CanB is a metabolic mediator of antibiotic resistance in Neisseria gonorrhoeae

Abstract

The evolution of the obligate human pathogen Neisseria gonorrhoeae has been shaped by selective pressures from diverse host niche environments as well as antibiotics. The varying prevalence of antibiotic resistance across N. gonorrhoeae lineages suggests that underlying metabolic differences may influence the likelihood of acquisition of specific resistance mutations. We hypothesized that the requirement for supplemental CO₂, present in approximately half of isolates, reflects one such example of metabolic variation. Here, using a genome-wide association study and experimental investigations, we show that CO₂-dependence is attributable to a single substitution in a β-carbonic anhydrase, CanB. CanB^19E is necessary and sufficient for growth in the absence of CO₂, and the hypomorphic CanB^19G variant confers CO₂-dependence. Furthermore, ciprofloxacin resistance is correlated with CanB^19G in clinical isolates, and the presence of CanB^19G increases the likelihood of acquisition of ciprofloxacin resistance. Together, our results suggest that metabolic variation has impacted the acquisition of fluoroquinolone resistance.

Introduction

Neisseria gonorrhoeae is the cause of the highly prevalent sexually transmitted infection gonorrhea, with an estimated 87 million new cases worldwide in 2016¹. N. gonorrhoeae can infect multiple anatomical sites, including the urethra, cervix, rectum, and oropharynx², and therefore interacts with and adapts to host immune and site-specific pressures^3,4. In addition to these pressures, the evolution of the gonococcus has been shaped by antibiotics. N. gonorrhoeae has acquired resistance to every first-line antibiotic used to treat it⁵, including the fluoroquinolone ciprofloxacin—with resistance around 35% in the US and >90% in several countries^6–9—and the current first-line therapy, the extended spectrum cephalosporin ceftriaxone^10,11.

Antibiotic resistance varies across the phylogenetic tree, with some lineages displaying more resistance than others, though the reasons for these differences are not known. N. gonorrhoeae has two major lineages, distinguished by sexual networks and levels of antibiotic resistance: lineage A, associated with men-who-have-sex-with-men and high levels of antibiotic resistance, and lineage B, associated with heterosexuals and lower levels of resistance¹² (Fig. 1A).

Figure 1.

A SNP in ngo2079, encoding a β-carbonic anhydrase, is necessary and sufficient to explain dependence on supplemental CO₂ in N. gonorrhoeae. (A) Maximum-likelihood phylogenetic tree of 5,007 strains of N. gonorrhoeae with a colored track representing CanB variant. Branch length represents total number of substitutions after removal of predicted recombinations. Lineage A represents isolates that arose phylogeographically After the introduction of N. gonorrhoeae into Asia, while Lineage B represents isolates that arose Before that breakpoint, as defined by Sánchez-Busó et al.¹² (B) Manhattan plot of SNPs associated with dependence on supplemental CO₂ across 30 strains of N. gonorrhoeae. Dotted line represents a p-value of 0.05 with Bonferroni correction for 57,691 unitigs. (C) Plating efficiencies in the absence and presence of 5% supplemental CO₂ of CanB variants introduced by kanamycin co-selection in N. gonorrhoeae strains FA1090 (parental CanB^19E) and FA19 (parental CanB^19G) (From left to right, N = 6, 6, 9, 9, from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.00512, p = 0.00062) determined by two-sided Mann-Whitney U test. (D) Plating efficiencies as in (C) of N. gonorrhoeae strains 28BL (parental CanB^19E) and FA19 (parental CanB^19G) with isogenic CanB mutants (from left to right, N = 9, 8, 10, 7, from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.00062, p = 0.00076) determined by two-sided Mann-Whitney U test. (E) NGO2079/CanB variants across sequenced Neisseria and related species represented alongside a 16S maximum-likelihood tree. Branch length represents number of substitutions per site. (F) Growth of FA19 CanB isogenic strains in the presence and absence of supplemental CO₂ (N=3, representative of two independent experiments, error bars represent SEM). Significance (from top to bottom, p = 0.45, p = 0.00069, p = 0.014) at 18 hours determined by unpaired two-sided t-test. *p<0.05, **p<0.01, 599 ***p<0.001

One hypothesis to explain the phylogenetic variation in resistance is that lineages may be subject to demographically or geographically variable antibiotic pressures. The overrepresentation of specific transmission networks based on sexual behavior within lineages lends support to this hypothesis¹², but the rapid global spread of N. gonorrhoeae and the extensive bridging across sexual networks argue against it^13,14. As metabolic factors can contribute to resistance¹⁵ and metabolic environment affects antibiotic resistance acquisition¹⁶, we hypothesized that metabolic variation may be one factor contributing to why some lineages more readily acquire resistance.

One metabolic factor with known variation is CO₂-dependence. In a clinical setting, protocols recommend growing N. gonorrhoeae in the presence of 5% supplemental CO₂¹⁷. However, only around half of clinical isolates require CO₂ for growth¹⁸, suggesting that that there may be a genetic basis for this dependency. Here, we found that CO₂-dependence in N. gonorrhoeae is driven by a single mutation in the N, gonorrhoeae β-carbonic anhydrase, CanB, that this mutation is associated with resistance to ciprofloxacin in clinical isolates, and that this mutation compensates for fitness costs incurred by gyrase mutations that confer ciprofloxacin resistance.

Results

A single mutation in CanB is necessary and sufficient for CO₂-dependence in N. gonorrhoeae

We first investigated the mechanism of CO₂-dependence using a genome-wide association study (GWAS). To generate phenotypes for GWAS, clinical and laboratory strains of N. gonorrhoeae were assessed for growth in the presence and absence of supplemental CO₂. The GWAS revealed a significant association with a single genomic site, a missense mutation in ngo2079 (Fig. 1B). In a parallel unbiased method, genomic DNA from FA6140, an N. gonorrhoeae strain that does not depend on supplemental CO₂, was used to transform FA19, a lab isolate that requires supplemental CO₂ (Extended Data 1A). Transformants that grew in the absence of supplemental CO₂ had acquired multiple SNPs in the region of ngo2079 (Extended Data 1B). As the SNP identified in the GWAS was also present in the undirected transformant, we proceeded with further investigation of the mutation in ngo2079.

