Mapping LexA regulons in E. coli
The full extent of LexA regulons was largely unexplored in any organism until computational and molecular genetic methods were combined with extensive genomic sequence data (Lewis et al., 1994). Systematic approaches to elucidate bona fide binding sites for a given regulator open many possibilities to explore the full extent of regulons such as LexA. Such studies have recently added to the list of LexA regulated genes in both E. coli and Bacillus subtilis, bringing the number of experimentally confirmed LexA regulated genes to 31 and 33 in these organisms respectively (Fernandez De Henestrossa et al., 2000; Au et al., 2005).
Extensive studies in E. coli using sophisticated sequence search algorithms coupled with northern analysis of post-induction expression profiles led to the discovery of seven new LexA-regulated genes (ysdAB, dinQ, hokE, ydjM, dinS and ydfE) and the experimental confirmation of three other genes (yjiW also known as sosC or dinL, ydjQ also known as sosD or dinM, and molR also known as sosF or dinO) previously hypothesized to be LexA-regulated (Fernandez De Henestrossa et al., 2000). A key feature of the experimental strategy was the examination of candidate gene expression by northern analysis using RNA extracted from three isogenic strains with variant lexA alleles. Application of this powerful genetic filter to nearly 70 putative LexA-regulated genes resulted in the reduction to only those 10 genes described above. Genes considered to be regulated by LexA would be expected to be: (i) induced in wild-type strains by DNA damage when exposed to mitomycin C, but show only basal expression levels in the absence of drug, (ii) fail to be induced in a strain with a dominant negative lexA3 Ind– (G85D that cannot autocleave and derepress), and (iii) be constitutively induced in a strain with lexA51 Def (G85D, but with a frameshift after amino acid 160 that impairs dimerization and stable DNA association). Most of these newly identified LexA-regulated genes unfortunately still have unknown function. An exception is ydjQ which was shown to be a UvrC analogue, now named Cho, and which now points to the exciting existence of an alternative nucleotide excision repair pathway (Moolenaar et al., 2002; see Van Houten et al., 2002 for detailed comments).
Other genome wide approaches have subsequently examined the E. coli LexA regulon or the global response to DNA damaging agents, but used different experimental conditions: for example, short UV pulses, or continuous exposure to mitomycin C (Courcelle et al., 2001; Khil and Camerini-Otero, 2002; Quillardet et al., 2003). The study conducted by Courcelle and coworkers is particularly noteworthy because it described a kinetic snapshot of gene expression changes over the course of an hour following exposure to a short UV pulse. The study revealed a range of induction amplitudes and temporal organization of gene expression. The compilation of differentially expressed genes that were induced by UV, but that were not induced in lexA1 strains, revealed a considerable number of hypothetical genes and genes with functions unrelated to DNA damage repair or tolerance. A future challenge will be to explain what role these LexA-regulated genes perform. Some progress may already be appreciated through work on molecular chaperones in E. coli. For instance, among those genes strongly upregulated in response to UV exposure is ybeW, also called hscC, and now known to encode a third heat shock protein (Hsp70) family member (Kluck et al., 2002). Interestingly, Kluck et al. (2002) first characterized HscC as a specialized stress chaperone and noted that hscC deletion strains were hypersensitive to UV irradiation and cadmium. An extensive genetic analysis of an hscC knockout showed a slow growth phenotype that was rapidly suppressed suggesting that under non-stressed conditions the gene plays in important role in bacterial fitness. It is therefore tempting to speculate that this chaperone interacts with proteins involved in orchestrating protection, or recovery, from endogenous or induced DNA damage.
