Identification of 24 different genes whose inactivation alters MarA function.
We used random transposome mutagenesis to find chromosomal genes affecting MarA-dependent MDR. We could not directly screen for mutants with altered MDR, since we would obtain many mutants that affected MDR by mechanisms unrelated to MarA. Therefore, we chose an alternative primary screening method which detected reduction of MarA function. As detailed in Materials and Methods, we used the HdeA100 strain (ΔmarRAB hdeABp-lacZ, with marA cloned in a plasmid) to select mutants in which MarA had lost its ability to repress the expression of the hdeABp-lacZ fusion (blue colonies). After screening 11,000 mutants (about 2.5 hits per gene), we found 24 different chromosomal genes whose inactivation prevented hdeAB repression by MarA. They encoded unrelated proteins, including transcriptional regulators (cspG, crp, hns, and ompR), enzymes (appA, cyaA, degP, maoC, metL, pcnB, recD, treC, and ynfE), transport proteins (acrA, acrB, mhpT, nikD, and tolC), and others (alx, damX, metV, yfdG, yibL, and yniD) (see Table ).
Effect of genes whose inactivation prevented repression of hdeAB by MarA on MarA-mediated MDR
Our assumption was that some of these genes would also affect MarA function on genes other than hdeAB
and, thus, would be important for MarA-mediated MDR. On the other hand, the effects of some of them on MarA may be specific for hdeAB
or may be the result of an indirect effect on the complex regulatory network of hdeAB
(see reference 42
), or some might affect the expression or copy number of the IPTG-inducible, marA
-containing plasmid in the strain HdeA100. Such genes, identified just because of the screening method used, were not expected to affect MarA-mediated MDR; however, some may still be relevant for MarA-mediated MDR if, independently, they were also involved in regulation of marA
expression or MarA function (for example, global regulators and other genes with pleiotropic effects).
Since the goal of this work was to find genes important for MarA-mediated MDR, the identified genes were not studied with regard to how they affected hdeAB repression by MarA. Instead, we performed a second screen, involving antibiotic resistance, to find which of them affected MarA-mediated MDR. This second screen was not performed in the HdeA100 background because of the aforementioned limitations and because, in this strain, marA is cloned in a multicopy IPTG-inducible plasmid. In the HdeA100 strain, therefore, the promoter and copy number of marA are not physiological, and ampicillin is needed in the plates to maintain the plasmid, which may alter the effects of the antibiotics tested in the second screen.
For the second screen, therefore, we transferred the mutations in the identified genes to the parental strains CR1000 (E. coli
; overexpresses marA
) and CR2000 (E. coli
; MarA free). These two strains were chosen for several reasons. There is almost no phenotypic difference in antibiotic susceptibility between a wild-type strain and a ΔmarRA
strain (Table ) because the expression of marA
in a wild-type strain (BW25113) is strongly repressed by MarR. A system in which marA
expression is not repressed is necessary to study MarA-mediated MDR. In a wild-type strain, such derepression can be achieved by adding to the culture medium chemicals such as salicylate, known to induce marA
expression by inactivating MarR (see the introduction). However, salicylate affects cell growth, MDR, and gene expression independently from MarA, which would complicate our studies and their interpretation. A more appropriate system for our screen involved the use of a ΔmarR
parental strain, which constitutively overexpresses marA
at physiological levels from its native location in the chromosome and, thus, is similar to MDR clinical isolates that overexpress marA
because they have mutations in marR
. As a comparing strain, we chose the ΔmarRA
parental strain instead of the wild-type strain because both strains are physiologically similar (similar antibiotic susceptibilities) and because the ΔmarRA
strain is completely MarA free so it is the best reference for antibiotic susceptibility and gene expression levels in the absence of MarA. Moreover, using this strain instead of the wild type prevents genes that indirectly affect marA
expression via MarR, e.g., proteins that interact with MarR, from being found in our screen (13
MICs and levels of MarA-mediated MDR in the wild-type and parental strains
After transferring the mutations in the identified genes to the ΔmarR and ΔmarRA parental strains, we examined the susceptibilities of the resulting strains to four structurally and functionally unrelated bactericidal (cefoxitin and norfloxacin) or bacteriostatic (chloramphenicol and minocycline) antibiotics whose increased MICs were MarA dependent (their MICs were increased in the ΔmarR parental strain compared to the level for the ΔmarRA parental strain) (Table ).
