Adaptation of prokaryotic toxins for negative selection and cloning-independent markerless mutagenesis (CIMM) in Streptococcus species

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 Xyl-S expression cassette 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 as IFDC4. IFDC4 was shown to exhibit exceptionally low background resistance, with 3 – 4 log reductions in cell number observed when plating on xylose-supplemented media. IFDC4 also functioned similarly in multiple strains of S. mutans as well as with S. gordonii and S. 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.

selection system that addresses both limitations, while still retaining the utility of the previous 4-CP-based 28 markerless mutagenesis system. We placed a variety of toxin-encoding genes under the control of the xylose-29 inducible Xyl-S expression cassette and found the Fst-sm and ParE toxins to be suitable candidates for inducible 30 negative selection. We combined the inducible toxins with an antibiotic resistance gene to create several different 31 counterselection cassettes. The most broadly useful of these contained a wild-type fst-sm open reading frame 32 transcriptionally fused to a point mutant form of the Xyl-S expression system, which we subsequently named as 33 IFDC4. IFDC4 was shown to exhibit exceptionally low background resistance, with 3 -4 log reductions in cell 34 number observed when plating on xylose-supplemented media. IFDC4 also functioned similarly in multiple 35 strains of S. mutans as well as with S. gordonii and S. sanguinis. We performed CIMM with IFDC4 and 36 successfully engineered a variety of different types of markerless mutations in all three species. The 37 hosts like E. coli during construct assembly, which significantly reduces the time and effort required to engineer 48 mutations of interest. The vast majority of targeted mutations in S. mutans (and other streptococci) are created 49 using marked mutagenesis, typically as allelic replacements with antibiotic resistance cassettes. While simple to 50 engineer, marked mutations have a couple fundamental limitations that can be highly problematic for genetic 51 studies. Firstly, antibiotic resistance cassettes normally contain promoters and/or transcription terminators, which 52 can introduce significant polar effects altering the expression patterns of genes downstream of a mutation site 53 (7). Secondly, the number of individual mutations that can be engineered into a single strain is inherently limited 54 by the number of unique selectable markers available for use in a particular organism. Both of these limitations 55 can by addressed through the creation of markerless mutations, but unfortunately, only a limited number of 56 organisms have markerless mutagenesis systems available for use. 57 58 Most markerless mutations are created by first employing marked mutagenesis to insert an antibiotic resistance 59 cassette onto a chromosome followed by a second step to subsequently remove the cassette to create the final 60 markerless mutant strain. Once an antibiotic resistance cassette has been removed, the same procedure can 61 be continually repeated to engineer any number of additional markerless mutations in the same strain (7). There 62 are several common strategies employed to remove antibiotic resistance cassettes from the chromosome and 63 all have been previously employed in S. mutans. The first is a two-step integration and excision strategy using a 64 conditionally replicating temperature sensitive plasmid (8). Temperature-sensitive plasmids are only available 65 for a limited number of organisms, and their temperature sensitivity often lacks stringency. The next strategy 66 uses site-specific recombinases, such as Cre/LoxP to remove antibiotic resistance cassettes following their 67 insertion (9). This approach has the major advantage of being theoretically adaptable for a wide range of species, 68 but also typically relies upon temperature sensitive vectors to provide transient cre expression. Cre-mediated 69 recombination between LoxP sites also generates scars on the chromosome that can interfere with the creation 70 of certain types of mutations like point mutations or specific gene truncations (10). The third common strategy 71 for markerless mutagenesis utilizes a combination of both positive and negative selection (i.e. counterselection). 72 When available for use, the counterselection approach is often preferred due to its ease of use and suitability for 73 the creation of all types of mutations, such as deletions, insertions, truncations, point mutations, etc. (7, 11). The 74 principal limitation is that few efficacious negative selection markers are available for use in most bacteria. 75

76
We previously developed the first counterselection-based markerless mutagenesis system available for S. 77 mutans, using induced sensitivity to galactose as a negative selection mechanism (12). By creating a mutant 78 recipient strain defective in galactose catabolism, it was possible to employ the endogenous S. mutans galK 79 gene (encoding galactokinase) as a negative selection marker in the presence of galactose-supplemented 80 media. The obvious drawback is that this approach is not compatible with wild-type strains of S. mutans or any 81

