Mismatch repair and the regulation of phase variation in Neisseria meningitidis



Neisseria meningitidis controls the expression of several genes involved in host adaptation by a process known as phase variation. The phase variation frequency of haemoglobin (Hb) receptors among clinical isolates of serogroups A, B and C differed drastically, ranging from ≈ 10−6 to 10−2 cfu−1. Frequencies of phase variation are a genetic trait of a particular strain, as two unlinked Hb receptors, hpuAB and hmbR, phase varied with similar frequencies within a given isolate. Based on these frequencies, six Neisserial clinical isolates could be grouped into three distinct classes; slow, medium and fast. An increase in phase variation frequency was accompanied by high rates of spontaneous mutation to rifampicin and nalidixic acid resistance in one medium and one fast strain. The remaining three medium strains displayed elevated levels of phase variation without increases in overall mutability, as they possessed low rates of spontaneous mutation to drug resistance. The mismatch repair system of N. meningitidis was found to play an important role in determining the overall mutability of the clinical isolates. Inactivation of mismatch repair in any strain, regardless of its original phenotype, increased mutability to a level seen in the fast strain. Insertional inactivation of mutS and mutL in the slow strain led to 500- and 250-fold increases in hmbR switching frequency respectively. Concurrently, the frequency of spontaneous point mutations of mutS and mutL mutants from the slow strain was increased 20- to 30-fold to the level seen in the high strain. The status of Dam methylation did not correlate with either the phase variation frequency of Hb receptors or the general mutability of Neisserial strains. Analysis of an expanded set of isolates identified defects in mismatch repair as the genetic basis for strains displaying both the fast Hb switching and high mutation rate phenotypes. In conclusion, elevated frequencies of phase variation were accompanied by increased overall mutability in some N. meningitidis isolates including strains shown to be mismatch repair defective. Other isolates have evolved mechanisms that seem to affect only the switching frequency of phase-variable genes without an accompanied increased accumulation of spontaneous mutations.


Neisseria meningitidis continues to be a major cause of endemic as well as epidemic meningitis and septicaemia. However, patients with meningococcal disease represent a rare subset of the estimated 500 million infected individuals worldwide: over 99% of all meningococcal infections result in asymptomatic nasopharyngeal carriage (Stephens, 1999). Therefore, N. meningitidis can be considered a highly successful microorganism given the number of infected individuals worldwide and the fact that meningococci are restricted to only one biological niche, the human nasopharynx. To ensure successful colonization, N. meningitidis must adhere to mucosal surfaces, use locally available nutrients and evade host defences. Although the requirements for adhesion and growth in the nasopharynx are similar between individuals, the immune surveillance systems differ significantly. Neisseriae and other mucosal pathogens have evolved a strategy for optimizing the adaptation to the unpredictable challenges in the new host environment. N. meningitidis phase varies the expression of over 30 surface-exposed molecules including lipooligosaccharide, receptors for iron acquisition, capsule, pili and adhesins (Parkhill et al., 2000; Tettelin et al., 2000). The phase variation process generally requires the presence of repeated nucleotide tracts within or near coding regions. The instability of these repeats during replication can shift reading frames or alter the strength of promoters, thus affecting gene expression (Moxon et al., 1994). This process allows meningococci to alter the expression of their surface components at high frequencies. Indeed, the frequencies of phase variation in Neisseriae have been consistently reported as 10−3 to 10−4 cfu−1 (Murphy et al., 1989; Jonsson et al., 1991; Sarkari et al., 1994; Hammerschmidt et al., 1996; Lewis et al., 1999). However, recent data have demonstrated that there is actually a large difference in frequencies of phase variation among Neisserial clinical isolates (Bucci et al., 1999; Richardson and Stojiljkovic, 1999). The magnitude of interstrain differences cannot be accounted for by varying levels of transcription or length of repeat tracts, although these were shown to modulate the frequency of phase variation to a small extent (Kunkel, 1990; Wierdl et al., 1996; Belland et al., 1997; De Bolle et al., 2000; Lavitola et al., 2000). Recently, the lack of Dam methylation has been offered as a source of hypermutability in N. meningitidis, suggesting that methyl-directed mismatch repair might play an important role in the regulation of phase variation in N. meningitidis (Bucci et al., 1999).

In this study, we conduct a rigorous examination of the phase variation frequency of two unlinked haemoglobin (Hb) receptors from several N. meningitidis clinical isolates. These unrelated strains represent serogroups A, B and C as well as different hmbR switching frequency phenotypes. Given the heterogeneity of Neisserial species (Maiden et al., 1996), all these strains were used to help define the Neisserial mismatch repair system and its role in mutability, with special attention to the frequency of phase variation. This analysis identified mutations in mismatch repair as a source of high-frequency phase variation (> 10−3 cfu−1). Finally, in contrast to the findings of Bucci et al. (1999), Dam methylation was shown to have no effect on the regulation of phase variation in N. meningitidis.


