Evolution, persistence, and host adaption of a gonococcal AMR plasmid that emerged in the pre-antibiotic era

Plasmids are diverse extrachromosomal elements significantly that contribute to interspecies dissemination of antimicrobial resistance (AMR) genes. However, within clinically important bacteria, plasmids can exhibit unexpected narrow host ranges, a phenomenon that has scarcely been examined. Here we show that pConj is largely restricted to the human-specific pathogen, Neisseria gonorrhoeae. pConj can confer tetracycline resistance and is central to the dissemination of other AMR plasmids. We tracked pConj evolution from the pre-antibiotic era 80 years ago to the modern day and demonstrate that, aside from limited gene acquisition and loss events, pConj is remarkably conserved. Notably, pConj has remained prevalent in gonococcal populations despite cessation of tetracycline use, thereby demonstrating pConj adaptation to its host. Equally, pConj imposes no measurable fitness costs and is stably inherited by the gonococcus. Its maintenance depends on the co-operative activity of plasmid-encoded Toxin:Antitoxin (TA) and partitioning systems rather than host factors. An orphan VapD toxin encoded on pConj forms a split TA with antitoxins expressed from an ancestral co-resident plasmid or a horizontally-acquired chromosomal island, potentially explaining pConj’s limited distribution. Finally, ciprofloxacin can induce loss of this highly stable plasmid, reflecting epidemiological evidence of transient reduction in pConj prevalence when fluoroquinolones were introduced to treat gonorrhoea.

We show that pConj is largely restricted to the gonococcus and track its evolution from when it first appeared over 80 years ago to now. Even though N. gonorrhoeae is a highly variable pathogen [24] with pConj present in multiple lineages, pConj has been remarkably conserved over this time except for acquisition of tetM and TA systems. The plasmid does not impose obvious fitness costs and is stably inherited, explaining the persistence of pConj after tetracycline was discontinued for treating gonococcal infection. Transposon insertion sequencing (TIS) indicates that pConj does not rely on any non-essential chromosomal gene for its maintenance. Instead, plasmid-encoded TA and partitioning systems act in concert to ensure pConj maintenance. Of note, we found that the orphan VapD encoded by pConj is part of a split TA system with VapX antitoxins encoded by other mobile genetic elements, pCryp or a horizontally-acquired Type IV secretion system (T4SS) genomic island, potentially explaining the restriction of pConj to Neisseria spp.. As the ParAB partitioning system is important for the vertical transmission of pConj, we tested whether ciprofloxacin, which can impair plasmid segregation through its effect on DNA gyrase [25], can be used to cure pConj. Importantly, exposure to ciprofloxacin enhances pConj loss from the gonococcus. This is consistent with the decrease in pConj prevalence that occurred when this antibiotic was introduced for treating gonococcal disease [26,27], and provides proof-in-principle that this highly stable and adapted plasmid can be eliminated from N. gonorrhoeae.

