sRNA-dependent control of curli biosynthesis in Escherichia coli: McaS directs endonucleolytic cleavage of csgD mRNA.

Production of curli, extracellular protein structures important for Escherichia coli biofilm formation, is governed by a highly complex regulatory mechanism that integrates multiple environmental signals through the involvement of numerous proteins and small non-coding RNAs (sRNAs). No less than seven sRNAs (McaS, RprA, GcvB, RydC, RybB, OmrA and OmrB) are known to repress the expression of the curli activator CsgD. Many of the sRNAs repress CsgD production by binding to the csgD mRNA at sites far upstream of the ribosomal binding site. The precise mechanism behind sRNA-mediated regulation of CsgD synthesis is largely unknown. In this study, we identify a conserved A/U-rich region in the csgD mRNA 5' untranslated region, which is cleaved upon binding of the small RNAs, McaS, RprA or GcvB, to sites located more than 30 nucleotides downstream. Mutational analysis shows that the A/U-rich region as well as an adjacent stem-loop structure are required for McaS-stimulated degradation, also serving as a binding platform for the RNA chaperone Hfq. Prevention of McaS-activated cleavage completely relieves repression, suggesting that endoribonucleolytic cleavage of csgD mRNA is the primary regulatory effect exerted by McaS. Moreover, we find that McaS-mediated degradation of the csgD 5' untranslated region requires RNase E.

INTRODUCTION

Biofilms are communities of microorganisms that aggregate and adhere to a surface (usually solid–liquid) encased in a self-produced matrix consisting of extracellular polymeric substances (EPS). This EPS is composed of polysaccharides, proteins, lipids and DNA (1). The sessile growth phenotype is entirely different from the independent single-cell growth mode of planktonic bacteria. The biofilm structure provides mechanical stability and protects the bacterial cells from various environmental conditions such as UV radiation, dehydration, and provides strong resilience against antibiotics (2,3). Biofilms provide an important bacterial reservoir that may serve as nidus for recolonization of infected hosts. This feature is believed to be the leading cause of recalcitrant chronic bacterial infections (4).

The main proteinaceous component of the extracellular matrix produced by Escherichia coli biofilms is curli (5). Curli are amyloid fibers that mediate surface adhesion and cell aggregation of E. coli and Salmonella spp. species and have been implicated in both intestinal and extraintestinal infections (6,7). The curli fibers are composed of the main structural protein, CsgA, and the nucleator protein CsgB, which are encoded by the csgBAC operon (8,9). CsgC acts as a periplasmic chaperone that prevents CsgA from forming toxic intracellular aggregates (10). A second operon, csgDEFG, encodes proteins required for the fiber assembly and secretion (CsgE, CsgF and CsgG) as well as a positive transcriptional regulator of curli synthesis CsgD, which in turn activates transcription of the csgBAC operon (11). The regulation of curli biosynthesis is extraordinarily complex. Multiple regulatory cascades process and integrate signals into the curli synthesis. The master regulator, CsgD, is crucially controlled by the stationary sigma-factor (RpoS) of RNA polymerase (12). Furthermore, the bacterial signalling molecule bis-(3′–5′)-cyclic di-GMP (c-di-GMP) is also required for activating CsgD expression. C-di-GMP is produced by diguanylate cyclease enzymes (DGCs) and the molecule is degraded by phosphodiesterase enzymes (PDEs). C-di-GMP is an important biofilm messenger in the motile-to-adhesive ‘lifestyle’ switch of E. coli (13,14).

In addition to the RpoS/c-di-GMP cascades leading to curli expression, at least three two-component systems regulate the expression of csgD. OmpR/EnvZ responds to changes in osmolarity and is essential for csgD expression. CpxA/R is a cell envelope stress response pathway responding to misfolded proteins repressing csgD expression. Similarly, the Rcs phosphorelay system is an envelope stress-controlled pathway that negatively affects the expression of csgD (15–20).

Besides the abundant transcriptional control of the curli operons, the csgD transcript serves as a hub for post transcriptional regulation by small non-coding RNAs (sRNAs). With the help from Hfq no less than seven sRNAs negatively influence the expression of csgD transcription—these are OmrA, OmrB, McaS, RprA, GcvB, RydC and RybB (21–26). The most common mechanism by which sRNAs regulate protein synthesis is by specific base pairing with sequences overlapping with the target mRNA ribosomal binding site (RBS), thereby interfering with translation initiation (27). Interestingly, the csgD mRNA contains an unusually long 5′ untranslated region (UTR) of 147 nucleotides (nt) and many of the sRNAs bind far upstream of the RBS. The exact mechanism by which many of these sRNAs inhibit CsgD synthesis and thereby curli production is not yet fully understood.

OmrA and OmrB are two redundant sRNAs located adjacent to each other on the chromosome. Whereas their 5′ and 3′ ends contain identical sequences, the central regions differ, indicating that they have both common and distinct targets (28). Both OmrA and OmrB bind csgD mRNA on the 5′ side of a conserved stem–loop structure, and base-pairing to nucleotides in a bulge is essential for regulation. In vitro, binding of OmrA and OmrB to csgD mRNA inhibits formation of the translation initiation complex in a manner independent of the RBS itself (21).

RprA binds csgD mRNA in three different regions: two distal sites far from the RBS and one overlapping with the Shine-Dalgarno (SD) sequence. Mutations in either of these sites do not alleviate csgD repression completely, whereas double mutants do (23). RybB, a major RpoE dependent sRNA known to downregulate multiple outer membrane porins, has two binding regions in the csgD UTR where one site overlaps with the SD sequence (26). Additionally, RydC has base pair complementarity to the SD region of csgD (25).

The precise mechanism by which the small RNAs McaS and GcvB regulate CsgD production is unknown. Both have csgD mRNA binding sites upstream of the RBS. Moreover, an in vitro structural probing assay suggests that McaS has extended base-paring with csgD mRNA. This extended base-paring allows McaS binding in proximity of the RBS and thereby directly prevents assembly of the ribosomal initiation complex (22).