LexA binding site in B. subtilis
An extensive mapping approach was also applied to chart the extent of the LexA regulon in B. subtilis, where until recently only five LexA-regulated genes were known (Au et al., 2005; Groban et al., 2005). In the first step, analytical methods were developed to define the LexA binding site using a quantitative gel shift assay. The thermodynamically derived consensus was then used in a genome-wide search for LexA operator sequences that turned up 39 sites in the genome and 33 located in promoter regions. To confirm LexA binding, the corresponding DNA regions were amplified and used in competition gel shift assay. Finally, candidate genes apparently regulated by LexA were tested for SOS induction by transcription profiling in microarray analyses using both UV- and mitomycin C-treated recA+ and recA deletion strains. The result of this heroic effort was the identification of 33 SOS-induced genes. Remarkably, comparative analysis with the E. coli LexA regulon revealed surprisingly little overlap with known LexA controlled genes. Only eight genes in the B. subtilis LexA regulon have homologous counterparts in E. coli (recA, lexA, ruvA, ruvB, uvrA, uvrB, uvrC and pcrA also known as uvrD) despite the presence of comparable numbers of LexA-regulated genes in both organisms. This important finding illustrates how regulatory circuits can easily diverge during the course of evolution and how little is known of the extent of LexA regulons if detailed study is conducted in only a handful of model organisms.
Three genes uncovered in the study of the B. subtilis LexA regulon do not encode products obviously involved in DNA damage repair. These are aprX, encoding a subtilisin-like protease, cwlD, encoding N-acetylmuramoyl-l-alanine amidase, and licA, encoding a component of the lichenan phophotransferase system (Au et al., 2005). Of the three, the cell-wall hydrolase appears to be the most interesting because it plays a pivotal role in the development of a mature spore cortex (Sekiguchi et al., 1995; Cowan et al., 2003). In the absence of cwlD, stable and normally dehydrated spores are produced but cannot germinate. This suggests an intriguing extension of the classic SOS responses to repair DNA damage by ensuring concomitant upregulation of a crucial gene product necessary for spore formation and survival.
Mapping E. coli LexA sites in vivo by chromatin immunoprecipitation and microarray
The application of in vitro methods to detect protein–DNA interaction suffers, of course, the inherent limitations of working with cell-free systems. The choice of buffer, probe, component concentrations, the role of additional proteins, and the absence of native chromosomal architecture impose constraints that limit the ability to detect potential binding sites. A landmark study using chromatin immunoprecipitation coupled with microarray analysis (so called ChIP on CHIP methodology) was recently reported in an effort to map the ensemble of E. coli LexA binding sites in vivo on a genome-wide scale (Wade et al., 2005). Three significant findings emerged from this study. First, 25 LexA sites were identified (termed Class I) which were concordant with previously described LexA binding sites identified using in vitro or genetic methods. Second, new LexA targets were identified that do not bind LexA in vitro and which were subdivided into two additional classes: those five targets which have LexA binding motifs (Class II), and 19 that do not have an obvious LexA motif (Class III). Classes II and III are therefore binding sites that may require additional factors and/or appropriate chromatin architecture to bind detectably in vivo. These in vivo results nearly double therefore, the known sites to which LexA binds and significantly extends our knowledge of the LexA regulon, perhaps in its entirety. Finally, the principal aim of the study revealed that the E. coli genome is globally accessible to the binding of LexA despite the presence of nucleoid-associated proteins, a finding in contrast to eukaryotic organisms where chromatin architecture restricts the binding of transcription factors and at a given time only a fraction of potential binding sites may be accessible. Indeed, the 24 base pair LexA binding site upstream of the E. coli sulA promoter could be introduced upstream of seven other widely dispersed regions in the chromosome, most transcriptionally inactive under the conditions of the assay, yet still bind LexA in vivo with less than a twofold difference when compared with LexA binding at the native sulA locus. The simple conclusion is that LexA binding in E. coli is not contingent upon the transcriptional activity of the region surrounding the binding site.
Studies such as this reveal that many global regulons may be defined in prokaryotic organisms, perhaps completely, if binding site information derived from in vitro data is merged with global genomic techniques to probe in vivo binding site occupancy. Reductions in microarray costs and the development of custom gene chips with intergenic regions or extensive tiled arrays for other microorganisms would certainly open the door to detailed analysis of other global transcriptional regulators.
Alternate LexA-like regulons
Although LexA is widespread among bacteria, it is not ubiquitous and other regulators might control SOS-like responses in those organisms that lack LexA (Eisen and Hanawalt, 1999). Some organisms appear have evolved a system that strikingly parallels LexA in regulatory features such as RecA-dependent autocleavage in response to DNA damage and derepression (Savijoki et al., 2003). For example, the HdiR protein (heat shock and DNA damage-induced regulator of Lactococcus lactis mediates responses to mitomycin C as well as heat shock. The extent of the HdiR regulon is unknown but it includes umuC although not recA, suggesting that some portions of an SOS regulon might fall under HdiR control in this organism.
