RNase Y Autoregulates Its Synthesis in Bacillus subtilis

The instability of messenger RNA is crucial to the control of gene expression. In Bacillus subtilis, RNase Y is the major decay-initiating endoribonuclease. Here, we show how this key enzyme regulates its own synthesis by modulating the longevity of its mRNA. Autoregulation is achieved through cleavages in two regions of the rny (RNase Y) transcript: (i) within the first ~100 nucleotides of the open reading frame, immediately inactivating the mRNA for further rounds of translation; (ii) cleavages in the rny 5′ UTR, primarily within the 5′-terminal 50 nucleotides, creating entry sites for the 5′ exonuclease J1 whose progression is blocked around position −15 of the rny mRNA, potentially by initiating ribosomes. This links the functional inactivation of the transcript by RNase J1 to translation efficiency, depending on the ribosome occupancy at the translation initiation site. By these mechanisms, RNase Y can initiate degradation of its own mRNA when the enzyme is not occupied with degradation of other RNAs and thus prevent its overexpression beyond the needs of RNA metabolism.


Introduction
Control of gene expression requires the instability of messenger RNA. In bacteria, mRNA half-lives vary from seconds to over an hour. The decay of mRNA generally follows an "all-or-none" pattern, implying that if control is to be efficient, it must occur at the rate-limiting initial step of the degradation process. Studies in Escherichia coli and Bacillus subtilis, two evolutionary very distant organisms, have provided essential knowledge on the initiation of mRNA decay by endonucleolytic cleavage that can be summarized by "different enzymes-similar strategies". The crucial role of the endoribonucleases RNase E (E. coli) and RNase Y (B. subtilis) in producing short-lived decay intermediates is now clearly recognized [1]. In B. subtilis, RNase Y affects global mRNA stability by cleaving its substrates with an RNase E-like single strand-specific cleavage specificity [2]. The intracellular levels of a majority of transcripts are affected in RNase Y-depleted strains [3][4][5]. An alternative degradation pathway has been described for some mRNAs where the decay of the primary transcript is triggered by the RNA pyrophosphohydrolase RppH which generates a monophosphorylated intermediate that becomes vulnerable to 5 -exonucleolytic digestion by RNase J1 [6]. RNase J, comprising RNase J1 and J2 subunits, is a multifunction enzyme with both endonuclease and 5 exonuclease activity; the latter activity being essentially a property of the J1 subunit that requires a monophosphorylated 5 -end (or 5 -OH) [7][8][9][10]. RNase J1 thus functions primarily in the 5 exonucleolytic degradation of RNA decay intermediates generated by endonucleolytic cleavage or 5 -end conversion. Interestingly, a monophosphorylated RNA 5 end also potentiates the endonucleolytic activity of RNase Y in vitro [2], but it is unknown whether this plays a significant role in the RNase Y-dependent degradation of RNAs in vivo.
The sequence specificity of RNase Y in B. subtilis and other Gram-positive organisms is rather loose with a preference for UA-rich single-stranded sequences often flanked by
Translational rny-lacZ fusions (see plasmids) are based on plasmid pHM14 [40] and were integrated in single copy at the amyE locus of the B. subtilis SSB1002 strain (see Table 1). Insertions were confirmed by PCR and sequencing. For the long rny-lacZ fusion and its mutated variants, the products of the ligation reaction were amplified with primers HP2248 and HP2249 and used to transform SSB1002 directly. Primers used in this study are listed in Table 2.

Plasmid Constructs
pHMD17. Plasmid pHMD17 contains the entire rny promoter region with the rny ORF up to codon 5 inserted as a BamHI-EcoRI fragment (primers HP1714-HP1715) in-frame into the lacZ fusion vector pHM14 [40].
pHMD22. Plasmid pHMD22 contains the rny gene including the Shine-Dalgarno sequence as a 1.6 kb PCR fragment (primers HP1832-HP1833) inserted into the inducible expression vector pWH353m between sites SphI and SalI.
pHMD28. Plasmid pHMD28 contains the entire rny promoter region with the rny ORF up to codon 13 inserted as a BamHI-EcoRI fragment (primers HP2164-HP2165) in-frame into the lacZ fusion vector pHM14 [40].
pHMD29. Plasmid pHMD29 contains the entire rny promoter region with the rny ORF up to codon 31 inserted as a BamHI-EcoRI fragment (primers HP1714-HP2166) in-frame into the lacZ fusion vector pHM14 [40].
pHMD33. Plasmid pHMD33 is similar to pHMD22 but contains a mutated rny gene which codes for an inactive RNase Y (mutations H368A and D369A in HD domain). Two overlapping PCR fragments (primer pairs HP1832-HP2228 and HP2227-HP1833) containing the rny double mutations at the overlap were amplified from chromosomal DNA using KOD DNA polymerase (Novagen). Both fragments were used as the template for amplification with primers HP1832-HP1833. The resulting fragment was digested with SalI and SphI and ligated into the respective sites of plasmid pWH353.
