Restriction-modification systems are required for Neisseria gonorrhoeae pilin antigenic variation

19 Restriction, modification, methylation, antigenic variation, Neisseria, gene conversion 20 21 22 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 2

23 Neisseria gonorrhoeae (Gc) is the causative agent of gonorrhea, a sexually transmitted infection 24 that remains a global health concern due to its development of antimicrobial resistance 1,2. Gc is a 25 human-restricted bacterium whose adhesion and invasion rely on pili and outer membrane 26 proteins (reviewed in 3). Among them, the type IV pilus (T4p) is a critical virulence factor that 27 facilitates bacterial colonization 4–6. Gc type IV pilin antigenic variation (Av) results in the existence 28 of multiple versions of its major subunit, pilin (or PilE) 7 within a lineage. During pilin Av one of ~19 29 silent copies (pilS) replaces the analogous sequences in the expressed pilE gene, while the pilS 30 gene remains unchanged 8,9. This gene conversion process alters the PilE amino acids sequence, 31 which generates a variable type IV pilus (T4p). Pilin Av can result in a fully expressed variant pilus, 32 reduced T4p expression, or a completely nonpiliated cell 10–12. This extensive variability may 33 promote persistence within the host, or more importantly, allow reinfection of a person who was 34 previously colonized by the same bacterium 13,14. 35 The pilE gene itself consists of a conserved region followed by semi-variable and hyper-variable 36 regions that share sequence identity with the pilS copies. There are two conserved elements called 37 cys1 and cys2 surrounding the hypervariable loop that form a disulfide bond in pilin. The pilE ORF is 38 flanked downstream by the SmaCla repeat, which is also found at the 3' end of all pilS loci. The 39 regions cys1, cys2, and SmaCla are involved in pilin Av 15,16. Pilin Av also involves transcription of the 40 gar (G4-associated RNA) noncoding sRNA upstream of pilE to form a DNA:RNA hybrid (R-loop) 17–19. 41 This R-loop captures the C-rich strand, enabling the formation of a guanine quadruplex (G4) 42 structure 17. Mutation of either the gar promoter or the G4-forming sequence through single-base 43 mutations abrogates pilin Av. 44 Pilin Av recombination tracts are defined by the first and last nucleotide changes that differ from the 45 starting recipient pilE sequence, and these recombination tracks are flanked by regions of 46 microhomology shared between pilE and the pilS donor 12,20. While pilin Av requires the recA gene 21, 47 and the RecF-like pathway genes recO, recR, recJ, and recQ 22–24, the 10 to 500 bp regions of variant 48 sequence transferred during pilin Av suggest that pilin AV requires other reactions besides 49 homologous recombination. We have previously proposed that pilin Av involves two steps: an initial 50 recombination between the pilE gene and a pilS copy at a region of microhomology, followed by 51 homologous recombination of the hybrid pilE-pilS intermediate into an intact pilE 20. 52 Restriction-modification (RM) systems are widespread in prokaryotic genomes, as 83% of available 53 genomes contain at least one of these modules 25. There are four major types of RMs 26, and type II 54 RMs are the most abundant, accounting for 39.2% of bacterial genomes 25. Gc isolate FA1090 55 possesses 15 RM systems 27–31. The 11 FA1090 type II RM systems each include a sequence specific 56 restriction endonuclease and a DNA methyltransferase. The endonuclease cleaves double-57 stranded DNA at its target motif unless the methyltransferase protects the site 32–34. Several host-58 restricted bacteria, such as Helicobacter, Haemophilus, Streptococcus, and Neisseria, all encode 59 multiple RM modules, despite having relatively small genomes. Being host-restricted, we predict 60 these organisms have fewer bacteriophage predators as compared to environmental bacteria; thus, 61 it is unknown why these organisms have multiple RM modules 35. In this work, we demonstrate that 62 specific 5’-CCGG sequences are frequently located at the microhomology bordering the 63 recombinant pilin Av sequences. We show that the pilE and pilS 5’-CCGG sites, as well as the 64 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 3 paralogous RM modules, are required for efficient pilin AV. Surprisingly, most genomic 5’-CCGG 65 sequences are undermethylated and a subset are cut by the paralogous restriction endonucleases. 66 Restriction of 5’CCGG sites in a subset of chromosomes results in a Gc fitness defect. These 67

