ABSTRACT

Neisseria gonorrhoeae engages in extensive intra-cellular gene conversion between the PilE-expression locus (pilE) and the transcriptionally-silent pil gene copies (pilS). In silico analyses were applied to investigate the extent of sequence heterogeneity between the various pilS gene copies. Analysis of synonymous and non-synonymous substitutions between the different pilS genes indicated that relatively few amino acid changes would occur due to nucleotide polymorphisms towards the 5’ end of the pilS genes whereas more frequent amino acid substitutions would be incorporated within the “hypervariable” region. The lack of non-synonymous substitutions at the 5’ end of the genes was found to be under selective pressure as indicated by a positive D_T score utilizing the Tajima test. The presence or absence of mismatch repair appeared to only impact recombination when non-identical DNAs recombined via the DNA transformation route, where small pil sequence heterogeneities were sufficient to terminate recombination tracts, with these sequence constraints being relieved in cells carrying a mutS mutation. Therefore, the data indicate that the effect of sequence heterogeneity on recombination within the pil system appears to depend upon the context with which the non-identical DNAs recombine.

Keywords:Recombination, Antigenic Variation, Pilin, Sequence Heterogeneity

1. Introduction

Neisseria gonorrhoeae (the gonococcus) causes the sexually transmitted disease gonorrhea, which is generally a non-complicated mucosal infection of humans that is characterized by a massive neutrophil infiltration. Though treatable by antibiotic therapies, vaccine development has been hampered due to the inability of the host to mount an effective immune response, in part, due to extreme antigenic and phase variation of several different gonococcal surface components (e.g., PilE polypeptide, opa gene expression, LOS variation; reviewed [1] ).

The major protein subunit of the pilus organelle, PilE polypeptide, is encoded by the pilE gene, and is capable undergoing antigenic variation in vivo [2] [3] . Besides pilE, the chromosome also contains other variable pil gene sequences in various silent loci (pilS) [4] [5] . pilS loci contain multiple pil gene copies arranged in a tandem array. The variable pilS gene copies differ from the pilE gene in that they are believed to be non-transcribed due to a lack of promoter elements, as well as lacking the conserved 5’ 150 bp found in all pilE genes. Despite these differences, all pil genes have common characteristics with variable segments being interspersed with small constant regions [6] . As the gonococcal chromosome contains multiple pil gene sequences (the number varies depending upon the strain), considerable sequence heterogeneity exists within the variable gene segments. Consequently, pilE/pilS recombination is a classic example of homeologous recombination.

PilE antigenic variants are produced following pilE recombination with a pilS gene copy (schematic model is presented in Figure 1). This intra-cellular event occurs as a RecA-dependent, non-reciprocal gene conversion with information being transferred uni-directionally from pilS to pilE with loss of the corresponding pilE gene sequence during the gene conversion event [4] [7] -[10] . Recently, a unique DNA structure has been identified at the pilE locus that has been proposed to initiate pilE/pilS recombination [11] . The formation of this guanine quartet structure apparently allows a nick to be introduced which is then acted upon by RecJ exonuclease to yield single-stranded DNAs that can then recombine with a pilS locus to yield pilE gene variants.

The release of the N. gonorrhoeae FA1090 genome sequence (GenBank accession number AE004969), followed by the release of numerous other Neisseria genome sequences, identified many of the recombination and repair proteins that have been described for Escherichia coli [12] . In particular, the RecA, RecB, RecC, and RecD proteins as well as a partial complement of the RecF pathway components. Gonococci also contain partial complements of other repair systems described for E. coli. They appear to utilize mismatch repair as they possess MutS and MutL homologues, yet lack the mutH and the dam methylase genes [13] [14] . Consequently, mismatch repair of single base mismatches or short insertions or deletions in the Neisseriae does not appear to be methyl-directed.

In this study, we examine, in silico, the extent of pil sequence heterogeneity within the pilS gene copies and determine its effect on the pil recombination system. The analysis demonstrates considerable sequence heterogeneity between the pilS gene copies, yet apparent selective pressure constrains 5’ nucleotide polymorphisms within pilS that would allow only synonymous amino acid changes to occur within this region of pilE following a pilE/pilS recombination events. In contrast, nucleotide substitutions leading to non-synonymous amino acid changes are predicted to occur towards the 3’ end of the gene within the so-called “hyper-variable” region. Previously, we had shown that the presence or absence of an active mismatch repair system did not seemingly affect the efficiency of intra-cellular pilE/pilS recombination [15] . However, it appears that short sequence nonidentities between variant pil segments impede DNA transformation-mediated recombination in rec+ bacteria, with such constraints being alleviated through inactivation of the mismatch repair system. Thus, the data indicate that the manner by which pil heteroduplexes are created dictates how the cell deals with recombination between non-identical pil DNAs.