Here, we present new insight into the mechanism by which curli-repressing sRNAs inhibit CsgD biosynthesis and thus biofilm formation. We show that McaS, and to some extent GcvB and RprA, induce endonucleolytic cleavage in a conserved A/U-rich region. This region is important for both binding of the RNA chaperone Hfq and induced degradation of csgD mRNA. In addition, we show that McaS-induced cleavage dependent on RNase E.

MATERIALS AND METHODS

Strains, plasmids and oligonucleotides

Bacterial strains and plasmids used in this study are listed in supplementary dataSupplementary Table S1 and S2; also their construction is described in supplementary data. Primers used for the construction of mutant strains, plasmids and DNA templates are listed in Supplementary Table S3.

Growth conditions

Cells of E. coli K12 were grown in standard Luria–Bertani (LB) broth at 37°C. When required, the medium was supplemented with antibiotics: ampicillin (30 μg/ml), chloramphenicol (30 μg/ml). The expression of curli was visualized on Congo red agar plates without the addition of NaCl (Bacto-Tryptone 10 mg/ml, yeast extract 5 mg/ml, Congo red 40 mg/ml, Coomassie Brilliant Blue G 10 mg/ml). The plates were incubated at 28°C for 48 h. Expression from the P promoter was induced by addition of 1 mM IPTG and the expression from the P promoter was induced by addition of 0.2% arabinose.

Transmission electron microscopy

To visualise extracellular curli fibers, samples were placed on 200-mesh carbon coated grids for 1 min. The samples were washed with deionized water and negatively stained with 3% (w/v) uranyl acetate for 30 s. Images were acquired using a JEM-1200EX II (Joel Germany) microscope at an acceleration of 120 kV.

RNA purification

Total RNA was extracted from 10 ml culture. Each sample was spun, and pellet was re-suspended in 150 μl solution 1 (10 mM Na-citrate, 10 mM Na-acetate pH 4.5 and 2 mM EDTA), transferred to a phase-lock tube from 5 PRIME containing 150 μl solution 2 (10 mM Na-acetate pH 4.5 and 2% SDS), 300 μl chloroform and 700 μl acidic phenol pH 4.5. The tube was heated at 80°C for 5 min and spun at 14 000 rcf for 5 min to separate the phases. The aqueous phase was transferred to 1 ml 96% ethanol to precipitate the RNA. The RNA pellet was resuspended in RNase-free H2O.

Northern blot experiments

Ten microgram total RNA was supplemented with 2 μl formamide loading buffer and separated on 6% polyacrylamide gels for 1.5–2.5 h at 300 V. Separated RNA was semi-dry blotted onto a Zeta-probe nylon membrane (Bio Rad) for 1 h at 400 mA, followed by cross-linking of RNA to the membrane with UV radiation. Probes listed in supplementary data were 5′-labeled with 32P-ATP using T4-polynucleotide kinase (NEB). Membranes were pre-hybridized for 10 min at 42°C before probing with 5′-labeled oligos overnight (ON). Probed membranes were washed three times in 2× SCC and 0.1% SDS for 10 min at 42°C, dried and visualized by phosphorimaging on a Typhoon Trio scanner (GE Healthcare).

Western blot experiments

Samples were spun and cell pellets were re-suspended in 1× SDS loading buffer (3% SDS, 10% glycerol, 50 mM Tris–HCl pH 6.8, 0.1% bromophenol blue, 12.5 mM EDTA, 100 mM DTT) to a concentration of approximately 106 cells/μl and boiled at 95°C for 5 min. 10 μl of each sample was separated in 4–12% NuPAGE® Bis-Tris gels (Invitrogen) at 190 V for 45 min, and blotted onto a nitrocellulose membrane (Invitrogen) for 1 h at 150 mA. Blocking and antibody addition was done using the SNAP i.d.® 2.0 system (Merck Millipore). α-FLAG (Sigma-Aldrich) and α-GroEL (Sigma-Aldrich) monoclonal antibodies were diluted 1:20 000 and 1:50 000, respectively. Mouse and rabbit horseradish peroxidase-conjugated secondary antibodies were diluted 1:2000 (Dako Cytomation).

Primer extension analysis

Primer extension analysis was carried out using the PXT primer listed in supplementary data. The primer was 5′-labeled with 32P-ATP using T4-polynucleotide kinase (NEB). Labeled primer was hybridized to 10 μg RNA, and extended with reverse transcriptase (Promega). Sanger sequencing was carried out with labeled PXT primer, PCR amplified csgD promoter and UTR as template and one of the ddNTPs (G, A, T, C) mixed with all four dNTPs in four different PCR reactions. The labeled cDNA and sequence of csgD were separated in 8% polyacrylamide gels, which were dried and visualized by phosphorimaging on a Typhoon Trio scanner (GE Healthcare).

In vitro synthesis and labeling of RNA

csgD RNAs were prepared by in vitro transcription (Megascript, Ambion) using 5 ug of DNA from a purified PCR product as template using the T7 oligos listed in Supplementary Table S3. The RNA transcripts were DNase I-treated before PAGE separation on a 4.5% 7 M urea denaturing gel. The RNA was extracted by electro elution using GeBAflex tubes (GerardBiotech). RNA transcripts were 5′-labeled with 32P-ATP using T4-polynucleotide kinase (NEB).

Electrophoretic mobility shift assays

The E. coli Hfq protein was expressed and purified using the Impact System (New England Biolabs) as described in (29).

Binding reactions were carried out in 10 μl volumes in 1× binding buffer (20 mM Tris, pH 8, 100 mM KCl, 1 mM MgCl2, 1 mM DTT) and incubated at 37°C for 10 min before addition of 5 μl of loading buffer (50% glycerol). Hfq-csgD mRNA interactions were carried out in 1× binding buffer using 2 nM 5′ end-labeled csgD mRNA, 10 ug tRNA in the presence of 0, 0.25, 0.5, 0.75, 1 or 2 μM Hfq (monomer concentration). Samples were separated by native electophoresis on 5% polyacrylamide gels at 4°C in 0.5× TBE. Samples were visualized by phosphoimaging on a Typhoon Trio scanner (GE Healthcare). The relative intensity of the shifted bands was quantified and fitted to a sigmoid curve, and dissociation constant (Kd) values were determined with SigmaPlot.

