Polynucleotide phosphorylase and RNA helicase CshA cooperate in Bacillus subtilis mRNA decay

ABSTRACT

Polynucleotide phosphorylase (PNPase), a 3ʹ exoribonuclease that degrades RNA in the 3ʹ-to-5ʹ direction, is the major mRNA decay activity in Bacillus subtilis. PNPase is known to be inhibited in vitro by strong RNA secondary structure, and rapid mRNA turnover in vivo is thought to require an RNA helicase activity working in conjunction with PNPase. The most abundant RNA helicase in B. subtilis is CshA. We found for three small, monocistronic mRNAs that, for some RNA sequences, PNPase processivity was unimpeded even without CshA, whereas others required CshA for efficient degradation. A novel colour screen for decay of mRNA in B. subtilis was created, using mRNA encoded by the slrA gene, which is degraded from its 3ʹ end by PNPase. A significant correlation between the predicted strength of a stem-loop structure, located in the body of the message, and PNPase processivity was observed. Northern blot analysis confirmed that PNPase processivity was greatly hindered by the internal RNA structure, and even more so in the absence of CshA. Three other B. subtilis RNA helicases did not appear to be involved in mRNA decay during vegetative growth. The results confirm the hypothesis that efficient 3ʹ exonucleolytic decay of B. subtilis RNA depends on the combined activity of PNPase and CshA.

Introduction

Bacterial mRNA turnover is an essential function and an important component of gene regulation [1]. For the majority of transcripts, which are terminated in a Rho-independent manner, exonucleolytic decay from the native 3ʹ end of the transcript is hindered by the strong secondary structure that constitutes the transcription termination signal. A model for mRNA decay that emerges from studies in E. coli and B. subtilis involves a decay-initiating endonucleolytic cleavage – by RNase E in E. coli; by RNase Y in B. subtilis – generating upstream and downstream mRNA fragments. Exonucleolytic decay of the upstream fragment proceeds in the 3ʹ-to-5ʹ direction from the 3ʹ hydroxyl end created by endonuclease cleavage. In E. coli, the downstream fragment, with a susceptible 5ʹ-monophosphate end, is a target for additional cleavage(s) by RNase E. In B. subtilis, the downstream fragment can be degraded in the 5ʹ-to-3ʹ direction by RNase J1 [2–4].

B. subtilis is known to have four 3ʹ exonucleases: polynucleotide phosphorylase (PNPase), RNase R, RNase PH, and YhaM. The major 3ʹ exonucleolytic mRNA turnover enzyme in B. subtilis appears to be PNPase [5–6]. In a strain with a deletion of the pnpA gene, encoding PNPase, 5ʹ-proximal mRNA decay intermediates accumulate, to high levels for some genes [7]. Approximately one-third of expressed genes showed a significant increase in the ratio of 5ʹ/3ʹ RNA-Seq reads in the ΔpnpA strain.

Processivity of PNPase 3ʹ exonuclease activity is inhibited by strong RNA secondary structure. In vitro RNA decay assays with purified B. subtilis PNPase showed that the enzyme is inhibited by stem-loop structures with >8 consecutive base-pairs in the stem [8,9]. Similar experiments have been performed with E. coli PNPase (which is 52% identical to B. subtilis PNPase), using RNA substrates with stems that consist solely of consecutive GC base-pairs [10]. However, such double-stranded structures are rarely found in mRNA coding sequences. Double-stranded structures containing internal loops and bulges are likely to be far more common in mRNAs, and little is known about the processivity of PNPase through such structures.

It is thought that 3ʹ exonuclease processivity on mRNA substrates is enhanced by an associated RNA helicase activity. B. subtilis contains four predicted RNA helicase genes: CshA, CshB, DeaD and YfmL [11]. Of the four helicases, CshA is the most abundant, is expressed constitutively, and its absence has the greatest effect on cell growth and morphology [12]. A pull-down experiment with a tagged version of CshA identified PNPase as an interacting partner in vivo [13]. These results suggest that CshA is the major B. subtilis helicase involved in mRNA turnover, and CshA may function in a complex with PNPase.

We wished to learn more about PNPase processivity and the requirement for RNA helicase activity, especially as these relate to decay through structures located in the body of a message. For this study, we capitalized on previous work in which we found that mRNA encoded by the slrA gene in B. subtilis is degraded primarily by PNPase [7]. We subsequently demonstrated that the accumulation of slrA mRNA in a ΔpnpA strain affected many other genes, via the regulatory role that SlrA has on expression of the fla/che operon [14]. In this report, we took advantage of PNPase-mediated slrA mRNA decay to construct a bacterial in vivo mRNA decay reporter assay. We studied the relationship of RNA secondary structure to PNPase resistance, and the contribution of RNA helicase activity to PNPase-mediated decay.

