Antitoxin MqsA Represses Curli Formation Through the Master Biofilm Regulator CsgD

MqsA, the antitoxin of the MqsR/MqsA toxin/antitoxin (TA) system, is a global regulator that reduces expression of several stress response genes (e.g., mqsRA, cspD, and rpoS) by binding to the promoter palindromic motif [5′-AACCT (N)3 AGGTT-3′]. We identified a similar mqsRA-like palindrome [5′-AACCT TA AGGTT-3′] 78 bp upstream of the transcription initiation site in the csgD promoter (p-csgD). CsgD is a master regulator for biofilm formation via its control of curli and cellulose production. We show here that MqsA binds to this palindrome in p-csgD to repress csgD transcription. As expected, p-csgD repression by MqsA resulted in reduced transcription from CsgD-regulated curli genes csgA and csgB (encoding the major and minor curlin subunits, respectively). Curli production was reduced in colonies and in planktonic cells upon MqsA production. Hence, MqsA directly represses p-csgD, and thereby influences curli formation. This demonstrates that TA systems can impact overall cell physiology by fine-tuning cellular stress responses.

B iofilms consist of bacterial populations adherent to each other, and often, to solid/liquid or air/liquid interfaces 1 . In Escherichia coli and Salmonella spp., a major extracellular component that promotes biofilm formation is curli [2][3][4] . Curli are thin proteinaceous, amyloid fibers (usually 4-12 nm in width and 100 to 10,000 nm in length) 5 that were initially identified as a surface organelle in E. coli that binds to host fibronectin 6 . Secretion of curli fibers to the bacterial membrane surface requires seven genes in two adjacent divergently transcribed operons: csgDEFG and csgBAC 7,8 . CsgE 9 , CsgF 10 , and CsgG 8 are accessory membrane proteins for efficient curli secretion, while CsgB and CsgA are structural subunits (curlin) that assemble into mature curli fibers 11 . The role of CsgC in curli biogenesis is less understood, although it has been suggested to participate in redox activity with CsgG 8 . In E. coli, both operons are activated by CsgD, a transcriptional regulator belonging to the FixJ/UhpA/LuxR family 7 . In addition to curli expression, CsgD also transcriptionally activates the gene of diguanylate cyclase AdrA, which synthesizes cyclic diguanylate (c-di-GMP) 12 . Both AdrA and c-di-GMP have been implicated in cellulose production 12,13 . As both curli and cellulose are components in biofilms, csgD regulation is thus an important determinant in microbial adaptation to different environments.
Curli production is highly responsive to environmental fluctuations such as low temperature 6 , low osmolarity 14 , and nutrient limitation 15 . These environmental cues influence the expression of no less than 10 transcriptional factors, which in turn regulate the expression of csgD 16 . For instance, csgD is activated by RNA polymerase containing the stationary phase sigma subunit s S (RpoS) during stationary growth phase 17 . This activation is further amplified by a positive feedback loop through CsgD-dependent transcription of iraP, which encodes a stabilizing factor for RpoS 18 . Another transcriptional factor, MlrA 19 , also stimulates CsgD expression through a signaling cascade of c-di-GMP generated by YegE and YdaM 20 . At the post-transcriptional level, the 59 untranslated region of csgD mRNA is also a regulatory hotspot 21 24 ] can directly bind to the 59 untranslated region of csgD to subsequently inhibit its expression. Overall, the multiple regulatory layers for CsgD expression underlie the complex regulation of curli production and biofilm formation.
Bacterial toxin/antitoxin (TA) systems are genetic elements that encode both toxic proteins that disrupt cellular processes, and antitoxins that attenuate this toxicity. TA systems are prevalent, with at least 38 TA systems identified in E. coli alone 25,26 . Among these, the MqsR/MqsA system is notable for its involvement in persistence 27 , quorum sensing 28 , biofilm formation 28,29 , direct control of another TA system 30 , and global regulation through the MqsR toxin 28 and the MqsA antitoxin 31 . MqsA possesses a C-terminal helix-turn-helix domain 32 that allows direct binding to a specific palindromic DNA motif [59-AACCT (N) 3 AGGTT-39] found in two copies in the promoter region of mqsRA 33 . Upon binding to this palindrome through its C-terminal domain, MqsA controls expression of various genes such as mqsRA 33 , rpoS 31 , and cspD 34 .