A global mutator phenotype is displayed by some clinical isolates of N. meningitidis

Inspection of hmbR phase variation frequencies in clinical isolates from serogroups A, B and C of N. meningitidis revealed significant, strain-specific differences. The phase variation frequencies of these isolates could be categorized into three classes: fast, medium or slow (Fig. 1A). The median hmbR switching frequency of IR2781(slow) was significantly lower than that of all other strains (P < 0.01), whereas IR2855(fast) had a median frequency that was significantly higher than that of the other five isolates (P < 0.01), as determined by the Wilcoxon rank sum test (Experimental procedures). The remaining four strains were classified as medium and were not significantly different from one another (P > 0.01).

Figure 1.

A global mutator phenotype exists in some clinical isolates of N. meningitidis.

A. Comparison of hmbR switching frequencies revealed significant differences across several clinical isolates. See Experimental procedures for classification of fast, medium or slow phenotypes. The number of G residues in the homopolymeric tract is indicated inside bars.

B. Comparison of hmbR and hpuAB switching frequencies within individual strains revealed no significant difference (P > 0.01). The number of G residues in the homopolymeric tract is indicated inside bars.

C. Comparison of the frequency of spontaneous mutation to rifampicin resistance across several strains revealed isolates displaying mutator phenotypes. All frequencies are represented as medians of at least 10 independent measurements with error bars representing ± one quartile.

Another locus was tested in order to determine whether the phase variation frequencies of other genes are similar to those of hmbR. hpuAB encodes an unlinked haptoglobin/Hb receptor that is known to undergo phase variation by an analogous mechanism, i.e. via mutation of a poly G tract within the receptor coding region (Lewis et al., 1999). In strain IR2855(fast), hmbR and hpuAB are out of frame, phase-OFF, with eight and nine consecutive guanine residues respectively. Both these genes phase vary at similarly high rates, 6.7 × 10−3 cfu−1 and 3.6 × 10−3 cfu−1, which are both statistically higher than frequencies seen in the other five strains (P < 0.01) (Fig. 1B). In strain IR2781(slow), hmbR and hpuAB are also out of frame with eight and 14 G residues respectively. Both genes phase vary at significantly lower frequencies in strain IR2781(slow) than in any of the other five medium or fast strains (P < 0.01) (Fig. 1B). Thus, within any given strain, the switching frequency phenotype is consistent across genes that phase vary by a similar mechanism. The 12-fold disparity between hmbR and hpuAB switching frequencies in IR2781(slow) is most probably caused by the longer homopolymeric tract in hpuA. Longer repeat tracts can lead to small increases in the frequency of phase variation (Kunkel, 1990; De Bolle et al., 2000).

As the transition from phase-OFF to phase-ON is essentially a specific type of mutation, the rates of other types of mutations were also determined. The spontaneous mutation rates to rifampicin and nalidixic acid resistance for the six isolates of N. meningitidis were assayed to determine the missense mutation frequency. Figure 1C shows the median frequencies of mutation to rifampicin resistance for all six strains. The IR2781(slow) strain displayed a low mutation rate to rifampicin resistance in addition to slow hmbR and hpuAB switching frequencies. It can be seen that some strains with elevated phase variation frequencies [e.g. IR2855(fast) and IR2860 (medium)] also possess mutation rates significantly higher than IR2781(slow) (P < 0.01) and are thus classified as high. However, strains IR2857(medium), IR2858(medium) and IR2863(medium) possess rifampicin resistance mutation rates that were not significantly different from IR2781(slow) (P > 0.01) and are thus classified as low, despite the elevated hmbR switching frequencies seen in these isolates. Rates of spontaneous mutation to nalidixic acid resistance confirmed the above results (data not shown).

These data demonstrate that some strains of N. meningitidis have low overall mutation rates [i.e. IR2781(slow)], whereas others have elevated levels of global mutability [i.e. IR2860(medium) and IR2855(fast)]. Furthermore, some strains [i.e. IR2857(medium), IR2858 (medium), IR2863(medium)] seem to have higher mutability only with respect to repeat tracts and not to missense mutations, as these strains possess low rates of spontaneous rifampicin resistance.

Dam methylation has no effect on mutability in several strains of N. meningitidis

Mismatch repair deficiencies were examined as a possible source of hypermutability in several isolates of N. meningitidis. As shown previously (Bucci et al., 1999), not all isolates of N. meningitidis possess a dam homologue. However, comparison of the Dam methylation patterns and phase variation frequencies in the six strains studied showed no correlation. For instance, isolates IR2858(medium), IR2860(medium) and IR2863 (medium) are all proficient in Dam methylation but still possess elevated rates of phase variation (Fig. 2A).

Figure 2.

Dam methylation has no effect on mutability in several N. meningitidis stains.

A. Comparison of hmbR switching frequencies among different clinical isolates shows no correlation between mutability and dam phenotype.

B. Comparison of hmbR switching frequencies between different strain derivatives shows that the dam phenotype has no effect on mutability. The Dam methylation profile, as determined by digestion of chromosomal DNA from each strain with MboI (left) and DpnI (right), is shown below. Frequencies are represented as medians of at least 10 independent measurements with error bars representing ± one quartile.