Materials and methods

mRNA decay screen

Plasmid pBL57 is a derivative of plasmid pAX01, an E. coli/B. subtilis shuttle vector that integrates in the B. subtilis lacA locus [15]. The plasmid specifies ampicillin resistance in E. coli and macrolide-lincosamide-streptogramin B (MLS) resistance in B. subtilis. The slrA gene was amplified by PCR, using oligonucleotides DHB1761 and DHB1763 (Table S3), directed upstream and downstream of the slrA transcription unit, and was cloned between the SacI and SacII sites of pAX01, replacing the xylR gene. In the course of this work, we discovered that a 209-base-pair (bp) duplication on pAX01 was a site for intramolecular recombination that resulted in a high incidence of plasmid deletion. To resolve this issue, we replaced the 1670-bp NotI fragment of pAX01, which contains the MLS resistance gene and one of the repeats, with a 1290-bp PCR fragment (using oligonucleotides DHB1803 and DHB1804) containing the MLS resistance gene from plasmid pMUTIN4 [16], yielding plasmid pBL57.

Mutagenesis of nine nucleotides (nts) in the variable sequence and cloning in pBL57 was done by Gibson cloning [17], using the mutagenic primer DHB1808 and oligonucleotides 1809–1811, and mutagenized plasmid DNA was used to transform DH5α cells. Transformed cells were selected on ampicillin-containing plates, and hundreds of colonies were pooled and grown up for plasmid DNA isolation. Plasmid DNA was linearized with restriction endonuclease PvuI and used to transform BG1110 (Table S2), with selection for lincomycin resistance on S-gal (Sigma-Aldrich) plates. In addition to the hag-lacZ fusion and slrA deletion mentioned in the results section, BG1110 is also deleted for the rok gene, which serves to increase the number of transformants [18]. We found that S-gal media interfered with the rather stringent selection of 12.5 μg/ml lincomycin, resulting in a hazy background of growth on which true colonies were clearly visible. We attributed this to the high iron concentration present in S-gal, which contains 500 μg/ml ferric ammonium sulphate [19], that likely increased the MIC for lincomycin [20]. From many hundreds of colonies obtained, 38 that were lighter in colour than the wild-type control were selected, and the sequence of the mutagenized region was determined (Table S1).

Mean grey measurements and phase microscopy

Overnight cultures were grown in Luria-Bertani (LB) medium with selection for chloramphenicol (4 μg/ml). Aliquots were diluted 10⁻⁵ and 10⁻⁶ for plating on S-gal solid media containing the above concentrations of erythromycin and lincomycin. Plates were incubated overnight at 37°C and photographed with a Nikon D7200 camera, using a Nikon 105 mm F2.8 lens at a distance of 2 feet and magnification of 1:4. Photos were shot at 1/200 of a second at f6.3 and ISO 100. An 18% grey card was used to normalize colour temperature and tint shift.

For phase microscopy of ΔcshA strains, cells were grown in LB broth to 0.8 OD₆₀₀. The culture was vortexed, and 10 µl was mounted on a microscope slide. Phase-contrast microscopy was performed with a Leica DMi8 widefield inverted fluorescence microscope, equipped with HC PL Fluotar 20x/0.40 Ph1 Corr lenses. An exposure time of 20 ms was set for the Leica DFC450C camera connected to the microscope.

RNA isolation

For steady-state analysis of rpsO, rapA, and cggR RNA, an overnight culture of BG1 was grown at 37°C in LB medium. The overnight culture was diluted 1:40 and grown to mid-exponential phase (80 Klett units). Total RNA was prepared by the hot phenol method [21].

For steady-state slrA analysis, bacitracin-inducible constructs were used. The bacitracin-inducible lia operon promoter [22] was amplified by overlap PCR (oligonucleotides DHB1882/1151) with a wild-type or mutated slrA transcription unit (oligonucleotides DHB1883/1884), and was cloned into the amyE integration vector, pDR66, between EcoRI and HindIII sites [23]. The slrA-containing pDR66 derivatives were used to transform a strain (BG973; Table S2) that was deleted for the native slrA gene. Overnight cultures were grown at 37°C in LB medium, in the presence of 4 μg/ml chloramphenicol. 250 μl of the overnight culture were used to inoculate 10 ml of LB without antibiotic selection. When cultures reached mid-exponential phase (80 Klett units), freshly prepared bacitracin solution (Sigma-Aldrich) was added to 50 µg/ml for induction of slrA expression. 9 ml of culture was harvested after 20 minutes of bacitracin induction, and total RNA was prepared using the RNeasy Mini Kit (Qiagen), according to the manufacturer’s instructions.

For mRNA half-life measurements, 35 ml cultures were grown and induced with bacitracin, as above. After bacitracin induction, freshly prepared rifampicin dissolved in 100% methanol was added to 150 µg/ml, to inhibit transcription. 8 ml aliquots were taken at indicated time points after rifampicin addition, and total RNA was extracted by the hot phenol method.

