Precision Genome Engineering in Streptococcus suis Based on a Broad-Host-Range Vector and CRISPR-Cas9 Technology

Streptococcussuis is an important zoonotic pathogen that causes severe invasive disease in pigs and humans. Current methods for genome engineering of S. suis rely on the insertion of antibiotic resistance markers, which is time-consuming and labor-intensive and does not allow the precise introduction of small genomic mutations. Here we developed a system for CRISPR-based genome editing in S. suis, utilizing linear DNA fragments for homologous recombination (HR) and a plasmid-based negative selection system for bacteria not edited by HR. To enable the use of this system in other bacteria, we engineered a broad-host-range replicon in the CRISPR plasmid. We demonstrated the utility of this system to rapidly introduce multiple gene deletions in successive rounds of genome editing and to make precise nucleotide changes in essential genes. Furthermore, we characterized a mechanism by which S. suis can escape killing by a targeted Cas9-sgRNA complex in the absence of HR. A characteristic of this new mechanism is the presence of very slow-growing colonies in a persister-like state that may allow for DNA repair or the introduction of mutations, alleviating Cas9 pressure. This does not impact the utility of CRISPR-based genome editing because the escape colonies are easily distinguished from genetically edited clones due to their small colony size. Our CRISPR-based editing system is a valuable addition to the genetic toolbox for engineering of S. suis, as it accelerates the process of mutant construction and simplifies the removal of antibiotic markers between successive rounds of genome editing.


■ INTRODUCTION
Streptococcus suis is one of the major porcine bacterial pathogens and a common colonizer of the porcine upper respiratory tract, the gastrointestinal tract, and the genital tract. 1 Although carriage of S. suis is endemic in farmed pigs, it mainly causes an invasive disease in young piglets around weaning. S. suis infections may lead to meningitis, arthritis, and septicaemia. 2−4 S. suis is also a zoonotic pathogen causing sepsis, meningitis (with deafness as a consequence), and endocarditis in humans. 5−8 Human-to-human transmission of S. suis has not been reported.Within the species S. suis, different serotypes and pathotypes have been identified, ranging from highly virulent to nonvirulent strains. 4Serotype 2 has frequently been reported to contain virulent strains, and most zoonotic strains belong to this serotype. 9o investigate S. suis virulence mechanisms and enhance our understanding of its pathobiology via mutant analysis, it is crucial that its genome can be efficiently edited.A recent advance in the genetic engineering of S. suis was the discovery of a peptide pheromone-inducible mechanism of extracellular DNA uptake (competence) and DNA integration into the genome. 10,11This method enables the high-efficiency genetic transformation of many S. suis strains.However, the obligatory use of antibiotic selection markers limits the number of possible genomic edits, and the insertion of a selection marker gene and promoter may cause polar effects on neighboring genes, requiring complementation approaches to confirm that the genetic deletion is responsible for an observed phenotype.While combined selection-counterselection methods have been used for marker-less mutant generation, these systems usually require time-consuming procedures involving multiple steps to remove the selection marker. 12Moreover, certain mutations, such as single nucleotide substitutions, cannot be constructed using these methods.
A recent development in the field of genome engineering was the advent of CRISPR-Cas9 technology, which allows targeted genome modification with unprecedented accuracy.CRISPR-Cas is a family of adaptive immune systems that is present in many prokaryotes. 13Naturally, this system acts as a defense mechanism against specific invading DNA elements such as plasmids, mediated via small RNAs�encoded in the CRISPR locus�that match corresponding sequences in the invading DNA. 14In type II CRISPR systems, the Cas9 enzyme cleaves invading DNA in a sequence-specific manner. 14,15Cas9 is guided to its target DNA sequence by a dual RNA complex consisting of a CRISPR RNA (crRNA) that contains a targetspecific spacer sequence and a trans-activating crRNA (tracrRNA) that plays a role in the maturation of crRNA.A protospacer adjacent motif (PAM) of 2−6 nt must immediately follow the target sequence to enable binding and cleavage by Cas9.Thus, the PAM sequence is an essential targeting motif distinguishing "nonself" (a sequence in the invading DNA) from "self" (the same sequence in the genomic CRISPR spacer lacking a PAM).To exploit this system for genome editing, the crRNA-tracrRNA complexes can be engineered into a single fusion guide RNA via a flexible linker, resulting in a chimeric single guide RNA (sgRNA).Since most prokaryotes lack a nonhomologous end-joining mechanism for DNA repair, they are unable to repair their chromosomal DNA after cleavage by Cas9, leading to cell death. 16,17This can be utilized for genome editing by providing a repair template (RT), which is a homologous DNA fragment containing the desired modifications but lacking the sequence that is targeted by Cas9.By incorporation of the modified DNA fragment into the genome via homologous recombination (HR), the bacteria can evade Cas9-mediated DNA cleavage and subsequent cell death.
Here, we describe the construction of a broad-host-range shuttle vector (pSStarget) expressing Streptococcus pyogenes Cas9 and demonstrate how this vector can be used for efficient genome editing in S. suis.The pSStarget vector facilitates highly efficient sgRNA cloning via replacement of a negative selection marker, and is easily lost from S. suis cells when cultured in the absence of antibiotic pressure.We show that pSStarget can be used for multiple successive rounds of highprecision Cas9-mediated genome editing in S. suis without genomic integration of selective markers.Moreover, we demonstrate the application of our pSStarget-based CRISPR-Cas system to construct a mutant with a single amino acid substitution in the essential eno gene.Finally, we describe small colony variants (SCVs) that are seemingly resistant to Cas9mediated genome editing in the absence of a RT.We propose a potential mechanism for the recovery of SCVs based on RNaseq and subsequent growth studies.