Structural probing assay

2 nM of 5′ end-labeled csgD 5′ UTR RNA was mixed with increasing concentrations of Hfq (0, 0.25, 0.5 and 2 μM), McaS (10, 20, 40 nM) or both (0, 0.25, 0.5, 2 μM Hfq mixed with 10, 20, 40, 80 nM McaS) in 1× RNA Structure Buffer (Ambion) and incubated at 37°C for 10 min. Samples were cooled at room temperature before addition of 2.5 nM lead(II)acetate or 1 U of Nuclease S1 (Promega). After incubation for 5 min at 37°C, 10 μl of the samples were transferred to 10 μl of cold formamide loading dye. RNase T1 digestions were carried out in 1× Sequence Buffer with 0.1 U of RNase T1 (Ambion) for 5 min at 37°C before transferring it to cold formamide loading dye. The samples were run at 55 W on 8% denaturing PAGE gel along with an alkaline hydrolysis ladder of the 5′ end-labeled csgD 5′ UTR RNA.

RESULTS

Induction of McaS inhibits the synthesis of curli amyloid fibers

We previously showed that ectopical co-expression of McaS with csgD mRNA rapidly downregulates the cellular levels of csgD mRNA. Furthermore, expression of McaS in E. coli K-12 wild type-cells showed a white morphotype when streaked on Congo red agar plates, indicative of a curli deficient phenotype (22). To see the direct effect of McaS on curli fiber biosynthesis we inspected a wild-type strain carrying the mcaS gene under control of an IPTG-inducible lac promoter on a low-copy-number R1 plasmid (pNDM-mcaS) by high-resolution transmission electron microscopy (TEM). When grown under curli inducing conditions, ectopic production of McaS in wild-type cells resulted in a complete loss of curli fibers and visibly naked cells (Figure 1). Moreover, these curli deficient cells showed an increase in flagella production. In contrast, wild-type cells harboring the empty vector pNDM220 produced curli fibers and only few visible flagella. These findings are consistent with the observation that McaS production leads to increased motility by activating FlhD the master regulator of flagella synthesis (24).

Figure 1.

McaS producing cells lack extracellular curli fibers. Close inspection of the inhibitory effect of McaS on curli fibers. Wild-type cells of SØ928ΔmcaS producing McaS from a P containing low-copy plasmid (pNDM-mcaS) and SØ928ΔmcaS cells carrying the empty vector (pNDM220) were grown in curli stimulating conditions on LB-agar plates (without NaCl) at 28°C. High-resolution EM images were taken after 48 hours of growth in the presence of 100 μM IPTG inducer.

The secondary structure of 5′ UTR of csgD mRNA is conserved among bacterial species

To identify regions important for sRNA-mediated regulation, we aligned the 5′ untranslated region (UTR) of csgD including the promoter regions from bacterial species encoding the sRNA McaS (22). The csgD promoter and +1 start site (identified with primer extension, Figure 8), Shine-Dalgarno sequence and the target sequence for McaS binding are conserved (Figure 2A). The alignment further revealed that three predicted stem–loop elements (I, II and III) were preserved, suggesting that the mRNA 5′ UTR structure is conserved too. This is supported by in vitro structural probing of the E. coli csgD mRNA (Figures 2B and 7) (21,25). Other than stem–loop I, II and III, the 5′end of the csgD UTR was not predicted to take up any consensus secondary structure (Figure 2C). Interestingly, however, this region contained a conserved A/U-rich region immediately upstream of stem–loop I (Figure 2A).

Figure 8.

McaS mediated csgD mRNA cleavage is RNase E dependent. Western blot of CsgDFLAG and primer extension analysis of csgD 5′end. Cultures of N3431 (rne-3071), the isogenic N3433 WT strain, containing the OmpR234 point mutation and a single or double knock-outs of rng, the gene encoding RNase G. The strains carry either the empty vector control pBAD33 or pBAD-mcaS. All cultures were grown in LB medium at 30°C in the presence of inducer (0.2% arabinose) to an OD600 of 1.2. At this point, the cultures were shifted to 42°C for 20 min before sampling for proteins and RNA. GroEL was probed for protein loading control and 5S, showing characteristic unprocessed bands in the rne-3071 allele, was probed as RNA loading control. Fold-change are of CsgDFLAG protein levels from three experiments.

Figure 2.

An A/U-rich stretch followed by a small stem–loop structure is conserved in the csgD mRNA. csgD from seven different species, all having a conserved mcaS gene, was aligned using Clustal Omega Multiple Sequence Alignment tool. (A) The promoter elements (−35 and −10) and transcriptional start site (+1) are highly conserved, as well as the McaS-binding site (dark gray), the RBS and stem–loop II. Furthermore, both an A/U-rich region (light gray) and a C and a G rich stretch (C-side and G-side, respectively) just downstream the A/U-rich region are conserved. Arrows indicate possible stem–loop structures (numbered I, II and III), bold letters indicate nucleotide residues engaging in the base-pairing of the stem–loops, and underlined lower-case letters indicate the translation start codon. (B) The E. coli csgD 5′UTR used in this study adopts an overall similar fold as the consensus folding shown in C. The E. coli structure is based on the structural probing presented in Figure 7 and supplementary data. The two McaS-binding sites are highlighted (C) RNAalifold was used to predict a generic consensus secondary structure of the csgD alignment obtained from Clustal Omega. The consensus sequence predicts csgD to contain three stem–loop structures: the first just downstream of the A/U-rich region (I), a second central large stem–loop structure upstream of the RBS containing a bulge (II), and a third stem–loop structure containing the RBS (III).

Figure 7.

Hfq interacts with the csgD 5′UTR at multiple positions. (A) Structural probing assays of csgD 5′ UTR RNA using Nuclease S1 with increasing concentrations of Hfq, McaS or both. Samples containing 5′end-labeled transcripts of 2 nM csgD 5′UTR RNA were pre-incubated at 37°C with Hfq, McaS or both for 10 min before incubation with Nuclease S1 for 5 min. An alkaline hydrolysis ladder (‘OH’) and a G-ladder (‘T1’, RNase T1 digestion) was used to determine nucleotide position. Untreated csgD 5′ UTR RNA was used as a control (‘C’). (B) Protected Hfq and McaS binding sites are underlined in the csgD 5′ UTR structure.

