Adaptation of Prokaryotic Toxins for Negative Selection and Cloning-Independent Markerless Mutagenesis in Streptococcus Species

ABSTRACT The Streptococcus mutans genetic system offers a variety of strategies to rapidly engineer targeted chromosomal mutations. Previously, we reported the first S. mutans negative selection system that functions in a wild-type background. This system utilizes induced sensitivity to the toxic amino acid analog p-chlorophenylalanine (4-CP) as a negative selection mechanism and was developed for counterselection-based cloning-independent markerless mutagenesis (CIMM). While we have employed this system extensively for our ongoing genetic studies, we have encountered a couple limitations with the system, mainly its narrow host range and the requirement for selection on a toxic substrate. Here, we report the development of a new negative selection system that addresses both limitations, while still retaining the utility of the previous 4-CP-based markerless mutagenesis system. We placed a variety of toxin-encoding genes under the control of the xylose-inducible gene expression cassette (Xyl-S) and found the Fst-sm and ParE toxins to be suitable candidates for inducible negative selection. We combined the inducible toxins with an antibiotic resistance gene to create several different counterselection cassettes. The most broadly useful of these contained a wild-type fst-sm open reading frame transcriptionally fused to a point mutant form of the Xyl-S expression system, which we subsequently named IFDC4. IFDC4 was shown to exhibit exceptionally low background resistance, with 3- to 4-log reductions in cell number observed when plating on xylose-supplemented medium. IFDC4 also functioned similarly in multiple strains of S. mutans as well as with Streptococcus gordonii and Streptococcus sanguinis. We performed CIMM with IFDC4 and successfully engineered a variety of different types of markerless mutations in all three species. The counterselection strategy described here provides a template approach that should be adaptable for the creation of similar counterselection systems in many other bacteria. IMPORTANCE Multiple medically significant Streptococcus species, such as S. mutans, have highly sophisticated genetic systems available, largely as a consequence of their amenability to genetic manipulation via natural competence. Despite this, few options are available for the creation of markerless mutations in streptococci, especially within wild-type strains. Markerless mutagenesis is a critical tool for genetic studies, as it allows the user to explore many fundamental questions that are not easily addressable using marked mutagenesis. Here, we describe a new approach for streptococcal markerless mutagenesis that offers a variety of advantages over the current approach, which employs induced sensitivity to the toxic substrate 4-CP. The approach employed here should be readily adaptable for the creation of similar markerless mutagenesis systems in other organisms.

M any Streptococcus species are highly amenable to genetic manipulation as a consequence of their efficient natural competence machinery (1,2). Accordingly, a variety of sophisticated genetic systems have also been developed for molecular pathogenesis studies of certain medically significant Streptococcus species, such as the prominent oral pathobiont Streptococcus mutans (3)(4)(5)(6). The S. mutans genetic system has now evolved to the point where targeted chromosomal mutations can be reliably engineered within a matter of days using various cloning-independent mutagenesis strategies. Such approaches circumvent the requirement for intermediate hosts like Escherichia coli during construct assembly, which significantly reduces the time and effort required to engineer mutations of interest. The vast majority of targeted mutations in S. mutans (and other streptococci) are created using marked mutagenesis, typically as allelic replacements with antibiotic resistance cassettes. While simple to engineer, marked mutations have a couple fundamental limitations that can be highly problematic for genetic studies. First, antibiotic resistance cassettes normally contain promoters and/or transcription terminators, which can introduce significant polar effects altering the expression patterns of genes downstream of a mutation site (7). Second, the number of individual mutations that can be engineered into a single strain is inherently limited by the number of unique selectable markers available for use in a particular organism. Both of these limitations can be addressed through the creation of markerless mutations, but unfortunately, only a limited number of organisms have markerless mutagenesis systems available for use.
Most markerless mutations are created by first employing marked mutagenesis to insert an antibiotic resistance cassette onto a chromosome, followed by a second step to subsequently remove the cassette to create the final markerless mutant strain. Once an antibiotic resistance cassette has been removed, the same procedure can be continually repeated to engineer any number of additional markerless mutations in the same strain (7). There are several common strategies employed to remove antibiotic resistance cassettes from the chromosome, and all have been previously employed in S. mutans. The first is a two-step integration and excision strategy using a conditionally replicating temperature-sensitive plasmid (8). Temperature-sensitive plasmids are available for only a limited number of organisms, and their temperature sensitivity often lacks stringency. The next strategy uses site-specific recombinases, such as Cre/LoxP, to remove antibiotic resistance cassettes following their insertion (9). This approach has the major advantage of being theoretically adaptable for a wide range of species, but also typically relies upon temperature-sensitive vectors to provide transient cre expression. Cre-mediated recombination between LoxP sites also generates scars on the chromosome that can interfere with the creation of certain types of mutations, like point mutations or specific gene truncations (10). The third common strategy for markerless mutagenesis utilizes a combination of both positive and negative selection (i.e., counterselection). When available for use, the counterselection approach is often preferred due to its ease of use and suitability for the creation of all types of mutations, such as deletions, insertions, truncations, point mutations, etc. (7,11). The principal limitation is that few efficacious negative selection markers are available for use in most bacteria.