pHMD37. Plasmid pHMD37 contains the entire rny promoter region with the rny ORF up to codon 210 inserted as a BamHI-EcoRI fragment (primers HP1714-HP1801) in-frame into the lacZ fusion vector pHM14 [40]. The ligation mix was amplified with primers HP2248-HP2249 and the resulting fragment used for transformation.
pHMD38. Plasmid pHMD38 is similar to pHMD37 but lacks a large part of the rny 5 UTR (deletion between positions −20 and −158). A PCR fragment (primers HP1714-HP2269) was amplified and used as a megaprimer in a second PCR reaction with primer HP1801 using plasmid pHMD37 as the template. The resulting fragment was inserted as a BamHI-EcoRI fragment in-frame into the lacZ fusion vector pHM14 [40].
pHMD39. Plasmid pHMD38 is similar to pHMD17 but lacks a large part of the rny 5 UTR (deletion between positions −20 and −158). A PCR fragment (primers HP1714-HP2268) was amplified plasmid pHMD17 as the template and inserted as a BamHI-EcoRI fragment in-frame into the lacZ fusion vector pHM14 [40].
pWH353m. Plasmid pWH353m is a modified version of the pWH353 vector [43] where the cat cassette was replaced between sites SmaI-StuI with the mls resistance cassette (sites StuI-SmaI) of plasmid pDG1663 [44].

Western Blot
For Western blot analysis, 20 µg of protein extract prepared as described previously [45] was separated by SDS-PAGE (10%). After electrophoretic transfer of the proteins, the nitrocellulose membrane (GE Healthcare, Chicago, IL, USA) was stained with amido black to check for equal transfer across all lanes. The membrane was blocked for 1 h with 5% skimmed milk in phosphate-buffered saline (PBS)-Tween buffer (100 mM NaH 2 PO 4 -Na 2 HPO 4 , pH 7.4, 100 mM NaCl, 0.1% Tween) and incubated with a monoclonal RNase Y antibody diluted in PBS-Tween for at least 4 h. Signals were detected by ECL Chemiluminescence (BIO-RAD, Clarity TM Western ECL) associated with a CCD camera (BIO-RAD ChemiDoc XR System+). When necessary, ECL-detected proteins were quantified using the ImageLab software version 6.0.1 (Bio-Rad, Hercules, CA, USA). The uncropped versions of the Western blots are shown in Supplementary Materials.

Northern Blot
RNA blot analysis was carried out using 5-10 µg of total RNA isolated as described previously [40], separated on a 1% agarose/0.22 M formaldehyde gel, and blotted onto Hybond N+ membranes (GE Healthcare). RNA was crosslinked to the membrane at 120 mJ cm -2 for 1 min. RNA probes were denatured at 80 • C for 3 min, kept on ice for 2 min, and then added to the prehybridized membranes (Amersham Rapid-hyb buffer) for hybridization at 70 • C for 2 h. Membranes were washed at 68 • C with washing solutions I (2 × SSC and 0.1% SDS), II (1 × SSC, 0.1% SDS), and III (0.5 × SSC, 0.1% SDS) for 15 min each. The signals were detected with a GE Healthcare Typhoon FLA 9500 imaging system. As lacZ and 5 UTR rny probes, T7 RNA polymerase in vitro transcripts were used which were generated from a PCR template obtained with oligonucleotides HP2270-HP2271 and HP2238-HP2239, respectively. Chromosomal DNA of the B. subtilis SSB530 strain was used as a template for PCR. 16S rRNA was used as loading control (hybridization with oligonucleotide HP2113). The uncropped versions of the Northern blots are shown in Supplementary Materials.

Half-Life Measurements
For half-life measurements, rifampicin (Sigma, St. Louis, MO, USA) was added at 150 µg/mL to cultures at OD 600~0 .5. At 0, 1, 2, 4, 6, and 10 min after addition of rifampicin, 10 mL of cell culture were mixed with 5 mL of frozen buffer (20 mM Tris-HCl pH 7.8, 5 mM MgCl 2 , 20 mM NaN 3 ), centrifuged, and frozen in dry ice/ethanol. Total RNA from the samples was purified and analyzed by Northern blot as described above.

Primer Extension Analysis
Primer extension was carried out using 5 µg of total RNA, 5 [ 32 P]-labelled primer HP905 complementary to sequences at the 5 end of the lacZ ORF and Superscript III reverse transcriptase (Invitrogen, Waltham, MA, USA) following the instructions of the manufacturer. Reactions were carried out at 55 • C for 10 min.

β-Galactosidase Measurements
For β-galactosidase measurements, cells were grown in the appropriate medium and cell extracts were prepared from 3 to 5 mL of cell culture. Extracts were used for measurements as described elsewhere [46]. All reported values are based on two to four independent measurements. Error bars represent standard deviation (sd).

