MarR, the negative regulator of the Escherichia coli multiple antibiotic resistance (marRAB) operon, is a member of a newly recognized family of regulatory proteins. The amino acid sequences of these proteins do not display any apparent homologies to the DNA binding domains of prokaryotic transcription regulators and a DNA binding motif for any one of the MarR homologues is currently unknown. In order to define regions of MarR required for DNA binding, mutant repressors, selected based on their ability to interfere with (negatively complement) the activity of wild-type MarR, were isolated. As determined using gel mobility shift assays, 13 out of 14 negative complementing mutants tested were unable to bind DNA in vitro. Three negative complementing alleles presumably specify truncated repressors and one of these proteins, a 120 residue MarR, can bind DNA in vitro. Most of the negative complementing mutations were clustered within two areas of MarR with features related to a helix–turn–helix DNA binding motif. These regions are presumed to be required for the DNA binding activity of the repressor.
The chromosomal multiple antibiotic resistance (mar) locus of Escherichia coli consisting of two divergently transcribed operons, marC and marRAB, controls an adaptational response to antibiotics and other environmental hazards (Alekshun and Levy, 1997). The first structural gene, marR, specifies the negative regulator of the marRAB operon while the second, marA, specifies a transcriptional activator that controls the expression of many genes on the E. coli chromosome (Alekshun and Levy, 1997).
The marR–d mutants isolated in this study define three areas which affect MarR activity. One appears to be required for oligomerization of the repressor and the other two reveal regions with homology to HTH binding motifs of known crystal structure, although other possibilities are addressed. Particular mutations in each of these two putative DNA binding regions yield proteins that interfere with wild-type MarR activity but show no or greatly reduced repressor activity in whole cells and no binding to DNA in vitro. While a three-dimensional structure of MarR is not available, these experiments strongly suggest that MarR interacts with operator DNA through at least two distinct regions of the protein.
Characterization of negative complementing MarR mutants
A low copy number wild-type MarR expression vector [pAC–MarR (WT) (Alekshun and Levy, 1999b)] was mutagenized in vitro and the mutated plasmids were transformed into the marR+ host SPC105 (Cohen et al., 1993) bearing a chromosomal PmarII::lacZ fusion (see Experimental procedures). Increased LacZ activity in SPC105 identified the putative negative complementing trans-dominant (MarR−d) mutant proteins. The selection depends on the ability of MarR to form multimers (Martin and Rosner, 1995; Seoane and Levy, 1995; Alekshun and Levy, 1999b) and to bind DNA. Twenty-five independently identified putative mutants were isolated (Table 1 and Fig. 1). Most were subsequently tested for repressor activity in SPC107 (Table 1), a strain deleted of the mar locus but containing a chromosomal PmarII::lacZ fusion (see Experimental procedures). Fourteen negative complementing proteins were identified by their ability to interfere with the functioning of the wild-type repressor, resulting in increased β-galactosidase expression in the SPC105 host (Table 1). As seen in SPC107, the mutant proteins showed reduced or no repressor activity (Table 1). The negative complementing phenotype was attributed to subunit mixing that would occur between the wild type and mutant proteins forming relatively inactive repressors.
Table 1. Effect of marR mutations on β-galactosidase activity from a PmarII–lacZ fusion in marR+ or ΔmarR backgrounds.
marR allele on plasmid
marR + (SPC105)
ND, not determined. a. Number refers to the site of the amino acid mutated as determined by DNA sequencing; letters refer to standard single-letter amino acid abbreviations.
pAC–MarR (WT) (marR+)
Mutants with moderate repressor activity
Mutants with near wild-type repressor activity
DNA sequencing of marR in each of the mutants was performed revealing the mutated sites (Table 1). Among one group of mutants (residues 69–77), the mutation at position 73 had the most profound negative complementing trans-dominant effect (Table 1). Other mutations, identified between amino acids 94 and 121 (Fig. 1, Tables 1 and 2), also specified proteins that displayed negative complementing trans-dominant phenotypes. It was postulated that the MarR−d mutations identified between residues 69–77 and 94–121 might reside in separate DNA binding domains or that these two regions function together in order to bind DNA (see below). When comparing the single (G69E) and double (R16H/G69R) mutants, it was apparent that a single non-conservative change at position 69 is sufficient to confer the negative dominant phenotype (Table 1).
a. See Table 1. b. Phenotype refers to the activity of the MarR proteins in SPC105 and SPC107 (see Table 1): -d, trans-dominant; w, near wild type repressor activity; m, moderate repressor activity (27–70%); r, recessive. c. Steady-state amounts of proteins as determined by Western blot analysis (Fig. 2 and as described in the text): C, similar to wild-type MarR; M, moderate; L, low; U, undetectable; ND, not determined. d. This allele also contains a silent mutation (ACC → ACT at amino acid 39). e. This allele also contains a silent mutation (ACG → ACA at amino acid 128).