pConj is conserved, restricted to Neisseria spp. and does not impose fitness costs
To investigate the evolution and restriction of pConj, we initially examined whole genome sequences (WGS) of gonococcal isolates deposited in PubMLST (n = 9,618, S1 Table) for the presence of pConj, pCryp and pbla. The earliest sequenced gonococcal isolate dates to 1928 and carries pCryp (isolate DO371). Subsequently, markerless pConj (i.e. lacking tetM) first appeared in 1940 (isolate DO12954) in isolates with pCryp. tetM + pConj was described later in 1983 [28]; shortly after, tetracycline monotherapy ceased in the USA [7] (Fig 1B). The first isolate with pbla (isolate 215/-02) dates from 1979 and harbours pConj and pCryp; this coincided with detection of pbla-carrying isolates in the USA in 1976 [29] (Fig 1B). Spread of pbla led to the discontinuation of penicillin treatment for gonorrhoea [8].
As pConj has existed in N. gonorrhoeae for over 80 years including the pre-antibiotic era, we next examined its evolution by analysing the sequences of 126 representative plasmids from all previously described seven pConj variants [9] in strains from different lineages (S2 Table). Phylogenetic analysis shows that there are two distinct pConj groups, I and II. Group I pConj (comprising pConj variants 4 to 7 [9]) appeared first, carries an ε:z3 TA system, and is markerless or tetM + , while group II pConj comprises variants 1 to 3 which acquired ε:z2 and tetM (Fig 1D).
z2 and z3 share 50-70% amino acid similarity and retain predicted active site residues (S1 Fig). Apart from tetM and ε:z, pConj has remained remarkably conserved: group I pConj from the 2000s are virtually indistinguishable from plasmids circulating in the 1940s (Fig 1D). An exception is a sub-set of recent group II N. gonorrhoeae pConj with mutations in genes involved in mating pair formation (Fig 1D). The conservation of gonococcal pConj contrasts with its diversity in N. meningitidis (S2 Fig). Despite its conservation, pConj is widely disseminated across multiple lineages within the gonococcal population (Fig 1E), consistent with previous reports [9], suggesting that pConj has remained conserved despite evolutionary pressure to the gonococcal chromosome.
pConj might be widespread in N. gonorrhoeae because of the use of tetracyclines to treat gonococcal infection [30]. To examine this, we determined the prevalence of pConj in gonococci isolated in the USA when tetracyclines were recommended as part of a dual therapy for treating gonococcal disease [30] (2010-2014, n = 1,467) and compared this with subsequent years when tetracyclines had been discontinued [31] (2015-2019, n = 1,050, S3 Table). Of note, the prevalence of tetM + pConj did not fall following the cessation of tetracycline use (p = 0.4547, unpaired t-test Fig 2A). Furthermore, the prevalence of strains harbouring markerless pConj has not changed over this time (p = 0.7846, Fig 2A). Therefore, pConj has persisted in the absence of selection imposed by tetracycline treatment for gonococcal disease. Similarly, pConj carriage has been steady in the UK since 2007 (n = 3,240, Fig 2B and S3  Table), even though tetracyclines have not been used to treat gonorrhoea in this country since 2004 [32].
The persistence of pConj in gonococcal populations is mirrored by its stable vertical transmission and lack of fitness costs. To determine the loss of pConj, group I and II pConj containing tetM (variants 4 and 1, respectively) were introduced into N. gonorrhoeae FA1090 and tagged with gfp:kan (generating pC1 GFP and pC4 GFP for variants 1 and 4, respectively) to allow detection of pConj loss at low frequency (i.e. � 0.3%). Introduction of gfp:kan did not affect the growth of strains (S3 Fig). No loss of pC1 GFP or pC4 GFP was observed over 160 generations in the absence of selection (S3 Fig). pC1 GFP was used in all subsequent experiments as variant 1 pConj is common in gonococci [9]. There was no difference in the growth of FA1090 +/-pConj in fastidious broth (FB) or gonococcal base media (GCBL, S3 Fig, two-way ANOVA with Sidak's multiple comparisons, p > 0.05). Therefore, pConj does not impose detectable fitness costs on the gonococcus, even though it accounts for approximately 10% of the genetic content of strains (the plasmid to chromosome copy number is approximately 4.6, S3 Fig). Taken together, pConj is a highly conserved plasmid that is largely restricted and adapted to the gonococcus in which it is stably inherited.

TIS does not reveal host genes for pConj maintenance
The adaptation and restriction of pConj to the gonococcus led us to hypothesise that the plasmid relies on host-specific genes for its maintenance. To identify chromosomal genes involved in pConj maintenance, we constructed a library of FA1090 mutants by in vitro transposon mutagenesis of FA1090 chromosomal DNA followed by uptake of mutagenised DNA via transformation. pConj was then introduced into the library. The resulting library containing >100,000 unique insertion sites (UIS) was grown with or without tetracycline in duplicate cultures for 56 generations; transposon insertion sites were determined using the Tradis-Xpress nucleotide sequencing method [33] after the 8 th , 16 th and 56 th generation. The number of UIS diminished over time, with 183,919 UIS in the initial library, 81,655-89,336 UIS by the 8 th generation and 54,708-70,973 UIS by the 16 th generation. By the 56 th generation, there were only between 5,815 and 6,442 UIS in the cultures, too few to identify genes involved in plasmid maintenance (S1 Dataset). The profile of insertions was highly consistent between cultures (S4 Fig). No chromosomal genes were found to contribute to plasmid maintenance after eight generations (S4 Table). Significant hits (i.e. enriched in cultures lacking tetracycline, log fold change < 0, q < 0.01) were only identified after 16 generations (Fig 3 and S4 Table).
However, all hits were due to differences in one or two insertions per gene (Fig 3), which is characteristic of false positive hits [34]. To determine whether they were false positives, mutants Δneis1066, Δneis1845, Δneis2592, which gave the most significant q-values or highest counts per million (logCPM) identified by Tn-Seq were constructed in pC1 GFP . Mutants were passaged for 16 generations, and plasmid loss was determined by detecting the absence of GFP. No plasmid loss was observed (LOD = 0.3%), confirming that TIS yielded only false positive hits. Therefore, after 16 generations we also inspected the distribution of transposon insertions in/around 16 genes encoding proteases and DNA replication machinery, which influence plasmid maintenance in other bacteria [35][36][37][38]. No differences were detected in the distribution or frequency of insertions around these genes (S4 Fig). Overall, these results suggest that pConj does not rely on a single non-essential chromosomal gene for its stable inheritance.

