Genetic Evidence for Pre-recruitment as the Mechanism of Transcription Activation by SoxS of Escherichia coli: The Dominance of DNA Binding Mutations of SoxS

doi:10.1016/j.jmb.2004.09.007

J. Mol. Biol. (2004) 344, 1–10

C OMMUNICATION

Genetic Evidence for Pre-recruitment as the Mechanism of Transcription Activation by SoxS of Escherichia coli: The Dominance of DNA Binding Mutations of SoxS Kevin L. Griffith and Richard E. Wolf Jr* Department of Biological Sciences, University of Maryland Baltimore County 1000 Hilltop Circle, Baltimore MD 21250, USA

SoxS, the direct transcriptional activator of the Escherichia coli superoxide (SoxRS) regulon, displays several unusual characteristics which suggest that it is unlikely to activate transcription by the ususal recruitment mechanism. Thus, agents that generate superoxide endogenously and thereby provoke the defense response elicit the de novo synthesis of SoxS, and with the SoxS binding site being highly degenerate, the number of SoxS binding sites per cell far exceeds the number of SoxS molecules per cell. To account for these distinctive features of the SoxRS system, we proposed “pre-recruitment” as the mechanism by which SoxS activates transcription of the regulon’s genes. In pre-recruitment, newly synthesized SoxS molecules form binary complexes with RNA polymerase in solution. These complexes provide the information content to allow the 2500 molecules of SoxS per cell to scan the 65,000 SoxS binding sites per cell for the 200 binding sites per cell that reside within SoxS-dependent promoters. As a test of whether SoxS activates transcription by recruitment or pre-recruitment, we determined the dominance relationships of SoxS mutations conferring defective DNA binding. We found that soxS DNA binding mutations are dominant to the wild-type allele, a result consistent with the pre-recruitment hypothesis, but opposite to that expected for an activator that functions by recruitment. Moreover, whereas positive control mutations of activators functioning by recruitment are usually dominant, a soxS positive control mutation was not. Lastly, with the SoxRS system as an example, we discuss the physiological requirement for stringent regulation of transcriptional activators that function by pre-recruitment. q 2004 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: gene regulation; DNA scanning; recruitment; SoxRS regulon; superoxide

Recruitment is the primary mechanism of transcription activation in bacteria. In recruitment, protein–protein interactions between a DNA-bound activator and RNA polymerase (RNAP) bring the polymerase to the promoter and thereby enhance open complex formation (Figure 1).1 Previously, we proposed that SoxS, the direct transcriptional activator of the Escherichia coli superoxide (SoxRS) Present address: K. L. Griffith, Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139, USA. Abbreviations used: RNAP, RNA polymerase; HTH, helix-turn-helix. E-mail address of the corresponding author: [email protected]

regulon,2,3 does not employ recruitment to activate target gene transcription, but rather uses a mechanism we termed “pre-recruitment”.4 In prerecruitment, the activator forms a binary complex with RNAP in solution and the binary complex then scans the chromosome for activator-dependent promoters using the DNA sequence recognition properties of the activator for its cognate binding site together with the promoter binding properties of the sigma factor for the K10 and K35 promoter hexamers (Figure 1). The utility of the pre-recruitment mechanism is that it provides a way for an activator that binds to a degenerate sequence and is present in limiting concentration relative to the number of its binding sites in the cell to discriminate between binding sites in activator-dependent

0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.

2

Figure 1. Two mechanisms of transcription activation: recruitment and pre-recruitment. In recruitment, the activator first binds to a specific sequence in the promoter region of the gene whose transcription will be activated. Through protein–protein interactions, the DNA-bound activator then recruits RNAP to the promoter and a stable open complex is formed. For the reasons given in the text, the activator is depicted as a monomeric recruitment activator (MRA). In pre-recruitment, protein–protein interactions between the activator and RNAP allow the formation of a binary complex in the absence of specific DNA binding. The binary complex then scans the chromosome for promoters with an activator binding site in the proper position and orientation to allow binding and stable open complex formation. SoxS is shown as the example of an activator that functions by pre-recruitment.

promoters and the excess of equivalent sites that are not located in promoters. The properties of SoxS are consistent with the advantages of the pre-recruitment mechanism. (1) The cellular concentration of SoxS is low under non-inducing conditions, !100 molecules per cell,4 and it is synthesized de novo in response to redoxcycling compounds like paraquat that generate superoxide endogenously.2,3 (2) The binding site for SoxS is highly degenerate (w9.5 bits of information).5 Computer analyses and direct measurements have shown that the E. coli genome contains w13,000 SoxS binding sites.4,6 Thus, since Escherichia coli cells growing rapidly in rich medium have four to six genomes,7 such cells would contain w65,000 SoxS binding sites per cell.4,6 (3) The E. coli genome contains less than 40 SoxS-dependent promoters,8 which is equivalent to about 200 SoxS-dependent promoters per fast-growing cell. Thus, SoxS binding sites lying within SoxSdependent promoters constitute !1% of the total SoxS binding sites in the cell. This ratio raises the question of how SoxS discriminates between the functional binding sites lying within SoxS-