Most of the negative complementing proteins, except the L100F and T101I mutants, showed little if any repressor activity in the SPC107 host (Table 1). That the negative complementing activity of the Q121Ochre mutant was low but detectable (Table 1) suggested that the heteromeric repressors formed in whole cells were able, to some extent, to bind marO DNA. This finding may reflect the mutation itself or a lower steady-state level of protein (see below). These results suggest that the lesion in each mutant resides in a region critical for DNA binding, but which does not interfere with the oligomerization of MarR.
The Q42Amber marR mutant specifies a 41 residue protein that is negative complementing in the marR+ host and has no repressor activity in SPC107 (Table 1). This finding implies that the first 41 amino acids of MarR contain the majority of the contacts necessary to mediate protein–protein interactions.
Analysis of mutant protein expression in whole cells
To generate sufficient quantities of these peptides for use in electrophoretic mobility shift assays (see below), the mutant alleles were cloned into pET13a and overexpressed in E. coli BL21(DE3). Total cell lysates were prepared from these strains and proteins were subjected to Western blot analysis.
In this high-level expression system, many of the mutants were easily detected (Fig. 2 and Table 2). However, synthesis of the proteins specified by the Q42Amber and Q110Ochre and R94C, G95D, and Q121Ochre alleles were undetectable. While apparently unstable in the cell, they are synthesized at a level sufficient to confer negative dominance.
DNA binding properties of MarR−d mutants
The DNA binding activities of the negative complementing mutants were assayed using electrophoretic mobility shift assays. These experiments were performed using total cellular lysates prepared from E. coli BL21(DE3) bearing the plasmid specified mutant marR alleles (Fig. 3). MarR formed a complex with marO DNA when a total cell lysate (5 ng) from a cell overexpressing the wild-type repressor was used (Fig. 3A and B, lane 2). The addition of a 250-fold molar excess of unlabelled marO eliminated this complex demonstrating the specificity of this interaction (Fig. 3A and B, lane 3). Using 100 ng of cell lysates, DNA binding activity was not detected from control BL21(DE3) cells bearing only the plasmid vector pET13a (Fig. 3A and B, lane 4) nor lysates containing the Q42Amber, G69E, R77H, L100F, T101I, Q110Ochre and G116S (Fig. 3A, lanes 5–11) or A70T, R94C, G95D, or T39I/T72I/L135K mutants (Fig. 3B, lanes 5–8 and 11). At this same concentration, the R73C and M74I mutants were also unable to bind to marO (data not shown). Even at much higher amounts (2.5 µg) a DNA binding activity for those mutants tested was still undetected (data not shown). Moreover, this lack of DNA binding activity cannot be attributed to lack of protein expression in whole cells because mutant proteins synthesized in large amounts [A70T, M74I, R77H and G116S (Fig. 2 and Table 2)] still did not bind DNA. These findings demonstrate that the DNA binding property of these 13 mutants was impaired.
A DNA binding activity for the Q121Ochre mutant was detected using 25 and 50 ng of protein (data not shown) and was more pronounced at a concentration (100 ng) that was 20-fold greater than that needed for the wild-type MarR; this mutant protein shifted ≈ 50% of the labelled probe (Fig. 3B, lane 9). Thus, while the other two nonsense mutants, Q42Amber and Q110Ochre, appear to be devoid of a DNA binding domain, Q121Ochre contains some of the elements required for DNA binding activity. Because this mutant yields a weak negative complementing phenotype (Table 1), it may also be defective in a step subsequent to repressor binding (see below).