TA systems co-operate to maintain pConj
As we did not identify any host genes contributing to pConj stability, we next examined plasmid genes. Initially, we determined whether the TA systems encoded by the GL region contribute to pConj stability. We replaced the entire GL region in pConj (neis2203 to vapD inclusive [20], Fig 4A) with gfp:kan, generating pCΔGL, and assessed plasmid loss after growing bacteria for 30-40 generations under non-selective conditions. Under these conditions, loss of pC1 GFP (pConj with the GL region) was below the limit of detection while deletion of the GL region reduced pConj stability (loss of pCΔGL, 8.5% ±3.1, Fig 4B).
To define which TA systems encoded in the GL region contribute to plasmid stability, we next deleted or inactivated the toxins individually (pCtΔvapD, pCt:z1 K115A , pCtΔz2) or in combination (pCtΔvapD:z1 and pCtΔTA, Fig 4C). trbC, which encodes the conjugative pilus, was removed from all plasmids to prevent pConj re-acquisition by conjugation [39]. No pConj loss was detected after growth for 30-40 generations following deletion of vapD or z2. However, inactivation of z1 led to low level plasmid loss (0.33% ±0.3) demonstrating that this TA system contributes to pConj maintenance ( Fig 4C).

Interactions between mobile genetic elements involved in pConj maintenance
To understand the mechanism by which VapD expressed by pConj (VapD pConj ) contributes to plasmid maintenance, we examined whether VapD pConj is toxic. VapD pConj shares homology with Helicobacter pylori [40] and Haemophilus influenzae [22] VapD (34.8% and 34.4% amino acid similarity respectively), and is also predicted to be a dimer by AlphaFold ( Fig 5A). VapD p-Conj expression in E. coli under an arabinose-inducible promoter [41] led to a marked reduction in bacterial survival (p < 0.0001, two-way ANOVA with Tukey's multiple comparisons test, Fig  5B), demonstrating that VapD pConj is toxic. We next investigated if VapD pConj interacts with VapX homologues encoded elsewhere in the gonococcal genome. Of note, pCryp, which was in the first gonococcal isolate found with pConj, encodes a vapXD TA system ( Fig 1B). pCryp, similar to pConj, is restricted to Neisseria spp., and is most common in N. gonorrhoeae (96.8% of isolates, 9,313/9,618, S1 Table). VapXD is also associated with a horizontally acquired chromosomal island containing a Type IV secretion system (T4SS) in some gonococci [9]. Due to the high prevalence of pCryp [9,42], we tested whether expression of vapX pCryp can prevent VapD pConj toxicity. Results demonstrate that VapX pCryp abrogates VapD pConj toxicity (p = 0.0069, Fig 5B), consistent with VapD pConj forming a split TA system with VapX pCryp .

Interplay between TA and partitioning systems stably maintain pConj
Apart from TA systems, partitioning systems can promote the maintenance of low copy number plasmids. We therefore examined the effect of the putative ParAB partitioning system on pConj maintenance (Fig 1A). We generated pConj lacking parB (pCtΔparB), and found that after only 6-9 generations approximately 15% of bacteria had lost the plasmid (Fig 6A). We reasoned that the dramatic impact of ParAB might be masked by the activity of TAs, so additionally inactivated the TA systems in this plasmid (generating pCtΔparBΔTA). This led to almost all bacteria losing pConj over only 6-9 generations (98% ±3.6, p = 0.0022 compared with pCtΔparB, Mann-Whitney test, Fig 6A), demonstrating that pConj TA and partitioning systems operate in concert to prevent plasmid loss.

