Activation of TnSmu1, an integrative and conjugative element, by an ImmR-like transcriptional regulator in Streptococcus mutans

Integrative and conjugative elements (ICEs) are chromosomally encoded mobile genetic elements that can transfer DNA between bacterial strains. Recently, as part of efforts to determine hypothetical gene functions, we have discovered an important regulatory module encoded on an ICE known as TnSmu1 on the Streptococcus mutans chromosome. The regulatory module consists of a cI-like repressor with a helix-turn-helix DNA binding domain immR Smu (immunity repressor) and a metalloprotease immA Smu (anti-repressor). It is not possible to create an in-frame deletion mutant of immR Smu and repression of immR Smu with CRISPRi (CRISPR interference) causes substantial cell defects. We used a bypass of essentiality (BoE) screen to discover genes that allow deletion of the regulatory module. This revealed that conjugation genes, located within TnSmu1, can restore the viability of an immR Smu mutant. Deletion of immR Smu also leads to production of a circular intermediate form of TnSmu1, which is also inducible by the genotoxic agent mitomycin C. To gain further insights into potential regulation of TnSmu1 by ImmRSmu and broader effects on S. mutans UA159 physiology, we used CRISPRi and RNA-seq. Strongly induced genes included all the TnSmu1 mobile element, genes involved in amino acid metabolism, transport systems and a type I-C CRISPR-Cas system. Lastly, bioinformatic analysis shows that the TnSmu1 mobile element and its associated genes are well distributed across S. mutans isolates. Taken together, our results show that activation of TnSmu1 is controlled by the immRA Smu module, and that activation is deleterious to S. mutans , highlighting the complex interplay between mobile elements and their host.


INTRODUCTION
Streptococcus mutans is a gram-positive bacterium that colonizes the human oral cavity [1]. Like many streptococci, S. mutans can cause disease when environmental conditions are favourable. Poor oral hygiene combined with frequent ingestion of simple sugars creates an environment in which S. mutans can cause dental caries (tooth decay) [2,3]. If this disease is allowed to progress, it causes the breakdown of teeth, with symptoms that include pain, difficulty eating and tooth loss [2]. Due to its widespread presence in the oral cavity and the resulting disease burden, it is necessary to identify targets for improved therapeutics and understand processes essential for the pathogen. With this goal in mind, we have recently applied transposon sequencing (Tn-seq) and CRISPR interference (CRISPRi), and have identified >200 S. mutans essential genes that are required for the viability of the organism [4,5]. For S. mutans, most essential genes can be broadly sorted into three categories: processing of genetic information, energy production and maintenance of the cell envelope [4].
It was during Tn-seq/CRISPRi experiments that we discovered an essential gene, SMU_218, which is annotated as a transcriptional regulator, and is designated immR Smu (immunity repressor) from here forwards. immR Smu resides in a two-gene operon with SMU_219/immA Smu (anti-repressor). Probing immR Smu and immA Smu with smart (Simple Modular Architecture Research Tool) [6], we found that the N-terminus of immR Smu contains a cI-like repressor of phage λ DNA-binding domain, often found in streptococcal phages. immA Smu contains a putative ImmA/IrrE family metallo-endopeptidase domain, based on possession of a conserved metalloprotease zinc-binding motif, HEXXH [7]. In addition, the toxin-antitoxin database rasta [8] annotates OPEN ACCESS this two-gene operon as a putative toxin-antitoxin module. We found this interesting as the antitoxin component (e.g. immR Smu ) of toxin-antitoxin modules are often essential, because they prevent the accumulation of the toxin protein. While we cannot exclude the hypothesis that this two-gene operon is a toxin-antitoxin module, exhaustive literature searches provide evidence for other functions of these genes. First, model λ cI repressors have an N-terminal domain that binds DNA and a C-terminal domain that functions in cI autoproteolysis, in conjunction with RecA [9]. In S. mutans, it is possible that these functions are separated into two genes, immR Smu and immA Smu . Next, cI-like repressors regulate prophage induction in bacteria, but they have also been shown to regulate integrative and conjugative element (ICE) expression [10,11]. Notably, immR Smu and immA Smu reside in the flanking region of the putative ICE designated as TnSmu1. This large region of DNA (23 kb) contains predicted conjugation genes, as well as many hypothetical proteins, and is found in several strains of S. mutans [12,13]. Very little is known about the activity or the mobility of this element, but it shares similarities with ICESt1, ICESt3, Tn916 and ICEBs1. Although the regulatory regions of these different ICEs are substantially re-arranged compared to TnSmu1, ICESt1 and ICEBs1 contain cI-like repressors that repress ICE expression [10,11,14]. Therefore, we have formulated two major hypotheses explaining the essential nature of immR Smu : (i) immR Smu is an antitoxin in a toxin-antitoxin module or (ii) immR Smu is involved in the activation of TnSmu1. For the first hypothesis, CRISPR-mediated knockdown of immR Smu would lead to over-accumulation of immA Smu , which would then act as a toxin. For the second hypothesis, loss of the repressor would lead to activation of TnSmu1 with loss of the element from the cell making immR Smu appear 'essential' .
Here, we investigate both of our hypotheses for explaining the essentiality of immR Smu . Using transposon mutagenesis, genome sequencing, RNA-seq, bioinformatics and standard molecular microbiology techniques, we find that immR Smu is very likely to be a repressor of the ICE TnSmu1 and is related to the ICEBs1 immunity repressor immR. Removal of immR Smu repression causes up-regulation of the TnSmu1 element, leading to excision from the genome and formation of a circular intermediate. We found that TnSmu1 is activated by DNA damage and we posit that this occurs via ImmA Smu -mediated cleavage of the ImmR Smu repressor. We also provide evidence of broader effects on S. mutans physiology, including slowed growth and disrupted cell morphology when TnSmu1 is activated. In summary, we have discovered an important regulatory module controlling TnSmu1 activation, which extends prior findings made in ICEBs1, and illustrates the complex relationship between mobile elements and their hosts.