The ngo2079 c.56G→A SNP encodes a G→E mutation at amino acid 19 of NGO2079, a β-carbonic anhydrase. N. gonorrhoeae contains two additional carbonic anhydrases: a well-characterized essential periplasmic ɑ-carbonic anhydrase^19–21 and a putative nonessential cytosolic ɣ-carbonic anhydrase²². All carbonic anhydrases catalyze the conversion of H₂O and gaseous CO₂ to aqueous bicarbonate. Aqueous bicarbonate is then used as a carbon building block in multiple pathways within the cell, including the synthesis of TCA cycle components, nucleotides, and phospholipids²³ (Extended Data 1C). As ngo2079 is the sole predicted N. gonorrhoeae β-carbonic anhydrase, we named this gene canB to parallel the corresponding carbonic anhydrase can in E. coli. The SNP is located outside of the highly conserved carbonic anhydrase domain (Extended Data 1D) and does not appear to be within a signal sequence domain by SignalP²⁴. Structures from Alphafold²⁵ indicate that the variant amino acid lies between two N-terminal alpha-helices, but it does not appear at the dimerization interface or the characterized active site (Extended Data 1E).

We validated our GWAS results by introducing the CanB variants into N. gonorrhoeae. Replacement by kanamycin co-selection of the CanB^19E in FA1090 with CanB^19G and of the CanB^19G in FA19 with CanB^19E showed that dependence on supplemental CO₂ was attributable to CanB^19G (Fig. 1C), while similar results for single point mutants in 28BL (19E→G) and FA19 (19G→E) (Fig. 1D) support this idea. These results indicate that CanB^19E is both necessary and sufficient for growth in the absence of supplemental CO₂. The concentration of CO₂ that differentiates this dependency is between atmospheric (~410 ppm) and 0.1% CO₂ (1000 ppm) (Extended Data 2A), slightly lower than has been found in other bacteria²⁶ and considerably below the 5% CO₂ standardly used in N. gonorrhoeae growth protocols²⁷.

CanB^19G, the allele that confers CO₂-dependence, appears to have arisen multiple times in N. gonorrhoeae, whereas genomes from Neisseria spp. and other closely related bacteria encode only CanB^19E (Fig. 1E). Analysis of the PubMLST database of more than 36,000 Neisseria spp. genomes showed that only one non-gonococcal Neisseria isolate carries CanB^19G, a urethral isolate of N. meningitidis (PubMLST id: 82427) that appears to have a acquired the canB region from N. gonorrhoeae (100% identity to sequenced N. gonorrhoeae isolates). Within N. gonorrhoeae, ~50% of sequenced isolates from across the phylogenetic tree have CanB^19G (Fig. 1A, E). The proportion of sequenced isolates with CanB^19G has remained relatively stable at ~50% during last twenty years (Extended Data 1F), though this allele may have been less common in the pre-antibiotic era of the 1930s (1/11 (9%) of pre-1930 N. gonorrhoeae isolates from the Pub-MLST database).

Although CanB^19E provides a growth advantage in the absence of supplemental CO₂, the growth of FA19 CanB^19G and CanB^G19E in the presence of supplemental CO₂ was indistinguishable and independent of inoculum size (Fig. 1F, left). Furthermore, in the presence of CO₂, kanamycin-labeled CanB isogenic FA19 strains remained at similar proportions in competition over time as compared to the input (Extended Data 2B). The CanB^19G variant, therefore, does not confer a general fitness defect but a specific growth disadvantage in the absence of CO₂. As shown in Fig. 1F, right, this disadvantage is concentration-dependent and is exacerbated at lower starting inocula. Although FA19 CanB^19E is able to grow at all starting CO₂ concentrations, both strains have a substantially longer lag phase in the absence of CO_2, suggesting that intracellular bicarbonate plays a critical role in lag phase, similar to N. meningitidis²⁸. This competitive disadvantage is not rescued by extracellular sodium bicarbonate or oxaloacetate, a downstream product of intracellular bicarbonate (Extended Data 2C). Thus, in contrast to many bacteria that are able to acquire bicarbonate from the environment via inorganic carbon transporters²⁹, N. gonorrhoeae does not appear to have a CO₂-concentrating mechanism apart from cytosolic carbonic anhydrases, explaining why exogenous bicarbonate does not rescue the competition defect. Supplementation with metabolites downstream of intracellular bicarbonate (Extended Data 1D), such as pyruvate, sub-minimum inhibitory concentration (MIC) palmitic acid, and pyrimidines were also unable to rescue this defect (Extended Data 2D).

CanB^19G has reduced enzymatic activity compared to CanB^19E

E. coli MG1655 was used to compare the activity of the CanB variants. Deletion of one of the two endogenous E. coli β-carbonic anhydrases, can, rendered MG1655 capnophilic and dependent on supplemental CO₂ (Fig. 2A, top two quadrants). Complementation with IPTG-inducible N. gonorrhoeae CanB variants restored growth in the absence of supplemental CO₂. However, lower concentrations of IPTG were required for complementation with CanB^19E versus CanB^19G (Fig. 2A, lower two quadrants), even though the abundance of CanB^19E and CanB^19G were similar at equivalent levels of induction (Extended Data 2E), suggesting that CanB^19G retains some enzymatic activity but does not restore growth as efficiently as CanB^19E.

Figure 2.

The 19G variant of CanB is a functional hypomorph. (A) Growth of parental E. coli MG1655, MG1655Δcan, and MG1655Δcan complemented with CanB variants under different levels of induction. (B) Survival curve under multiple low pH conditions of FA19 isogenic CanB isolates in monoculture (N=3, representative of three independent experiments, error bars represent SEM). (C) As in (B), survival at 20 minutes of exposure to low pH conditions (N=6, from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.81, p = 0.002) by two-sided Mann-Whitney U test. (D) Intracellular pH of FA19 CanB^19G and FA19 CanB^G19E as measured by flourimetric dye in the presence and absence of supplemental CO₂ (N=3, representative of two independent experiments, error bars represent SEM). Significance (p = 0.016) by unpaired two-sided t-test. (E) Competition under anaerobic conditions between FA19 isogenic CanB strains on GCB-K agar (N=3, representative of two independent experiments, error bars represent SEM). *p<0.05, **p<0.01, ***p<0.001

Since CanB^19G is a hypomorph, we hypothesized that the CanB variants result in altered gonococcal physiology. At pH 4.5, an acidic condition similar to healthy vaginal pH³⁰, FA19 CanB^19G died more quickly than the isogenic FA19 CanB^19E (Fig. 2B, C), consistent with the finding that, in the absence of supplemental CO₂, FA19 CanB^19G is less able to buffer intracellular pH (Fig. 2D). The CanB^19E survival advantage at low pH did not extend, however, to intracellular survival within immortalized macrophages (Extended Data 3A). At higher pH and in aerobic conditions, neither the CanB^19E or CanB^19G variant displayed a substantial advantage (Extended Data 2B, Extended Data 3B). In contrast, under anaerobic conditions, FA19 CanB^19E had a distinct growth advantage in competition (Fig. 2E). Our data thus show that the CanB^19E allele provides an advantage at low pH and under anaerobic conditions, suggesting that CanB^19E may allow for buffering in the acidic vaginal environment.