indicate that Gc has co-opted these RM modules to facilitate pilin Av, albeit at the cost of 68 reduced fitness. 69 70

71 72 A 5’-CCGG site is prevalent at the borders of pilE recombination tracts 73 Gc isolate FA1090 encodes 15 RM systems, which are conserved in most Gc isolates. Among all the 74 putative RM systems present in FA1090, the recognition sites that are the most represented in the 75 genome are 5’-GCSGC (16173 sites) and 5’-CCGG (11103 sites) (Table S1). Some sites overlap, 76 such as 5’-GCCGGC, which also contains 5’-CCGG. 77 The 5’-CCGG motif shows a differential distribution across the Gc genome, with a frequency higher 78 than one site per 100 bp in 167 out of the 2,211 annotated ORFs (Supplementary Table 3). All 19 pilS 79 copies and the pilE variable region were among these high-density loci. In contrast, 5’-CCGG sites 80 were completely absent from the garP/G4 sequences and the pilE conserved N-terminal coding 81 sequences (Fig. 1A, Table S2). 82 83 84 85 86 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 4 Fig. 1. A restriction and modification target sequence is overrepresented in the pilE gene. A. 87 Number of 5’-CCGG sites across Gc FA1090 genome per 1000 bp bin. B. Zoom-in on the pilE region 88 from panel A. C. Distribution frequency of 5’-CCGG sites per 1000 bp bin across the genome. D. A 89 schematic of the pilE locus. The cartoon shows the garP sRNA promoter, the guanine quadruplex-90 forming sequence (G4), the pilE open reading frame that includes the conserved N-terminus coding 91 sequences and the semi-variable and hypervariable regions (that are also in the pilS copies), 92 containing the conserved cys1 and cys2 repeats, followed by the SmaCla repeat. The SmaCla 93 repeat is found at the end of each pilS locus. The 5’-CCGG is the most frequent in the pilE locus and 94 is found only within the pilE variable region. E. Quantification of the RM target sites at pilin variant 95 recombination tracts that were identified in the analysis of FA1090 pilin variants12. The 5’-CCGG 96 motif is present at the site of recombination in 82% of the variant strains. 97 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 5 The pattern of variant pilE sequence changes has been extensively studied 21,36–38. We reanalyzed 98 100 previously described variant pilE sequences, identified their pilS donors, and mapped each 99 recombination tract 12. The sequences bordering the recombination tracts upstream and 100 downstream of the pilS insert indicate where the recombination event occurred. 82% of the 101 analyzed pilE variants contained a 5’-CCGG site at one or both recombination track border (Fig. 1B). 102 In contrast, the six pilE 5’-GGCC sites were only localized at the recombination tracts border of 37% 103 of the variant strains (Fig 1A, B). Based on this reanalysis of published sequencing data, we 104 postulate that the 5’-CCGG sequences could have a role in pilin Av. 105 The pilE 5’-CCGG sites are important for pilin Av 106 There are several methods for measuring pilin Av frequencies 22,39. One method is the pilus-107 dependent colony morphology change (PDCMC) assay, which records the number of nonpiliated or 108 underpiliated blebs that emerge from a piliated colony over time 22. When pilus phase variation 109 occurs by pilin Av, they are captured by the PDCMC score 22. One limitation of this assay is that it is 110 sensitive to growth rates since slower growth decreases antigenic variation 39,40. We introduced 111 silent mutations that do not change the pilE coding sequence in all four 5’-CCGG sites (strain 112 SM598). The PDCMC assay of strain SM598 with the mutated 5’-CCGG sites revealed a significant 113 decrease in pilin Av (Fig. 2 B, E). After 30h of growth, the colonies of the pilE CCGG-disrupted 114 mutant strain SM598 did not show any significant growth defect (Fig. 2A). Since the decrease in 115 SM598 pilin Av frequencies is not due to decreased growth. These results support the hypothesis 116 that the sites found at the borders of recombination tracts are important for pilin Av (Fig. 2A). 117 Despite the lower level of variation in the pilE 5’-CCGG mutant, some pilin variants were observed. 