Bacterial strains and culture conditions
S. mutans strains were cultured from single colonies in brain heart infusion (BHI) broth (Difco). Unless otherwise stated, S. mutans was routinely cultured at 37 °C in a 5 % CO 2 , microaerophilic atmosphere. Escherichia coli strains were routinely cultured in LB broth (Lennox formula; 10 g tryptone l −1 , 5 g yeast extract l −1 and 5 g NaCl l −1 ) at 37 °C with aeration. Antibiotics were added to growth media at the following concentrations: kanamycin (1.0 mg ml −1 for S. mutans, 50 µg ml −1 for E. coli), spectinomycin (1.0 mg ml −1 for S. mutans), ampicillin (100 µg ml −1 for E. coli). A list of strains and plasmids (Table S1) and oligonucleotide primers (Table S2) can be found in the supplementary material.

Gene mutagenesis and plasmid cloning
Standard DNA manipulation techniques were used to engineer deletion mutant strains [15]. A PCR ligation mutagenesis method was used to replace genes with non-polar kanamycin markers [16]. For each gene deletion, primers A and B were designed to amplify 500-600 bp upstream of the coding sequence (with approximately 50 bp overlapping the coding sequence of the gene). Primers C and D were designed to amplify 500 to 600 bp downstream of the coding sequence (with approximately 50 bp overlapping the coding sequence of the gene). Primers B and C contained BamHI restriction enzyme sites for ligation of the AB and CD fragments to a non-polar kanamycin cassette digested from plasmid pALH124 [17]. Transformants were selected on BHI agar containing kanamycin. Double-crossover recombination, without introduction of nearby secondary mutations, was confirmed by PCR and Sanger sequencing using primers E and F, away from the site of recombination.
Plasmid cloning was conducted using the protein expression plasmid pBAD/His/A that contains an arabinose-inducible promoter for tightly regulated protein production. PCR products for SMU_218, SMU_219 and SMU_218-219 were amplified and digested, before being ligated into digested pBAD/His/A plasmid. Correct in-frame insertion was verified with Sanger sequencing.

Transmission electron microscopy (TEM)
Overnight cultures of CRISPRi strains were diluted 1 : 100 into 200 µl FMC-maltose without or with 0.025 % xylose, and then incubated at 37 °C in a 5 % CO 2 incubator for 16 h. Afterwards, cells were rinsed with 0.1M sodium cacodylate buffer and then fixed in 3 % glutaraldehyde overnight at 4 °C. The following day, cells were treated with 1.5 % osmium tetroxide in the dark for 1 h at 4 °C. Afterwards, the cells were mixed in equal parts with 5 % agarose in PBS, collected by centrifugation at 2000 g, and cooled to 4 °C. Small chunks of the bacterial pellet plus agarose were then dehydrated in ethanol via the following steps: 30%, 50%, 70%, 80%, 90 % each for 15 min, 99 % 10 min, and then absolute ethanol 2×10 min. The dehydrated cells were embedded in an epoxy resin, sectioned and stained with uranyl acetate and lead citrate. Microscopy was conducted using a Hitachi H7600 transmission electron microscope.

Transposon interaction screen
To look for genetic interactions of immR Smu with other genes, a previously created Tn-seq library [4] was transformed with an immRA Smu deletion product (described above). Transformation of the library was completed in triplicate and each transformation reaction was screened on at least ten agar plates containing spectinomycin (transposon cassette) and kanamycin (immRA Smu deletion construct). Colonies that grew and were potentially viable immRA Smu deletions (plus a transposon cassette) were picked and plated onto fresh antibiotic-containing media. After this, colonies were cultured overnight and frozen in glycerol at −80 °C. In addition, control experiments where conducted. Controls included attempting to delete immRA Smu with the deletion construct and selection of the Tn-seq library on kanamycin (to observe background spontaneous resistance).

Interaction screen genome sequencing analysis
Genomic DNA was isolated from strains using a MasterPure Gram positive DNA (Epicentre) purification kit with modifications as previously described [18]. After DNA purification, total DNA concentration and purity was measured using a NanoDrop spectrophotometer (Thermo Fisher Scientific). Purified genomic DNA was sent to the SeqCenter (Pittsburgh, USA), and samples were sequenced according to their protocol. Sequencing reads were compared to a S. mutans UA159 GenBank file (accession no. NC_004350.2) using Breseq. Following genomic sequencing, transposon insertion locations were verified with PCR and Sanger sequencing (primers are detailed in Table S2). Correct deletion of immRA Smu was also verified using the primer pair 218_219E and 219F, followed by gel electrophoresis and comparison to expected product sizes.