Inactivation of dam in IR2781(slow) had no effect on hmbR switching frequency (Fig. 2B). Similarly, the introduction of a functional dam allele by gene replacement into strains IR2855(fast) and IR2857(medium) did not affect the frequency of hmbR phase variation in these isolates (Fig. 2B). Moreover, Dam-dependent methylation had no effect on rates of spontaneous mutation to rifampicin resistance in these strains (data not shown).

In Escherichia coli, Dam methylase and MutH work together in directing mismatch repair to the newly synthesized strand; therefore, an attempt was made to identify a Neisserial MutH homologue. However, neither degenerate primers made against conserved domains of mutH homologues of E. coli, Haemophilus influenzae, Salmonella typhimurium and Vibrio cholerae nor Western blot analysis with polyclonal anti-MutH (E. coli) antibody detected homologues of MutH in any Neisserial strain tested (data not shown). In control experiments, the degenerate primers were able to detect mutH sequences from both E. coli and H. influenzae and the anti-MutH (E. coli) antibody detected MutH homologues from E. coli, S. typhimurium, H. influenzae and V. cholerae (data not shown). These results were confirmed by the completion of the genome sequences for two strains of N. meningitidis, neither of which possessed a mutH locus (Parkhill et al., 2000; Tettelin et al., 2000).

As a genetic test for Dam methyl-directed mismatch repair in these strains, an attempt was made to disrupt recombinational repair in dam backgrounds. In E. coli, recombinational repair is crucial for viability of dam strains, as they suffer an increased incidence of double-strand breaks as a result of the lack of mismatch repair strand discrimination (Marinus, 2000). A recA::SpR allele was introduced in three Neisserial dam mutants with similar transformation efficiencies to their isogenic dam+ derivatives (P > 0.05) (data not shown). Furthermore, these recA dam double mutants grew at rates comparable with those of their recA parental strains (data not shown). This suggests that the lack of Dam activity does not result in an increased incidence of double-strand breaks as recombinational repair is dispensable in dam mutants. This is assuming that there is no recA-independent repair pathway for double-strand DNA breaks in N. meningitidis.

MutS and MutL play the major role in the regulation of phase variation frequency in N. meningitidis

Inactivation of dam had no effect on mutation rates in our collection of isolates, although its role in the mismatch repair of E. coli is well understood (Modrich, 1989; Au et al., 1992). To define the mismatch repair system of N. meningitidis, mutS and mutL were inactivated independently via the insertion of a SpR cassette. Inactivation of mutS or mutL in IR2781(slow) increased the hmbR switching frequency by more than two orders of magnitude (Fig. 3A), whereas the frequency of mutation to rifampicin resistance was increased over 20-fold (Fig. 3B). Therefore, as in E. coli, neisserial mismatch repair has a greater influence in maintaining repeat tract fidelity than on preventing the accumulation of missense mutations (Strauss et al., 1997).

Figure 3.

Inactivation of mismatch repair in N. meningitidis mimics the mutator phenotype of IR2855(fast).

A. Effect of mismatch repair inactivation on hmbR switching frequency.

B. Effect of mismatch repair inactivation on the frequency of spontaneous mutation to rifampicin resistance. Frequencies are represented as medians of at least 10 independent measurements with error bars representing ± one quartile.

The difference in phase variation frequencies between IR2781(slow) and IR2855(fast) (Fig. 1A) is comparable with the increase in switching frequency seen upon inactivation of mutL or mutS in IR2781(slow) (Fig. 3A). This pattern also holds true when comparing frequencies of mutation to rifampicin resistance (Fig. 3B). The similarity between the phenotypes of the mutS or mutL mutant derivatives of IR2781(slow) and the phenotype of IR2855(fast) suggests a role for mismatch repair in the elevated mutability of the latter isolate. Indeed, inactivation of mutS and mutL in IR2855(fast) had no effect on the frequencies of phase variation or rifampicin resistance in this strain (Fig. 3A and B).

Upon inactivation of mutS or mutL, medium switching strains also showed an increase in mutability to levels similar to that of IR2855(fast) and the mutS or mutL derivative of IR2871(slow) (Fig. 3A and B). In essence, inactivation of mutS or mutL increased the mutability of all isolates to a level indistinguishable from that of IR2855(fast).

Complementation of a mutator phenotype in IR2855(fast) by a cloned mutLIR2781(slow) allele

The effects of mismatch repair inactivation on both phase variation and spontaneous mutation frequencies indicated that IR2855(fast) might be defective in mismatch repair (Fig. 3A and B). However, all strains discussed thus far, including IR2855(fast), possessed both mutS and mutL homologous sequences as assayed via Southern blot analysis (data not shown).