Identification of sRNA-stimulated cleavage sites in csgD mRNA

Commonly, trans-encoded sRNAs induce degradation of their target mRNAs, and McaS co-degrades with csgD (22,30). We decided to identify the exact csgD 5′ UTR cleavage sites during in vivo induction of the regulatory sRNAs McaS, RprA, GcvB and OmrA by primer extension analysis. We constructed a minimal target molecule consisting of the csgD 5′ UTR plus 87 nucleotides of the coding region fused to the small RNA ChiX for stabilization of the 3′ end. This short target, termed csgD, allowed for better separation of cleavage products in polyacrylamide gels. Expression of sRNAs were induced from P by addition of IPTG from a low-copy number mini R1 plasmid (pNDM220), and the target csgD RNA was expressed from a pBAD33 derivative by addition of arabinose. To reduce interference from the chromosomally encoded sRNAs, these experiments were performed in a strain deleted for the sRNAs McaS, RprA, GcvB, OmrA, and OmrB. For simplicity we call this strain SØ928Δ5. McaS strongly induced RNA degradation while OmrA did not stimulate degradation. RprA and GcvB also stimulated cleavage but to a lesser extent than McaS. The most apparent cleavage sites were identified between residues G-132 and U-131, and U-131 and U-130, which correspond to the 5′ end of the conserved A/U-rich region (Figure 3A and B). Several other less prominent cleavage sites were identified, all located within the A/U-stretch (A-127 to U-123, A-116 to U-114 and U-102 to A-101), except for two cleavage sites induced specifically by GcvB (A-93 to A-91) (Figure 3A). Thus, McaS, RprA and GcvB appear, at least in part, to regulate csgD expression by a common mechanism – one that is different from that of OmrA and OmrB.

Figure 3.

McaS, RprA, and GcvB stimulate csgD mRNA cleavage in the conserved A/U-rich region. (A) Primer extension and northern blot with RNA purified from SØ928Δ5 carrying pBAD-csgDchiX and either pNDM-mcaS, pNDM-rprA, pNDM-gcvB, pNDM-omrA or the empty vector pNDM220. Cultures were grown in LB broth at 37° to an OD600 of 0.6 (−/−). The cultures were induced with 1 mM IPTG for 10 min (+/−), followed by 5 min induction with 0.2% arabinose (+/+). Primer extension and northern blot was performed as described in methods. (B) csgD mRNA sequence with cleavage sites indicated. The most apparent cleavage sites () are marked with black arrows and the remaining () with grey arrows. is specific to McaS, while is specific to GcvB. McaS, RprA and GcvB all induce cleavage in the conserved A/U-rich region (gray bar).

Only the distally located McaS-binding site within the csgD 5′UTR is necessary in vivo

Previously, we identified two McaS-binding sites on csgD 5′UTR. Mutational analysis and structural probning experiments showed that McaS directly interacts with csgD 5′UTR at two sites (22). The first site (Binding Site 1) is located far upstream the AUG start codon and the second site (Binding Site 2) is located close to, and overlapping with, the ribosomal binding site (Figure 4A). Moreover, a toeprinting assay revealed that tRNAfMet-dependent 30S complex at the csgD TIR (translation initiation region) was lost upon McaS pre-incubation in vitro. This suggested that McaS binding to Binding Site 2 on csgD inhibits translation initiation. Thus, two possibilities might explain the regulatory mechanism by which McaS regulates CsgD synthesis. One mechanism relies on the conventional pattern where the sRNA binds to the translational initiation region of the target mRNA and prevents translation initiation. The other mechanism results in coupled degradation of sRNA and the target mRNA to control CsgD expression.

Figure 4.

Only the upstream binding site is necessary for McaS-mediated regulation of CsgD protein levels. (A) Schematic presentation of the csgD 5′UTR indicating the two sites of csgD complementary to McaS (underlined). Deletion mutants in binding sites 1 and 2 respectively are highlighted by black circles. (B) E. coli strain SØ928 carrying either a wild-type csgDFLAG construct or mutant alleles on a pBAD plasmid vector. Expression from the empty vector control pNDM220 (−) or the isogenic plasmid borne McaS (+) was induced by addition of 1 mM IPTG. After 10 min of incubation, 0,2% arabinose was added to induced expression of the csgDgenes. After 5 min of induction samples for western blot analysis was taken. GroEL was probed as loading control.

We decided to test the importance of the two McaS base-paring sites on CsgD production in vivo. We examined the ability of McaS to downregulate the expression of two mutant csgD alleles by tagging the very end of the csgD gene with a sequence encoding a triple FLAG epitope. The mutant alleles each carried a deletion, which interrupted the McaS binding site with Binding Site 1 or Binding Site 2, respectively (Figure 4A). We transiently induced expression of McaS from pNDM220 for 10 min followed by short-term expression of csgD and mutant alleles. Western analysis showed that McaS regulated the CsgDFLAG expression with a Binding Site 2 truncation as well as wild-type csgD, whereas disruption of Binding Site1 completely disrupted regulation, indicating that Binding Site 2 plays little role in McaS dependent CsgD regulation (Figure 4B).

