11050–11066 Nucleic Acids Research, 2021, Vol. 49, No. 19 Published online 6 October 2021 https://doi.org/10.1093/nar/gkab773 Non-canonical LexA proteins regulate the SOS response in the Bacteroidetes 1 1 2 2 1,* Miquel Sánchez-Osuna , Pilar Cortés , Mark Lee , Aaron T. Smith , Jordi Barbé and Ivan Erill 1,3,* 1 Departament de Genètica i de Microbiologia, Universitat Autònoma de Barcelona, 08192 Bellaterra, Spain, 2 Department of Chemistry and Biochemistry, University of Maryland Baltimore County, Baltimore, MD 21250, USA and 3 Department of Biological Sciences, University of Maryland Baltimore County, Baltimore, MD 21250, USA Received May 28, 2021; Revised August 18, 2021; Editorial Decision August 22, 2021; Accepted October 04, 2021 ABSTRACT sponse (1,2). First described in Escherichia coli, the SOS response involves the coordinated expression of over 40 Lesions to DNA compromise chromosome integrity, genes encoding primarily DNA repair and recombination posing a direct threat to cell survival. The bacterial enzymes, translesion synthesis polymerases and cell divi- SOS response is a widespread transcriptional regu- sion inhibitors (3,4). This process is governed by the LexA latory mechanism to address DNA damage. This re- repressor. This protein forms dimers that bind to operator sponse is coordinated by the LexA transcriptional sites in the promoter region of regulated operons by target- repressor, which controls genes involved in DNA re- ing a highly specific palindromic motif (CTGT-N8-ACAG pair, mutagenesis and cell-cycle control. To date, the in E. coli) (3). Following DNA damage, single-stranded SOS response has been characterized in most major DNA (ssDNA) fragments stemming from stalled replica- bacterial groups, with the notable exception of the tion forks are bound by the recombination protein RecA, Bacteroidetes. No LexA homologs had been iden- resulting in active nucleoprotein filaments that are capable of inducing the autocatalytic cleavage of the LexA protein tified in this large, diverse and ecologically impor- (5,6). Upon self-cleavage, the LexA dimer detaches from its tant phylum, suggesting that it lacked an inducible operator sites, derepressing regulated operons and inducing mechanism to address DNA damage. Here, we re- the SOS response (4). port the identification of a novel family of transcrip- Over the last three decades, the SOS response has been tional repressors in the Bacteroidetes that orches- documented in multiple bacterial groups (1,2,7–10). This trate a canonical response to DNA damage in this broad taxonomic perspective has revealed a minimal set phylum. These proteins belong to the S24 peptidase of core genes that are predominantly regulated by LexA family, but are structurally different from LexA. Their across phyla. These encompass the lexA and recA genes, N-terminal domain is most closely related to CI-type as well as those encoding the type IV (dinB) and type bacteriophage repressors, suggesting that they may V (umuDC) error-prone DNA polymerases (1). The sys- have originated from phage lytic phase repressors. tematic analysis of this system has also uncovered that, in contrast with many other transcriptional regulators, Given their role as SOS regulators, however, we pro- the LexA repressor has significantly changed its binding pose to designate them as non-canonical LexA pro- motif through evolution. Reported LexA-binding motifs teins. The identification of a new class of repressors range from short (TTAC-N3-GTAA; Bdellovibrio bacteri- orchestrating the SOS response illuminates long- ovorus) to large palindromes (GGTT-N10-AACC; Geobac- standing questions regarding the origin and plastic- ter sulfurreducens) and include several direct-repeat mo- ity of this transcriptional network. tifs (GTTC-N7-GTTC; Alphaproteobacteria) (1). Dupli- cations of the lexA gene are often associated with LexA- binding motif divergence, enabling the redundant lexA gene INTRODUCTION to significantly alter its binding motif (11). It has been pos- Many environmental insults and endogenous processes can tulated that, upon loss of the primary lexA gene, the di- cause DNA lesions that pose a direct threat to cell survival. verged LexA protein will gradually retake control of the In most bacterial species, DNA damage is addressed by a SOS regulatory network via a process of convergent evo- transcriptional regulatory process known as the SOS re- lution (7). * To whom correspondence should be addressed. Tel: +1 410 455 2470; Fax: +1 410 455 3875; Email: erill@umbc.edu Correspondence may also be addressed to Jordi Barbé. Email: jordi.barbe@uab.cat.  C The Author(s) 2021. Published by Oxford University Press on behalf of Nucleic Acids Research. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Nucleic Acids Research, 2021, Vol. 49, No. 19 11051 LexA monomers contain an N-terminal winged helix– the Bacteroidetes play important roles, and are essential for turn–helix (wHTH) DNA-binding domain with three ␣ the degradation of complex carbohydrate-based biomass helices (PF01726) and a C-terminal autolysis and dimer- (24). Due to the wide variety of habitats they occupy, mem- ization domain (PF00717). These two domains are con- bers of the Bacteroidetes phylum are frequently exposed to nected by a flexible linker region (12). Several other proteins many DNA damaging agents including antibiotics and re- share the S24 serine peptidase family (PF00717) catalytic active oxygen species (ROS). However, the presence of a C-terminal domain of LexA. The Enterobacteria phage DNA repair system similar to the SOS network has not been Lambda CI and Salmonella phage P22 C2 repressors bind reported to date in this bacterial phylum. The sequencing DNA via an N-terminal HTH domain encompassing five of the first representative Bacteroidetes complete genomes alpha helices and repress genes that participate in the lytic (Bacteroides thetaiotaomicron and Bacteroides fragilis) ex- cycle of temperate bacteriophages. Like LexA, CI and C2 posed the lack of LexA homologs in these bacterial species, repressors are capable of undergoing self-catalytic cleavage suggesting that this phylum lacked a conventional SOS re- via interaction with RecA nucleoprotein filaments, trigger- sponse (1). The recent description of a conventional LexA ing the phage lytic cycle (13). The E. coli UmuD protein regulon in the Bacteroidetes sister phylum Balneolaeota forms part of the DNA polymerase V complex with UmuC. prompted us to investigate the SOS response in the Bac- RecA-mediated self-catalytic cleavage of UmuD activates teroidetes (10). the UmuDC complex, providing a post-translational level In this work we combine comparative genomics ap- of control on this highly mutagenic polymerase in addition proaches with in vitro and in vivo analyses to elucidate the to its transcriptional regulation by LexA (14,15). SOS response in the Bacteroidetes. We report that a novel Given its importance in maintaining cell viability, the family of LexA-like repressors, targeting two distinct palin- SOS response is often assumed to be universal. The absence dromic motifs, controls a conventional SOS response in this of LexA homologs, however, has been reported in several phylum. Our results highlight the evolutionary plasticity of bacterial groups. Beyond endosymbiotic bacteria that have this transcriptional network and support the notion that undergone substantial genomic reduction, such as the Rick- the SOS response has reevolved several times through con- ettsiae (16), the absence of LexA homologs has been par- vergent evolution using S24-family peptidases with distinct ticularly well documented in the Streptococcaceae. In the DNA-binding domains. absence of LexA, different Streptococcaceae species have been shown to orchestrate a basic response to DNA dam- MATERIALS AND METHODS age using a LexA-like repressor, HdiR. In Lactococcus lac- tis, where it was first reported, HdiR was shown to regulate Genome data and ortholog detection itself and a UmuC homolog in response to DNA damage Bacteroidetes complete genome assemblies were down- (17). As is the case for LexA and phage repressors, RecA nu- loaded from the RefSeq database (25) both in protein cleoprotein filaments trigger self-catalytic cleavage of HdiR, multi-FASTA and nucleotide GenBank formats. Orthologs but further processing of HdiR by the Clp proteolytic com- for UmuD and DinB/UmuC were identified in all the plex is required for full derepression (17). Subsequent work Bacteroidetes RefSeq complete genome assemblies using revealed basic SOS-like regulatory networks under control HMMER (hmmsearch, E-value 1e-10 ) (26), using Hidden of HdiR in Streptococcus species. In Streptococcus uberis, Markov Models (HMM) derived for the Clusters of Or- HdiR regulates only a mutagenic gene cassette, while in thologous Groups (COGs) mapping to E. coli UmuD Streptococcus thermophilus, it regulates both the cinA-recA (COG1974) and DinB/UmuC (COG0389) as queries operon and two umuC loci (18,19). Interestingly, non-SOS (27,28). Orthologs for the Bacteroidetes RecA protein and regulatory functions have also been reported for HdiR in for the two putative Bacteroidetes SOS regulators were ob- the Streptococcaceae. In L. lactis, HdiR was shown to re- tained by searching all available Bacteroidetes complete spond to heat shock, whereas in S. thermophilus HdiR was proteomes with BLASTP (limiting e-value 1e–20 and query found to interfere with natural transformation. coverage > 75%), using as queries all the instances of these The uptake of a minimal SOS-like network by proteins proteins identified in the comparative genomics analyses, other than LexA has also been reported in the Moraxel- and