Reassessing the Role of the Type II MqsRA Toxin-Antitoxin System in Stress Response and Biofilm Formation: mqsA Is Transcriptionally Uncoupled from mqsR

There is growing controversy regarding the role of chromosomal toxin-antitoxin systems in bacterial physiology. mqsRA is a peculiar toxin-antitoxin system, as the gene encoding the toxin precedes that of the antitoxin. This system was previously shown to play a role in stress response and biofilm formation. In this work, we identified two promoters specifically driving the constitutive expression of the antitoxin, thereby decoupling the expression of antitoxin from the toxin. We also showed that mqsRA contributes neither to the regulation of biofilm formation nor to the sensitivity to oxidative stress and bile salts. Finally, we were unable to confirm that the MqsA antitoxin is a global regulator. Altogether, our data are ruling out the involvement of the mqsRA system in Escherichia coli regulatory networks.

TA systems are heavily represented in the accessory genome of most bacterial species, with representatives identified in plasmids, phages, transposons, superintegrons, and genomic islands (15,16). They were first discovered on low-copy-number plasmids and were shown to participate in plasmid maintenance through a process known as postsegregational killing or addiction (17)(18)(19)(20). However, most characterized and predicted TA systems are encoded in bacterial chromosomes and can constitute up to 4% of the total open reading frames (ORFs) in the genome of some species, as seen in Mycobacterium tuberculosis or Microcystis aeruginosa (21)(22)(23). The biological functions of these chromosomal TA systems remain ambiguous and are highly debated; previous studies reported a role for chromosomal TA systems in stress responses (24), programmed cell death (25), generation of specialized ribosomes (26), and antimicrobial persistence (27). However, contradictory studies failed to show an implication of TA systems in stress responses (28)(29)(30), programmed cell death (28,31,32), generation of specialized ribosomes (33,34), or antimicrobial persistence (35)(36)(37). Other established functions in the literature, which are reminiscent of the mobile and addictive nature of TA systems, consist of stabilization of mobile genetic elements such as integrative conjugative elements (38) or superintegrons (39) but also protection against phages through abortive infection and against conjugative plasmids through anti-addiction and plasmid exclusion (40)(41)(42).
A well-characterized chromosomal TA system in Escherichia coli, mqsRA, has been associated with stress resistance (43,44), biofilm formation (45,46), and persister formation (47,48). While most TA operons adopt a promoter-antitoxin-toxin organization, allowing the production of excess antitoxin supposedly through translational coupling, mqsRA adopts an unconventional promoter-toxin-antitoxin configuration (49)(50)(51). The MqsR toxin, an endoribonuclease, was shown to regulate motility (45), biofilm formation (45), and deoxycholate resistance (44). On the other hand, the MqsA antitoxin was described as a pleiotropic regulator that represses the expression of at least three known genes, as follows: csgD, which encodes a transcriptional regulator that modulates the production of cell surface-associated structures (i.e., curli fibers) and biofilm formation (46); cspD, which encodes a toxic protein linked with bacterial persistence (47); and rpoS, which encodes the general stress response sigma factor (43). MqsA was shown to be posttranslationally regulated by oxidative stress since hydrogen peroxide induces rapid degradation of the MqsA antitoxin by the Lon protease, leading to the derepression of MqsA-regulated genes like rpoS (43). Derepression of rpoS by MqsA would induce catalase expression and detoxification of reactive oxygen species, thus confirming the role of mqsRA as a stress response regulator (43). The autoregulatory properties of mqsRA differ from those of other TA systems since the MqsR toxin was shown to alleviate the binding of MqsA to the promoter of the system rather than act as a corepressor (52). Thus, a high level of production of MqsR could also result in the upregulation of MqsA-regulated genes (52). Transcript levels of mqsR were also shown to be upregulated by starvation in serine, valine, or glucose, although the mechanism of this upregulation is unknown (51).
Since the mqsRA system is organized in reverse order, i.e., with the toxin gene preceding the antitoxin one, it raises the question of how the expression of this system is regulated. We found that in addition to the promoter driving the expression of the mqsRA operon, mqsA transcription is driven by two constitutive promoters located in the mqsR coding sequence. These promoters are not regulated by MqsA and are together stronger than the promoter that drives the expression of the whole operon, suggesting that the expression level of the antitoxin exceeds that of the toxin. While investigating the signals that might regulate the transcription activity of the three promoters, we failed to demonstrate any regulation of these promoters by previously reported stress conditions. We also could not show any effect of mqsRA and other TA systems on resistance or tolerance to oxidative stress and deoxycholate or on biofilm formation.