Previously, we identified an mqsRA-like palindromic motif [59-AACCT TA  in the promoter of csgD using a wholegenome bioinformatic search 31 . Here, we show that MqsA binds to this mqsRA-like palindrome to repress csgD expression, which eventually results in reduced curli formation in E. coli. Taken together, MqsA behaves as a negative determinant in biofilm formation and as a regulator of an important regulator (CsgD).

Results
We have shown that MqsA is a global regulator that represses rpoS transcription by binding at the mqsRA-like palindrome of p-rpoS 31 . RpoS is the master regulator of stress response 35 . In the presence of oxidative stress, which would normally induce genes positively controlled by RpoS such as those encoding curli, curli/cellulose production was reduced by 13 6 2 fold in cells producing MqsA, and the csgD transcript was decreased by 3 6 1 fold 31 . Hence, under stressful conditions with overproduction of MqsA, the reduction in curli/ cellulose is at least partially a result of a lack of induction of p-csgD by RpoS due to MqsA repressing p-rpoS. Our previous bioinformatic analysis also identified a mqsRA-like palindrome 78 bp upstream of the transcriptional start site of csgD 31 (Fig. 1). We hypothesized that MqsA decreases curli formation by a direct binding to this mqsRAlike palindrome to subsequently repress csgD at a transcriptional level. Therefore, we examined curli production in the absence of oxidative stress to reduce the effect of RpoS on p-csgD which allowed us to see the effect of MqsA directly on p-csgD. Note that curli production is RpoS-dependent, so an rpoS deletion strain could not be used for phenotypic assays. Since curli is formed in laboratory E. coli strains at temperatures between 26uC to 32uC 6,36 , 28uC was used here to promote curli formation.
Curli production is reduced in MqsA-producing strains. As Congo Red (CR) is a dye that binds curli and cellulose 13 , colonies with high amounts of curli will appear red on salt-free agar. Note that E. coli K-12 does not produce cellulose 13,37 . As expected, the csgD deletion strain appeared as white colonies (Fig. 2a). In an mqsRA deletion strain, producing MqsA from a plasmid rendered the colonies less red than an isogenic strain harboring the empty plasmid after prolonged incubation (Fig. 2a). By quantifying the amount of CR bound to planktonic cells, we determined that curli production was 1.9 6 0.2 fold and 1.7 6 0.2 fold less in the MqsA-producing cells after 3 h and 6 h incubation, respectively (Fig. 2b). Therefore, MqsA reduces CsgD activity via two pathways: (i) indirect repression of p-csgD through repression of p-rpoS under oxidative stress; and (ii) direct repression of p-csgD in the absence of oxidative stress.
To corroborate these results, we further examined the content of curli at cellular level using SEM. In cells harboring the empty plasmid, curli fibers were present (Fig. 3, left panel) in considerable amounts after 2 days of incubation, with curli forming extracellular matrix that traps individual cells to form biofilms. Cells also showed a rougher surface with tiny lumps. In contrast, curli were essentially absent in MqsA-producing cells (Fig. 3, right panel). We estimated that the curli content in 400 MqsA-producing cells was approximately 6 6 7 fold less in comparison to the same number of cells harboring empty plasmid. Hence, production of MqsA reduces curli production. Given that curli production is positively correlated with biofilm formation, this reduced amount of curli in MqsA-producing strain is consistent with the previous observation that biofilm formation was decreased by 2 fold in cells expressing MqsA 31 .

Curli-related gene transcripts are reduced in cells expressing
MqsA. Since p-csgD is repressed by MqsA, we reasoned that genes controlled by CsgD, such as csgB (curli-related) and adrA (celluloserelated), will also be repressed upon production of MqsA. To investigate this possibility, we tested the expression of csgB and adrA in various growth conditions using quantitative real-time reverse-transcription PCR (qRT-PCR). csgB encodes the minor curlin subunit, while adrA (encoding a cyclic diguanylase) is part of the regulatory network in cellulose production. In the five growth conditions tested, the csgD transcript in MqsA-expressing cells, whose mqsRA loci were deleted, was consistently decreased by 2 to  (Table 3). Similarly, csgB and adrA transcripts were also reduced under the same conditions. The largest reductions in csgD and csgB transcripts were seen with prolonged MqsA overexpression; in particular, after ,6 h of MqsA overexpression in LB, csgD and csgB were repressed by nearly 6 fold and 109 fold, respectively. In contrast, adrA repression was more apparent when MqsA production was induced for a short duration. Under 30 min induction, adrA transcript was decreased by nearly 5 fold, but the reductions were less than 3 fold under long inductions (.1 h). This suggests that CsgD does not activate csgB and adrA in the same manner 38 , which further implies that curli and cellulose production are regulated differently during biofilm formation.