Genetic Evidence for Pre-recruitment by SoxS

dependent promoters and the vast excess of sequence-equivalent but non-functional sites that are scattered throughout the genome. (4) For recruitment to satisfy the conditions under which SoxS functions, the number of SoxS molecules produced per cell should approach the number of SoxS binding sites. However, even though only 2500 molecules of SoxS per cell are produced upon treatment with paraquat, this small number of molecules per cell relative to the number of SoxS binding sites per cell is sufficient to activate transcription of the genes of the SoxRS regulon maximally4 and within minutes after administering an inducer.4,6 Moreover, increasing the number of SoxS molecules per cell by w100-fold, by hyper-expression of soxS from a multi-copy plasmid, does not further enhance transcription of target genes;4,9,10 this absence of effect of significantly elevated SoxS level on transcription activation strongly suggests that the concentration of SoxS produced upon paraquat treatment is not limiting for target gene expression. Consistent with this idea and with the prerecruitment hypothesis, the basal level of AraC/XylS family members11 SoxS and MarA but not Rob (all three proteins share w45% amino acid sequence identity along the common length of SoxS and activate transcription of the same set of genes, albeit to different degrees), is sufficient for enhanced activation of transcription of a zwf-lac fusion when the native SoxS binding site is replaced with an optimal DNA binding sequence (Table 1). Interestingly, although Rob is expressed constitutively at 5000–10,000 molecules per cell, transcription of the fusion carrying an optimal binding site is not enhanced (Table 1, compare strains RA4468 and GC4468). This failure to respond to an improved DNA binding site may be a consequence of Rob being sequestered into several distinct, immuno-stainable foci,12 which presumably blocks its access to the transcriptional machinery; in addition, Rob poorly activates transcription from class I promoters like zwf (K.L.G. & R.E.W., Jr, unpublished results). In previous work, we estimated that a single E. coli bacterium contains !200 molecules of SoxS and MarA combined,4 yet despite the vast excess of sequence-equivalent binding sites in the cell, this low concentration of the activators is able to enhance in vivo transcription of the “super zwf” promoter by a factor of 4 relative to transcription from the wild-type zwf promoter (Table 1). Although enhancing the affinity between an activator and its DNA target sequence might be expected to lead to increased transcription regardless of the mechanism by which the activator functioned, the fact that a total of !200 molecules of SoxS and MarA combined significantly enhanced activation despite a 325-fold excess of nonfunctional binding sites relative to functional sites is more consistent with the hypothesis of initial formation of an activator–RNAP binary complex that then scans the chromosome for activatordependent promoters, i.e. the pre-recruitment mechanism, than with the recruitment mechanism.

3

Genetic Evidence for Pre-recruitment by SoxS

Table 1. Effects of improved SoxS binding sites on transcription activation of the zwf promoter by the basal, uninduced levels of SoxS and MarA

zwf Double zwf Triple zwf Super zwf

b-Galactosidase activity (Miller units)

Sequencea

Soxbox

N8453 (soxTcat, Dmar, robTkan)

RA4468b (robTkan)

GC4468b (wild-type)

41 60 54 51

38 (0.9) 82 (1.4) 181 (3.4) 173 (3.4)

50 (1.2) 83 (1.4) 210 (3.9) 198 (3.9)

ATCGCACGGGTGGATAAGCG ATCGCACGGGTGGACAAACG ATCGCACGGATGGACAAACG ATCGCACTTAATGACAAACG

Add-on PCR was used to replace the wild-type SoxS binding site of the zwf gene of plasmid pZ519 with the optimal binding sites double, triple, or super zwf described by Griffith & Wolf.9,10 Plasmid pZ5 contains a zwf-lacZ transcriptional fusion that was recombined with phage lRS45 to create the respective fusion phages that were then used to prepare single-copy lysogens of strains GC4468 (wild-type), RA4468 (robTkan), and N8453 (soxTcat, Dmar, robTkan), as described by Griffith & Wolf.9,10 To determine the effects of the basal expression of SoxS, MarA, and Rob on zwf transcription, strains were grown in LB medium, aliquots were taken at various times, and bgalactosidase activity was determined by the high-throughput assay method of Griffith & Wolf.20 a Sequence of SoxS binding sites. Bold nucleotides are the critical elements for SoxS DNA binding, the “invariant A” at position 1, recognition element 1 (optimal sequence GCAC), and recognition element 2 (optimal sequence CAAA).9,10 Sequential modifications to the sequence are underlined. b Effects of the basal expression of SoxS and MarA on transcription of zwf. The values shown are Miller units and the values in parentheses are the induction ratios relative to the basal expression level of the triple mutant strain N8453 (soxTcat, Dmar, robTkan).

Importantly, Martin et al. have directly demonstrated the ability of SoxS and MarA to form binary complexes with RNAP in solution in vitro and they determined the Kd of the MarA–RNAP interaction to be w0.3 mM.6 Moreover, in a model which they referred to as “DNA scanning”, Martin et al. proposed that the binary complexes scan the chromosome for promoters of the regulon and that binding of the complexes to these promoters is more efficient than the independent binding of the activator and RNAP that would occur if activation was effected by recruitment.6 Although the above-described in vivo evidence that led to and is in support of the pre-recruitment hypothesis is largely circumstantial, it is consistent with the in vitro evidence for binary complex formation presented by Martin et al.6 What is lacking, however, is direct evidence for binary complex formation in vivo. In this report, we provide genetic evidence in support of the prerecruitment hypothesis by determining the dominance relationships of SoxS mutations defective in DNA binding or in positive control with respect to the wild-type soxS allele.

Experimental design The recruitment and pre-recruitment mechanisms of transcription activation make very different predictions of the dominance or recessiveness of mutations conferring defects in DNA binding or in positive control (Figure 2). With a monomeric transcription activator that functions by pre-recruitment, a mutant protein defective in DNA binding and overexpressed to a sufficient level is expected to be dominant to the wild-type protein, because the abundance of the mutant protein will be high enough to sequester most or all of the 3300 transcription–initiation-competent molecules of holo-RNAP per cell13 into binary complexes that cannot bind to and activate transcription of target

Figure 2. The type of mutation of a monomeric transcription activator that confers dominance depends on the mechanism by which it functions. With an activator that functions by the recruitment mechanism, positive control mutations are dominant to the wild-type allele. As shown, a positive control mutation (PC) of the monomeric recruitment activator (MRA) produces a protein that is able to bind normally to its specific sequence in the promoter region of the target gene but is defective in the protein–protein interaction necessary for recruitment of RNAP to the promoter. When the positive control mutant protein (MRAPC) is produced in excess to the wild-type activator (MRAWT), it binds to its specific sequence and thereby blocks recruitment of RNAP to the promoter because it excludes the binding of the wild-type activator. In contrast, a DNA binding mutation of a monomeric activator that functions by prerecruitment is dominant to the wild-type allele. With SoxS as the example, the activator defective in DNA binding (SoxSD) is able to form a normal binary complex with RNAP, but the activator–RNAP complex is unable to form a stable open complex on the promoter because of the activator ’s defect in binding to its specific target sequence. Thus, when produced in excess to the wildtype activator (SoxSWT), the mutant activator sequesters all of the available RNAP into non-functional binary complexes and thereby blocks the ability of the wild-type activator to form binary complexes with RNAP that can activate transcription from the target promoter.