RESULTS
Newly identified promoters uncouple mqsA transcription from mqsR. Due to the reverse configuration of the mqsRA system, we reasoned that additional promoters might be present in the mqsRA operon to allow excess antitoxin production. Putative promoters were predicted using the bprom software (53). A potential sigma70-binding site located inside the mqsR ORF (pA1) that could drive the expression of mqsA independently of mqsR was identified ( Fig. 1A; see also Fig. S1A in the supplemental material). RNA sequencing data sets characterizing transcription start sites (TSS) in E. coli corroborate the functionality of this promoter, as a TSS can be detected 8 nucleotides (nt) downstream of this putative promoter (54). Another TSS can be found 111 nt upstream of this first TSS (54). A fairly conserved sigma70-binding sequence (pA2) can also be found 8 nt upstream of this second TSS ( Fig. 1A and S1A).
To test whether these putative promoters are transcriptionally active, two fragments containing either pA1 or pA2 plus pA1 (referred to as pA1 ϩ 2) and the native mqsA ribosome binding site were cloned in a single-copy reporter plasmid, upstream of mScarlet-I, which encodes a fast-maturing and bright-red fluorescent protein (55) (Fig. 1A and S1A). These constructs were introduced in the MG1655 strain and the isogenic ΔmqsRA mutant. Fluorescence was analyzed by flow cytometry to study the effect of autoregulation on these two promoters as well as on the promoter driving the transcription of the full-length messenger that encodes the whole system (referred to as pRA).
Our results show that the pRA promoter has very low transcriptional activity, as this construct shows fluorescence (111 Ϯ 1 arbitrary units [AU]) very close to the background level (104 Ϯ 1 AU) (Fig. 1B). The activity of the pRA promoter is higher in the ΔmqsRA mutant strain (369 Ϯ 9 AU), confirming that MqsA acts as a transcriptional repressor of this promoter (Fig. 1B). The repression of pRA in the ΔmqsRA mutant strain is restored by introducing the mqsRA operon in trans, confirming that this system is autoregulated (Fig. 1C). The pA1 and pA1 ϩ 2 fragments showed stronger activities than did pRA, with pA1 rivaling a derepressed pRA (354 Ϯ 8 AU) and pA1 ϩ 2 being twice as strong (675 Ϯ 12 AU), confirming the functionality of these predicted promoters in the mqsR ORF (Fig. 1B). However, deleting mqsRA had no effect on the transcriptional activity of pA1 (352 Ϯ 5 AU) or pA1 ϩ 2 (680 Ϯ 13 AU), showing that these promoters are not subjected to autoregulation (Fig. 1B). Note that in all of these cases, the fluorescence distribution of the population was monodisperse (Fig. S1B).

Stress does not regulate mqsRA at the transcriptional and posttranslational levels.
Since mqsA transcription appears to be mainly driven by two internal promoters under the testing conditions described above, we investigated whether these promoters could be regulated by different signals and therefore lead to the modulation of the ratio of toxin to antitoxin proteins. The transcriptional activities of pRA, pA1, and pA1 ϩ 2 were measured under various stress conditions previously assayed in the literature, i.e., amino acid starvation (51) and oxidative stress (43). Cells were starved for serine and isoleucine by using serine hydroxamate or valine, respectively, while oxidative stress was investigated using hydrogen peroxide and methyl viologen. Hydrogen peroxide induced rapid bleaching of the mScarlet-I reporter and is not presented here. Deoxycholate, a bile salt whose resistance is mediated by MqsR (44), was also investigated for its effects on these promoters.
Surprisingly, none of the studied promoters, including pRA, showed appreciable changes in activity after 1 h of treatment with serine hydroxamate, valine, methyl viologen, or deoxycholate compared to activity under untreated conditions (P ϭ 0.36) ( Fig. 2A). Fluorescence distribution did not change under the treated conditions either (Fig. S2A). Upregulation of pRA under serine starvation was also not observed under conditions where mqsR was previously shown to be upregulated, i.e., M9 medium with glucose and amino acids (51) (Fig. S2B).
MqsA stability was also assessed during previously mentioned treatments. Since our anti-MqsA antibody did not detect endogenous MqsA, a low-copy-number plasmid bearing the mqsA gene under the inducible lacL8.UV5 promoter was used to mildly overproduce MqsA. Our results show that MqsA is moderately unstable but is not degraded at a different rate when cells are treated with serine hydroxamate, valine, methyl viologen, hydrogen peroxide, or deoxycholate ( Fig. 2B and S2C). Using conditions identical to those previously published, i.e., growth in LB medium (43), we also failed to observe degradation of MqsA in response to hydrogen peroxide (Fig. S2D). These unexpected results question the stress-responsive nature of mqsRA that was previously described in the literature (43).