19
Transformation 20 DNA constructs were introduced into S. mutans using a previously described methodology (2). Briefly, S. 21 mutans cultures were diluted 1:40 from overnight cultures and grown to an optical density of OD600 ~0.1 in THYE 22 before the addition of transforming DNA and 1 μg ml −1 Competence Stimulating Peptide (CSP; GenScript). The 23 cultures were subsequently incubated for an additional 2 h and then plated on antibiotic-supplemented THYE 24 plates. Transformation of Streptococcus sanguinis and Streptococcus gordonii were performed as described 25 previously (16). Briefly, cultures were diluted 1:40 from overnight cultures and grown to an optical density of 26 OD600 ~0.07 in BHI medium before the addition of transforming DNA and 1 μg ml −1 of the appropriate CSP 27 (ChemPep). The cultures were subsequently incubated for an additional 2 h and then plated on antibiotic-28 supplemented BHI plates. For strains undergoing negative selection, THYE or BHI plates were supplemented 29 with 1% (wt vol -1 ) xylose (Sigma-Aldrich). 30

DNA manipulation 32
Phusion DNA polymerase (Thermo Scientific) or AccuPrime Polymerase (Invitrogen) were used to amplify 33 individual PCR amplicons and overlap extension PCR (OE-PCR) products. 34 35 Generation of strains harboring candidate counterselection cassettes 36 Each candidate counterselection cassette was created using OE-PCR to transcriptionally fuse a toxic gene 37 product to the Xyl-S xylose induction cassette (17) for negative selection and then it was subsequently combined 38 with a downstream spectinomycin resistance gene aad9 for positive selection. Each of the candidate 39 counterselection cassettes was inserted into the brsRM locus of S. mutans (18) to assay its functionality. The 40 Fst-sm-encoding counterselection cassette was the first to be constructed. The Xyl-S induction cassette and 41 spectinomycin resistance gene aad9 were both PCR amplified from the plasmid pZX9 (17) using the primer pairs 42 1925-2 Fwd + 1925-2 Rvs and 1925-4 Fwd + 1925 respectively. The resulting PCR amplicons were mixed with that of the previously assembled inducible fst-sm 49 construct and assembled into a final construct via OE-PCR using the primer pair 1925-1 brsRM159-LF + 1925-50 5 brsRM159-RR. The resulting full-length construct was then transformed to UA159 and selected on antibiotic-51 supplemented agar plates. Several clones of the resulting transformation were tested for negative selection in 52 the presence of xylose. Clones of interest were sequenced to confirm the expected counterselection cassette 53 genotype. Functional counterselection cassettes contained either of two point mutations located within codon 7 54 of the xylR open reading frame (ORF), resulting in either an A7S or A7T mutation within XylR. These two strains 55 were subsequently named as 159MfstS (XylR A7S) and 159MfstT (XylR A7T). To generate counterselection 56 cassettes encoding MazF, SmuT, and RNaseH, similar assembly strategies were performed as described for 57 strains 159MfstS and 159MfstT, except that the following primers pairs were employed. The mazF construct 58 utilized primers 1925-2 Fwd + 1915-2M Rvs, 1915-3 MazF Fwd + 1915-3 MazF Rvs, and 1915-4M Fwd + 1925 IFDC-Rvs to generate the strain 159MmazF. The smuT construct utilized primers 1925-2 Fwd + 1915-2S Rvs, 60 1915-3 SmuT Fwd + 1915-3 SmuT Rvs, and 1915-4S Fwd + 1925-IFDC-Rvs to generate the strain 159MsmuT. 61 The rnH construct utilized primers 1925-2 Fwd + 1917-2R Rvs, 1917-3 RnaseH Fwd + 1917-3 RnaseH Rvs, and 62 1917-4R Fwd + 1925Rvs to generate the strain 159MrnH. To create a counterselection cassette 63 containing the parE toxin, strain 159MfstS was used as a template to amplify the upstream homologous fragment 64 and Xyl-S induction cassette using the primers 1925-1 brsRM159-LF + 2019-1-Rvs. The spectinomycin 65 resistance gene aad9 and the downstream homologous fragment were similarly amplified from strain 159MfstS 66 using the primers 2019-3 Fwd + 1925-5 brsRM159-RR. The parE ORF was PCR amplified with the primers 67 2019-2 ORF5-Fwd + 2019-2 ORF5-Rvs together with lysates of S. mutans strain UA140, which naturally harbors 68 the cryptic plasmid pUA140 containing the parE addiction module. Each of these PCR amplicons was mixed and 69 assembled via OE-PCR using the primers 1925-1 brsRM159-LF + 1925-5 brsRM159-RR. The assembled 70 construct was subsequently transformed to UA159 and selected on antibiotic-supplemented agar plates. Several 71 clones of the resulting transformation were tested for negative selection in the presence of xylose. Clones of 72 interest were sequenced to confirm the expected counterselection cassette genotype. Functional 73 counterselection cassettes were found to contain one of two point mutations located either within the parE 74 ribosome binding site (RBS) or within the parE ORF, encoding a ParE S56G mutation. These two strains were 75 subsequently named 159MparE1 (ParE RBS) and 159MparE2 (ParE S56G). To compare the functionality of the 76 fst-sm-and parE-containing counterselection cassettes in other S. mutans strains, the cassettes were first PCR 77 amplified from strains 159MfstS, 159MfstT, 159MparE1, and 159MparE2 using the primers 1925-1 brsRM159-78 LF + 1925-5 brsRM159-RR. These amplicons were subsequently transformed into S. mutans wild-type strains 79 CL1, JF243, and UA140 and selected on antibiotic-supplemented agar plates. The resulting strains were named 80 CL1MfstS, CL1MfstT, JF243MfstS, JF243MfstT, 140MfstS, and 140MfstT (Fst-sm) as well as CL1MparE1, 81 CL1MparE2, JF243MparE1, JF243MparE2, 140MparE1, and 140MparE2 (ParE). 82