Quantitative PCR (qPCR) measurements of TnSmu1 excision and circularization
qPCR was used to calculate the excision and circularization frequency of TnSmu1. Specific primers, which only create products during excision or circularization, were used in the qPCR assay (Table S2). Primers that amplified a chromosomal single copy gene, sloR, close to TnSmu1 were also used to standardize results. Primers were designed using the qPCR settings in the Primer3plus online application [19]. Cell samples were grown to mid-log phase in rich media and DNA was collected using a MasterPure Gram positive DNA (Epicentre) purification kit with modifications as previously described [18]. Standard curves for each primer pair were generated using eight 10-fold dilutions of PCR products, starting with 10 8 copies µl −1 . Triplicates of standard curve DNAs, samples and cDNA controls were added to wells containing iQ SYBR Green Supermix (Bio-Rad) with primers (0.4 µM). Thermocycling was carried out using a CFX Touch Real-Time PCR detection system (Bio-Rad) set to the following protocol: 40 cycles of 95 °C for 10 s and 60 °C for 45 s, with a starting cycle of 95 °C for 30 s.

RNA-seq analysis
RNA was extracted from OD 600 ~0.4-0.6 bacterial cultures using the RNeasy Mini kit. RNA was extracted from biological triplicates of CRISPRi strains carrying a short guide RNA (sgRNA) targeting SMU_218 in the presence or absence of the dead Cas9 (dCas9) inducing molecule xylose. RNA was also extracted from wild-type UA159, SMU_197 c::Tn ΔimmRA Smu and SMU_201 c::Tn ΔimmRA Smu . Next, RNA was sent to SeqCenter who generated RNA-seq libraries using Illumina stranded RNA library preparation with RiboZero Plus rRNA depletion. Sequencing was conducted on an Illumina platform providing up to 12 million paired end reads (2×51 bp) per sample. After sequencing, SeqCenter conducted a basic RNA-seq analysis pipeline that provided raw transcript level quantification. Afterwards, gene expression changes between samples were quantified with Degust (http://degust.erc.monash.edu/) using the edgeR methodology [20]. The original RNA-seq data from this study was uploaded to the GEO database (https://www.ncbi.nlm.nih.gov/geo/) with the accession number GSE202804.

TnSmu1 distribution bioinformatics
Upon acknowledging the TnSmu1 operon presence in S. mutans UA159, we isolated the gene sequences of all the individual genes by constructing a BED type file, which was parsed to the getfasta command from the Bedtools suite [21]. The resulting individual fasta files with each gene sequence were blasted against a collection of over 600 clinical isolate genomes in our possession. The blast output was then filtered with the objective of keeping only those strains showing both a per cent identity greater than 75 % and a ratio between the query length and the sequence length greater than 0.9. From this filtered blast output, we isolated the PID column for each isolate across all genes and generated a matrix as input for the heatmap. In parallel, we conducted a MLST analysis with all resulting strains in the curated blast output. The MLST sequence was then parsed to phyml [22] to reconstruct a maximum-likelihood phylogeny tree. Finally, the phylogenies and the heatmap were plot together in R statistical language [23] with packages ggplot2 [24] and ggtree [25].

Evidence that immR Smu is not an antitoxin
Following our initial observations, we began by investigating the possibility that immRA Smu is a toxin-antitoxin module. We reasoned that we would be able to mutagenize both immR Smu and immA Smu at the same time, as this would not lead to the accumulation of the putative toxin. For other toxin-antitoxin modules (e.g. mazEF or relBE) double knockout of both genes is possible, whereas single knockout of the antitoxin module is not permitted. Under the conditions tested, we have not been able to obtain a double mutant of immRA Smu (Fig. S1, available with the online version of this article). However, deletion of the putative metalloprotease, immA Smu , is permitted by S. mutans cells (Fig. S1). A lack of viability for a double knockout provides some evidence that immRA Smu is not a toxin-antitoxin module. To confirm, we also created over-expression strains in E. coli to test the toxicity of each protein. The immR Smu and immA Smu genes were cloned into a pBAD protein expression vector, where the genes are induced by the addition of the monosaccharide arabinose. When the expression of immA Smu was induced, there was only a minor impact on the growth rate of E. coli, suggesting that immA Smu accumulation in E. coli is not toxic (Fig. S2). Over-expression of the ImmR Smu protein did cause a moderate growth phenotype in E. coli. The growth phenotype was absent when immR Smu expression was not induced by arabinose. To summarize, the data do not support the hypothesis that ImmRA Smu constitute a toxin-antitoxin module.

Phenotypic defects caused by repression of immR Smu
Next, we wanted to employ CRISPRi as a tool to characterize any phenotypic impacts caused by repression of immR Smu . Knockdown of gene expression is performed by targeting the immR Smu gene with an sgRNA, which acts together with dCas9 to block gene transcription; this system is also inducible with xylose [5]. Depletion of immR Smu causes a substantial growth defect compared with the control strain (targeting a non-essential gene, lacG) (Fig. 1a). Having shown that immR Smu repression leads to a strong growth defect, we next investigated whether repression causes any cell morphology impacts. Using TEM, we were able to gather ultrastructural insights into cells experiencing immR Smu depletion. As shown in Fig. 1(b), when immR Smu was depleted, there was a sub-population of sgRNA-immR Smu cells with extreme cell morphology defects. Compared to the control strain (sgRNA-lacG), immR Smu depleted strains were bloated and appeared to be not dividing or not dividing correctly (at the cell septum) (Fig. 1b). Overall, repression of immR Smu has a substantial impact on the normal physiology of S. mutans.