Most HTH motifs are 20–21 residues in length and the following parameters generally apply: stereochemical requirements exist for residues 4–5, 8–10 and 15 (italicized boldfaced numbers in Fig. 4); positions 4 and 15 are most often completely buried and, thus, should be non-polar with valine or isoleucine usually being found at position 15; position 5 is generally a small residue, such as alanine or glycine, and should not be a branched chain amino acid (Val, Leu or Ile) because these larger residues would interfere with the conformation of the HTH motif; and residue 9 is usually a small amino acid residue, either glycine (most common) or alanine (Pabo and Sauer, 1984; Branden and Tooze, 1991; Pabo and Sauer, 1992).
While the above parameters apply for many HTHs of known structure, there are exceptions and salient examples of HTH motifs of atypical primary sequence are listed in Fig. 4. For example, the lysine residue at position 5 in Trp is uncommon and its presence strains the overall structure of the HTH motif (Pabo and Sauer, 1984) and the lengths of both recognition helices in MarA (Rhee et al., 1998) are longer than those normally encountered (Fig. 4).
A putative HTH motif in MarR was identified among residues 61–80 (MarR–M) in which a hydrophobic residue (leucine) is present at positions 4 and 15 and residue 9 is glycine (Fig. 4). Unusually, serine was found at residue 5; however, MarA [in the first of two HTH motifs (Rhee et al., 1998)], TrpR, and γδ resolvase contain non-consensus amino acids (boldfaced residues in Fig. 4) at this position (Pabo and Sauer, 1984; Rhee et al., 1998).
The clustering of MarR−d mutations in the C-terminus of MarR and the trans-dominant negative complementing phenotype of the Q110Ochre and Q121Ochre mutants (considering that these alleles should specify truncated repressors with an intact HTH at MarR–M) was not anticipated if MarR had only a single region critical for DNA binding. However, upon further analysis it was found that amino acids 97–116 (MarR–C, Fig. 4) also resemble a HTH. The amino acids at positions 4 (hydrophobic, leucine) and 9 (small, alanine) within this putative HTH would adhere to the consensus, while those at positions 5 (weakly polar, threonine) and 15 (weakly polar, cysteine) do not (Fig. 4). An amino acid with a large side chain (position 5), relative to glycine or alanine, and a weakly polar residue (position 15) at these positions were not anticipated. However, the HTH of TrpR contains a lysine residue at position 5, which imposes a structural strain on the positioning of the helices within the motif (Harrison and Aggarwal, 1990) and the second HTH in MarA (Rhee et al., 1998) contains a moderately polar amino acid (threonine) at position 15 (boldfaced residues in Fig. 4).
Within known HTH motifs, amino acids at positions 11–13, 16–17 and 20 have been found to make contacts with DNA (arrows in Fig. 4) (Branden and Tooze, 1991; Pabo and Sauer, 1992). This characteristic would explain why mutations R73C (position 13) and R77H (position 17) of MarR–M and Q110Ochre (position 14) and G116S (position 20) of MarR–C (see model in Fig. 5) have a negative complementing trans-dominant phenotype. That the R73C mutation has the greatest effect suggests that the wild-type residue at this position makes a specific contact(s) with a base(s) in the operator.
Gly69 and Ala70 of MarR correspond with positions 9 and 10 of MarR–M and Gly104 corresponds to position 8 of MarR–C (Fig. 5). These residues would occur in the turns of their respective putative HTHs and the amino acids at these positions of HTH motifs of known crystal structures are critical for the correct orientation of the two helices in the motif (Pabo and Sauer, 1984; Pabo and Sauer, 1992). The size and charge of the amino acid side chains in these mutants may distort the helices and hinder DNA binding. Likewise, the L100F and T101I mutants found within MarR–C correspond to positions that are subject to stereochemical constraints and these mutations may distort the orientation of the motif (Fig. 5).
Characterization of other mutations within MarR
During the course of these experiments additional MarR mutants were identified with near wild type or moderate repressor activity (Fig. 1, Tables 1 and 2). The C47Y and P88S MarR mutants exhibit near wild type activity and the S34F/P35, T39I/T72I/L135F, A40T, V45M, S48F, A52T/V84M, T72A and V84M alleles specify moderately active repressors with activities that are between 27 and 70% of the wild-type MarR (Table 1). In addition, the T72A mutant binds marO in vitro (Fig. 3B, lane 10). For the other proteins, the results are not unexpected because most of the mutations reside either outside of the region presumed to participate in protein–protein interactions or between MarR–M and MarR–C (Fig. 1). Because the double (A52T/V84M) is slightly less active than the single (V84M) mutant, it is suggested that mutations at both residues function cumulatively to lower MarR activity or that the former is less stable in the cell (see below).