The MqsA antitoxin is not a pleiotropic regulator. MqsA was shown to bind and repress several promoters, including csgD, cspD, and rpoS promoters, allowing this antitoxin to regulate biofilm formation, persistence, and the stress response (43,46,47). Since these regulations were demonstrated by overproducing ample amounts of the MqsA antitoxin (43,46,47), which can potentially lead to unspecific binding, repression of these promoters was investigated under more relevant conditions, especially since mqsA-binding operators found in these promoters significantly differ from those found in pRA (Fig. S3A). Transcriptional reporters of csgD, cspD, and rpoS promoter activities were constructed and validated by measuring their activities under conditions known to affect the transcription of these genes, as follows: pcsgD is downregulated in high-osmolarity medium (56), pcspD is upregulated upon glucose starvation (57), and prpoS is upregulated in an ΔarcA mutant during growth in LB medium (58) (Fig. S3B). Our results show that the activities of these three constructs are similar in wild-type and ΔmqsRA mutant strains (Fig. 3A). Mild overproduction of MqsA under a lacL8.UV5 promoter was used to assess whether an excess of MqsA would repress these three promoters. However, promoter activities showed no appreciable changes when MqsA was overproduced (Fig. 3B). No changes in fluorescence distribution were observed under these conditions (Fig. S3C). Altogether, these results show that MqsA does not regulate the activity of the csgD, rpoS, and cspD promoters as previously shown (43,46,47). mqsRA and other TA systems do not foster stress resistance or tolerance. Since mqsRA was shown to play a pivotal role in survival to oxidative stress (43) and bile salts treatments (44), resistance and tolerance to hydrogen peroxide and sodium deoxycholate of a ΔmqsRA mutant strain were measured and compared to the wild-type strain. These experiments were also performed with two independently constructed Δ10TA mutant strains deleted for 10 type II TA systems, mqsRA included (35,36). The minimal inhibitory concentrations (MICs) of peroxide and deoxycholate are similar in the wild-type, ΔmqsRA mutant, and Δ10TA mutant strains (Fig. 4A). Overexpression of mqsA did not affect peroxide or deoxycholate sensitivity either (Fig. 4A). Furthermore, no appreciable differences in survival were observed in a ΔmqsRA mutant or in Δ10TA mutants compared to the wild-type strain (Fig. 4B). Overexpression of mqsA did not affect the killing rate by these two compounds either (Fig. 4B). We used ΔrpoS and ΔtolC mutants as internal controls since they are hypersensitive to peroxide and deoxycholate, respectively (59,60). Catalase activity, which was shown to be regulated by MqsA through RpoS (43), was also measured in exponentially growing cultures, and no change in activity was detected when mqsRA and other TA systems were deleted or when mqsA was overexpressed (Fig. S4). Thus, our results show that mqsRA and other TA systems do not play a role in tolerance or resistance to oxidative stress and bile salts as previously shown (43,44). mqsRA and other TA systems do not regulate biofilm formation. The role of mqsRA and other TA systems in biofilm formation was also reassessed since mqsRA was shown to regulate adherence (45) and the production of curli, amyloid fibers that play a pivotal role in biofilm formation and multicellular behaviors (46). Our results show that the ΔmqsRA and Δ10TA mutant strains generate similar amounts of adherent biomass as the isogenic wild-type strain (Fig. 5A). Moreover, mild overexpression of mqsA did not affect the amount of adherent biomass either (Fig. 5B). The presence of curli was quantified by measuring the binding of Congo red to planktonic cells. No differences in Congo red binding were found between the wild-type, ΔmqsRA mutant, and Δ10TA mutant strains (Fig. 5C) or upon mild overexpression of mqsA (Fig. 5D). In a macrocolony biofilm model, which consists of week-old colonies grown on a lowosmolarity medium containing Congo red (61), no morphological differences between the wild-type, ΔmqsRA mutant, and Δ10TA mutant strains, as well as no differences in coloration, were observed (Fig. 5E). Overexpression of mqsA did not affect these parameters either (Fig. 5F). A curli-deficient mutant (ΔcsgA) was used as a control in all these experiments since it poorly forms biofilms and does not bind Congo red. Altogether, these results suggest that mqsRA and other type II toxin-antitoxin systems do not play a role in biofilm formation as previously shown (45,46).