Generation of markerless galK nonsense point mutations in S. mutans and S. gordonii 35
To engineer a markerless point mutation into the S. mutans galK gene, we first inserted the fst-sm (A7T) 36 counterselection cassette between the galR and galK ORFs on the chromosome of S. mutans strain UA159. 37 The upstream and downstream homologous fragments were PCR amplified from UA159 using the primer pairs The preparation of H2O2 indicator plates and detection of bacterial H2O2 production were performed using a 70 previously described methodology (19). Briefly, overnight cultures of each strain were washed twice with PBS 71 and then adjusted to an optical density OD600 0.5. A volume of 10 µl was pipetted onto BHI Prussian-blue indicator 72 plates and incubated overnight at 37 °C in a 5% CO2 atmosphere. 73 74

Luciferase assays 75
Luciferase assays were performed using a previously described methodology (16,20). Briefly, strains were 76 this approach that can be problematic, mainly its requirement for a toxic substrate (i.e. 4-CP) and a narrow host 91 range for pheS-containing counterselection cassettes. Here, we aimed to improve upon these by developing a 92 new negative selection strategy to facilitate cloning-independent markerless mutagenesis (CIMM). We 93 previously developed a xylose-based gene induction system (called Xyl-S) that was shown to exhibit an 94 exceptionally wide dynamic range with low basal expression, and it also functioned similarly in several different 95 oral streptococci with no evidence of toxicity (17). In fact, we had even employed Xyl-S to engineer multiple 96 conditional lethal mutations in S. mutans (17). Thus, we reasoned that the Xyl-S system may also be appropriate 97 to control the expression of a toxic gene product to confer negative selection in S. mutans (Fig. 1). As shown in 98 Table 1, we selected several candidate chromosomal toxins from verified S. mutans toxin-antitoxin modules 99 (mazF, smuT, and fst-sm) as well as the plasmid addiction module toxin parE from the S. mutans cryptic plasmid 00 pUA140 and RNase H (rnH) from E. coli. Each of these genes was transcriptionally fused to the Xyl-S cassette 01 and transformed into S. mutans strain UA159 to test for xylose-inducible negative selection. Following 02 transformation of the constructs, we noted that most seemed to exhibit high levels of basal uninduced toxicity, 03 as some transformants grew slowly and we observed fewer transformants than expected. Regardless, we 04 selected candidate transformants from each reaction and then quantified their viability on agar plates  1% (wt 05 vol -1 ) xylose. As shown in Figure 2A, only the fst-sm and parE constructs inhibited cell growth in the presence of 06 xylose, with each exhibiting potent negative selection. We sequenced several clones of the fst-sm and parE 07 strains that exhibited inducible negative selection and found all to contain at least one point mutation. For the 08 strains harboring the fst-sm construct, they each contained either of two missense mutations within the same 09 codon of the xylose repressor gene xylR, conferring either a XylR A7S or A7T mutation (Fig. 2B). The strains 10 harboring parE constructs all contained the same XylR A7S mutation because one of the mutated fst-sm clones 11 had been employed as a PCR template during the assembly of the parE construct (Fig. 2C). However, the parE 12 strains also contained one of two additional point mutations, either within the parE ribosome binding site (RBS) 13 or within the parE ORF, conferring a ParE S56G missense mutation (Fig. 2C). While the impact of the ParE 14 S56G mutation was unclear, the parE RBS mutation almost certainly reduced the efficiency of parE translation, 15 as the original construct contained a consensus Shine-Dalgarno sequence. Interestingly, we reassembled the 16 parE construct with a wild-type xylR ORF together with either of the newly identified parE RBS or S56G mutations 17 and found that the point mutant xylR was indeed required for construct stability. Thus, it would appear that the 18 XylR A7S (and presumably A7T) mutations are responsible for reducing the basal uninduced expression of fst-19 sm and parE. The additional compensatory point mutations required for the proper functionality of the parE 20 construct suggests that S. mutans strains encoding the ParE toxin are likely to be even more sensitive to toxic 21 leaky gene expression as compared to fst-sm. 22