Repression of immR Smu with CRISPRi strongly up-regulates TnSmu1
To begin to understand the function of immR Smu , we performed RNA-seq profiling of an S. mutans strain experiencing knockdown of the immR Smu gene using CRISPRi. As shown, when immR Smu is repressed by CRISPRi there is a strong growth defect (Fig. 1a). RNA was collected from uninduced and induced (with 0.1 % xylose) cultures grown to OD 600 ~0.4, and this RNA was then processed for RNA-seq. The RNA-seq analysis identified 134 genes whose log 2 fold change was ≥2 after repression of SMU_218 (False Discovery Rate (FDR)<0.05; Table S3). Of these genes, 34 were repressed and 100 were induced (Fig. 2a).
There are several notable trends identified by transcriptome analysis of the immR Smu depleted strain. Firstly, every gene within the TnSmu1 element is strongly induced (Fig. 2a, Table S3). Most TnSmu1 genes were up-regulated by greater than 20-fold, with SMU_204 c (hypothetical protein) induced by over 1000-fold (log 2 fold change 10.03). Of the 20 most strongly induced genes, all belonged to the TnSmu1 region. Visual inspection of the TnSmu1 region, with reads shown and comparable to a non-induced strain, highlights the strong induction across the element (Fig. 2b). This region is predicted to be an ICE and exists in the ICE database ICEberg 2.0 [26], but to our knowledge has not been experimentally verified. To be considered an ICE, these mobile elements must contain several important features: a recombinase (also known as an integrase), a specific site of attachment (typically a tRNA gene), conjugation machinery, a relaxase and an oriT site [27]. Several of these features exist within the TnSmu1 region (Fig. 2b). SMU_191 c encodes a putative tyrosine recombinase that may function to catalyse DNA breakage and rejoining. Immediately adjacent to the 5′ end of SMU_191 c is tRNA-leu, which is likely the integration site. The attachment (att) sites attL and attR have the sequence CTATACCGGCGGCCG. A putative MOB T family relaxase is encoded by the gene SMU_207 c and may function to convert a circular form of TnSmu1 into ssDNA prior to conjugation. Several genes within TnSmu1 are predicted to be members of a type IV secretion system (SMU_196 c, SMU_197 c, SMU_198 c, SMU_199 c, SMU_201 c and SMU_208 c). A putative oriT site is located between SMU_207 c and SMU_208 c, with the sequence ACCC CCCT ATTA GTAT CGGGGGG. This shares similarities with the oriT site in ICEBs1 ( ACCCCC CCACGCTA A CA GGGGGGT) and ICESt3 ( ACCCCC GATTTCTA A TA GGGGGGT); conserved nucleotides are underlined. Thus, the relaxase encoded by SMU_207 c may recognize this oriT sequence, where it would nick the ICE DNA to form transfer DNA (T-DNA). Altogether, TnSmu1 contains many of the genes that are required for an ICE, and may allow excision, replication, transfer and integration of the element. In addition, strong up-regulation of these genes upon immR Smu depletion is suggestive of ImmR Smu acting as a repressor of TnSmu1.
Worth highlighting is that two intergenic regions within TnSmu1 are strongly induced (Fig. 2b). The first region is between SMU_217 c and immR Smu , and the second region is between immA Smu and SMU_220 c. For the first region, we do not currently know what is being produced (e.g. small RNA, small peptide, antitermination system, etc.). The second region contains a chromosomally encoded type I toxin-antitoxin system [28]. Expression of a toxic peptide known as Fst-Sm, encoded by this system, is toxic when over-expressed in S. mutans (in the absence of the antitoxin) [28]. The mechanism of action of the toxic peptide is unknown but plasmid-encoded Fst peptides affect the integrity of the membrane, cause defects in cell division and sensitize cells to nisin (by altering the cell membrane) [29,30].
Several other genes are up-or down-regulated after immR Smu depletion. A broad overview of the effect on the S. mutans transcriptome is shown in Fig. 2(c), with categorization by Clusters of Orthologous Groups (COGs). Several genes involved in amino acid transport or metabolism are up-regulated, including a cysteine transport operon tcyDEFGH, argD, cysK, opcD, pepQ, ilvE, gatA, hipO, SMU_1216 c (putative cystine transporter), SMU_1486 c (putative histidinol-phosphatase) and SMU_1938 c (putative methionine transporter). Other transport systems that are also up-regulated include treB (trehalose PTS), fruA (fructose PTS), ammonium transporter (nrgA), bacteriocin/competence transporters (comB, SMU_1889 c, SMU_1897, comG, mutE2 and mutG), tauC and msmK. The expression of several purine metabolism genes is altered with purDEK up-regulated and purC down-regulated. All the genes in the type I-C CRISPR-Cas system are up-regulated when immR Smu is repressed. Several genes are down-regulated after CRISPRi repression of immR Smu , many of which are hypothetical proteins. Six genes with roles in translation are down-regulated, predominately 50S and 30S ribosomal proteins. Some, but not all, genes located in the genomic island TnSmu2 are down-regulated (mubD, mubC, mubH, mubG and mubE). This region encodes a hybrid nonribosomal peptide (NRP) and polyketide (PK) system that produces mutanobactin [31]. A putative acyl carrier protein (SMU_27, acpP), which may participate in PK synthesis, is also down-regulated [32].