The expression of most of these other mutants except one (A52T/V84M) was easily detected from total cell lysates of SPC107 bearing the mutants in trans (data not shown). This result demonstrates that their lack of effect on the functioning of wild type MarR cannot be attributed to poor expression in whole cells.
Previous amino acid homology alignments suggested the presence of a 22 residue putative HTH DNA binding motif within CinR, a MarR homologue from Butyrivibrio fibrisolvens E14 and other MarR family members (Dalrymple and Swadling, 1997). This sequence would correspond to amino acids 64–85 and 48–69 of MarR and CinR respectively (Dalrymple and Swadling, 1997). However, this region is uncharacteristic of known HTH motifs at many critical residues. In the case of MarR, a charged polar (glutamic acid) residue would be present at position 4 and a branched chain (leucine) amino acid would occupy positions 5 and 9. With respect to the positioning of the helices within a HTH motif, a leucine residue at position 9 (within the turn) would most likely impose a structurally strain (Harrison and Aggarwal, 1990). Moreover, data regarding particular residues of CinR and other MarR family members which are required for DNA binding were lacking.
At total protein concentrations of up to 2.5 µg, we were unable to detect DNA binding for the majority of the negative complementing mutants tested (data not shown). That mutations in the first putative HTH, MarR–M, were more detrimental to the function of MarR in whole cells (Table 1) especially for the R73C mutant, which may make a specific operator contact(s), suggests that MarR–M may play a more fundamental role in DNA binding, i.e determine specificity.
A single L135F mutation in MarR results in a protein that exhibits a greater affinity for marO than wild-type MarR (Alekshun and Levy, 1999b). The T39I/T72I/L135F mutant was unable to bind marO in vitro (Fig. 3B, lane 11) and it was not negative complementing (Table 1). Because the first 42 residues, at least, of MarR are expected to participate in protein–protein interactions, it is possible that the T39I prevents the association of the mutant protein with wild-type MarR in the cell and thus does not confer a trans-dominant phenotype (Table 1). The lack of DNA binding of this mutant in vitro result is reasonable if the role of MarR–M is to determine specificity. In the triple mutant, a large hydrophobic amino acid (isoleucine) at position 72 might interfere with a specific contact(s) between arginine 73 and the operator. According to this proposal, the regions distal to MarR–M, i.e. super-repressor mutants V132M and L135F (Alekshun and Levy, 1999b), might function as non-specific or accessory binding points.
A G95S mutation in MarR greatly increases the affinity of the repressor for marO and its expression in whole cells, while substantially less (20%) than that of wild-type MarR, is easily detected (Alekshun and Levy, 1999b). However, the G95D mutation in the wild-type strain is recessive and it has no function by itself (Table 1). These results, and that of R94C, may be attributed to increased proteolysis within the cell and demonstrate the importance of these two residues in the function of MarR.
The presence of two regions required for DNA binding is also supported by the phenotype of the nonsense mutations at residues 110 (Q110Ochre) and 121 (Q121Ochre) (Table 2) and the behaviour of L100F, T101I, Q110Ochre and G116S mutants in vitro (Fig. 3A, lanes 8–11). The Q110Ochre and Q121Ochre alleles would specify a truncated MarR but each would contain an intact HTH (MarR–M) between residues 61–80. Because these mutants confer a negative complementing phenotype, it is suggested that both regions (MarR–M and MarR–C) are necessary for the DNA binding activity of MarR.
Seven of the negative complementing mutations isolated in this study (R94C, L100F, T101I, G104D, Q110Ochre, G116S and Q121Ochre) were found very near to, or in, what appears to be a second HTH, in the C-terminus of MarR (MarR–C, Fig. 5). It is currently unknown whether MarR functions by preventing binding of RNA polymerase to the mar promoter or whether the repressor interferes with the functioning of RNA polymerase after it binds to the promoter. The Arc repressor inhibits RNA polymerase isomerization (Vershon et al., 1986). The Q121Ochre mutant can still bind DNA (Fig. 3B, lane 9) but may lack the ability to interfere with transcriptional initiation or may prevent or interfere with the binding of wild-type MarR to marO.