118 Analysis of individually derived nonpiliated or underpiliated (P-) variants of SM598 revealed that 119 11/30 had a pilE gene deletion. Some P- variants (5/30) had the parental pilE sequence, indicating 120 another process than pilin Av produced the nonpiliated phenotype. The remaining 14 variants were 121 pilin antigenic variants. In these variant sequences, reconstituted 5’-CCGG sites were found in the 122 pilS inserts, but not outside of the recombination tract. However, ten out of twelve pilS variant 123 sequences contained 5’-CCGG sites at the variant border. Six had sites on both sides of the insert, 124 while four had a site on one side. This analysis suggests that 5’-CCGG motifs in pilS may also 125 contribute to pilin Av, even when they are no longer present in pilE. 126 pilS copy 5’-CCGG Sites are critical for pilin Av 127 We tested whether the 5’-CCGG sites in the pilS donors are also involved in pilin Av. Given the 128 presence of 19 pilS copies in the genome, mutating all 5’-CCGG sites within these loci proved 129 challenging. Therefore, we generated a strain where all 19 pilS copies were deleted (the pilS 130 hexadeleted mutant, SM564, Fig. 2D). We then reintroduced an unaltered pilS3C1 copy into its 131 original locus (Strain SM567) and a pilS3C1 copy with five mutated 5’-CCGG sites (strain SM570). In 132 these original and mutated pilS3C1 strains, we also introduced the mutated pilE with or without 5’-133 CCGG sites (respectively, strains SM584 and SM585) (Fig. 2D). 134 Because the frequency of pilin Av is lower with a single pilS copy, we could not use the PDCMC 135 assay to measure pilin Av. Instead, we used a modified PCR-based sequencing assay to measure 136 pilin Av 39,41. The pilS3C1-mutated SM564 strain did not exhibit any detectable pilE variation (Fig. 137 2E), while the non-mutated pilS3C1 strain SM567 showed typical pilin Av frequency (Fig. 2E). 138 Mutating the pilE 5’-CCGG motifs with the nonmutated pilS3C1 donor also reduced pilin Av 139 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 6 (SM584). Finally, disruption of the 5’-CCGG motifs in both pilE and pilS reduced the level of 140 antigenic variation to undetectable levels (SM585), identical to the hexadeleted control (SM564, Fig. 141 2E). These results show that 5’-CCGG sites in both the donor and recipient gene are critical for pilin 142 Av. 143 144 145 146 Fig. 2. Pilin antigenic variation levels of the parental strain recA6 A. Pilin antigenic variation 147 levels of pilE CCGG disrupted mutant (SM598) measured by the surrogate PDCMC assay in the 148 absence or presence of 1 mM IPTG. B. Colony forming units (CFU) per colony at 30h of growth in the 149 parental strain and the pilE CCGG-disrupted mutant (SM598) in the presence and absence of 1 mM 150 IPTG. C. Pictures of colonies of the parental recA6 strain and the pilE CCGG-disrupted mutant 151 (SM598) after 32h of growth on GCB + IPTG. D. Schematic representation of the pilE (±CCGG 152 disruption) and pilS (±CCGG disruption) constructs in a pilS hexadeleted mutant (SM564). E. PCR-153 based assay using Illumina sequencing to measure the variation level in a pilE amplicon. 154 Approximately 500 colonies of the pilS hexadeleted strain were grown with 1mM IPTG. pilE PCR was 155 performed, and Illumina sequencing of the native or mutated pilE and/or pilS loci. Fischer’s LSD 156 uncorrected test was used; only p values <0.05 were plotted. 157 158 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 7 Two RM Modules targeting 5’-CCGG are required for Pilin Av 159 The pilin Av phenotypes of the pilE and pilS 5’-CCGG mutants suggested that these palindromic 160 sites are recognized by factors that are critical for pilin Av. Gc encodes several RM systems 27,42. We 161 identified three type II RM modules that are predicted to recognize 5’-CCGG and 5’-GCCGGC 162 sequences. Two of these methylases were previously reported to methylate their predicted sites: 163 m.NgoAXIV (RM1M, 5’-CCGG) and m.NgoAIV (RM3M, 5’-GCCGGC) 27. The m.NgoAXIII (RM2M) 164 methylase is predicted to recognize 5’-CCGG 42, but this prediction has not been experimentally 165 confirmed (Table S1). The ngoAXIV (RM1) operon encodes a restriction enzyme with two predicted 166 ORFs (RM1R1 and RM1R2) and a complete methylase (RM1M), while the ngoAXIII (RM2) operon 167 includes a full restriction enzyme (RM2R) and a methylase split into two predicted ORFs (RM2M1, 168 RM2M2) (Fig. S2). Analysis of 47 complete genomes from PubMLST database shows that both 169 operons are fully conserved across all analyzed Gc strains. Notably, 71% of Neisseria meningitidis 170 genomes, a closely related species that also undergoes pilin Av, encoded a 5’-CCGG restriction 171 enzyme (RM1R). In contrast, neither RM1R nor RM2R was detected in Neisseria lactamica, which 172 does not undergo pilin Av (Supplementary Table 2). 