Bypass of essentiality (BoE) screen reveals genes that when inactivated allow the deletion of immR Smu
Under the conditions we have tested, we are not able to obtain mutants of immR Smu or immRA Smu but deletion of immA Smu is permitted (Fig. S1). We hypothesized that deletion of the immRA Smu module might be possible if introduced into a transposon library previously generated by our laboratory [4]. This is because the Tn-seq library might contain mutants that interrupt a gene responsible for immRA Smu essentiality and, thus, would tolerate immRA Smu deletion. This is known as BoE, with systematic studies of essential gene bypass having been conducted in yeast [33,34]. BoE has been shown for several essential genes in bacteria, including zipA, c-di-AMP null mutants, RNase E, bamD and ftsH [35][36][37][38][39]. Utilizing a BoE screen, where we transformed an immRA Smu deletion construct into the Tn-seq library, we were able to isolate eight double mutants capable of tolerating immRA Smu deletion (Fig. 3a). These mutants were then subjected to complete genome sequencing, to identify the location of the transposon insertion and identify suppressor mutations in the event these were also generated. Genome sequencing revealed the transposon insertions to be within the following regions: SMU_197 c, SMU_201 c (isolated twice), SMU_141 c, recN, perR, the murE and SMU_1678 intergenic region, and the intergenic region between 16S rDNA and SMU_1750 c (Table S4). The growth phenotypes of these strains were variable (Fig. 3b). The intergenic transposon insertion into murE caused a considerable growth phenotype, as did an insertion in SMU_197 c. Certain double mutant isolates grew almost the same as wild-type S. mutans, including perR::Tn, recN::Tn and SMU_141 c::Tn.
To account for ΔimmRA Smu strains having a wild-type copy of immRA Smu , we used PCR to visualize the size of the immRA Smu region. Gene duplication events occurred in the strains perR::Tn, recN::Tn, and the strain with a transposon insertion between 16S rDNA and SMU_1750 c (Fig. S3, Table S4). Gene duplication is the simultaneous deletion of immRA Smu with a copy of wild-type immRA Smu still present. Double mutants with an immRA Smu gene duplication grew better than strains that fully lacked immRA Smu (Fig. 3b). For the other five mutants, only one band for immRA Smu was amplified and the size was as expected for a deletion mutant. As an additional control, PCR was also conducted to confirm that the Magellan6 transposon was inserted in the regions identified by whole-genome sequencing. Transposon insertions were apparent for all strains except for SMU_141 c::Tn, where the results were inconclusive (Fig. S4, Table S4).

Evidence of a circular intermediate form of TnSmu1
Next, we noticed that in several double mutant backgrounds, read coverage (after genomic sequencing) was significantly elevated for a region between SMU_191 c and SMU_220 c (Figs 4 and S5). Read coverage depth in this region was ~300 for S. mutans UA159, but was elevated to as much as ~5000 reads in double mutant backgrounds. This increased read depth corresponds exactly to a region of the genome that is predicted to contain the putative ICE, TnSmu1. When excised from the host chromosome, ICEs exist as circular DNA molecules [40]. For some ICEs, these circular intermediate forms are capable of rolling-circle replication [41]. Increased read coverage depth suggests that loss of immR Smu is leading to excision and circularization of TnSmu1. In addition to the observation that sequencing reads increase when immR Smu is deleted, there was also evidence of unexpected genomic junctions in strains of S. mutans missing the immR Smu gene. There was evidence from the genome sequencing that attL and attR are combining, which is leading to a junction forming between SMU_191 c and SMU_220 c. This is additional evidence that TnSmu1 is excising from the host chromosome and forming a circular DNA molecule. To confirm these observations, we designed primers to test excision and circularization. For excision, these primers will only amplify a PCR product if TnSmu1 is no longer present on the host chromosome. For circularization, a PCR product will form if there is a junction between the ends of the ICE (i.e. SMU_191 c and SMU_220 c). Excision and circularization of TnSmu1 in wild-type S. mutans, grown to mid-log phase, occurs close to the limit of detection for the assay (Fig. 5a, b). However, excision and circularization of TnSmu1 is apparent in immR Smu double mutant strains. For ΔimmR Smu strains carrying transposon insertions within putative TnSmu1 conjugation genes (SMU_197 c and SMU_201 c), there are multiple copies of TnSmu1 per cell. In these strains, excision is occurring 1000-fold more compared to wild-type S. mutans. We also observed circularization/excision in two strains experiencing duplication of immR Smu , perR::Tn and recN::Tn. We chose to examine these strains because they are much healthier than the strains with disrupted conjugation genes (Fig. 3b). For these two strains, circularization of TnSmu1 was occurring, but excision of the element occurred at levels close to the limit of detection for the assay. Taken together, our observations of excision and circularization are consistent with TnSmu1 having this important life-cycle feature of ICEs.

Induction of TnSmu1 by the genotoxic agent mitomycin C (MMC)
The immRA Smu module shares similarities to cI-like phage repressors that are sensitive to DNA damage (via induction of the SOS response) [42]. Other ICEs contain similar regulatory modules, ICEBs1 being the most extensively characterized, but also ICESt3 and SXT [10,43,44]. To determine whether DNA damage was an inducer of TnSmu1, possibly via cleavage of ImmR Smu , we incubated S. mutans UA159 in the presence of genotoxic agent MMC. When treated with MMC, there was a clear dosedependent response, with clear growth hindrance at 25 ng ml −1 (Fig. 5c). Induction of TnSmu1 circularization was investigated at MMC concentrations of 3.125, 6.25 and 12.5 ng ml −1 . With increasing concentrations of MMC, the relative abundance of circular TnSmu1 increased in a dose-dependent manner (Fig. 5d). At a concentration of 12.5 ng ml −1 not all cells have a circular copy of TnSmu1 but circularization is 1000-fold above levels seen in untreated S. mutans cells.