subsequently removing duplicates. laceae, a bacterial family with no known LexA homologs. In Acinetobacter baumannii and Acinetobacter baylyi, a vari- Motif discovery and comparative genomics analysis ant of the UmuD protein containing an N-terminal DNA- binding domain (UmuDAb) has been shown to regulate the The upstream regions (from −250 to +2 bp of the pre- expression of its own and several other umuDC operons dicted translational start site (TLS)) of identified UmuD (20,21). Like LexA and HdiR, UmuDAb undergoes auto- and DinB/UmuC orthologs, and subsequently for putative catalytic cleavage upon interaction with RecA nucleopro- Bacteroidetes SOS regulators, were obtained from the re- tein filaments, but its regulatory function appears to be re- spective complete genome sequences. Redundant upstream stricted to the control of type V error-prone polymerases sequences (those with nucleotide sequence identity > 75 %) (21,22). were removed, and the resulting non-redundant panel was The Bacteroidetes phylum comprises a large and diverse used to perform motif discovery. Motifs were inferred with group of bacteria present in many different ecosystems, such MEME (29) using a 12–26 bp motif size, the Any Num- as freshwater and marine habitats, soil, plants and the mam- ber of Repetitions (ANR) site distribution model and oth- malian gastrointestinal tract, and ranging from temperate erwise default parameters. Refined motif discovery was per- to tropical and polar climates (23). In all these ecosystems, formed by activating the -pal option in MEME to restrict 11052 Nucleic Acids Research, 2021, Vol. 49, No. 19 the search to palindromic motifs and otherwise identical (37). For RecA, the resulting amino acid sequence align- parameters. Comparative genomics analyses of the regu- ment was processed with Gblocks using the half-gap set- latory networks defined by the identified motifs were per- ting (38). Phylogenetic inference was carried out with Mr- formed using the CGB comparative genomics platform, Bayes (39). Four Metropolis-coupled Markov chain Monte using COG and PFAM for functional annotation of or- Carlo simulations with four independent chains were run thologous groups (10). CGB configuration files for regu- for 5 000 000 (SOS regulators) and 10 000 000 (RecA) gen- lon reconstruction are provided as supplementary material erations, using a mixed four-category gamma distributed in JSON format (Supplementary Data S1, Supplementary rate plus proportion of invariable sites model [invgamma] Data S2). and a JTT (Jones–Taylor–Thornton) amino acid substitu- tion model (40). Convergence was monitored with Tracer (41), imposing the restriction that the estimated sample size Protein sequence and in silico structural analyses (ESS) be above 200 and that the potential scale reduction A multiple sequence alignment of the C-terminal region of factor (PSRF) be within 0.005 of 1, and burn-in was set at the putative Bacteroidetes SOS regulators and known S24 25% of iterations. A consensus tree was generated with the catalytic domains was performed with CLUSTALW in pro- all-compat option. Tree visualization and annotation were file alignment mode, using the E. coli LexA protein struc- performed with iTOL (42). ture (P0A7C2) to define a gap penalty mask (30). Con- sensus sequences for each group of putative Bacteroidetes Protein purification and electro-mobility shift assays SOS regulators were inferred with the EMBOSS Cons ser- vice (31), and the multiple sequence alignment was plotted The Salegentibacter agarivorans [BM084 RS03395], with BioEdit (32). HMM for the two clusters of putative Pontibacter actiniarum [CA264 RS07585] and Eliza- Bacteroidetes SOS regulators were obtained by performing bethkingia anophelis [BBD30 RS12205] genes encoding independent multiple sequence alignments of each cluster putative Bacteroidetes SOS regulators (WP 075324850.1, of sequences with CLUSTALW, and generating the HMM WP 025606010.1 and WP 078407279.1, respectively) with the hmmbuild command of the HMMER suite. were synthesized by ATG:biosyntheticsGmbH, Germany, Experimentally-determined structures of E. coli LexA and cloned into a dephosphorylated pUA1108 vector (PDB ID: 3JSO), full-length Enterobacteria phage Lambda (43) using an NdeI and BamHI (New England Biolabs) (PDB ID: 3BDN) and the C-terminal domain of Lambda double digest procedure. Genes with internal restriction Cl (PDB ID: 1F39) were obtained from the Protein Data- sites for any of these two enzymes were subcloned into bank (PDB). Calculated structures were determined via the pUA1108 vector using the HiFi DNA assembly kit the Robetta suite (33,34) using the TrRosetta, comparative (New England Biolabs), following the manufacturer’s modelling, and ab initio strategies independently. All three protocol. Cloned genes were overexpressed in E. coli strategies demonstrated a qualitative similarity regardless BL21-CodonPlus(DE3)-RIL (Stratagene) cells and the of modelling pipeline. After observation of all outputs, Tr- resulting His-tagged proteins were purified following the Rosetta produced models with the highest confidence inter- previously described protocol for LexA proteins (44). val across the entirety of the polypeptide and was therefore EMSAs were carried out using 100 bp-long DNA probes, used for comparative analysis against the experimentally- which were generated using two complementary synthetic determined structures. Structural superpositions and C␣ oligos centered on predicted binding-motifs and perform- root-mean-square deviations (RMSDs) were calculated us- ing PCR with M13 universal digoxigenin-labeled primers ing PyMol (35) after manual trimming to remove unstruc- (Supplementary Table S1) (9). EMSAs were performed on tured regions belonging to the flexible linker connecting the a mixture containing 25 ng of each digoxigenin-marked N- and C-terminal domains. The numbers of C␣s used for DNA probes and 30 nM of the purified Bacteroidetes SOS each pairwise RMSD calculation are indicated in the figure regulators as described previously (9). For competition legend. assays, 400-fold molar excess of the same unlabeled probe was used as a specific competitor fragment (9). Samples were loaded onto 6% non-denaturing Tris-glycine poly- Phylogenetic inference acrylamide gels and DNA-protein complexes were detected For phylogenetic inference of putative SOS regulators, the using the manufacturer’s protocol (Roche NimbleGen) (9). amino acid sequences of all putative SOS regulators iden- The DNA sequencing of all genes and probes was carried tified in the comparative genomics analyses were com- out by Macrogen. bined with all the representative sequences for COG1974 and COG2932 available in the COG database (27). For RT-qPCR gene expression analyses RecA, records for one representative member of each Bac- teroidetes family, as well as for all the species used in mo- Gene expression was determined by RT-qPCR as previ- tif discovery, were selected for phylogenetic inference. In ously described (45), using Lightcycler RNA Master SYBR both cases, a protein sequence multiple sequence align- green I (Roche) on a Lightcycler 480 instrument (LC480; ment was generated using T-COFFEE (36), combining Roche), following the manufacturer’s instructions. Expo- three CLUSTALW amino acid sequence alignments with nential cultures of S. agarivorans DSM 23515 and P. ac- different (5,10,25) gap opening penalties and a single T- tiniarum DSM 19842 were grown in Marine Broth at 28◦ C, COFFEE Lalign method amino acid sequence alignment and cultures of E. anophelis DSM 29660 were grown in Nucleic Acids Research, 2021, Vol. 49, No. 19 11053 Tryptone Soya Broth at 28◦ C. Cultures were inoculated with of different core SOS proteins strongly suggested that these 7.5 ␮g/ml mitomycin C (Sigma-Aldrich), a concentration motifs could define a SOS regulatory network. To investi- that has been frequently used in the past to study the SOS gate this possibility we used both motifs to perform com- response in multiple organisms (3,46–51). After 3 h, RNA parative genomic analyses of their putative regulatory net- samples were extracted using the RNeasy mini-kit (Qiagen) work in the Bacteroidetes species in which the motifs had following the manufacturer’s instructions. Specific oligonu- been identified. The results of the comparative analysis with cleotides were used to validate the expression of genes of in- the GGA-N5-TCC motif (Figure 1A) revealed a putative terest (Supplementary Table S1). The relative mRNA con- regulatory network dominated by genes encoding DinB ho- centrations of selected genes were determined according to mologs (COG0389), often associated with the gene encod- a standard curve generated by amplifying a fragment of the ing the DNA polymerase III alpha subunit (COG0587) in gyrB gene. This gene is one of the most stable internal con- an operon arrangement that has been reported before in trols for RT-qPCR in multiple organisms, and it is known multiple SOS systems (40). not to be inducible by DNA damage in E. coli (52–56). The The conserved elements of this putative network also en- gene expression factor was calculated as the ratio of the nor- compass the SplB radical SAM protein involved in DNA malized mRNA concentration of each target gene in the repair (COG1533), which has been shown to be SOS reg- treated strain versus the non-treated strain. ulated in several species (57,7,10), as well as a two-gene operon encoding a putative DNA base excision repair sys- tem (COG1573-COG4277) involving a uracil-DNA glyco- sylase (UDG) (7). Other putative members of the inferred RESULTS regulatory network include a umuDC operon (COG0389- COG1974) (40) and a SOS-response