Northern blot analysis

Total RNA (6 or 10 µg) was separated on a 6% denaturing polyacrylamide gel as described previously. The gel was electroblotted to a BioBond-Plus nylon membrane (Sigma-Aldrich) and electroblotted overnight at 20 V and then one hour at 30 V. 5ʹ-end-labelled oligonucleotide probes were prepared using T4 polynucleotide kinase (New England Biolabs) and [γ-³²P] ATP (Perkin Elmer). Probe sequence and location of complementary targets are listed in Table S3. The membrane was blocked for 1 hour at 68°C in QuikHyb solution (Agilent Technologies) containing sheared salmon-sperm DNA (100 μg/ml) and hybridized overnight at 42°C. The membrane was washed for 20 min in 2X saline-sodium citrate (SSC), 0.1% sodium dodecyl sulphate (SDS), followed by three washes for 30 min each in 0.1X SSC, 0.1% SDS. Membranes that were probed for 5S rRNA were washed a fourth time. Washes were done at 58°C, except for veg and 5S rRNA blots, which were done at 55°C. Membranes were stripped and reprobed as described previously [24].

Radioactivity in RNA bands was quantitated using a Typhoon Phosphorimager (Typhoon FLA 9500/7000, GE Healthcare). The amount of full-length and decay intermediate RNA was normalized to veg mRNA for each strain, and this value, in turn, was normalized to the amount of veg-normalized full-length RNA in the strain containing construct 12 (wild-type 38-nt sequence). The relative RNA abundance from three independent experiments was used to calculate the mean ± SD for each construct.

The loading control for the steady-state measurements in Fig. 4A was veg mRNA, rather than the customary 16S or 5S rRNA. For 6% denaturing polyacrylamide gels, we found it difficult to quantitate 16S rRNA, which migrates near the top of the gel. Since RNA for this experiment was isolated by the Qiagen RNeasy kit, small RNAs, including 5S rRNA, were largely excluded. veg mRNA (~300 nts) is constitutively expressed during vegetative growth [25], and it migrates the appropriate distance for this type of gel. However, we observed consistently that veg mRNA is present at higher levels in the cshA mutant than the wild type. This necessitated adjusting RNA abundance separately within each genetic background (wild-type or cshA mutant), as explained above.

Figure 4.

Northern blot analysis of five selected clones. (A) Steady-state slrA RNA in wild-type (left) and cshA mutant (right) strains. Above each lane is the number of the slrA construct. Migration of prominent slrA RNAs indicated on the right: FL, full-length RNA; diamond, 520-nt decay intermediate; DI, decay intermediate with a 3ʹ end at the downstream edge of the 38-nt variable sequence. The blot was stripped and reprobed with the veg probe. Marker lane contained RNA Century-Plus markers, with the indicated RNA sizes in nucleotides. (B) Quantitation of results of three independent Northern blot experiments for steady-state slrA RNA. (C) Northern blot analysis of slrA RNA decay in strains with constructs 12 and 14, after addition of rifampicin at time zero. Aliquots were taken after rifampicin addition at times (min) indicated above each lane. Marker lane contained RNA Century-Plus markers, with the indicated RNA sizes in nucleotides. (D) Half-life determination of full-length slrA RNA in strain with construct 12 (circles, solid line), full-length slrA RNA in strain with construct 14 (squares, dashed line) and decay intermediate from construct 14 (triangles, dashed line). (E) Northern blot analysis of slrA RNA in single and triple helicase mutants. Above each lane is the number of the slrA construct. Migration of RNA Century Plus marker RNAs indicated on the left

For RNA half-life measurements, the percent of slrA RNA remaining was normalized to the 5S rRNA loading control. Experimental RNA half-lives were determined by an exponential regression analysis of percent RNA remaining vs. time. R² values of the full-length data were > 0.97. Since the decay intermediate is simultaneously decaying and being generated from decay of full-length RNA slrA, the corrected half-life of the decay intermediate was calculated with the Bateman equation for radioactive decay, as described [26], using the Solver feature of Microsoft Excel. Corrected values are plotted in Fig. 4D.

Western blot analysis

The strain with a FLAG-tagged PNPase was constructed using the pMiniMAD protocol [27]. The C-terminal half of the pnpA coding sequence and an additional 1000 base-pairs downstream were cloned into pMiniMAD. The FLAG-tag sequence was added to the end of the coding sequence using the QuikChange II Site-Directed Mutagenesis protocol (Agilent). Transformation and growth of the wild-type strain to replace the native pnpA gene with the FLAG-tagged version was as described [27]. The same method was used to make the FLAG-tagged CshA construct in the context of the PNPase-FLAG strain. Total protein isolation and Western blotting was as described [28], with anti-FLAG M2 antibody (Sigma-Aldrich) used at 1:5000 dilution.