Transformation of a plasmid encoding the mutL allele of IR2781(slow) into IR2855(fast) significantly reduced the mutability of this strain (Table 1) (P < 0.01). Insertion of the mutLIR2781(slow) allele into the IR2855(fast) chromosome also reversed the high mutability of this strain (data not shown), indicating that mutLIR2781(slow) could complement IR2855(fast) even in single copy. Furthermore, replacing the mutS allele of IR2781(slow) with mutSIR2855(fast) (see Experimental procedures) did not increase mutability in this strain, indicating that MutSIR2855(fast) is fully functional (Table 1). These data show that the MutL protein of IR2855(fast) is non-functional and is the sole mismatch repair defect in this strain.

Table 1. Effect of mismatch repair gene complementation on Hb receptor switching frequencies of IR2781(slow) and IR2855(fast)ab.
StrainWT mutS::SpR mutL::SpR mutS IR2855 c mutS IR2781 c WT + pmutLIR2855dWT + pmutLIR2781d
  • a . Values represent the medians of at least 10 independent measurements rounded to two significant digits and are given × 10 −6.

  • b

    . mutS complementation was achieved by allele replacement (Experimental procedures), and mutL trans complementation was achieved by cloning into the pMGC18.1 vector (Experimental procedures).

  • c . Assays were performed in hmbR::Km R backgrounds and represent hpuAB switching frequencies.

  • d . Assays were performed in hpuAB::Erm R backgrounds and represent hmbR switching frequencies.


mutL of IR2855(fast) has suffered several missense mutations

The nucleotide sequence of mutLIR2855(fast), as well as of the functional mutLIR2781(slow), was determined and compared with the published alleles from strains Z2491 and MC58 (Parkhill et al., 2000; Tettelin et al., 2000). No frameshift mutations or premature stop codons were detected in the non-functional mutLIR2855(fast) allele. Comparison of the putative promoter sequences revealed no base substitutions unique to IR2855(fast), ruling out promoter-down mutations in mutLIR2855(fast) (data not shown). Alignment of the MutLIR2855(fast) sequence with that of the functional MutLIR2781(slow) revealed 32 amino acid substitutions (Fig. 4A). The isolate Z2491 (IR3474) does not display a mutL phenotype. This medium switching and low mutability strain cannot be complemented by MutLIR2781(slow), and therefore MutLZ2491(medium) is assumed to be functional (see below). Alignment of the MutL sequence of IR2855(fast) to MutLIR2781(slow) and MutLZ2491(medium) revealed only 14 unique amino acid substitutions in MutLIR2855(fast)(Fig. 4A, filled flags). A chimeric mutL allele constructed by replacing an internal ClaI fragment of mutLIR2855(fast) with the same fragment from mutLIR2781(slow) also complemented the defect in IR2855(fast) (Fig. 4A, bottom). This indicates that one or more of the 11 amino acid substitutions between residues 39 and 473 is responsible for rendering MutLIR2855(fast) non-functional (Fig. 4A). The highly conserved N-terminal 40 kDa portion of MutL in IR2855(fast) contains an R177L substitution that seemed to be a possible candidate for the debilitating mutation in MutLIR2855(fast) (Ban and Yang, 1998). However, site-directed mutagenesis of MutLIR2781(slow) revealed that an R177L mutation alone does not render MutL non-functional (data not shown).

Figure 4.

The MutL protein of IR2855(fast) contains several amino acid substitutions.

A. MutLIR2855(fast) was compared with the sequences of functional MutLIR2781(slow) and MutLZ2491(medium). The 32 amino acid differences between MutLIR2855(fast) and MutLIR2781(slow) are indicated. The 14 substitutions unique to MutLIR2855(fast) alone [when aligned with both MutLIR2781(slow) and MutLZ2491(medium)] are indicated (dark symbols). The functional chimeric MutL (see text) is shown at the bottom.

B. Comparison of MutLIR2855(fast) with MutL proteins from IR2781(slow) and the sequences of N. meningitidis strains Z2491, MC58 and N. gonorrhoeae strain FA1090. The upper right half of the matrix lists sequence identity; the lower left indicates sequence divergence (megalign software; DNASTAR 1993–2000).

An expanded analysis of N. meningitidis isolates confirms different mutability phenotypes

To ensure that the mutability phenotypes (i.e. slow, medium or fast Hb receptor switching frequencies and high or low rates of spontaneous rifampicin resistance) are a fair representation of N. meningitidis isolates, the mutability phenotypes were determined for an additional 20 strains. This set of isolates represents serogroups A, B and C and consists of disease as well as carriage strains. Twelve of these isolates possessed at least one phase-OFF Hb receptor, thus allowing the determination of Hb receptor phase variation rates. Table 2 describes the comprehensive results from the analysis of mutability phenotypes in all isolates studied thus far. Five (28%) of the 18 isolates were classified as fast switchers, 11 (61%) as medium switchers, and two (11%) strains fell into the slow switching category (Table 2). The spontaneous rifampicin resistance rates of these 18 strains fell into two statistically significant classes, low and high. As with the original six isolates, this analysis revealed that missense mutation rates can correlate with repeat instability. Of the 16 medium/fast strains, six also possessed high rates of spontaneous mutation to rifampicin resistance (Table 2). Conversely, 10 of the medium/fast isolates had low levels of missense mutation accumulation (Table 2). Both slow strains possessed low rifampicin resistance rates (Table 2).