Ciprofloxacin can promote pConj loss
Given the importance of the ParAB partitioning system, we attempted to eliminate pConj from the gonococcus by exposing bacteria to ciprofloxacin which can affect plasmid segregation by targeting DNA gyrase and/or topoisomerase [44]; spectinomycin was included as a control [45], and pCΔGL was used in these experiments to exclude the effects of TA systems. . Assays were carried out three times and analysed with Mann-Whitney test, LOD = 0.3%, shown as mean ± SD. ** p � 0.01. Treatment of (B) FA1090 or (C) 60755 with ciprofloxacin promotes pConj loss. Bacteria were grown overnight on GCB before incubation in 0.5 MIC of ciprofloxacin or spectinomycin. Colonies (n � 200; LOD = 0.5%) were tested for pConj carriage. Assays were carried out three times and analysed with student's unpaired ttest, shown as mean ± SD. * p � 0.05, *** p � 0.001. (D) As pConj enters N. gonorrhoeae, interactions between pCryp and pConj via VapXD contribute to the stabilisation of pConj in the gonococcus as well as its maintenance. pConj encodes TA and partitioning systems-therefore, pConj has developed multiple fail-safe mechanisms to ensure it is maintained in N. gonorrhoeae. While pConj is stably maintained in N. gonorrhoeae, antibiotic use facilitated spread of both itself and pbla, resulting in the spread of plasmid-mediated antimicrobial resistance. Currently, despite the lack of tetracycline use, pConj persists in N. gonorrhoeae; as such, treatment regimens no longer suggest the use of penicillin or tetracycline. https://doi.org/10.1371/journal.pgen.1010743.g006 We exposed bacteria to 0.5 Minimal Inhibitory Concentration (MIC) of ciprofloxacin or spectinomycin for 48 hours and assessed plasmid loss from FA1090 and 60755, an isolate from Kenya that naturally carries pConj; bacterial survival was similar in the presence of either antibiotics (S5 Fig). Compared with spectinomycin, exposure to ciprofloxacin resulted in significant pConj loss from both strains (for FA1090, 7.6% ±0.3 for ciprofloxacin vs. 1.6% ±1.1 for spectinomycin, p = 0.0009; for 60755, 4.7% ±1.7 vs. 1.3% ±0.7, p = 0.03 unpaired t-test, Fig 6B  and 6C). These data demonstrate that ciprofloxacin, and other related antibiotics, could eliminate pConj from the gonococcus.