Other curiosities are bound to emerge as more genomes are sequenced. Phylogenetic studies showed that lexA is duplicated in several species each of Pseudomonas and Xanthomonas (Abella et al., 2004; Yang et al., 2005; Erill et al., 2004). An intriguing finding in Pseudomonas putida is that a second lexA (lexA2) is tightly associated with, and in some cases, regulates a multigene DNA damage-inducible cassette (Abella et al., 2004). This finding suggests that LexA regulons have considerable evolutionary complexity and, indeed, might prove useful in charting the emergence and distribution of DNA repair gene clusters (Campoy et al., 2005; Erill et al., 2006). In yet another curious discovery, the study of inducible DNA repair genes in Mycobacterium tuberculosis showed that numerous genes are under the dual control of both a RecA/LexA-dependent pathway and a RecA-independent pathway (Davis et al., 2002; Rand et al., 2003).
A LexA paralogue regulates carbon-controlled genes in Synechocystis
The finding that some LexA-regulated genes are ostensibly unrelated to DNA repair and damage tolerance functions is an emerging theme as the studies described above revealed when LexA regulons were examined at the genome-wide level. New studies also point to evolutionary divergence of a LexA-like protein that is not SOS-inducible and which exerts control over other gene sets. Domain and coworkers (Domain et al., 2004) reported the finding that a LexA paralogue in the cyanobacterium Synechocystis spp. PCC6803 regulates a set of carbon-controlled genes and is crucial to cells facing carbon starvation. Careful examination of the LexA protein sequence in this organism revealed the absence of a canonical autocleavage site and the absence of a key reactive serine that sets it apart from most other LexA proteins. Their study identified 57 genes whose expression is altered upon LexA depletion. Of these, 11 genes are known to be involved in carbon assimilation, while 30 other genes, many of which are uncharacterized, are nevertheless regulated by carbon availability.
LexA is commonly regarded as a transcriptional repressor. Another peculiarity in studies with Synechocytis PCC6803 suggests that the LexA paralogue can function as a transcriptional activator (Tapias et al., 2002; Gutenkunst et al., 2005). Most recently, Gutenkunst and colleagues (Gutenkunst et al., 2005) reported studies of the promoter region upstream of the Synechocystis spp. PCC6803 hoxEFUYH operon encoding the bidirectional NiFe hydrogenase. A series of promoter fragments fused to a luciferase reporter revealed that in vivo activity could not be detected unless two essential regions were present: an untranslated leader adjacent to the hoxE ATG and an upstream region that turned out to bind LexA. The exact mechanism of promoter regulation is unknown but it is reminiscent of action at a distance via long-range protein–protein interaction or transmission of topological changes that affect promoter function.
Endogenous triggers and population studies of the SOS response
Understanding what turns on SOS is a fundamentally important question that includes the need to understand both exogenous and endogenous triggers. Exogenous triggers are plentiful (quinolones, organic mutagens, UV and mitomycin C) and have been extensively described. Hydrostatic-pressure as a physical stress is also now known to induce an SOS response in E. coli by acting through a novel type IV restriction endonuclease (Aertson and Michiels, 2005). Endogenous triggers are much more poorly documented and, curiously a comprehensive screen for SOS constitutive mutants in a single organism was not attempted until O'Reilly and Kreuzer (O'Reilly and Kreuzer, 2004) designed a systematic screen for transposon insertions that lead to SOS constitutive mutants in E. coli. Their study succeeded in identifying insertions in 42 non-essential genes that caused significant upregulation of a dinD1::lacZ reporter. Their study is particularly significant because it described the first quantitative comparison of SOS constitutive mutants in the same genetic background. Classification of the gene disruptions led to the assignment of disrupted genes to six categories: replication/recombination/repair, dimer resolution, transcription and regulation, nucleoside metabolism, membrane structure/function and miscellaneous. The ensemble of mutants represents the collection of probably all, or nearly all, non-essential genes that are required to maintain genome stability. Aside from the evident wealth of information to be mined from the data, one especially promising offshoot of this study is the development of new genetic tools to screen for genotoxic metabolites.