173 To test the involvement of the RM modules in pilin Av, we constructed two mutant strains. We 174 deleted the RM1 and RM2 operons that are predicted to recognize 5’-CCGG (strain SM500). We also 175 deleted the RM3 operon that is predicted to recognize 5’-GCCGGC (strain SM604). The SM604 176 mutant showed no differences in growth or pilin Av frequency and was not analyzed further (Fig. 177 3B). In contrast, the RM1RM2 double mutant SM500 displayed a significantly lower level of PDCMC 178 and increased growth (Fig. 3B, C). Moreover, deleting each restriction gene RM1R1 (SM323) and 179 RM2R (SM327) also reduced pilin variation levels (Fig. 4A, B). Pilin Av was restored upon 180 complementation of the RM1R1 or RM2R mutants with the mutated gene with a TetR-regulated 181 promoter gene at an ectopic site (Fig. S3). 182 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 8 183 Fig. 3. Mutation of RM genes alters pilin antigenic variation. A. Pilin Av levels of the parental 184 strain and the SM500 and SM604 mutants measured by the PDCMC assay. B. Growth of the 185 parental strain and the SM500 and SM604 mutants without IPTG on solid media. Fischer’s LSD test 186 was used for multiple comparisons; only the p-values <0.05 were plotted. C. Pictures of the 187 colonies grown with and without IPTG, showing the appearance of P- blebs (white arrows). 188 189 190 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 9 191 Fig. 4. Deletion of RM system operons reduces pilin Av. PDCMC assay results (graphs A,B &C) 192 and picture of the pilin Av defective colonies (D). A, B & C. Shown are the PDCMC scores for each 193 single deleted mutant compared to the parental strain grown on GCB + IPTG for 30h. Multiple 194 comparisons were performed using Fisher's LSD test. D. Pictures of the colonies of the parental and 195 the two mutants with a lower PDCMC than the parental, ΔR1.RM1 and ΔR.RM2 grown on GCB and 196 GCB IPTG for 30h. 197 198 199 The observed reduction in PDCMC in the ΔRM1R1 (SM323) and ΔRM2R (SM327) mutants was not 200 due to a growth deficiency (Fig. S3). We also observed that the deletion of the entire RM2 operon 201 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 10

in similar levels of pilin Av to those of the RM2R mutant, whereas this was not the case for 202 the RM1 operon (Fig. 3B, C). RNA sequencing data showed that the restriction gene RM2R has 203 significantly higher transcript levels than its methylase43. Differences in transcription and stability 204 between RM1 and RM2 effectors could explain why deleting the RM2 restriction enzyme has the 205 same effect as deleting the whole operon. These results indicate that the restriction enzymes 206 RM1R1 and RM2 are necessary for efficient pilin Av. 207 The pilE and pilS 5’-CCGG motifs are cut in genomic DNA 208 To test the hypothesis that the restriction endonucleases RM1R1 and RM2R2 digest 5’-CCGG 209 sequence, we designed an adapter-DNA with a GC overhang to ligate with the two nucleotide 5’-CG 210 overhangs resulting from a 5’-CCGG cut. We ligated the adapter directly to purified genomic DNA, 211 and PCR amplification was performed using a primer recognizing the adaptor paired with an 212 upstream or downstream pilE primer (pilRBS or pilEREV primers 44). We detected multiple amplified 213 products whose sizes corresponded to the location of pilE 5’-CCGG sites (Fig. S4). 214 We sequenced the adaptor-ligated DNA using ONT Nanopore long-read sequencing, which enabled 215 us to distinguish the pilE and pilS loci. Among the four 5’-CCGG sites found in pilE, the adapter 216 mapping revealed cleavage at the second and third 5’-CCGG sites. The first site, 5’-GCCGGC, and 217 the fourth 5’-CCGG site (located 9 bp after the third site) showed no cuts (Fig. S6). In the pilS 218 copies, cuts were detected at 44 of the 54 5’-CCGG sites across all 19 pilS copies, and no ligation 219 was found at the 32 5’-GCCGGC sites. These cleavage patterns matched those observed in the 220 positive control pretreated with HpaII, an endonuclease that also targets 5’-CCGG (Fig. 5, 221 Supplementary Table 1). No cuts were detected in the RM1RM2-deleted strain (SM500), but 222 cleavage was restored in the HpaII-treated control, confirming that deletion of these operons 223 inactivates their cognate restriction enzymes. Notably, in the RM1RM2 double mutant pretreated 224 with HpaII, we detected a low-level cut at the last 5’-CCGG site, suggesting that the residual loss of 225 methylation upon RM1M and RM2M deletions implicates a minor protective role for at least one of 226 the two 5’-CCGG-specific RM methylases (Fig. S5). As anticipated, the pilE CCGG-disrupted mutant 227 (SM598) showed no cuts in pilE. However, all pilS had cleaved 5’-CCGG sites in SM598, as in the 228 parental strain (Fig. 5, Supplementary Table 1). 