McaS induces csgD mRNA cleavage at the conserved A/U-rich region

To investigate the importance of sequence and structure motifs for the action of McaS, we constructed a set of five csgD mutants and evaluated target RNA stability by northern blotting using a ChiX-specific probe (Figure 5A). The csgD probe binds the 3′ end of the csgD-chiX fusion and thus only the full length and 3′end degradation products of csgD are able to hybridize to the probe. In a parallel set of experiments, we monitored CsgDFLAG protein levels (Figure 5C). We performed transient expression assays where McaS was turned on by addition of IPTG for 10 min followed by short-term induction of csgD by addition of arabinose for 5 min. As observed for full-length csgD 5′UTR, short-term induction of McaS resulted in clear csgD degradation products and a corresponding decrease in CsgD protein levels (Figure 5B and C, lanes 2 and 3). Deletion of the entire conserved A/U-rich region (ΔA/U) completely prevented cleavage at this site and made the target RNA less responsive to McaS-mediated degradation and consequently, western analysis of CsgD protein levels showed that downregulation by McaS was lost (Figure 5B and C, lanes 4 and 5). Mutations in the conserved A/U-rich region by substitution of four nucleotides with four guanines (GGGG), changed the target RNA degradation pattern and decreased translational inhibition, but to a lesser extent than ΔA/U (Figure 5B and C, lanes 6 and 7). Deletion of stem–loop I (Δstem) inhibited McaS-dependent cleavage in the conserved A/U-rich region and decreased its translational inhibition, however to a lesser degree than ΔA/U (Figure 5B and C, lanes 8 and 9). The two remaining mutants were created in order to force an alternative secondary structure within the csgD 5′ UTR (Forced). In both mutants, the 3′ half of the conserved A/U-rich region and the 5′ half of the conserved stem–loop was replaced with a stable tetraloop to induce formation of a large stem masking the cleavage sites. Further substitution of a single uracil for cytosin in the 5′ end of the conserved A/U-rich region (Forced+C) completely prevented cleavage at that site and greatly decreased McaS-dependent translational inhibition (Figure 5B and C, lanes 10 and 11). Mutation of two adjacent uracils located on the descending strand of the stem just upstream of the sRNA binding site (corresponding to cleavage site in Figure 3) into guanines, inhibited cleavage in the conserved A/U-rich region and decreased McaS-dependent translational inhibition greatly (Figure 5B and C, lanes 12 and 13). Taken together, the Northern- and Western blots demonstrate that prevention of cleavage in the conserved A/U-rich region alleviates McaS-dependent repression of CsgD biosynthesis, suggesting that McaS induces csgD mRNA cleavage at the conserved A/U-rich region thereby inhibiting translation. Furthermore, we conclude that the conserved A/U-rich region itself as well as the local secondary structure are important for efficient McaS repression of CsgD expression in vivo.

Figure 5.

The A/U-rich region and secondary structure is important for efficient mRNA cleavage and translational inhibition. (A) Graphical representation of wild-type (WT) and mutant csgD mRNAs, shown with consensus secondary structure (Figure 2) indicating the conserved A/U-rich region (grey line), nucleotide substitutions (asterisk) and cleavage site and from Figure 3 (arrows). (B) Northern blot with RNA purified from SØ928Δ5 carrying either wild-type or mutant pBAD-csgDchiX derivatives (this plasmid contains the csgD 5′UTR and the first 87 nucleotides of the coding region and the small RNA ChiX for stabilisation of the construct) and either pNDM-mcaS or the empty vector pNDM220. Cultures were grown in LB media at 37°C to an OD600 of 0.6 at which point a sample was taken from the culture with wild type csgD and empty pNDM220 vector as a negative control. The cultures were induced with 1 mM IPTG for 10 min, followed by 5 min induction with 0.2% arabinose at which point samples were taken from all cultures. Northern blot was performed as described in methods. (C) Northern and Western blot with RNA and proteins purified from SØ928Δ5 carrying either wild type or mutant pBAD-csgDFLAG and either pNDM-mcaS or the empty vector pNDM220. Cultures were grown and samples were taken as for panel B and Western and northern blots were performed as described in methods. The CsgD-FLAG levels were quantified from three experiments and normalized to the loading control (GroEL). The efficiency of McaS-mediated repression was determined by dividing the signal from samples without McaS by the signal from samples with McaS. Efficiency of McaS on wild type csgD was defined as 1.

csgD mRNA degradation is Hfq-dependent

The function of most sRNAs in E. coli is greatly affected by Hfq, which binds and stabilises sRNA molecules and mediates their base-pairing with cognate target mRNAs (29). Hfq further exhibits RNA chaperone activity affecting RNA secondary structure (31). Hfq recognizes single-stranded A/U-rich or ARNn motifs (where A is adenine, R is a purine, and N is any nucleotide) sequences, which are often preceded or followed by stem–loop structures (32,33). It interacts directly with RNase E and PNPase and has therefore been suggested to recruit RNases to cleave the sRNA–mRNA complexes (34,35).

We used electrophoretic mobility shift assays (EMSA) to investigate the interaction between Hfq and the csgD 5′UTR in vitro. To examine the interaction in detail, we measured the dissociation constant (Kd) to describe the affinity between Hfq and csgD. Deletion of the conserved A/U-rich region as well as forcing an alternative secondary structure (ΔA/U, Forced+C and Forced+GG described above) reduced the affinity between Hfq and csgD mRNA molecules (Figure 6). Deletion of the stem–loop I structure (Δstem) did not impact complex formation and substitution of the conserved A/U-rich region with four guanine nucleotides (GGGG) slightly reduced complex formation. Taken together with the results from Figure 5, McaS-stimulated cleavage in the conserved A/U-rich region and the efficiency of McaS-dependent translational inhibition of CsgD does not correlate directly with changes in Hfq binding affinities. However, CsgD synthesis inhibition is at least in part dependent on Hfq binding to a region close to or at the conserved A/U-rich region. This is consistent with Hfq involvement in the recruitment of the RNA degradation machinery to the cleavage sites at mRNA substrates (30). Interestingly, at high Hfq concentrations, we see higher order multimers of csgD and Hfq complexes suggesting that Hfq interacts with csgD at multiple sites.

Figure 6.

Hfq binds the csgD 5′UTR constructs with different affinities assisted by the adjacent conserved stem–loop. Electrophoretic mobility shift assays of Hfq binding to csgD mRNA. Samples containing 5′ end-labeled transcripts of 2 nM csgD wild type or csgD mutant mRNAs (Figure 4A) were incubated with increasing amounts of Hfq. Final monomeric concentrations of Hfq were 0, 0.25, 0.5, 0.75, 1 and 2 μM from left to right.