Fifteen identified genes affect MarA-mediated antibiotic resistance.
Compared to the levels for the ΔmarR and ΔmarRA parental strains, inactivation of 15 of the 24 identified genes reduced MarA-mediated resistance to at least one antibiotic by 50% or more; in contrast, inactivation of 9 of the 24 genes (cspG, ompR, appA, metL, yfnE, mhpT, alx, metV, and yfdG) did not significantly alter MarA-mediated antibiotic resistance (Table ).
Inactivation of 7 of the 15 genes affecting resistance reduced MarA-mediated resistance to only one antibiotic: these genes were nikD, degP, recD, and yibL (chloramphenicol), maoC and yniD (minocycline), and treC (norfloxacin) (Table ). This finding suggests that these genes do not affect marA expression or MarA function on genes essential for the Mar phenotype, such as acrAB or tolC. In contrast, these seven genes may have a role in some MarA-mediated cell responses that are important only for certain antibiotics or may be necessary for such responses to be effective. These genes were not further studied.
Inactivation of the other eight genes (crp, hns, cyaA, pcnB, acrA, acrB, tolC, and damX) decreased MarA-mediated MDR to all the antibiotics tested (Table ). This was expected for acrA, acrB and tolC because of the major role of AcrAB-TolC in the Mar phenotype but was not known for the other five genes. Because of their general effect in MarA-mediated MDR, these genes were studied in more detail (see Fig. to ).
FIG. 1. Complementation of genes whose inactivation reduced MarA-mediated MDR to all antibiotics tested. Those genes whose inactivation reduced MarA-mediated MDR to all antibiotics tested (Table ) were added back on a plasmid to their respective (more ...)
FIG. 4. Effects of crp, cyaA, and hns on adaptation of MarA-mediated MDR to environmental stimuli. The percentage of MarA-mediated MDR was calculated as explained in Table , footnote a. The results are presented as the average ± the standard (more ...) crp and cyaA.
CRP is one of the major global transcriptional regulators in E. coli
. It is involved in regulation of catabolic operons in response to the energetic status of the cell and in other functions. CRP directly activates or represses the transcription of about 200 genes, although it probably interacts with many other low-affinity binding sites in the chromosome. The activity of CRP is triggered by binding of the second messenger cAMP, whose conversion from ATP is catalyzed by the adenylate cyclase enzyme CyaA in response to glucose starvation and other stresses (references 20
and references therein).
Inactivation of crp
moderately reduced MarA-mediated MDR to all the antibiotics tested (Table ). Complementation of these mutants by the crp
gene on a plasmid restored MarA-mediated MDR (Fig. ). Also, addition of 1 mM cAMP to the culture medium complemented the cyaA
inactivation, whereas it had no effect on the parental or crp
strains (data not shown). Inactivation of crp
had previously been shown to increase MDR to oxacillin, macrolides, and crystal violet in a strain also inactivated for the major pump AcrAB-TolC because CRP-cAMP represses the expression of the accessory multidrug efflux pump MdtEF (38
). In contrast, we show here that CRP-cAMP contributes to MDR in E. coli
in the presence of MarA and the AcrAB-TolC pump.
The reason for such a reduction in MarA-mediated MDR in the crp
mutants seems to be that the expression of marA
(Fig. ) and, in consequence, the MarA induction of acrA
, and micF
(Fig. ) were significantly reduced in both mutants. Earlier studies had demonstrated that CRP-cAMP binds to and activates marRAB
transcription in vitro
but not in cells grown in minimal medium (20
). Our results agree with those in vitro
results and suggest that activation of marA
expression by CRP-cAMP does occur in cells grown in rich medium, which is important for full acrA
, and micF
activation by MarA and thus for MarA-mediated MDR. However, indirect effects of CRP-cAMP on regulation of marA
expression cannot be ruled out, given the many genes regulated by CRP-cAMP.