229 Methylation profiles of the pilE and pilS 5’-CCGG sites 230 Using nanopore sequencing, we quantified methylated residues (5mC) at each site with Modkit 231 (Oxford Nanopore Technologies, 2023). Remarkably, across pilE and all pilS copies, most 5-232 GCCGGC were methylated (32/33 methylated sites), whereas most 5’-CCGG were not methylated 233 (4/58 methylated sites) (Fig. 5 A,B). A more in-depth analysis of the methylated 5’CCGG showed 234 these sites in particular overlap with other sites known to be a methylase target (5’-GGNNCC, 235 NgoAXV; 5’-RGCGCY; NgoAI; 5’-GGCC, NgoAII). We found no cuts at methylated 5’-GCCGGC sites, 236 confirming their protection. Similarly, methylated 5’-CCGG sites remained intact, and we didn’t 237 detect methylation at the 5’-CCGG sites that were cleaved. Only five pilE 5’-CCGG sites were 238 neither methylated nor cut. For this analysis, a cut was defined as detectable if it occurred in 239 ≥0.25% of the reads at a specific site. In general, the 5’-CCGG sites are found to be 240 undermethylated on the whole genome (Supplementary Table 4). This low level of genome-wide 241 methylation at 5’-CCGG sites is consistent with PacBio data for Gc FA1090 deposited to the Rebase 242 database 42. Additionally, the RM2 methylase (M.NgoAXIII) is predicted to target 5’-CCGG sites as 243 an isoschizomer of HpaIIM, previous studies were unable to identify its target site based on 244 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 11 methylation data and concluded that this methylase is inactive 27. Taken together, these results 245 show that there is an undermethylation of some but not all sites within the genome. 246 247 248 Fig. 5. Mapping of the 5’-CCGG and 5’-GCCGGC cuts and methylation status within pilE and 249 pilS copies A. Schematic of the pilE locus with 5’-CCGG sites in orange and the 5’-GCCGGC site in 250 pink. HVL corresponds to the hypervariable loop of pilE and HVT. B Schematic of the 19 aligned pilS 251 with 5’-CCGG sites in orange and the 5’-GCCGGC site in pink. In A & C, the sites that exhibited cuts 252 are hatched and the sites that were methylated are boxed. Sites are considered as methylated with 253 a fraction Nmodified/Nvalid > 3, and as cuts when the adapter was detected in more than 0.25% of the 254 total reads at that genomic position. 255 256 257 258 259 260 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 12 Restriction at 5’-CCGG sites impacts Gc growth 261 The RM1RM2 double mutant (SM500) displayed an enhanced growth phenotype after 30h of growth 262 (Fig. 3A). The pilE CCGG-disrupted mutant (SM598) also displayed bigger colonies than the parental 263 strain, although difference in CFUs was not significant (Fig. 2A). Since reduced growth can equalize 264 after extended growth, we measured the CFU/colony of these strains during a growth curve on solid 265 media. This solid media growth curve confirmed that SM598 showed enhanced growth compared to 266 the parental strain (Fig. 6 C, D). In addition, strains lacking the 5’-CCGG-targeting restriction 267 enzymes (RM1R1 and RM2R) also showed significantly higher CFU/colony counts, and growth was 268 reduced by expressing RM1R or RM2R from an ectopic locus (Fig. 6D, F). 269 We measured CFU/colony in hexadeleted strains carrying native pilin loci and the reintroduced pilS 270 gene, with and without 5’-CCGG-mutation (Fig. 2A, B). Interestingly, the two strains that showed the 271 highest CFU/colony were the pilE CCGG-disrupted mutant lacking pilS (SM582) or with a 5’-CCGG 272 mutant pilS locus (SM585). The strain with native pilE and pilS sequences (SM578) demonstrated 273 the lowest growth (Fig. 6A). Strains with either an intact pilS or pilE locus showed intermediate 274 growth, confirming that the elevated CFU/colony count in the pilE 5’-CCGG-disrupted mutant is 275 attributable to pilE and pilS disruption. 