Cloning-independent markerless mutagenesis (CIMM) in S. mutans 24
Based upon the extreme potency of the negative selection observed from the fst-sm and parE counterselection 25 cassettes ( Fig. 2A), we were next curious to test the utility of these cassettes for CIMM. As a simple phenotypic 26 readout, we employed the fst-sm and parE cassettes to engineer markerless gusA (-glucuronidase) 27 replacements of the S. mutans brsM ORF. Under normal growth conditions, the gene product of brsM inhibits 28 the expression of the brsRM operon by preventing BrsR positive feedback autoregulation, leading to extremely 29 low basal expression in the wild-type (Fig. 3A) (18, 21, 22). Consequently, insertion of the gusA ORF at the 3' 30 end of the brsRM operon results a white colony phenotype (i.e. low basal gusA expression) because the brsRM 31 operon remains weakly expressed similar to the parent wild-type (Fig. 3A). Conversely, a gusA replacement of 32 brsM will stimulate BrsR positive feedback autoregulation of the operon, resulting in strong gusA expression and 33 a corresponding blue colony phenotype (Fig. 3A). Using an existing brsRM-gusA reporter strain, we first replaced 34 the gusA ORF downstream of brsM with the fst-sm and parE counterselection cassettes and subsequently 35 performed a second transformation to replace both brsM and the counterselection cassette with gusA. As 36 expected, the original parent brsRM-gusA reporter strain grew normally on the xylose plates with undetectable 37 -glucuronidase activity (Fig. 3B). In agreement with the strong xylose-inducible negative selection previously 38 observed with the fst-sm and parE counterselection cassettes ( Fig. 2A), we observed minimal background 39 growth from the negative control transformation reactions (Fig. 3B). In contrast, strains transformed with gusA 40 DNA to replace both brsM and the counterselection cassettes all yielded numerous dark blue colonies, further 41 confirming the efficacy of negative selection as well as the expected gusA allelic replacement of brsM (Fig. 3B). 42 Importantly, despite the presence of some detectable background growth in the negative control reactions, the 43 markerless gusA transformations all yielded few, if any, white colonies, indicating that nearly 100% of the 44 transformants exhibited the expected brsM gusA + genotype (Fig. 3B). 45 46

Comparison of the fst-sm and parE counterselection cassettes in multiple streptococci 47
As previously mentioned, one of our goals was to create a counterselection cassette exhibiting a broader host 48 range than the previous pheS-based IFDC2 system. Since the fst-sm and parE cassettes were suitable for 49 performing CIMM in S. mutans strain UA159, we next tested these same cassettes in three additional strains of 50 S. mutans to determine whether they exhibit any evidence of strain-specificity. Both of the fst-sm cassettes 51 performed quite similarly in each of the three S. mutans strains as previously observed with UA159. On xylose-52 supplemented agar plates, we detected 3 -4 log reductions in total cell number, which corresponds to 0.1% 53 background growth (Table 2). Unlike the fst-sm constructs, the results with the parE cassettes were mixed. The 54 parE RBS point mutant version only exhibited noticeable negative selection in strain CL1 (Table 2), which is a 55 serotype K clinical isolate (18). However, even in this strain, growth was partially inhibited on plates lacking 56 xylose, which indicated that the basal expression of this toxin was still negatively affecting its growth. The ParE 57 S56G mutant cassette performed better than the RBS mutant cassette, yielding xylose-inducible 3 -4 log 58 reductions in strains CL1 and JF243 with no obvious toxicity on the xylose-free plates (Table 2). Interestingly, 59 strain UA140 was completely resistant to negative selection using both of the parE cassettes (Table 2). It is not 60 yet clear why parE selection failed in UA140, but we suspect it is because this strain naturally harbors the 61 pUA140 cryptic plasmid that encodes the ParE addiction module. From these results, we conclude that both of 62 the fst-sm counterselection cassettes are likely to function well in most strains of S. mutans, while the ParE S56G 63 cassette is slightly less universal, possibly due to the presence of the pUA140 plasmid in some strains. 64