To demonstrate the direct effect of MqsA on p-csgD, we tested the expression of csgD, csgB, and adrA in an rpoS-deleted strain. In the absence of RpoS, transcription of csgD, csgB, and adrA in MqsAoverproducing cells remain repressed (Table 3). In comparison with an rpoS 1 strain with MqsA produced from plasmid (BW25113 DmqsRA/pBS(Kan)-mqsA), the csgD, csgB, and adrA transcripts were repressed ,50% less in the rpoS-deleted strain with MqsA produced.
To corroborate this direct binding of MqsA to p-csgD, we produced MqsA via pCA24N-mqsA in MG1655 D6 R3 PrpoS that harbors a mutated mqsRA-like palindrome (59-ACCT TGC TCAC-39) upstream of chromosomal rpoS 31 , and measured the transcription of csgD, csgB, and adrA. In this background, MqsA is unable to affect chromosomal rpoS transcription due to the mutated palindrome in the rpoS promoter 31 . In comparison to the isogenic strain harboring an empty plasmid, csgD, csgB, and adrA were reduced by ,2 fold upon MqsA production (Table 3). csgD transcription was further repressed by nearly 5 fold in MG1655 D6 R1 PrpoS, a strain that harbors the wild-type mqsRA-like palindrome (59-ACCT TGC AGGT-39) in the rpoS promoter 31 . Hence, MqsA represses csgD transcription in the absence of its effect on rpoS transcription, and there is a greater reduction in transcription of csgD when both the promoter of rpoS and csgD are repressed. Therefore, these results confirm direct p-csgD repression by MqsA and demonstrate that   repression of curli synthesis by MqsA is a result of repression of both p-csgD and p-rpoS.
MqsA binds the mqsRA-like palindrome in p-csgD. To investigate whether MqsA binds the mqsRA-like palindrome in p-csgD to mediate gene repression, a 312-bp fragment (p-csgD) was amplified from the csgD promoter of E. coli (Fig. 1), and incubated with MqsA in EMSA reactions. We found that MqsA binds p-csgD (Fig. 4a, lane  2), and that this binding could be reversed by adding unlabeled p-csgD in excess (Fig. 4a, lane 3). For the positive control, MqsA bound the p-mqsRA double palindrome and formed three distinct bands: the most prominent and largest DNA-MqsA band is where MqsA binds both palindromes of p-mqsRA whereas the two smaller bands are for single MqsRA binding each of the individual palindromes (Fig. 4a, lane 5).
We investigated the specificity of the MqsA binding to p-csgD at the mqsRA-like palindrome by incubating MqsA with a 30-bp fragment that corresponds to either the native mqsRA-like palindrome (59-AACCT TA AGGTT-39) or its mutated counterpart with five nucleotides changed and is not able to form a palindrome (59-AACCT TA TCACC-39) ( Table 2). MqsA-bound native palindromes in p-csgD were shifted upon adding a 50-fold, a 100-fold or a 200-fold excess MqsA (Figure 4b, lanes 4, 6, 8). However, when the p-csgD mutated palindrome was used, the binding was drastically reduced (Figure 4c, lanes 4, 6, 8). This shows that MqsA binding to the mqsRA-like palindrome in p-csgD is specific, and this binding mediates CsgD repression.