Table 2. Survey of switching and mutation rate phenotypes in N. meningitidis isolates.




Mismatch repair
  • a . Hb receptor phase variation frequencies were classified as described in Experimental procedures. Strains classified as slow had median frequencies below 2.0 × 10 −5, medium strains had frequencies between 1.0 × 10−3 and 1.0 × 10−4, and fast strains had frequencies > 1.0 × 10−3.

  • b . Frequencies of spontaneous mutation to rifampicin resistance were classified as described in Experimental procedures. Strains classified as low had median frequencies below 5.0 × 10 −9, strains classified as high had median frequencies > 7.0 × 10−9.

  • c

    . Complementation of mismatch repair defects was accomplished as described in Experimental procedures.

IR2855ADisease hmbR FastHighMutL
IR2849ADisease hpuAB FastHighMutS
IR2854ADisease hpuAB FastLowNone
IR2862CDisease hmbR FastLowNone
IR2859CDisease hpuAB FastLowNone
IR3547BCarriage hpuAB MediumHighNone
IR3545BCarriage hmbR MediumHighNone
IR2856ADisease hpuAB MediumHighNone
IR2860CDisease hmbR MediumHighNone
IR3474ADisease hpuAB MediumLowNone
IR2851ADisease hpuAB MediumLowNone
IR2857CDisease hmbR MediumLowNone
IR2863CDisease hmbR MediumLowNone
IR3543BCarriage hpuAB MediumLowNone
IR3549BCarriage hpuAB MediumLowNone
IR2858CDisease hmbR MediumLowNone
IR3546BCarriage hmbR SlowLowNA
IR2781BDisease hmbR SlowLowNA

Each strain with elevated Hb receptor switching frequencies was assayed for defects in mismatch repair by complementation with the mutS or mutL locus from IR2781(slow). This screen identified the two strains displaying the fast switching, high mutability phenotypes as being defective in mismatch repair (Table 2). Strain IR2849 was identified by complementation with the mutSIR2781(slow), which lowered the hpuAB switching frequency from 3.7 × 10−3 to 2.8 × 10−5 and the rifampicin resistance frequency from 2.18 × 10−8 to 1.08 × 10−9. The other mismatch repair-defective strain, IR2855(fast), has been described previously (Table 1). None of the other 14 medium/fast strains could be complemented (i.e. restoring slow rates of OFF to ON Hb receptor phase variation) by MutS or MutL alleles of IR2781(slow).


Examination of phase-variable genes in several isolates revealed large differences in switching frequency among different strains (Bucci et al., 1999; Richardson et al., 1999). These differences are not limited to any particular gene, but appear to be a general trait of any strain, as they have been demonstrated in two unlinked loci, hmbR and hpuAB. Furthermore, in some strains [i.e. IR2855(fast) and IR2860(medium)], increased phase variation frequency appears to be associated with an increase in overall mutability. The absence of Dam methylase has been proposed to increase overall mutability in N. meningitidis, presumably by eliminating the strand discrimination of mismatch repair (Bucci et al., 1999). Our data did not show any correlation between Dam methylation and the frequency of phase variation. Neither inactivation of dam nor the introduction of a functional copy of dam into isolates lacking this gene had any effect on mutability. In E. coli, Dam methylation directs the single-strand nicking of the newly synthesized, unmethylated strand via the MutH endonuclease (Lahue et al., 1989). None of the sequenced Neisserial genomes contained a mutH homologue (Parkhill et al., 2000; Tettelin et al., 2000). Furthermore, all attempts to detect a MutH homologue in both pathogenic and commensal isolates of Neisseriae failed. However, these attempts might not detect MutH homologues with highly divergent primary sequences. In conclusion, not only does Dam activity not affect mutability in N. meningitidis, but the endonuclease involved in creating Dam-directed strand nicks may not exist in Neisseriae.

Unlike Dam methylase, MutS and MutL play a major role in the modulation of phase variation frequencies in N. meningitidis. Inactivation of mutS or mutL in IR2781(slow) led to mutability at levels similar to that of IR2855(fast), i.e. an increase of over two orders of magnitude with respect to phase variation frequencies and over 20-fold for missense mutation rates. This dichotomous effect of mismatch repair on repeat tract instability versus mismatch correction has been described in many organisms and is attributed to the differential preferences of the proofreading subunit of DNA polymerase III (Greene and Jinks-Robertson, 1997; Strauss et al., 1997; Sahger et al., 1999).