Discussion
Gonorrhoea is a major public health problem due to increasing AMR [7], and N. gonorrhoeae carries multiple plasmids with limited host range, including pConj which contributes to the spread of high-level beta-lactam and tetracycline resistance [11]. Here, we show that pConj arose decades earlier than previously appreciated [46], is prevalent, remarkably conserved and imposes no detectable fitness cost on the gonococcus, despite accounting for around 10% of total genetic content. The stable inheritance of pConj relies on redundant and co-operative effects of partitioning and TA systems. The orphan VapD toxin on pConj forms a split TA system with antitoxins encoded on other mobile genetic elements, which may explain the plasmid's restriction to the gonococcus. Finally, we used ciprofloxacin, which interferes with partitioning, to provide a proof-in-principle that this highly adapted, stable plasmid can be cured through understanding mechanisms underlying its inheritance.
Toxicity and antitoxicity assays and epidemiological evidence indicate that pConj and pCryp form a stable association likely due to the interaction between VapD pConj and VapX. This is consistent with the initial appearance of pConj over 80 years ago in a gonococcal isolate with pCryp, and the observation that introduction of pCryp vapX into E. coli increases its ability to acquire pConj by conjugation [20]. The genetic location of vapX may be important, as both N. lactamica and N. meningitidis, which have low pConj carriage, do not carry vapX on the high-copy-number plasmid pCryp found in N. gonorrhoeae [47]. Furthermore, there could be limited exposure of both N. meningitidis and N. lactamica to pConj as they occupy different niches to N. gonorrhoeae [48]. The relationship between VapD pConj and VapX pCryp offers a potential explanation for the restricted distribution of pConj to Neisseria spp., even though its P1 replicon is typical of broad host-range plasmids [49].
Since its acquisition by the gonococcus, pConj has been remarkably conserved, unlike most other AMR plasmids [50,51]; major changes in pConj are limited to acquisition of tetM, ε:z and trbJ/K/L, which contrasts with how the gonococcal genome has evolved rapidly under antibiotic pressure [24]. tetM was previously shown to be acquired from a streptococcal transposon Tn916 [52] and was suggested to be acquired independently twice, resulting in two tetM alleles [53]. Both group I and II pConj are prevalent and inherited stably, suggesting that both ε:z2 or ε:z3 promote pConj maintenance; differences in their activity is under further investigation. We also identified the sub-group in group II pConj as the lineage-specific pConj variant 2 previously characterised [9] and hypothesised that the extensive mutations could be due to horizontal acquisition or lineage-specific evolution of trb genes. Regardless, the long-term success of pConj is evident from its continued prevalence after tetracycline treatment for gonococcal infection ceased. Although gonococci could still be exposed to tetracyclines given to treat co-infections with Chlamydia trachomatis [54], this does not explain the persistence of markerless pConj. Of note, all analyses of pConj prevalence and emergence are dependent on collections and datasets available, thus increased sampling and collection from clinics in the future is paramount to understanding pConj spread and maintenance in the future.
Host genes can influence the fitness cost of plasmids [55], and plasmid-host co-evolution often results in plasmid stabilisation [56]. Our TIS analysis did not reveal any chromosomal factor required for pConj maintenance even after inspection of genes implicated in plasmid replication and segregation in other species [35][36][37][38]. We confirmed that this is unlikely due to reacquisition of pConj during liquid growth, as conjugation is low during liquid culture (conjugation rate, 2.6 x10 -3 over 24 hr). However, TIS cannot assess the role of essential genes or the concerted effect of multiple genes. Additionally, the number of UIS decreased over time, limiting the ability of TIS to detect significant hits in later generations. Instead, we found that the plasmid possesses TA and partitioning system for its stable inheritance. This has significant implications for understanding the persistence of low copy plasmids which express both TA and partitioning systems; plasmids such as pConj are segregated by partitioning systems during cell division, with the TA systems acting as a fail-safe mechanism should segregation fail. In this way, pConj appears to be entirely self-sufficient for its high prevalence in the gonococcus.
As our findings highlighted the importance of ParAB in pConj stability, we reasoned that the plasmid could be eliminated by drugs such as ciprofloxacin which interfere with DNA segregation [57]. We were able to eliminate pConj in a proportion of bacteria by exposing them to sub-lethal concentrations of ciprofloxacin. These findings mirror the observation that pConj carriage decreased following the emergence of ciprofloxacin resistance due to its introduction to treat gonorrhoea [26,27]. Furthermore, it provides proof in principle that molecules targeting TA/partitioning, including next-generation gyrase inhibitors, could be used to eliminate pConj from the gonococcus.
Overall, we show how plasmid-mediated resistance has evolved and is stably inherited in the high priority pathogen N. gonorrhoeae (Fig 6D). We found previously that tetM + pConj is more frequently found in LMICs than in wealthier countries, probably due to extensive and inappropriate use of antibiotics [9]. However, stopping tetracycline treatment has not reduced pConj prevalence, highlighting the exquisite adaptation of this plasmid to the gonococcus. Our work also provides a further example that AMR might not be eliminated by simply reducing antibiotic use. However, we show that it might be possible in the future to eliminate pConj, the main driver of gonococcal plasmid-mediated AMR, by understanding then targeting plasmid maintenance systems.