Rny-lacZ Expression Is Responsive to RNase Y Levels In Vivo
We first assessed the ability of RNase Y to regulate its own expression using two translational lacZ fusions. Both contained the entire 164 nt 5 UTR of the rny gene identified previously by a transcriptomic analysis [47] and confirmed by primer extension analysis (see below). Transcription is driven by a typical sigma A-type promoter ( Figure 1B). The two lacZ constructs differed in the length of the rny ORF fused in-frame to lacZ, i.e., 5 and 210 codons, referred to as short and long fusions, respectively ( Figure 1B). The fusions were integrated in single copy at the amyE locus. galactosidase activity and the level of fusion protein in the cell by Western blot, respectively ( Figure 1D, lanes 9 and 10). These data indicated that RNase Y might autoregulate its own expression and that control primarily involves rny mRNA sequences between codon 5 and 210.  Regulated overexpression of RNase Y was achieved by further transforming rny-lacZcontaining strains with the rny expression plasmid pHMD22 that allows transcriptional control of the additional plasmid-borne rny gene by a tetracycline-inducible promoter. Growth in the presence of inducer increased the intracellular RNase Y level up to about 8-to 10-fold ( Figure 1C, lower panels, confer lanes 1, 2 and 4, 5). This overexpression of RNase Y only slightly reduced the activity of the short rny-lacZ fusion (~1.3-fold) while the long rny-lacZ fusion was strongly (~6-fold) repressed ( Figure 1C, β-gal activities at the top).
To test whether this control requires the catalytic activity of RNase Y we also overproduced, an RNase Y protein where the conserved HD motif in the active site was mutated (H368A/D369A, Figure 1A). We have previously shown that single mutations of the HD motif almost abolish the endonucleolytic activity of RNase Y [2]. In contrast to the wildtype enzyme, overexpressing the inactive form of RNase Y ( Figure 1C, lanes 3 and 6) does not decrease the activity of the short nor the long form of the rny-lacZ fusions. On the contrary, expression of the long lacZ fusion was significantly increased under these conditions, suggesting that inactive RNase Y interferes with the activity of the chromosomeencoded wild-type form of the enzyme ( Figure 1C, see also Discussion). This increase in β-galactosidase activity is also visible in the Western blot analysis aimed at quantifying RNase Y overproduction ( Figure 1C, Western blots at the bottom, cf. lanes 5 and 6). Since the monoclonal antibody is directed against a peptide in the N-proximal 210 aa (but not within the N-terminal 5 aa) of RNase Y ( Figure 1A), we could also detect the fusion protein synthesized from the long but not the short form of the rny-lacZ fusion construct ( Figure 1C, lower panel, lane 6; see also lower WB panel of Figure 1D).
We next analyzed the effect of the absence of RNase Y on rny-lacZ expression. For this purpose, the entire rny ORF was replaced with a spectinomycin resistance cassette. Expression of the short rny-lacZ fusion remained unchanged in the absence of RNase Y. However, expression of the long fusion was induced 2-to 3-fold when measuring both βgalactosidase activity and the level of fusion protein in the cell by Western blot, respectively ( Figure 1D, lanes 9 and 10).
These data indicated that RNase Y might autoregulate its own expression and that control primarily involves rny mRNA sequences between codon 5 and 210.

RNase Y Affects rny and rny-lacZ Transcript Levels Alike
The data obtained with the short rny-lacZ fusion described above strongly suggested that the lacZ mRNA part of the fusion mRNA does not contribute to autoregulation (see also discussion). Nevertheless, to test how RNase Y affects its own expression, we analyzed the effect of RNase Y on both rny and rny-lacZ mRNA levels. First, we transformed the wild-type strain (SSB1002) with plasmid pHMD22 for inducible RNase Y overexpression (pY) or with the empty vector as control (pC). The rny messenger transcribed from the chromosomal gene was detected with a probe specific for the rny 5 UTR which does not detect rny mRNA transcribed from the plasmid since this gene copy lacked the 5 UTR. The rny mRNA level transcribed from the chromosome was very sensitive to the intracellular RNase Y concentration (Figure 2A). In the absence of inducer, we observed a two-to three-fold overproduction of RNase Y due to leaky expression. This was already sufficient to strongly reduce chromosomal rny mRNA about four-fold (Figure 2A). In the presence of inducer RNase Y was overproduced~ten-fold, causing rny mRNA to drop below detectable levels ( Figure 2A). In contrast, a similar overproduction of the inactive RNase Y mutant led to a more than two-fold increase in rny transcripts (Figure 2A), mirroring a similar effect on the β-galactosidase expression of the long rny-lacZ fusion under the same conditions ( Figure 1C). The transcripts of the short and long rny-lacZ fusions, detected using a lacZspecific probe, responded in much the same way as the rny mRNA to overproduction of RNase Y ( Figure 2B). While