CanB^19G is associated with ciprofloxacin resistance in clinical isolates

Carbonic anhydrases have been previously shown to be targets of sulfa antibiotics²¹ and to mediate natural competence in N. meningitidis³¹. Sulfa antibiotics were one of the first antibiotics to be used to treat gonorrhea³² and primarily target dihydropteroate synthase in the folate synthesis pathway. Both CanB variants had similar sulfamethoxazole MICs across two genetic backgrounds. Similarly, the FA19 CanB isogenic pair had similar trimethoprim/sulfamethoxazole MICs, though in 28BL, there was a modest twofold increase in the CanB^E19G strain (Extended Data 4A). Furthermore, MG1655Δcan complemented with CanB^19G or CanB^19E also had similar sulfamethoxazole MICs (Extended Data 4B). The CanB variants therefore do not have different susceptibility to sulfa antibiotics. With respect to competence, while a canB knockout was dependent on supplemental CO₂ (Extended Data 5A), both the wild-type and knockout had the same level of natural competence, as measured by the introduction of a single SNP in gyrB conferring nalidixic acid resistance (Extended Data 5B). Similarly, competence was not associated with the CanB isogenic pairs of FA19 and 28BL (Extended Data 5B), though there were small differences in competence likely due to increased piliation from strain construction. Furthermore, there was no difference in nalidixic acid susceptibility at baseline between the FA19 isogenic strains (Extended Data 4C), suggesting that the competence assay was not biased by underlying changes in nalidixic acid resistance. Finally, as an additional control, introduction of a kanamycin cassette led to similar transformation efficiencies between isogenic pairs (Extended data 5C). Thus, selection for CanB variants does not appear to be driven by associations with sulfa susceptibility or natural competence.

We instead hypothesized that genomic and associated metabolic variation could modify the likelihoods of acquiring and maintaining resistance. Support for this idea comes from previous research showing that intracellular metabolite perturbations can shape the likelihood of developing resistance^16,33. To begin to address whether the CanB variants have shifted the evolutionary landscape of N. gonorrhoeae antibiotic resistance, we analyzed whether the variants are associated with antibiotic resistance on the population level. We found that the CanB^19G variant is associated with >4-fold higher MICs for ciprofloxacin (average log₂ MIC 0.43 vs. 0.10 μg/mL) (Fig. 3A), whereas MICS for azithromycin (0.32 μg/mL vs. 0.24 μg/mL), penicillin (0.41 vs. 0.70 μg/mL), and ceftriaxone (.0093 μg/mL vs. .013 μg/mL) did not differ (Extended Data 6A). CanB^19G is also associated with the most common resistance determinant for ciprofloxacin, GyrA^91F, as 52.3% of CanB^19G isolates have GyrA^91F, as compared to only 30.4% of CanB^19E isolates (p<0.0001, Pearson’s chi-square) (Fig. 3B). This relationship holds true across years, as the proportion of GyrA^91F and CanB^19G isolates per year in the United States have tracked closely since 2000 (Fig. 3C). On a country-by-country level, isolates with CanB^19G have a comparatively higher ciprofloxacin MIC (Extended Data 6B), indicating that the association is true across datasets and geography. One possible explanation is that CanB^19G directly increases ciprofloxacin resistance. However, the isogenic CanB FA19 strains have comparable levels of resistance to ciprofloxacin (Extended Data 4C).

Figure 3.

The CanB^19G variant is associated with ciprofloxacin resistance (A) Violin plot of ciprofloxacin MICs for 8,912 N. gonorrhoeae clinical isolates. Thick lines represent median, thin lines represent 25^th/75^th percentiles. Significance (p = 1.3e-41) by two-sided Mann-Whitney U test. (B) Maximum-likelihood tree, as in Figure 1C, overlaid with tracks representing ciprofloxacin conferring variants of GyrA. (C) Proportion of sequenced U.S. isolates stratified by year that have the CanB^19G variant and the ciprofloxacin resistance-conferring GyrA^91F variant. *p<0.05, **p<0.01, ***p<0.001

CanB^19G compensates for gyrase mutation-related fitness costs

We reasoned that there were multiple potential mechanistic explanations for the differences in ciprofloxacin associated with the CanB^19G and CanB^19E variant strains. The first is that CanB^19G has a higher mutation rate, manifesting as a higher rate of resistance to all antibiotics and therefore leading to more ciprofloxacin resistance. This does not appear to be the case, since FA19 CanB^19E had a modestly higher mutation rate to rifampin, an antibiotic that targets RNA polymerase, by fluctuation assay (Extended Data 6C, D). There was additionally no difference in killing by ciprofloxacin for either strain, suggesting that the population-level difference in ciprofloxacin resistance between the strains is not due to differential viability in the presence of drug (Extended Data 7A).

A further possibility is that the CanB^19G strains have an increased ability to tolerate gyrA mutations. If this were true, we would expect that CanB^19E gyrA mutants would have a relative competitive disadvantage compared to the parental strains. To investigate this possibility, we generated GyrA^91F and GyrA^91F/95G mutants in both FA19 CanB backgrounds. For both the GyrA^91F and GyrA^91F/95G mutants, the CanB^19E strain was significantly disadvantaged against the parental strain (Fig. 4A). In monoculture, FA19 CanB^19E GyrA^91F/95G also grew significantly slower than FA19 CanB^19G GyrA^91F/95G (Fig. 4B, Extended Data 7B), as did the equivalent mutants in the 28BL background (Extended Data 7C). We generated spontaneous ciprofloxacin resistant mutants, all of which bore GyrA⁹⁵ variants seen in clinical isolates (Extended Data 4D). The ciprofloxacin-resistant FA19 CanB^19G strains competed better than ciprofloxacin-resistant CanB^19E strains against the susceptible parental strains (Extended Data 7D) and showed faster growth in monoculture (Extended Data 7E). Furthermore, acquisition of the specific GyrA^91F/95G allele was not different in a natural transformation experiment (Extended Data 5D). Importantly, this experiment was performed on a limited time-scale, minimizing any differences in fitness between the strains. Taken together, these results suggest that the increased rate of acquisition of ciprofloxacin resistance in the CanB^19G background is due to its higher relative fitness in the background of gyrase mutations. This finding also suggests that, particularly with the recent expansion of GyrA^91F isolates (Fig. 3C), there may be pressure in the gyrA mutant background to select for the CanB^19G variant. Thus, while canB variation is present in the N. gonorrhoeae population prior to the introduction of fluoroquinolones, we cannot rule out that CanB^19G may also represent a compensatory mutation acquired following gyrA mutation in at least some cases.

Figure 4.