Discussion
Elucidating the synthesis of bacterial curli amyloids, and its regulation, is important for biofilm research, particularly from a clinical perspective. Bacterial curli fibers share structural, biochemical and biophysical properties with protein amyloids 39 , which are   notoriously implicated in chronic neurodegenerative disorders such as Alzheimer's disease and Parkinson's disease 40 . Thus, curli formation in bacteria has been proposed as a model system to study amyloid formation and regulation 41 . Similar with eukaryotic amyloids, enterobacterial curli fibers are also able to trigger inflammatory responses in mice 42 . Given that curli amyloids (and other extracellular polymeric substances) are important for the biofilm matrix 43 , curli formation is an attractive target for the development of anti-biofilm and anti-virulence drugs 44 . Regulation of curli synthesis is complex. Here, we add new insight into the regulation of curli, i.e., MqsA decreases curli formation by repressing p-csgD. The 300-bp p-csgD has binding sites for over 10 transcription factors 16 that individually or synergistically respond to various stimuli. Among these, the binding site (the mqsRA-like palindrome) for MqsA in p-csgD (278 bp from the transcription start site) overlaps with the binding sites of H-NS (128 to 2201) 45 , CpxR (268 to 280; 290 to 2102) 19,45 , and one of the two sites of IHF (237 to 296) 19,45 . In particular, the mqsRA-palindrome is flanked by two CpxR binding boxes 45 (Fig. 1). As both MqsA (this work) and CpxR 46 negatively regulate curli expression, it is likely that MqsA or CpxR binding will respond to different stimuli (such as oxidative and acid stress for MqsA 31 , or envelope stress for CpxR 47 ).
Overall, our results support the notion that TA systems have physiological roles such as biofilm formation 29,48 , and antitoxin MqsA behaves as a global regulator intimately related to biofilm formation 31 . When nutrients are plentiful, MqsA increases motility by increasing flhD (the master regulator of motility) partly through rpoS 31 inhibition, and partly through csgD inhibition; hence, the role of MqsA would be to inhibit biofilm formation in the absence of stress. Under stressful conditions, however, MqsA is degraded by proteases 31 , and MqsR is activated 49 . MqsA degradation leads to derepression of rpoS and csgD, inhibition of flhD, and subsequently, results in increased biofilm 31 since stress increases biofilm formation 50 . Therefore, this global regulative behavior of MqsA cements the role of antitoxins as far more than regulators of their own loci.

Methods
Bacterial strains, plasmids, and culture conditions. All strains and plasmids used in this study are summarized in Table 1. All strains were grown in lysogeny broth (LB, 10 g/L tryptone, 5 g/L yeast extract, and 10 g/L NaCl), unless specifically indicated. Strains MG1655 D6 R1 PrpoS and MG1655 D6 R1 PrpoS (Table 1) were cultured using 50 mg/mL kanamycin and 5 mg/mL tetracycline. Kanamycin (50 mg/mL) was also used to maintain pBS(Kan)-based plasmids 51 and to select for E. coli BW25113 DcsgD. Chloramphenicol (30 mg/mL) was used to maintain pCA24N-based plasmids 52 . The CmqsRA-f and CmqsRA-r primers ( Table 2) were used to confirm the mqsRA deletion in BW25113 DmqsRA via PCR and DNA sequencing. Similarly, the rpoS-f and ygbN-r primers ( Table 2) were used to verify the rpoS deletion in BW25113 DrpoS via PCR and DNA sequencing. BW25113 DrpoS DKm R was created by eliminating the Km R cassette in BW25113 DrpoS V Km R 53 using FLP recombinase encoded by pCP20 54 .
Congo red (CR) assay. Curli production was examined by CR-binding assay at 28uC using agar plates 13 and planktonic cells 55 . Two mL of each overnight culture was spotted on salt-free CR plates (10 g/L tryptone, 5 g/L yeast extract, 1 mM IPTG, 40 mg/mL CR, 20 mg/mL Coomassie Blue). Plates were incubated for 7 days, and the appearance of red colonies indicates binding to CR. Quantification of CR-binding was performed by measuring the amount of CR binding in planktonic cells. Briefly, 1 mL of each overnight culture was harvested at 13,000 3 g centrifugation for 2 min, and washed with 1 mL of T-broth (10 g/L tryptone). Cells were resuspended in Tbroth containing 1 mM IPTG and 40 mg/mL CR, and were incubated at 28uC for 3 to 24 h. Prior to incubation, an aliquot of cell suspension was removed, and CR was measured spectrophometrically at 490 nm (Abs 490 initial ). At specified time points, cells were harvested at 13,000 3 g centrifugation for 10 min, and the supernatant was spectrophotometrically measured at 490 nm (Abs 490 unbound ). The Abs 490 bound was therefore calculated from Abs 490 initial -Abs 490 unbound , and the amount of CR bound to cells (mg) was obtained from a standard curve constructed using 0 mg to 40 mg of CR dissolved in T-broth. All CR values were normalized using cell densities (OD 600 ). 56 was determined using two-day old colonies grown on salt-free LB agar with 1 mM IPTG at 28uC. Colony cells were suspended gently in 1 mL of 100 mM sodium cacodylate buffer (pH 7.4), and collected by filtering 57 with a 0.22-mm polycarbonate filter (GE Healthcare, Little Chalfont, UK). Cell-containing filters were placed in specimen baskets. Filters were fixed with 2.5% (v/v) glutaraldehyde and 2% (v/v) formaldehyde in sodium cacodylate buffer for 15 min, followed by three washes with buffer alone for 5 min each wash. Filters were then fixed with 1% (w/v) osmium tetroxide in sodium cacodylate buffer for 30 min in dark. After two washes with buffer alone, filters were dehydrated in a graded ethanol series (50%, 70%, 85%, 95% and 3 times with 100% ethanol, for 5 min each). The dehydrated filters were dried using a critical point dryer (Bal-tec CPD-030) with liquid carbon dioxide as a transitional fluid. After affixing the filters on SEM stubs using double-sided carbon adhesive tape, the filters were sputter coated with white gold (Au/Pd) for 3 min (Baltec SCD-050 Sputter Coater). Coated specimens were examined with a field emission SEM (FEI NanoSEM 630 FESEM) operating under low vacuum mode (5 kV).