the long fusion mRNA became almost undetectable when overproducing RNase Y, the short fusion mRNA was also reduced two-fold suggesting that RNase Y does not only cleave within the rny ORF but also in the 5′ UTR ( Figure 2B). In the absence of RNase Y (Figure 2C), the short rny-lacZ fusion mRNA is barely increased (~1.1-fold) while induction of the long rny-lacZ fusion mRNA is upregulated by a factor of 5-fold compared to a wild-type strain, again suggesting that 5′ proximal sequences in the rny ORF play an important role for controlling the rny transcript levels in the cell. The similar response of rny mRNA and the long rny-lacZ transcript to changes in the cellular level of RNase Y Figure 2. Effect of RNase Y on wild-type rny and rny-lacZ transcript levels. (A). Northern blot analysis of rny mRNA (1.72 kb) in the wild-type (SSB1002) and strains containing either the control plasmid (pC, SSB614) or plasmids conditionally overexpressing wild-type RNase Y (pY, SSB615) or an inactive version of RNase Y (pY*, SSB616). Cultures were grown with or without inducer (0.2 µg/mL tetracycline) as indicated. RNase Y levels were determined by Western blot (lower panel) (B). Northern blot analysis of short and long rny-lacZ transcripts (3.2 and 3.8 kb, respectively) in strains containing either the control (pC, SSB1082 and SSB1084) or the RNase Y conditionally overexpressing plasmid (pY, SSB1083 and SSB1085). Transcripts were detected using a lacZ-specific probe. Cultures were grown with or without inducer (0.2 µg/mL tetracycline) as indicated. RNase Y levels were determined by Western blot (lower panel) (C). Northern blot analysis of rny-lacZ transcripts in wild-type and a ∆rny background using a lacZ-specific probe. Wild-type strains (lanes 1, 3, 5, 7): SSB525, SSB530, SSB599, SSB600. Delta-rny strains (lanes 2, 4, 6, 8): SSB593, SSB594, SSB602, SSB603. All cultures were grown in LB medium. Hybridization with a 16S rRNA specific probe was used as a loading control. Numbers below the Northern and Western blots show relative densitometry readings based on the control value arbitrarily set to 100 in each gel. Measurements for Northern blots are already corrected for the 16S loading controls. NB: Northern blot. WB: Western blot.
While the long fusion mRNA became almost undetectable when overproducing RNase Y, the short fusion mRNA was also reduced two-fold suggesting that RNase Y does not only cleave within the rny ORF but also in the 5 UTR ( Figure 2B). In the absence of RNase Y ( Figure 2C), the short rny-lacZ fusion mRNA is barely increased (~1.1-fold) while induction of the long rny-lacZ fusion mRNA is upregulated by a factor of 5-fold compared to a wild-type strain, again suggesting that 5 proximal sequences in the rny ORF play an important role for controlling the rny transcript levels in the cell. The similar response of rny mRNA and the long rny-lacZ transcript to changes in the cellular level of RNase Y indicated that the regulation of the long rny-lacZ fusion by RNase Y quite accurately mimics autoregulation of the wild-type rny gene. Intermediate length rny-lacZ fusions containing 13 and 31 rny codons, respectively, showed that with increasing length regulation becomes more efficient. The 31 codon fusion was regulated to the same extent as the long fusion suggesting that cleavages within the first 100 nucleotides of the rny ORF are important for control.

RNase Y Affects the Stability of Its Own mRNA
As shown above, it is the endonucleolytic activity of RNase Y which is required for the observed auto-control of rny expression. We therefore investigated whether changes in rny transcript levels involve modulation of rny mRNA stability. The half-life of the chromosomal rny transcript was measured in presence of physiological levels of RNase Y, i.e., B. subtilis cells containing the empty control plasmid (pC), or when the nuclease was overproduced (pY). Overproduction of RNase Y from pY is induced with tetracycline but since the high amount of the nuclease expressed under these conditions lead to undetectable rny mRNA levels (confer Figure 2A) we grew cells without inducer. As already mentioned, this still leads to~two-fold excess of RNase Y in the cell. This overproduction of RNase Y caused a 3.5-fold decrease in rny half-life (1.2 min vs. 4 min) as shown in Figure 3A. We observed an equivalent four-fold effect when measuring the half-life of the long rny-lacZ fusion ( Figure 3B), but a smaller two-fold reduction in half-life for the short rny-lacZ fusion ( Figure 3C). This suggested that cleavage(s) in the 5 proximal sequences of the rny ORF make the most significant contribution to auto-control. indicated that the regulation of the long rny-lacZ fusion by RNase Y quite accurately mimics autoregulation of the wild-type rny gene. Intermediate length rny-lacZ fusions containing 13 and 31 rny codons, respectively, showed that with increasing length regulation becomes more efficient. The 31 codon fusion was regulated to the same extent as the long fusion suggesting that cleavages within the first 100 nucleotides of the rny ORF are important for control.