The HTH motif is not the only DNA binding motif used by prokaryotic transcription factors. The Arc (Breg et al., 1990), Met (Rafferty et al., 1989), Pur (Schumacher et al., 1994) and Lac (Lewis et al., 1996) repressors use α-helices, β-sheets, or randomly coiled ‘hinge’ regions, in addition to, or separate from, a HTH, to bind DNA. For example, LacR and PurR use both a HTH to contact bases in the major groove and a ‘hinge’ to make minor groove contacts (Schumacher et al., 1994; Lewis et al., 1996). Mutations in each of these regions confer a trans-dominant negative complementing phenotype (Kleina and Miller, 1990; Choi and Zalkin, 1994). Compared with MarR–M, the alignment of MarR–C with the sequences of HTH motifs of known structure (Fig. 4) is suboptimal. Thus, we cannot rule out the possibility that one of these alternative structures among residues 94–121 exists in MarR.
Because the phenotype of the Q42Amber mutant is negative complementing (lacks repressor activity) (Table 1), the majority of the contacts needed to maintain protein–protein interactions between subunits of the multimer must reside within the first 41 amino acids of MarR. The lack of activity by itself in the whole cell assay (Table 1), may reflect too little protein present, e.g. instability in the cell (Table 2), or lack of DNA binding activity.
That MarR contains two HTH motifs is unusual for prokaryotic transcription factors. However, a genetic and biochemical approach was originally used to suggest the existence of two HTH motifs in AraC (Francklyn and Lee, 1988) and this proposal was later confirmed (Niland et al., 1996). Moreover, the crystal structure of MarA, an AraC homolog, provided the first direct proof of a prokaryotic transcription factor with two HTH motifs (Rhee et al., 1998).
The organization of marO also suggests that recognition by a protein with a dual HTH motif (or with two DNA binding regions) is possible. MarR protects both strands of marO, in two separate regions (sites I and II), from nuclease attack in footprinting experiments (Martin and Rosner, 1995). Based on the length of these two regions (21 bp) and the size of MarR (16 kDa) it was suggested that MarR functioned as a dimer at each of these sites (Martin and Rosner, 1995). Sites I and II are also composed of two pentameric subelements (inverted repeats) (Martin and Rosner, 1995) and therefore, exhibit two-fold rotational symmetry. Because these subelements are separated by a short (2 bp) distance, they would not occur on the same face of the DNA helix. MarR might use one HTH motif to bind nucleotides in one pentamer and the other HTH [or an alternative DNA binding structure (see above)] to contact the second pentameric sequence. Because sites I and II are symmetrical, this would allow MarR to bind to both faces of marO. While our data are suggestive, more definitive information will be gained by the structural determination of MarR.
Bacterial strains, plasmids and genetic techniques
SPC105 (Cohen et al., 1993) contains a wild-type mar locus and SPC107 has a 39 kb deletion that includes the mar locus (Seoane and Levy, 1995); both strains bear a chromosomal (AmpR) PmarII::lacZ fusion at the λ attachment site (Cohen et al., 1993; Seoane and Levy, 1995). E. coli BL21(DE3) (Novagen) was used for high level MarR expression. Mutant marR alleles were sequenced from plasmids purified using the Qiagen plasmid purification kit (Qiagen) or from PCR products generated using the mutant plasmids as template DNA and Platinum Taq DNA polymerase high fidelity according to the manufacturerís protocols (Life Technologies). DNA sequence analysis was performed in-house using an ABI automated DNA sequencer. Competent cells were prepared as previously described (Tang et al., 1994). Plasmid pAC–MarR (WT) (KanR) was constructed from pACT7 (Maneewannakul et al., 1992), a pACYC184 derivative, and is a low copy number wild-type MarR expression vector that has a p15 A origin of replication (Alekshun and Levy, 1999b). In this plasmid, synthesis of MarR is governed by the lacP1 promoter (IPTG inducible) and the wild-type MarR ribosome binding site (AGGG) and start codon (GTG). To create high level MarR expression vectors, the wild type and marR–d mutant alleles were amplified by PCR from the mutant low copy number vectors [pAC–MarR (WT) derivatives] and subsequently cloned into pET13a (Studier et al., 1990) as previously described (Alekshun and Levy, 1999b). In these plasmids, expression of marR is regulated by the T7 RNA polymerase promoter, a near consensus ribosome binding site, and a canonical start codon.