We previously developed the first counterselection-based markerless mutagenesis system available for S. mutans, using induced sensitivity to galactose as a negative selection mechanism (12). By creating a mutant recipient strain defective in galactose catabolism, it was possible to employ the endogenous S. mutans galK gene (encoding galactokinase) as a negative selection marker in the presence of galactose-supplemented media. The obvious drawback is that this approach is not compatible with wild-type strains of S. mutans or any of the numerous other bacterial species that naturally catabolize galactose. This issue has limited the widespread adoption of the galK negative selection system. To improve upon this major shortcoming, we next adapted the PheS/p-chlorophenylalanine (4-CP) negative selection system for use in S. mutans markerless mutagenesis (7,13,14). For this approach, an A314G mutant version of the endogenous S. mutans PheS protein, PheS (A314G), is expressed for negative selection on medium supplemented with the toxic phenylalanine analog 4-CP. S. mutans strains producing the PheS(A314G) mutant protein exhibit much higher sensitivity to 4-CP toxicity than the wild type, thus facilitating negative selection. The major advantage to this approach is that the pheS gene is highly conserved among prokaryotes, and consequently, 4-CP negative selection is theoretically adaptable for use in most wild-type bacteria, unlike other negative selection approaches (7). To create a counterselection cassette, we combined an antibiotic resistance gene and a point mutant S. mutans pheS gene into a single cassette referred to as IFDC2. We further illustrated how to employ the IFDC2 cassette for the first demonstration of cloning-independent markerless mutagenesis (CIMM) (7). Despite the major improvements of the CIMM approach over previous markerless mutagenesis strategies, several limitations were still observed. Background 4-CP resistance was often higher than desired, and the IFDC2 cassette functioned only in S. mutans. In our experience, reliable 4-CP negative selection requires mutant pheS genes to be derived from the organism of interest. In a subsequent study, we were able to further increase the stringency of 4-CP negative selection in S. mutans by adding a second point mutation to the pheS gene contained on IFDC2, resulting in a PheS(T60S) mutation in addition to the prior A314G mutation (15). However, this updated version (referred to as IFDC3) still retained the requirement for species-specific pheS genes to be employed. In addition, while using 4-CPbased negative selection, we have encountered instances in which a particular chromosomal mutation sensitized the strain to 4-CP, thus rendering the markerless mutant difficult to isolate on 4-CP plates. For this reason, we were interested in searching for an alternative negative selection mechanism that does not require plating on medium supplemented with a toxic substrate like 4-CP. Additionally, we were interested in developing a single counterselection cassette that functions in multiple streptococci, unlike our pheS-based systems. Here, we describe the development of a toxin-based counterselection mechanism that satisfies these requirements, while also exhibiting exceptionally low background resistance.

RESULTS
Development of a toxin-based negative selection system in S. mutans. Despite the proven utility of 4-CP-based counterselection in S. mutans, there are still a couple limitations with this approach that can be problematic, mainly its requirement for a toxic substrate (i.e., 4-CP) and a narrow host range for pheS-containing counterselection cassettes. Here, we aimed to improve upon these by developing a new negative selection strategy to facilitate cloning-independent markerless mutagenesis (CIMM). We previously developed a xylose-based gene induction system (called Xyl-S) that was shown to exhibit an exceptionally wide dynamic range with low basal expression, and it also functioned similarly in several different oral streptococci, with no evidence of toxicity (16). In fact, we had even employed Xyl-S to engineer multiple conditional lethal mutations in S. mutans (16). Thus, we reasoned that the Xyl-S system may also be appropriate to control the expression of a toxic gene product to confer negative selection in S. mutans (Fig. 1). As shown in Table 1, we selected several candidate chromosomal toxins from verified S. mutans toxinantitoxin modules (mazF, smuT, and fst-sm) as well as the plasmid addiction module toxin gene parE from the S. mutans cryptic plasmid pUA140 and RNase H gene (rnH) from E. coli. Each of these genes was transcriptionally fused to the Xyl-S cassette and transformed into S. mutans strain UA159 to test for xylose-inducible negative selection. Following transformation of the constructs, we noted that most seemed to exhibit high levels of basal uninduced toxicity, as some transformants grew slowly and we observed fewer transformants than expected. Regardless, we selected candidate transformants from each reaction and then quantified their viability on agar plates 6 1% (wt/vol) xylose. As shown in Fig. 2A, only the fst-sm and parE constructs inhibited cell growth in the presence of xylose, with each exhibiting potent negative selection. We sequenced several clones of the fst-sm and parE strains that exhibited inducible negative selection and found all to contain at least one point mutation. For the strains harboring the fst-sm construct, they each contained either of two missense mutations within the same codon of the xylose repressor gene xylR, conferring in XylR either an A7S or A7T mutation [XylR(A7S) or XylR(A7T)] (Fig. 2B). The strains harboring parE constructs all contained the same XylR(A7S) mutation because one of the mutated fst-sm clones had been employed as a PCR template during the assembly of the parE construct (Fig. 2C). However, the parE strains also contained one of two additional point mutations, either within the parE ribosome binding site (RBS) or within the parE ORF, conferring a ParE(S56G) missense mutation (Fig. 2C). While the impact of the ParE(S56G) mutation was unclear, the parE(RBS) mutation almost certainly reduced the efficiency of parE translation, as the original construct contained a consensus Shine-Dalgarno sequence. Interestingly, we reassembled the parE construct with a wild-type xylR ORF together with either of the newly identified parE(RBS) or parE(S56G) mutations and found that the point mutant xylR was indeed required for construct stability. Thus, it would appear that the XylR(A7S) and, presumably, XylR(A7T) mutations are responsible for reducing the basal uninduced expression of fst-sm and parE. The requirement of additional compensatory point mutations for the proper functionality of the parE construct suggests that S. mutans strains encoding the ParE toxin are likely to be even more sensitive to toxic leaky gene expression compared to fst-sm.