Transcriptome analysis in S. mutans strains lacking immR Smu
In order to provide further evidence that immR Smu represses activation of TnSmu1, we examined the transcriptomes of two strains generated during the BoE screens. RNA was extracted from SMU_197 c::Tn ΔimmRA Smu and SMU_201 c::Tn ΔimmRA Smu cultures grown to OD 600 ~0.4, and this RNA was then processed for RNA-seq. For the SMU_197 c::Tn ΔimmRA Smu strain, RNA-seq analysis identified 54 genes whose log 2 fold change was ≥2 (FDR<0.05; Table S5). Of these genes, 7 were repressed and 47 were induced (Fig. 6a). Similar results were obtained for the SMU_201 c::Tn ΔimmRA Smu strain, with RNA-seq analysis identifying 51 genes whose log 2 fold change was ≥2 (FDR<0.05; Table S6); 7 genes were repressed and 44 were induced (Fig. 6a). As with CRISPRi repression of immR Smu , TnSmu1 expression is significantly induced in both ΔimmRA Smu strains. In both strains, SMU_205 c (hypothetical protein residing in TnSmu1) was the most strongly up-regulated gene by 968-fold (SMU_201 c::Tn ΔimmRA Smu ) and 399-fold (SMU_197 c::Tn ΔimmRA Smu ). Next, we generated a heat map of selected genes with Fig. 5. TnSmu1 is capable of circularization and excision. qPCR was employed to measure the relative abundance of circular (a) and excised (b) copies of TnSmu1. The number of copies was compared to a nearby single-copy chromosomal gene, sloR. Induction of TnSmu1 circularization was also measured by inducing TnSmu1 with the genotoxic agent MMC in wild-type S. mutans UA159. The impact of different concentrations of MMC was quantified using a growth assay (c), and then different doses of MMC were used to induce TnSmu1 circularization (d).
the aim of comparing gene expression between ΔimmRA Smu strains and the sgRNA-218 strain (Fig. 6b). Notably, up-regulation of TnSmu1, the type I-C CRISPR-Cas system and the CslAB transporter system is consistent among all three strains. In addition, SMU_40 (hypothetical protein with a RelE toxin-antitoxin system domain) is moderately up-regulated (~5 fold) across the three strains. For the malolactic fermentation (mle) locus (SMU_137 to SMU_141) [45], down-regulation (~10 fold) was observed in the ΔimmRA Smu strains, and up-regulation (~3.5 fold) in the sgRNA-218 strain. The mle locus contains a malolactic enzyme (mleS; SMU_137), a malate permease (mleP; SMU_138), an oxalate decarboxylase (oxdC; SMU_139), a glutathione reductase (gshR; SMU_140) and a hypothetical protein (SMU_141); the role of this operon is in malolactic fermentation (conversion of l-malate to l-lactate). Taken together, this additional RNA-seq analysis shows that loss of ΔimmRA Smu is leading to activation of TnSmu1. In addition, there is a consistent trend showing activation of the type I-C CRISPR-Cas system and the cslAB transporter system.

Distribution and conjugative transfer of TnSmu1 genes between S. mutans strains
To begin to understand whether TnSmu1 is capable of conjugative transfer, we wanted to determine whether closely and distantly related strains of S. mutans carry this ICE. Notably, we have recently discovered that S. mutans strains carry CRISPR spacers against this element [46]. This finding suggests that S. mutans strains encounter TnSmu1, or genes from related ICEs, and CRISPR-Cas is recording these horizontal transfer events. Fig. 7 shows strains that carry TnSmu1 genes and is organized by strain relatedness. Distantly related strains, such as smu342 and UA159, carry full length TnSmu1 elements, which might be indicative of conjugal transfer. Although many S. mutans strains carry TnSmu1 genes, only nine strains carry the full element as organized on S. mutans UA159. ICEs have plasticity and lose/gain genes, often interacting with other mobile elements such as transposons, lysogenic phages and genomic islands. It is, therefore, not surprising that the structure of TnSmu1 changes as it transfers within the S. mutans species, particularly if its expression is lethal in other S. mutans hosts. Under normal conditions, we anticipate that a very small sub-fraction of S. mutans cells would express TnSmu1 because of ImmR Smu -mediated repression of the element. Therefore, it is likely that transfer of TnSmu1 from a donor strain to a recipient strain occurs at very low efficiency. Similar ICEs in other 'Firmicutes' bacteria, like ICEBs1 and Tn916, exhibit considerable differences in conjugation rates. In the future, we plan to explore conjugative transfer of this element between strains in greater detail.