Hydroxylamine and nitrosoguanidine mutagenesis of pAC–MarR (WT) were performed as previously described (Miller, 1972). In order to maximize the isolation of independent mutants, the entire transformation mix (1 ml) was divided into 10 equal portions prior to incubation and plating. Subsequently, transformants were selected at either 30°C or 37°C and a maximum of four colonies per plate were retained for further analysis.
Isolation of mutants
Mutant marR alleles which encode peptides that can interact with and inactivate wild-type MarR are defined as negative complementing and are trans-dominant. Hydroxylamine and nitrosoguanidine mutagenized pAC–MarR (WT) plasmids were transformed into SPC105 and plated on MacConkey lactose agar containing ampicillin (100 µg ml−1), kanamycin (30 µg ml−1) and IPTG (50 µM). The host alone or containing pACT7 (Maneewannakul et al., 1992), the parent plasmid of pAC–MarR (WT), yielded weak Lac+ colonies, while cells containing pAC–MarR (WT) were Lac−. Plasmids bearing negative complementing marR mutations were initially identified as strong Lac+ colonies.
β-galactosidase assays for repressor activity
Low copy number plasmids bearing wild type or mutant marR were isolated from SPC105 and used to transform E. coli SPC107 (ΔmarR). SPC105 or SPC107 containing these plasmids were grown at 37°C to mid-logarithmic phase in LB broth, without glucose, containing the appropriate antibiotics and IPTG (50 µM). β-galactosidase assays were performed in cells permeabilized with chloroform–SDS as previously described (Miller, 1972; Seoane and Levy, 1995).
Preparation of total cell lysates and Western blot analysis to determine protein expression in whole cells
BL21(DE3) bearing marR alleles on plasmids were grown in LB broth containing the appropriate antibiotics to mid-logarithmic phase and induced with 1 mM IPTG. Cells were collected by centrifugation, washed, resuspended in 50 mM sodium phosphate (pH 7.4), 2 mM 4-(2-aminoethyl)-benzenesulfonylfluoride hydrochloride [(AEBSF, serine protease inhibitor) (Sigma)], and 5 mM EDTA, and sonicated on ice. Insoluble proteins were removed by centrifugation at 14 900 g at 4°C and total soluble protein concentration determinations were performed using the Bio-Rad protein assay kit (Bio-Rad).
Equivalent amounts of protein from control and experimental cultures were separated by SDS–PAGE on 15% gels. Western blot analysis using anti-MarR polyclonal antibodies (Alekshun and Levy, 1999b) was performed essentially as described (McDermott et al., 1998) and the reactive proteins were visualized. In some instances, the overexpressed proteins were seen directly on the Coomassie blue-stained gel.
Electrophoretic mobility shift assays
An 163 bp DNA fragment containing the entire PmarII-marO region was labelled at the 3′ end by incubation with terminal transferase and digoxigenin-11-dUTP (DIG-11-dUTP) according to the manufacturer’s protocols (Roche Molecular). DNA binding reactions (20 µl) contained 20 mM HEPES (pH 7.6), 1 mM EDTA, 10 mM (NH4)2SO4, 5 mM DTT, 0.2% Tween 20, 30 mM KCl, 50 fmol of DIG-11-dUTP-labelled DNA fragment, 1.5 µg poly(dI-dC) (Roche Molecular Biochemicals) and soluble protein from control (containing wild-type MarR) (5 ng) or experimental cellular lysates (0.1–2.5 µg). The reaction mixtures were incubated at room temperature for 15 min and then subjected to electrophoresis in prerun 6% polyacrylamide gels [22.3 mM Tris, 22.3 mM boric acid, 0.5 mM EDTA, pH 8.0 (0.25∞ TBE)] at 200 V for 2 h at 4°C. The gels were processed and detection of DIG-11-dUTP-labelled DNA was performed according to the manufacturer’s protocols (Roche Molecular Biochemicals).
This work was supported by NIH grant GM 51661. The authors wish to thank Laura McMurry and Michael Malamy for their comments on the manuscript.