■ RESULTS
Construction of pSStarget.We used the previously published pLABtargetc vector, which was originally developed for CRISPR-targeting in Lactococcus lactis, 18 as a starting point for the construction of pSStarget.The broad-host-range pIL253 origin of replication (ori) allows its use in a wide variety of Gram-positive organisms, including lactobacilli, bacilli, and Streptococcus species. 19−22 pLABtargetc contains a chloramphenicol acetyltransferase (cat) gene conferring chloramphenicol resistance, wild-type (wt) CRISPR-associated endonuclease (cas9) from S. pyogenes under the control of its native promoter, and an sgRNA handle (the portion that binds Cas9) with BsaI restriction sites that can be used for cloning of a specific spacer sequence.Expression of sgRNA is driven by the lactococcal eps promoter.
To generate a shuttle vector for cloning in Escherichia coli, we cloned the Gram-negative p15A ori and a tetracycline resistance marker (tetA) into pLABtargetc, resulting in the plasmid pSStarget-NT.The p15A ori has a low copy number in E. coli, minimizing the toxic effects of Cas9 expression during cloning.To facilitate efficient cloning of the spacer sequence, we inserted the negative selection marker ccdB (encoding a toxin protein which selectively inhibits DNA gyrase in the E. colicloning host) between two BsaI restriction sites to generate the plasmid pSStarget (Figure 1A).Target-specific spacer sequences were generated by annealing partially complemen-tary oligonucleotides, resulting in 4 nt overhangs that anneal to the overhangs generated by BsaI digestion of pSStarget (Figure 1C).Insertion of the spacer sequence eliminates the BsaI sites and removes the ccdB gene (Figure 1B,D).The CcdB-resistant E. colistrain ccdB Survival 2 was used for propagation of the pSStarget plasmid.
Cas9-Mediated Genome Editing and "CRISPR Escape" Mutants.The use of the CRISPR-Cas9 system for genome engineering relies on HR carried out by S. suis using a linear RT, followed by Cas9-mediated counter-selection (Figure 2).RTs contained the modified target region (e.g., deletion or single nucleotide changes) flanked by approximately 1000 bp of homologous DNA on each side to allow for efficient HR in the corresponding S. suis genome.Transformation of bacteria with the RT and a target-specific pSStarget vector will lead to HR in a subset of cells.This results in the genomic integration of the RT and the simultaneous loss of the WT sequence containing the target site for the sgRNA.In cells that do not incorporate the RT into the genome, the Cas9/sgRNA complex cleaves DNA sequences matching the target-specific sgRNA, leading to double-stranded breaks in the genome.Therefore, Cas9 can be utilized as an efficient negative Meanwhile, Cas9 and sgRNA are expressed and assembled to form a complex.(C) Mature Cas9/sgRNA complex scans the genome for target sites.Edited cells that incorporated the RT by recombination lack the target site, making them immune to Cas9-mediated DNA cleavage.In the absence of HR, the target site is cleaved by Cas9, leading to cell death.counterselection tool, as double-stranded breaks in genomic DNA are lethal to most bacteria.
To demonstrate the use of this system for genome engineering in S. suis, we chose three previously characterized virulence factors with known deletion phenotypes: (1) the cpsE and cpsF genes that are part of the capsular polysaccharide locus; (2) the gene encoding the hemolytic toxin Suilysin (sly); and (3) the lgt gene, which encodes the Lgt transferase involved in anchoring lipoproteins to the cell membrane.
To our surprise, we observed two different colony morphologies upon transformation of S. suis P1/7 with pSStarget plasmids containing a target sgRNA and corresponding RT, but not with a nontargeting plasmid lacking a sgRNA spacer (pSStarget-NT).Transformation of S. suis with pSStarget-NT yielded approximately 3.2 × 10 4 normal-sized colonies.Co-transformation of S. suis with pSStarget containing target-specific sgRNAs and their corresponding RTs yielded a mixture of normal-sized colonies and small colonies that were barely visible to the naked eye (Figure 3A).
When S. suis was transformed with only the pSStarget-sgRNA plasmids, predominantly small colony variants (SCVs) were recovered (Figure 3B).The number of small colonies recovered was highly dependent on the sgRNA sequence, with sg6 yielding many more small colonies than sg4 and sg42, giving rise to an intermediate amount of small colonies (data not shown).Moreover, small colony counts were lower when a corresponding RT was cotransformed than when only pSStarget was transformed.
More than 80% of the normal-sized colonies contained the desired deletion, confirmed by colony PCR.It was not always possible to generate an amplicon from small colonies, presumably due to the low number of bacteria.When a PCR amplicon was obtained from SCVs transformed with pSStarget-sg4, the amplicon size was approximately 3.5 kb, corresponding to the length expected for the WT sly sequence.As no mutations were identified in the target site amplicons, we hypothesized the SCVs were able to resist Cas9-mediated killing by an unknown mechanism.Although large numbers of the small "CRISPR-tolerant" SCVs can be present after transformation, it does not pose a problem for mutant selection, as SCVs are easily distinguished from the large colonies of bacteria with successfully edited genomes.