Studies at the population and single cell level are also revealing new features of the SOS response (Ronen et al., 2002; McCool et al., 2004; Friedman et al., 2005). Monitoring the transcriptional activation of eight different SOS-regulated promoters fused to a rapidly maturing green fluorescent protein (GFPmut3) reporter gene in E. coli strains exposed to a short UV pulse revealed a rapid and co-ordinate activation, but also revealed clear differential inactivation. Apparently, genes required in early steps of repair are the first to be inactivated, followed by the inactivation of genes required in later steps in recovery and adaptation (Ronen et al., 2002). When this study was repeated at the single cell level using only recA, lexA and umuDC promoter-GFP reporter strains and time-lapse fluorescence microscopy a remarkably precise temporal regulation of promoter activity was observed (Friedman et al., 2005). The SOS response apparently synchronizes the repair process and is highly structured despite the initial degree of triggering damage. The umuDC operon was also identified in this study as a key contributor maintaining the temporal precision of the SOS response.
In another provocative study, McCool and coworkers used a sulA promoter-GFP reporter to monitor SOS induction in E. coli and discovered that a fraction of wild-type cells (0.3%) are highly induced in a population (McCool et al., 2004). Additional studies need to be conducted, but it is possible that a robust SOS response is induced by endogenous triggers at low frequencies within a population and that ultimately this may reflect a source of mutation and enhanced heterogeneity.
Enhanced dispersion of mobile elements and virulence factors by SOS induction
From the standpoint of epidemiology and infection control, important lessons are emerging from the study of SOS induction by therapeutic drug use and its consequences for the dissemination of virulence factors in human pathogens. While it is well appreciated that drugs are developed and administered to prevent or control infection, little is known of antibiotics' biological effects, especially when cells are exposed to low, or subinhibitory, levels. Provocative arguments have been raised suggesting that antibiotics in the natural environment rarely achieve therapeutic concentrations and thus their true purpose is more likely to modulate metabolic functions in microbial populations or tailor responses to changing environments (Davies, 2006; Yim et al., 2006). In this respect, although it is commonplace to relate the emergence of drug resistance and in many cases detailed descriptions of underlying mechanism, it is less common to encounter examples that highlight how drugs actually provoke a myriad of genetic responses, often with unsettling consequences. As illustrative examples, we focus on more recent studies concerning S. aureus and extend the discussion to Vibrio cholerae and E. coli.
Antibiotic-induced SOS responses and their consequences in S. aureus
Goerke and coworkers (Goerke et al., 2006) noted that subinhibitory levels of antibiotics of two distinct classes, ciprofloxacin and trimethoprim, could derepress bacteriophage Φ13 lysogens, or Φ13-like lysogens, in a variety of S. aureus strains including clinical specimens isolated from cystic fibrosis patients. Φ13 inserts into the chromosomal gene encoding haemolysin β (hlb) and is part of a broad class of hlb-converting phage. Simple measurement of haemolytic colonies on sheep blood agar or evidence of phage excision by a polymerase chain reaction-based assay of attP sites per chromosome equivalent revealed prophage induction that was not observed in untreated controls. Exposure to ciprofloxacin was also demonstrated to lead to increased transcription of sak, the phage-encoded staphylokinase, another virulence factor and was strongly correlated with upregulation of recA. Exposure to trimethoprim, in contrast, resulted in only minor induction of sak and no detectable induction of recA indicating that drugs may have considerably different effects or thresholds. To what extent this mechanism acts in nature is largely unknown, but the data strongly hint at the necessity for further understanding and controlling potential adverse consequences of antibiotic therapy.