Gonococcal strain construction
Primers used in this study are listed in S6 Table. In general, gene deletion was achieved by amplifying approximately 1 kb up-and downstream of the target gene and fused with an intervening marker for selection and a flanking vector by Gibson assembly (New England Biolabs). The assembly reaction was then transformed into E. coli DH5α for selection and sequencing, before transformation into E. coli.
The gfp:kan cassette was generated from 280 bp of the opaB promoter [60], sfGFP [41], and the kanamycin resistance cassette aph(3)-Ia [39]. The insertion of the gfp:kan cassette included a duplication of the intergenic region between vapD and res which contains a predicted terminator to flank the cassette. The z1 K115A substitution [21] was introduced by site-directed mutagenesis of AAA to GCC at nt. 343-345 of z1; fragments up-and downstream of z1 K115A were amplified from pC GFP using primers 259/284 and 253/283, respectively.
Constructs were amplified from plasmids with relevant primers (S6 Table) before transformation into gonococci; between 100 ng and 1000 ng of each linear construct was spotted to solid media and left to dry. Bacteria were streaked over the DNA or water control dry spots and incubated for 8-10 hours at 37˚C in 5% CO 2 . Transformants were selected by plating on solid media containing appropriate antibiotics, and confirmed by sequencing.
For construction of FA1090 carrying wildtype pConj variant 1 or 4, FA1090 was conjugated with strains 60755 or 55496 respectively for 24 hours, before selection on tetracycline as previously described [39]. Individual colonies were then picked, and the porB gene was sequenced for strain confirmation.

Plasmid stability and fitness costs
For plasmid stability assays, gonococcal strains were grown overnight on solid media and used to inoculate 15 ml FB at an OD 600 0.001. Cultures were passaged every 24 hours, starting at OD 600 0.001; approximately 300 individual colonies were plated every two passages (20 generations) to assess pConj loss using a UV transilluminator to visualise fluorescence (Invitrogen), resulting in a detection limit of 0.3%. After 160 generations, 40 individual colonies were restreaked on media with or without tetracycline to confirm pConj carriage.
For assessing fitness cost, strains were grown overnight on solid media then used to inoculate GCBL at an OD 600 0.1 in 96 well plates. The OD 600 was measured every 20 min over 8 hours using a FLUOstar Omega Microplate Reader (BMG Labtech), and plated out at 8 hours; 40 individual colonies were re-streaked on media with or without tetracycline to determine pConj carriage. Separately, for assays > 8 hours, strains were inoculated into 15 ml FB at a concentration of 5 x 10 4 CFU/ml and sampled over 24 hours to follow bacterial viability.

Plasmid copy number
To generate standards for assessment of pConj copy number, a 160 bp fragment of chromosomal recA was amplified with Q7 and Q8, and primers 121 and 119 were used to amplify a 965 bp fragment containing res from pConj. These fragments were cloned into pCR 2.1-TOPO (Thermofisher), which was used as a control plasmid. Plasmid concentration was measured using Qubit, serially diluted and added to qPCR mastermix for qPCR. Primers Q7/ Q8 for recA and Q21/22 for res were used to generate standard curves, and amplify a 160 bp and a 130 bp fragment, respectively. All primer sequences are listed in S6 Table. Primer efficiencies, defined as the fraction of target molecules that are amplified in one PCR cycle, were 102.4 and 90.1% for recA and res, respectively.
To measure pConj copy number, bacteria were grown in 5ml GCBL until OD 600 0.5-0.8. Two 10 μl samples were collected, washed in nuclease-free water, then resuspended in 100 μl nuclease-free water and boiled at 95˚C for 5 min. For qPCR, 1 μl of purified control plasmid or lysate was added to 19 μl of qPCR mastermix containing SYBR GREEN (Applied Biosystems), and 0.5 μM of each primer. Reactions were run on a QuantStudio Real-Time PCR System (Thermofisher). Ct values were compared to standard curves as described previously [61], and the relative plasmid copy number ratio (P:C ratio) determined as the ratio of res to recA.

Plasmid maintenance assays
Bacteria were grown overnight on solid media containing either kanamycin or tetracycline to ensure all colonies contain pConj, then used to inoculate 5 ml FB at OD 600 0.01. Cultures were passaged every 20 hours as required to a final 30-40 generations. For shorter assays, two individual colonies were used to inoculate 5 ml FB and grown for 20 hours (6-9 generations). To detect plasmid loss, strains with pConj harbouring the gfp:kan cassette were serially diluted, plated, visualised with a UV transilluminator (Invitrogen), and the number of green and white colonies counted (approximately 300 colonies in total). In every experiment, all white colonies and 10-20 green colonies (chosen at random) were replica plated onto GCB plates containing tetracycline or kanamycin to confirm the absence/presence of pConj.