Another explanation for the induction of cleavage in the A/U-rich sequence could be a conformational change within csgD upon McaS and Hfq binding (36). To pinpoint the exact Hfq-binding sites on csgD 5′UTR we performed structural probing experiments using Nuclease S1, which specifically degrades single-stranded RNA (or DNA). This simultaneously allowed Hfq and McaS binding site identifications as well as probing the overall structure of the csgD 5′UTR. Figure 7 shows the probing of csgD 5′UTR alone and with increasing concentrations of Hfq, McaS, or both. The probing was visualized on different acrylamide gels to improve the resolution of both the very 5′end and the far upstream located translation initiation region (Figure 7A). We observed several Hfq-binding sites within the csgD 5′ UTR (Figure 7B). We identified one site at the A/U rich sequence in the 5′end extending to include the AUUUA hairpin (Site I). This site is located at nucleotides –128 to –112 between the cleavage region and the McaS base-paring region. This site is close to, but not overlapping, the strong cleavage sites identified in Figure 3. Moreover, Hfq cooperates in complex formation and facilitate csgD-McaS interaction (Figure 7, lanes 11–14). In addition, we observed two additional sites located close to the translational initiation region, Site II is positioned in the single-stranded A-tract region preceding the Shine-Dalgarno sequence, and Site III is located right after the AUG start codon in the open reading frame (Figure 7B). Indeed, Hfq binding Site I and II were recently suggested as functional Hfq binding sites (33). The protection observed with Nuclease S1 at Site II and Site III were weak, so we decided to validate and confirm these sites with Pb2+-probing (Supplementary data Figure S4). We also detected protection in a bulge in the conserved stem–loop structure (*). However, we note that the protection is weak and only present at high Hfq concentrations, and the protected sequence does not constitute the typical Hfq recognition sequences (i.e. an A/U tract or an ARN motif) (33). In addition, the Nuclease S1 probing suggested that Hfq binds to the hairpin of the conserved stem–loop (**), however, we could not confirm this site by Pb2+-probing. Furthermore, binding of Hfq and McaS to the csgD 5′UTR did not result in an overall structural rearrangement (Figure 7).

RNase E is required for McaS-dependent degradation of csgD

McaS, RprA and GcvB all stimulate endoribonucleolytic cleavage of csgD mRNA. To further our understanding of the sRNA dependent cleavage of csgD mRNA we examined several ribonucleases as specific mediators of McaS-csgD decay. Several sRNA-mRNA complex responsive ribonucleases exist in E. coli including RNase III and RNase E (37). In addition, a recently identified RNase, YbeY, was considered as a candidate for sRNA-stimulated csgD mRNA cleavage since it modulates Hfq dependent and independent sRNA-mRNA interactions (38).

RNase III exists as a dimer and cuts specifically at double-stranded RNA, which allows further processing of the resulting fragments (39,40). RNase III and YbeY are not essential in E. coli, and McaS-stimulated regulation of curli synthesis could therefore be assayed in the corresponding mutant backgrounds on agar plates containing Congo red, which binds to the amyloid curli fibers to produce a red color. RNase III- and YbeY-deficient strains were cultured alongside their respective isogenic strains on Congo red plates under McaS-inducing conditions. In all strains McaS stimulated down-regulation of curli production (Supplementary Figure S2). The ybeY mutant is characterised by a slow growth phenotype and poor red colouring on Congo red agar plates. We therefore validated the plate assay with western blot analysis (Supplementary Figure S3).

RNase E is essential in E. coli, and therefore a primer extension assay was set up using strain N3431, which carries the rne-3071 allele, a temperature-sensitive mutation of RNase E, and its isogenic wild type strain N3433 (41). Since pNDM220 exhibits greatly elevated copy numbers due to run-away replication at 42°C (42) (the non-permissive temperature for RNase E in N3431), the arabinose inducible pBAD33 vector derivative was used for sRNA induction, while the chromosomal csgD transcript level was assayed. Natural levels of csgD proved insufficient for primer extension (data not shown) and therefore a hyperactive OmpR mutant with a point mutation, OmpR234 (43), was introduced to N3431 and N3433 to increase transcription of csgD from the chromosome (Supplementary Figure S3). We also deleted RNAse G, encoded by rng, in this strain background for several reasons: First, RNase G is a paralog of RNase E with a close catalytic domain resemblance; Second, RNase E is not essential in a strain overproducing RNase G indicating that they have redundant targets; Third, McaS co-purifies with RNase G in a recent immunoprecipitation study (40,44,45). Finally, we introduced the triple FLAG epitope tag sequence to the 3′end of chromosomally encoded csgD to monitor protein levels as well as transcript abundance.

The strains were grown to late-log phase in the presence of McaS inducer (0.2% arabinose) to keep csgD mRNA levels to a minimum. The cultures were shifted to the non-permissive temperature of RNase E for 20 min and the samples were assayed for CsgD protein and transcript accumulation (Figure 8). A WT strain and the corresponding rne-3071, rng, and the double mutants harbouring the empty pBAD33 control vector all showed a clear 5′end csgD product and a corresponding clear CsgDFLAG protein band when shifted to 42°C (Figure 8, lanes 1–4). However, csgD degradation products disappeared in the rne mutant allele (Figure 8, lanes 2 and 4). These degradation products presumably stem from the chromosomally encoded sRNAs McaS, RprA and GcvB (Figure 3). The co-expression of McaS from P showed significant downregulation of csgD 5′end mRNA and a dramatic decrease in protein levels (Figure 8, lanes 5–8). Interestingly, CsgDFLAG protein levels accumulated only in strains lacking RNase E (Figure 8, lanes 6 and 8). Consistently, these strains also showed elevated 5′end transcript levels. Taken together we conclude that McaS dependent inhibition of CsgD synthesis depend on RNase E. Furthermore, RNase E cuts at A/U rich single stranded regions and we observe that the degradation products are located in the single stranded A/U rich region of csgD 5′UTR. Finally, we note that RNase G does not contribute to McaS mediated degradation of csgD (Figure 8, lanes 7–8 and Supplementary Figure S2).

DISCUSSION

The switch from a motile single cell to a sessile lifestyle in a biofilm is a highly regulated process and a prominent adaptation programme. This lifestyle decision must be strictly coordinated, because once the decision to form a biofilm is made, the individual cells cannot easily reverse to single planktonic cells again. It is therefore not a surprise that the master regulator of curli biosynthesis, CsgD, is controlled at many levels. Regulation takes place at both the transcriptional and post-transcriptional level by a cohort of regulatory proteins and small non-coding RNAs. The csgD transcript has an unusually long 5′ UTR, which is largely conserved. No less than seven sRNAs bind the csgD 5′UTR to regulate the expression of curli biosynthesis. Each sRNA relays a signal to the genetic network thereby cross-linking individual genetic regulons and adaptation strategies. Consistent with this concept, many of the sRNAs show largely non-redundant expression patterns (46).