FIG. 2. Gene expression in mutants with reduced MarA-mediated MDR to all antibiotics tested. The effect of each inactivated gene on the expression of marA, acrA, tolC, micF, and gapA was measured by RT-qPCR (see Materials and Methods). The results are presented (more ...)
Therefore, crp and cyaA play a role in MarA-mediated MDR by being involved in upregulation of marA expression and, in consequence, of MarA-regulated genes.
H-NS is a nucleoid-associated DNA-binding protein that plays a major role in the organization of the bacterial chromosome and in regulation of gene expression. It affects the transcription of over 5% of E. coli
genes, usually acting as a repressor or as a gene silencer. H-NS plays a pleiotropic role in bacterial response to environmental stimuli such as temperature, osmolarity, pH, and starvation (17
Inactivation of hns
strongly reduced MarA-mediated MDR to all the antibiotics tested (Table ). Addition of hns
on a plasmid restored (cefoxitin) or significantly increased (norfloxacin, chloramphenicol, and minocycline) MarA-mediated MDR (Fig. ). H-NS has previously been shown to decrease MDR in a strain inactivated for the major multidrug efflux pump AcrAB-TolC. This occurred because H-NS represses acrEF
, and emrKY
expression. These three operons are not expressed in a wild-type strain grown in LB and encode three TolC-dependent MDR efflux pumps whose activity is detected only in the absence of AcrAB-TolC (37
). In contrast, we found that H-NS increases MDR in E. coli
in the presence of MarA and an intact AcrAB-TolC pump.
When we studied the hns
mutant at the level of gene expression, we found that marA
expression and MarA-dependent induction of tolC
expression were unaffected (Fig. ). However, the level of acrA
induction by MarA was half that in the parental strain (Fig. ). This finding indicates that the reduced MarA-mediated MDR observed in the hns
mutant is caused by a decrease in the AcrAB component of the AcrAB-TolC multidrug efflux pump. Such a specific contribution by H-NS to MarA induction of acrA
, but not the other MarA-regulated genes tested, is of interest because H-NS is known to have no effect on regulation of acrA per se
). Moreover, such an effect also occurs on hdeAB
, which is regulated by MarA and H-NS both independently and synergistically (42
). Decreased induction of acrA
expression by MarA in the hns
mutant (Fig. ) did not involve alteration of marA
expression (Fig. ), and no protein-protein interaction between H-NS and MarA was found in two-hybrid experiments (data not shown). Thus, this effect is either indirectly mediated by other H-NS-regulated genes or attributable to synergic MarA binding or function at the acrA
promoter caused by H-NS effects on DNA topology. An effect of DNA topology changes on acrA
expression has not been reported before; however, DNA looping or changes in DNA supercoiling produced by H-NS, even when bound far from a promoter, can alter the DNA affinity of other transcription factors (reference 17
and references therein).
Therefore, hns plays an essential role in MarA-mediated MDR by being necessary for full MarA induction of acrA expression.
gene encodes the enzyme poly(A) polymerase I (PAP I), which, together with polynucleotide phosphorylase and other proteins, is involved in polyadenylation, processing, and degradation of the vast majority of E. coli
mRNA transcripts in exponential phase (reference 36
and references therein).
Inactivation of pcnB
strongly reduced MarA-mediated MDR to all the antibiotics tested (Table ); addition of pcnB
on a plasmid to this mutant restored MarA-mediated MDR (Fig. ). The reason for such a reduction in MarA-mediated MDR in the pcnB
mutant seems to be that the levels of marA
transcripts (Fig. ) and, by consequence, MarA induction of acrA
, and micF
(Fig. ) were strongly reduced in this mutant compared to the levels for the parental strain. Further experiments showed that marA
transcripts were polyadenylated both in the parental strain and in the pcnB
mutant (RT-PCR products of the same size were obtained; data not shown). Moreover, we found no significant differences in the half-lives of the marA
mRNA transcripts between the parental strain and the pcnB
mutant (Fig. ). These findings suggest that the reduction in marA
transcript levels found in the pcnB
mutant compared to the level for the parental strain is not the result of a decrease in marA
mRNA stability but rather the result of a decrease in marA
transcription. Considering that a role for PAP I in transcription has not been reported, the effect of PAP I on marA
expression may be indirectly mediated by other proteins, such as SoxS and Rob, two activators of marA
transcription whose mRNA stabilities have been found to be increased after overproduction of PAP I (36
FIG. 3. Effect of pcnB inactivation on marA mRNA stability. The absolute numbers of marA transcripts per ng of total RNA before addition of rifampin (time zero) were 2.3·105 for the ΔmarR parental strain (shown in blue), 1.1·105 for the (more ...)