276 Collectively, these data indicate that RM1R or RM2R activity at 5’-CCGG sites in pilS or pilE results 277 in reduced growth. This observation explains why unpiliated colonies tend to have higher CFU 278 counts 46. To test this hypothesis, we compared the CFU per colony between two nonpiliated 279 mutants: a ΔpilE mutant, which does not undergo pilin Av, and a pilE promoter mutant, a mutation 280 that doesn’t affect pilin Av frequency 47. Interestingly, the pilE promoter mutation did not alter CFU 281 per colony compared to the parental strain, whereas the ΔpilE mutant exhibited a significantly 282 higher CFU per colony count (Fig. S7). We introduced the pilE promoter mutation in the CCGG-283 disrupted SM598 mutant, but it didn’t affect growth. This result shows that the increased growth we 284 observed is most likely due to the CCGG mutation rather than pilin or pilus expression. These 285

suggest that cleavage of the pilE and pilS copies, and, by extension, pilin Av, negatively 286 impacts fitness. 287 288 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 13 14 16 18 20 10 100 1000 10000 100000 Time (hours) CFU/colony pilS3C1pilEStrain -native SM576 native native SM578 CCGG-disruptednative SM580 -CCGG-disrupted SM582 native CCGG-disrupted SM584 CCGG-disrupted CCGG-disrupted SM585 12 14 16 18 20 22 24 1 10 100 1000 10000 Time (hours) CFU/colony parental strain SM323 ΔR1.RM1 SM527 ΔR1.RM1 R.RM1::NICS * * ns 12 14 16 18 20 22 24 1 10 100 1000 10000 Time (hours) CFU/colony parental strain SM327 ΔR.RM2 SM560 ΔR.RM2::NICSns * * 12 14 16 18 20 22 24 1 10 100 1000 10000 100000 Time (hours) CFU/colony Parental strain IPTG SM598 SM598 IPTG Parental strain * * * SM576 SM578 SM580 SM582 SM584 SM585 1 10 100 1000 10000 CFU/colony 0.0001 0.0006 <0.0001 0.0038 <0.0001 0.0042 <0.0001 16h timepoint A B C D E F Parental strain GCB Parental strain GCB IPTG CCGG-disrupted pilE GCB CCGG-disrupted pilE GCB IPTG 289 Fig. 6 – Growth on solid media. A. Strains from the pilS hexadeleted background with and without 290 disruption of the pilE or pilS1C3 sequences grown in the presence of 1mM IPTG. B. Statistical 291 analysis using Fisher’s LSD multiple comparison of the 16h timepoint from panel A, only 292 statistically significant comparisons are plotted. C. Growth of the parental strain and SM598 (pilE 293 CCGG disrupted) in the absence or presence of 1mM IPTG. D. Pictures of the parental strain and 294 the SM598 strain (pilE CCGG disrupted) grown for 18h on GCB in the absence or presence of 1mM 295 IPTG. E. Complementation of the SM323 growth in the presence or absence of 1mM IPTG. F. 296 Complementation of the SM327 growth in the presence or absence of 1mM IPTG. 297 298 299 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 14

300 Bacterial adaptation is shaped by a fundamental balance: traits such as virulence factors, 301 antibiotic tolerance, or antigenic variation that may reduce individual fitness but allow populations 302 to overcome selective pressures – whether from host immunity, microbial competition, or 303 environmental challenges. Antigenic variation (Av) illustrates this paradox: by generating diverse 304 surface antigen repertoires, some cells escape immune detection, while those with less variation 305 are more likely to be eliminated 48,49. 306 We demonstrate that the paralogous restriction enzymes RM1R1 (R1.NgoAXIV) and RM2R 307 (R.NgoAXIII), which each cleave the unmethylated 5’-CCGG sites across the genome to induce cuts 308 within the pilE and pilS loci. We propose that these cuts ligate to create one or two junctions 309 between the recombining genes, and resulting in regions of microhomology at one or both 310 recombination junctions. These results help explain why small regions of DNA can be transferred 311 during the pilin Av gene conversion reactions. However, these results do not explain the 312 requirement for homologous recombination (e.g., RecA, RecO, and RecR) for pilin AV. Previous data 313 suggested that there is a multi-step process in which a junction is formed between pilE and a pilS 314 copy 20. We postulate that the hybrid locus forms on a donor chromosome and recombines with the 315 pilE on the recipient chromosome 20,50. This idea is supported by the fact that Gc and the closely 316 related N. meningitidis, both of which undergo pilin Av 51,52 are polyploid, while N. lactamica that 317 does not undergo pilin Av is monoploid 50. This model is consistent with the observation that some 318 pilE recombinants have only one 5’-CCGG site at the recombination tract border. Ultimately, 319 