Inactivation of mutS or mutL in strains with elevated frequencies of phase variation alone [i.e. IR2857(medium), IR2858(medium) and IR2863(medium)] increased overall mutability to levels similar to IR2855(fast) or mutS and mutL derivatives of IR2781(slow). This indicates that the majority of medium strains must have a mechanism that allows increases in the frequency of phase variation while keeping the overall mutability of the strains low. Increased levels of transcription as well as longer repeat tracts have been implicated as causing increased phase variation frequencies (Belland et al., 1997; Lavitola et al., 1999; De Bolle et al., 2000). However, the described effects of these factors do not account for the magnitude of increased phase variation frequency, i.e. 31-fold difference in IR2863(medium) over IR2781(slow) switching frequencies. Currently, the mechanism by which bacteria exclusively increase the frequencies of their phase variable genes is unknown.

The hypermutable phenotype displayed by IR2855(fast) was attributed to one or more of the amino acid substitutions between amino acids 39 and 473 of the MutL protein. These data clearly implicate defects in mismatch repair as being a source of hypermutable phenotypes in naturally occurring isolates. N. meningitidis is not the first organism in which this phenomenon has been described. Mutator phenotypes resulting from mismatch repair defects have been observed in pathogenic and commensal E. coli and S. typhimurium, as well as in Pseudomonas aeruginosa (Taddei et al., 1997; Oliver et al., 2000).

N. meningitidis is found only in a very restricted niche of the human nasopharynx. Direct transmission between hosts is critical in maintaining a perpetual colonization of the human population. Differences between individuals, such as genetic polymorphism and immune surveillance, are probably major obstacles to successful colonization. As these differences are largely unpredictable, Neisseriae must randomly alter the expression of several genes to ensure that at least a fraction of a bacterial population expresses a phenotype that is compatible with a new environment. This phase variation results in a stochastically dynamic population of bacteria that maximizes the likelihood of successful establishment in the new host (Jerse et al., 1994; Akopyants et al., 1995; Weiser and Pan, 1998). It can be estimated that at least 30 genes of N. meningitidis use phase variation to modulate their expression between phase-ON and phase-OFF states (Parkhill et al., 2000; Tettelin et al., 2000). Therefore, individual Neisserial colonies are genetically and phenotypically heterogeneous as a result of the presence of these phase-variable genes. An increase in phase variation frequencies would lead to even more heterogeneity by expanding the number of different phenotypes available to a finite population of colonizing N. meningitidis (De Bolle et al., 2000).

In this study, 16 (89%) of the 18 Neisserial isolates possessed elevated Hb receptor switching frequencies (i.e. > 1.0 × 10−4). Furthermore, six (38%) of these isolates with increase phase variation were essentially mutators, in that they also possessed increased overall mutability (i.e. rifampicin resistance rates > 7.0 × 10−9). This subset of strains includes two mismatch repair-deficient mutants, IR2855(fast) and IR2849(fast), which were shown to be defective in MutL and MutS respectively. The fact that N. meningitidis isolates exhibiting a general mutator phenotype exist emphasizes the magnitude of the selective pressure for more variability. Mutators, such as those strains with defects in mismatch repair, would also suffer the increase in accumulation of deleterious mutations, thus lowering the fitness of the cell (Taddei et al., 1997; Rainey, 1999; Fuchain et al., 2000). The high prevalence of mutators among isolates of N. meningitidis may result in part from the extra benefit gained by large increases in phase variation frequencies. Furthermore, the cost of a mutator phenotype may be lessened by the fact that Neisseriae are naturally competent and undergo frequent extrachromosomal recombination (Feil et al., 1999). This common phenomenon could potentially sever the linkage between mutator alleles and deleterious mutation that may have occurred, thus lowering the cost of a mutator phenotype (Radman et al., 2000).

The majority of strains analysed were isolated from patients, and it is possible that the high prevalence of medium and fast strains is a result of a bias in strain selection. However, strains with increases in the frequency of phase variation were also found in carriers, ruling out the absolute correlation between elevated phase variability and increased virulence. Instead, these data suggest that the selective advantage of fast and medium strains may be in their faster adaptation to new environments and thus to more efficient transmissibility. Further studies must be carried out in order to determine the prevalence of strains with increases in phase variation frequencies and their role in the virulence and transmissibility of N. meningitidis.

Experimental procedures

Bacterial strains

Neisserial strains used in this study are described in Table 3. The 18 disease isolates examined were obtained from Dr M. Reeves (Centers for Disease Control and Prevention) with a minimal number of passages. These isolates are phylogenetically unrelated and represent serogroups A, B and C. The eight serogroup B carriage isolates studied here were obtained from Dr D. Stephens (Veterans Administration Medical Center, Atlanta, GA, USA). Neisseriae were grown on GCB (Difco) agar containing Kellog's supplements and incubated at 37°C and 5% CO2. E. coli were grown in Luria–Bertani (LB) broth at 37°C. When necessary, Neisseriae (E. coli) were grown in the presence of 3 µg ml−1 (300 µg ml−1) erythromycin, 100 µg ml−1 (100 µg ml−1) spectinomycin, 100 µg ml−1 (50 µg ml−1) kanamycin, 750 µg ml−1 (100 µg ml−1) streptomycin, 50 µM desferoxamine mesylate and 100 µg ml−1 human Hb for selection. Kanamycin selection in Neisseriae was performed in brain–heart infusion (BHI) media (Difco) supplemented with 2.5% fetal bovine serum (heat inactivated).