Therefore, pcnB-encoded PAP I plays an important role in MarA-mediated MDR by being necessary, probably indirectly, for full marA expression.
acrA, acrB, and tolC.
Upregulation of the expression of the AcrAB-TolC multidrug efflux pump is the main basis for MarA-mediated MDR (28
), which explains why inactivation of acrA
, or tolC
dramatically reduced MarA-mediated MDR to all the antibiotics tested (Table ). Such reduction was similar for each of these three genes (Table ). This fact suggests that other minor multidrug efflux pumps in which AcrA or TolC are involved (28
) do not play a role in MarA-mediated MDR under the conditions tested here.
These three genes may have additional effects on the MarA system besides their role in antibiotic efflux, since we found in our first screen that their inactivation also reduced MarA repression of hdeAB by an unknown mechanism. Therefore, we studied them in more detail to determine if they have other effects on MarA-mediated MDR.
To study additional effects of the AcrAB component of the AcrAB-TolC pump, we focused on the acrB mutant. As expected, complementation of this mutant with the acrB gene on a plasmid restored MarA-mediated MDR (Fig. ). However, such a complementation was only partial for two antibiotics (cefoxitin and norfloxacin), which suggested that the inactivation of acrB might be also affecting the expression of the upstream acrA gene. In fact, when we studied the acrB mutant at the level of gene expression (RT-qPCR), we found that acrA transcriptional activation by MarA was indeed significantly reduced (Fig. ). However, marA expression (Fig. ) and tolC and micF activation by MarA (Fig. ) were not affected. To study whether this reduction in acrA activation by MarA was produced by the insertion of the transposome in acrB or by the absence of the AcrB protein, we added a single-copy plasmid bearing an acrA-lacZ transcriptional fusion to the parental and acrB mutant strains. In the parental strain, MarA activation of acrA-lacZ expression obtained by β-galactosidase assays (not shown) was similar to MarA activation of acrA expression found by RT-qPCR, showing the equivalence of both assays; however, in the acrB mutant, acrA-lacZ activation by MarA was no longer reduced compared to the level for the parental strain (not shown), in contrast to the RT-qPCR results. This finding shows that the reduced acrA activation by MarA in the acrB mutant compared to the level for the parental strain observed by RT-qPCR was produced artifactually by the cis transposome insertion in acrB and not by the absence of the AcrB protein.
Complementation of the tolC
mutant with tolC
on a plasmid restored MarA-mediated MDR, as expected (Fig. ). When we studied this mutant at the level of gene expression, we found that marA
expression was significantly increased (2.7-fold) compared to the level for the parental strain (Fig. ). This result agrees with a recent publication by Rosner and Martin (41
). Using a different method (lacZ
fusions) to study gene expression, they found that marA
expression, Rob activity, and the expression of several genes belonging to the mar
regulon were increased in a tolC
). They did not find the mechanism by which inactivation of tolC
expression, but they propose that TolC is involved in one or more pumps responsible for effluxing cellular metabolites; in a tolC
mutant, these metabolites would not be eliminated and would trigger the activation of the MarA/SoxS/Rob system in order to upregulate TolC-mediated efflux and restore homeostasis (41
). The metabolites and pump(s) involved are still unknown (41
). However, they found that the mechanism by which tolC
inactivation alters marA
expression is independent of the major AcrAB-TolC pump, since inactivation of acrAB
did not produce the changes in marA
expression and other changes found when tolC
is inactivated (41
). Our findings agree with their result, since we also found no alteration of marA
expression in the acrB
mutant (Fig. ).