RNase Y Affects the Stability of Its Own mRNA
As shown above, it is the endonucleolytic activity of RNase Y which is required for the observed auto-control of rny expression. We therefore investigated whether changes in rny transcript levels involve modulation of rny mRNA stability. The half-life of the chromosomal rny transcript was measured in presence of physiological levels of RNase Y, i.e., B. subtilis cells containing the empty control plasmid (pC), or when the nuclease was overproduced (pY). Overproduction of RNase Y from pY is induced with tetracycline but since the high amount of the nuclease expressed under these conditions lead to undetectable rny mRNA levels (confer Figure 2A) we grew cells without inducer. As already mentioned, this still leads to ~two-fold excess of RNase Y in the cell. This overproduction of RNase Y caused a 3.5-fold decrease in rny half-life (1.2 min vs. 4 min) as shown in Figure  3A. We observed an equivalent four-fold effect when measuring the half-life of the long rny-lacZ fusion ( Figure 3B), but a smaller two-fold reduction in half-life for the short rny-lacZ fusion ( Figure 3C). This suggested that cleavage(s) in the 5′ proximal sequences of the rny ORF make the most significant contribution to auto-control.  SSB615). Rifampicin was added to mid-log cultures (time 0′) and RNA isolated at the indicated time points was hybridized to a probe specific for the rny 5′ UTR; this probe detects the chromosomally encoded rny mRNA and the rny-lacZ fusion mRNAs, but not the plasmid-encoded rny mRNA. The intracellular RNase Y level at time 0 was and RNA isolated at the indicated time points was hybridized to a probe specific for the rny 5 UTR; this probe detects the chromosomally encoded rny mRNA and the rny-lacZ fusion mRNAs, but not the plasmid-encoded rny mRNA. The intracellular RNase Y level at time 0 was determined by Western blot (lower panels). The position of the rny transcript on the Northern blots is indicated. (B). Half-life measurement of the long rny-lacZ fusion mRNA in wild-type (SSB530) and ∆rny (SSB594) strains. Total RNA was isolated at the indicated time points and rny-lacZ transcripts were detected with a lacZ-specific probe. (C). Half-life measurement of the short rny-lacZ fusion mRNA in wild-type (SSB525) and ∆rny (SSB593) strains. Transcripts were detected with a lacZ-specific probe. All cultures were grown in LB medium and no inducer was added for RNase Y overexpression. Hybridization with a 16S rRNA specific probe was used as a loading control. Numbers below the Northern and Western blots (Dens) show relative densitometry readings based on the control value arbitrarily set to 100 in each gel. Measurements for Northern blots are already corrected for the 16S loading controls.

RNase Y Can Cleave Efficiently in the 5 UTR of its mRNA
The half-life and the intracellular level of the short rny-lacZ mRNA was still affected in the ∆rny strain or by increased levels of RNase Y, albeit to a lesser extent than the long fusion mRNA. This implied that RNase Y might not only cleave within the rny open reading frame but also in the 5 UTR. In order to identify such cleavages, we analyzed the short fusion mRNA by primer extension in strains containing the inducible RNase Y overexpression plasmid pHMD22 (pY) grown with or without inducer or the control plasmid. The amount of full-length cDNA (238 nt) was reduced in the RNase Y overexpressing strain, especially when grown in the presence of inducer ( Figure 4A). Under these conditions, a~90 nt cDNA accumulated as would be expected if RNase Y cleavage products were trimmed by RNase J1 to the point when the 5 exonuclease encounters a ribosome bound to the translation initiation site. Since the ribosome protects about −15 to +15 nt on the mRNA with respect to the initiation codon [51], a 5 -end at position −15 would result in a cDNA of 88 nt which is in line with the detected~90 nt cDNA product.
In order to find out where RNase Y might cleave within the 5 UTR, we carried out the primer extension analysis in an RNase J1 deletion mutant where the 5 ends of cleavage products should be stabilized. Surprisingly, we found a large number of cleavages all along the 164 nt 5 UTR ( Figure 4B). In accordance, the multiple new 5 ends detected within the 5 UTR are almost absent in the wild-type strain expressing RNase J1. Instead, the major 5 end identified within the 5 UTR of the rny mRNA likely results from trimming by RNase J1 up to the position of a ribosome bound at the Shine-Dalgarno sequence as described above ( Figure 4B).
Deletion of almost the entire 5 UTR, residues −21 to −158 of the 164 nt leader, in the long rny-lacZ construct still allowed auto-control when overexpressing RNase Y, albeit at some reduced level as judged by Northern analysis (Figure 4C, left panel). However, in the absence of the 5 UTR the short rny-lacZ fusion mRNA level was completely insensitive to RNase Y overexpression ( Figure 4C, right panel), as would be expected from the absence of both regions we had identified as being sensitive to cleavage by RNase Y.