2. Materials and Methods

2.1. Strains and Growth Conditions

Neisseria gonorrhoeae strain MS11 was used for all aspects of this study. Gonococci were grown on gonococcal typing medium [16] at 37˚C in a 5% CO₂ atmosphere. When grown in liquid culture 420 ng/ml NaHCO₃ was added to the culture medium and the agar was omitted. Where appropriate, antibiotics were added at the following concentrations; chloramphenicol 10 μg/ml, kanamycin 80 μg/ml, and erythromycin 5 μg/ml. The various pilE variants and mutants have previously been described [15] [17] [18] . Colony morphology variants were assessed using a phase contrast microscope as previously described [2] [8] . The recB growth suppressor mutant arose on plates and has previously been described [19] .

All cloning was performed with Escherichia coli DH 5α using standard protocols. Transformants were selected on plates containing antibiotics at the following concentrations; ampicillin, 100 μg/ml, erythromycin, 200 μg/ml, kanamycin 80 μg/ml, and chloramphenicol, 20 μg/ml.

2.2. DNA Manipulations

The donor DNA (pCLPX-4) used for DNA transformation was constructed as follows; a SmaI/EcoRI fragment from pLP61 [18] carrying the pilE gene was cloned into pUC19. A promoter-less cat gene was then blunt-end cloned into the unique Bsu36 site immediately downstream of the pilE promoter. Consequently, transcription of the cat gene is under the control of the pilE promoter. For transformation experiments with donor DNAs containing 5’ and 3’ pilE flanking homologies plasmid pNG3005 was used, which contains a drug resistance marker (ermC) within opaE located immediately downstream of the pilE locus [20] .

2.3. Transformation Protocols

Gonococci were lifted from plates on sterile Whatman paper fragments and the cells were re-suspended in 1 ml liquid growth medium containing 2 mM IPTG. The cells were then pre-incubated for 20 min to express RecA protein [21] . After 20 min pre-incubation, 1 μg of donor DNA was added to the culture and the cells were incubated for 3 hrs in a 5% CO₂ atmosphere. The cells were then pelleted by centrifugation and were plated on selective medium in the absence of IPTG. Following overnight growth, the individual drug resistant colonies were passed for an additional day on selective medium without IPTG before the total DNA was prepared. The pilE gene was then PCR amplified for DNA sequencing using the following primers: 5’-TCCCCTTTCAATTAGGAGT-3’ and 5’-CCGATATATTATTTCCACC-3’.

Transformation frequencies were determined using the previously described single-colony transformation procedure [22] . The transformation frequencies were corrected for the initial piliation status of the recipent population as pilus minus bacteria are non-transformable. For transformation studies with mutS mutants, such corrections had a significant impact. Statistical analysis utilized Student’s t test and reflect deviations from the standard mean. All other transformation experiments utilized the previously described plate-transformation protocol [23] .

2.4. DNA Sequencing

DNA sequencing was performed either on an ABI DNA sequencer or using the manual dideoxy sequencing technique with the fragments being resolved on 8% polyacrylamide sequencing gels.

2.5. In Silico Analyses

Sequences were obtained from the NCBI database and aligned using the multiple sequence program MAFFT [24] . Thirteen pilS gene copies, that varied in size from 144 - 420 bp (mean = 340 bp), allowed a nucleotide alignment measuring 480 nucleotides in length (Supplemental Table 1). The alignment incorporated 366 gaps (due largely to the relatively small size of pilS6 copy 2 (144 bp) and pilS1 copy 6 (216 bp)) along with a few, short (~2 - 18 bp) regions that could not be aligned between all base pairs. This alignment was used to identify the sequence polymorphisms within N. gonorrhoeae MS11 pilS gene copies. A sequence logo was created from the constructed alignment in order to visualize the nucleotide divergence within the pilS alleles [25] . In order to determine the number of parsimony informative (each variant is present in at least two alleles) and singleton (variant appears in only one sequence), the DNASP program was used [26] . Additionally, a script was created to identify singleton nucleotide divergences within the alignment. The PAML program was then used to find the average number of synonymous polymorphisms per synonymous site (Ks) and the average number of non-  synonymous polymorphisms per non-synonymous site (Ka) [27] . DNASP was also used to perform the Tajima test in order to determine if non-neutral evolution is evident in any of the pilS alleles. The Poison Random Field model was used to detect if the nucleotide polymorphisms identified within the alignment were evolving under neutral expectations. To achieve this, the PRFMLE program was used to analyse the pilS alleles [28] .