Cloning-independent markerless mutagenesis in S. mutans. Based upon the extreme potency of the negative selection observed from the fst-sm and parE counterselection cassettes ( Fig. 2A), we were next curious to test the utility of these cassettes for CIMM. As a simple phenotypic readout, we employed the fst-sm and parE cassettes to engineer markerless gusA (b-glucuronidase) replacements of the S. mutans brsM ORF. Under normal growth conditions, the gene product of brsM inhibits the expression of the brsRM operon by preventing BrsR positive feedback autoregulation, leading to extremely low basal expression in the wild type ( Fig. 3A) (17)(18)(19). Consequently, insertion of the gusA ORF at the 39 end of the brsRM operon results a white colony phenotype (i.e., low basal gusA expression) because  Xylose is a nontoxic sugar that is not metabolizable by streptococci, but is efficiently transported into the cells via still unknown mechanisms. Xylose will bind to the xylose repressor XylR and prevent it from inhibiting target gene expression via the xylose promoter operator xylA O . To create markerless mutations via the two-step CIMM approach, the counterselection cassette is first created using OE-PCR to assemble the inducible toxin together with a positive selection marker, such as the aad9 gene conferring spectinomycin resistance. The assembled counterselection cassette is then introduced onto the chromosome by creating a typical allelic replacement mutation using positive selection. To remove the counterselection cassette and create a markerless mutation, a second unmarked OE-PCR fragment is subsequently transformed into the spectinomycinresistant strain and then plated on xylose-supplemented medium to initiate negative selection. The resulting xylose-resistant colonies should all contain the expected markerless mutant genotype.
the brsRM operon remains weakly expressed, similar to the parent wild type (Fig. 3A). Conversely, a gusA replacement of brsM will stimulate BrsR positive feedback autoregulation of the operon, resulting in strong gusA expression and a corresponding blue colony phenotype ( Fig. 3A). Using an existing brsRM-gusA reporter strain (18), we first replaced the gusA ORF downstream of brsM with the fst-sm and parE counterselection cassettes and subsequently performed a second transformation to replace both brsM and the counterselection cassette with gusA. As expected, the original parent brsRM-gusA reporter strain grew normally on the xylose plates, with undetectable b-glucuronidase activity (Fig. 3B). In agreement with the strong xylose-inducible negative selection previously observed with the fst-sm and parE counterselection cassettes ( Fig. 2A), we observed minimal background growth from the negative-control transformation reactions (Fig. 3B). In contrast, strains transformed with gusA DNA to replace both brsM and the counterselection cassettes all yielded numerous dark blue colonies, further confirming the efficacy of negative selection as well as the expected gusA allelic replacement of brsM (Fig. 3B). Importantly, despite the presence of some detectable background growth in the negative-control reactions, the markerless gusA transformations all yielded few, if any, white colonies, indicating that nearly 100% of the transformants exhibited the expected DbrsM gusA 1 genotype (Fig. 3B).
Comparison of the fst-sm and parE counterselection cassettes in multiple streptococci. As previously mentioned, one of our goals was to create a counterselection cassette exhibiting a broader host range than the previous pheS-based IFDC2 system. Since the fst-sm and parE cassettes were suitable for performing CIMM in S. mutans strain UA159, we next tested these same cassettes in three additional strains of S. mutans to determine whether they exhibit any evidence of strain specificity. Both of the fst-sm cassettes performed quite similarly in each of the three S. mutans strains as previously observed with UA159. On xylose-supplemented agar plates, we detected 3-to 4-log reductions in total cell number, which corresponds to #0.1% background growth ( Table 2). Unlike the fst-sm constructs, the results with the parE cassettes were mixed. The parE(RBS) point mutant version only exhibited noticeable negative selection in strain CL1 (Table 2), which is a serotype K clinical isolate (17). However, even in this strain, growth was partially inhibited on plates lacking xylose, which indicated that the basal expression of this toxin was still negatively affecting its growth. The parE(S56G) mutant cassette performed better than the RBS mutant cassette, yielding xylose-inducible 3-to 4-log reductions in strains CL1 and JF243, with no obvious toxicity on the xylose-free plates (Table 2). Interestingly, strain UA140 was completely resistant to negative selection using both of the parE cassettes (Table 2). It is not yet clear why parE selection failed in UA140, but we suspect it is because this strain naturally harbors the pUA140 cryptic plasmid that carries the ParE addiction module. From these results, we conclude that both of the fst-sm counterselection cassettes are likely to function well in most strains of S. mutans, while the ParE(S56G) cassette is slightly less universal, possibly due to the presence of the pUA140 plasmid in some strains.