CanB^19G provides a fitness advantage in the context of gyrase mutations. (A) Competition between ciprofloxacin resistant isogenic CanB strains and susceptible parental strains (N=3, representative of two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.00063, p = 0.0019) by unpaired two-sided t-test. (B) Growth curves of CanB isogenic FA19 strains with ciprofloxacin resistance-determining gyrA alleles (N=12, data from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.018, 0.048, 5.0e-5, 9.7e-5, 0.00041, 0.00019) by unpaired two- sided t-test. (C) Volcano plot of RNA-seq showing transcript levels compared between FA19 CanB^G19E and FA19 CanB^G19E GyrA^91F by DeSeq2. Specific transcripts of interest are highlighted in color. (D) Competition between FA19 CanB^G19E and FA19 CanB^G19E GyrA^S91F after 16 hours along with media supplementation in liquid GCP medium (N=6, representative of two independent experiments, error bars represent SEM). Significance (p = 0.99, p = 2.2e-6, p = 3.3e-9) by one-way ANOVA with Dunnett’s multiple comparisons test. *p<0.05, **p<0.01, ***p<0.001

To investigate the mechanism of compensation and the fitness effects of GyrA in the FA19 CanB^19E background, we performed RNA-sequencing (RNA-seq) on FA19 CanB^19G, FA19 CanB^19E, and FA19 CanB^19E GyrA^91F growing in log-phase liquid culture with 5% CO₂ and focused on transcripts that were differentially regulated between FA19 CanB^19E and FA19 CanB^19E GyrA^91F, but not differentially expressed between FA19 CanB^19G and FA19 CanB^19E. We identified 151 such transcripts, including the downregulation in FA19 CanB^19E GyrA^91F of purK, encoding an N5-carboxyaminoimidazole ribonucleotide synthase involved in purine synthesis (Supp. Table 1, Fig. 4C).

Based on this finding, we hypothesized that purine availability modulates fitness in a manner dependent on carbonic anhydrase and gyrase. Purine nucleosides, including inosine (Fig. 4D), adenosine, and guanosine (Extended Data 7F) were sufficient to rescue the growth defect of FA19 CanB^19E GyrA^91F in competition with FA19 CanB^19E. In contrast, supplementation with pyrimidines had no effect on competition between the two strains (Extended Data 7F). Furthermore, purine supplementation had no effect on competition between FA19 CanB^19G GyrA^91F and FA19 CanB^19G (Extended Data 7G). Thus, purine nucleosides rescued the growth defect conferred by GyrA^91F in the CanB^19E background. CanB^19E also shows lineage-specific background dependence, as the alleviation of CO₂-dependence in clinical isolate NY0195 requires further introduction of AasN^Q321E, a mutation in an acyl-ACP synthetase involved in fatty acid uptake³⁴ (Extended Data 8A). The AasN^321Q variant is restricted to a single lineage that carries the ceftriaxone-resistance determining PenA34 allele (Extended Data 8B), suggesting further complex connections between central metabolism and antibiotic resistance.

Discussion

Collectively, our findings suggest that CanB represents a metabolic mediator of drug resistance acquisition in N. gonorrhoeae. CanB^19G is present in about half of clinical isolates of N. gonorrhoeae and explains the growth requirement for supplemental CO₂; in contrast, isolates with CanB^19E can grow without this requirement. Although canB is not essential for in vitro growth (Extended Data 5A), we observed no evidence of premature stop codons in a collection of >16,000 genomes from clinical isolates, indicating its important role in vivo. The hypomorphic CanB^19G allele distinguishes N. gonorrhoeae from the other Neisseriae, which primarily inhabit the oropharyngeal niche as commensals, and it is also found in N. gonorrhoeae specimens pre-dating the introduction of antibiotics, together suggesting it reflects adaptation to the urogenital niche or sexual transmission.

The two CanB alleles generated distinct metabolic landscapes on which subsequent pressures acted, as CanB^19G facilitates the acquisition of ciprofloxacin resistance when compared with CanB^19E, a process mediated via purine availability. There are three possible explanations for the relationship among purines, GyrA mutations, and CanB mutations. The first is that changes in supercoiling from gyrase mutations may lead directly to transcriptional differences of purine synthesis genes such as purK and thereby cause a purine deficiency and decreased fitness. A second possibility is that gyrase mutations indirectly drive changes in nucleotide pools via changes in mutational burden, leading to a secondary effect on purine synthesis. This hypothesis is supported by the finding that the mismatch repair enzyme MutS is upregulated in the GyrA^91F mutant (Fig. 4C). A third possibility is that the gyrase mutant strains are not purine deficient, but instead purine supplementation has other effects on cellular metabolism, such as aerobic respiration³⁵, supported by the observation that NADH dehydrogenase components are downregulated in this mutant (Fig. 4C). As PurK is bicarbonate-dependent³⁶, any of these three mechanisms would be consistent with a complex interaction between canB and gyrA in N. gonorrhoeae. Future work will be needed to elucidate how gyrase mutations alter the metabolic landscape of the cell.

CanB^19G is therefore an example of a widespread variant that facilitates acquisition of drug-specific resistance through shifts in the gonococcal metabolic landscape and without directly affecting drug susceptibility or tolerance. This work builds on research showing that gene interactions are critical in compensating for the fitness costs incurred by resistance mutations^37–40, shows how pre-existing genetic variation can modulate the emergence and spread of antibiotic resistance mutations, and expands on recent findings that mutations in metabolic genes can change the drug-specific resistance profile of clinical isolates¹⁵. By showing how specific lineages are more likely to acquire specific resistance determinants, our work provides a new perspective for understanding the evolution of antibiotic resistance.

Methods

Bacterial strains, culture conditions, MIC measurements, and figure design

Strains, plasmids, and primers are presented in Supplementary Table 1^41–50. N. gonorrhoeae was cultured on GCB agar (Difco) with Kellogg’s supplements⁵¹ (GCB-K) at 37°C. Strains were grown at concentrations of CO₂ as indicated. Antibiotic susceptibility testing was performed on GCB agar with Kellogg’s supplements via Etest (BioMerieux (azithromycin, ciprofloxacin, benzylpenicillin, ceftriaxone, trimethoprim/sulfamethoxazole) or agar dilution (for trimethoprim and sulfamethoxazole). For plating efficiency experiments, we made serial dilutions, then plated in parallel on GCB agar with Kellogg’s supplements. For anaerobic growth on solid medium, N. gonorrhoeae strains were first grown overnight on GCB agar with Kellogg’s supplement and additional supplementation with 1.2 mM NaNO₂. Growth curves and competitions in liquid culture were performed in GCP medium (15 g/L proteose peptone 3 (ThermoFisher), 1 g/L soluble starch (ThermoFisher), 4 g/L K₂HPO₄, 1 g/L KH₂PO₄, 5 g/L NaCl)²⁷ with Kellogg’s supplement and additional metabolite supplementation as shown in the figures. Doubling time was determined by linear regression on log₂-transformed CFU/mL data from 2–10 hours. Supplementation with metabolites did not affect pH unless otherwise noted. Anaerobic conditions were ensured using an anerobic hood and a gas mix consisting of 80% N₂, 10% CO₂, and 10% H_2. Optical density was measured at 600 nm. E. coli strains were cultured in LB. Figures were designed in Prism (v4.7) (GraphPad) and Illustrator (v26.5) (Adobe). Extended Data Figure 1D was adapted from the ”Lac operon” template,” by BioRender.com (2022, retrieved from https://app.biorender.com/biorender-templates).