3. Results

3.1. In silico Analysis of Pil Sequence Heterogeneity

Previous analysis of pilE sequence divergence had revealed the presence of semiand hyper-variant gene segments within the pilE gene [4] -[6] . In order to determine the total extent of pil gene sequence heterogeneity, the pilS gene copies were examined using various in silico tools. Thirteen pilS gene copies were aligned (Table 1; supplemental data) and used to determine the nucleotide polymorphisms as well as the intensity and direction of selection.

In order to visualize the regions containing prevalent nucleotide polymorphisms, the multiple sequence alignment of the 13 pilS copies was visualized as a sequence logo (Figure 2) [25] . Taller nucleotide stacks within the sequence logo denotes greater conservation, thus providing a visual represention of the relative frequency of the nucleotides in that position; areas with missing nucleotides correspond to regions where all, or most, sequences failed to align (i.e., alignment gaps). As evident in Figure 2, pilS genes contain constant and

Summary of the descriptive statistics depicting the degree of polymorphism within each pilS allele. Informative sites are parsimony informative, where each variant is present in at least two alleles. Singleton sites are where the variant occurs in only one sequence among the set [26] [27] . constant represents the average number of synonymous polymorphisms per synonymous site (i.e., those that would code for the same amino acid). represents the average number of non-synonymous polymorphisms per non-synonymous site (i.e., those that would code for a different amino acid). Informative sites are parsimony informative, where each variant is present in at least two alleles. Singleton sites are where the variant occurs in only one sequence among the set.

variable regions, with the the 5’ end to the mid-gene region being relatively constant; the 3’ region, however, contains an elevated degree of sequence diversity, especially between base pairs 320 - 385. As such, this initial analysis conforms to the notion of variable “mini-cassettes” interspersed with constant regions and reveals the so-called “hyper-variable” region towards the 3’ end of the pil gene copies. This analysis also reveals several short stretches of non-identity towards the 5’ end of the pilS gene copies.

To better envision the single nucleotide polymorphisms identified with the sequence logo algorithm, singleton base changes were extracted. A singleton base change is one that occurs in only a single sequence in the alignment. If the base change from the consensus sequence occurs in two or more pilS sequences, it is not considered a singleton base change and has not been included in Figure 3. Every pilS gene copy shows a single nucleotide polymorphism within the alignment except pilS1 copy 5. Therefore, this particular pil gene copy reflects the consensus sequence which suggests that pilS1 copy 5 shows similarity among the entire length of its sequence to the majority of the other pilS sequences. Most of the singleton changes occur between bases 320 - 385 (highlighted in grey) which is also observed within the sequence logo and corresponds to changes within the so-called “hyper-variable” region. Also evident in Figure 3, the 5’ region also contains a number of singleton changes.

The pilS alignment was then analyzed to determine the degree of polymorphism between and within each pilS locus that contained more than two pil gene copies (intra-genic polymorphisms are excluded for pilS5 and pilS7). As can be seen in Table 1, the number of nucleotide changes that occur within two or more pilS sequences (referred to as parsimony-informative sites and designated as Informative) is indicated, along with the number of polymorphisms that are found only within a single pilS sequence (designated as Singleton); sites containing alignment gaps were excluded to ensure correct analysis (the number of bases analyzed are designated as No. Sites). Both the parsimony-informative and singleton statistics indicate the number of different nucleotides found at each site (2, 3 or 4). The constant K_s represents the average number of synonymous polymorphisms per synonymous site (i.e., those that would code for the same amino acid), whereas the constant K_a represents the average number of non-synonymous polymorphisms per non-synonymous site (i.e., those that would code for a different amino acid). Consequently, the more sequences within the alignment that contain a polymorphism at a designated synonymous or non-synonymous site will increase the K_s or K_a ratio, allowing the ratio of these two constants (K_a/K_s) to be used as an indicator of selection. A high ratio is indicative of positive selection. The data show that between all the pilS gene copies there is a large number of parsimony-informative sites, accounting for approximately 30% of all sites analyzed (2 = ~19%, 3 = ~10% and 4 = ~ 2%). When all 13 pilS gene copies were analyzed the K_a value was similar to the K_s value along the entire length of the alignment, indicating a lack of non-synonymous substitutions, which would account for relatively small changes in the resulting amino acid sequence upon unidirectional recombination with pilE. However, as pilS1 copy 6 and pilS6 copy 2 are much shorter than the other pilS copies and introduce large gaps at the 5’ end of the alignment, analyses performed on the sequence alignment excluding these two pilS copies shows that there is a disparity between the amount of synonymous and non-synonymous polymorphisms, with the non-synonymous polymorphisms being more fre-