Since we had previously demonstrated the utility of the Xyl-S system for xylose-inducible gene expression in both Streptococcus gordonii and Streptococcus sanguinis (16), we were Streptococcal Negative Selection System mSphere also curious to determine whether the xylose-inducible fst-sm and parE counterselection cassettes would function in these organisms. We introduced the four fst-sm and parE counterselection cassettes into S. gordonii strain DL1 and S. sanguinis strain SK36, which are among the most commonly utilized strains for genetic studies of both species. Only the fst-sm(A7T) cassette functioned in both organisms, yielding 3-log reductions in the presence of xylose (Table 2). Surprisingly, the RBS mutant parE cassette functioned even better than fst-sm(A7T) in S. sanguinis, but it completely failed to inhibit S. gordonii ( Table 2). The fst-sm(A7S) and parE(S56G) cassettes did not function in either S. gordonii or S. sanguinis. Overall, the negative selection results indicated that the fst-sm(A7T) cassette is likely to be the most broadly useful counterselection cassette for streptococci, although the fst-sm(A7S) and parE cassettes may also function quite well for certain species or strains. To further confirm the utility of the fst-sm(A7T) counterselection cassette in S. gordonii and S. sanguinis, we performed CIMM with this cassette to markerlessly insert the green renilla luciferase ORF (renG) immediately downstream of the pyruvate oxidase-encoding gene spxB in both species (Fig. 4A). Since the pyruvate oxidase enzyme is responsible for generating the majority of the H 2 O 2 excreted by these two streptococci (20,21), the spxB mutant phenotype is readily observable on Prussian blue plates as this dye reacts with H 2 O 2 to produce a blue precipitate. As shown in Fig. 4B, both S. gordonii and S. sanguinis produced much less blue precipitate when spxB was replaced with the fst-sm(A7T) counterselection cassette. However, wild-type levels of H 2 O 2 were once again detectable when spxB was markerlessly inserted back to its original locus together with a downstream renG ORF (Fig. 4B). We also measured luciferase activity from both of the markerless luciferase reporter strains and observed ;4-log-increased reporter activity over the background, unlike the parent spxB deletion strains, which lacked the renG ORF (Fig. 3C). Thus, we conclude that the fst-sm (A7T) counterselection cassette (henceforth referred to as IFDC4) is suitable for both markerless deletions and insertions in S. gordonii and S. sanguinis. Markerless point mutagenesis. As previously described, one potential advantage of counterselection-based markerless mutagenesis is that it supports the creation of targeted point mutations due to scarless excision of the counterselection cassettes. Therefore, as a final confirmation of IFDC4 utility, we designed CIMM constructs to create nonsense point mutations within the galK genes of S. mutans, S. gordonii, and S. sanguinis. We chose to mutate galK because this gene is both nonessential in all three species and should confer resistance to the toxic effects of the galactose analog deoxygalactose, yielding an easily observable growth phenotype. As shown in Fig. 5A, deoxygalactose resistance was indeed created after markerlessly introducing an ochre (CAA!TAA) nonsense point mutation into codon 8 of the S. mutans UA159 galK gene. We repeated the same experiment using S. gordonii DL1 and were able to successfully introduce a similar ochre (CAA!TAA) nonsense mutation into codon 9 of its galK gene (Fig. 5B). However, we were surprised to discover that wild-type DL1 is naturally resistant to the toxic effects of deoxygalactose. Therefore, unlike S. mutans, we were unable to observe any difference in growth between the wild-type and galK point mutant strains. For unknown reasons, we could not achieve any discernible negative selection when trying to mutate the galK gene of S. sanguinis SK36, despite the efficacy of negative selection with our other SK36 mutant constructs using the same IFDC4   (Table 2 and Fig. 4B and C). We even repeated the experiment using the ParE(RBS) mutant counterselection cassette, which exhibited even stronger negative selection in SK36 compared to IFDC4 (Table 2), yet we still observed the same problem. This issue appeared unique to the galK locus, as negative selection was quite stringent for our other SK36 constructs (Table 2 and Fig. 4B and C), and we have also recently used IFDC4 to engineer Streptococcal Negative Selection System mSphere additional markerless mutations in SK36 for our other ongoing research. It is unclear why the SK36 galK locus is specifically problematic, but based upon our results, it would appear that this mutation affects the utility of xylose induction. However, additional studies would be required to determine whether this is indeed the cause. Regardless, our success with both S. mutans and S. gordonii suggests that point mutagenesis would be unlikely to fail in most instances in SK36.