Results

Decay intermediates in pnpA and cshA deletion strains

The loss of PNPase activity affects the decay of many mRNAs, including three small mRNAs encoded by the rpsO, rapA and cggR genes [7,29]. Decay intermediates of these mRNAs, which are undetectable in the wild-type strain, accumulate in the ΔpnpA strain. We wished to determine the effect of the loss of helicase activity on decay-intermediate accumulation. We hypothesized that CshA is the major B. subtilis helicase involved in mRNA turnover, based on its abundance relative to the other three helicases and its constitutive expression through all phases of growth [13]. Total RNA was isolated from mid-exponential phase cultures of the wild-type strain and of strains that were deleted for either the pnpA gene, the cshA gene, or both. The results of Northern blot analyses using oligonucleotide probes complementary to 5ʹ-proximal mRNA sequences are shown in Fig. 1. For rpsO (Fig. 1A), no decay intermediates were detected in either the wild type strain or the ΔcshA strain, while the prominent 180-nt decay intermediate [29,30] and other RNAs accumulated in the ΔpnpA strain and to a somewhat higher degree in the ΔpnpA ΔcshA double mutant strain. The rpsO results suggested that CshA was not required for efficient mRNA decay mediated by PNPase. Results with rapA and cggR mRNAs (Fig. 1B,Fig. 1C), on the other hand, led to a different conclusion. Unlike rpsO mRNA, which codes for a ribosomal protein, rapA and cggR mRNAs code for regulatory proteins (control of sporulation initiation and transcriptional regulator of gapA operon, respectively), and, as such, are expected to be present in much lower abundance. While the full-length rapA and cggR transcripts were barely visible on these exposures, decay intermediates were easily detected not only in the pnpA mutant strain, as discovered previously, but also in the cshA mutant strain (arrows in Fig. 1B,Fig. 1C). These results indicate that both PNPase and CshA were required for efficient turnover of these RNA fragments, and demonstrate directly that helicase activity is needed in the turnover of native mRNA decay intermediates.

Figure 1.

(A-C) Northern blot analysis of rpsO (A), rapA (B), and cggR (C) RNA in wild-type (lanes 1), pnpA (lanes 2), cshA (lanes 3), and pnpA cshA (lanes 4) strains. Migration of full-length RNA is indicated by ‘FL.’ Marker lane in (A) contained RNA Century-Plus marker (Ambion), with the indicated RNA sizes in nucleotides. Values to the left in (B) and (C) indicate migration of 5ʹ-end-labelled fragments (sizes in nucleotides) of a TaqI digest of plasmid pSE420 [49]. Arrows point to RNA decay intermediates that accumulate in the cshA mutant strain. (D) Determination of relative amounts of FLAG-tagged PNPase and CshA. The four lanes are from independent protein isolations, grown under identical growth conditions

Relative levels of PNPase and CshA

PNPase enzyme functions as a homotrimer, with the subunits forming a central channel into which single-stranded RNA is fed [31,32]. If an association with CshA helicase is required for efficient PNPase-mediated mRNA decay, one would expect that the cellular concentration of CshA would be either similar to, or in excess of, that of PNPase, enabling PNPase/CshA interaction for all or most PNPase homotrimeric complexes. Previously-published data on the cellular levels of these proteins in B. subtilis indicated that PNPase ranked in the top 10% of all proteins, whereas CshA ranked either in the top 10% or the top 25% [33]. In addition, the regulation of expression (as measured by transcript level) of the pnpA and cshA genes under all growth conditions tested is almost identical [34]. To measure the relative levels of these proteins directly, we constructed a strain in which the native PNPase and CshA genes were replaced with C-terminal FLAG-tagged versions. We assumed that the FLAG-tagged versions were expressed at the same level as the native genes. Western blot analysis of the strain carrying the FLAG-tagged PNPase and CshA showed that there were roughly equal amounts of these proteins in the cell (Fig. 1D); the average CshA:PNPase ratio was 0.9 ± 0.1. Since CshA is believed to function as a monomer, as is the case for RhlB of E. coli [35], there is approximately a 3-fold excess of CshA monomer over PNPase trimer. Thus, CshA would be readily available to function with PNPase in mRNA decay.