associated peptidase Two divergent DNA motifs define SOS-like regulatory net- (COG2135) regulated by LexA in the Balneolaeota (10,58). works in the Bacteroidetes Evidence of regulation for other genes encoding canonical The presence of LexA homologs in the Bacteroidetes had SOS proteins (3), such as RecA (COG0468) and the mis- been cursorily assessed when few complete genomes for this match repair enzymes UvrC (excinuclease; COG0322) and phylum were available (1). To ascertain whether members UvrD (helicase; COG0210) is also found in several species of this phylum encode LexA homologs, we used a reference (Supplementary Figure S3). dataset of experimentally-validated LexA proteins (Supple- The comparative analysis with the CTAA-N5-TTAG mentary Table S2) to query the proteins encoded by all the motif outlined a canonical SOS regulatory network, with Bacteroidetes complete genomes available in NCBI RefSeq core SOS genes (umuDC, recA, recN, dinB, ruvB and splB) with BLASTP. No significant hits were obtained, confirm- presenting conserved instances of this motif in their up- ing the previous observation that Bacteroidetes genomes do stream region. Genes encoding additional SOS proteins not encode LexA homologs. However, BLAST searches us- also showed substantial evidence of regulation by this mo- ing the E. coli UmuDC and DinB proteins revealed the pres- tif in several Bacteroidetes species. These include the mu- ence of putative translesion synthesis polymerases in several tagenic cassette imuA-imuB-dnaE2 (COG4544-COG0389- members of this phylum (Supplementary Table S3). Given COG0587) (40) and the UDG DNA base excision repair their high mutagenic activity, these polymerases tend to be system enzymes (COG1573-COG4277). The analysis also tightly regulated (15). reveals evidence of regulation in several species of this To investigate their possible regulation in the Bac- phylum for the genes encoding the aforementioned SOS- teroidetes, we implemented a computational pipeline to de- associated peptidase (COG2135), the single-stranded bind- tect orthologs of these proteins, automatically infer putative ing protein Ssb (COG0629), the RecG DNA translocase binding motifs and assess the regulatory network such mo- (COG1200), the RecQ helicase (COG0514) and the DNA tifs control (Supplementary Figure S1). We used HMMER repair protein RadC (COG2003) (59) (Supplementary Fig- to search Bacteroidetes proteomes using the HMM pro- ure S4). files for the COGs mapping to E. coli UmuD (COG1974) In addition, the analyses with both motifs identified other and DinB/UmuC (COG0389) (Supplementary Table S4). proteins that could be potentially associated with a response We obtained the nucleotide sequence upstream of the genes to DNA damage (Figure 1AB). In particular, evidence of encoding putative UmuD and DinB/UmuC homologs and regulation under both motifs was observed for an GIY-YIG used MEME to detect overrepresented motifs in these se- domain-containing exonuclease (COG0847) with homol- quences (Supplementary Data S3, Supplementary Data ogy to the proofreading subunit of DNA polymerase III, S4). MEME returned two significant motifs with appar- and for a phage-like repressor (COG2932). Furthermore, ent palindromic structure (Supplementary Figure S2), and putative regulation by the CTAA-N5-TTAG motif was also these were subsequently refined by activating the palin- consistently detected for an operon encoding DNA topoi- dromic constraint in MEME (Figure 1A, B). One of the somerase IV subunits A (COG0188) and B (COG0187). motifs (GGA-N5-TCC) was identified only in the upstream For both motifs, two or more motif instances were identi- regions of DinB/UmuC-homolog encoding genes, whereas fied in the promoter region of several putatively regulated the other motif (CTAA-N5-TTAG) was detected in the up- genes, a common feature in the SOS regulatory network stream regions of both DinB/UmuC- and UmuD-homolog (Supplementary Table S5) (3,7,60). Overall, these results encoding genes. are consistent with, and strongly suggestive of, SOS reg- The detection, across multiple species, of well-defined ulatory networks mediated via the identified motifs in the motifs in the upstream region of genes encoding homologs Bacteroidetes. 11054 Nucleic Acids Research, 2021, Vol. 49, No. 19 Figure 1. Comparative analysis of SOS Bacteroidetes regulons. (A) Sequence logo of the GGA-N5-TTC motif identified by MEME upstream of genes encoding UmuC/DinB proteins. CGB-generated heatmap of the posterior probability regulation for orthologous groups using the identified motif. Cells are colored from green (regulation) to red (no regulation), with blue indicating absence of ortholog. Only genes predicted to be regulated in at least 15% of the species analyzed are shown. (B) Sequence logo of the CTAA-N5-TTAG motif identified by MEME upstream of genes encoding UmuC/DinB and UmuD proteins. CGB-generated heatmap of regulation for orthologous groups using the CTAA-N5-TTAG motif. Color scheme is the same as in panel A. Only genes predicted to be regulated in at least 15% of the species analyzed are shown. (C) Comparative analysis of S24 peptidase domains. Multiple sequence alignment including the C-terminal segment of the E. coli (Eco) LexA and UmuD sequences, the Mycobacterium tuberculosis (Mtu) LexA sequence, the Lambda phage CI repressor sequence and the consensus sequences of the two Bacteroidetes SOS regulator clusters. A novel family of LexA-like repressors in the Bacteroidetes analysis. Furthermore, several species with motifs identified upstream of dinB lack UmuD homologs (Supplementary The detection of putative SOS regulatory networks in the Table S6), suggesting that a UmuD-like protein is not the Bacteroidetes implicitly entailed the presence of transcrip- regulator targeting these dinB promoters. tional regulators targeting the identified motifs. Given the As mentioned above, a salient feature of the compara- absence of canonical LexA homologs, an obvious candi- tive analyses in Figure 1AB was the presence of an ortholog date for this role were UmuDAb-type proteins, which have group mapping to COG2932 (Phage repressor protein C). been described to target palindromic motifs (21), but no Members of this ortholog group were present in all the DNA-binding domains were detected in any of the Bac- species in which the GGA-N5-TCC and CTAA-N5-TTAG teroidetes UmuD homologs identified in the comparative Nucleic Acids Research, 2021, Vol. 49, No. 19 11055 motifs were identified, and the genes coding for them pre- tative Bacteroidetes SOS regulators define a well-supported sented strong evidence of regulation under both motifs (Fig- clade, clearly differentiated from those defined by COG1974 ure 1A, B). This suggested that proteins from this ortholo- and COG2932 representatives. This result is in contrast with gous group might be regulating the putative SOS regulatory HdiR, which forms a well-defined cluster with Streptococ- networks defined by these two motifs. BLASTP searches cus phage repressors. The tree also reveals that these puta- against representative genomes from all other orders in the tive Bacteroidetes SOS regulators form two distinct clus- Bacteria domain did not return any significant matches, in- ters with significant intercluster distance, indicative of an dicating that these putative transcriptional regulators are ancient evolutionary split. Furthermore, the promoter re- exclusive to the Bacteroidetes phylum. gions of the genes encoding members of these two clusters A feature common to all SOS-like S24-family regulators presented predicted binding sites for the GGA-N5-TCC or (LexA, HdiR and UmuDAb) is their ability to undergo CTAA-N5-TTAG motifs (Supplementary Table S8) in the self-catalytic cleavage mediated by RecA nucleoprotein fila- comparative analyses of Figure 1AB, suggesting that the ments. The sequence determinants for this process are well- evolution of these DNA motifs was associated to the di- conserved in the S24 serine peptidase family and have been versification of the putative Bacteroidetes SOS regulators amply documented (61). They encompass the presence of an in these two clusters. Ala-Gly (Cys-Gly in UmuD) peptide bond that defines the Taken together, these data suggested that the putative cleavage site and a structurally adjacent Ser-Lys catalytic Bacteroidetes SOS regulators define a novel family of LexA- dyad (61). To investigate whether these sequence features like proteins controlling a conventional SOS network in were conserved in the putative SOS regulators mapping to multiple Bacteroidetes genera. Furthermore, even though COG2932 (Figure 1AB), we performed a multiple sequence they share a common ancestor, these LexA-like regulators alignment including the C-terminal regions of these pro- show evidence of ancient diversification, leading to their tar- teins and of S24 family proteins for which the structure geting of distinct palindromic motifs in the Bacteroidetes. of the C-terminal domain has been experimentally deter- To further explore the relationship between putative mined. Bacteroidetes SOS regulators and other S24 family mem- The alignment (Figure 1C; Supplementary Data S5) re- bers, we performed structural predictions of two repre- vealed the presence of two distinct groups of sequences sentative species from each cluster (S. agarivorans DSM among these putative SOS regulators, and highlighted that 23515 [WP 075324850.1] and P. actiniarum DSM 19842 all the key residues involved in the autocatalytic cleavage of [WP 025606010.1]) using the Robetta suite (Figure 3; Sup- S24 peptidases are conserved in both groups. To perform plementary Data S7). The predicted three-dimensional fold their