DISCUSSION
In the current study, we describe a new approach to perform CIMM in streptococci. This system was developed to address a couple of the limitations encountered when using the previous pheS-based IFDC2 counterselection cassette, mainly its requirement for a toxic substrate during negative selection (i.e., 4-CP) and its narrow host range. Here, we employed the Xyl-S induction system to control the expression of a toxic gene product as a negative selection mechanism. We have previously used the Xyl-S system to introduce controllable gene expression in a number of in vitro and in vivo studies (16,22), so this expression system has proven utility in multiple streptococci with no evidence of xylose toxicity. Furthermore, our previous studies had already suggested that the Xyl-S system would be suitable for engineering conditional lethality (16), which we repurposed here as a negative selection mechanism for CIMM. As shown in Table 2, negative selection was highly efficient with exceptionally low background, resulting in an extremely high percentage of markerless mutants with the expected genotypes (Fig. 3B). It is currently unclear why the fst-sm(A7S) and parE(S56G) cassettes did not function in either S. gordonii or S. sanguinis, whereas both yielded strong negative selection in S. mutans. However, it is worth noting that the fst-sm (A7S) cassette is apparently less efficacious in some S. mutans strains than the fst-sm(A7T) cassette (Table 2). Therefore, we suspect that the XylR(A7T) mutation likely yields a more stringent repressor than both the wild-type and XylR(A7S) versions. This difference in regulatory stringency may be even more exaggerated in S. gordonii and S. sanguinis. Furthermore, unlike the IFDC2 cassette, IFDC4 [encoding Fst-sm(A7T)] also allowed us to employ a single cassette to engineer a variety of different types of markerless mutations in S. mutans, S. gordonii, and S. sanguinis, thus confirming its broader host range. For unknown reasons, we were unable to induce negative selection when mutating the galK gene of S. sanguinis. This problem seemed specific to the S. sanguinis galK locus, as we did not encounter similar issues with either of the S. mutans or S. gordonii galK mutations (Fig. 5A and B), nor did we experience issues with negative selection in other S. sanguinis loci ( Table 2 and  Fig. 4B and C). We tried a variety of different construct designs to create the galK mutant, but all failed to exhibit any evidence of xylose-inducible negative selection. Therefore, we suspect that xylose is either modified or exhibits defective transport following mutagenesis of the S. sanguinis galK locus.
Our results provide a general strategy for negative selection that should be adaptable for use in other organisms. Most bacteria encode endogenous toxin-antitoxin modules on their chromosomes, while many others also have particular strains that naturally host cryptic plasmids, which are highly likely to contain addiction modules (23,24). Thus, there is a strong reservoir of potential toxins that could be exploited for use in most bacteria. This approach also requires a reliable regulated gene expression system, which may be a limiting factor for certain organisms, especially those with minimal genetic tools available. From our experience, successful toxin-based negative selection requires both low basal uninduced expression of the toxin gene as well as strong inducibility (i.e., a wide dynamic range of expression). However, the specific expression characteristics required to create an efficacious inducible negative selection system are likely to vary widely. Thus, the challenge is to find an appropriate match between the available expression system(s) and the toxicity of a particular toxic gene product. For example, our results suggest that ParE is likely to be a more potent toxin in S. mutans than Fst-sm, as fst-sm could be stably expressed from the Xyl-S expression system after acquiring a single point mutation within the xylR gene, whereas the parE constructs required the same xylR point mutation as well as additional compensatory parE mutations (Fig. 2C). Thus, when developing a new toxin-based negative selection system, it is advisable to screen a variety of candidate toxins to increase the likelihood of identifying the appropriate gene to pair with an available expression system. In our case, none of the toxin genes we selected was immediately ideal for our Xyl-S expression system. However, the fst-sm construct was apparently a sufficiently close match that it allowed us to isolate spontaneous compensatory xylose repressor mutations, resulting in a mutant counterselection cassette with lower basal uninduced toxicity, but stringent xylose-inducible negative selection. It seems unlikely that we would have identified the appropriate compensatory parE mutations if the strain did not already contain the XylR (A7S) mutation within the Xyl-S induction system, which apparently reduced its leakiness. This may also explain why we did not isolate spontaneous compensatory mutations for the other toxins in our initial screen of candidates, as these were constructed without the XylR(A7S) or XylR(A7T) mutations. Presumably, the mutants we did isolate with these other toxins all contained inactivating mutations because none of the isolates exhibited any evidence of xylose-inducible negative selection ( Fig. 2A). Since we had already observed effective negative selection from the fst-sm and parE constructs, we did not perform additional tests to determine whether the other toxin genes might function in combination with the XylR(A7S) mutation. However, it is certainly conceivable that at least some of them would, perhaps requiring additional compensatory mutations similar to the parE construct. As our results demonstrate, it is not essential for a given gene induction system to perfectly pair with a toxin gene, provided the initial basal uninduced toxicity of the construct is moderate enough that it gives the cell a chance to develop the appropriate compensatory mutations. From there, one can screen the isolates to identify those that may be suitable for stable negative selection. Finally, it is worth mentioning that we sequenced a strain harboring the IFDC4 cassette and did not detect any single nucleotide polymorphisms (SNPs) elsewhere on the chromosome, suggesting that the cassette is relatively stable in its uninduced state (see Table S3, Table S4, and Table S5 in the supplemental material). When developing novel negative selection cassettes, the magnitude of uninduced basal toxicity from the cassettes is likely to play a major role in determining the frequency of acquired secondary mutations. Therefore, as a general rule, strains harboring an optimized negative selection cassette should exhibit normal growth rates in the absence of inducer. Other negative selection approaches that employ media containing toxic chemicals like 4-CP can also affect normal cell growth and share the same potential to promote off-target secondary mutations. Therefore, it is advisable to include controls like genetic complementation and genotyping when performing genetic studies employing counterselection-based markerless mutagenesis. With the appropriate optimization, it should be possible to create a highly reliable toxin-based negative selection system that supports the facile creation of various types of mutant strains that would otherwise be typically unattainable with marked mutagenesis.