The amount of full-length cDNA (238 nt) was reduced in the RNase Y overexpressing strain, especially when grown in the presence of inducer ( Figure 4A). Under these conditions, a ~90 nt cDNA accumulated as would be expected if RNase Y cleavage products were trimmed by RNase J1 to the point when the 5′ exonuclease encounters a ribosome bound to the translation initiation site. Since the ribosome protects about −15 to +15 nt on the mRNA with respect to the initiation codon [51], a 5′-end at position −15 would result in a cDNA of 88 nt which is in line with the detected ~90 nt cDNA product. . Primer extension analysis of the short rny-lacZ fusion mRNA using a lacZ-specific primer (HP905, the 5′ end of the primer is complementary to lacZ sequences 57 nt downstream of the last rny codon) in strains harboring the empty control plasmid (pC, SSB1082) or the conditionally RNase Y overexpressing plasmid (pY, SSB1083). Cells were grown in the absence or presence of inducer (0.2 mg/mL of tetracycline). An arrow indicates the putative mRNA 5′ end generated by the 5′ exonuclease activity of RNase J1 and arrested by a ribosome bound to the Shine-Dalgarno sequence (15 nucleotides upstream of the AUG initiation codon corresponds to a cDNA of 88 nt). (B). Primer extension analysis of the short rny-lacZ fusion mRNA using a lacZ-specific primer (HP905) in a wild-type (SSB525) and RNase J1 mutant strain (∆J1, SSB1086). Extension of the primer to the original 5′ end leads to a product of 238 nt. The major 5′ end within the 5′ UTR likely generated by 5′ exonuclease trimming is indicated by an arrow. M: 25 bp ladder. (C). Effect of RNase Y overexpression on long and short rny-lacZ fusion transcripts . Primer extension analysis of the short rny-lacZ fusion mRNA using a lacZ-specific primer (HP905, the 5 end of the primer is complementary to lacZ sequences 57 nt downstream of the last rny codon) in strains harboring the empty control plasmid (pC, SSB1082) or the conditionally RNase Y overexpressing plasmid (pY, SSB1083). Cells were grown in the absence or presence of inducer (0.2 mg/mL of tetracycline). An arrow indicates the putative mRNA 5 end generated by the 5 exonuclease activity of RNase J1 and arrested by a ribosome bound to the Shine-Dalgarno sequence (15 nucleotides upstream of the AUG initiation codon corresponds to a cDNA of 88 nt). (B). Primer extension analysis of the short rny-lacZ fusion mRNA using a lacZ-specific primer (HP905) in a wild-type (SSB525) and RNase J1 mutant strain (∆J1, SSB1086). Extension of the primer to the original 5 end leads to a product of 238 nt. The major 5 end within the 5 UTR likely generated by 5 exonuclease trimming is indicated by an arrow. M: 25 bp ladder. (C). Effect of RNase Y overexpression on long and short rny-lacZ fusion transcripts that lack the rny 5 UTR. Total RNA from strains carrying the long ∆5 UTR rny-lacZ fusion and the empty control plasmid (pC, SSB638) or the RNase Y overexpression plasmid (pY, SSB639), and from strains carrying the short ∆5 UTR rny-lacZ fusion and the empty control plasmid (pC, SSB630) or the RNase Y overexpression plasmid (pY, SSB631) was hybridized with a lacZ-specific probe. Numbers below the Northern blot show relative densitometry readings of the lacZ fusion mRNA based on the control value arbitrarily set to 100 in each gel. Measurements for Northern blots are already corrected for the 16S loading controls. Asterisks mark the position of ribosomal RNA hybridizing non-specifically to the lacZ probe.

Discussion
In this study, we have shown that B. subtilis RNase Y can regulate its expression by cleaving its own mRNA. The use of rny-lacZ translational fusions was helpful as notably the long fusion encoding the 210 N-terminal amino acids of RNase Y behaved very similar when compared to the wild-type rny transcript. Auto-control is very sensitive, a 1.5-fold overexpression of RNase Y is sufficient to strongly reduce rny mRNA levels and the same rny transcripts are strongly increased in a strain lacking RNase Y.
This regulation is most likely direct and strictly requires the endonucleolytic activity of the enzyme. Even a strong overexpression of a catalytic mutant of RNase Y does not downregulate rny expression. Interestingly, rny transcript levels or rny-lacZ fusion activity are not only not reduced under these conditions but strongly induced. This is exactly the phenotype we observed in a ∆rny strain and implies that inactive RNase Y has a dominant negative effect on global RNase Y activity. Soluble forms of RNase Y lacking the N-terminal transmembrane domain are active in vitro [2] and form mainly dimers and tetramers [49]. The N-terminal non-catalytic domain of RNase Y which is intrinsically disordered contributes significantly to dimerization [50]. In vivo, RNase Y also forms dynamic higher order complexes whose assembly status can be shifted by regulatory proteins, with smaller complexes likely being the most active form of the enzyme [27]. Since RNase Y occurs, at a minimum, as a dimer, overproduction of inactive RNase Y proteins should lead to mixed forms of RNase Y oligomers where the mutated entity might also inhibit the activity of the wild-type form present in the same complex. Alternatively, but not mutually exclusive, mutated RNase Y might simply compete with the wild-type enzyme for binding to the rny mRNA and other substrates, imitating an apparent lack of RNase Y in the cell.