4 promoters. Moreover, the pre-recruitment mechanism predicts that an overexpressed mutant protein defective in positive control will be recessive to the wild-type protein because the positive control mutant protein would be unable to form binary complexes with RNAP and, as such, would be unable to interfere with binary complex formation and transcription activation by the wild-type protein. Only when expressed to an extraordinarily high level, a level high enough to saturate all the potential DNA binding sites in the cell, would a positive control mutant of an activator functioning by pre-recruitment be dominant to the wild-type protein. Thus, with 65,000 SoxS binding sites per cell and only about 3300 molecules of holo-RNAP per cell available to initiate transcription,13 the number of molecules of the positive control mutant required to achieve dominance, w65,000,4,6 would have to far exceed the number of molecules of the DNA binding mutant, w3300, necessary to achieve dominance. In contrast to the predicted dominance of DNA binding mutants of activators that function by prerecruitment, overexpression of a DNA binding mutant of a monomeric activator that functions by recruitment is predicted to be recessive to the wildtype protein, because the mutant protein would be unable to interfere with DNA binding and transcription activation by the wild-type protein. Moreover, overexpression of a positive control mutant of an activator that functions by recruitment is expected to be dominant to the wild-type allele, as the positive control mutant protein would bind normally to the target sites in activatable promoters and, provided the abundance equals or exceeds the number of DNA binding sites, would occupy the target sites to the exclusion of the wild-type protein. Indeed, the predicted dominance of positive control mutations of activators thought to function by recruitment has been used to select for them genetically.14 In summary, for an activator functioning by recruitment, positive control mutations are transdominant negative, and DNA binding mutations are recessive, while for an activator functioning by pre-recruitment, DNA binding mutations are transdominant negative and positive control mutations are recessive. We note that these arguments apply best to monomeric activators like SoxS because oligomeric activators may form mixed oligomers when two alleles are present in the cell and these hetero-oligomers may confer a dominant-negative phenotype that would not occur with a mixture of the two types of homo-oligomers; lacIKD mutations are an example.15 In previous work, we used in vivo and in vitro tests to identify single alanine substitutions of SoxS that are defective in DNA binding or positive control.16 The wild-type and mutant soxS genes were carried on plasmid pBAD18-his6-SoxS, where their expression is under control of the arabinoseinducible promoter, PBAD, and where SoxS carries a his6 tag at its N terminus. Although the mutations

Genetic Evidence for Pre-recruitment by SoxS

were initially introduced into the his6-tagged gene to facilitate purification of mutant proteins, we subsequently determined that the tag increases the half-life of SoxS from two minutes to 26 minutes.17 For these studies, we decided to retain the his6 tag in the plasmid-borne soxS genes because, as mentioned above, the ability of a mutant allele to achieve dominance over the wild-type allele may require that the concentration of the protein product of the mutant soxS allele be not only in excess over the concentration of protein produced from the wild-type allele but, more importantly, in excess over the concentration of RNAP with which it may putatively form binary complexes.

Dominance tests of DNA binding and positive control mutations To determine the dominance relationships between wild-type soxS and soxS genes carrying mutations altering DNA binding (G32A, F42A, or F88A) or transcription activation (K30A), we introduced plasmid pBAD18-his6-SoxS carrying the respective mutations into a DaraBAD derivative of strain RGM9087, which carries a zwf-lac transcriptional fusion.16 As controls, we also introduced the empty vector, plasmid pBAD18,18 and plasmid pBAD18-his6-SoxS carrying the wild-type soxS gene, into the fusion strain. We treated brothgrown, exponential phase cultures of the strains in the following three ways: (i) with 0.5 mM paraquat alone, as a control to determine the amount of zwf-lac transcription upon induction of the chromosomally encoded wild-type SoxS protein; (ii) with 0.02% (w/v) arabinose alone, as a control to determine the amount of activity produced upon induction of the plasmid-encoded his6-SoxS proteins; and (iii) with both inducers simultaneously for determination of the amount of reporter transcription when the chromosomally encoded wild-type protein is produced in cells overexpressing the plasmid-encoded wild-type or mutant SoxS protein. Importantly, the concentrations of paraquat and arabinose were chosen as the minimal concentrations that produced maximal activation of reporter gene transcription.16,19 After one hour of inducing treatment, the amount of b-galactosidase activity produced from the zwf-lac fusion was determined in the cultures.20 We also determined the amount of basal reporter gene transcription in uninduced cultures of the respective strains. Table 2 shows the results of the analysis with the zwf-lac fusion. The first data column shows that the basal level of reporter gene transcription is independent of the soxS allele carried by the plasmid, except the level is slightly higher in the strain with the wild-type allele, perhaps because of leaky transcription from the PBAD promoter. The second column shows that the amount of transcription activation by the paraquat-induced, chromosomally encoded wild-type SoxS protein is unaffected by the

5

Genetic Evidence for Pre-recruitment by SoxS

Table 2. Dominance relationships of DNA binding and positive control mutations of soxS evaluated at the zwf promoter SoxS allele

None Wild-type K30A G32A F42A F88A

Locationa

Helix 2 Turn of HTH 1 Recognition helix of HTH 1 Recognition helix of HTH 2

b-Galactosidase specific activity (Miller units)