MATERIALS AND METHODS
Primers, bacterial strains, and growth conditions. The primers and bacterial strains used in this study are shown in Table S1 and Table S2, respectively, in the supplemental material. S. mutans strains UA159, UA140, JF243, CL1, and their derivatives were cultured anaerobically (in an atmosphere consisting of 85% N 2 , 10% CO 2 , and 5% H 2 ) at 37°C in Todd-Hewitt medium (Difco) supplemented with 0.3% (wt/vol) yeast extract (THYE) or on THYE agar plates. For the selection of antibiotic-resistant colonies, 1 mg mL 21 spectinomycin (Sigma-Aldrich) was added to growth media. Streptococcus sanguinis strain SK36, Streptococcus gordonii strain DL1, and their derivatives were cultured in 5% CO 2 at 37°C in brain heart infusion medium (BHI) (Difco) or on BHI agar plates. For the selection of antibiotic-resistant colonies, 400 mg mL 21 spectinomycin (Sigma-Aldrich) was added to growth media.
Transformation. DNA constructs were introduced into S. mutans using a previously described methodology (2). Briefly, S. mutans cultures were diluted 1:40 from overnight cultures and grown to an optical density at 600 nm (OD 600 ) of ;0.1 in THYE before the addition of transforming DNA and 1 mg mL 21 competence-stimulating peptide (CSP) (GenScript). The cultures were subsequently incubated for an additional 2 h and then plated on antibiotic-supplemented THYE plates. Transformation of Streptococcus sanguinis and Streptococcus gordonii was performed as described previously (22). Briefly, cultures were diluted 1:40 from overnight cultures and grown to an OD 600 of ;0.07 in BHI medium before the addition of transforming DNA and 1 mg mL 21 of the appropriate species-specific CSP (ChemPep). The cultures were subsequently incubated for an additional 2 h and then plated on antibiotic-supplemented BHI plates. For strains undergoing negative selection, THYE or BHI plates were supplemented with 1% (wt/vol) xylose (Sigma-Aldrich).
DNA manipulation. Phusion DNA polymerase (Thermo Scientific) or AccuPrime polymerase (Invitrogen) was used to amplify individual PCR amplicons and overlap extension PCR (OE-PCR) products.
Generation of strains harboring candidate counterselection cassettes. Each candidate counterselection cassette was created using OE-PCR to transcriptionally fuse a toxic gene product to the Xyl-S xylose induction cassette (16) for negative selection, and then it was subsequently combined with a downstream spectinomycin resistance gene, aad9, for positive selection (see Fig. S1A in the supplemental material). Each of the candidate counterselection cassettes was inserted into the brsRM locus of S. mutans (17) to assay its functionality (Fig. S1B). The Fst-sm-encoding counterselection cassette was the first to be constructed. The Xyl-S induction cassette and spectinomycin resistance gene aad9 were both PCR amplified from the plasmid pZX9 (16) using the primer pairs 1925-2 Fwd and 1925-2 Rvs and 1925-4 Fwd and 1925-IFDC-Rvs, respectively. The fst-sm gene was PCR amplified using UA159 genomic DNA (gDNA) and the primer pair 1925-3 Fst-sm Fwd and 1925-3 Fst-sm Rvs. Each of the resulting PCR amplicons contains segments of sequence complementarity facilitating their subsequent assembly via OE-PCR with the primer pair 1925-2 Fwd and 1925-IFDC-Rvs. Next, the upstream and downstream homologous fragments used for targeting recombination within the brsRM locus were PCR amplified from UA159 gDNA using the primer pairs 1925-1 brsRM159-LF and 1925-1 Rvs and 2018-2 Fwd and 1925-5 brsRM159-RR, respectively. The resulting PCR amplicons were mixed with that of the previously assembled inducible fst-sm construct and assembled into a final construct via OE-PCR using the primer pair 1925-1 brsRM159-LF and 1925-5 brsRM159-RR. The resulting full-length construct was then transformed to UA159 and selected on antibiotic-supplemented agar plates. Several clones of the resulting transformation were tested for negative selection in the presence of xylose. Clones of interest were sequenced to confirm the expected counterselection cassette genotype. To compare the functionality of the fst-sm-and parE-containing counterselection cassettes in other S. mutans strains, the cassettes were first PCR amplified from strains 159MfstS, 159MfstT, 159MparE1, and 159MparE2 using the primers 1925-1 brsRM159-LF and 1925-5 brsRM159-RR. These amplicons were subsequently transformed into S. mutans wildtype strains CL1, JF243, and UA140 and selected on antibiotic-supplemented agar plates. The resulting strains were named CL1MfstS, CL1MfstT, JF243MfstS, JF243MfstT, 140MfstS, and 140MfstT (Fst-sm) as well as CL1MparE1, CL1MparE2, JF243MparE1, JF243MparE2, 140MparE1, and 140MparE2 (ParE).