Interestingly, despite marA
expression being higher in the tolC
mutant (Fig. ), the degree of activation of acrA
expression by MarA in this mutant was equal to or less than that observed in the parental strain (Fig. ). However, in both the ΔmarR
and the ΔmarRA
backgrounds, the absolute numbers of acrA
transcripts per ng of RNA in the tolC
mutant were 2- to 4-fold higher than those observed in the parental strain (data not shown). These findings may be explained by the increased soxS
expression and Rob activity also found in a tolC
). SoxS and Rob are two MarA homologs that are functionally similar to MarA with respect to their binding sites and the genes that they regulate. Thus, their increased amount/activity in the tolC
) would explain the higher levels of acrA
expression in absolute numbers that we found in the tolC
mutant compared to the level for the parental strain. Moreover, the functional overlapping of these regulators with MarA may explain why the level of marA
expression was higher in the tolC
mutant than in the parental strain but the level of MarA induction of acrA
was not higher. Induction of the expression of acrA
by additional amounts/activity of SoxS and/or Rob (41
) would impede additional induction of these genes by MarA, since all three regulators have the same binding site. These findings may be also explained by the different response that some genes have in vivo
to changes in the concentration of MarA (34
Therefore, acrA, acrB, and tolC are essential for MarA-mediated MDR because of the role of the AcrAB-TolC pump in antibiotic efflux. Moreover, tolC is involved in regulation of marA expression by an unknown and probably indirect mechanism.
gene encodes a predicted membrane-anchored protein conserved among many Enterobacteriaceae
. Little is known about its regulation and function. Lyngstadaas et al. (29
) found that overproduction of DamX interfered with cell division. However, they suggested that such an effect could be indirect or nonspecifically produced by nonphysiological amounts of DamX, since they found no differences between a wild-type and a damX
-inactivated strain. Leclerc et al. (25
) found that the damX
homolog in Salmonella
might be involved in adherence to and invasion of human intestinal cells.
We found that inactivation of damX moderately decreased MarA-mediated MDR to all the antibiotics tested (Table ). This is the first report of a role for damX in E. coli when expressed at physiological levels. Addition of damX on a plasmid restored MarA-induced MDR (Fig. ). Inactivation of damX had no effect on marA expression or on MarA induction of tolC and micF (Fig. ). However, it dramatically decreased MarA induction of acrA expression (Fig. ). This finding indicates that the reduction in MarA-mediated MDR observed in this mutant was the result of a decrease in the AcrAB-TolC multidrug efflux pump. Considering that marA expression was unaffected (Fig. ), that no protein-protein interaction between DamX and MarA was found in two-hybrid experiments (data not shown), and that DamX is a predicted membrane protein, the effect of DamX on acrA induction by MarA is likely to be indirect.
Therefore, damX contributes to MarA-mediated MDR by specifically enhancing, probably indirectly, MarA induction of acrA expression.
The identified genes can mediate the adaptation of MarA-mediated MDR to environmental stimuli.
Understanding how the genes found here affect MarA-mediated MDR provides a better knowledge of how marA functions in E. coli. Since the expression/activity of many of these genes/proteins are known to be regulated by different environmental stimuli, we studied whether these genes played a role in adaptation of MarA-mediated MDR to different environmental conditions. As examples, we focused on crp, cyaA, and hns (Fig. ).
As mentioned above, CRP-cAMP regulates bacterial response to glucose starvation and other stress conditions. In the presence of glucose, CyaA does not catalyze the conversion of ATP to cAMP and CRP is not activated. As shown in Fig. , we found that addition of glucose (0.4%, wt/vol) to the medium indeed reduced MarA-mediated MDR to all the antibiotics tested in the parental strain. Such a reduction was similar to that observed for the crp and cyaA mutants in the absence of glucose. On the contrary, addition of glucose did not significantly alter MarA-mediated MDR in the crp and cyaA mutants (Fig. ). Thus, MarA-mediated MDR is greater under glucose starvation conditions than in the presence of glucose, and such an effect is dependent on crp and cyaA.