Rapid Loss of pSStarget under Nonselective Growth Conditions Enables Multiple Rounds of Genome
Editing.We then sought to investigate the possibility of generating double-knockout mutants by conducting successive rounds of genome editing.To allow the transformation of a new pSStarget plasmid targeting a different gene, the obtained mutants first had to lose the original pSStarget plasmid to restore their sensitivity to chloramphenicol.To achieve this, we grew the obtained mutants overnight in a large volume (30 mL) of THY in the absence of antibiotics.The next day an aliquot of the culture was spread on selective agar plates to verify the loss of the plasmid.In most cases (>90%), all cells had lost the plasmid after one night of culture.In the remaining cases, the plasmid was lost after a second night of culture.The Δcps mutant was used as a parental strain for the construction of two double-knockout strains, resulting in strains Δcps Δsly and Δcps Δlgt.
To demonstrate the absence of off-target effects during CRISPR-Cas9-mediated genome engineering, we sequenced the whole genome of the CRISPR-edited strains Δcps, Δsly, and Δcps Δsly, as well as the genome of the parental strain P1/  7. We identified a deletion of 5 bp in an intergenic region (660,128−660,132, GCAGA) and two SNPs in our laboratory stock of P1/7 relative to the published P1/7 reference sequence (AM946016.1).One mutation (1151748A > T) is located in an intergenic region, and the second (1406781G > T) results in an amino acid change (G47 V) in gene SSU1381.When comparing the mutant strains with our new P1/7 sequence, the exact deletion of the target regions was confirmed, and no additional mutations or structural rearrangements were found, demonstrating the high fidelity and accuracy of our CRISPR-Cas9 genome editing procedure.(Figure 4) Phenotyping of Selected Mutants.We conducted multiple assays to confirm that the constructed mutants had the expected phenotypes (Figure 5).For the Δsly mutant, we assayed hemolytic activity after growth in pullulan, which was previously shown to induce sly expression. 23Hemolysis was absent in the Δsly strain, while the WT strain resulted in complete hemolysis across a wide range of dilutions (Figure 5A,B).Transmission electron microscopy was used to visualize the morphological effects of the cps and lgt deletions on the S. suis capsule (Figure 5C,D).The polysaccharide capsule was clearly visible in wild-type P1/7 and in the Δlgt mutant, whereas both Δcps mutant strains lack the polysaccharide capsule.There was no clear morphological difference between the Δlgt mutant strains and the corresponding parental strain expressing the lgt gene (data not shown).
In addition, we tested the activation of human TLR2/6 in transfected HEK-293 reporter cell lines by Δcps, Δlgt, and Δcps Δlgt strains (Figure 5E,F).These reporter cell lines were transfected with the pNiFty2 reporter plasmid, which expresses luciferase upon NF-κB signaling.As activation of TLRs leads to NF-κB signaling, the activation can be quantified by measuring the luminescent activity.A HEK-293 reporter cell line transfected with the pNiFty2 reporter plasmid but lacking TLRs was used as a control.The different HEK-293 cell lines were stimulated with each bacterial strain in penicillincontaining medium to release bacterial cell wall and membrane components. 24n the cell line expressing hTLR2/6, we observed a statistically significant difference (adjusted p-value = 0.0305) between luminescence upon stimulation with the Δcps and Δlgt strains (Figure 5E).Although the differences between the other samples did not reach statistical significance, a general trend could be observed.The deletion of lgt resulted in reduced TLR2/6 signaling, whereas the deletion of cpsEF increased TLR2/6 activation.When the reporter cells were stimulated with the Δcps Δlgt double mutant strain, we measured similar values as for the wildtype, suggesting that the opposing effects observed for the single gene mutants cancel each other out.No increased luminescence was measured in control HEK-293 cells that did not express TLRs for any tested condition, except for the positive control TNFα, which activates the NF-κB pathway independently of TLR signaling (Figure 5F).The results of the TLR activation assay are presented as luminescent activity relative to the values measured for the unstimulated baseline controls of the same cell line.
Our findings correspond to earlier research on the function of these genes.The main ligand leading to activation of TLR2/ selective media, individual large colonies appeared among the SCVs, suggesting the occurrence of a mutation in the target site.