To test the functionality of the counterselection cassettes in both Streptococcus sanguinis and Streptococcus gordonii, candidate cassettes were used to create allelic replacement mutants of the spxB genes from both species. For S. sanguinis, the spxB upstream and downstream homologous fragments were PCR amplified from wild-type strain SK36 using the primer pairs 2010-1-LF and 2110-1-LR and 2110-3-RF and 2110-3-RR, respectively. The counterselection cassettes were PCR amplified from S. mutans strains 159MfstS, 159MfstT, 159MparE1, and 159MparE2 using the primers 2110-2 DspxB Fwd and 2110-2 DspxB Rvs. The upstream and downstream homologous fragments were mixed with each of the counterselection cassette amplicons and assembled via OE-PCR using the primers 2010-1-LF and 2110-3-RR. These amplicons were subsequently transformed into S. sanguinis strain SK36 and selected on antibiotic-supplemented agar plates. The resulting strains were named SK36SBfstS and SK36SBfstT (Fst-sm) as well as SK36SBparE1 and SK36SBparE2 Single nucleotide polymorphism analysis. The genotype of an S. mutans strain harboring the IFDC4 cassette in the brsRM locus was compared to its parental UA159 wild-type as well as another UA159 derivative strain harboring the aad9 spectinomycin resistance cassette inserted into the brsRM locus. All three strains were subjected to whole-genome sequencing and then analyzed for the presence of single nucleotide polymorphisms. Adapters were trimmed and low-quality reads were filtered out using Trimmomatic (v.0.39) (25). SPAdes genome assembler (v.3.15.5) (26) was used to assemble reads from the UA159 wild-type strain into 63 contigs (total length, 2,006,195 bp; maximum contig length, 237,927 bp). Snippy (v.4.6.0) (27) was used to compare the trimmed fastq reads to the reference UA159 contig file for both the UA159-aad9 and UA159-IDFC4 strains. The parental UA159 sequence data were used as a positive control for snippy analysis. The bwa mem algorithm (28) was used to map the trimmed fastq reads from the UA159-aad9 and UA159-IFDC4 strains to the UA159 contigs to confirm adequate genome coverage (100% coverage for all contigs of .1,000 bp). Additional SNP mapping statistics are provided in Table S3, Table S4, and Table S5.
Markerless replacement of brsM with the b-glucuronidase-encoding gene gusA. In order to check the efficiency of markerless allelic replacements, gDNA from strain ifdLRS/brsRM-gusA (18) was used a template for PCR using the primer pair 1925-1 brsRM159-LF and 1925-5 brsRM159-RR. This PCR amplicon was transformed into strains 159MfstS, 159MfstT, 159MparE1, and 159MparE2 to replace the brsM open reading frame (ORF) with the gusA ORF and remove the respective counterselection cassettes. Transformants were selected on THYE plates supplemented with 1% (wt/vol) xylose and 200 mg mL 21 X-Gluc (5-bromo-4-chloro-3indolyl-b-D-glucuronic acid). The generation of the expected DbrsM gusA 1 genotype results in transformants with blue color, while background growth of the parent strain results in white colonies.
Creation of markerless spxB-renG green renilla luciferase reporters in S. sanguinis and S. gordonii. To create the markerless spxB-renG transcription fusion in S. sanguinis strain SK36, we transformed the renG construct into a previously constructed fst-sm counterselection cassette allelic replacement of the spxB ORF. The SK36 spxB upstream homologous fragment was PCR amplified from SK36 gDNA using the primer pair 2110-1-LF and 2021-17-1-Rvs, while the downstream homologous fragment was PCR amplified using the same gDNA template with the primer pair 2021-17-3-Fwd and 2021-17-3-Rvs. The renG ORF was PCR amplified from the gDNA of a previously constructed comX-renG reporter strain of S. mutans using the primer pair 2021-17-2 Fwd renG and 2021-17-2 Rvs renG. The resulting PCR amplicons were mixed and assembled via OE-PCR using the primer pair 2110-1-LF and 2021-17-3-Rvs. The OE-PCR amplicon was subsequently transformed into strain SK36SBfstT and selected on BHI agar plates supplemented with 1% (wt/vol) xylose to generate the markerless spxB-renG reporter strain SK36RGfstT.
To create the markerless spxB-renG transcription fusion in S. gordonii strain DL1, we transformed the renG construct into a previously constructed fst-sm counterselection cassette allelic replacement of the spxB ORF. The DL1 spxB upstream homologous fragment was PCR amplified from DL1 gDNA using the primer pair 2021-22-1-LF and 2021-23-1-Rvs, while the downstream homologous fragment was PCR amplified using the same gDNA template with the primer pair 2021-23-3-Fwd and 2021-23-3-Rvs. The renG ORF was PCR amplified from the gDNA of a previously constructed comX-renG reporter strain of S. mutans using the primer pair 2021-17-2 Fwd renG and 2021-17-2 Rvs renG. The resulting PCR amplicons were mixed and assembled via OE-PCR using the primer pair 2021-22-1-LF and 2021-23-3-Rvs. The OE-PCR amplicon was subsequently transformed into strain DL1SBfstT and selected on BHI agar plates supplemented with 1% (wt vol) xylose to generate the markerless spxB-renG reporter strain DL1RGfstT.