H-NS regulates bacterial response to osmolarity and other stresses. H-NS activity strongly decreases at high osmolarity (e.g., 300 mM NaCl) because changes in H-NS protein structure and/or DNA curvature reduce H-NS affinity for its DNA-binding sites (reference 14
and references therein). We found that addition of 300 mM NaCl to the medium reduced MarA-mediated MDR to all the antibiotics tested in the parental strain (Fig. ). This reduction was similar to that observed in the hns
mutant without added NaCl. On the contrary, addition of NaCl did not significantly alter MarA-mediated MDR in the hns
mutant (Fig. ). Thus, high osmolarity reduces MarA-mediated MDR in a hns
The results presented here demonstrate the importance of 15 chromosomal genes in modulating antibiotic resistance mediated by the transcriptional regulator MarA in E. coli. These genes encode unrelated transcriptional regulators, enzymes, transporters, and unknown proteins and can affect MarA response in general (by altering marA expression or MarA function) or specifically only for some antibiotics. Moreover, some of these genes can mediate the adaptation of MarA-mediated MDR to different environmental stimuli.
These and previous findings show that MarA-mediated MDR is regulated at multiple levels, including marA
transcription, MarA protein stability, and MarA function on specific promoters (Fig. ). They also show an interconnection of the MarA system with global regulation and cell metabolism. Why does the MarA system have such a multifaceted regulation? It may be a consequence of the complexity of antibiotic action and cell responses to antibiotics, including secondary targets, signaling effects, and indirect effects, such as antibiotic-induced generation of hydroxyl radicals or changes in gene expression and cell metabolism (8
). On the other hand, such a regulation would allow integration of different signals as well as transitory and fine-tuned adaptation of the MarA-mediated response to different conditions, which would ensure that such a response is always adjusted to the cell needs. This idea agrees with reports showing that marA
expression changes when cells are grown in different media or growth phases (4
FIG. 5. Regulation of MarA-mediated MDR. The figure was produced using data from the literature (see main text) and the results obtained here. Functional interactions, either direct or indirect, are represented as arrows for activation/induction and as “” (more ...)
This complex regulation has additional implications. Modification of marA
expression (directly or indirectly, e.g., via crp
, or tolC
) is a mechanism of generally altering the expression of the MarA regulon, which may thus affect other MarA-regulated functions such as virulence (1
). On the other hand, altering MarA activity only on some genes allows a specific modification of only certain MarA functions. Here, we found that hns
specifically alter MarA activation of acrA
expression and, thus, MarA-mediated MDR. We had previously found that other regulators of hdeAB
were able to modify MarA activity on hdeAB
). Therefore, gene-specific alteration of MarA activity by other proteins seems not unusual, which may explain why some genes are differently activated in vivo
by the same concentrations of MarA, especially when these differences are poorly correlated with the in vitro
affinity of MarA for their promoters (34
The complexity and adaptability of the MarA response may favor the spread of MDR strains with mutations in the autorepressor marR
. Also, the different levels of MDR found among different clinical isolates mutated in marR
(e.g., reference 31
) may be explained by additional mutations in the genes found here to modulate MarA-mediated MDR. Moreover, these findings raise the question of whether the effect of other MDR regulators, such as AcrR, a repressor of acrA
, or the MarA-homologs SoxS and Rob, as well as marA
orthologs or marA
-like regulators in other clinically relevant pathogens, are also subject to this kind of fine-tuned regulation. Such a complex regulation of MDR may exist also in other bacteria, especially among Enterobacteriaceae
. However, the specific regulatory details may be different since other species have different genes and regulatory networks, including additional MarA homologs and additional MDR regulators, and the genes found here to alter MarA-mediated MDR may have different effects. For example, in contrast with what was found for E. coli
, inactivation of tolC
in S. enterica
does not affect marA
expression but results in overexpression of the marA
Finally, inhibitors of multidrug efflux pumps and inhibitors of MDR and virulence regulators, including MarA, have been proposed to enhance the effectiveness of antibiotics (7
). A potential alternative would involve the use of inhibitors of the expression or activity of genes that enhance marA
expression or function. This approach would offer additional advantages, such as the possibility of inhibiting only some MarA functions or the possibility of synergistic effects, since some of these genes also have a role in MDR or virulence.