In second study, Ubeda and coworkers (Ubeda et al., 2005) described elegant experiments where either mitomycin C or ciprofloxacin exposure induced not only derepression of resident S. aureus prophages, but also induced mobilization and high frequency horizontal transmission of coresident pathogenicity islands (SaPIs) that are governed by the resident prophage. A key experiment showed that inactivation of either host recA, or use of a bacteriophage Φ11 lysogen harbouring an engineered cI repressor variant that cannot undergo autocleavage, dramatically reduced the horizontal transfer frequency of SaPIbov1. SOS-mediated derepression of resident prophages is likely commonplace, and so the enhanced induction of resident SaPIs by a drug stimulus that triggers SOS will no doubt depend significantly upon the particular bacterial strain genetic composition. The relatively easy demonstration of horizontal transfer in the laboratory points to the potential of what might occur in nature. Fortunately, barriers to transfer might include, for example, the chance combination of appropriate compatible mobilizing prophage and pathogenicity islands, or the existence of strong restriction modification barriers among currently circulating S. aureus lineages (Waldron and Lindsay, 2006).
In a follow-up to this fascinating study, additional subinhibitory concentrations of other antibiotic classes commonly in clinical use were tested for their ability to induce an SOS response in S. aureus (Maiques et al., 2006). Of those classes examined, which included macrolide, aminoglycoside, chloramphenicol and tetracycline, only the tested β-lactams (ampicillin, penicillin G, cloxacillin and ceftriaxole) were shown to promote enhanced SaPI induction. A strong dependence upon a functional recA gene was noted in this study, but the use of a non-cleavable lexA allele (lexAG94E) reduced, but did not entirely abolish SaPI induction, suggesting that a LexA-independent pathway may exist. Future experiments should examine this pathway.
The collective conclusion from these studies is that a classical DNA damaging agent such as mitomycin C, or even SOS-inducing ciprofloxacin, or β-lactam treatment, can promote enhanced horizontal gene transfer of virulence factors encoded on mobile elements that are themselves induced by SOS triggers. As was succinctly mentioned in the Ubeda study, most clinically relevant toxinoses are almost universally linked to mobile elements (Novick, 2003). The intriguing and open question is to what extent mobile element transfer is modulated by antibiotic therapy and whether this alters the course of clinical case resolution. In nature, mobile elements serve as a vast conduit and motor for genetic diversity. These observations lay groundwork for future investigation of SOS-induced mobilization of genetic elements modulating virulence factors in a variety of human and animal pathogens.
Distinct from the drug-inducted induction of mobile elements or prophage, our laboratory described the discovery in S. aureus that fnbB, one of two chromosomal genes encoding a fibronectin binding protein (FnbB, a known virulence factor in this organism) was part of the LexA regulon, and could be induced by subinhibitory concentrations of ciprofloxacin in highly quinolone resistant strains (Bisognano et al., 2004). The demonstration that a drug-induced SOS response could provoke enhanced production of an MSCRAMM (microbial surface components recognizing adhesive matrix molecules) virulence factor protein that mediates tissue attachment and invasion into non-myeloid cells is of particular interest. S. aureus is not traditionally considered to be an intracellular pathogen, although there is mounting evidence suggesting that S. aureus can invade and persist transiently in a variety of cell types (Clément et al., 2005; Haslinger-Löffler et al., 2005).
Fibronectin attachment is a crucial step for invasion that does not otherwise require active bacterial processes. Fibronectin binding proteins (FnBPs) promote the attachment of S. aureus to host cell integrins using surface or soluble fibronectin as a bridging molecule (Sinha et al., 1999). Remarkably, heterologous expression of S. aureus fnbB in non-invasive staphylococci (Staphylococcus carnosus), Lactococcus cremoris, or clinical isolates of non-invasive S. aureus rendered them invasive (Sinha et al., 2000). In addition, a particularly insidious variety of S. aureus, known as small colony variants, which show altered drug resistance profiles and give rise to persistent infections in osteomyeltis or implanted devices, exhibit enhanced production of FnBPs that favour cell invasion kinetics (Vaudaux et al., 2002). Collectively, these findings suggest that factors that modulate FnBP expression ultimately affect bacterial tissue invasion the establishment of persistent infection. The use of fluoroquinolones per se might also favour the carrier state of methicillin-resistant S. aureus (MRSA) that would increase the risk of contracting severe infection (Weber et al., 2003). If low level ciprofloxacin exposure can enhance the production of a key MSCRAMM colonizing factor via SOS induction, then certainly the SOS response needs to be examined in greater detail in this organism, together with a thorough evaluation of the consequences of this particular antibiotic therapy.