65
Since we had previously demonstrated the utility of the Xyl-S system for xylose-inducible gene expression in 66 both S. gordonii and S. sanguinis (17), we were also curious to determine whether the xylose-inducible fst-sm 67 and parE counterselection cassettes would function in these organisms. We introduced the four fst-sm and parE 68 counterselection cassettes into S. gordonii strain DL1 and S. sanguinis strain SK36, which are among the most 69 commonly utilized strains for genetic studies of both species. Only the fst-sm (A7T) cassette functioned in both 70 organisms, yielding 3-log reductions in the presence of xylose (Table 2). Surprisingly, the RBS mutant parE 71 cassette functioned even better than fst-sm (A7T) in S. sanguinis, but it completely failed to inhibit S. gordonii 72 ( Table 2). The fst-sm (A7S) and parE (S56G) cassettes did not function in either S. gordonii or S. sanguinis. 73 Overall, the negative selection results indicated that the fst-sm (A7T) cassette is likely to be the most broadly 74 useful counterselection cassette for streptococci, although the fst-sm (A7S) and the parE cassettes may also 75 function quite well for certain species or strains. To further confirm the utility of the fst-sm (A7T) counterselection 76 cassette in S. gordonii and S. sanguinis, we performed CIMM with this cassette to markerlessly insert the green 77 renilla luciferase ORF (renG) immediately downstream of the pyruvate oxidase-encoding gene spxB in both 78 species (Fig. 4A). Since the pyruvate oxidase enzyme is responsible for generating the majority of the H2O2 79 excreted by these two streptococci (19, 23), the spxBphenotype is readily observable on Prussian blue plates, 80 as this dye reacts with H2O2 to produce a blue precipitate. As shown in figure 4B, both S. gordonii and S. 81 sanguinis produced much less blue precipitate when spxB was replaced with the fst-sm (A7T) counterselection 82 cassette. However, wild-type levels of H2O2 were once again detectable when spxB was markerlessly inserted 83 back to its original locus together with a downstream renG ORF (Fig. 4B). We also measured luciferase activity 84 from both of the markerless luciferase reporter strains and observed ~4-log increased reporter activity over the 85 background, unlike the parent spxB deletion strains, which lacked the renG ORF (Fig. 3C). Thus, we conclude 86 that the fst-sm (A7T) counterselection cassette (henceforth referred to as IFDC4) is suitable for both markerless 87 deletions and insertions in S. gordonii and S. sanguinis. 88 89

Markerless point mutagenesis 90
As previously described, one potential advantage of counterselection-based markerless mutagenesis is that it 91 supports the creation of targeted point mutations due to scarless excision of the counterselection cassettes. 92 Therefore, as a final confirmation of IFDC4 utility, we designed CIMM constructs to create nonsense point 93 mutations within the galK genes of S. mutans, S. gordonii, and S. sanguinis. We chose to mutate galK because 94 this gene is both nonessential in all three species and should confer resistance to the toxic effects of the 95 galactose analog deoxygalactose, yielding an easily observable growth phenotype. As shown in figure 5A, 96 deoxygalactose resistance was indeed created after markerlessly introducing an ochre (CAA → TAA) nonsense 97 point mutation into codon 8 of the S. mutans UA159 galK gene. We repeated the same experiment using S. 98 gordonii DL1 and were able to successfully introduce a similar ochre (CAA → TAA) nonsense mutation into 99 codon 9 of its galK gene (Fig. 5B). However, we were surprised to discover that wild-type DL1 is naturally 00 resistant to the toxic effects of deoxygalactose. Therefore, unlike S. mutans, we were unable to observe any 01 difference in growth between the wild-type and galK point mutant strains. For unknown reasons, we could not 02 achieve any discernable negative selection when trying to mutate the galK gene of S. sanguinis SK36, despite 03 the efficacy of negative selection with our other SK36 mutant constructs using the same IFDC4 cassette (Table  04 2 and Fig. 4B -C). We even repeated the experiment using the ParE RBS mutant counterselection cassette, 05 which exhibited even stronger negative selection in SK36 compared to IFDC4 (Table 2), yet we still observed 06 the same problem. This issue appeared unique to the galK locus, as negative selection was quite stringent for 07 our other SK36 constructs (Table 2 and Fig. 4B -C), and we have also recently used IFDC4 to engineer 08 additional markerless mutations in SK36 for our other ongoing research. It is unclear why the SK36 galK locus 09 is specifically problematic, but based upon our results, it would appear that this mutation affects the utility of 10 xylose induction. However, additional studies would be required to determine whether this is indeed the cause. 11 Regardless, our success with both S. mutans and S. gordonii suggests that point mutagenesis would be unlikely 12 to fail in most instances in SK36. 13