quent. This would result in a relatively high degree of sequence diversity in the resulting amino acid sequence upon unidirectional recombination with pilE (Table 1).

In order to visualize the number of synonymous and non-synonymous polymorphisms at different positions within the pilS genes, a sliding window approach was used to graphically represent the lack of uniformity. This was performed separately for synonymous and non-synonymous sites on all pilS alleles excluding pilS1 copy 6 and pilS6 copy 2. As the number of non-synonymous and synonymous sites is comparatively constant among the entire set of pilS genes, the same window sizes and sliding increments were used for both analyses. Interestingly, the 5’ region of pilS showed a relatively high degree of synonymous substitutions when compared to the 3’ end of the pilS alleles, with non-synonymous polymorphisms being more focused towards the 3’ end of the genes (cf. Figure 4(a) and Figure 4(b)). This indicates that the sequence divergence seen with both the sequence logo (Figure 2) and the singleton analysis (Figure 3) towards the 5’ end are enriched for synonymous base changes, while the sequences within the “hyper-variable” region (starting around base pair 310) contained a higher degree of non-synonymous substitutions.

In order to identify whether selection pressure was being applied near the 5’ end of pilS sequences (with pilS1 copy 6 and pilS6 copy2 being excluded), the Tajima test was used to determine whether the nucleotide polymorphisms occurring within the pilS alignment are consistent with neutral expectations [26] . This test uses Tajima’s D_T statistic, which is a scaled version of the ratio of the heterozygosity (k) of the population over the relative number of segregating sites based on a constant (S/a₁). Purifying selection is indicated if Tajima’s D_T statistic is significantly less than zero, while a D_T value greater than zero is indicative of balancing selection. As seen in Table 2, the values of Tajima’s D_T revealed evidence for balancing selection in the 5’ region of the analyzed pilS alignment, while purifying selection appears to have taken place in the center of the so called “hypervariable” region.

The Poisson Random Field model was also used to test if the polymorphisms are evolving under neutral expectations [28] . This test determines if polymorphisms occur under neutral expectations, at an estimated rate of µ > 0 with a selective advantage of γ > 0, with a null hypothesis of γ = 0. When using this test on the pilS alleles, γ can be used to estimate the selective advantage and the nature of the selection pressure can be determined by using the sign (+ or −) of this value. When the Poisson Random Field model test was applied to the same pilS alignment excluding pilS1 copy 6 and pilS6 copy2, the analysis corroborated the Tajima test in that balancing

Regions showing significant D_T values on the pilS alignment excluding pilS1 copy 6 and pilS6 copy 2 with a sliding window of 5 bp moved in 1 bp increments [26] .

selection was indicated between base pairs 1 - 300 (Table 3). As there was only a short region undergoing purifying selection within the so-called “hyper-variable” region, this analysis was unable to detect any selection from base pairs 300 - 428 (not indicated in Table 3).

Overall, the above analyses were able to identify the constant and variable regions within the pilS gene copies; confirm that considerable sequence variation is present within the pilS gene copies; depending upon the position of the nucleotide polymorphism within the pilS gene copy, recombination with pilE may or may not lead to amino acid changes within the variant PilE polypeptides; and, the data also show that distinct regions within pilS are also evolving independently.