Elements of the in vivo mRNA decay screen

As mentioned in the Introduction, one of the known targets of PNPase is slrA mRNA. Turnover of slrA mRNA occurs by PNPase degradation from the native slrA mRNA 3ʹ end, which is susceptible to PNPase activity because transcription termination occurs in a Rho-dependent manner [14]. Although the slrA 3ʹ end is predicted to form an extended secondary structure, this is not a hindrance to PNPase-mediated decay, as evidenced by the < 1.0 min half-life of full-length slrA mRNA [14]. SlrA protein is a negative regulator of the large, 32-gene fla/che operon, whose penultimate gene is sigD, the gene for transcription sigma factor D [36]. In the absence of PNPase, there is an accumulation of slrA mRNA and its decay intermediates that include the coding sequence (CDS), and presumably an increase in SlrA protein. The elevated level of SlrA protein results in decreased sigD expression and, therefore, decreased expression of the sigD regulon, which includes the hag gene [37]. Our previous RNA-Seq data showed a 4-fold decrease in sigD reads in the ΔpnpA strain, which has increased slrA expression [7]. Thus, it was possible to use a P_hag-lacZ fusion gene to report on the level of SlrA protein, which, in turn, would be affected by the efficiency of slrA mRNA decay (Fig. 2A). The assay strain consisted of (1) a P_hag-lacZ fusion integrated at the amyE locus [38], (2) a deletion of the native slrA gene [36], and (3) an slrA gene construct integrated in the lacA locus. (Disruption of the lacA locus was important to reduce background β-galactosidase activity.)

Figure 2.

(A) Diagram of the mRNA decay screen. The top line is a schematic of the slrA transcription unit, showing the CDS and 5ʹ and 3ʹ UTRs (not drawn to scale). The structured 38-nt variable sequence is shown schematically, as well as the 3ʹ-proximal predicted secondary structure in a region that is putatively required for Rho-dependent termination. The characteristics of a Rho-dependent terminator in B. subtilis are still not well-defined [50]. (B and C) Structure of the 38-nt variable sequence in constructs 12 and 14, and colony colour on S-gal plates in the wild-type background. (D) Correlation of mean grey value with predicted strength of secondary structure for 40 clones in the wild-type background

The native slrA CDS is small (156 base-pairs), but the slrA transcript contains fairly long 5ʹ and 3ʹ untranslated regions (UTRs) (Fig. 2A), the functions of which are unknown [14]. The slrA construct used in this study (pBL57) contained the slrA CDS with a ~ 170 base-pair 5ʹ UTR and a ~ 340 base-pair 3ʹ UTR, cloned into pAX01 [15], a lacA integration vector that encodes macrolide-lincosamide-streptogramin B (MLS) resistance. A 38-nt sequence (‘variable sequence’ in Fig. 2A), located from 63–101 nts downstream of the slrA CDS, was chosen as a test sequence for PNPase resistance. This sequence is far enough downstream of the CDS that it would not be subject to ribosome transit, thus eliminating the complicating effect of translation on mRNA decay [39]. The 38-nt sequence is also far enough upstream of the slrA mRNA 3’ end such that alteration of this sequence would likely not affect Rho-dependent transcription termination. The 38-nt sequence has a predicted secondary structure that contains two internal loops and has a ΔG₀ value of −9.3 kcal/mol (Fig. 2B). As mentioned above, slrA mRNA is rapidly degraded in the wild-type strain, and we did not previously observe a processing intermediate whose 3ʹ end would map to the downstream edge of this predicted secondary structure.

The level of β-galactosidase expression in a strain carrying the wild-type slrA construct (Fig. 2B; BG1112) was assayed qualitatively on plates containing the S-gal indicator (3,4-Cyclohexenoesculetin β-D-galactopyranoside; Sigma-Aldrich), which is more sensitive than X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside) for detection of β-galactosidase expression. As shown in Fig. 2B, BG1112 formed colonies that were black at the centre, with lighter colouring as the colony extended to its outer rim. This suggested a low level of slrA mRNA due to rapid mRNA decay, as expected, and, consequently, a high level of expression from the P_hag-lacZ reporter gene. A control construct was made in which the 38-nt sequence was predicted to have a stable lower stem (‘DS stem’), with an overall predicted ΔG₀ value of −18.8 kcal/mol (Fig. 2C; BG1114). BG1114 colonies, which contain the DS stem version of the 38-nt sequence, showed no colour (Fig. 2C). This suggested an accumulation of slrA RNA capable of expressing SlrA protein, and, therefore, a low level of P_hag-lacZ expression. We hypothesized that the accumulation of slrA RNA was due to resistance to PNPase-mediated decay conferred by the stable 38-nt sequence. For the remainder of this report, strains with the wild-type construct will be designated as strain 12, and strains with the DS stem construct will be designated as strain 14.

Screening for susceptibility/resistance to PNPase activity

In the context of plasmid pBL57 (described above), 9 nts on the left side of the stem, shown in italics in Fig. 2B, were subjected to random mutagenesis. Mutagenized plasmid DNA was used to transform B. subtilis to MLS resistance, with selection on S-gal plates for lincomycin resistance. Many hundreds of transformants were screened for colour intensity, relative to the controls shown in Fig. 2B,Fig. 2C. As expected, most of these transformants had the same colour intensity as BG1112, indicating that the mutagenized 38-nt in these colonies did not provide protection against PNPase processivity. However, a number of transformants were lighter in colour, and 38 of these clones were selected for further analysis. The mutagenized 9-nt sequence was determined for each clone, and these strains were grown overnight, diluted, and plated on S-gal plates. Table S1 contains data on mean grey value (higher number indicates lighter colony colour) for these 38 clones, as well as the 12 and 14 control strains. These data are presented graphically in Fig. 2D, from which we conclude that there was a statistically significant Spearman correlation between predicted secondary structure and mean grey value data (r = – 0.702; p < 0.0001).