We sought to dissect the mechanism by which several sRNAs downregulate the expression of CsgD. Our results demonstrate that at least two different mechanisms for CsgD repression are used by the sRNAs; one prevents ribosomes from translating CsgD, either through direct binding to the RBS or via structural rearrangement of a preceding stem–loop (21,25), and the second mechanism induces endoribonucleolytic cleavage of the csgD mRNA in a conserved A/U-rich region located in the 5′ UTR.

Short transient ectopic expression of OmrA does not result in any visible csgD degradation products. According to our data, the OmrA-mediated regulation of curli biosynthesis is therefore not dependent on rapid cleavage of the csgD transcript. This is consistent with previous studies, showing that OmrA and OmrB bind the csgD mRNA in vitro and ectopic expression of the sRNAs regulate CsgD production in vivo. The two small RNAs bind upstream the translational binding site (Figure 9A) to inhibit the access of initiating ribosomes (21).

Figure 9.

Model for sRNA-mediated regulation of csgD. (A) Binding/cleavage sites of the sRNAs, Hfq and RNase E on the csgD mRNA. McaS, RprA, GcvB and OmRA/B all have extensive base pairing located distally from the ribosomal binding site. While OmrA and OmrB cause translational inhibition, McaS, RprA and GcvB induce ribonucleolytic cleavage of the mRNA. The most significant cleavage site is located near the 5′-end, just upstream of an Hfq binding site. Mutations of this Hfq site inhibits RNase E cleavage and alleviates McaS induced repression of CsgD expression. RydC and RybB bind csgD at the ribosomal binding site to inhibit the formation of the translational initiation complex. (B) Model for the two sRNA mediated mechanisms for posttranscriptional regulation of csgD. Left) McaS (RprA and GcvB) is first bound and stabilized by Hfq. The sRNA-Hfq binary complex interacts with RNase E of the degradosome and its substrate mRNA through base-pairing between the RNAs as well as protein-RNA and protein-protein interactions, in which Hfq binds to Site I of csgD. Complex formation leads to ribonucleolytic cleavage of csgD and McaS. Right) RydC (omrA/B and rybB) is bound by Hfq. This sRNA-Hfq complex will bind near the ribosomal binding site by base-pairing between the two RNAs as well as Hfq interactions with csgD at site II and site III. Formation of this complex leads to translational inhibition.

Previously, we showed that GcvB downregulates the expression of csgD. Here, we discovered a specific degradation pattern of csgD mRNA following short-term GcvB expression (Figure 3A). The degradation pattern resembles that caused by McaS, except for one cleavage product that is characteristic to GcvB. It is presently not known whether the regulatory effect of GcvB is exerted primarily through translation inhibition or transcript degradation. Other studies show that GcvB prevents ribosomal binding on target mRNAs (47). However, based on the almost identical csgD mRNA breakdown products caused by McaS and GcvB induction, we speculate that GcvB may affect CsgD expression through a similar mechanism as McaS. However, the precise mechanism behind GcvB-mediated control of curli biosynthesis and the origin of the specific degradation product, remains to be determined (Figure 3A).

In contrast, RprA-mediated repression has been investigated previously. Hengge suggested in 2012 that two RprA binding sites of csgD are involved in a partially redundant manner. This was based on the observation that a single deletion mutation in the common sRNA binding site (shared with McaS) in the proximity of the RBS did not relieve RprA-mediated repression of CsgD biosynthesis downregulation, while double mutations did (23). However, we have previously shown, in a different experimental setup, that a deletion mutation in the common sRNA binding site completely relieved RprA-mediated repression (22). The study conducted by Hengge used constitutively expressed csgD and RprA from low-copy and high-copy number plasmids, respectively, while we used short-term ectopic expression of csgD and RprA from medium-copy and low-copy number plasmids, respectively. Furthermore, the deletion mutants used by the two studies were slightly different. Due to the differences in experimental setup in these studies, the exact nature and the primary regulatory mechanism of RprA-controlled repression remain to be resolved. However, given that RprA also induces cleavage in the A/U-rich region of csgD, we suspect that the mechanism of RprA-mediated control of csgD is similar to that of McaS (see below).

Short-term ectopic expression of McaS leads to degradation of csgD mRNA expressed ectopically as a csgD fusion transcript with specific degradation patterns (Figure 5). Primer extension analysis identified sRNA mediated cleavage sites within the 5′ UTR of the csgD mRNA (Figure 3). We performed mutational analysis of these sites via nucleotide substitution, deletion, or forced secondary structure formation (Figure 5A). We cannot exclude the possibility that the changes in cleavage patterns in the mutants are at different sites than in the wild-type construct. However, we observe that specific cleavage products disappear in the mutant constructs and other products remain, indicating that the change in cleavage patterns reflect the introduction of specific mutations. Combined we find that the single stranded A/U-rich tract and the adjacent stem loop at the 5′ end of csgD mRNA are necessary for McaS-mediated degradation in vivo.

Further analysis by structural probing show that the A/U-rich region and the adjacent hairpin in the csgD 5′-end most likely serve as a platform for Hfq binding to both sites (Figure 7). Disruption of Hfq binding site I by deletion of the stem–loop structure, reduced McaS dependent control of CsgD synthesis in vivo (Figure 5) but did not change the dissociation constant (Kd) between Hfq and csgD in vitro (Figure 6). However, we no longer observe a higher mobility complex corresponding to csgD-Hfq3 in electric mobility shift assays. In general, we do not know the exact identity of the nucleoprotein complexes in our mobility shift assays. Hfq interacts with csgD at multiple sites making the interpretation of the results more difficult. It is entirely plausible that the observed mobility shifts are a mixture of Hfq binding at different csgD binding sites. However, given that we have only made mutations close to Hfq-binding Site I, suggests that the changes in Kd values are a consequence of Hfq binding to this site.