To analyze the growth phenotype of the SCVs, we measured growth in selective THY medium inoculated with SCVs recovered after transformation with pSStarget-sg4 (targeting sly) or the NT control.Colonies of bacteria transformed with NT control displayed normal growth, while cultures inoculated with small colonies did not grow after 24 h (Figure 7A).However, after incubation for 36−48 h, growth was observed in three of the eight wells (i.e., D6, F5, and F6) inoculated with SCVs (Figure 7A).Plating the cultures where growth was observed on selective agar media produced exclusively large colonies, while plating the medium from nongrowing colonies (e.g., E6) gave a small colony phenotype on agar plates.Sanger sequencing of the sly gene in D6, F5, and F6 revealed "CRISPR-escape" mutations in the target sites or adjacent PAM sequences, which would prevent Cas9-mediated DNA cleavage (Figure 7B).Bacteria in F6 acquired four mutations in the protospacer of the target site (983C > T, 987T > G, 989T > G, and 901G > T), which allowed normal growth under antibiotic selection for pSStarget-sg4.Bacteria in well D6 acquired a single-nucleotide substitution in the PAM site, changing the PAM sequence from CGG to CGA.Interestingly, well D6 had a lower growth rate during the exponential phase and reached a lower maximal density (Figure 7A, green curve).These results suggest that the SCVs might resemble persister cells, as the growth of the respective cells is temporarily arrested in response to stress.
To gain further insights into the mechanism of CRISPRtolerance in SCVs, we conducted a transcriptomic analysis comparing gene expression in SCVs recovered after transformation with pSStarget-sg6 (SCV) to the gene expression in bacteria recovered after transformation with pSStarget-NT (NT), lacking a targeting spacer as a control.At the time of sample collection, we plated an aliquot of each culture to Figure 8. Mapping of RNaseq reads to the region of pSStarget containing the cat and cas9 genes.The top panel shows the location of the cat and cas9 genes.The two lower panels show the mapping of RNaseq reads to these two genes in the targeting (SCV) and nontargeting (NT) conditions, respectively.As in the case of the cat gene, when too many reads were present to plot each read individually, the reads were visualized as a pileup plot.Both genes were strongly downregulated in the SCV condition (upper panel) compared to the NT condition (lower panel) ensure that the small colony phenotype had been maintained.Mapping of the RNaseq reads to the plasmid sequence verified the expression of all critical plasmid components in the NT group, including the sgRNA expression cassette (Figure S4).We found that all plasmid-encoded genes were strongly downregulated (50-fold to 113-fold reduction) in the SCVs compared to the NT control, suggesting a reduction in the plasmid copy number.The most downregulated gene was the chloramphenicol (Cml) resistance marker (cat) (113-fold reduction), suggesting growth inhibition due to the bacterio-static effect of Cml.Moreover, the cas9 gene was downregulated 51-fold, which may not be sufficient to induce DNA cleavage (Figure 8).This observation led us to hypothesize that SCV formation might depend on maintaining a balance between sufficient cat expression to maintain Cml resistance and a low level of cas9 expression.Indeed, plating the transformed bacteria on agar containing an increased Cml concentration of 7.5 μg/mL resulted in a clear reduction of SCVs relative to the usual Cml concentration of 5 μg/mL (Figure S5).Other downregulated genes included growth-related genes such as the fab and acc operons (SSU1597−1609) which are involved in lipid biosynthesis, genes encoding enzymes from the tricarboxylic acid cycle (SSU1040−SSU1042), and genes involved in glycolysis (Table 1).The most upregulated genes in the SCVs included genes involved in pyrimidine metabolism (SSU0735−SSU0738 and SSU0860−SSU0868) and genes involved in pentose and glucuronate interconversion (SSU0998-SSU1008).Additionally, multiple peptidoglycanbinding genes were upregulated, including two genes encoding murein hydrolases (SSU1854 and SSU1911) that are involved in competence-induced cell death (fratricide) in other streptococci.Four of the six chromosomally encoded toxinantitoxin systems were upregulated in the SCVs (SSU0571/ 572, SSU1318/1319, SSU1795/1796, and SSU1798/1799).Furthermore, two genes involved in cell division (SSU0106 and SSU0859) and genes involved in DNA repair, including the mismatch repair proteins MutT and MutX, the uracil-DNA-glycosylase Ung (involved in base excision repair), and the recombinase RecF, had higher expression levels in the SCVs than in the nontargeting control.Also, the smpB gene (SSU1218), which is involved in rescue of stalled ribosomes, was upregulated.Finally, transport proteins, including a uracil permease (SSU1729), multiple putative drug export pumps (SSU0573, SSU0744, SSU0745, and SSU1242), three MFS transporters (SSU1181, SSU1555, and SSU1912), and a variety of ABC-transporters and PTS systems, were upregulated in the SCVs.The complete list of all up-and downregulated genes can be found in the Supporting Information Table S3.