Northern blot analysis of slrA RNA in selected clones

We next used Northern blotting to analyse directly the slrA mRNA from five selected clones. These included the wild-type slrA construct (strain 12, Fig. 2B), the construct with the DS stem (strain 14, Fig. 2C), and three more clones with constructs (designated 16, 18, and 21) that had intermediate mean grey values (Table S1) and colour phenotypes on S-gal plates (Fig. 3A). The predicted secondary structures of constructs 16, 18, and 21 are shown in Fig. S1. Since we observed a significant effect on growth of mutant strains containing slrA constructs that expressed stable mRNA (see below), we constructed a version of each slrA construct that had transcription under the control of a bacitracin-inducible promoter [22]. As we found earlier, full induction of transcription was achieved by adding bacitracin to 50 μg/ml, 20 minutes prior to RNA isolation [30]. For this report, all Northern blot analyses were performed in strains containing the bacitracin-inducible constructs, but for simplicity sake we refer to them with the same designations as above (i.e., 12, 16, 18, 21, and 14).

Figure 3.

Phenotypes of five clones selected for Northern blot analysis, containing the indicated slrA constructs. (A) Colony colour of wild-type strains. Predicted ΔG₀ (kcal/mol) of the 38-nt sequence and mean grey (MG) value of the colony are shown. (B) Colony colour of cshA deletion mutant strains. (C) Overnight cultures of cshA deletion mutant strains. (D) Phase microscopy (20X) of c shA deletion mutant strains. Bar, 10 µM

RNA was isolated from the five strains containing different 38-nt sequences by the RNeasy method (Qiagen). Northern blot analysis of slrA RNA in the wild-type background is shown in Fig. 4A, left side. The full-length slrA transcript is about ~670 nts, and this band is observed in all strains at about the same intensity (labelled ‘FL’ in Fig. 4A). An mRNA decay intermediate that is due to inhibition of PNPase 3ʹ-to-5ʹ processivity at the 38-nt sequence is predicted to be 420 nts (labelled ‘DI’ in Fig. 4A). Such a band is not detectable in strain 12, which contains the wild-type 38-nt sequence that we know from colony phenotype does not hinder PNPase-mediated decay (Fig. 2B). On the other hand, the 420-nt decay intermediate is detectable in strains 16, 18, and 21, and to a much greater degree in strain 14 (Fig. 4A,Fig. 4B). That this band is due to inhibition of 3ʹ-to-5ʹ processivity rather than transcription termination is evident from the fact that there is little or no change in the amount of full-length RNA as the amount of decay intermediate increases. There is an additional band at about 520 nts, which we assume is due to inhibition of 3ʹ-to-5ʹ decay – or transcription termination – by a predicted secondary structure (ΔG₀ = −17.7 kcal/mol) that begins 40 nts downstream of the 38-nt sequence (Fig. S2). This intermediate is present in all strains at about the same level. (For notes about the veg mRNA loading control, see Materials and Methods.)

Determination of slrA RNA half-life

The hypothesis from the data presented so far was that turnover of slrA mRNA can be inhibited by the presence of strong, internal secondary structure. As such, we expected this to be reflected in the relative half-life of the decay intermediate. This was shown directly by Northern blot analysis of a strain containing construct 12 (wild-type 38-nt sequence) compared to a strain containing construct 14 (DS stem sequence). Native slrA mRNA has a short half-life of less than a minute [14]. RNA was isolated by the hot phenol method after addition of rifampicin to stop new transcription, and the amount of slrA RNA remaining after 1, 2, and 3 minutes was determined by Northern blotting (Fig. 4C). While the half-life of full-length slrA mRNA was virtually the same for these two strains (0.66 and 0.70 min), the half-life of the decay intermediate in strain 14 was about five-fold longer than the full-length transcript (3.56 min). These results confirm that the increased level of the decay intermediate with a strongly structured 38-nt sequence is due to a block to processive decay, presumably catalysed by PNPase, resulting in greatly increased RNA half-life.

