed RNase J and RNase Y phenotypically to restore the lethality due to Eco-RNase E deficiency in E. coli, but this restoration was limited to M9 minimal medium and did not occur on LB medium. As reported previously, CFA in Δrne E. coli on LB is restored either by the deaD::Tn10 insertion mutation or by Eco-RNase G overproduction, but not by the ydfV::Tn10 insertion mutation [28,60]. These results suggest that the genetic factor(s) necessary for supporting growth on LB is genetically distinct from the factor(s) required for growth on minimal medium and it is not restored by RNase J or RNase Y complementation (see Fig 2). The genetic factor(s) responsible for LB growth by E. coli (which cannot be explained by nutrient auxotrophy) remains unknown at present.
In summary, the culture of Δrne E. coli on M9 minimal medium was used to evaluate the functional similarity between Eco-RNase E and RNase J or RNase Y from other organisms. We cloned and genetically characterized Wpi-RNase E/G as RNase G and Wpi-RNase J as RNase J. The various phenotypes observed after complementation by the ribonucleases tested in this study suggest that the normal function of Eco-RNase E is diverse during E. coli carbon source utilization. Our results indicate that the fundamental function of Eco-RNase E in conferring CFA in Δrne E. coli is also mediated by RNase J and RNase Y in various bacterial species. MT696 was inoculated into LB medium in the presence of 0.1% L-(+)-arabinose, cultured at 37˚C to mid-log phase, harvested, washed once with LB medium, and reinoculated into LB medium at an OD 600 of 0.05 in the presence of each concentration of IPTG (as indicated) lacking 0.1% L-(+)-arabinose, harvested, and the amount of 0.1 OD 600 per well was separated on a 10% SDS polyacrylamide gel. The gel was stained with CBB. The arrowhead indicates the expected size of GFPuv protein with a calculated molecular weight of 26.8 kDa. The size of protein marker is shown on the left side. (b) Culture of MT696 was spread onto LB and M9-Glyce plates (0.6% gellan gum) with various concentration of IPTG as indicated. Plates were scanned after incubation at 37˚C for 6 days. (PDF)

S6 Fig. Confirmation of in vivo protein expression of wild-type and mutated RNases in
Δrne E. coli by SDS-PAGE. (a) All bacterial strains were inoculated into LB medium in the presence of 0.1% L-(+)-arabinose, cultured at 37˚C to mid-log phase, harvested, washed once with LB medium, and reinoculated into LB medium at an OD 600 of 0.05 in the absence of 0.1% L-(+)-arabinose containing the appropriate antibiotics and IPTG (10 μM for MT956 and MT983; 50 μM for MT1072 and MT1315; no IPTG (leaky expression) for MT696, MT1070, MT1200, MT1266, and MT1288). Cultures were grown for 6 h and then harvested. The amount of 0.1 OD 600 per well was separated on a 10% SDS polyacrylamide gel. The gel was stained with CBB. The asterisk shown on the right side indicates the expected band of each protein. The size of protein marker is shown on the left side. (b) MT1200 was inoculated into LB medium in the presence of 0.1% L-(+)-arabinose, cultured at 37˚C to mid-log phase, harvested, washed once with LB medium, and reinoculated into LB medium at an OD 600 of 0.05 in the absence of 0.1% L-(+)-arabinose containing the appropriate antibiotics and IPTG as indicated. Cultures were grown for 6 h and then harvested. The amount of 0.1 OD 600 per well was separated on a 10% SDS polyacrylamide gel. The gel was stained with CBB. The arrowhead shown on the right side indicates the expected size (59. α-proteobacterial lineages. The evolutionary history was inferred by using the Maximum Likelihood method (See Materials and methods for details). The percentage of trees in which the associated taxa clustered together (bootstrap values) is shown next to the branches. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. Nodes with less than 50% bootstrap support are collapsed. KEGG organism codes are given in square brackets. Amino acid lengths of RNase E are given in parentheses. Taxa with short RNase E (< 750 a.a.) are highlighted with shading. (PDF) S1 Table. Length of RNase E/G homologues in α-proteobacteria. (PDF)