Table 3. Strains, plasmids and oligonucleotides used in this study.
IR0540 S. typhimurium strain LT2Laboratory collection
IR1113 N. gonorrhoeae strain MS11W. Shafer
IR2059 V. cholerae strain DH11S. Payne
IR2363–65 N. lactamica strains O9238, E6040 and O1748S. Barish
IR2781 N. meningitidis clinical isolateD. Stephens
IR2848–63 N. meningitidis clinical isolatesLaboratory collection
IR2865 N. gonorrhoeae strain FA1090W. Shafer
IR3366 E. coli strain BS40 metD+/XLB. Strauss
IR3446 H. influenzae strain RdM. Farley
IR3474 N. meningitidis strain Z2491D. Stephens
IR3542–49 N. meningitidis carriage isolatesD. Stephens
IR4114 E. coli strain GW3773 mutH::KmR E. coli Genetic Stock (ECGS)
IR4638 N. meningitidis strain ctrA::Tn916, recA::SpRD. Stephens
 pSK-SpRSpR cassette in SspI of pBluescript SK+This study
 pMGC18.1 E. coli Neisseriae shuttle vectorNassif et al. (1991)
 pIRS525 hmbR::KmR (NotI) Richardson and Stojiljkovic (1999)
 pARR1500 hpuB::ErmR (ClaI) Richardson and Stojiljkovic (1999)
 pARR1569 mutS::SpR (EcoRI)This study
 pARR1731 mutL::SpR (NdeI)This study
 pARR1736 dam::KmR (engineered PstI)This study
 pARR1798 mutS::ermC-rpsL (NdeI)This study
 pARR2023 drg::ermC-rpsL (engineered BamHI)This study
 pARR2027 mutL IR2855(fast) in pMGC18.1This study
 pARR2030 mutL IR2781(slow) in pMGC18.1This study
 pARR2062 mutL IR2781(slow) in pSK-SpRThis study
 pARR2063 mutS IR2781(slow) in pSK-SpRThis study
 mutH.1a5′-AAACGNGACAAAGGNTGGGT-3′This study
 mutH.1b5′-GTCCATNARTTCCCARTC-3′This study

Plasmid construction

The sequences for all oligonucleotides are listed in Table 3. mutS and mutL were amplified by polymerase chain reaction (PCR) from strain IR2781(slow) with oligonucleotides mutS.1a/mutS.1b and mutL.1a/mutL.1b and cloned into pCR2.1 (Invitrogen). A SpR cassette from pHP45Ω (Prentki and Krisch, 1984) was inserted into the EcoRI site of mutS and a NdeI site of mutL to give pARR1569 and pARR1731 respectively. In addition, the mutL and mutS loci from IR2781(slow) were amplified with oligonucleotides mutL.3a/mutL.1b and mutS.2a/mutS.3b, respectively, cloned into pCR2.1 and, finally, subcloned into pSK-SpR by EcoRV–HindIII to give pARR2062 and pARR2063 respectively. The mutL locus from pARR2062 was subcloned into pMGC18.1 via BamHI to give pARR2030. The mutL locus from IR2855 was also amplified with mutL.3a/mutL.1b and subcloned from pCR2.1 into pMGC18.1 with EcoRV–HindIII to give pARR2027. A dam::KmR construct was generated by amplifying a 5′dam segment with the primer pair leuS.X/damP.rev, a 3′dam segment with the primer pair damP.for/dam.2b. A KmR cassette from pKSAC (Pharmacia) was inserted into the engineered internal PstI site to give pARR1736. The mutS and dam allelic exchange constructs, pARR1798 and pARR2023, were generated by cloning an EcoRI–HindIII fragment containing an ermCrpsL cassette from pFLOB4300 (Johnston and Cannon, 1999) into both a blunted NdeI site of mutS and an engineered BamHI site in drg respectively.

DNA sequencing

All DNA sequencing was performed at the Emory University Core Sequencing Facility on an ABI model 377 automated sequencer using the dye terminator cycle sequencing technology. mutLIR2855(fast) and mutLIR2781(slow) from two independently cloned fragments amplified with mutL.3a/mutL.1b were sequenced using oligonucleotides mutL.1a.1b.2a.2b and mutL.3a. mutH homologous sequence from various species was PCR amplified using degenerate oligonucleotides mutH.1a/mutH.1b, and the products were cloned directly into pCR2.1 for sequencing. mutH sequence was determined by sequencing cloned fragments with primers M13-Forward and M13-Reverse. DNA sequence confirmed the presence of mutH homologues from E. coli and H. influenzae; however, no mutH homologous sequence was detected from any Neisserial strain tested.