regulatory function, S24-family regulators use differ- of each protein is composed of three domains: an N- ent variants of the HTH DNA-binding domain (62). HH- terminal domain, an unstructured linker region, and a C- pred searches using a multiple sequence alignment of the terminal domain, similar to the observed three-dimensional putative Bacteroidetes SOS regulators N-terminal region folds of both Enterobacteria phage Lambda CI (full length matched members of the penta-helical HTH motif char- PDB ID: 3BDN; C-terminal domain only PDB ID: 1F39) acteristic of phage repressors (PF01381), and subsequent and E. coli LexA (PDB ID: 3JSO) (Figure 3A). We then search with HMMER identified this domain in all these pu- compared the predicted C- and N-terminal domains of the tative SOS regulators (Supplementary Table S7). The data representative species to the crystallized C- and N-terminal hence indicate that these putative Bacteroidetes SOS regu- domains of Enterobacteria phage Lambda CI and E. coli lators encompass an S24 peptidase domain, containing the LexA (Figure 3BCDE). conserved residues involved in RecA-mediated self-catalytic Qualitatively, and consistent with HHpred results, the cleavage of S24 peptidases. This supports the hypothesis folds of the structurally-predicted C-terminal domains that these proteins contain functional S24 peptidase do- (CTDs) bore a strong resemblance to the folds of both En- mains, capable of undergoing self-catalytic cleavage medi- terobacteria phage Lambda CI and E. coli LexA CTDs, ated by RecA nucleoprotein filaments. Furthermore, the with the exception of a terminating ␣-helical segment of presence of an N-terminal HTH DNA-binding domain in- unknown function in the putative Bacteroidetes SOS regu- dicated that, like other S24 family members, these proteins lators (Figure 3CE). In contrast, the structurally-predicted could potentially bind DNA and operate as transcriptional N-terminal domains (NTDs) of the putative Bacteroidetes regulators. SOS regulators were dissimilar from the E. coli LexA NTD (Figure 3C, D). In fact, these calculations predict that the NTDs of the S. agarivorans and P. actiniarum SOS regu- Phylogeny and structure of putative Bacteroidetes SOS reg- lators are multi-helical HTH domains (composed of only ulators ␣-helices) strongly resembling the Enterobacteria phage To elucidate the relationship between these putative Bac- Lambda CI NTD rather than the winged-helix HTH NTD teroidetes SOS regulators and other S24 family members (composed of both ␣-helical and ␤-sheet segments) that is known to regulate SOS-like networks, we inferred their phy- characteristic of E. coli LexA. logeny using a multiple sequence alignment of these pro- These observations are underscored by RMSD calcula- teins and representative members of COG1974 (encompass- tions of the N-terminal superpositions of the known and ing E. coli LexA and UmuD) and COG2932 (encompass- predicted NTD structures (Figure 3D), in which only 3 ing L. lactis HdiR and the repressors of Enterobacteria ␣-helices of calculated NTDs superpose modestly with a phage Lambda and Salmonella phage P22). The resulting strong mismatch of the E. coli LexA winged-helix NTD, tree (Figure 2; Supplementary Data S6) shows that the pu- compared to a satisfactory superposition of all 4 ␣-helices 11056 Nucleic Acids Research, 2021, Vol. 49, No. 19 Figure 2. Unrooted Bayesian consensus tree of S24 peptidase family protein sequences. Sequences of putative SOS regulators (orange) were combined with the representative sequences for COG1974 (green for LexA and blue for UmuD) and COG2932 available in the COG database. Phage repressor coloring (purple) was assigned to COG2932 representatives if close homologs could be detected (>95% sequence similarity) in complete bacteriophage genomes. Branch width denotes support values. The inferred binding motifs for the Bacteroidetes SOS regulators are superimposed on the corresponding clades in the tree. of the calculated NTDs with the Lambda CI NTD. In SOS regulons using two distinct motifs. These analyses orig- contrast, RMSD calculations of the CTD superpositions inated with homology searches of error-prone polymerases (Figure 3E) are excellent irrespective of the organism of DinB and UmuD, leaving open the possibility that ho- origin. These results remained qualitatively the same re- mologs of these regulators in Bacteroidetes species not har- gardless of whether the three-dimensional structural pre- boring these error-prone polymerases might target other dictions were calculated based on homology or through ab motifs. To ascertain whether this was the case, we identi- initio strategies. Thus, the three-dimensional fold of the N- fied additional homologs of these regulators in the Bac- terminal DNA-binding domain of the Bacteroidetes SOS teroidetes using BLASTP (Supplementary Table S9) and regulators is predicted to have striking similarity to the we performed motif discovery with MEME using their pro- N-terminal multi-helical HTH domain of Enterobacteria moter regions. As expected, MEME elicited only the two phage Lambda CI. previously identified DNA motifs, which associate preferen- tially with members of the two putative Bacteroidetes SOS regulators clusters, as illustrated in Figure 2 (Supplemen- Sequence-specific binding and regulation of SOS genes by tary Figure S5). Bacteroidetes SOS regulators To assess whether the predicted regulation of genes by Bioinformatics analyses indicated that a new family of pu- these putative SOS regulators had a SOS-like regulatory ef- tative Bacteroidetes SOS regulators controlled conventional fect, we analyzed the expression of several genes predicted Nucleic Acids Research, 2021, Vol. 49, No. 19 11057 Figure 3. Comparisons of experimentally determined and calculated models of S24-family repressors. (A) X-ray crystal structures of full-length E. coli LexA (blue; PDB ID: 3JSO) and Lambda CI (red; PDB ID: 3BDN) compared to Robetta-calculated models of Bacteroidetes SOS repressors from S. agarivorans (green) and P. actiniarum (orange). (B) N-terminal domains (NTDs) of the proteins in panel A color-coded identically. (C) C-terminal Domains (CTDs) of the proteins in panel A, color-coded identically with the exception of the Lambda Cl CTD (yellow; PDB ID: 1F39), which is poorly resolved in the full-length structure (PDB ID: 3BDN). (D) Superpositions of NTDs and their RMSD values color-coded as in panel A. The numbers of superposed C␣s are as follows: 33 C␣s for E. coli and S. agarivorans; 38 C␣s for E. coli and P. actiniarum; 55 C␣s for Lambda Cl and S. agarivorans; and 33 C␣s for Lambda Cl and P. actiniarum. (E) Superpositions of CTDs and their RMSD values color-coded as in panel B. The number of superposed C␣s are as follows: 63 C␣s for E. coli and S. agarivorans; 67 C␣s for E. coli and P. actiniarum; 61 C␣s for Lambda Cl and S. agarivorans; and 56 C␣s for Lambda Cl and P. actiniarum. to be regulated in each species under SOS-inducing con- promoter DNA as a control. EMSAs with both putative ditions by RT-qPCR. The results of the RT-qPCR assays SOS regulators (Figure 5) reveal a retardation band con- following mitomycin C treatment for both species (Figure sistent with binding of the protein to the target promot- 4) show significant induction for the genes encoding the P. ers containing predicted binding sites. Furthermore, the re- actiniarum and S. agarivorans putative SOS regulators, as tardation band is abolished by the addition of unlabeled well as for canonical and putative SOS genes predicted to promoter DNA, demonstrating that both regulators bind be regulated in the comparative analysis (Figure 1AB). The specifically to these promoter sequences. only exception is the S. agarivorans recA gene (Figure 4A), To further delineate the region specifically bound by these which does not show apparent induction as reported before putative SOS regulators and identify the specific regions for other organisms (48,49,63). In both species, the largest involved in binding, we performed EMSAs coupled with induction values correspond to genes encoding type IV and site-directed mutagenesis on both promoters. The results type V error-prone polymerases. demonstrate that the S. agarivorans putative SOS regula- To determine whether the putative SOS regulators in each tor (WP 075324850.1) binds specifically to the palindromic of the two clusters bound specifically to their predicted mo- GGA-N5-TCC motif (Figure 6A). Substitutions on either tifs, we purified the corresponding proteins for two rep- motif dyad, as well as changes in spacer length, of the mo- resentative species from each cluster (S. agarivorans DSM tif completely abolished binding, whereas substitutions in 23515 [WP 075324850.1] and P. actiniarum DSM 19842 spacer and dyad-adjacent positions had no significant ef- [WP 025606010.1]) and performed electromobility-shift as- fect on binding. Similarly, the P. actiniarum putative SOS says (EMSA) on promoters (umuD and recA, respectively) regulator (WP 025606010.1) was found to bind specifically predicted to be regulated by each protein using unlabeled to the CTAA-N5-TTAG motif (Figure 6B). As in the case 11058 Nucleic Acids Research, 2021, Vol. 49, No. 19 Figure 4. In vivo validation of the SOS Bacteroidetes regulators. Induction factor (log2 ) of (A) S. agarivorans and (B) P. actiniarum genes as the ratio of normalized mRNA between cultures treated with mitomycin C and untreated cultures, determined by RT-qPCR. The error bars represent the standard error of the mean of two independent biological replicates. Figure 5. Binding specificity of SOS Bacteroidetes regulators. EMSA with purified (A) S. agarivorans (Saga ) and (B) P. actiniarum (Pact ) putative SOS regulator proteins on the promoter region of the genes encoding UmuD and RecA orthologs, respectively. The ‘–’ symbol denotes absence of protein and ‘+’ the presence of protein in the mixture. The ‘C’ symbol indicates the presence of unlabeled competitor DNA. of S. agarivorans, substitutions on either motif dyad and Together, these results demonstrate that the two putative changes to the spacer resulted in the abolishment of bind- Bacteroidetes SOS-like regulators assessed here control the ing by the putative regulator, while substitutions in dyad- expression of canonical and putative SOS genes in response adjacent and spacer positions had no effect. Furthermore, to DNA-damage by binding specifically to their predicted electromobility-shift assays on the upstream region of the binding motifs and that, like LexA, they act as global tran- genes encoding the P. actiniarum and S. agarivorans putative scriptional repressors. The presence of homologs for these SOS regulators, containing instances, respectively, of the two proteins and of predicted SOS-like networks using their GGA-N5-TCC and CTAA-N5-TTAG motifs, determined respective motifs in multiple Bacteroidetes species suggests that both proteins are capable of binding their respective that these proteins are likely bona fide regulators of the SOS promoters (Supplementary Figure S6). response in the Bacteroidetes phylum. Nucleic Acids Research, 2021, Vol. 49, No. 19 11059 Figure 6. In vitro validation of the SOS Bacteroidetes regulators binding motifs. EMSA with purified (A) S. agarivorans (Saga ) and (B) P. actiniarum (Pact ) putative SOS regulator proteins on the wild-type and site-directed mutagenesis variants of the promoter region of the genes encoding UmuD and RecA orthologs, respectively. The ‘–’ symbol denotes absence of protein and ‘+’ the presence of protein in the mixture. Distribution of the SOS response and its regulators in the The analysis also revealed extensive paralogy in the Bacteroidetes CTAA-N5-TTAG ortholog group, with some species in the Weeksellaceae family harboring up to five homologs The results presented above established the presence of two of these SOS regulators. Paralogy in the CTAA-N5-TTAG clearly differentiated members of a novel family of regu- group regulators is positively correlated (Spearman ␳ = lators controlling large SOS regulatory networks via dis- 0.35, P < 0.001) with the loss of the CTAA-N5-TTAG tinct motifs. Homology searches, however, revealed that sev- motif associated with this ortholog group. This suggested eral Bacteroidetes clades did not harbor homologs of these that duplication events might be facilitating divergence in regulators. To investigate the evolutionary history of this SOS regulator homologs and their associated binding mo- transcriptional response in the Bacteroidetes, we performed tif. To investigate this hypothesis, we examined the promoter Bayesian phylogenetic inference on RecA protein sequences region of Bacteroidetes SOS regulators that did not con- from representative species harboring these novel SOS reg- tain instances of the CTAA-N5-TTAG motif for the pres- ulators and representative members of all Bacteroidetes or- ence of alternative palindromes that could constitute di- ders. On the inferred tree, we annotated the presence of SOS vergent binding motifs. We identified several instances of regulators mapping to the GGA-N5-TCC and CTAA-N5- putative diverged binding motifs following the dyad-spacer TTAG motif clusters, their putative regulation by the cor- palindromic structure of the CTAA-N5-TTAG motif (Fig- responding motif and the inferred size of the regulon based ure 8A; Supplementary Table S10). To assess whether these on the number of known SOS genes with evidence of regu- predicted palindromes could be functional binding sites for lation under the corresponding motif. their cognate SOS regulators, we experimentally confirmed The resulting tree (Figure 7, Supplementary Data S8), the specific binding of the E. anophelis strain DSM 29660 in broad agreement with previously reported Bacteroidetes SOS regulator (WP 078407279.1) to the predicted binding phylogenies (64), revealed remarkable diversity in the dis- site TTAC-AAATT-GTAA in its promoter sequence (Fig- tribution of these two SOS regulator ortholog groups ure 8B). We also determined that the gene encoding this across the Bacteroidetes. Orthologs mapping to the CTAA- transcriptional regulator and its tandem opposite, encod- N5-TTAG group are widely distributed across the Bac- ing an error-prone polymerase, are induced by mitomycin teroidetes, whereas the GGA-N5-TCC group is found pri- C treatment (Figure 8C). marily in the Flavobacteriaceae family. Both groups of orthologs are absent in the Bacteroidales order, and in the Capnocytophaga and Cytophagaceae sequence diver- DISCUSSION gence in the identified homologs prevented their unequiv- ocal assignment to either cluster. Regulon size varies sig- A canonical SOS response under control of non-canonical LexA proteins nificantly, but is fairly cohesive within phylogenetic clades, with several families presenting large (8.90 ± 3.27 oper- Comparative analyses have established that error-prone ons) regulons under the CTAA-N5-TTAG (Hymenobac- polymerases constitute the conserved core SOS response teraceae, Sphingobacteriaceae and Chitinophagaceae) and across bacteria, presumably due to the need to mitigate GGA-N5-TCC (Flavobacteriaceae) motifs. Interestingly, the high mutagenic load of unregulated error-prone poly- regulon size is markedly small in species related to clades merases (1,10,15). Leveraging this insight, we performed a where SOS regulators are absent or highly diverged, sug- systematic comparative analysis of error-prone polymerase gesting that loss of regulation is associated with regu- regulation in the Bacteroidetes. Motif discovery (Figure lon contraction. Within the Flavobacteriaceae, a significant 1AB) elicited two putative regulatory motifs (GGA-N5- number of species present homologs mapping to both or- TTC and CTAA-N5-TTAG) upstream of genes encoding tholog groups, but the predicted SOS regulon is under con- error-prone polymerases in the Bacteroidetes. Compara- trol of the GGA-N5-TCC group regulator. tive genomics analyses of the putative regulatory networks 11060 Nucleic Acids Research, 2021, Vol. 49, No. 19 Figure 7. Rooted Bayesian consensus tree of Bacteroidetes RecA protein sequences. Branch width denotes support values. Next to each tip label, infor- mation for both Bacteroidetes SOS regulator types (pink/green for the regulators targeting GGA-N5-TTC and CTAA-N5-TTAG motifs, respectively) is represented with three items: colored numbers indicate the number of encoded SOS regulator genes; pie charts represent the fraction of regulated (blue) and not regulated (yellow) SOS regulators; colored circles denote the number of additional genes predicted to be regulated by the respective SOS regulators. Gray numbers stand for unclassified SOS regulators. Families of Bacteroidetes that are discussed in the text are highlighted with circled numbers. The tree was rooted using the Balneolaeota vulgaris RecA protein sequence as an outgroup. Nucleic Acids Research, 2021, Vol. 49, No. 19 11061 Figure 8. Analysis of divergent palindromes in the promoter region of apparently unregulated Bacteroidetes SOS regulators. (A) Inferred phylogeny for the Weeksellaceae family (Figure 7) showing predicted binding sites (MEME predicted CTAA-N5-TTAG sites and identified perfect palindromes) for each SOS regulator paralog. MEME predicted sites are highlighted in green. (B) Electromobility-shift assays with the purified E. anophelis SOS regulator on the wild-type and site-directed mutagenesis variants of the promoter region of its encoding gene. The ‘–’ symbol denotes absence of protein and ‘+’ the presence of protein in the mixture. (C) Gene expression factor (log2 ) of E. anophelis genes after mitomycin C induction as determined by RT-qPCR. The genomic environment of the SOS regulator encoding gene is also shown. encoded by these motifs (Figure 1AB) revealed a large reg- cation forks or acting as an alternative repair pathway for ulatory network encompassing many well-documented ele- double-stranded breaks (67,68). ments of the SOS response (recA, dinB, umuDC, uvrC, uvrD, In the absence of bona fide LexA homologs, the compara- imuA-imuB-dnaE2, ssb, splB and radC; (3,7,40,57,59)). tive genomics analysis identified two orthologous groups of Beyond canonical SOS genes, the analysis also con- putative phage repressors, each one showing clear evidence firmed the regulation of several genes that have been pre- of regulation by one of the identified motifs (Figure 1AB), viously identified as members of the SOS response only as possible regulators of the inferred SOS-like networks. in a few species. The SOS regulation of a putative uracil- Both groups of repressors present the signature residues of DNA glycosylase-based DNA base excision repair system S24 peptidases (Figure 1C), and we confirmed that exem- has been reported in the Verrucomicrobia and the Al- plars of each group are DNA-damage inducible (Figure 4). phaproteobacteria (7,65). Our findings in the Bacteroidetes Furthermore, both regulators are capable of binding specif- suggest that uracil-DNA glycosylases may be a relatively ically to their assigned motifs, and target genes were shown common feature of the SOS response, aimed at address- to be induced by mitomycin C treatment (Figure 4). Taken ing ribonucleotide misincorporation by SOS-induced error- together, these results demonstrate that these S24-family prone polymerases. In a similar vein, regulation