Genomics pipeline and GWAS

Genomic analysis was conducted as in published work³. Reads were downloaded from sources also described³. These reads were inspected using FastQC (v0.11.7) (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and removed for poor base quality or if GC content was divergent from expected values (~52–54%). Reads were then mapped to the NCCP11945 genome (RefSeq accession: NC_011035.1) using BWA-MEM (v0.7.17-r1188)⁵². We deduplicated using Picard (v2.8.0) (https://github.com/broadinstitute/picard). Samples with <70% reads aligned or with less than 40% coverage were identified with BamQC in Qualimap (v2.2.1)⁵³ and discarded. Pilon (v1.16)⁵⁴ was used to call variants with a mindepth of 10 and a minmq of 20 and pseudogenomes were generated from Pilon VCFs by including all PASS sites and alternate alleles with AF>0.9, while all other sites were set to “N” and samples with >15% of sites called as missing were excluded. De novo assemblies were created using SPAdes (v3.12.0, with 8 threads, the --careful flag, and paired end reads where available)⁵⁵. Contigs were then filtered for quality (coverage >10X, length >500 bp, total genome size ~2.0–2.3 Mbp). Assemblies were annotated using Prokka (v1.13)⁵⁶. Core genes were clustered using Roary (v3.12, using flags -z -e -n -v -s -i 92)⁵⁷ and core intergenic regions were clustered using piggy (v1.2)⁵⁸. Gubbins (v 2.3.4)⁵⁹ was then used to construct a recombination-corrected phylogeny with the aligned pseudogenomes and the resulting phylogeny was visualized in iTOL (v6)⁶⁰. Metadata, including year, resistance determinants, and available susceptibility data have been collected and published³. For country level data, isolates with MICs that were annotated as greater than or less than that value were set to that value.

For the determination of GWAS phenotypes, thirty strains of N. gonorrhoeae were streaked on GCB agar with Kellogg’s supplements in the presence and absence of supplemental 5% CO₂. These plates were then scored on a binary phenotype of growth/no growth in the absence of CO₂. For the computational aspect of the regression-based GWAS, we used a linear mixed model with a random effect to control for possible confounding by population structure. This linear mixed-model GWAS was run using Pyseer (v1.2.0 with default allele frequency filters)⁶¹ based on 57,691 unitigs generated from the genome assemblies using unitig-counter (https://github.com/johnlees/unitig-counter). The Gubbins recombination-corrected phylogeny was used to parameterize the population structure random effects model in Pyseer. The resulting unitigs were mapped to the WHO F strain reference genome (Genbank accession: GCA_900087635.2) edited to contain only one locus of 23S rRNA using BWA-MEM (modified parameters -B 2 and -O 3). Significant unitigs were annotated through Pyseer’s annotation pipeline. Unitigs were then further analyzed in Geneious Prime (v2021.0, http://www.geneious.com). The pipeline was run using Anaconda (v2021.11) and Python (v3.10.0).

RNA-Sequencing (RNA-seq)

For RNA-sequencing, strains of interest as indicated in the text were grown overnight in duplicate at 37C with 5% CO2 on GCB-K agar. Streaks were then scraped into liquid GCP medium with Kellogg’s supplement and diluted to an OD of 0.1, followed by incubation with shaking at 200 rpm in a 37C incubator with 5% CO2. After the culture reached an OD of 0.2, cultures were pelleted. RNA was purified using the Direct-zol RNA Miniprep Kit (Zymo Research) according to the manufacturer’s instructions. RNAseq library prep was performed by SeqCenter (seqcenter.com). As per SeqCenter, the isolated RNA was treated with DNAse (Invitrogen), and libraries were prepped with the Ribo-Zero Plus kit (Illumina) using Stranded Total RNA Prep Ligation with 10bp IDT for Illumina indices. Sequencing was then performed using a NextSeq2000, producing 2×50bp reads. The resulting reads were then demultiplexed with bcl-convert (v3.9.3) (https://support.illumina.com/sequencing/sequencing_software/bcl-convert.html) followed by quality control and adapter trimming. The resulting demultiplexed reads were then then mapped to the FA19 genome using bowtie2 (v2.4.5)⁶². Mappings were used to generate a vcd to check for additional mutations using bcftools (v1.15)⁶³; none of the isolates had additional mutations. Features were mapped to the annotated FA19 genome using FeatureCounts (from the Subread package, v2.0.3)⁶⁴. Counts were compared using DeSeq2 (v3.15)⁶⁵. Transcripts of interest were filtered by both having a |l2fc| > 1 and p_adj < 0.01 between FA19 CanB^19E and FA19 CanB^19E GyrA^91F, while also having a |l2fc| < 1 and p_adj > 0.01 between FA19 CanB^19G and FA19 CanB^19E. Significance is determined by the Wald test with Benjamini-Hochberg correction for multiple comparisons using default settings via DeSeq2 (v3.15). Data were analyzed via Rstudio (v1.2.1335, rstudio.com).

Undirected transformation of N. gonorrhoeae and determination of CO₂-dependent plating efficiency

For identification of the genetic modulator of CO₂ dependence (determined to be NGO2079), the CO₂-dependent laboratory strain FA19 was transformed with genomic DNA from the laboratory strain FA1090, which is CO₂-independent. Similarly, for identification of an additional mediator of CO₂-dependency (determined to be aasN), clinical isolate NY0195 was transformed with genomic DNA from the clinical isolate GCGS0457 using the following procedure. Liquid transformations were performed according to previously published protocols^27,50. Piliated N. gonorrhoeae were grown on GCB-K agar overnight. The colonies were then resuspended in GCP medium with Kellogg’s supplement, 10 mM MgCl₂, and approximately 400 ng genomic DNA. These suspensions were incubated at 37°C for 10 minutes in the presence of 5% CO₂ then rescued on non-selective GCB-K agar for 4 hours at 37°C with 5% CO₂. The recovered transformants were then re-plated on GCB-K agar and incubated overnight at 37°C without supplemental CO₂. The process was carried out in parallel without gDNA as a control for spontaneous mutation. Finally, colonies were subcultured for further analysis. To determine capacity for natural competence, strains of N. gonorrhoeae were transformed with pSY6, a plasmid conferring an allele of GyrB that confers nalidixic acid resistance⁶⁶, pDR1, a plasmid conferring Kan^R(Supp. Table 1), or a PCR product containing the allele for GyrA^91F/95G amplified using primers DR_380 and DR_383, conferring ciprofloxacin resistance. Finally, serial dilutions of the transformants and negative control were plated in parallel on GCB-K agar plus Kellogg’s supplement either without antibiotics, with 1 μg/mL nalidixic acid, 50 μg/mL kanamycin, or 0.03 μg/mL ciprofloxacin.