Generation of markerless galK nonsense point mutations in S. mutans and S. gordonii. To engineer a markerless point mutation into the S. mutans galK gene, we first inserted the fst-sm(A7T) counterselection cassette between the galR and galK ORFs on the chromosome of S. mutans strain UA159. The upstream and downstream homologous fragments were PCR amplified from UA159 using the primer pairs 2130-1Fwd LAM and 2130-1Rvs LAM and 2130-3Fwd RAM and 2130-3 Rvs RAM, respectively. The fst-sm(A7T) counterselection cassette was amplified from S. mutans strain 159MfstT using the primer pair 1925-IFDC-Fwd and 1925-IFDC-Rvs. The three PCR amplicons were mixed and subsequently assembled using OE-PCR with the primer pair 2130-4 nest Fwd and 2130-4 nest Rvs. The assembled construct was then transformed to UA159 and selected on antibiotic-supplemented agar plates to generate the intermediate strain 159GKfstT. A second construct was created to replace the counterselection cassette with a mutagenized version of the galK ORF in which codon 8 was changed from CAA to TAA. The upstream and downstream homologous fragments were PCR amplified from UA159 using the primer pairs 2130-1 Fwd LAM and 2130-5 Rvs galk mut and 2130-6 Fwd galk mut and 2130-3 Rvs RAM, respectively. The PCR amplicons were mixed and assembled via OE-PCR using the primer pair 2130-1Fwd LAM and 2130-3 Rvs RAM to generate the final construct, which was transformed into strain 159GKfstT and selected on xylose-supplemented agar plates to generate the strain 159galk22. Several transformants were sequenced to confirm the expected point mutant genotypes.
To generate a markerless galK point mutation in S. gordonii strain DL1, a similar strategy to that described for strain 159galk22 was employed. We first inserted the fst-sm(A7T) counterselection cassette between the galR and galK ORFs on the chromosome of DL1. The upstream and downstream homologous fragments were PCR amplified from DL1 using the primer pairs 2201-1 Fwd LAM and 2201-1 Rvs LAM and 2201-3 Fwd RAM and 2201-3 Rvs RAM, respectively. The fst-sm(A7T) counterselection cassette was PCR amplified from S. mutans strain 159MfstT using the primer pair 1925-IFDC-Fwd and 1925-IFDC-Rvs. The three PCR amplicons were mixed and subsequently assembled using OE-PCR with the primer pair 2201-4 nest Fwd and 2201-4 nest Rvs. The assembled construct was then transformed to DL1 and selected on antibiotic-supplemented agar plates to generate the intermediate strain DL1GKfstT. A second construct was created to replace the counterselection cassette with a mutagenized version of the galK ORF in which codon 9 was changed from CAA to TAA. The upstream and downstream homologous fragments were PCR amplified from DL1 using the primer pairs 2201-1 Fwd LAM and 2201-5 Rvs galk mut and 2201-6 Fwd galk mut and 2201-3 Rvs RAM, respectively. The PCR amplicons were mixed and assembled via OE-PCR using the primer pair 2201-1 Fwd LAM and 2201-3 Rvs RAM to generate the final construct, which was transformed into intermediate strain DL1GKfstT and selected on xylose-supplemented agar plates to generate the final strain DL1galk25. Several transformants were sequenced to confirm the expected point mutant genotypes.
Detection of hydrogen peroxide production on Prussian blue indicator plates. The preparation of H 2 O 2 indicator plates and detection of bacterial H 2 O 2 production were performed using a previously described methodology (20). Briefly, overnight cultures of each strain were washed twice with phosphate-buffered saline (PBS) and then adjusted to an OD 600 of 0.5. A volume of 10 mL was pipetted onto BHI Prussian blue indicator plates and incubated overnight at 37°C in a 5% CO 2 atmosphere.
Luciferase assays. Luciferase assays were performed using a previously described methodology (22,29). Briefly, strains were diluted 1:40 from overnight liquid cultures and then incubated for 4 h at 37°C. Optical densities (OD 600 ) were subsequently measured for each sample to normalize luciferase values. Coelenterazine-H solution (Prolume) was added to each sample at a final concentration of 7.5 mg mL 21 , immediately followed by measuring the resulting luciferase activity using a Promega Glomax Discover luminometer. Normalized luciferase activity was determined by dividing luciferase relative light units (RLU) by the measured OD 600 values.
Measurement of deoxygalactose sensitivity. Overnight cultures were washed twice with PBS and adjusted to an OD 600 of 0.5. The cultures were serially diluted 10-fold, and 10 mL of each dilution was pipetted onto BHI agar plates 6 1% (wt/vol) deoxygalactose. The plates were incubated anaerobically at 37°C for 24 h.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, TIF file, 5 MB.