Resistance to PNPase processivity in the cshA knockout strain

A derivative of the strain used for the in vivo PNPase sensitivity assay was constructed in which the gene coding for CshA helicase was replaced with a neomycin-resistance gene, as described [12]. Growth curves of the cshA wild-type and cshA knockout strains showed that deficiency of this single helicase had a significant effect on growth (Fig. S3), as observed previously [12]. The cshA mutant grew considerably slower than even the pnpA mutant strain. Chromosomal DNA from 27 constructs (Table S1) was used to transform the ΔcshA strain to MLS resistance, and strains were plated on S-gal. Mean grey value for the cshA mutant clones are shown in Table S1, and the colour phenotypes for the five selected slrA constructs are shown in Fig. 3B. While the wild-type slrA sequence was still susceptible to PNPase decay, as evidenced by the dark-coloured colonies, the four strains in which the 38-nt structure was more strongly structured showed no colour, indicating a higher level of SlrA due to PNPase resistance. The slower growth of the cshA mutant apparently affected the colour phenotype (compare strains 12 in the two backgrounds in Fig. 3A), such that the mean grey values in the wild-type and cshA mutant backgrounds cannot be compared. Nevertheless, among clones in the cshA mutant background, there was little correlation between predicted strength of secondary structure and mean grey value (Spearman correlation coefficient = – 0.369; p = 0.058).

We observed previously that B. subtilis strains containing an elevated level of SlrA protein (e.g., a PNPase mutant or an slrA overexpression mutant) grow in long chains [14]. This is due to decreased expression of autolysis genes, which are under the control of SigD-dependent promoters [7]. In overnight cultures of the ΔpnpA strain, which has a ~ 2-fold increase in slrA mRNA, tiny clumps of cells can be observed. Similarly, the level of SlrA was clearly reflected in overnight cultures of the cshA mutant strain (Fig. 3C). While strain 12 grew normally, the other four strains, with a more structured 38-nt sequence, showed increasing amounts of clumping. The strain 14 overnight culture had a tight ball of growth in an otherwise clear medium.

To confirm that the clumping was due to chain growth, phase microscopy of the ΔcshA strains grown to 0.8 OD₆₀₀ was performed (Fig. 3D). Strain 12 (wild-type slrA) showed primarily single or dividing cells and a few short chains. The short chains were likely due to the increased amount of full-length slrA mRNA in the ΔcshA background (Fig. 4A, lanes 12). Strain 16 showed some single and dividing cells, but mostly medium-length chains. Strains 18 and 21 showed growth in long chains. Strain 14 clumping was so intense that it was not possible to separate the clump for microscopy. Presumably, the degree of clumping or chaining in these strains correlated with the level of slrA RNA accumulation.

Northern blot analysis of slrA RNA in cshA mutant strains

To determine directly the effect of RNA secondary structure on PNPase processivity in the absence of CshA, RNA was isolated after bacitracin induction from ΔcshA strains containing the five slrA constructs examined above. The right side of Fig. 4A showed that there was a considerable increase (~10-fold; Fig. 4B) in the amount full-length slrA mRNA for all five constructs. This is likely due to the inability of PNPase alone to degrade past the secondary structure that is near the native 3ʹ end. Although this structure does not have the classic elements of a Rho-independent terminator, one can predict a larger structure (encompassing 56 nts) that has a ΔG₀ of −18 kcal/mol [14]. The 520-nt RNA was also present in higher amounts in the cshA mutant strains. The 420-nt decay intermediate was not detected for the wild-type 38-nt sequence (Fig. 4A, right side, lane 12), indicating that PNPase can proceed efficiently through this sequence even in the absence of CshA. However, constructs with greater predicted strength of the 38-nt sequence had a graded increase in the amount of 420-nt decay intermediate RNA (Fig. 4A,Fig. 4B). Although there was a considerable amount of this RNA in strain 14 even in the wild-type background (Fig. 4B, left), there was almost 3-fold more of this RNA when CshA was absent (Fig. 4B, right).

Other RNA helicases are not involved in slrA mRNA decay

As mentioned in the Introduction, the B. subtilis genome is predicted to contain four DEAD box RNA helicases – CshA, CshB, DeaD, and YfmL. We tested whether the three RNA helicases other than CshA were required for efficient slrA mRNA decay. A ΔcshB ΔdeaD ΔyfmL strain was obtained from J. Stuelke [12]. The native slrA gene was deleted from this strain, and the five slrA constructs were moved into the lacA locus. RNA was isolated after bacitracin induction and probed with an slrA probe. It is apparent from a comparison of the Northern blots shown in Fig. 4A,Fig. 4E that the pattern of slrA RNA in the triple helicase mutant was the same as in the wild-type. Thus, we can conclude that only CshA is required for efficient, PNPase-mediated RNA decay.