Scanning electron microscopy (SEM). Curli formation in cells in colony biofilms
RNA isolation and qRT-PCR. Total RNA was isolated from planktonic cells using a Qiagen RNeasy Mini kit, as described previously 58 . IPTG (1 mM) was used to produce MqsA in planktonic cells (IPTG was also added to the empty plasmid controls) for various durations and in various growth medium at 28uC (Table 3). RNA integrity was checked by agarose gel electrophoresis, and by the Abs 260 /Abs 280 ratio of 1.85 to 2.25. qRT-PCR was performed according to manufacturer's instructions (Power SYBR Green RNA-to-C T 1-Step kit, Life Technologies, Carlsbad,  www.nature.com/scientificreports SCIENTIFIC REPORTS | 3 : 3186 | DOI: 10.1038/srep03186 CA) using 100 ng of total RNA as the template. Primers were annealed at 60uC, and rrsG was used to normalize all data. The specificity of all qRT-PCR primers ( Table 2) was verified using normal PCR. Fold changes in various gene transcripts in MqsAproducing strain in relative to strain harboring empty vector were calculated using the 2 2DDCT formula 59 .
Electrophoretic mobility shift assay (EMSA). Gene promoters were PCR-amplified from the genomic DNA of E. coli BW25113 and 39-labeled with biotin (Biotin 39-End DNA Labeling kit, Thermo Scientific, Waltham, MA). Primers p-mqsR-f and p-mqsRr (Table 2) were used to amplify p-mqsRA (corresponds to position 3166541 bp to 3166774 bp of E. coli MG1655; Genbank accession number: U00096.2). Primers EMSA-csgD-f and EMSA-csgD-r (Table 2) were used to amplify p-csgD (corresponds to position 1102383 bp to 1102694 bp of E. coli MG1655). The mqsA-N fragment (that corresponds to the 253 bp at the 59 coding sequence of mqsA) was generated by primers mqsAep-f and mqsA-N-r (Table 2). To investigate MqsA binding specificity, complimentary oligonucleotides with their 39 end labeled with biotin were synthesized, solubilized, and annealed as previously described 31 . Palindrome 1 of p-mqsRA is at position 3166595 bp to 3166609 bp; palindrome 2 is at position 3166629 bp to 3166639 bp; Genbank accession number: U00096.2).
Binding reactions were performed in 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 1 mM DTT and 1 mg of poly (dI.dC). Briefly, 30 to 50 fmol of labeled DNA probe was incubated with purified MqsA in excess at ambient temperature for 1 h. Purified MqsA was sequentially obtained by His 6 -tagged purification, removal of His 6 tag using TEV cleavage, and size exclusion chromatography 32 . The reactions were electrophoresed on a 6% DNA retardation gel (Life Technologies, Carlsbad, CA) at 100 V in 0.53 TBE buffer for 90 min. Samples in the gel were electroblotted onto a nylon membrane (GE Healthcare, Little Chalfont, UK) at 380 mA in 13 TBE buffer for 1 h. The membrane was UV-crosslinked at 302 nm for 20 min, and detection was carried out using the protocol described in the Chemiluminescent Nucleic Acid Detection Module (Thermo Scientific, Waltham, MA).