In this work, we provide evidence that the mqsRA type II toxin-antitoxin system does not play a role in the stress response and biofilm formation as previously shown (43)(44)(45)(46)(47). Our results show that under all of the stress conditions in which mqsRA has been suggested to play a role, the transcription of both mqsR and mqsA remains stable. The half-life of the MqsA protein also remains unchanged under these stress conditions. It was hypothesized that MqsA regulates stress responses by repressing the transcription of the rpoS general stress sigma factor gene (43). Our results show that MqsA levels have no effect on rpoS expression. We also showed that deleting mqsRA has no effect on resistance or tolerance to two of these stresses which are bactericidal, namely, oxidative stress and bile salts. Therefore, we conclude that the mqsRA system does not respond to these various stress conditions and is likely not implicated in a response to these stresses. As no changes in mqsA expression under stress conditions were observed, both at the transcriptional and posttranslational levels, it is likely that MqsA is maintained at a steady and constant level. This is corroborated by the constitutive nature of the two newly identified mqsA promoters which would allow for a constant and steady level of mqsA transcription.
We also showed that mqsRA and nine other type II TA systems (relBE, yefM-yoeB, mazEF, dinJ-yafQ, yafNO, chpB, higBA, hicAB, and prlF-yhaV) do not play a role in biofilm formation using two different biofilm models, i.e., adherence in microplates and macrocolonies, and by showing that mqsA does not regulate rpoS and csgD expression, Revisiting the Biological Roles of mqsRA ® which are two pivotal factors in biofilm physiology (62). Moreover, we showed that mqsRA does not affect the biosynthesis of curli fibers, which are a pivotal component of the extracellular matrix in biofilms (61,62). In light of these contradictory results, we reexamined the literature covering the biological functions of mqsRA. The mqsR (motility and quorum-sensing regulator) gene was first identified as an autoinducer 2-responsive regulator of motility and biofilm formation (45). Indeed, it was shown that a Tn5 insertion mutant of mqsR was less motile and formed less adherent biomass (45). Interestingly, this low-motility phenotype could be complemented by overexpressing mqsR under a strong inducible promoter on a multicopy plasmid compared to a wild-type strain (45), despite being later shown to be a growth-inhibiting toxin (50). It was also reported that the effects of mqsR on motility and biofilm formation were mediated through the qseBC twocomponent system, which is located around 1 kb upstream of mqsR (45). We hypothesize that a Tn5 insertion in mqsR could alter mqsRA-mediated transcriptional regulation of the pRA promoter and induce polar effects in its close vicinity. The presence of an frt-flanked kanamycin resistance cassette or an frt scar in lieu of mqsR had contradictory effects on biofilm formation, supporting the hypothesis that a polar effect might be responsible for these effects on motility and biofilm formation (50). It was also reported that deleting mqsRA in a strain already deleted for five TA systems increased biofilm formation instead of reducing it (43). Another explanation for these discrepancies would be that the aforementioned MG1655 mqsR::Tn5 strain is not isogenic to the wild-type strain to which it was compared in this study (45). Indeed, this mqsR mutant shows a drastic decrease in the levels of Crl (26-fold), an RNA polymerase-RpoS holoenzyme assembly factor responsible for curli biosynthesis and biofilm formation. It is known that various clones that are labeled as MG1655 show genetical differences, especially with regard to insertion sequence (IS) dynamics (63). The crl gene is known to be disrupted by an IS1 insertion in a specific clone of MG1655, CGSC7740 (64,65). Increased expression of the opp and glp operons as well as reduced gatC expression were also observed in the mqsR::Tn5 mutant when compared to a wild-type strain (45). The CGSC7740 clone harbors several mutations that can explain these differences. First, an IS5 insertion in the opp promoter disrupts an operator known to be bound by the Lrp repressor, thus likely causing increased expression of this operon (65). A nonsense mutation in the glpR gene encoding the repressor of the glp operon can also be found in this clone and is known to induce derepression of the glp operon (65,66). Finally, another nonsense mutation can be found in the gatC gene, thus likely causing reduced expression of this gene (65). Also, many transcripts from the e14 cryptic prophage (ymfI, lit, ymfD, ymfK, ymfG, ymfJ, ymfT, ymfL, intE, and mcrA) are detected more abundantly in the mqsR::Tn5 mutant (45). Since e14 can readily excise from its attachment site (67), this could suggest that this prophage was excised from the wild-type strain that was compared to the mqsR::Tn5 mutant. An apparent decrease in flhDC expression in the mqsR::Tn5 mutant, and thus, reduced motility, can also be explained by the high propensity of the flhDC promoter to undergo IS1/IS5 insertions and excisions, which are known to affect motility in supposedly isogenic strains (63,64,68). Altogether, we suggest that the mqsR::Tn5 mutant might have been constructed from clone CGSC7740 but compared to another, nonisogenic wild-type MG1655 strain. Differences that were observed between these two strains are likely due to differences in background rather than to the disruption of mqsR since motility and biofilm formation are known to vary between various backgrounds of E. coli K-12 (45,50,64,69). This is also supported by the fact that we did not observe changes in any protein levels in strains deleted for 10 TA systems, including mqsRA, compared to an isogenic wild type (36). Other sources for discrepancies could be phase variations; indeed, mqsR was previously shown to regulate the flu gene encoding the antigen 43 adhesin (45,47). However, a subsequent study showed that mqsR does not regulate flu and that phase variations in activation of the flu promoter through dam methylation could be responsible for these contradictory reports (70).