14
Discussion 15 In the current study, we describe a new approach to perform CIMM in streptococci. This system was developed 16 to address a couple of the limitations encountered when using the previous pheS-based IFDC2 counterselection 17 cassette, mainly its requirement for a toxic substrate during negative selection (i.e. 4-CP) and its narrow host 18 range. Here, we employed the Xyl-S induction system to control the expression of a toxic gene product as a 19 negative selection mechanism. We have previously used the Xyl-S system to introduce controllable gene 20 expression in a number of in vitro and in vivo studies (16, 17), so this expression system has proven utility in 21 multiple streptococci with no evidence of xylose toxicity. Furthermore, our previous studies had already 22 suggested that the Xyl-S system would be suitable for engineering conditional lethality (17), which we repurposed 23 here as a negative selection mechanism for CIMM. As shown in Table 2, negative selection was highly efficient 24 with exceptionally low background, resulting in an extremely high percentage of markerless mutants with the 25 expected genotypes (Fig. 3B). Unlike the IFDC2 cassette, IFDC4 also allowed us to employ a single cassette to 26 engineer a variety of different types of markerless mutations in S. mutans, S. gordonii, and S. sanguinis, thus 27 confirming its broader host range. For unknown reasons, we were unable to induce negative selection when 28 mutating the galK gene of S. sanguinis. This problem seemed specific to the S. sanguinis galK locus, as we did 29 not encounter similar issues with either of the S. mutans or S. gordonii galK mutations (Fig. 5A -B), nor did we 30 experience issues with negative selection in other S. sanguinis loci (Table 2 and Fig. 4B -C). We tried a variety 31 of different construct designs to create the galK mutant, but all failed to exhibit any evidence of xylose-inducible 32 negative selection. Therefore, we suspect that xylose is either modified or exhibits defective transport following 33 mutagenesis of the S. sanguinis galK locus. 34