Further inspection of the structural probing experiment (Figure 7) suggests that McaS and Hfq cooperates at Hfq-binding Site II and Site III, and not just to Site I. We have previously seen that McaS forms extendend base-paring with csgD (McaS-binding site 2, Figure 4A) (22). This extendend base-paring brings McaS close to Hfq-binding Site II and the RBS. Thus, the additional protection at these sites probably reflects McaS association to this region in vitro. However, this region is not important in vivo since deletion of McaS-binding Site 2 has no effect on McaS mediated repression of CsgD protein levels (Figure 4). Whilst McaS inhibits translation initiation in vitro (22), our new data does not support this observation but suggests that the extended base-paring at McaS-binding Site 2 is not functional in vivo.

Interestingly, the A/U tract is cleaved by RNase E. McaS producing cells have no or only minor CsgD production in an ompR234 background strain, however, upon shift to the non-permissive temperature of a temperature sensitive mutation of RNase E, csgD mRNA accumulates and consequently CsgD protein level increases. The fact that McaS-mediated regulation of CsgD is lost in the absence of RNase E support our view that inhibition of translation initiation has no effect in vivo (Figures 4 and 8). Introducing GGGG residues in the A/U rich region, as well as deleting it, decrease cleavage efficiency and thus McaS mediated control of CsgD biosynthesis (Figure 5, Δstem and ΔA/U). Replacing the adjacent stem–loop structure with a stable tetra-loop and simultaneous masking the A/U region by base-paring (Figure 5, Forced C and Forced+GG) completely eliminates McaS control of CsgD in vivo. Combined, we therefore suggest that the primary regulatory mechanism relies on endoribonucleolytic cleavage in the 5′end of csgD mRNA.

We have schematically summarized our findings in Figure 9A. To complete the overview, we have included the known small RNAs regulating curli production via the CsgD transcriptional regulator. From the illustration, it is clear that the csgD 5′UTR serves as a hub for post-transcription regulation with the involvement of many small non-coding RNAs and with multiple Hfq binding sites. Several of the small RNAs have multiple base-paring regions along the csgD sequence. In this study, we focused on the small RNA McaS and the mechanism by which it controls CsgD expression. The McaS dependent RNase E induced cleavage sites are highlighted by arrows with intense and weak cleavage products, respectively.

In some cases, the sRNA-binding to csgD overlaps with our detected Hfq binding sites. A previous study using Spot 42 small RNA as a model and its association with the entB mRNA demonstrated that the Hfq-binding and sRNA targeting sites cannot overlap (48). How exactly the sRNAs and Hfq interact with csgD at these overlapping sites remain unclear. In addition, it is equally unclear how the different Hfq-binding sites serve as regulatory elements and how they individually contribute to sRNA recognition. We are currently investigating the importance of the individual Hfq-binding sites and how they each aid in regulating CsgD in detail.

It seems likely that the two downstream located Hfq-binding sites, at the TIR are involved in the base-paring of the small RNAs that inhibit translation initiation (e.g. OmrA/B, RydC and RybB) and that the first Hfq-binding site (located at the A/U rich sequence), together with the small RNAs that leads to csgD decay (e.g. McaS, RprA and GcvB), is involved in the recruitment of RNase E, to cleave the csgD substrate (Figure 9B). For simplicity, we have only added RNase E itself to our illustration, though it is part of the degradosome. The degradosome is a multienzyme complex essential for RNA metabolism and consists of RNase E, the helicase, RhlB, the polynucleotide posphorylase, PNPase, and enolase. The N-terminal half of RNase E is highly conserved and contain the endoribonuclease activity. The C-terminal domain is natively disordered, interacts with RNA and provide a scaffold for degradosome assembly (49). The C-terminal scaffold interacts with Hfq-sRNA binary complexes facilitating mRNA target recognition (50). Hfq is likely to dissociate from the complex at some point before or after the initial degradation process. In such a scenario, RNase E could induce cleavage at the A/U rich sequence after the dissociation of Hfq. Then, the weak overlapping cleavage sites in the A/U rich sequence does not exclude Hfq association to this site as well.

Another explanation for the induction of cleavage at the A/U-rich sequence could be a conformational change within the csgD mRNA upon McaS binding. It was previously reported that the binding of MicF deep within the coding region of lprX mRNA decreased mRNA stability by inducing a conformational change that induced RNase E mediated cleavage by exposing the mRNA sequence (36). Our data does not support a model where the csgD mRNA undergo structural rearrangement upon McaS/Hfq binding (Figure 7 and Supplementary Figure S4). We thus favour a model where McaS, together with Hfq, recruit RNase E directly to the A/U rich sequence to induce cleavage of the csgD mRNA.

Interestingly, RNase G does not contribute to csgD mRNA decay but RNase G co-purifies with McaS in pull-down experiments (45). McaS is a dual-functioning sRNA. It activates the production of exopolysaccharide β-1,6 N-acetyl-D-glucosamine (PGA) by interacting with the carbon storage regulator CsrA (51). It also serves as a canonical sRNA regulating gene expression in a Hfq dependent manner. In addition to regulating csgD expression, McaS also stimulates the synthesis of flagella by activating flhD in an Hfq dependent manner (24). However, it seems unlikely that RNase G is involved in these processes, suggesting that McaS has additional uncharacterized targets.

Tight control of csgD expression is an important and integral aspect in E. coli physiology. The switch from a motile single cell lifestyle to that of a sessile non-motile lifestyle involves a complex cascade where many environmental conditions signal the rewiring of the regulatory network. At the centre of this network is the CsgD protein. Our study demonstrates that the McaS mediated control of CsgD biosynthesis depend on mRNA degradation. CsgD has an unusual long 5′UTR where McaS (but also RprA and GvcB) binds far upstream the RBS and induces mRNA cleavage at sites located additionally upstream. Thus, the long csgD 5′UTR is not just required for stabilisation of transcript but also serves as a conserved highly regulatory element. To begin to understand the significance of long untranslated leaders it is important to understand how other long UTRs are involved in transcript regulation. It would be interesting in the future to study if many other UTRs also serve as hubs for sRNA mediated signal integration. Indeed, it has been suggested that long 5′UTRs are important in the control of many virulence genes in L. monocytogenes and it highlights the importance of post-transcriptional control of key regulatory pathways (52).

DATA AVAILABILITY

RNAalifold: http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAalifold.cgi. Clustal Omega: https://www.ebi.ac.uk/Tools/msa/clustalo/.

Click here for additional data file.