Transposon insertion sequencing (TIS)
A library of N. gonorrhoeae FA1090 mutants was generated by in vitro mutagenesis. pHsk1 was constructed by Gibson assembly to contain the kanamycin resistance gene flanked by Hsmar inverted repeats, with backbone of pCE005 (gift from Dr Cara Ellison) carrying erythromycin resistance. In each 100 μl transposition reaction, 30 μg genomic DNA was subjected to mutagenesis with 125 nM Hsmar transposase (gift from Professor Ronald Chalmers) with 5 μg pHsk1 in reaction buffer (20 mM Tris-HCl pH 8, 100 mM NaCl, 10% v/v glycerol, 2 mM DTT, 2.5 mM MgCl 2 ) for 18 hours at 30˚C. Products of eight transposition reactions were pooled, precipitated by the addition of 10 μl 3 M sodium acetate pH 5.2 and 800 μl isopropanol, resuspended in ddH 2 O, and repaired in T4 DNA ligase buffer (50 mM NaCl, 0.5mM dNTPs) with 50 U T4 DNA polymerase (Thermofisher) and 10 U T4 DNA ligase (Invitrogen) at room temperature for 90 min. The repaired DNA (375 ng in 25 μl) was used to transform N. gonorrhoeae in 81 independent transformations.
The mutant library was used as the recipient in conjugation with FA1090 containing pConj. The library and donor were grown to OD 600 0.6-0.8 in 20 ml GCBL, mixed at a 1:1 ratio, and added as 10 μl spots onto GCB plates before incubating for 24 hours. A total of 72 conjugations were performed. Transconjugants were harvested from plates containing kanamycin and tetracycline, and grown in 15 ml GCBL in duplicate, with or without tetracycline, for 20 hours starting at an OD 600~0 .3. Every 12 hours bacteria were subcultured at OD 600 0.1 in 15 ml FB with kanamycin +/-tetracycline. In total, bacteria were passaged 14 times (approx. 56 generations) over 7 days; the tetracycline concentration was increased to 10 μg/ml for the last passage. Genomic DNA from 3 x 10 9 bacteria from each condition was extracted using Wizard Genomic DNA Purification kit (Promega, USA) every 24 hours. Nucleotide sequences were determined for genomic DNA from the initial library, 2 nd , 4 th and 14 th passage (generations 0, 8,16,56) using the TraDIS-Xpress sequencing method [33].
For sequencing library preparation, the Nextera Tagmentation Enzyme kit (Illumina) was used. Custom oligonucleotides (S6 Table) were used instead of i5 index primers with 28 PCR cycles to add sequencing primer sequences and indexes [33,62]. Fragment sizes were checked using a 4200 TapeStation System (Agilent). The resulting DNA was purified using AMPure XP beads (Beckman Coulter), then sent for paired-end sequencing on a NextSeq500 machine (Illumina) using a High Output PE 150 (150 cycles), mixed with standard whole genome sequencing (WGS) shotgun bacterial libraries at Quadram Institute (Norwich, UK).
For the stability assays, bacteria were grown overnight on solid media containing erythromycin and tetracycline then harvested and inoculated into 5 ml FB + 1% vitox at an OD 600 0.1. Cultures were passaged every 12 hours for 16 generations. Plasmid loss was detected as previously by the detection of GFP visualised with a UV transilluminator. As all colonies were green, 20 green colonies (chosen at random) were replica plated onto GCB plates containing tetracycline or GCB to confirm the presence of pConj.

Minimal inhibitory concentration assays
Minimal inhibitory concentrations (MICs) were determined in accordance to previously published protocols [63]. Briefly, strains were grown overnight on GCB agar then resuspended in GCBL to OD 600 1.0; 10 4 CFU in 5 μl were spotted onto GCB plates containing the appropriate antibiotics at two-fold dilutions. Once spots were dried, plates were incubated for 20-24 hours before MIC determination.

pConj curing
Strains carrying pCΔGL with gfp:kan were grown on non-selective GCB agar for 18-22 hours before starting liquid cultures at 5 x 10 4 CFU/ml in 5ml FB containing 0.5 MIC of each antibiotic. Cultures were then left static for 48 hours as described previously [25], and mixed well before plating on GCB. Approximately 200 colonies were inspected for loss of plasmid using a UV transilluminator (Invitrogen), resulting in a limit of detection of 0.5%. All white colonies and 10-20 green colonies were picked for replica plating. Cultures were also serially diluted at the end of the experiment and 5 μl were spotted onto GCB plates for CFU determination.