To determine the plating efficiency in the presence and absence of CO₂, N. gonorrhoeae was streaked overnight on GCB-K agar at 37°C with 5% CO₂. The resulting streaks were then serially diluted in at least triplicate and plated on GCB-K agar, with parallel dilutions being grown in the presence and absence of 5% CO₂ at 37°C. Plating efficiency was defined as the quotient of CFUs in the absence of CO₂ and CFUs in the presence of CO₂.

Sequencing and structural prediction

Following undirected transformation, genomic DNA from transformants was isolated using the PureLink Genomic DNA Mini kit (Life Technologies). Genomic DNA libraries were prepared and sequenced by the Microbial Genome Sequencing Center (https://www.migscenter.com/) on a NextSeq 2000. Resulting paired-end reads were then merged with BBMerge (v38.84)⁶⁷ and trimmed with BBDuk (v1.0). Reads were then mapped to the parental genomic DNA assembly and variants predicted in Geneious 2021.0 using default settings. Signal sequences were predicted using SignalP 6.0²⁴ and dimeric structures were predicted using AlphaFold (via Colab, https://colab.research.google.com/github/deepmind/alphafold/blob/main/notebooks/AlphaFold.ipynb)²⁵.

The resulting structures were then visualized in PyMOL (v2.5) (https://pymol.org).

CanB variant composition within Neisseria

Variant composition within Neisseria and closely related species at NGO2079/NEIS2004 was aggregated from multiple sources, including a previously published set of 12,111 N. gonorrhoeae isolates⁶⁸, PubMLST (https://pubmlst.org/, using NEIS2004 as reference, to obtain variant composition from N. meningitidis, N. polysaccharea, N. cinerea, N. subflava, N. mucosa, N. lactamica strains) and blastp (https://blast.ncbi.nlm.nih.gov/, using NGO2079 from N. gonorrhoeae isolate FA1090 as query, to obtain variant composition from N. sicca, B. denitrificans, N. elongata, E. corrodens). Genomes from strains representing a diverse set of species, listed in Figure 1C, were assembled as above or downloaded from Genbank and annotated 23S rRNA sequences were extracted from each genome. Each genome contained only a single annotataed 23S rRNA or all annotated 23S rRNA sequences were identical. Finally, a phylogenetic tree was constructed from these 23S rRNA sequences using RAxML (v8, 100 replicates, GTRGAMMA nucleotide model, Rapid Bootstrapping algorithm)⁶⁹. The resulting tree was visualized in iTOL.

Targeted transformation of CanB and AasN

A ~1 kb PCR fragment surrounding the specific SNP of interest in canB was amplified using the primer pair (DR_250/DR_251) (see Supplementary Table 1). For aasN, the specific SNP of interest in aasN was amplified using primer pair (DR_341/DR_342). CO₂-dependent strains were transformed with approximately 600 ng of resulting PCR product and transformants were selected for growth on the absence of CO₂ as above. For selection of CO₂-dependent transformants from CO₂-independent parental strains, liquid transformations were performed as above and colonies were plated overnight on non-selective GCB-K agar. Colonies were then replica plated onto GCB-K agar and grown in parallel in the presence and absence of 5% CO₂. Finally, colonies that appeared to have CO₂-dependence were subcultured for further analysis. Transformation reactions using the alternative canB variant were used as a negative control to account for spontaneous mutation. For kanamycin co-selection, pDR50/51 was constructed using the following primer pairs: (DR_343/DR_344), (DR_345/DR_346), (DR_347/DR_348), (DR_349/DR_350), and (DR_194/DR_195) using Gibson assembly⁷⁰. Liquid transformations were then performed on the above, followed by selection on GCB-K agar supplemented with 70 μg/mL kanamycin. Transformations without DNA were used as negative controls. All transformants were then checked by Sanger sequencing.

To knockout CanB, pDR35 was constructed using the following primer pairs:(DR_266/DR_272), (DR_273/DR_274), (DR_275/DR_276), (DR_277/DR_271), and (DR_194/195) using Gibson assembly and transformed as above. Similarly, to knockout AasN, pDR46 was constructed using primer pairs (DR_359/DR_360), (DR_361/DR362), (DR_363/DR_364), and (DR_194/DR_195). pDR46 was then transformed into N. gonorrhoeae followed by selection on 8 μg/mL chloramphenicol in GCB-K agar and checked by Sanger sequencing. Finally, for complementation of AasN, pDR75, a kanamycin resistant derivative of pMR32 was cloned, using primer pairs (DR_214/DR_215), (DR_216/DR_217), and (DR_218/DR_219) and Gibson assembly. Complementation constructs pDR48 and pDR49 were then cloned from pDR75 using restriction enzyme cloning with PacI/XbaI and an insertion product amplified by primer pair (DR_341/DR_342). These products were transformed into the knockout strains, followed by selection on GCB-K agar supplemented with 70 μg/mL kanamycin.

Competition of CanB variants

Both FA19 CanB^19G and CanB^G19E were transformed with pDR1, a kanamycin cassette derivative (constructed with primer pairs (DR_62/DR_63), (DR_64/DR_65), and (DR_66/DR_67)) of pKH37⁴¹. The resulting transformants were selected on GCB-K agar supplemented with 50 μg/mL kanamycin and the resulting transformants were checked by colony PCR using primer pair (DR_62/DR_63). During competition experiments, FA19 CanB^19G was grown at a 1:1 starting ratio with kanamycin-resistant FA19 CanB^G19E, while in parallel FA19 CanB^G19E was grown at a 1:1 starting ratio with kanamycin resistant FA19 CanB^19G. At each timepoint, cultures were serially diluted and plated on both GCB-K agar and GCB-K agar supplemented with 50 μg/mL kanamycin. Finally, dilutions on both plates were quantified and the proportion of kanamycin resistant bacteria was quantified at each timepoint.

pH or antibiotic kill curves and determination of intracellular pH

Strains were plated overnight on GCB-K agar then scraped after 16–20 hours into liquid GCP medium with Kellogg’s supplement. The cells were spun down in a microcentrifuge (13,500 × g for one minute) then resuspended in GCP medium with Kellogg’s supplement. Cells were then diluted to an optical density of 1.00 and incubated in either pH-adjusted GCP medium with Kellogg’s supplement (or, if noted, Graver-Wade medium⁷¹) or GCP medium with the defined concentration of antibiotics and Kellogg’s supplement at 37°C in the absence of CO₂ in triplicate.