Defectb None

Paraquat

Arabinose

ParaquatC arabinose

Relative zwf transcriptionc

None Positive control DNA binding DNA binding

288G31 390G43 306G57 314G35 300G25

2372G313 2587G172 2621G155 2616G260 2451G156

290G18 3176G601 1451G230 318G47 309G52

2192G167 3787G296 2549G122 1369G117 1310G215

1.7 1.2 0.6 0.6

DNA binding

281G33

2569G243

453G68

1606G237

0.7

Plasmids were introduced into the previously described D(araBAD)714 derivative of D(argF-lac) Dmar strain RGM9087, which carries a transcriptional fusion of lac to the zwf promoter on a single-copy l prophage.16 Plasmids were derivatives of pBAD1818 carrying a soxS gene encoding a hexa-histidine (his6) tag at the N terminus. In these plasmids, expression of SoxS is under control of the arabinoseinducible promoter, PBAD. Plasmid pBAD18-his6-SoxS and mutant pBAD18-his6-SoxS plasmids carrying single alanine substitutions conferring defects in DNA binding, viz. mutations G32A, F42A, or F88A, or positive control, viz. mutation K30A, have been described.16 In addition to being defective in DNA binding in vivo and in vitro,16 a structural model of SoxS based on the co-crystal structure of MarA bound to mar DNA22 indicates that SoxS mutations G32A and F42A reside in the turn and the recognition helix, respectively, of the Nterminal HTH motif, while mutation F88A resides in the recognition helix of the C-terminal HTH motif. Strains harboring vector plasmid pBAD18 or plasmid pBAD18-his6-SoxS with one of the indicated soxS alleles were grown in LB broth containing ampicillin (100 mg/ml), treated with paraquat (0.5 mM), to induce synthesis of chromosomally encoded SoxS, and/or L-arabinose (0.02%), to induce plasmid-encoded his6-SoxS synthesis, and the effects of the various mutations and treatments on expression of the zwf-lac transcriptional fusion were determined by assay of b-galactosidase activity by the method of Miller33 using a micro-plate reader under high-throughput conditions,20 all as described previously.16 a Predicted position of the amino acid residue of SoxS in a model based on the crystal structure of MarA in complex with mar DNA.22 b Defect conferred by respective single alanine substitutions of his6-SoxS as described by Griffith & Wolf.16 c Specific b-galactosidase activity of strain containing pBAD18-his6-SoxS divided by specific b-galactosidase activity of strain containing pBAD18 vector in cultures treated with both paraquat and arabinose. A ratio less than unity indicates that the plasmidencoded his6-SoxS allele is dominant to the chromosomally encoded wild-type allele, while a ratio of unity or higher indicates the plasmid-encoded allele is not dominant.

presence of any of the pBAD18 plasmids. The third column shows that the extent of activation of reporter transcription by the arabinose-induced, plasmid-encoded wild-type SoxS protein is approximately the same as that produced by paraquat-mediated induction of the chromosomally encoded wild-type protein. In agreement with previous results,16 the third column also shows that the DNA binding mutations G32A, F42A, and F88A reduce reporter transcription by about tenfold, whereas positive control mutation K30A, the most severe single alanine substitution conferring a defect in positive control, reduces zwf-lac transcription about twofold. The fourth column shows the effect on reporter transcription of the simultaneous induction of the chromosomal wild-type SoxS protein and the plasmid encoded wild-type or mutant protein, i.e. the potential dominance of the mutant alleles. To quantify the potential dominant-negative effects of the plasmid-borne soxS alleles on transcription of the zwf-lac fusion by the chromosomally encoded wild-type SoxS protein, we normalized the amount of b-galactosidase activity produced by the strains with plasmid-borne wild-type or mutant soxS alleles treated with both inducers to the amount of b-galactosidase activity produced by the strain carrying the vector control treated in the same manner. The ratios are shown in the fifth column. In the strain carrying pBAD18-his6-SoxS with the wild-type soxS allele, the amount of b-galactosidase activity produced by treatment with both paraquat

and arabinose (3787 Miller units) was 1.7-fold greater than the amount of activity produced by treatment of the vector control with both inducers (2192 Miller units). We do not know why the simultaneous induction of his6-SoxS expression from plasmid pBAD18-his6-SoxS and wild-type SoxS expression from the chromosome produces more zwf transcription than induction of the chromosomally encoded soxS gene alone, since the concentration of each inducer was set at the concentration that produced maximal induction of reporter gene transcription. Similarly, we do not know why treatment of the strain carrying plasmid pBAD18-his6-SoxS with both inducers produced an amount of b-galactosidase activity (3787 Miller units) greater than the activity produced when the strain was treated with paraquat alone (2587 Miller units) or arabinose alone (3176 Miller units). Nonetheless, we expected that the amount of zwf transcription produced by expression of the chromosomally encoded wild-type SoxS protein in the presence of a greater abundance of a mutant SoxS protein expressed from a strongly dominant-negative allele of soxS would be significantly less than the amount of transcription produced by expression of the chromosomally encoded wild-type allele alone; i.e. expression of a dominant-negative soxS allele should interfere with expression of the wildtype allele and therefore produce a value of relative transcription less than unity. In contrast, a soxS allele that is not dominant should produce a value of relative transcription equal to or greater than unity. As shown in Table 2, soxS mutations conferring

6

Genetic Evidence for Pre-recruitment by SoxS

severe defects in DNA binding were dominant while positive control mutation K30A was not. Specifically, simultaneous induction of the chromosomally encoded wild-type SoxS protein and plasmid encoded SoxS proteins with DNA binding mutations G32A, and F42A, which reside in the N-terminal helix-turn-helix (HTH) motif of SoxS, and F88A, which resides in the C-terminal HTH motif, produced relative transcription values significantly less than unity (0.6, 0.6, and 0.7, respectively), while simultaneous induction of the wild-type protein and the mutant protein with positive control mutation K30A produced a relative transcription value slightly greater than unity (1.2). In other words, overexpression of mutant SoxS proteins with defects in DNA binding interfered with transcription activation by wild-type SoxS while overexpression of the positive control mutant of SoxS did not. We also carried out similar experiments with an fpr-lac transcriptional fusion (Table 3). As with zwf, DNA binding mutations G32A, F42A, and F88A were dominant-negative to the wild-type allele while positive control mutation K30A was not. In summary, the dominant-negative effects of three DNA binding mutations of soxS on transcription activation of two SoxS-dependent promoters are consistent with pre-recruitment as the mechanism of transcription activation by SoxS and inconsistent with the recruitment mechanism.