invoking separate donor and recipient chromosomes explains why this represents a gene 320 conversion process since the silent copies on the recipient chromosome are not involved. Notably, 321 vlsE antigenic variation in Borrelia species, which are also polyploid 53 and possess several RM 322 systems on their plasmids 42, occurs via trans recombination, a process in which genetic exchange 323 takes place between physically distinct DNA molecules 54. It is interesting to speculate that Borrelia 324 may also use RM system-mediated recombination between donor and recipient sequences as a 325 step during antigenic variation. One unexpected observation from this study is the increased growth 326 in RM mutants or strains with mutated pilE or pilS 5’-CCGG sites. These results suggest that 327 restriction in the pilE and pilS 5’-CCGG sites interferes with Gc fitness. We knew that a ΔpilE mutant 328 exhibits increased growth, and we assumed this was a fitness cost of producing the pilin protein 329 and pilus fiber. However, since a pilE promoter -10 mutant (which abolishes pilE expression and 330 thus piliation) displays parental-growth, we conclude that the reduced growth when the 5’-CCGG 331 sites are cleaved is due to chromosomal loss during pilin AV. 332 There are many unknown parts of the pilin Av process. Our results do not explain how the digested 333 pilE and pilS copies associate, whether multiple digestions events occur within the pilin loci on a 334 single chromosome, and if other parts of the digested chromosomes would ligate together. Given 335 that most pilin loci are clustered within a 30 kb region of the 2.1 Mb chromosome, hybrid formation 336 may simply result from the proximity of these loci and their high density of 5’-CCGG sites. While we 337 postulate that recombination occurs between a hybrid formed between a donor chromosome and a 338 recipient, this would require protection of the second chromosome from digestion. Alternatively, 339 hybrid locus formation could occur in one half of a diplococcus, with the recipient chromosome 340 residing in the second coccal unit. 341 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 15 Taken together, these results imply that RM-mediated cleavage at 5’-CCGG sites in pilE is an early 342 and critical step in pilin Av. This improvement in fitness in laboratory conditions was inversely 343 correlated with the decrease in Pilin Av frequencies, which implies that pilin Av has a fitness cost 344 even without immune pressure. This aligns with observations in other pathogens, such as malaria 345 parasites and viruses, where antigenic diversity involves fitness costs in vitro 55,56. These findings 346 advocate that the pilin Av mechanism is adapted to limit the proliferation of non-varying cells, 347 thereby maintaining population-level diversity. These observations reinforce a key evolutionary 348 principle: population resilience often depends on limiting individual fitness. 349 350 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 16

351 Bacterial strains and growth 352 Bacterial strains used in this study (listed in Table S3) are derivatives of the N. gonorrhoeae FA1090 353 isolate. All strains were checked to have the pilE 1-81-S2 variant sequence developed during 354 human volunteer colonization 44. The strains were grown in 37°C at 5% CO2 on Gonococcal (Gc) 355 Medium Base (BD Difco) (36.25 g/L), agar (1.25g/L), Kellogg Supplement I (22.2 mM glucose, 0.68 356 mM glutamine, 0.45 mM cocarboxylase), and Kellogg Supplement II [1.23 µM Fe(NO3)3] 11. 357 Pilin antigenic variation measurements 358 PDCMC pilin Av assay 359 The PDCMC assay was performed as previously published 22. The Gc strains were revived on GCB 360 from a frozen stock overnight. The next day, a single colony was reisolated onto GCB and GCB 1 mM 361 IPTG plates and incubated overnight. After 18h of growth, 20 colonies were selected, and their 362 morphologies were monitored to score the appearance of nonpiliated blebs. After 30h of growth, 363 the number of blebs in the selected colonies was counted. The variation score was counted e.g., 1 364 for 1 bleb, 2 for 2 blebs, 3 for 3 blebs, and 4 for 4 or more blebs. The PDCMC score corresponds to 365 this variation score divided by the number of colonies (20). 366 Analysis of pilE variants 367 After 30h of growth on GCB 1mM IPTG, a bleb was reisolated from each colony on GCB. The next 368 day, a single colony was used to perform a PCR with the primers PilRBS and S3PA 44. The amplicon 369 was then sequenced using Genewiz Azenta Sanger services. The alignment of the sequences was 370 performed using Jalview 57. 