Discussion

The current study was designed to address the effect of internal RNA sequences/structures on PNPase processivity in B. subtilis, and the contribution of RNA helicase activity to PNPase-mediated RNA decay. The role of RNA helicase in RNA turnover has been studied primarily in E. coli. E. coli contains six known RNA helicases, of which RhlB is the major one involved in mRNA turnover under vegetative growth conditions [40]. RhlB helicase is required for efficient 3ʹ exonucleolytic decay of repetitive extragenic palindromic (REP) sequences [41] and from a 3ʹ end past a transcription terminator sequence [42]. REP and transcription terminator sequences are predicted to form extremely stable RNA secondary structures located at transcript 3ʹ ends, with multiple, successive Watson-Crick base pairs. A parallel functions is likely not important in B. subtilis, which has no REP sequences [43] and whose 3ʹ-terminal fragments we have shown are likely degraded primarily by the 5ʹ exonuclease, RNase J1 [44]. The effect of an RNA helicase on RNA decay was also reported for a decay intermediate of a lacZ message transcribed by T7 RNA polymerase in E. coli [45]. The transcriptome-wide effect of major helicase deletion mutants has been reported for several organisms, including E. coli [46], B. subtilis [12], and Staphylococcus aureus [47]. A microarray analysis of the B. subtilis ΔcshA strain detected almost equal numbers of transcripts that decreased and increased in the absence of CshA [12]. This observation likely reflects indirect transcriptional effects, as an increase (rather than a decrease) in transcript abundance would be expected in a strain that is deleted for an RNA decay-related activity.

Here, we showed that certain decay intermediates accumulate in otherwise isogenic strains that are missing either PNPase or CshA (Fig. 1). We believe this is the first observation of accumulation in a helicase mutant of native mRNA decay intermediates, in the absence of special stabilizing sequences. Although we have not mapped their precise 3ʹ ends, the similar size of these decay intermediates in both the ΔpnpA and the ΔcshA strains suggest that the combined action of both enzymes is required for efficient decay. Interestingly, the most prominent rpsO 180-nt decay intermediate that is easily detected in the absence of PNPase [29,48], and whose 3ʹ end was mapped to the base of a predicted RNA secondary structure [29], is undetectable when PNPase is present but CshA is absent (Fig. 1A). On the other hand, decay intermediates of other mRNAs are clearly evident not only when PNPase alone is absent but also when CshA alone is absent (Fig. 1B,Fig. 1C). These results suggest that the putative collaboration of PNPase and CshA in the efficient decay of mRNA is not a simple matter of RNA structure requiring unwinding by the helicase for PNPase to proceed. We also observed an apparent increase in the amount of some RNA decay intermediates in the ΔpnpA ΔcshA mutant (Fig. 1, lanes 2 and 4). We have not yet quantitated this effect.

In the case of slrA mRNA, we took advantage of the fact that transcription termination is Rho-dependent, leaving a 3ʹ end at which mRNA decay could initiate, and providing a test for the processivity of PNPase through the upstream 38-nt variable sequence. To screen a pool of randomly mutagenized versions of the 38-nt sequence, we created a novel β-galactosidase reporter assay (Fig. 2A). We are not aware of a previous report of a bacterial in vivo mRNA decay screening tool. In this screen, the overwhelming majority of colonies showed the same colour phenotype as the strain with the wild-type 38-nt sequence, indicating that changes in this sequence rarely led to a block to PNPase processivity. For 40 clones that showed a range of colony colour, there was a significant correlation between the predicted strength of secondary structure and susceptibility to 3ʹ-to-5ʹ decay (Fig. 2D). However, RNA secondary structure was clearly not the sole determinant of resistance to PNPase.

We found that the DS stem (construct 14) conferred a five-fold increase in half-life on the decay intermediate (Fig. 4D). This stabilization could account entirely for the difference between the measured mean steady-state amount of strain 14 full-length RNA (2.00 ± 1.27) and decay-intermediate RNA (12.49 ± 5.27), in the wild-type background (Fig. 4B, left, lanes 14). Because the decay-intermediate RNA includes the entire slrA coding sequence, in principle a strain carrying the DS stem would contain at least five times as much SlrA protein as the wild type, explaining the stark difference in colony colour between strains 12 and 14 (Fig. 3A). The extreme chaining phenotype of strain 14 in the ΔcshA background (Fig. 3C,Fig. 3D) is a reminder of the significance of efficient mRNA decay in maintaining normal cellular function.

Growth curve data (Fig. S3) showed that the cshA mutant grew even slower than the pnpA mutant. It is expected that the absence of a major enzymatic activity involved in RNA turnover would affect cellular physiology significantly, either directly or through indirect transcriptional effects. The lesser effect on growth of a PNPase deletion could be explained by the presence of other 3ʹ exoribonuclease activities that compensate in mRNA turnover, although there is currently no evidence that the three known 3ʹ exoribonucleases (RNase PH, RNase R, and YhaM) are involved in general mRNA decay. In the case of RNA helicase activity, our result indicating that RNA helicases other than CshA are not important for PNPase-mediated slrA mRNA turnover (Fig. 4E) suggests that CshA alone functions in mRNA decay with PNPase, at least during vegetative growth. Future work will involve transcriptome analysis to address the nature of RNA sequences/structures that require CshA to allow PNPase-mediated decay.

Funding Statement

This work was supported by the National Institutes of Health [1R01GM125655].

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Supplementary material

Supplemental data for this article can be accessed here.