■ DISCUSSION
In recent years, several tools have been developed for the genetic engineering of S. suis, including the discovery of the competence-inducing peptide XIP and the development of combined selection-counterselection cassettes for marker removal. 10,12Recently, we have shown that strains of S. suis grown in active porcine or human serum become competent for DNA transformation. 28Although this method is less efficient than peptide-induced natural competence, it is simple compared to other methods and works with strains that contain different competence peptide alleles or which are not competent via the peptide-induced pathway. 28o enrich the toolbox for genetic manipulation of S. suis and enable efficient "markerless" genome editing, we developed a plasmid-based CRISPR-Cas9 system.The method involves transformation with the plasmid pSStarget that expresses Cas9 and a sgRNA in combination with a homologous RT.Our CRISPR-Cas9 system cleaves the targeted WT sequence and efficiently counter-selects against nonedited cells that contain the targeted WT sequence in the genome.The high efficiency of this system drastically reduces the number of colonies to be screened by colony PCR.
−31 Due to the broad host range of this vector, the system can likely be used in other organisms, pending minor adaptations where necessary.In S. suis, we showed that pIL253 is segregationally unstable in the absence of antibiotic selection and is easily cured by subculture in nonselective medium.A ccdB gene encoding the gyrase inhibitor CcdB, a potent toxic protein flanked by BsaI recognition sites, was introduced in the sgRNA cloning site.The ccdB gene counter-selects against maintenance of plasmids in common laboratory strains of E. coli,which are sensitive to killing by CcdB.Digestion of pSStarget with BsaI and ligation of the annealed spacer oligonucleotides into the plasmid replaces the ccdB gene.Transformation of the original (ccdBcontaining) plasmid leads to production of the CcdB toxin, resulting in the death of such cells and thereby ensuring that all recovered colonies contain the desired sgRNA insert.pSStarget also contains the p15A replicon, which facilitates a quick and easy workflow for cloning and plasmid preparation in a CcdBresistant strain of E. coli.
The segregationally unstable plasmid allows for multiple sequential rounds of genome editing.Subculture overnight in nonselective medium ensures a rapid loss of the plasmid and associated selection marker, allowing the introduction of a second plasmid on the following day.We also demonstrated the use of this CRISPR system to edit essential genes.The unprecedented accuracy of this genome editing system allows the introduction of single base pair mutations leading to amino acid changes in genes that are essential for growth in laboratory medium.This facilitates the study of essential genes by modification of different parts and domains of the encoded protein.
In accordance with previous reports of CRISPR-mediated genome editing in bacteria, we found a few colonies that survived CRISPR selection without incorporation of the RT.Sequence analysis of such colonies showed random inactivating mutations in the target site or CRISPR system. 18,32We also discovered a different novel mechanism by which S. suis can escape killing by a targeted Cas9-sgRNA complex in the absence of a homologous RT.A visual characteristic of this novel mechanism is the presence of very small colonies (SCVs) after transformation and plating.These small colonies grew extremely slowly in liquid culture, typical of a dormant, persister-like state.However, after 36 to 48 h, some isolates resumed growth due to mutations in the target sequence for the sgRNA or PAM site (Figure 7), suggesting that slow growth may have allowed for DNA repair or the introduction of mutations interfering with lethal Cas9 nuclease activity.When growth resumed, two cultures followed a growth pattern similar to that of the NT control, while a third culture (D6) appeared to have a lower growth rate and final density.Sequencing of target sites revealed a single-nucleotide substitution in the PAM site that had altered the corresponding PAM sequence from CGG to CGA.Previous reports have shown that S. pyogenes (Sp) Cas9 can recognize and cleave targets flanked by a NGA PAM with efficiencies of up to 15% compared to targets flanked by a canonical NGG PAM, 33,34 suggesting that the observed differences in growth phenotype of culture D6 could be a consequence of the residual Cas9activity, leading to the death of a subset of cells.The number of SCVs was highly dependent on the sgRNA sequence.While pSStarget-sg4 resulted in very few small colonies, transformation with pSStarget-sg6 yielded substantially more small colonies.Both sgRNAs had a comparable on-target score (66.5 vs 65.5) but differed substantially in GC content (60 vs 20%).This suggests that the on-target score should not be used as the only metric for sgRNA selection but that additional criteria, such as GC content, should be taken into consideration during sgRNA selection.
RNaseq analysis of SCVs revealed that all pSStarget plasmidencoded genes were strongly downregulated, suggesting that the plasmid copy number had been reduced.This reduction of pSStarget copy number may have reduced Cas9-induced stress and resistance to Chloramphenicol (Cml), thereby affecting growth and colony size.We found a variety of multidrug efflux pumps and other transport proteins up-regulated in the SCVs, which could contribute to lowering intracellular Cml concentrations.A transcriptomic analysis of Bacillus subtilis exposed to subinhibitory concentrations of Cml showed an upregulation of multiple transport proteins, including a multidrug efflux pump and the uracil permease pyrP 35 together with upregulation of the pyrimidine biosynthesis gene cluster, which were upregulated in our experiments as well.
Furthermore, we saw the induction of four toxin-antitoxin (TA) systems, members of the relEB and yoeB/yef M families, that inhibit translation via cleavage of translated mRNAs, leading to nonstop mRNAs and stalling of ribosomes.In line with this, we observed a 3-fold upregulation of the smpB gene, which plays a crucial role in the rescue of stalled ribosomes via trans−translation. 36We also noted upregulation of the gene encoding RNase R that is recruited to the rescue of stalled ribosomes for specific degradation of nonstop mRNA. 37,38−41 Specifically, the RelEtype toxins induce a transient bacteriostatic state through potent inhibition of the translation machinery, and induction of the corresponding antitoxin rapidly reverts the inhibition phenotype. 42In the pathogenic bacteria Edwardsiella piscicida, a homologue of yoeB (SSU1795) has been shown to increase resistance to chloramphenicol and other antibiotics and plays an important role in persister cell formation at lethal chloramphenicol concentrations. 43Induction of chromosomal TA systems could have led to the dormant state observed for SCVs in our experiments.
Translational inhibition and reduced plasmid copy numbers could underpin the mechanisms of CRISPR-escape.Cas9 is a large protein (1371 AA) and its production would likely be more affected by translational inhibition compared to that of smaller proteins.Combined with a 50-fold reduction in mRNA expression, it is likely that very few, if any, functional Cas9 proteins can be produced in this condition, preventing the lethal nuclease activity.
Overall, the findings of our transcriptome analysis of SCVs are consistent with the observed growth inhibition.Downregulation of specific genes involved in lipid biosynthesis, the TCA cycle, and glycolysis corresponds to reduced metabolism and growth rate and underpins a state of dormancy.A state of reduced metabolic activity and growth rate has been linked to increased antibiotic tolerance and persister cell formation in Staphylococcus aureus. 44Genes involved in DNA repair and recombination have been upregulated in and a cold-shock protein (SSU0368), that was shown to be involved in compensating for a mutator phenotype, was downregulated in the SCVs. 45,46In combination with the results from the prolonged growth curve experiment, it is intriguing to suggest that the induced mechanism of dormancy allowed mutations in the targeted DNA site to accumulate, leading to escape from the CRISPR endonuclease activity and reversion to the normal growth phenotype.
The insights into the mechanisms underlying SCV formation were used to improve the efficiency of the editing process.For example, sgRNA targets with a low GC content seem to lead to more SCV formation and should be avoided when possible.To account for the variable sgRNA efficiency, it can be useful to include a "background control" by transforming only the pSStarget plasmid without the corresponding RT.A further recommendation is to maintain chloramphenicol (Cml) selection pressure for a few passages to avoid the growth of nonedited SCVs.This strategy is especially useful in cases where only low-GC-content targets are available for sgRNA design, and thus the formation of many SCVs in addition to larger (normal-sized) colonies can be expected, making it more difficult to select a single larger colony.Another approach could be to adjust the concentration of Cml to reduce the formation of SCVs.We theorize that SCVs are formed by lowering the copy number of the pSStarget plasmid (or expression levels of the encoded genes) to allow sufficient cat expression to grow slowly in the presence of Cml stress but not sufficient expression of the cas9 gene to induce a doublestranded break in the genomic DNA by Cas9.When the Cml concentration is increased from 5 to 7.5 μg/mL, the SCV formation is greatly reduced in strain P1/7.The optimal concentration of Cml required to suppress SCV formation is likely to be strain-dependent and might need to be optimized when this system is adaptable for use in other strains.