3.2. Pil Sequence Heterogeneity and the Role of the Mismatch Repair System

Given the variable nature of pil gene sequences, pilE/pilS recombination is a classic example of homeologous recombination, with pil recombination requiring the pairing and exchange of non-identical DNAs. An RNAbased assay was previously used to assess whether mismatch repair influences intra-cellular recombination between pilE and pilS in N. gonorrhoeae strain MS11 and showed that comparisons between isogenic mutS+/MutS bacteria, as well as isogenic recB_supmutS+/recB_supmutS bacteria, equivalent signal intensities were observed, suggesting that the presence of an active mismatch repair does not appear to influence the efficiency of intracellular pilE/pilS recombination in strain MS11 [15] . Therefore, the mechanism for pilE/pilS recombination

The Poisson Random Field model was applied to detect if the polymorphisms within pilS are evolving under neutral expectations. This test was applied to the entire alignment of pilS along with each gene copy in the different loci. The Poisson Random Field model was able to detect balancing selection pressure on the polymorphisms within bp 1 - 300 of the pilS alignment when pilS1 copy 6 and pilS6 copy 2 were excluded [28] .

appears to be “blind” to the inherent sequence non-identities.

We next examined whether pil sequence heterogeneity affected recombination when heteroduplex DNAs pair and recombine via the DNA transformation route. Transformation of recipients containing an IPTG-regulatable recA gene [21] with a defined pilE::cat construct allowed defined pil/pil recombinations to be established, with recombinant pilE genes being “locked-in” by plating transformant populations on plates lacking IPTG. DNA sequencing of pilE from chloramphenicol-resistant transformants allowed us to assess the extent of incorporation of the variant pilE gene sequence carried on the donor plasmid DNA into the recipient pilE gene. Following transformation of recipients expressing the pilE variant 7:30:2 allele with HincII-digested pCLPX-4 donor DNA, 20/21 recombinant pilEs that were sequenced (from several different transformation experiments) showed that incorporation of the plasmid donor DNA into the recipient pilE gene occurred until a 9 bp sequence non-identity was encountered between the two DNA molecules (Figure 5). The single exception was found to simply crossin the drug resistance marker using regions of sequence identity that surrounds the promoter region. Therefore, the short sequence non-identities that were identified towards the 5’ end of the sequence alignment logo (Figure 2) may be sufficient to impede recombination when non-identical pil DNAs recombine via the DNA transformation route.

In contrast to the observations presented above, when the same donor DNA was used to transform cells carrying a mutS mutation, in addition to the inducible-recA gene, two different types of pilE recombinant were recovered. These included those that simply crossed in the drug resistance marker and maintained the recipient pilE sequence and those that crossed in the marker as well as incorporating long tracts of the plasmid donor DNA, even across regions that contained multiple DNA sequence non-identities between the recombining DNAs (data not shown). Together, these observations suggest that with wild type bacteria, when non-identical pil DNAs recombine via the DNA transformation route, the recipient mismatch repair system will terminate the extent of recombination tracts at small regions of non-identity, whereas in the absence of mismatch repair homology constraints appear to be removed.

3.3. Sequence Heterogeneity, Mismatch Repair And Transformation Efficiency

In the absence of RecA protein (using the IPTG-regulatable strain), a N. gonorrhoeae strain MS11 mutS mutation elevated the phase transition rates within non-selected recipient populations (Table 4); in a rec+/mutS comparison, a 18.75X increase was observed, p < 0.05, n = 10; in a recB growth suppressor/recB growth suppressor mutS comparison, a 10X increase was observed, p < 0.1, n = 10. Although we did not explore this phenomenon further, pilus phase transitions in the absence of RecA are believed to be caused by frameshifting within the pilC locus [29] .

We next examined whether heterologous pilE/pilE transformations are also more efficient in mutS mutants when the donor DNA carries a pilE gene with 5’ and 3’ pilE flanking sequences (pNG3005; [20] ). The data presented in Table 4 show that in a rec+/mutS comparison, there was a 15.6X increase in transformation efficiency with the mutS recipients (2.73 × 10⁻³ vs 42.6 × 10⁻³ for rec+ and mutS respectively; p < 0.05, n = 10). Similarly, a 5.75X increase in transformation efficiency was observed in recB growth suppressor mutants that also carried a mutS mutation when compared to the recB suppressor parental strain (7.46 × 10⁻³ vs 42.9 × 10⁻³ for recB sup and recB sup mutS respectively; p < 0.2, n = 10). pilE sequence analysis of the mutS transformant populations also showed that pilE allelic exchange occurred in approximately 50% of those pilE genes that were sequenced (n = 20). Moreover, the sequencing data also indicated that intact pilE genes were exchanged without the incorporation of errors within the mutS transformant population. Despite this later observation indicating that transformant populations should be predominantly piliated, when these populations were scored microscopically for

^(a)frequency = # transformants/μg donor DNA/cfu (×10⁻³); ^(b)the colony phenotype (the percentage of piliated cells) of the recipient population was determined on cells plated at time 0 on plates without IPTG being present; the colony phenotype of the transformant population was determined on cells plated on selective medium. n = 10.

their piliation status, wild type rec+ bacteria showed a higher percentage of colonies that retained the piliated phenotype when compared to the isogenic mutS strain (32% vs 20%; Table 4).