Caveats, limitations, and alternative interpretations of the dominance relationships There are several possible caveats to this conclusion. Firstly, the observed dominance of the

DNA binding mutations could be due to the presence of the his6 tag in the mutant protein, but this seems unlikely because the positive control mutation was not dominant even though the mutant protein also carried the his6 tag. Secondly, the dominance of the DNA binding mutations could be an artifact if the mutations somehow increased the half-life, and hence the abundance of the mutant protein to a level so far above that of the chromosomally encoded wild-type SoxS protein, that the mutant protein was able to interfere with the action of the wild-type protein, e.g. by aggregating with it. Similarly, the failure to observe dominance of the positive control mutation might also be an artifact if the mutation were to decrease the half-life and hence the abundance of the his6SoxS mutant protein carrying the K30A substitution to a level too low to allow it to exert its putative dominance. To test these latter possibilities, we used our previously described Western blotting procedure4 to determine the abundance of the plasmidencoded wild-type and mutant his6-SoxS proteins in the various strains relative to the abundance of the chromosomally encoded wild-type SoxS. Figure 3(A) shows a representative Western blot of the native SoxS protein produced when the strains were induced with paraquat and the plasmidencoded his6-SoxS proteins produced when the strains were induced with arabinose. The relative abundance data shown in Figure 3(B) were compiled as the averages of five independent Western blot experiments. We found that the abundance of the plasmid-encoded, wild-type his6-SoxS protein was 52-fold higher than that of the chromosomally encoded wild-type protein (Figure 3B). The observed over-abundance of the

Table 3. Dominance relationships of DNA binding and positive control mutations of soxS evaluated at the fpr promoter SoxS allelea

None Wild-type K30A G32A F42A F88A

Locationb

Helix 2 Turn of HTH 1 Recognition helix of HTH 1 Recognition helix of HTH 2

b-Galactosidase specific activity (Miller units)d

Defectc None

Paraquat

Arabinose

ParaquatC Arabinose

Relative fpr Transcriptione

None Positive control DNA binding DNA binding

195G13 319G56 233G33 214G28 201G26

2892G196 2851G225 2991G193 2944G205 3099G337

175G13 3660G802 993G108 167G19 252G41

2473G93 4450G277 2604G542 785G92 1359G206

1.8 1.1 0.3 0.6

DNA binding

247G10

2876G186

215G13

1674G291

0.7

a The fusion strain carried plasmid pBAD18-his6-SoxS with either the wild-type allele of his6-SoxS or single alanine substitutions of it; the empty vector, plasmid pBAD18, served as a control. b Predicted position of the amino acid residue of SoxS in a model based on the crystal structure of MarA in complex with mar DNA.22 HTH 1 is comprised of helices 2 and 3 while HTH 2 is comprised of helices 5 and 6. c Defect conferred by respective single alanine substitutions of his6-SoxS as described by Griffith & Wolf.16 d Cultures of a DaraBAD derivative of strain RGM9082 carrying an fpr-lac transcriptional fusion and harboring vector plasmid pBAD18 or plasmid pBAD18-his6-SoxS with the indicated soxS allele were grown in LB medium and treated with 0.5 mM paraquat or 0.02% L-arabinose or both. The effects of the various mutations and treatments on expression of the fpr-lac fusion were determined by assay of b-galactosidase activity as described in Table 2. e Specific b-galactosidase activity of strain containing pBAD18-his6-SoxS divided by specific b-galactosidase activity of the strain containing pBAD18 vector in cultures treated with both paraquat and arabinose. A ratio less than unity indicates that the plasmidencoded his6-SoxS allele is dominant to the chromosomally encoded wild-type allele, while a ratio of unity or higher indicates the plasmid-encoded allele is not dominant.

Genetic Evidence for Pre-recruitment by SoxS

7

Figure 3. Determination of the abundance of plasmid-encoded wild-type and mutant his6-SoxS proteins relative to the abundance of chromosomally encoded native SoxS. A, Western blot of native SoxS induced with paraquat and plasmidencoded wild-type and mutant his6-SoxS induced with arabinose. Except for the absence of urea in the sonication buffer, the cells were grown, extracts were prepared, and Western blotting was carried out as described previously.4,17 Briefly, cultures of strain RGM9087 harboring plasmids pBAD18, pBAD18-his6-SoxS, or pBAD18-his6-SoxS carrying either a DNA binding mutation (G32A, F42A, F88A) or the positive control mutation (K30A) were grown overnight in LB medium containing ampicillin (100 mg/ml) at 37 8C. The cultures were diluted 1 : 100 into fresh medium in duplicate and grown to A600Z0.1, at which point, one culture of each pair was treated with 0.5 mM paraquat to induce chromosomally encoded native SoxS while the other was treated with 0.02% arabinose to induce plasmid pBAD18-encoded wild-type or mutant his6-SoxS. After one hour incubation at 37 8C, 1 ml samples were transferred to ice-cold tubes for determination of the A600 of the cultures; in addition, 10 ml samples were transferred to ice-cold tubes and the cells were collected immediately by centrifugation. The cell pellets were resuspended in 400 ml of sonication buffer (50 mM Tris–HCl (pH 7.9), 3 mM DTT, 1 mM EDTA) and subjected to sonication with a Branson sonifier for two pulses of one minute. The sonicates were centrifuged at 13,000g for 30 minutes to remove insoluble SoxS and cell debris, and the supernatant fluid was transferred to a microfuge tube containing 200 ml of Laemmli gel loading buffer. Then 10 ml of each sample was loaded into the well of an 18% acrylamide Tris-glycine gel and subjected to SDS-PAGE. Western blotting with polyclonal anti-SoxS serum and quantification of the bands containing SoxS or his6-SoxS were carried out as described previously.4,17 Control experiments were conducted to ensure that the signal intensity was proportional to protein input. The intensity of the SoxS signals in the Western blots (in arbitrary units) was normalized to the A600 of the culture. B, The abundance of plasmid-encoded wild-type or mutant his6-SoxS induced with arabinose to the abundance of chromosomally encoded native SoxS induced with paraquat. For each strain, the abundance of the his6-SoxS protein relative to that of the native SoxS protein was determined by dividing the value of signal intensity per A600 for the arabinose-induced culture by the value of signal intensity per A600 for the paraquat-induced culture.