Intracellular pH was determined by using the BCECF-AM dye (ThermoFisher). Strains of interest were grown overnight on GCB-K agar with Kellogg’s supplements. After 16–20 hours, streaks were resuspended in liquid GCP medium with Kellogg’s supplement and diluted to a starting OD of 0.2. These strains were then grown in parallel in the presence and absence of 5% supplemental CO₂ for four hours in triplicate. Cells were then washed in PBS, resuspended in PBS with 1 μg/mL BCECF-AM dye, and incubated for thirty minutes. Finally, cells were measured for excitation/emission at both 440nm/535nm and 485nm/535nm in a plate reader. The fluorescence at both excitation/emission wavelengths from a control with dye but no cells was then subtracted from the measured 440nm/535nm and 485nm/535nm values. Finally, as the 440nm/535nm value is a proxy for cell density, the final (485nm/535nm)/(440nm/535nm) ratio was compared to a normal curve created using laboratory strain FA19 in the presence of Graver-Wade medium at multiple pHs and in the presence of 30 μg/mL of the decoupler CCCP (Sigma).

Intracellular macrophage survival

RAW264 macrophages were plated overnight in a 12-well plate at ~1 million cells/well in DMEM with 4 mM L-glutamine, 1.25 mM sodium pyruvate, and 10% heat-inactivated FBS (ThermoFisher). Cells were then infected with the strains of interest at an MOI of ~100. The plate was then centrifuged at 500 x g for 4 minutes to promote adherence and was incubated for 1 hour at 37°C with 5% CO₂. Cells were then washed three times in DMEM and finally replated in the supplemented DMEM with 100 μg/mL gentamicin. At 0, 1, 2, and 4 hours, three wells for each bacterial strain were resuspended in 300 mL of 0.01% Triton-X-100 in sterile PBS by vigorous pipetting. The Triton-X-100 solution was incubated for ten minutes then vortexed vigorously, followed by dilution plating on GCB-K agar to determine the total intracellular bacterial burden.

Knockout of can and complementation of CanB variants in E. coli

pDR55, a derivative of the sacB allelic exchange vector pRE107⁴⁴, was cloned with primer pairs (DR_404/DR_405), (DR_406/DR_407), (DR_408/DR_409), and (DR_410/DR_411), to insert into can and transformed into the E. coli λpir DAP auxotroph β2155. The plasmid was then conjugated into E. coli MG1655 by incubating the two E. coli strains together at a 1:1 ratio on LB with 0.3 mM DAP in the presence of 5% supplemental CO₂. The resulting transformants were then plated onto LB without DAP in the presence of 5% supplemental CO₂. Finally, the resulting transformants were checked by colony PCR using primer pair (DR_402/DR_403) for successful knockouts. The resulting strains were then transformed with pDR42 and pDR43, both derived from pMR33⁴² using restriction-enzyme cloning with enzymes PacI/XbaI and canB-FLAG amplified with primer pair (DR_280/DR_306), as well as empty-vector pMR33 and selected on LB agar with 20 μg/mL chloramphenicol. Finally, these strains were assessed for growth in the presence and absence of CO₂ at different levels of IPTG induction.

Proteins were quantified by Western blot. Briefly, MG1655Δcan::pDR42/43 was grown overnight in LB with 20 μg/mL chloramphenicol and 5% CO₂ at 37°C. Each strain was then diluted to OD 0.01 and then grown in the presence of different levels of IPTG inducer. After 4 hours, 1.5 mL of each culture was centrifuged (20,000 x g for 3 minutes) and resuspended in lysis buffer (1% v/v Triton-X-100, 150 mM NaCl, 2 mM EDTA, 20 mM Tris-HCl pH 7.5, 1X v/v cOmplete protease inhibitor (Sigma). After 30 minutes of incubation at 4°C with agitation, lysed cells were centrifuged at 4°C (2,800 x g for 5 minutes). The resulting supernatant was used for further protein analysis. Whole cell extracts were electrophoresed on 4–12 μg/mL Tris-Glycine gels (Invitrogen) and transferred to a PVDF membrane with semi dry transfer. The membranes were then blocked in 5% milk, incubated with primary antibodies to G6PD (1:2500, AssayPro) and FLAG-tag (1:2500, ThermoFisher), then HRP-conjugated goat anti-rabbit secondary antibodies (1:5000, ThermoFisher). Finally, the membranes were visualized with Pierce ECL Western Blotting Substrate (ThermoFisher) and visualized by chemiluminescent and colorimetric imaging.

Fluctuation analysis and gyrase mutants

gyrA^91F and gyrA^91F/95G alleles were amplified by PCR using primer pair (DR_380/DR_383) from clinical N. gonorrhoeae isolate NY0195. These alleles were then introduced into strains FA19 and 28BL using liquid transformation as described above and selected on GCB-K agar with .003 and .006 μg/mL ciprofloxacin respectively.

For fluctuation analysis, N. gonorrhoeae strains were first streaked overnight. The resulting streaks were then resuspended in liquid GCP with Kellogg’s supplement, diluted to OD 0.1, then re-plated onto GCB-K agar with Kellogg’s supplement and plated for CFUs with 6 replicates. After 16–20 hours, each well was resuspended in 200 μl of liquid GCP and agitated with beads. From each well, 40 μl was plated onto overdried GCB-K agar with Kellogg’s supplement and the antibiotic of interest. Colonies were then counted after 24 hours. Furthermore, six wells were used to determine the number of plated CFUs by serial dilution on nonselective GCB-K agar with Kellogg’s supplement. Mutation rates were determined using the R package flan (v0.9, default conditions, plating efficiency 0.2)⁷². For generation of Cip^R mutants, we generated unique mutants using the same method as in the fluctuation assay with plating on GCB-K agar with 0.01 μg/mL ciprofloxacin, followed by Sanger sequencing to validate gyrase mutations.

Extended Data

Extended Data Fig. 1.

(A) Schematic of undirected transformation of FA19 (parental CanB^19G) to identify causative factors for CO₂-dependence (B) Plating efficiency in the presence and absence of CO₂ following undirected transformation of gDNA from N. gonorrhoeae strain FA6140 into the N. gonorrhoeae lab strain FA19 (N = 10, 10, 9 from left to right from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.00018, p = 0.00028) determined by two-sided Mann-Whitney U test. (C) Putative reaction catalyzed by NGO2079/CanB and downstream metabolic products. (D) SNPs present in whole-genome sequencing of two transformants in (B). Arrow indicates the SNP identified in Figure 1. (E) (Top) Alphafold predicted homodimeric structures of the CanB^19E variant (teal) and the CanB^19G variant (green). (Bottom) Magnified Alphafold predicted structures overlaying the glutamate (maroon) and glycine (red) at position 19. (F) Proportion of sequenced isolates by year with the CanB^19G variant. *p<0.05, **p<0.01, ***p<0.001

Extended Data Fig. 2.