Antiobiotic-induced SOS responses in V. cholera and E. coli
The link of SOS induction and the horizontal dissemination of virulence determinants is not limited to S. aureus. Drug stress induces SOS responses in V. cholerae and E. coli, notably, with important epidemiological consequences. Beaber and coworkers (Beaber et al., 2004) reported conditions where DNA damaging agents, such as mitomycin C, or ciprofloxacin, could promote the spread of antibiotic resistance through induction of mobile elements in V. cholerae O139. SXT is a self-transmissible, integrative conjugative element (ICE) that harbours a cluster of genes conferring resistance to chloramphenicol, sulphamethaoxazole, trimethoprim and streptomycin. In this ICE, genes involved in transfer are repressed by SetR, itself encoded by SXT. The similarity of SetR to bacteriophage λ CI repressor naturally leads to the hypothesis that agents that could induce an SOS response might concomitantly promote SetR derepression and enhance SXT transfer. In an elegant series of genetic experiments, Beaber et al. indeed demonstrated that SXT transfer depends upon SOS induction. Induction could be abolished by deleting donor cell recA, engineering a RecA coprotease deficient mutant (recA430), or introducing a non-cleavable SetR repressor (SetR G94E). The mechanism appears to involve an SOS-trigger leading to RecA activation and RecA-promoted cleavage of SetR, an event that culminates in derepression of genes coding for transcription factors SetC and SetD that control conjugal transfer and integrase genes.
The SXT element can also be easily established in and transferred from E. coli, which suggests the potential for a broad range of transmissibility. When one considers that the Set element architecture is conserved in other SXT-like elements, it is not hard to imagine scenarios where encounter with a variety of environmental stresses likely promotes additional antibiotic resistance spread.
The temperate filamentous bacteriophage of V. cholerae, CTXΦ, which encodes and transmits cholera toxin, is also induced during an SOS response (Quinones et al., 2005). Quinones et al. discovered that RecA-dependent cleavage of host-LexA is required for CTX induction by either UV or mitomycin C treatment. A non-cleavable LexA variant lacking the conserved Ala-Gly autocleavage site, blocks CTX induction. The phage repressor, RtsR, acts together with host LexA to repress the phage rtsA promoter that controls genes necessary for virion production. A detailed understanding of promoter regulation is still necessary, but it is clear that both LexA and RtsR bind to the rtsA promoter region and are required for repression. RtsR itself is apparently not a target for activated RecA, nor does it contain a protease domain or canonical autocleavage motif. Thus, in an unusual regulatory twist, the CTXΦ phage apparently has come under the control of host LexA repression, a condition unlike more familiar cases where the phage repressor itself is cleaved following SOS-mediated activation of RecA.
The enhanced production of CTX virions following SOS induction raises the important question whether cholera toxin production is concomitantly enhanced. Recent detailed investigation of the cholera toxin ctxA promoter itself suggests that SOS induction does not play a significant regulatory role in augmenting ctx expression in vivo in the suckling mouse intestine model (Quinones et al., 2006). Thus, although experimentally evoked SOS clearly can augment ctx transcription, most likely via read-through from activation of the upstream rtsA promoter, no SOS induction occurs in the course of model infection. Under conditions where an SOS response occurs, the downstream effects might be limited to promoting dissemination of infectious virions, although not dramatically altering toxin formation. To what extent this model is applicable in the course of human infection is unresolved. In sharp contrast, studies with shiga-toxin producing E. coli indicate that therapeutic regimens (i.e. fluoroquinolones) that provoke an SOS response may significantly disrupt clinical evolution of disease by enhancing toxin production and dissemination (Kimmitt et al., 2000; Zhang et al., 2000).
Additional studies in E. coli now reveal that subinhibitory concentrations of ciprofloxacin also induce SOS-dependent synthesis of colicins (Jerman et al., 2005). Indeed, the exposure of E. coli to purified colicin E9 alone can induce an SOS response perhaps not surprisingly through its activity as an endonuclease and ensuing DNA damage detection (Walker et al., 2004). While colicins can kill closely related bacteria, recent reports detail the ability of colicins to regulate microbial diversity, perhaps via induction of SOS, or evoked transcriptional upregulation of regulatory genes in mobile elements (Walker et al., 2004).