Insertional inactivation of mutS, mutL, dam and recA genes

Inactivation of dam, mutL and mutS genes was achieved by transformation of N. meningitidis with XbaI-linearized pARR1736, pARR1731 and pARR1569 respectively (Table 3). Constructs were HaeIII methylated (New England Biolabs) before transformation. Approximately 1 µg of linearized plasmid DNA was incubated with 1×HaeIII reaction buffer (provided), 80 µM SAM (provided) and 2.5 U HaeIII methylase at 37°C for 1 h. Transformation of N. meningitidis was performed as described previously (Richardson et al., 1999). mutL and mutS inactivation was confirmed by Southern blot analysis. dam inactivation was confirmed by chromosomal DNA digestion with DpnI and MboI. recA was inactivated by transformation with IR4638 chromosomal DNA (Table 3). Inability to be transformed with chromosomal DNA from IR3298 (IR2781(slow) hpuB::KmR) confirmed the phenotype of recA::SpR transformants.

Allelic replacement and trans complementation

Replacement of mutS and drg alleles was performed by transforming StrR isolates with XbaI-linearized pARR1798 and pARR2023, respectively, followed by selection for erythromycin resistance (Table 3). These ErmRStrS alleles were replaced by transformation with chromosomal DNA from strains with the desired genotype [i.e. IR2781(slow) chromosomal DNA for dam+ and mutSIR2781(slow) replacements, and IR2855(fast) chromosomal DNA for mutSIR2855(fast) replacement], followed by selection for streptomycin resistance and screening for erythromycin-sensitive transformants. Trans complementation of mutL was achieved by cloning alleles into the pMGC18.1 shuttle vector (Table 3). mutL and mutS complementation was also accomplished by chromosomal co-integration of undigested plasmids pARR2062 and pARR2063 respectively. Constructs were HaeIII methylated before transformation.

Mutant frequency and transformation efficiency assays

Mutant frequencies have been used throughout the study to describe rates of phase variation and mutation to drug resistance. These frequencies allowed for valid comparisons between isolates, as the average number of generations/colony differed by less than 10% among all strains. hmbR phase variation frequencies were determined in an hpuB::ErmR background, and hpuAB phase variation frequencies were determined in an hmbR::KmR background. Single colonies were picked from an overnight GCB agar plate and resuspended in 200 µl of broth. Serial dilutions were plated on GCB agar plates supplemented with 50 µM desferoxamine mesylate (Sigma) and 100 µg ml−1 human Hb (Sigma). Hb receptor switching frequencies were obtained by counting the number of Hb+ cfu/colony and dividing by the total viable cfu/colony (Richardson et al., 1999). These frequencies are represented as the median value of at least 10 independent measurements with error bars representing plus or minus quartiles. Spontaneous mutation rates to rifampicin and nalidixic acid frequencies were performed by pelleting ≈ 1010 bacteria, resuspending in 250 µl of broth and plating serial dilutions on GCB agar plus rifampicin (3 µg ml−1) or nalidixic acid (5 µg ml−1). These concentrations were chosen as they are roughly five times the average minimal bactericidal concentration (MBC) for this collection of strains (0.5 µg ml−1 for rifampicin and 1.25 µg ml−1 for nalidixic acid; data not shown). Frequencies were determined by dividing resistant cfu ml−1 by the total viable cfu ml−1, and rates are represented as medians of at least 10 independent measurements. Transformation efficiencies were determined by incubating a known viable count of bacteria with DNA for 30 min, followed by a 6 h growth period. Viable counts were obtained again along with the number of transformants. The number of transformants was corrected for the generations during outgrowth, and efficiencies were determined as the number of transformation events cfu−1 in the original inoculum. The medians of three independent reactions were used for comparison.

Statistical analysis

Mutant frequencies were determined (as mentioned above), and median values were used to represent data sets from each of the different N. meningitidis isolates. Statistical significance across different data sets was determined via the Wilcoxon rank sum test. Divisions between fast, medium and slow Hb receptor switching frequency phenotypes were derived by analysing pairwise comparisons of the data sets from each of the 18 isolates. Using P < 0.01 as a cut-off, three distinct classes emerged. Divisions between high and low rates of spontaneous resistance to rifampicin were determined similarly. Pairwise analysis using the data sets from 18 strains could be grouped into two distinct classes based on a cut-off of P < 0.01.

Western blot analysis

Whole-cell lysates were obtained by boiling 1 ml of mid-log cell culture for 10 min in SDS cracking buffer (Sambrook et al., 1989), then cooling on ice for at least 30 s. The samples were loaded directly to SDS–PAGE and blotted as described previously (Richardson et al., 1999). Blots were probed with a poly αMutH (E. coli) antibody obtained from Dr P. Modrich and detected as described previously (Richardson et al., 1999).


We thank Drs S. Berish, M. Farley, P. Modrich, M. Reeves, W. Shafer, D. Stephens and B. Strauss for providing strains used in this study. We thank P. Modrich for anti-MutH antibody. We thank S. Jinks-Robertson, B. Levin, M. G. Marinus, J. Scott, B. Shafer and D. Stephens for comments and suggestions. This work was supported by the Public Health Service grant AI472870-01A1 to I.S. and the National Institutes of Health (NIH) training grant 2T32 AI07470 to A.R.R.