DISCUSSION
In this study, we sought to determine the role of SMU_218/immR Smu , a gene we previously identified as being essential or nonmutable by Tn-seq. We hypothesized based on genomic context and protein domain similarities that immR Smu was either an antitoxin gene in a toxin-antitoxin system, or a transcriptional regulator involved in activation of the ICE TnSmu1. We provide evidence that immR Smu is not an antitoxin. Instead, we have concluded that immR Smu is a repressor that keeps TnSmu1 silent, except under certain environmental conditions. In exploring immR Smu , we provide evidence that TnSmu1 can be induced, that it produces a circular intermediate, and the element, or parts of it, are distributed among S. mutans strains. We also uncovered that activation of TnSmu1 has a considerable impact on S. mutans physiology, leading to changes in gene expression, slowed growth and abnormal cell morphology. An independent study by McLellan et al. [47], which was completed at the same time as our study, has also described features of TnSmu1, including transfer into recipient strains and growth arrest caused by activation of the element. Our conclusion that immR Smu is a repressor involved in the activation of TnSmu1 is based on several lines of evidence. Firstly, blast and other related bioinformatics searches show that immR Smu is a predicted helix-turn-helix DNA binding protein, and the downstream gene immA Smu is a predicted Zn 2+ metalloprotease. blast searches also reveal that immRA Smu shares similarities with other described transcriptional regulator and metalloprotease regulatory modules. Most importantly, the mobile element ICEBs1 contains a repressor, ImmR, that is cleaved by a protease ImmA, in a two gene organization like immRA Smu but in a different location within the ICE (Fig. 8a) [11,44]. Due to the similarities of the system described here, and ICEBs1 immRA, we have named SMU_218 as immR Smu and SMU_219 as immA Smu . Activation of ICEBs1 occurs when ImmR is cleaved by the protease ImmA, either through a cell-cell signalling pathway involving RapI and PhrI or via a RecA-dependent DNA damage response [11,44]. When the signalling peptide PhrI is abundant it binds to RapI, which stops RapI from being able to inactivate ImmRmediated repression of ICEBs1 [44]. TnSmu1 does not appear to contain a RapI/PhrI cell-cell signalling system but additional regulatory systems on top of immRA Smu cannot be ruled out, as TnSmu1 contains several hypothetical genes with unknown functions. ImmRA-like systems are encoded on other putative and studied mobile genetic elements. ICESt1 and ICESt3 contain a regulation module with a ImmR-like gene (arp2), a metalloprotease gene (orfQ) and a cI gene (arp1) [14]. In this module, the metalloprotease gene, orfQ, is not downstream of the immR-like gene, arp2, but as with ICEBs1 these elements are activated by DNA damage, which is probably regulated via repressor proteolysis [14]. An ImmRA-like system is also required for the activation of a staphylococcal pathogenicity island SaPI3 [48]. For this regulation module, ImmR repression is partially alleviated when Sis (SaPI inducer of SaPIs), a protein produced by other SaPIs, binds to ImmR but full activation only occurs with ImmA-mediated proteolytic cleavage of ImmR [48]. ImmRA modules are also encoded by lysogenic phages known to be activated by DNA damage [11]. These modules are, therefore, a common regulatory system governing activation of mobile genetic elements, and several have additional regulatory complexities in addition to ImmRA.
Aside from inducing excision of ICE elements, there is evidence that similar transcriptional regulator and metalloprotease modules are encoded in other systems. Recently, a system that parallels ImmRA was identified on CBASS (cyclic oligonucleotide based anti-phage signalling system) anti-phage systems [49]. Called CapH/CapP, this module regulates expression of a CBASS in response to DNA damage. The transcriptional regulator CapH represses CBASS transcription until it is cleaved by the metallopeptidase CapP, which is stimulated by the presence of ssDNA. Once the CBASS is derepressed, a cell killing pathway is induced Both ICEs integrate at tRNA-leu. Note that the immRA Smu regulatory module is in a different location compared with ICEBs1. (b) A diagram depicting the ImmRA Smu regulatory pathway for TnSmu1 activation is shown. Our model draws heavily from ICEBs1 ImmRA, and closely related systems, where DNA damage leads to activation of the protease ImmA. Activated ImmA cleaves the ImmR repressor leading to activation of TnSmu1. It is likely that stochasticity causes sub-population responses so that TnSmu1 activation can lead to transfer of the element, cell arrest/cell death and other phenotypes. The SOS pathway, leading to RecA-mediated activation of ImmA Smu , is not well known in S. mutans and will need to be investigated further. Yet to be discovered signals may also lead to ImmA Smu -mediated cleavage of ImmR Smu . that kills the host bacteria. Another ImmRA-like system, known as DdrO/IrrE, is encoded by Deinococcus spp. [50]. This system regulates a DNA damage response, with the repressor DdrO being cleaved by IrrE during radiation, leading to expression of DNA repair genes and an apoptotic-like cell death pathway [50]. There is a common theme among the ImmRA, CapHP and DdrO/ IrrE modules in that they cause expression of genes upon sensing of DNA damage, with cleavage of a transcriptional repressor by an activated metalloprotease. In addition, both CapHP and DdrO/IrrE lead to host genome killing. That these systems have pathways that lead to cell killing is notable because activation of TnSmu1 (via CRISPRi) caused growth defects and abnormal cell morphology. However, additional studies are warranted to determine whether this is mediated directly or indirectly by TnSmu1.