Another study showed that mqsA regulates biofilm formation by repressing the transcription of the csgD master regulator of curli biosynthesis (46). This was shown by overexpressing mqsA using high-copy-number vectors, such as pBS(Kan) mqsA (46), a Bluescript-based vector with a lac promoter (71), or pCA24N mqsA (43,47), a rop-pBR322-based vector with a T5-lac promoter (72). These vectors are replicated by origins that are mutant derivatives of the pMB1 replicon, which are known to be maintained at several hundreds of copies per cell (71,72). Using such high-copynumber vectors could lead to a disproportionate production of MqsA, which would bind to low-affinity and biologically irrelevant operator-like sequences like the ones present in the promoters of csgD, cspD, and rpoS. Using a pSC101-based vector (5 copies per chromosome) as well as a deletion mutant of mqsRA, we showed that MqsA levels, whether they are reduced by deletion or increased by reasonable overexpression, do not affect the transcriptional activities of the csgD, cspD, and rpoS promoters. A gene found directly downstream of mqsRA, ygiS, was shown to be an important factor for deoxycholate resistance (44). The ygiS transcript, encoding a protein promoting deoxycholate import, is cleaved by MqsR, thus promoting its downregulation and reducing deoxycholate import and sensitivity (44). However, we failed to show that mqsRA was responsive to deoxycholate at both the transcriptional and posttranslational levels or had any effect on deoxycholate tolerance or resistance. Since ygiS is located directly downstream of mqsRA, we cannot exclude the possibility that the mqsRA deletion performed by the authors had polar or readthrough effects on ygiS transcription. It is also worth to note that the overexpression of mqsR induces cleavage of 30% of the messenger transcriptome as well as in rRNA precursors, showing that mqsR is an RNase promoting general downregulation of translation by cleaving RNA indiscriminately rather than being a posttranscriptional regulator that cleaves specific transcripts to promote their downregulation (34,73). Moreover, the overexpression of mqsR was shown to increase ygiS transcript abundance rather than decrease it (34).
Another publication showed that mqsA was mediating the general stress response by repressing rpoS transcription (43). Under oxidative stress, MqsA was degraded by the Lon protease, leading to derepression of rpoS and increased stress resistance (43). However, this regulation was tested again by using ample overexpression of 6ϫHis-MqsA using a pCA24N vector. As stated above, overproduction studies are not an appropriate experimental setup. Moreover, the authors showed that 6ϫHis-MqsA overexpressed using this vector was rapidly degraded under oxidative stress, contrary to what we showed with reasonably overexpressed and untagged MqsA. The reason of this discrepancy is unclear, but the presence of a His tag or the high level of production, which might possibly induce misfolding, could increase the proteolytic sensitivity of MqsA under oxidative stress.
The idea that toxin-antitoxin systems respond to stress and regulate biological processes is not recent (24,25,74). For example, many TA systems in E. coli have been shown to be transcriptionally upregulated under stress conditions (e.g., serine hydroxamate-induced amino acid starvation) (51,75). This has been attributed to stress-induced degradation of antitoxins, leading to the derepression and transcriptional upregulation of TA operons. This would allow the liberation of toxins from TA complexes, which would then be able to exert bacteriostatic activity, promoting dormancy and, thus, stress tolerance. However, in the current state of the literature, we think there is no strong evidence of TA-mediated stress responses. Previous research led in our group showed that deleting five TA systems did not lead to viability or fitness changes under various stress conditions, supporting the idea that these modules are not implicated in the stress response (28). Moreover, proteomic analysis of strains deleted for 10 TA systems failed to show changes in any protein levels, questioning the functionality of these systems beyond their own regulation and sustainability (36).
To conclude, our data do not support any implications of mqsRA and other toxinantitoxin systems in core biological functions of E. coli, such as stress response and biofilm formation. We think that genetic modules presenting so much diversity and abundance and that raise so many questions deserve to be studied with due standards.