plasmid-encoded protein is likely due to a combination of the increased stability of his6-SoxS relative to that of native SoxS17 and the increased dosage of the plasmid-encoded gene. The relative abundance obtained is an underestimate because the values were determined for SoxS present in the soluble fraction of the cell extract, whereas a significant

portion of plasmid-encoded his6-SoxS is found in inclusion bodies. We chose to determine the relative abundance of soluble SoxS (both native SoxS and his6-SoxS) rather than total SoxS because soluble SoxS is the form of the protein most likely to be accessible to chromosomal DNA and RNA polymerase in the cell, and hence to be most meaningful

8 in the analysis of the putative effects of the relative abundance of the respective proteins on the dominance relationships. In contrast to the possibility suggested above, the DNA binding mutations G32A, F42A, and F88A reduced the abundances of the plasmid-encoded his6-SoxS mutant proteins relative to that of the chromosomally encoded wild-type protein to 31-fold (0.6 of wild-type his6-SoxS), 17-fold (0.3 of wild-type his6-SoxS), and 49-fold (0.9 of wild-type his6-SoxS), respectively. Moreover, the positive control mutation K30A increased the relative abundance of the plasmid-encoded his6-SoxS mutant protein to 61-fold (1.2 of wild-type his6-SoxS). Thus, with the relative abundance of the DNA binding mutant proteins being about half that of the positive control mutant protein, the dominance of the former and the recessiveness of the latter are not likely to be due to differences emanating from differences in their relative abundances. In work to be reported elsewhere (I. M. Shah & R.E.W., Jr, unpublished results), we show that the effects of the mutations on relative abundance are due to effects of the mutations on the half-life of the his6-SoxS proteins. As mentioned above, dominance tests like those carried out here are most meaningful if the DNA binding protein functions as a monomer. For example, a DNA binding mutation of a homodimeric activator functioning by recruitment might appear dominant if the overexpression of the mutant protein relative to the wild-type protein sequestered all of the wild-type monomers into heterodimers, with the result that no wild-type, homodimeric activator able to bind DNA efficiently would be formed. Two lines of evidence indicate that SoxS binds DNA as a monomer. Firstly, with an electrophoretic mobility shift assay, we showed that SoxS and the closely related Rob protein (55% identity over the length of SoxS) bind DNA as monomers.21 Secondly, the crystal structure of MarA in complex with target DNA,22 as well as the structure of a Rob-DNA co-crystal,23 show only monomers bound to DNA; we infer that a co-crystal structure of SoxS in complex with target DNA would also show SoxS binding DNA as a monomer. Lastly, we point out that overexpression of one allele relative to a second allele is not the classical manner in which dominance relationships have been determined, since usually the two gene products are expressed to approximately equal levels. However, as mentioned above, it was necessary in these experiments to overproduce the SoxS mutants defective in DNA binding to levels high enough to allow them to potentially sequester most or all of the RNAP into binary complexes. Nonetheless, and despite this imbalance, the results of the genetic tests presented here (Tables 2 and 3) are consistent with pre-recruitment as the mechanism used by SoxS to activate transcription and inconsistent with recruitment as the activation mechanism. Moreover, to our knowledge, no example has been reported in the literature of the

Genetic Evidence for Pre-recruitment by SoxS

dominance of a DNA binding mutation for any monomeric transcription activator, including activators thought to function by recruitment, nor has recessiveness been reported for a positive control mutation of a monomeric activator that appears to function by recruitment.

Putative requirement for stringent regulation of activators that function by pre-recruitment: the SoxRS, Mar, and Rob systems as examples We argue here that the level and/or accessibility of activators that function by the pre-recruitment mechanism are likely to be subject to tight regulation. While most activators function by recruitment,1 are expressed constitutively (i.e. present under all growth conditions), and are activated/deactivated by either small molecule effectors or by modifications conferred by signal transduction pathways, SoxS and MarA are synthesized de novo in response to their respective inducing signals and have no known ligands that affect their activity.2,3,24,25 Moreover, following removal of the specific inducer and alleviation of the imposed stress, expression of the member genes of the SoxRS and Mar regulons quickly returns to the basal level, because the respective systems are reset by active processes: first, de novo synthesis of the activator ceases; then, the residual activator, which is intrinsically unstable (t½Z1–2 minutes), is rapidly degraded by Lon protease.17 Similarly, while most activators are soluble and accessible to the chromosome and the transcription machinery, Rob is synthesized constitutively26 but is sequestered into a small number of discrete aggregates12 that presumably prevent its access to the transcription machinery. Recent work has shown that transcription of Rob-activatable promoters can be enhanced several fold by treatment of cells with dipyridyl or bile salts and that these ligands interact with the C-terminal region of the protein.27,28 Thus, it is reasonable to infer that transcription activation by constitutively expressed, physically sequestered Rob is regulated by small molecule inducers that release it from the aggregated state. In work to be presented elsewhere, we provide evidence in support of this mechanism. Other parameters may distinguish activators that function by recruitment from those that function by pre-recruitment. For example, although quantitative examples are lacking, it is likely that the number of molecules per cell of most activators functioning by recruitment approaches the number of binding sites per cell; in contrast, the number of molecules per cell of SoxS and MarA is far less than the number of binding sites per cell.4,6 In addition, activators functioning by pre-recruitment might be expected to have a lower intrinsic binding affinity for their DNA targets than activators that function by recruitment, as the overall binding energy of the pre-recruitment activators would be enhanced by combining the binding energy of the