Additional evidence that immRA Smu is a regulatory module was provided by RNA-seq, which helped to identify genes activated in response to either depletion or deletion of immR Smu . When immR Smu is repressed or deleted, there is a clear and strong up-regulation of TnSmu1-associated genes by as much as 1000-fold. Although derepression of TnSmu1 genes by ImmR Smu would be most strongly confirmed by DNA-binding assays (e.g. electrophoretic mobility shift assay), activation of TnSmu1 genes upon loss of the ImmR repressor is consistent with similar systems described in ICEBs1, CapH/CapP and DdrO/IrrE. In addition, deletion of immR Smu was found to activate excision/circularization of TnSmu1. Again, this would be consistent with a hypothesis that loss of the repressor leads to constitutive derepression of TnSmu1, followed by activation of the mobile element. Based on ICEBs1 ImmRA (and similar systems), we predict that ImmA Smu is a metalloprotease that is activated by DNA damage, leading to cleavage of ImmR Smu and derepession of TnSmu1. Although we do not provide direct evidence confirming this hypothesis, we do show that TnSmu1 circularization is induced in response to DNA damage by the genotoxic agent MMC. Taken together, activation by TnSmu1 as shown in RNA-seq studies and in response to MMC provides reliable evidence that ImmRA Smu functions as a regulatory module that has similarities to ICEBs1 ImmRA.
We were initially drawn to studying immR Smu because of Tn-seq data showing that the gene was essential. With the evidence gathered here, we are able to conclude that immR Smu is not a classical essential gene and does not participate in a core, required for survival, biological pathway. Instead, immR Smu appears essential because deletion of immR Smu activates TnSmu1 leading to excision and eventual loss of the mobile element from the cell; thereby making it not possible to select a deletion strain. The observation that inactivation of nearby conjugation-related genes (SMU_197 c and SMU_201 c) allows immR Smu deletion provides evidence for this hypothesis. Here, disruption of the conjugation genes likely leads to TnSmu1/ΔimmR Smu becoming 'trapped' in the cell, and selectable through antibiotic-cassette replacement. However, we were able to describe significant impacts on cell growth, cell morphology and the transcriptome when immR Smu was depleted with CRISPRi. Despite in many cases ICEs carrying beneficial genes, such as virulence factors or antibiotic-resistance genes, expression of these elements has been found to have a major impact on certain hosts. Induction of ICEclc in Pseudomonas spp. causes a sub-population of cells to differentiate into 'transfer competent' cells that have arrested cell growth and cell lysis [51]. Activation of Tn916 causes severe cell growth defects in Enterococcus faecalis and Bacillus subtilis [52]. Tn916 only activates in a small percentage (0.1-3 %) of cells and the lethality of Tn916 was discovered by activating it in a much larger proportion of cells. The deleterious effects of Tn916 were less impactful in cells lacking conjugation genes that reside in Tn916, and a gene, yqaR, found within a defective phage-like element skin [52]. Intriguingly, Tn916 can also cause cell death in Enterococcus faecalis and B. subtilis, species that do not encode a yqaR-like gene. Future studies of the growth arrest/cell death mechanism in TnSmu1 would be helpful in understanding this generally conserved process for ICE transfer, and whether it is beneficial or costly for transfer efficiency.
We expect that under most conditions TnSmu1 will be quiescent because of ImmR Smu -mediated repression of the mobile genetic element. This silence helps maintain vertical transmission of the element during non-stressful conditions. We anticipate that activation of TnSmu1 occurs when ImmA Smu becomes active and cleaves ImmR Smu leading to de-repression of TnSmu1 (Fig. 8b). As with other ICEs, such as ICEBs1, DNA damage causes activation of TnSmu1, as shown here with MMC treatment. For ICEBs1, during DNA damage, ImmA is stimulated via an SOS (save our souls) response that causes RecA to activate ImmA cleavage of ImmR [11,44]. Although our model of ImmRA Smu -mediated regulation of TnSmu1 includes activation of ImmA by RecA, there may be substantial differences in the SOS response between S. mutans and B. subtilis. Streptococcus spp. have generally been thought to lack a classical SOS response because of work performed in Streptococcus pneumoniae, which lacks a LexA homologue, where genetic competence plays a central role in responding to DNA damage [53]. However, other streptococci such as Streptococcus thermophilus encode LexA-like repressors and competence development is antagonistic to the SOS response [54]. The S. mutans UA159 genome encodes two LexA-like repressors, SMU_1398 (irvR) and SMU_2027 (hdiR). IrvR does not participate in the SOS-response and instead is a stress-responsive biofilm regulator [55]. In comparison, hdiR is induced by SOS stress (MMC) in a recA-dependent manner [55]. Genetic competence could play a role in an SOS-like response because recA is up-regulated as part of the late competence regulon [56]. Notably, other pathways may be involved in TnSmu1 regulation as RNA-seq studies show TnSmu1 up-or down-regulation in different conditions or mutant backgrounds [57][58][59][60][61]. Although our core model of TnSmu1 activation via ImmR Smu proteolysis borrows heavily from the ICEBs1 system, we anticipate notable differences in how ImmA Smu becomes active.
To conclude, we show that S. mutans contains an ICE, TnSmu1, whose activation is likely controlled by an ImmRA-like regulatory module. Supporting this hypothesis is strong up-regulation of TnSmu1 genes and excision/circularization of the element upon loss or depletion of immR Smu . The mobile element is also inducible via DNA damage, which is a conserved feature of ImmRA-like regulatory systems. Future studies will be directed to better understanding the steps involved in TnSmu1 activation, as well as identifying the mechanism that causes activation to be deleterious to host fitness and cell biology.

Funding information
This work was supported by the National Institute of Dental and Craniofacial Research (NIDCR) grant DE029882 awarded to R.C.S.