(A) Plating efficiency at different CO₂ concentrations for N. gonorrhoeae strains 28BL (parental CanB^19E) and FA19 (parental CanB^19G) with isogenic CanB mutants (N = 6 for 0.1% CO₂, N = 3 for air, from two independent experiments, error bars represent SEM). (B) Competition experiment in the presence of CO₂ between FA19 CanB^19G and FA19 CanB^E19G with (left) the FA19 CanB^19G strain kanamycin-labeled and (right) the FA19 CanB^E19G kanamycin-labeled. The proportion of colony forming units (CFUs) that are kanamycin resistant is graphed against time (N=3, representative of two independent experiments, error bars represent SEM). (C) Similar to (B), timecourse of competition between isogenic CanB FA19 strains in the presence of media supplementation and the absence of supplemental CO₂ (N=3, representative of two independent experiments, error bars represent SEM). (D) As in (C), competition between unlabeled FA19 CanB^G19E and kanamycin-labeled FA19 CanB^19G with selected metabolite supplementation assayed at 16 hours in the absence of supplemental CO₂ (N=6, from two independent experiments, error bars represent SEM). (E) Western blot of E. coli MG1655Δcan complemented with IPTG-inducible CanB variants C-terminally FLAG-tagged with G6PD as loading control (representative of two independent experiments).

Extended Data Fig. 3.

The 19E variant of NGO2079 is not advantaged at moderately low pH or within macrophages. (A) Gentamicin intracellular protection assay in RAW 264.7 macrophages as measured by CFUs (N=3/timepoint, representative of two independent experiments, error bars represent SEM). (B) Competition experiment as in Extended Data 2B in pH-adjusted Graver-Wade media (N=3, error bars represent SEM).

Extended Data Fig. 4.

(A) MICs for sulfamethoxazole, trimethoprim, and trimethoprim/sulfamethoxazole for the FA19 and 28BL pairs isogenic CanB variants. (B) Sulfamethoxazole susceptibility of MG1655Δcan complemented with isogenic CanB variants and induced with IPTG. (C) MICs for clinically relevant antibiotics for FA19 CanB^19G and FA19 CanB^E19G as determined by E-test and agar dilution plating. (D) gyrA mutations for spontaneous ciprofloxacin escapees or clean mutants of FA19 CanB^19G and FA19 CanB^19E, along with associated ciprofloxacin MIC.

Extended Data Fig. 5.

Loss of CanB does not appear to confer a loss of natural competence. (A) Plating efficiency in the presence and absence of CO₂ for a knockout of CanB in the N. gonorrhoeae lab strain FA1090 (N=6, from two independent experiments, error bars represent SEM). Significance (p=0.0006) by two-sided Mann-Whitney U. (B) Natural competence of strains of N. gonorrhoeae as measured by uptake of a nalidixic acid resistance-conferring integrating plasmid (N=3, representative of two independent experiments, error bars represent SEM). (C and D) As in (B), natural competence as measured by (C) uptake of Kan^R-conferring plasmid DR1 and (D) a PCR product containing the allele encoding for GyrA^91F/95G (N=6, representative of two independent experiments, error bars represent SEM). *p<0.05, **p<0.01, ***p<0.001

Extended Data Fig. 6.

(A) Violin plots of drug MICs for ~10,000 N. gonorrhoeae clinical isolates. Black bar represents median, dotted lines represent 25^th/75^th percentiles. (B) Average log₂ of ciprofloxacin MIC in μg/mL by country and by allele of CanB. Weighted average of all datasets is represented by black dots/line. Size of data point represents size of dataset. Error bars represent standard deviation as determined by weighted variance. Significance (p = 0.0120) by unpaired two-sided t-test. (C) Calculated mutation rates of resistance acquisition for to rifampin by fluctuation analysis of isogenic CanB strains. (N=192, representative of two independent experiments, error bars represent 95% confidence interval). Significance (p=0.029) by unpaired two-sided t-test (N=192, representative of two independent experiments, error bars represent SD). (D) Cumulative distribution of mutants in experiments in (C) along with data on the inoculum size and final population size. *p<0.05, **p<0.01, ***p<0.001

Extended Data Fig. 7.

The CanB^19G variant does not affect killing by ciprofloxacin, but CanB^19G provides an advantage in the presence of gyrA mutations across multiple gyrA alleles and strain backgrounds. (A) Kill curve of isogenic CanB FA19 at 2x MIC and 32x MIC (N=3, representative of two independent experiments). (B) Doubling time of replicates in Fig. 3E as determined by linear regression on log-transformed CFU/mL counts from 2–10 hours of growth (N=12, from two independent experiments, error bars represent SEM). Significance (p = 1.9e-5) by unpaired two-sided t-test. (C) Growth of CanB isogenic isogenic 28BL strains bearing ciprofloxacin resistance-determining GyrA^91F/95G (N=3, representative of two independent experiments). Significance determined by unpaired two-sided t-test. (D) Competition between spontaneously ciprofloxacin resistant isogenic CanB strains (see Extended Data 2) and susceptible parental strains (N=3, representative of two independent experiments). (E) Growth curves of CanB isogenic FA19 strains with spontaneous ciprofloxacin resistance-determining gyrA alleles (N=3, representative of two independent experiments). (F and G) Competition between (F) FA19 CanB^G19E and FA19 CanB^G19E GyrA^91F (N=6, except UMP and 100 μM Adenosine, N=4) and (G) FA19 CanB^19G and FA19 CanB^19G GyrA^91F (N=3) after 16 hours along with media supplementation in liquid GCP-K (representative of two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.0002, p = 5.7e-13, p = 3.86e-5), by oneway ANOVA with Dunnett’s multiple comparisons test. *p<0.05, **p<0.01, ***p<0.001

Extended Data Fig. 8.

(A) Plating efficiency in the absence and presence of supplemental CO₂ of clinical N. gonorrhoeae isolate NY0195 following the introduction of mutations in CanB and AasN (N=6, from two independent experiments, error bars represent SEM). Significance (from left to right, p = 0.00034 p = 0.00028) by two-sided Mann-Whitney U test. (B) Maximum-likelihood tree as in Figure 1A overlaid with tracks indicating AasN allele and the ceftriaxone resistance determinant mosaic PenA34 allele. *p<0.05, **p<0.01, ***p<0.001

Data Availability

The datasets generated during and/or analyzed during the current study can be found in the source data associated with this work. RNA-seq reads are available publicly at SRA Bioproject PRJNA869861. The NCCP11945 genome is accessible via RefSeq accession NC_011035.1. Genomic data, metadata, and accession numbers are publicly available in previously published work.⁶⁸