35
Our results provide a general strategy for negative selection that should be adaptable for use in other organisms. 36 Most bacteria encode endogenous toxin-antitoxin modules on their chromosomes, while many others also have 37 particular strains that naturally host cryptic plasmids, which are highly likely to contain addiction modules (24, 38 25). Thus, there is a strong reservoir of potential toxins that could be exploited for use in most bacteria. This 39 approach also requires a reliable regulated gene expression system, which may be a limiting factor for certain 40 organisms, especially those with minimal genetic tools available. From our experience, successful toxin-based 41 negative selection requires both low basal uninduced expression of the toxin gene as well as strong inducibility 42 (i.e. wide dynamic range of expression). However, the specific expression characteristics required to create an 43 efficacious inducible negative selection system are likely to vary widely. Thus, the challenge is to find an 44 appropriate match between the available expression system(s) and the toxicity of a particular toxic gene product. 45 For example, our results suggest that ParE is likely to be a more potent toxin in S. mutans compared to 46 as fst-sm could be stably expressed from the Xyl-S expression system after acquiring a single point mutation 47 within the xylR gene, whereas the parE constructs required the same xylR point mutation as well as additional 48 compensatory parE mutations (Fig. 2C). Thus, when developing a new toxin-based negative selection system, 49 it is advisable to screen a variety of candidate toxins to increase the likelihood of identifying the appropriate gene 50 to pair with an available expression system. In our case, none of the toxin genes we selected were immediately 51 ideal for our Xyl-S expression system. However, the fst-sm construct was apparently a sufficiently close match 52 that it allowed us to isolate spontaneous compensatory xylose repressor mutations, resulting in a mutant 53 counterselection cassette with lower basal uninduced toxicity, but stringent xylose-inducible negative selection. 54 It seems unlikely that we would have identified the appropriate compensatory parE mutations if the strain did not 55 already contain the XylR A7S mutation within the Xyl-S induction system, which apparently reduced its leakiness. 56 This may also explain why we did not isolate spontaneous compensatory mutations for the other toxins in our 57 initial screen of candidates, as these were constructed without the XylR A7S or A7T mutations. Presumably, the 58 mutants we did isolate with these other toxins all contained inactivating mutations because none of the isolates 59 exhibited any evidence of xylose-inducible negative selection ( Fig. 2A). Since we had already observed effective 60 negative selection from the fst-sm and parE constructs, we did not perform additional tests to determine whether 61 the other toxin genes might function in combination with the XylR A7S mutation. However, it is certainly 62 conceivable that at least some of them would, perhaps requiring additional compensatory mutations similar to 63 the parE construct. As our results demonstrate, it is not essential for a given gene induction system to perfectly 64 pair with a toxin gene, provided the initial basal uninduced toxicity of the construct is moderate enough that it 65 gives the cell a chance to develop the appropriate compensatory mutations. From there, one can screen the 66 isolates to identify those that may be suitable for stable negative selection.   and exogenous toxic ORFs were transcriptionally fused to the Xyl-S cassette and then transformed into S. 29 mutans strain UA159. The resulting strains were tested for growth on agar plates  xylose. B) Several clones 30 harboring the fst-sm counterselection cassette were sequenced and found to contain one of two different 31 missense mutations within the xylR ORF. Both mutations targeted codon 7 of xylR, resulting in either A7S or 32 A7T substitutions in XylR. C) Several clones harboring the parE counterselection cassette were sequenced and 33 found to contain one of two different point mutations in addition to the same XylR A7S mutation found in some 34 of the fst-sm counterselection cassettes.  gusA strain that was used as a template during construction of the markerless mutant strains. Since this strain 54 has a wild-type brsM ORF, expression of the brsRM operon remains in its basal state due to the inhibitory 55 function of BrsM towards the operon activator protein BrsR. Consequently, this strain will not exhibit detectable 56 -glucuronidase activity due to low gusA expression. The middle panel shows the genotype of strains 57 transformed with the different counterselection cassettes (+/-), replacing the gusA ORF from the parent brsRM-58 gusA reporter strain. The bottom panel shows the genotype of the expected markerless reporter strains created 59 by replacing both brsM and the different counterselection cassettes with the gusA ORF. After deleting brsM, 60 inhibition of BrsR is relieved, resulting in potent autoactivation of the operon promoter and high levels of gusA 61 expression. The markerless brsM brsR-gusA reporter strains will exhibit a dark blue colony phenotype due to 62 the large amount of -glucuronidase activity produced from the reporters.  Prussian blue agar plates due to the H2O2 produced primarily from the spxB-encoded enzyme pyruvate oxidase. 78 The middle panel shows an allelic replacement of spxB with the counterselection cassettes (+/-), resulting in a 79 major reduction in H2O2 production. A subsequent transformation of these strains with spxB-renG DNA replaces 80 IFDC4 and restores spxB to its original locus along with a transcriptional fusion to the renG ORF. The markerless 81 spxB-renG reporter strains should produce similar levels of H2O2 as the original wild-type strains. B) 82 Representative strains of the wild-type (WT), spxB mutant (spxB), and spxB-renG reporter (spxB + renG + ) of 83 both S. gordonii (DL1) and S. sanguinis (SK36) were spotted onto Prussian blue agar plates to observe the H2O2 84 production phenotypes of each. C) Luciferase activity was compared between the spxB strains harboring the 85 counterselection cassettes (IFDC4) vs. the markerless spxB-renG reporter strains (spxB-renG). Results from the 86 S. gordonii strains are shown in red, while S. sanguinis results are shown in blue. Luciferase data are presented 87 relative to the cell-free background luciferase values, which were arbitrarily assigned a value of 1. Luciferase 88 data are derived from five independent clones of each strain, which were averaged and presented together with 89 their corresponding standard deviations .  90  91  92  93  94  95  96  97  98  99  00  01  02  03  04  05  06  07  08  09  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39 40 41