4. Discussion

In this study, we examined the extent of pil sequence heterogeneity within N. gonorrhoeae strain MS11 and determined the impact of sequence divergence on various recombination profiles. In silico analysis revealed considerable sequence heterogeneity with non-synonymous nucleotide substitutions being primarily confined to the so-called “hyper-variable” region. Nonetheless, sequence heterogeneity was observed towards the 5’ end of the pilS gene copies, however, they were predominantly synonymous substitutions which would conserve the primary amino acid sequence of the variant PilE polypeptides that are created during a pilE/pilS recombination event. Two statistical analyzes indicated that positive selective pressure is being applied on the 5’ nucleotide polymorphisms in order to maintain synonymous substitutions. Consequently, the 5’ and 3’ ends of the pilS gene copies appear to be evolving independently. Despite the considerable observed sequence heterogeneity, and that pilE/pilS recombination involves recombination between these non-identical DNAs, we had previously found the presence or absence of an active mismatch repair system appeared to have little effect on the efficiency of intra-cellular gene conversion [15] . However, we did find that short sequence non-identities impeded incorporation of divergent donor DNA when delivered by the DNA transformation route, and that inactivating the mismatch repair system allowed longer tracts of divergent DNA to be incorporated during the transformation process. Consequently, the molecular mechanisms by which non-identical pil DNAs interact appears to dictate the recombination profile. Therefore, given the considerable sequence divergence within the pil system, this demands that any model for gene conversion be able to accommodate sequence heterogeneity.

A previous study on a different strain has also examined the effects of mismatch repair on pilE gene conversion and concluded that longer pilS gene tracts were incorporated into pilE recombinants in the absence of mismatch repair [29] , which was also observed in our heteroallelic transformation experiments when MS11 mutS recipients were used. In the previous study, a kinetic assay, that scores the number of pilus minus outgrowths from a pilus plus colony over time, also indicated increased levels of antigenic variation in mutS mutants [29] . In contrast, we found little effect of a mutS mutation on the efficiency of pilE/pilS using a RNA based intra-cellular assay [15] . A possible explanation for the discrepancies between the two studies could be that frameshifting within pilC, which is elevated with mismatch repair mutants [29] , could lead to more pilus minus outgrowths in the kinetic assay, which are then scored as antigenic variants. Differences between the two studies were also observed on the impact of mismatch repair on transformation efficiency where we found that inactivation of mismatch repair caused elevated transformation frequencies. This difference could be due to, 1) our correction for the percentage of pilus minus bacteria in the recipient population as non-piliated gonococci do not engage in transformation; 2) our use of single colony transformations (each colony had between 10⁵ - 10⁶ cells), as a previous study had demonstrated that heavy recipient cell densities negatively impacted transformation frequencies [22] ; and 3) differences in the donor DNA concentrations that were used in the two studies (2 ng vs 1 μg) [29] ; this study].

The observation that inactivating the mismatch repair system can relieve sequence constraints on incorporation of non-identical DNAs may have some implications with regard to recombination that occurs during an infection. Mutator clones frequently arise through the accumulation of mutations within the mismatch repair system genes in an in vivo setting and have been proposed to drive adaptive mutation in bacteria [30] . Indeed, mutator clones apparently account for the emergence of epidemic N. meningitidis strains, which may facilitate quicker responses to changing environmental conditions [31] [32] . Therefore, the emergence of Neisseria mutator clones in vivo may allow some relaxation of the sequence constraints imposed by mismatch repair following horizontal transmission of chromosomal DNA via DNA transformation and could possibly expand the repertoire of pil genetic diversity. However, as a mismatch repair mutation significantly influenced pilus phase transition rates, the emergence of mutator clones in vivo is more likely to facilitate niche adaptation allowing an expansion of the disease profile (e.g., promote an invasive phenotype which is enhanced with pilus minus bacteria).

Acknowledgements

This work was supported in part by NIH grant 1R15 AI072720-01A1 to SAH.

References

NOTES

^*Corresponding author.