■ MATERIALS AND METHODS
Growth of Bacterial Strains.All strains used in this study are listed in Table 2. Liquid cultures of S. suis were grown in Todd-Hewitt broth (Oxoid) supplemented with 0.2% Bacto yeast extract (BD Biosciences) (THY) without agitation.For hemolysis assays, a complex medium supplemented with 1% (w/v) Pullulan (Sigma-Aldrich) (CM + Pul) has been prepared as previously described. 47For agar plate cultures, THY supplemented with 1.5% agar (Thermo Fisher) or Columbia agar plates (Thermo Fisher) supplemented with 5% sheep blood (Thermo Fisher) were used.Cultures were incubated in a humidified incubator at 37 °C and a 5% CO 2 level.Unless otherwise noted, chloramphenicol (Sigma-Aldrich) was added to a concentration of 5 μg/mL.
E. coliNEBturbo was used as a general cloning host.E. coliwas cultured in a Luria−Bertani (LB) medium (Merck) for liquid cultures.For agar plate cultures, LB was supplemented with 1.5% agar.When appropriate, antibiotics were added to the medium at following concentrations: chloramphenicol (Cml) 25 μg/mL; kanamycin (Merck) (Kan) 50 μg/mL; and tetracycline (Sigma-Aldrich) (Tet) 10 μg/mL.The cultures were incubated at 37 °C in a dry incubator without adding  S1 and   S2, respectively.Plasmid DNA was routinely isolated from E. colicultures using the QIAprep Spin Miniprep Kit (Qiagen).When large amounts of plasmid were required, the Plasmid Plus Midi Kit (Qiagen) was used for purification.Cloning PCRs were performed using Q5 High-Fidelity DNA Polymerase (New England Biolabs).Oligonucleotides were synthesized by IDT.PCR products were purified using the MSB Spin PCRapace kit (Invitek).For the isolation of Genomic DNA (gDNA), bacterial cells were transferred to a tube containing 0.1 mm silica beads (MP Biomedicals) and lysed by bead beating for 40 s at 4.0 m/s using a FastPrep-24 5G (MP Biomedicals).The lysates were cleared by centrifugation 16.000× for 10 min, and the gDNA has been isolated using the PowerSoil DNA Isolation Kit (Qiagen) according to manufacturer's instructions.Sanger sequencing was performed using the LightRun sequencing service according to instructions (Eurofins Genomics).
Chemically competent E. colistrains were transformed by heat shock.S. suis strains were transformed as previously described by inducing the natural competence pathway via the addition of the XIP peptide. 10Briefly, S. suis overnight cultures were diluted 40-fold and grown to an OD 600 of approximately 0.04−0.05.Aliquots of 100 μL of bacterial culture were transferred to a sterile microcentrifuge tube, and 1−10 μg of transforming DNA was added to each tube.The XIP peptide (Genscript) was added to a final concentration of 250 μM, and the cultures were incubated for 2 h at 37 °C with 5% CO 2 before plating 100 μL of transformation mix on selective agar plates.Appropriate dilutions were plated to allow the selection of single colonies.
Construction of pSStarget.For the construction of pSStarget, we used the previously published pLABtargetc as a starting point. 18To enable replication of this vector in E. coli, a fragment containing the p15A origin of replication (ori) and the tetracycline resistance gene tetA was amplified from pACYC184 (ATCC) using primers P1/P2.The pLABtargetc was linearized using SmaI (Thermo Fisher), and the purified PCR product was cloned into the linearized vector by blunt cloning using T4 Ligase (Promega) and transformed to E. coliTop 10 competent cells, resulting in vector pSStarget-NT.Next, pSStarget-NT was digested using BsaI-HFv2 (New England Biolabs).The ccdB region was amplified from pDONR221 (Invitrogen) using primers P3/P4.The purified fragments were assembled using the NEBuilder HiFi DNA Assembly Kit (New England Biolabs) and transformed to E. coliOne Shot ccdB Survival 2 T1R Competent Cells (Invitrogen).The final plasmid was sequence-verified using the certified Oxford Nanopore Technology (ONT) plasmid sequencing service offered by Plasmidsaurus.
Design and Assembly of sgRNAs in pSStarget.The sgRNAs were designed using the CRISPR tool in Benchling. 48he desired target region was selected, and the computed sgRNAs were sorted according to their on-target score. 49otential off-target sites in the S. suis P1/7 genome were predicted using Cas-Designer. 50For each target gene, multiple sgRNAs with the highest on-target scores were selected, and sgRNAs with potential off-target sites were omitted.To allow cloning into pSStarget, the nucleotide sequence "TGAT" was appended upstream of the target-specific sgRNA sequence, and "GTTT" was appended downstream of that sequence.In cases in which the sgRNA did not start with a G, the sequence "TGATG" was appended upstream instead, ensuring that each sgRNA started with a G.This sequence was used to design partially complementary oligonucleotides, which result in a dsDNA product with 4 bp overhangs on each side upon annealing that are complementary to the overhangs created on the pSStarget vector upon BsaI digestion.
The complementary oligonucleotides were mixed in annealing buffer (10 mM Tris pH 8, 50 mM NaCl) to yield a final concentration of 10 μM for each oligonucleotide.For annealing, this mixture was heated to 95 °C for 2 min and gradually cooled down to 25 °C over the course of 45 min.The resulting dsDNA fragments were cloned into pSStarget in a Golden Gate-like reaction.Each reaction contained 75 ng of pSStarget, 1 μL of annealed oligonucleotides (sgRNA), 2.5 μL of T4 DNA ligase buffer (10×), 1000 units of T4 DNA ligase (NEB#M0202, New England Biolabs), and 30 units of BsaI-HFv2 (NEB#R3733, New England Biolabs) in a total volume of 25 μL.This assembly mixture was then incubated in a thermocycler for 60 cycles (37 °C for 5 min, 16 °C for 5 min), followed by an inactivation step for 5 min at 65 °C.The reaction mixtures were then transformed to competent E. coliNEBturbo.
Construction of Homologous Repair Templates.Homologous RTs were assembled by SOEing (Splicing by Overlap Extension) PCR. 51For each deletion, flanking regions of approximately 1000 bp were selected directly upstream (US) and downstream (DS) of the region to be deleted.The NEBuilder assembly tool was used for primer design with matching overlaps on the internal primers, using a minimal overlap of 20 nt and unchecking the "circularize" box. 52The US and DS fragments were amplified by PCR using P1/7 gDNA as a template.The purified fragments were mixed in equimolar amounts and joined in a PCR lacking primers.The reaction products were diluted 50-fold and amplified by PCR using distal primers (US_Fw/DS_Rv).
For site-directed mutagenesis of the eno gene, a circular assembly was made, including a synthetic fragment containing the desired mutations (GeneArt), the homologous flanks, and a pUC57-kan (Genscript) fragment encoding kanamycin resistance and a replicon.All primers (P27−P32) were designed using the NEBuilder assembly tool, and the fragments