9

Genetic Evidence for Pre-recruitment by SoxS

activator–DNA interaction with the binding energy of the interaction between holo-RNAP and the K10 and K35 promoter elements. In turn, as stated above, binary complex formation would allow the activator to identify the binding sites that reside in activator-dependent promoters and distinguish them from the 325-fold excess of equivalent sites that do not. Thus, if SoxS and MarA can indeed form binary complexes with RNAP in the absence of DNA, any physiological condition affecting their regulation or the intrinsic properties described above might adversely affect cell growth or survival, because target genes would become activated under inappropriate conditions and, as described below, RNAP would be diverted from transcription of essential to non-essential genes.

Perspectives Given that mechanisms of gene regulation in bacteria have been studied for more than 40 years, it is interesting to consider why other examples of pre-recruitment have not been documented. One possibility is that pre-recruitment is unique to the SoxS/MarA/Rob subset of proteins within the AraC/XylS family.11,29,30 Alternatively, the caveats mentioned above may have obscured the identification of other transcription activators that function by pre-recruitment. For example, the binding sequences of many eukaryotic transcription factors have relatively low information content and DNA binding specificity is often gained by the formation of multi-protein complexes, e.g. the enhanceosome.31 Thus, such eukaryotic transcription factors can be thought of as using the pre-recruitment pathway of transcription activation and, in this respect, SoxS, MarA, and Rob in their role as cosigma factors may better resemble eukaryotic transcription factors than the typical bacterial transcription activator. Indeed, in work presented elsewhere,32 we show that SoxS interacts specifically with the “265 DNA binding determinant” of the C-terminal domain of the RNAP a subunit, the RNAP domain previously known to be important only for binding the UP element, e.g. in ribosomal RNA promoters, and for binding DNA in factormediated transcription activation, e.g. CAP-dependent activation of lac. This unusual protein–protein interaction with the a-CTD occurring during oxidative stress allows SoxS to re-direct RNAP from the “strong” promoters representing 60% of total transcription in fast growing cells to the SoxSdependent promoters of genes whose products will eliminate the oxidative stress, repair the damage caused by it, and restore the cell’s redox potential. The benefit to cell survival of the redeployment of RNAP by formation of SoxS-RNAP binary complexes is yet another bit of evidence in support of the pre-recruitment hypothesis. Moreover, by this mechanism of action, SoxS and MarA can be considered as stress-inducible, co-sigma factors.

Acknowledgements We thank I. M. Shah for carrying out the experiment shown in Figure 3, and we thank R. G. Martin and J. L. Rosner for sharing information prior to publication. We also thank D. M. Eisenmann, R. H. Ebright, S. J. W. Busby, and M. C. O’Neill for critical reading of the manuscript. This work was supported by Public Health Service grant GM27113 from the National Institutes of Health.

References 1. Ptashne, M. & Gann, A. (1997). Transcriptional activation by recruitment. Nature, 386, 569–577. 2. Ama´bile Cuevas, C. F. & Demple, B. (1991). Molecular characterization of the soxRS genes of Escherichia coli: two genes control a superoxide stress regulon. Nucl. Acids Res. 19, 4479–4484. 3. Wu, J. & Weiss, B. (1991). Two divergently transcribed genes, soxR and soxS, control a superoxide response regulon of Escherichia coli. J. Bacteriol. 173, 2864–2871. 4. Griffith, K. L., Shah, I. M., Myers, T. E., O’Neill, M. C. & Wolf, R. E., Jr (2002). Evidence for “pre-recruitment” as a new mechanism for transcription activation in Escherichia coli: the large excess of SoxS binding sites per cell relative to the number of SoxS molecules per cell. Biochem. Biophys. Res. Commun. 291, 979–986. 5. Wood, T. I., Griffith, K. L., Fawcett, W. P., Jair, K.-W., Schneider, T. D. & Wolf, R. E., Jr (1999). Interdependence of the position and orientation of SoxS binding sites in the transcriptional activation of the class I subset of Escherichia coli superoxide-inducible promoters. Mol. Microbiol. 34, 414–430. 6. Martin, R. G., Gillette, W. K., Martin, N. I. & Rosner, J. L. (2002). Complex formation between activator and RNA polymerase as the basis for transcription activation by MarA and SoxS in Escherichia coli. Mol. Microbiol. 43, 355–370. 7. Bremer, H. & Dennis, P. P. (1996). Modulations of chemical composition and other parameters of the cell by growth rate. In Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology (Neidhardt, F. C., Curtiss, R., Ingraham, J. L., Lin, E. C. C., Low, K. B., Magasanik, B. et al., eds), vol. 2.2, American Society for Microbiology, Washington, DC. 8. Martin, R. G. & Rosner, J. L. (2002). Genomics of the marA/soxS/rob regulon of Escherichia coli: identification of directly activated promoters by application of molecular genetics and informatics to microarray data. Mol. Microbiol. 44, 1611–1624. 9. Griffith, K. L. & Wolf, R. E., Jr (2001). Systematic mutagenesis of the DNA binding sites for SoxS in the Escherichia coli zwf and fpr promoters: identifying nucleotides required for DNA binding and transcription activation. Mol. Microbiol. 40, 1141–1154. 10. Griffith, K. L. & Wolf, R. E., Jr (2001). Systematic mutagenesis of the DNA binding sites for SoxS in the Escherichia coli zwf and fpr promoters: identifying nucleotides required for DNA binding and transcription activation. Erratum. Mol. Microbiol. 42, 571. 11. Gallegos, M.-T., Schleif, R., Bairoch, A., Hofmann, K. & Ramos, J. L. (1997). AraC/XylS family of transcriptional regulators. Microbiol. Mol. Biol. Rev. 61, 393–410. 12. Talukder, A. A., Hiraga, S. & Ishihama, A. (2000). Two

10

13. 14. 15.