was also phage-like repressors control the expression of SOS genes predicted for an operon encoding the A (parC) and B (parE) in multiple Bacteroidetes species. LexA proteins have been DNA topoisomerase IV subunits, and their induction by reported to act as transcriptional activators (69) and, in mitomycin C was confirmed by RT-qPCR (Figure 4A). the absence of non-cleavable mutant data, it cannot be dis- The regulation of topoisomerases by the SOS response has carded that these S24-family phage-like repressors activate been described in the Vibrionaceae (59,66), and the regu- expression of SOS genes following DNA damage induction lation of parCE in the Bacteroidetes indicates that topoi- via a hitherto unknown mechanism. The sequence conser- somerases may also be a recurring component of the SOS vation of both S24 signature elements (the scissile Ala-Gly response, participating in the resolution of converging repli- bond and the Ser-Lys catalytic dyad; Figure 1C) in these 11062 Nucleic Acids Research, 2021, Vol. 49, No. 19 proteins and the robust overlay of their catalytic C-terminal by the wing of the winged HTH domain (12). Given the domains with that of self-cleavable S24 enzymes (Figure absence of such a wing in Bacteroidetes SOS regulators, but 3E), however, strongly suggests that these S24-family phage- their consistent targeting of fixed-spacer motifs (Figure 6), like repressors operate as SOS transcriptional repressors in we speculate that the additional C-terminal ␣-helix may be this phylum. involved in dimer stabilization, leading to rigid recognition DNA-damage inducible control of SOS genes by S24 of fixed-spacer motifs. peptidases other than LexA has been reported before in two Our results indicate that repressors of putative phage ori- bacterial families: the Streptococcaceae and the Moraxel- gin have taken up SOS regulatory functions at least twice laceae. However, in these families the corresponding S24- (in the Streptococcaceae and the Bacteroidetes) through a family regulators (HdiR and UmuDab) control only the in- process of convergent evolution. In the Bacteroidetes, this duction of a minimal SOS regulon, composed primarily of has resulted in phage-like repressors orchestrating a classi- error-prone polymerases (17,20). Their independent evolu- cal SOS response. Given that the canonical LexA is also a tion in two unrelated bacterial clades suggests that regula- member of the S24 family, and that convergent evolutionary tion of error-prone polymerases by S24 peptidases is a con- processes seem to be capable of replicating the conventional vergent evolutionary process driven by the need to silence SOS response with alternative repressors, it is plausible to their expression in non-stressed conditions. Our results indi- assume that the canonical LexA repressor may also have cate that this convergent evolutionary process does not nec- originated from the capture of a bacteriophage repressor by essarily stop with the regulation of error-prone polymerases, an early ancestor of extant bacterial groups (Figure 9). but can continue with the incorporation of DNA repair and Direct links between the SOS response and bacterio- recombination enzymes into a full-fledged SOS response. phages have been noted before (1). The lytic cycle of tem- perate phages is usually controlled by CI-like repressors, but in several instances the host LexA protein is used to re- Functional definition of SOS transcriptional repressors press lytic genes, either directly or through its interaction The discovery of a group of repressors with structural simi- with repressors or anti-repressors (73–76). These examples, larity to phage repressors controlling a canonical SOS re- together with additional instances of LexA regulating the sponse in the Bacteroidetes poses relevant questions re- mobilization of other genetic elements (1), blur the lines re- garding the origin of this transcriptional response and its garding the definition of the SOS response. The common repressor. The origin of S24-family repressors previously denominator in all these cases is the adaptive advantage of shown to regulate SOS genes is well-defined. UmuDAb responding to DNA damage conferred by the LexA repres- has been shown to cluster with regular UmuD proteins in sor both to bacterial cells (which can activate repair/bypass the Moraxellaceae (70), whereas HdiR clearly clusters with mechanisms) and to bacteriophages and other mobile ge- Streptococcus phage repressors in Figure 2. In contrast, netic elements (which can seek out undamaged hosts). The phylogenetic and multiple sequence alignment analyses of here-reported ability of putative phage repressors to coordi- the Bacteroidetes SOS regulators and other S24 peptidases nate the regulation of DNA repair and translesion synthesis (Figure 1C, Figure 2) indicate that these regulators define a pathways indicates that the connection between the bacte- new family of S24 regulators restricted to the Bacteroidetes. rial SOS response and bacteriophages is a two-way process, Structural comparisons with the reference S24 pepti- with both entities capable of co-opting the other’s DNA dase structures of E. coli LexA and Enterobacteria phage damage response mechanisms. Lambda CI show marked divergence from key domains of In this context, and given the inability of structural in- the predicted structures of Bacteroidetes SOS regulators. formation to clearly distinguish between the different tran- The structurally-predicted C-terminal domains overlay re- scriptional repressors involved, we advocate the adoption markably well with both the LexA and CI C-terminal do- of a functional nomenclature. The term lex was first coined mains (Figure 3CE), responsible for dimerization and au- to designate the ‘locus for X-ray sensitivity’ correspond- tocatalytic cleavage, suggesting that functional constraints ing to lexA in the E. coli genome (77) and, subsequently, have preserved the structure of this domain in spite of sub- the name lexA has been consistently used to refer to the stantial sequence divergence, as has been noted before for DNA-damage inducible repressor of the SOS response. LexA, UmuD and CI (71). We hence propose to designate hereafter as ‘LexA’ those In contrast with the C-terminal domains, the chromosomally-encoded repressors in control of a canon- structurally-predicted N-terminal domains of Bac- ical SOS network (DNA repair, translesion synthesis and teroidetes SOS regulators do not superpose well with the cell-cycle arrest), including the Bacteroidetes SOS repres- winged HTH N-terminal domain of LexA (Figure 3BD). sors here described. Instead, these N-terminal domains contain four predicted ␣-helices that overlay well with the N-terminal domain Motif divergence and convergent evolution of SOS regulation of CI. This difference suggests that, like HdiR, these Bacteroidetes regulators are most likely of phage origin. In A prominent feature of the SOS response is the documented this context, it is worth noting that the C-terminal domain tendency of the LexA transcriptional repressor to dramati- of these regulators harbors a predicted ␣-helix that is not cally alter its binding motif throughout the course of evolu- present in LexA or CI structures. While CI repressors can tion, with 17 distinct LexA-binding motifs reported to date target dyad motifs with variable spacing (72), LexA pro- in different bacterial groups (1,8,7,9,10). Hypervariability teins are known to target motifs with rigid spacers, and this in the LexA-binding motif could in principle be explained feature has been attributed to stabilizing contacts mediated by inherent lability in the LexA DNA-binding domain, but Nucleic Acids Research, 2021, Vol. 49, No. 19 11063 Figure 9. Uptake of the SOS regulatory network by a phage lytic cycle repressor. (1) A temperate phage infects a bacterial cell containing unregulated SOS genes (green) and integrates in its chromosome as a prophage. (2 & 3 ) Standard pathway for prophage lytic development under control of the phage repressor (pink). (2) Mutational events disrupt the prophage, rendering it inactive. (3) The prophage repressor (pink arrow) gradually takes up regulation of cellular SOS genes (green arrows), yielding a functional SOS response. This figure was constructed using BioRender templates. Figure 10. Model for SOS network evolution. (1) LexA controls the expression of the SOS network, regulating itself and other SOS genes. (2) A lexA gene duplication takes place. (3) One of the LexA proteins diverges, altering its LexA-binding motif. (4) Upon deletion of the primary LexA, the diverged LexA protein gradually reuptakes the regulon, using the novel LexA-binding motif. 11064 Nucleic Acids Research, 2021, Vol. 49, No. 19 the LexA-binding motif has been documented to be ex- Accession numbers for all other sequences and structures tremely stable across broad phylogenetic groups, suggest- used in this study are provided in the supplementary mate- ing that other factors are involved (16,78,79). Recent and rials. ancient duplications of the lexA gene have been reported in multiple bacterial groups, and shown to correlate with SUPPLEMENTARY DATA the emergence of novel motifs (63,75,80,8). Duplications in transcription factors can rapidly lead to motif divergence Supplementary Data are available at NAR Online. and the uptake of novel regulatory functions (81,82). In the case of LexA, it has been proposed that, after duplication, a ACKNOWLEDGEMENTS diverged LexA protein may reuptake SOS regulatory func- tions following the loss of the primary lexA gene, resulting The authors wish to thank Joan Ruiz and Susana Escribano in an apparent change to the LexA-binding motif (7) (Fig- for their technical support during some of the experimental ure 10). procedures, as well as Ángela Martı́nez-Mateos for her con- The results reported here depict a strikingly similar pic- tinued support. ture for a novel SOS repressor that is not a homolog of LexA. As in the case of conventional LexA proteins, the FUNDING Bacteroidetes LexA shows clear evidence of gene duplica- tion, with several groups harboring multiple paralogs of the Spanish Ministerio de Economı́a y Competitividad CTAA-N5-TTAG group LexA, as well as coexistence of [BIO2016-77011-R to J.B.]; National Science Foundation Bacteroidetes LexA proteins targeting the GGA-N5-TCC (NSF) [CHE1844624 to A.T.S]. M.S.