Figure 1 .
Figure 1.Overview of pSStarget and the sgRNA cloning strategy (A) plasmid map of pSStarget showing the main plasmid features.Protein coding sequences are shown in purple, promoters in green, and the p15A ori in yellow.The sgRNA expression cassette, which is divided into subparts, is colored light blue.The sgRNA expression cassette contains three BsaI sites but for clarity the internal one was omitted from panels (B−D) Schematic representation of the sgRNA cloning strategy.The BsaI restriction and ligation steps are depicted separately but can be combined in a single reaction tube because the correctly assembled constructs lack the BsaI sites.(B) Close-up of the sgRNA expression cassette.The BsaI restriction enzyme recognizes its binding site (purple) and cleaves the DNA, thereby removing the ccdB toxin gene together with the BsaI sites and generating 4 bp overhangs at the end of the linearized plasmid.(C) Annealing of two partially complementary oligonucleotides generates a dsDNA fragment encoding the spacer sequence with 4 bp overhangs on each side that are complementary to those on the linearized plasmid.(D) The spacer fragment is ligated into the sgRNA expression cassette, making use of the matching overhangs.

Figure 2 .
Figure 2. Schematic overview of the principle behind Cas9-mediated genome editing in S. suis.(A) S. suis is cotransformed with plasmid pSStarget and a linear repair template (RT) (B) homologous recombination of the RT with the chromosome occurs in a subset of the transformed cells.Meanwhile, Cas9 and sgRNA are expressed and assembled to form a complex.(C) Mature Cas9/sgRNA complex scans the genome for target sites.Edited cells that incorporated the RT by recombination lack the target site, making them immune to Cas9-mediated DNA cleavage.In the absence of HR, the target site is cleaved by Cas9, leading to cell death.