16.

17.

18.

19.

20.

21.

22.

Genetic Evidence for Pre-recruitment by SoxS

types of localization of the DNA-binding proteins with the Escherichia coli nucleoid. Genes Cells, 5, 613–626. Ishihama, I. (2000). Functional modulation of Escherichia coli RNA polymerase. Annu. Rev. Microbiol. 54, 499–518. Bell, A. & Busby, S. (1994). Location and orientation of the activating region in the Escherichia coli transcription factor FNR. Mol. Microbiol. 11, 383–390. Gilbert, W. & Muller-Hill, B. (1970). The lactose repressor. In The Lactose Operon (Beckwith, J. R. & Zipser, D., eds), pp. 93–109, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY. Griffith, K. L. & Wolf, R. E., Jr (2002). A comprehensive alanine scanning mutagenesis of the Escherichia coli transcriptional activator SoxS: identifying amino acids important for DNA binding and transcription activation. J. Mol. Biol. 322, 237–257. Griffith, K. L., Shah, I. M. & Wolf, R. E., Jr (2004). Proteolytic degradation of Escherichia coli transcription activators SoxS and MarA as the mechanism for reversing the induciton of the superoxide (SoxRS) and multiple antibiotic resistance (Mar) regulons. Mol. Microbiol. 51, 1801–1816. Guzman, L.-M., Belin, D., Carson, M. J. & Beckwith, J. R. (1995). Tight regulation, modulation, and highlevel expression by vectors containing the arabinose PBAD promoter. J. Bacteriol. 177, 4121–4130. Fawcett, W. P. & Wolf, R. E., Jr (1995). Genetic definition of the Escherichia coli zwf “soxbox,” the DNA binding site for SoxS-mediated induction of glucose 6-phosphate dehydrogenase in response to superoxide. J. Bacteriol. 177, 1742–1750. Griffith, K. L. & Wolf, R. E., Jr (2002). Measuring b-galactosidase activity in bacteria: cell growth, permeabilization, and enzyme assays in 96-well arrays. Biochem. Biophys. Res. Commun. 290, 397–402. Jair, K.-W., Yu, X., Skarstad, K., Tho¨ny, B., Fujita, N., Ishihama, A. & Wolf, R. E., Jr (1996). Ambidextrous transcriptional activation of promoters of the superoxide and multiple antibiotic resistance regulons by Rob, a binding protein of the Escherichia coli origin of chromosomal replication. J. Bacteriol. 178, 2507–2513. Rhee, S., Martin, R. G., Rosner, J. L. & Davies, D. R. (1998). A novel DNA-binding motif in MarA: the first structure for an AraC family transcriptional activator. Proc. Natl Acad. Sci. USA, 95, 10413–10418.

23. Kwon, H. J., Bennik, M. H. J., Demple, B. & Ellenberger, T. (2000). Crystal structure of the Escherichia coli Rob transcription factor in complex with DNA. Nature Struct. Biol. 7, 424–430. 24. Cohen, S. P., Levy, S. B., Foulds, J. & Rosner, J. L. (1993). Salicylate induction of antibiotic resistance in Escherichia coli: activation of the mar operon and a mar-independent pathway. J. Bacteriol. 175, 7856–7862. 25. Ariza, R. R., Cohen, S. P., Bachhawat, N., Levy, S. B. & Demple, B. (1994). Repressor mutations in the marRAB operon that activate oxidative stress genes and multiple antibiotic resistance in Escherichia coli. J. Bacteriol. 176, 143–148. 26. Skarstad, K., Tho¨ny, B., Hwang, D. & Kornberg, A. (1993). A novel binding protein of the origin of the Escherichia coli chromosome. J. Biol. Chem. 268, 5365–5370. 27. Rosenberg, E. Y., Bertenthal, D., Nilles, M. L., Bertrand, K. P. & Nikaido, H. (2003). Bile salts and fatty acids induce the expression of Escherichia coli AcrAB multidrug efflux pump through their interaction with Rob regulatory protein. Mol. Microbiol. 48, 1609–1619. 28. Rosner, J. L., Dangi, B., Gronenborn, A. M. & Martin, R. G. (2002). Posttranscriptional activation of the transcriptional activator Rob by dipyridyl in Escherichia coli. J. Bacteriol. 184, 1407–1416. 29. Martin, R. G. & Rosner, J. L. (2001). The AraC transcriptional activators. Curr. Microbiol. 4, 132–137. 30. Egan, S. M. (2002). Growing repertoire of AraC/XylS activators. J. Bacteriol. 184, 5529–5532. 31. Carey, M. (1998). The enhanceosome and transcriptional synergy. Cell, 92, 5–8. 32. Shah, I. M. & Wolf, R. E., Jr (2004). Novel protein– protein interaction between Escherichia coli SoxS and the DNA binding determinant of the RNA polymerase a subunit: SoxS functions as a co-sigma factor and redeploys RNA polymerase from UP-element-containing promoters to SoxS-dependent promoters during oxidative stress. J. Mol. Biol. 343, 513–532. 33. Miller, J. H. (1972). Experiments in Molecular Genetics. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY.

Edited by R. Ebright (Received 1 June 2004; received in revised form 3 August 2004; accepted 7 September 2004)