The use of translational rny-lacZ fusions proved to be very useful for characterizing the regions of the rny transcript susceptible to RNase Y cleavage. Early studies in B. subtilis on the role of stabilizing elements like secondary structure or a ribosome binding site at the 5 end of an mRNA had shown that the E. coli lacZ mRNA fused to these upstream elements could actually be stabilized to attain half-lives exceeding 40 min [52,53]. This suggested that in B. subtilis the heterologous lacZ mRNA is quite resistant to endonucleolytic cleavage, notably by RNase Y, an enzyme unknown at the time. Here, we could clearly show that even a strong overexpression of RNase Y cannot reduce the intracellular level of a rny-lacZ fusion mRNA when those rny regions that are susceptible to RNase Y cleavage are replaced or deleted ( Figure 4C).
RNase Y can cleave its mRNA at multiple positions to control the steady-state level of the functional transcript. Cleavages within the open reading frame would inactivate the transcript for a new translational round. We tried to identify specific cleavage sites within the rny ORF by a primer extension analysis of rny-lacZ fusions of various length in an RNase J1 mutant strain and obtained multiple but not well reproducible signals. While the long fusion containing 600 nt of rny sequences was the most susceptible to RNase Y cleavage, a significant effect of the absence of RNase Y on the level of rny-lacZ fusion mRNA was already observed with a fusion containing only 39 nt (13 codons) of rny. The presence of 93 nt (31 codons) of rny gave results equivalent to those of the long fusion. This indicates that RNase Y controls its own expression to a large part by cleaving within the first 100 nucleotides of the rny ORF. This might involve multiple cleavages as the absence of a 5 exonucleolytic activity (RNase J1 mutant) was not enough to precisely identify new 5 ends within the ORF. However, internal fragments created by more than one cleavage would also be degraded by 3 exoribonucleases complicating their detection. Translation of the rny ORF does not appear to play an important role in regulation. A long lacZ fusion construct with an AUC start codon mutation was downregulated when RNase Y was overexpressed, similarly to the wild-type fusion albeit at much lower expression levels.
A second more indirect control involves RNase Y cleavages within the rny 5 UTR. The many new 5 ends detected in the RNase J1 mutant may not all be genuine RNase Y cleavage sites. However, overexpression of RNase Y clearly destabilizes the 5 UTR producing entry sites for the 5 exonuclease J1 which trims the mRNA until encountering an initiating ribosome. It remains a surprising finding that a 3 kb mRNA molecule like the lacZ transcript appears completely resistant to cleavage by RNase Y while the 160 nt rny 5 UTR is cleaved efficiently at multiple positions. RNase Y is thought to cleave with little sequence specificity in single-stranded AU-rich sequences with a secondary structure in the up-or downstream vicinity and with a certain preference for cleavage at or close to a G residue [2,11,12,14]. We have observed that the strongest cleavages in the rny 5 UTR occur within the fifty 5 proximal nucleotides, a region composed of 76% U and A residues with a single stem-loop in the middle part. The 5 proximal sequences of the rny ORF also contain several long A stretches interrupted by a few G or C residues. However, as expected we found no clearly defined characteristics that would allow a valid prediction of potential RNase Y cleavage sites, or explain why the lacZ mRNA is not a substrate.
The β-gal activity of the rny-lacZ fusions was generally in line with the transcript levels determined by Northern blot except in one case. Induced overexpression of RNase Y caused only a minor 1.3-fold effect on the β-gal activity of the short fusion reporter while the same conditions led to a 4-fold reduction in the levels of the short fusion mRNA ( Figure 2B, confer lanes 1 to 4). This apparent difference might be linked to the length of the rny-lacZ transcript (3.2 kb). Assuming an average elongation rate of~40 nucleotides/s, it should take about 80 s to synthesize the 3.2 kb transcript. Notably under conditions of RNase Y overexpression, most nascent rny-lacZ transcripts should begin to decay before their synthesis is completed. With the short rny-lacZ fusion, RNase Y cannot directly inactivate the transcript because the rny sequences within the ORF susceptible to cleavage are absent. However, efficient cleavage within the rny 5 UTR by RNase Y still occurs. In this case, inactivation of the transcript relies on the progression of the incoming 5 exonuclease RNase J1 which can be hampered by a bound ribosome, notably at the Shine-Dalgarno sequence. In this configuration, the translation of most rny-lacZ mRNAs can probably proceed to the end abidingly, even though the levels of the full-length transcripts are constantly decreased due to shortening of the mRNA from the 5 end. Since actively translating transcripts would be less efficiently degraded from the 5 end this could explain the discrepancy between β-gal activity and full-length short rny-lacZ mRNA levels. In accordance, the 5 dependent degradation is relieved in short rny-lacZ transcripts when the 5 UTR rny sequences that are susceptible to RNase Y cleavage are not present.