Phylogenetic analysis
Isolates (n = 126) representing all seven pConj variants (S2 Table) were randomly selected from gonococcal pConj-carrying strains on the PubMLST database [64]. The Tn916 transposon sequence was identified by first aligning the consensus sequence of all seven pConj variants; tetM-containing sequence and its transposon was removed from the analysis. For generation of N. meningitidis pConj phylogenetic tree, all isolates carrying pConj on a single contig (35 kb-44 kb, n = 70) were selected. pConj sequences were aligned with the Multiple Alignment using Fast Fourier Transform (MAFFT) program on Galaxy (www.usegalaxy.org). RAXML-NG (version 1.1.0) [65] using the GTRGAMMA model of nucleotide substitution was used to generate a phylogenetic tree. The program was run thrice independently with the default settings until either convergence was reached, or 1000 replicates were completed. The resulting tree was then visualised on MEGA-X (version 11.0.11).
For visualisation of pConj dissemination within the gonococcal population, N. gonorrhoeae isolates from S1 Table were selected on PubMLST [64]. Genetic relationships between gonococcal isolates were determined through the core genome multilocus sequence typing scheme (Ng_cgMLST) then visualised using a minimum spanning algorithm using Grapetree software [66] as done previously [9]. The presence of pConj in each isolate was determined through genes unique for this plasmid (traM and trbC) [9].

Analysis of plasmid distribution
Data analysis involving pConj, pCryp and pbla carriage in all Neisseria spp. populations was carried out using isolate sequences available on the PubMLST database [64]. All isolates were confirmed to have a sequence bin > 1 Mb and contigs � 200 to ensure that sequences available were complete and well sequenced. For UK and US analyses, isolates prior to 2000s were not analysed due to inadequate number of samples. Presence of traM or trbC denoted pConj carriage (as above), while presence of mobC or repA denoted pCryp presence. Presence of pbla was denoted by the bla TEM gene (neis2357). BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) was used to analyse for presence of sequences similar to pConj in non-Neisseria species.
Statistical significance using either unpaired t-test, one-way ANOVA and two-way ANOVA with multiple comparisons if required (as indicated in figure legends) and normality tests for assays were performed with GraphPad PRISM v 9.0. Odds ratio was also similarly calculated using default parameters on PRISM v 9.0. Supporting information S1 Fig. Alignment of z3 and z2 amino acid sequences. The two most prevalent protein variants of each z3 and z2 (with their respective neis2213 alleles in brackets) were used for the alignment. Two other characterised Streptococcus pneumoniae z proteins were included in the alignment (PDB numbers, 1GVN and 2P5T, are in brackets). Shading intensity is proportionate to percentage conservation. All z protein variants have conserved active site residues (catalytic residue and the corresponding amino acid on 1GVN is indicated by the arrow). (TIF) S2 Fig. N. gonorrhoeae pConj is conserved compared to pConj in N. meningitidis. Phylogenetic tree of N. meningitidis and N. gonorrhoeae pConj drawn separately but at the same scale. Each dot represents an isolate, colour-coded according to year of its isolation. (TIF) S3 Fig. pConj is a low copy plasmid that does not impose a fitness cost. (A) gfp:kan cassette on pConj does not affect growth rates. FA1090 carrying the respective plasmids were first grown overnight before liquid culture was set up using gonococcal base media (GCBL) and monitored over 8 hours. (B) No pConj loss was observed with FA1090 carrying pC1 GFP or pC4 GFP . FA1090 +/-pConj grows at the same rate in (C) fastidious broth and (D) GCBL over 24 and 8 hours respectively. (E) Ratio of pConj to chromosome is 4.60 ±1.63, as determined by qPCR of res in pConj and recA on the chromosome using bacteria grown to mid-log phase. Results of three independent experiments were analysed with two-way ANOVA with Sidak's multiple comparisons and shown as mean ± SD. ns, p > 0.05. (TIF)   Table. List of statistical output from Bio-Tradis for all genes from Bio-Tradis at passages (A) 2 and (B) 4 comparing the +Tet and control conditions. LogFC refers to log fold change of the +Tet condition compared to the control condition, LogCPM refers to the logarithmic count per million reads, while the q value refers to the corrected p value. Available to view or download from https://figshare.com/s/94a4394060ce5d13a07a (XLSX) S5