-O. was the recip- and CTAA-N5-TTAG motifs in the same species (Figure ient of a predoctoral fellowship from the Ministerio de 7). Furthermore, our data also demonstrate that degener- Educación, Cultura y Deporte de España. Funding for ate variants of the CTAA-N5-TTAG motif are functional open access charge: Spanish Ministerio de Economı́a y in clades with multiple parologs (Figure 8; Supplementary Competitividad. Table S10), and support the notion that gradual regulon re- Conflict of interest statement. None declared. duction, and the eventual loss of regulation, are strongly as- sociated with duplication events. The fact that the evolution REFERENCES of the SOS response in the Bacteroidetes, under control of 1. Erill,I., Campoy,S. and Barbe,J. (2007) Aeons of distress: an a non-canonical LexA protein with a distinct DNA bind- evolutionary perspective on the bacterial SOS response. FEMS ing domain (Figure 3B, D), parallels that of previously re- Microbiol. Rev., 31, 637–656. ported SOS systems under control of canonical LexA pro- 2. Baharoglu,Z. and Mazel,D. (2014) SOS, the formidable strategy of bacteria against aggressions. FEMS Microbiol. Rev., 38, 1126–1145. teins suggests that hypervariability in LexA-binding motifs 3. Fernandez De Henestrosa,A.R., Ogi,T., Aoyagi,S., Chafin,D., is not due to specific DNA-binding particularities of the Hayes,J.J., Ohmori,H. and Woodgate,R. (2000) Identification of LexA repressor, but rather it is an intrinsic property of the additional genes belonging to the LexA regulon in Escherichia coli. SOS response. Mol. Microbiol., 35, 1560–1572. 4. Walker,G.C., Smith,B.T. and Sutton,M.D. (2000) The SOS response to DNA damage. In: Storz,G. and Hengge-Aronis,R. (eds). Bacterial DATA AVAILABILITY Stress Responses. Washington, D.C., pp. 131–144. 5. Sassanfar,M. and Roberts,J.W. (1990) Nature of the SOS-inducing Python scripts used for obtaining and analyzing genomic signal in Escherichia coli. The involvement of DNA replication. J. data are available in the GitHub repository (https://github. Mol. Biol., 212, 79–96. 6. Kovačič,L., Paulič,N., Leonardi,A., Hodnik,V., Anderluh,G., com/ErillLab/SOS Bacteroidetes/). All the genome and Podlesek,Z., Žgur-Bertok,D., Križaj,I. and Butala,M. (2013) protein sequence data used in this work is available in Structural insight into LexA–RecA* interaction. Nucleic Acids Res., the NCBI RefSeq database (https://www.ncbi.nlm.nih.gov/ 41, 9901–9910. refseq/). The protein structures used in this study are avail- 7. Erill,I., Campoy,S., Kılıç,S. and Barbé,J. (2016) The Verrucomicrobia LexA-binding motif: insights into the evolutionary dynamics of the able at the Protein Data Bank (https://www.rcsb.org/). The SOS response. Front. Mol. Biosci., 3, 33. T-COFFEE multiple sequence alignment suite is available 8. Sanchez-Alberola,N., Campoy,S., Emerson,D., Barbe,J. and Erill,I. at http://www.tcoffee.org/. The HMMER suite is available (2015) An SOS regulon under control of a noncanonical at http://hmmer.org/. CGB is a software suite for compara- LexA-binding motif in the Betaproteobacteria. J. Bacteriol., 197, tive analysis of prokaryotic regulatory networks available in 2622–2630. 9. Sánchez-Osuna,M., Barbé,J. and Erill,I. (2017) Comparative the GitHub repository (https://github.com/ErillLab/cgb/). genomics of the DNA damage-inducible network in the The multiple sequence aligner CLUSTALW is available at Patescibacteria. Environ. Microbiol., 19, 3465–3474. http://www.clustal.org/. The PyMol molecular visualization 10. Kılıç,S., Sánchez-Osuna,M., Collado-Padilla,A., Barbé,J. and Erill,I. suite is available at https://pymol.org/. The Robetta protein (2020) Flexible comparative genomics of prokaryotic transcriptional regulatory networks. BMC Genomics, 21, 466. structure prediction service is available at https://robetta. 11. Voordeckers,K., Pougach,K. and Verstrepen,K.J. (2015) How do bakerlab.org/. The phylogenetic inference software Mr- regulatory networks evolve and expand throughout evolution? Curr. Bayes is available at https://nbisweden.github.io/MrBayes/. Opin. Biotechnol., 34, 180–188. TrRosetta models for S. agarivorans DSM 23515 12. Zhang,A.P.P., Pigli,Y.Z. and Rice,P.A. (2010) Structure of the [WP 075324850.1] and P. actiniarum DSM 19842 LexA–DNA complex and implications for SOS box measurement. Nature, 466, 883–886. [WP 025606010.1] are available in ModelArchive 13. Phizicky,E.M. and Roberts,J.W. (1980) Kinetics of RecA (https://www.modelarchive.org/) under accession num- protein-directed inactivation of repressors of phage lambda and bers ma-gno0n and ma-l80i0, respectively. phage P22. J. Mol. Biol., 139, 319–328. Nucleic Acids Research, 2021, Vol. 49, No. 19 11065 14. Reuven,N.B., Arad,G., Maor-Shoshani,A. and Livneh,Z. (1999) The 33. Song,Y., DiMaio,F., Wang,R.Y.-R., Kim,D., Miles,C., Brunette,T., mutagenesis protein UmuC is a DNA polymerase activated by Thompson,J. and Baker,D. (2013) High-resolution comparative UmuD’, RecA, and SSB and is specialized for translesion replication. modeling with RosettaCM. Struct. Lond. Engl., 21, 1735–1742. J. Biol. Chem., 274, 31763–31766. 34. Raman,S., Vernon,R., Thompson,J., Tyka,M., Sadreyev,R., Pei,J., 15. Jaszczur,M., Bertram,J.G., Robinson,A., van Oijen,A.M., Kim,D., Kellogg,E., DiMaio,F., Lange,O. et al. (2009) Structure Woodgate,R., Cox,M.M. and Goodman,M.F. (2016) Mutations for prediction for CASP8 with all-atom refinement using Rosetta. worse or better: low-fidelity DNA synthesis by SOS DNA polymerase Proteins, 77, 89–99. V is a tightly regulated double-edged sword. Biochemistry, 55, 35. DeLano,W.L. (2002) Pymol: an open-source molecular graphics tool. 2309–2318. CCP4 Newsl. Protein Crystallogr., 40, 82–92. 16. Erill,I., Jara,M., Salvador,N., Escribano,M., Campoy,S. and Barbe,J. 36. Notredame,C., Higgins,D.G. and Heringa,J. (2000) T-Coffee: a novel (2004) Differences in LexA regulon structure among Proteobacteria method for fast and accurate multiple sequence alignment. J. Mol. through in vivo assisted comparative genomics. Nucleic. Acids. Res., Biol., 302, 205–217. 32, 6617–6626. 37. Huang,X. and Miller,W. (1991) A time-efficient, linear-space local 17. Savijoki,K., Ingmer,H., Frees,D., Vogensen,F.K., Palva,A. and similarity algorithm. Adv. Appl. Math., 12, 337–357. Varmanen,P. (2003) Heat and DNA damage induction of the 38. Castresana,J. (2000) Selection of conserved blocks from multiple LexA-like regulator HdiR from Lactococcus lactis is mediated by alignments for their use in phylogenetic analysis. Mol. Biol. Evol., 17, RecA and ClpP. Mol. Microbiol., 50, 609–621. 540–552. 18. Varhimo,E., Savijoki,K., Jalava,J., Kuipers,O.P. and Varmanen,P. 39. Ronquist,F. and Huelsenbeck,J.P. (2003) MrBayes 3: Bayesian (2007) Identification of a novel streptococcal gene cassette mediating phylogenetic inference under mixed models. Bioinformatics, 19, SOS mutagenesis in Streptococcus uberis. J. Bacteriol., 189, 1572–1574. 5210–5222. 40. Erill,I. (2006) Dispersal and regulation of an adaptive mutagenesis 19. Boutry,C., Delplace,B., Clippe,A., Fontaine,L. and Hols,P. (2013) cassette in the bacteria domain. Nucleic. Acids. Res., 34, 66–77. SOS response activation and competence development are 41. Rambaut,A., Drummond,A.J., Xie,D., Baele,G. and Suchard,M.A. antagonistic mechanisms in Streptococcus thermophilus. J. Bacteriol., (2018) Posterior summarization in Bayesian phylogenetics using 195, 696–707. tracer 1.7. Syst. Biol., 67, 901–904. 20. Hare,J.M., Perkins,S.N. and Gregg-Jolly,L.A. (2006) A constitutively 42. Letunic,I. and Bork,P. (2019) Interactive Tree Of Life (iTOL) v4: expressed, truncated umuDC operon regulates the recA-dependent recent updates and new developments. Nucleic. Acids. Res., 47, DNA damage induction of a gene in Acinetobacter baylyi strain W256–W259. ADP1. Appl. Environ. Microbiol., 72, 4036–4043. 43. Mayola,A., Irazoki,O., Martı́nez,I.A., Petrov,D., Menolascina,F., 21. Aranda,J., Poza,M., Shingu-Vázquez,M., Cortés,P., Boyce,J.D., Stocker,R., Reyes-Darias,J.A., Krell,T., Barbé,J. and Campoy,S. Adler,B., Barbé,J. and Bou,G. (2013) Identification of a (2014) RecA protein plays a role in the chemotactic response and DNA-damage-inducible regulon in Acinetobacter baumannii. J. chemoreceptor clustering of Salmonella enterica. PLoS One, 9, Bacteriol., 195, 5577–5582. e105578. 22. Hare,J.M., Ferrell,J.C., Witkowski,T.A. and Grice,A.N. (2014) 44. Cambray,G., Sanchez-Alberola,N., Campoy,S., Guerin,E., Da Re,S., Prophage induction and differential RecA and UmuDAb Gonzalez-Zorn,B., Ploy,M.C., Barbe,J., Mazel,D. and Erill,I. (2011) transcriptome regulation in the DNA damage responses of Prevalence of SOS-mediated control of integron integrase expression Acinetobacter baumannii and Acinetobacter baylyi. PLoS One, 9, as an adaptive trait of chromosomal and mobile integrons. Mob. e93861. DNA, 2, 6. 23. Hahnke,R.L., Meier-Kolthoff,J.P., Garcı́a-López,M., Mukherjee,S., 45. Aranda,J., Garrido,M.E., Cortés,P., Llagostera,M. and Barbé,J. Huntemann,M., Ivanova,N.N., Woyke,T., Kyrpides,N.C., (2008) Analysis of the protective capacity of three Streptococcus suis Klenk,H.