Figure 3 .
Figure 3. Representative pictures of colonies observed after plating the transformation mix on Columbia Blood Agar (CBA) supplemented with 5% sheep blood and 5 μg/mL chloramphenicol (Cml).Normal-sized colonies are indicated with black arrows, and SCVs are indicated using yellow arrows.(A) S. suis P1/7 transformed with pSS-sg42 and the corresponding repair template (1.2 μg each) forms many SCVs besides the normal-sized colonies.(B) S. suis P1/7 transformed only with pSS-sg42 (1.2 μg) was used as a "background control" to assess the amount of SCVs formed in the absence of a corresponding repair template.Mainly, SCVs are formed in the absence of an RT, and only a single normal-sized colony was observed.

Figure 4 .
Figure 4. Coverage graph showing the alignment of ONT long-read sequencing data generated for strain P1/7 and the mutant strains Δcps, Δsly, and Δcps Δsly to the cps (A) and sly (B) loci of the reference sequence for strain P1/7 (AM946016.1).The read coverage drops substantially in the regions targeted for deletion, confirming successful genome editing.

Figure 5 .
Figure 5. Phenotypic characterization of mutant strains.(A) Graph showing hemolytic activity of S. suis P1/7 and P1/7 Δsly.All values are the means of three experiments.Values are expressed as percent lysis relative to the complete lysis control (1% Triton X-100).(B) Image of a representative hemolysis assay plate.(C) Representative electron microscopy image of the WT P1/7 at 12,000× magnification.The polysaccharide capsule is clearly visible surrounding the bacterial cells.(D) Representative electron microscopy image of strain P1/7 Δcps at 12,000× magnification.No polysaccharide capsule is present on the surface of the cells, confirming the successful deletion of the cpsEF genes.(E) Quantification of NF-κB signaling resulting from hTLR2/6 activation by S. suis P1/7 and the Δcps, Δlgt, and Δcps Δlgt mutant strains.Deletion of the lgt gene resulted in decreased hTLR2/6 activation, whereas deletion of the cpsEF gene resulted in increased hTLR2/6 activation.Results are shown as relative luminescence values, with values of unstimulated hTLR2/6 cells set to 1. (F) Quantification of NF-κB signaling by S. suis P1/7 and the Δcps, Δlgt, and Δcps Δlgt mutant strains in control cells lacking TLRs.No differences between the tested strains were observed in this cell line, confirming the absence of TLR-independent NF-κB signaling.Results are shown as relative luminescence values, with values of unstimulated cells set to 1.

Figure 6 .
Figure 6.Alignment of Sanger sequencing reads showing the WT sequence and eno K261A .The top panel shows the WT sequence modified in silico by introducing silent mutations that do not affect the amino acid sequence.The middle panel shows Sanger sequencing results of the WT sequence, with the bases highlighted (pink boxes) that were changed to introduce the silent mutations displayed in the upper panel.The bottom panel shows the sequencing read corresponding to the successfully constructed mutant P1/7 eno K261A ; the boxed sequence AGC indicates the K261A replacement.

Figure 7 .
Figure 7. Growth of small-colony variants.(A) Growth curves recorded over 60 h in wells inoculated with colonies transformed with the nontargeting control pSStarget-NT (NT1, NT2) or SCVs recovered after transformation with pSStarget-sg4 (D6, E6, F5, and F6).While the control cultures displayed normal growth to maximum density over 24 h, the growth of SCVs was arrested or delayed (as shown in D6, E6, F5, and F6).(B) Sequencing of the target site revealed single-nucleotide substitutions (highlighted in pink) occurring in the protospacer (Well F6; 983C > T, 987T > G, 989T > G, and 901G > T) or PAM site (well D6, 998 G > A) of the sequence.No mutations were found in the NT control.The numbers below the nucleotide sequences refer to the nucleotide position in the sly coding sequence.

Table 1 .
Log2 Fold-Change (FC) in Expression Values of Selected Differentially Expressed Genes (SCV vs NT)

Table 2 .
Bacterial Strains Used in This Study All plasmids, oligonucleotides, and sgRNAs used in this study are listed in Table 3 and Supporting Information Tables

Table 3 .
Plasmids Used in This Study