We have summarized the current data in a model for the regulation of expression of the rny gene ( Figure 5A). It contrasts with that proposed for the autoregulation of RNase E in E. coli where binding of a stem-loop in the rne mRNA 5 UTR, without being cleaved, seems required for expediting cleavage elsewhere within the rne transcript [54]. discrepancy between β-gal activity and full-length short rny-lacZ mRNA levels. In accordance, the 5′ dependent degradation is relieved in short rny-lacZ transcripts when the 5′ UTR rny sequences that are susceptible to RNase Y cleavage are not present.
We have summarized the current data in a model for the regulation of expression of the rny gene ( Figure 5A). It contrasts with that proposed for the autoregulation of RNase E in E. coli where binding of a stem-loop in the rne mRNA 5′ UTR, without being cleaved, seems required for expediting cleavage elsewhere within the rne transcript [54]. . Freely available RNase Y can downregulate its own synthesis by cleaving within the ORF and the 5′ UTR. Cleavage in the ORF likely at multiple positions immediately inactivates the transcript. On the contrary, cleavage within the 5′ UTR generates 5′ monophosphorylated entry sites for the 5′ exoribonuclease activity of RNase J1. In this case, inactivation of the rny transcript depends on ribosome occupancy at the translation initiation site, hence linking the degradation of the rny mRNA to translation efficiency. (B). Potential secondary structure of the translation initiation region predicted by RNAfold [55]. The Shine-Dalgarno sequence is boxed in red.
In this context, it is noteworthy that the rny Shine-Dalgarno sequence can form 10 canonical Watson-Crick and one G•U base pair with the 3′-end of 16S rRNA, with a ∆G of −17.5 kcal mol −1 . This actually represents one of the strongest ribosomal binding sites present in B. subtilis mRNAs [56]. A strong interaction between the ribosome and the Shine-Dalgarno sequence might be necessary for the ribosome to efficiently block RNase J1 progression, but also to provide for efficient translation initiation especially since the RBS can be part of a strong secondary structure that extends up to codon 13 of the rny ORF ( Figure 5B). Similar structures are also found in the rny transcripts of closely related species. Indeed, the secondary structure itself may also decelerate RNase J1 progression and such a possibility would be compatible with the observed ~90 nt cDNA product from the primer extension experiment. On the other hand, formation of the stem-loop structure could also play a role for the susceptibility of the translated mRNA to RNase Y cleavages . Freely available RNase Y can downregulate its own synthesis by cleaving within the ORF and the 5 UTR. Cleavage in the ORF likely at multiple positions immediately inactivates the transcript. On the contrary, cleavage within the 5 UTR generates 5 monophosphorylated entry sites for the 5 exoribonuclease activity of RNase J1. In this case, inactivation of the rny transcript depends on ribosome occupancy at the translation initiation site, hence linking the degradation of the rny mRNA to translation efficiency. (B). Potential secondary structure of the translation initiation region predicted by RNAfold [55]. The Shine-Dalgarno sequence is boxed in red.
In this context, it is noteworthy that the rny Shine-Dalgarno sequence can form 10 canonical Watson-Crick and one G•U base pair with the 3 -end of 16S rRNA, with a ∆G of −17.5 kcal mol −1 . This actually represents one of the strongest ribosomal binding sites present in B. subtilis mRNAs [56]. A strong interaction between the ribosome and the Shine-Dalgarno sequence might be necessary for the ribosome to efficiently block RNase J1 progression, but also to provide for efficient translation initiation especially since the RBS can be part of a strong secondary structure that extends up to codon 13 of the rny ORF ( Figure 5B). Similar structures are also found in the rny transcripts of closely related species. Indeed, the secondary structure itself may also decelerate RNase J1 progression and such a possibility would be compatible with the observed~90 nt cDNA product from the primer extension experiment. On the other hand, formation of the stem-loop structure could also play a role for the susceptibility of the translated mRNA to RNase Y cleavages occurring in the 5 proximal coding sequences. Mutations predicted to destabilize the structure without altering the RBS had no significant effect on expression but slightly reduced the capacity of overexpressed RNase Y to downregulate expression. The protection of the rny transcript from 5 exonucleolytic degradation by ribosome binding provides a sensitive link between translation initiation and the functional stability of the RNA. A number of aspects of this regulation need to be addressed in the future, notably the role of secondary structure formation at the translation initiation site and whether other factors potentially interacting with RNase Y can interfere with autoregulation.  Data Availability Statement: Data sharing is not applicable for this article as no datasets were generated or analyzed during the current study.