-P. and Göker,M. (2016) Genome-based taxonomic proteins induced under divalent-cation-limited conditions. Infect. classification of Bacteroidetes. Front. Microbiol., 7, 2003. Immun., 76, 1590–1598. 24. Thomas,F., Hehemann,J.-H., Rebuffet,E., Czjzek,M. and Michel,G. 46. Tapias,A., Campoy,S. and Barbe,J. (2000) Analysis of the expression (2011) Environmental and gut Bacteroidetes: the food connection. of the Rhodobacter sphaeroides lexA gene. Mol. Gen. Genet., 263, Front. Microbiol., 2, 93. 957–965. 25. Haft,D.H., DiCuccio,M., Badretdin,A., Brover,V., Chetvernin,V., 47. Campoy,S., Mazon,G., Fernandez de Henestrosa,A.R., O’Neill,K., Li,W., Chitsaz,F., Derbyshire,M.K., Gonzales,N.R. et al. Llagostera,M., Monteiro,P.B. and Barbe,J. (2002) A new regulatory (2018) RefSeq: an update on prokaryotic genome annotation and DNA motif of the gamma subclass Proteobacteria: identification of curation. Nucleic Acids Res., 46, D851–D860. the LexA protein binding site of the plant pathogen Xylella fastidiosa. 26. Eddy,S.R. (2011) Accelerated profile HMM searches. PLoS Comput. Microbiology, 148, 3583–3597. Biol., 7, e1002195. 48. Campoy,S., Fontes,M., Padmanabhan,S., Cortes,P., Llagostera,M. 27. Tatusov,R.L., Fedorova,N.D., Jackson,J.D., Jacobs,A.R., and Barbe,J. (2003) LexA-independent DNA damage-mediated Kiryutin,B., Koonin,E.V., Krylov,D.M., Mazumder,R., induction of gene expression in Myxococcus xanthus. Mol. Mekhedov,S.L., Nikolskaya,A.N. et al. (2003) The COG database: an Microbiol., 49, 769–781. updated version includes eukaryotes. BMC Bioinformatics, 4, 41. 49. Campoy,S., Salvador,N., Cortes,P., Erill,I. and Barbe,J. (2005) 28. Dibrova,D.V., Konovalov,K.A., Perekhvatov,V.V., Skulachev,K.V. Expression of canonical SOS genes is not under LexA repression in and Mulkidjanian,A.Y. (2017) COGcollator: a web server for analysis Bdellovibrio bacteriovorus. J. Bacteriol., 187, 5367–5375. of distant relationships between homologous protein families. Biol. 50. Menouni,R., Champ,S., Espinosa,L., Boudvillain,M. and Ansaldi,M. Direct, 12, 29. (2013) Transcription termination controls prophage maintenance in 29. Bailey,T.L. and Elkan,C. (1994) Fitting a mixture model by Escherichia coli genomes. Proc. Natl. Acad. Sci. U.S.A., 110, expectation maximization to discover motifs in biopolymers. Proc. 14414–14419. Int. Conf. Intell. Syst. Mol. Biol., 2, 28–36. 51. Vickridge,E., Planchenault,C., Cockram,C., Junceda,I.G. and 30. Thompson,J.D., Higgins,D.G. and Gibson,T.J. (1994) CLUSTAL W: Espéli,O. (2017) Management of E. coli sister chromatid cohesion in improving the sensitivity of progressive multiple sequence alignment response to genotoxic stress. Nat. Commun., 8, 14618. through sequence weighting, position-specific gap penalties and 52. Courcelle,J., Khodursky,A., Peter,B., Brown,P.O. and Hanawalt,P.C. weight matrix choice. Nucleic Acids Res., 22, 4673–4680. (2001) Comparative gene expression profiles following UV exposure 31. Madeira,F., Park,Y., Lee,J., Buso,N., Gur,T., Madhusoodanan,N., in wild-type and SOS-deficient Escherichia coli. Genetics, 158, 41–64. Basutkar,P., Tivey,A.R.N., Potter,S.C., Finn,R.D. et al. (2019) The 53. Galisa,P.S., da Silva,H.A.P., Macedo,A.V.M., Reis,V.M., Vidal,M.S., EMBL-EBI search and sequence analysis tools APIs in 2019. Nucleic Baldani,J.I. and Simões-Araújo,J.L. (2012) Identification and Acids Res., 47, W636–W641. validation of reference genes to study the gene expression in 32. Hall,T.T. (1999) BioEdit: a user-friendly biological sequence Gluconacetobacter diazotrophicus grown in different carbon sources alignment editor and analysis program for Windows 95/98/NT. using RT-qPCR. J. Microbiol. Methods, 91, 1–7. Nucleic Acids Symp. Ser., 41, 95. 11066 Nucleic Acids Research, 2021, Vol. 49, No. 19 54. Rocha,D.J.P., Santos,C.S. and Pacheco,L.G.C. (2015) Bacterial (2002) Identification of a DNA nonhomologous end-joining complex reference genes for gene expression studies by RT-qPCR: survey and in bacteria. Science, 297, 1686–1689. analysis. Antonie Van Leeuwenhoek, 108, 685–693. 68. Suski,C. and Marians,K.J. (2008) Resolution of converging 55. Wen,S., Chen,X., Xu,F. and Sun,H. (2016) Validation of reference replication forks by RecQ and topoisomerase III. Mol. Cell, 30, genes for real-time quantitative PCR (qPCR) analysis of 779–789. Avibacterium paragallinarum. PLoS One, 11, e0167736. 69. Tapias,A., Fernandez,S., Alonso,J.C. and Barbe,J. (2002) 56. Liu,S., Meng,C., Xu,G., Jian,H. and Wang,F. (2018) Validation of Rhodobacter sphaeroides LexA has dual activity: optimising and reference genes for reverse transcription real-time quantitative PCR repressing recA gene transcription. Nucleic Acids Res., 30, 1539–1546. analysis in the deep-sea bacterium Shewanella psychrophila WP2. 70. Hare,J.M., Bradley,J.A., Lin,C. and Elam,T.J. (2012) Diverse FEMS Microbiol. Lett., 365, https://doi.org/10.1093/femsle/fny229. responses to UV light exposure in Acinetobacter include the capacity 57. Cirz,R.T., O’Neill,B.M., Hammond,J.A., Head,S.R. and for DNA damage-induced mutagenesis in the opportunistic Romesberg,F.E. (2006) Defining the Pseudomonas aeruginosa SOS pathogens Acinetobacter baumannii and Acinetobacter ursingii. response and its role in the global response to the antibiotic Microbiology, 158, 601–611. ciprofloxacin. J. Bacteriol., 188, 7101–7110. 71. Galkin,V.E., Yu,X., Bielnicki,J., Ndjonka,D., Bell,C.E. and 58. Aravind,L., Anand,S. and Iyer,L.M. (2013) Novel autoproteolytic Egelman,E.H. (2009) Cleavage of Bacteriophage ␭ cI repressor and DNA-damage sensing components in the bacterial SOS response involves the RecA C-terminal domain. J. Mol. Biol., 385, 779–787. and oxidized methylcytosine-induced eukaryotic DNA demethylation 72. Shearwin,K.E., Dodd,I.B. and Egan,J.B. (2002) The helix-turn-helix systems. Biol. Direct, 8, 20. motif of the Coliphage 186 immunity repressor binds to two distinct 59. Krin,E., Pierlé,S.A., Sismeiro,O., Jagla,B., Dillies,M.-A., Varet,H., recognition sequences. J. Biol. Chem., 277, 3186–3194. Irazoki,O., Campoy,S., Rouy,Z., Cruveiller,S. et al. (2018) Expansion 73. Shearwin,K.E., Brumby,A.M. and Egan,J.B. (1998) The Tum protein of the SOS regulon of Vibrio cholerae through extensive transcriptome of coliphage 186 is an antirepressor. J. Biol. Chem., 273, 5708–5715. analysis and experimental validation. BMC Genomics, 19, 73. 74. Quinones,M., Davis,B.M. and Waldor,M.K. (2006) Activation of the 60. Dullaghan,E.M., Brooks,P.C. and Davis,E.O. (2002) The role of Vibrio cholerae SOS response is not required for intestinal cholera multiple SOS boxes upstream of the Mycobacterium tuberculosis lexA toxin production or colonization. Infect. Immun., 74, 927–930. gene–identification of a novel DNA-damage-inducible gene. 75. Abella,M., Campoy,S., Erill,I., Rojo,F. and Barbe,J. (2007) Microbiology, 148, 3609–3615. Cohabitation of two different lexA regulons in Pseudomonas putida. J. 61. Luo,Y., Pfuetzner,R.A., Mosimann,S., Paetzel,M., Frey,E.A., Bacteriol., 189, 8855–8862. Cherney,M., Kim,B., Little,J.W. and Strynadka,N.C. (2001) Crystal 76. Fornelos,N., Bamford,J.K.H. and Mahillon,J. (2011) Phage-borne structure of LexA: a conformational switch for regulation of factors and host LexA regulate the lytic switch in phage GIL01. J. self-cleavage. Cell, 106, 585–594. Bacteriol., 193, 6008–6019. 62. Wintjens,R. and Rooman,M. (1996) Structural classification of HTH 77. Howard-Flanders,P. and Boyce,R.P. (1966) DNA repair and genetic DNA-binding domains and protein-DNA interaction modes. J. Mol. recombination: studies on mutants of Escherichia coli defective in Biol., 262, 294–313. these processes. Radiat. Res. 1966, 156–184. 63. Jara,M., Nunez,C., Campoy,S., Fernandez de Henestrosa,A.R., 78. Erill,I., Escribano,M., Campoy,S. and Barbe,J. (2003) In silico Lovley,D.R. and Barbe,J. (2003) Geobacter sulfurreducens has two analysis reveals substantial variability in the gene contents of the autoregulated lexA genes whose products do not bind the recA gamma proteobacteria LexA-regulon. Bioinformatics, 19, 2225–2236. promoter: differing responses of lexA and recA to DNA damage. J. 79. Cornish,J.P., Matthews,F., Thomas,J.R. and Erill,I. (2012) Inference Bacteriol., 185, 2493–2502. of self-regulated transcriptional networks by comparative genomics. 64. Garcı́a-López,M., Meier-Kolthoff,J.P., Tindall,B.J., Gronow,S., Evol Bioinform Online, 8, 449–461. Woyke,T., Kyrpides,N.C., Hahnke,R.L. and Göker,M. (2019) 80. Yang,M.-K., Hsu,C.-H. and Sung,V.-L. (2008) Analyses of binding Analysis of 1,000 type-strain genomes improves taxonomic sequences of the two LexA proteins of Xanthomonas axonopodis classification of Bacteroidetes. Front. Microbiol., 10, 2083. pathovar citri. Mol. Genet. Genomics MGG, 280, 49–58. 65. Paes da Rocha,R. (2008) Characterization of the SOS regulon of 81. Pérez,J.C., Fordyce,P.M., Lohse,M.B., Hanson-Smith,V., DeRisi,J.L. Caulobacter crescentus. J. Bacteriol., 190, 1209–1218. and Johnson,A.D. (2014) How duplicated transcription regulators 66. Sanchez-Alberola,N., Campoy,S., Barbe,J. and Erill,I. (2012) can diversify to govern the expression of nonoverlapping sets of Analysis of the SOS response of Vibrio and other bacteria with genes. Genes Dev., 28, 1272–1277. multiple chromosomes. BMC Genomics, 13, 58. 82. Pougach,K., Voet,A., Kondrashov,F.A., Voordeckers,K., 67. Weller,G.R., Kysela,B., Roy,R., Tonkin,L.M., Scanlan,E., Della,M., Christiaens,J.F., Baying,B., Benes,V., Sakai,R., Aerts,J., Zhu,B. et al. Devine,S.K., Day,J.P., Wilkinson,A., d’Adda di Fagagna,F. et al. (2014) Duplication of a promiscuous transcription factor drives the emergence of a new regulatory network. Nat. Commun., 5, 4868.