371 Pilin Av DNA sequencing assay 372 To enable measuring of pilin Av frequencies by Illumina sequencing, recA6 strains were grown for 373 22 h on 1 mM IPTG GCB, which corresponds to about 19 or 20 generations 39. About 500 colonies 374 were collected, and genomic DNA was extracted. The KOD Hot Start (Novagen, Toyobo) PCR was 375 performed according to the supplied protocol, with 100 ng of template genomic DNA. False 376 “variants” can occur by in vitro recombination during the PCR amplification 39. To limit these PCR-377 generated hybrids, we limited the number of cycles to 30. The PCR product was then purified using 378 Qiagen PCR purification kit, and Illumina sequencing (SeqCenter). The sequencing data were 379 aligned with pilE 1-81-S2 sequence as a reference genome using Hisat2 58 and Samtools 59 . The 380 variant calling was then run using Pysam 60 in a custom script 61. 381 Detection of restricted 5’-CCGG sites 382 A synthetic adaptor was produced with a 5’-GC overhang and a single phosphorylated 5’-end. To 383 prepare the adapter-ligated DNA, we annealed the oligonucleotides SM243 TOP 384 (CGGGTCGGCAGTAACGTATTGATGCATACC) and SM244 BOT 385 (GGTCGGCAGTAACGTATTGATGCATACCCG) to create a CG overhang by mixing SM244 BOT and 386 SM243 TOP, heating the mixture to 98°C, and gradually decreasing the temperature by 5°C every 10 387 minutes until reaching 4°C to allow efficient annealing. We purified the product using the column 388 cleanup PCR purification kit (Qiagen). We ligated the purified adapters with purified SM18 recA6 389 genomic DNA using T4 DNA ligase at 16°C overnight (NEB). For the positive control of this 390 experiment, we used genomic DNA pre-treated with the restriction endonuclease Hpa II (NEB). 391 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 17 PCR was performed using forward and reversed primers corresponding to the adapter sequences 392 and the primers pilRBS 44, pilEREV 44, SM241 (GGCGTTACCGCGCCCGTC) and SM242 393 (ATCCATGAACCCGACCGCACAACG). The reaction was performed using GoTaq (Promega) with the 394 adapter-ligated gDNA as the template. The amplified products were analyzed on agarose gels. 395 We performed nanopore sequencing of the adapter-ligated gDNA (SeqCenter). We analyzed the 396 sequencing data using a custom script that mapped the adapters across the genome 61. The script 397 processes nanopore sequencing data to identify adapter positions by mapping the FASTQ reads to 398 a reference genome using minimap2, detecting adapters with edlib, and splitting reads at adapter 399 sites. Adapter positions were filtered and the location of adaptors was visualized using matplotlib. 400 The workflow utilizes pysam for BAM/FASTA file handling and Bio.Seq for sequence manipulation, 401 providing a comprehensive analysis of sequencing data quality and biological significance. 402 Methylation analysis 403 Methylated bases were identified using Modkit dist_modkit_v0.4.4_7cf558c 45, a tool designed for 404 analyzing nanopore sequencing data. Raw sequencing reads were aligned to the reference genome 405 using EPI2ME with default parameters. The resulting BAM files were processed with Modkit module 406 for detection of the 5mC methylations. Methylation frequencies were visualized, and regions of 407 interest were extracted for downstream analysis, comparing treated and control samples. 408 QUANTIFICATION AND STATISTICAL ANALYSIS 409 The statistical analysis was conducted using GraphPad Prism. The growth and PDCMC statistical 410 analysis were executed using an ANOVA test with the multiple comparison Fisher's LSD 411 parameters. Each bar graph represents the individual values that are represented by symbols. The 412 bar graphs correspond to the mean, with the SD indicated as an error bar. 413

414 We would like to thank Dr. Shaohui Yin for providing the hexadeleted mutant. We are also grateful to 415 Dr. Linda Hu for her insightful and constructive feedback during the revision of this manuscript, and 416 to all the members of the Seifert lab for their valuable discussions throughout the study. 417 Research reported in this publication was supported by the National Institute of Allergy and 418 Infectious Disease (NIAID) of the National Institutes of Health under grant numbers R37 AI033493, 419 R01AI146073, and R21AI148981. 420 DETAILS. 421 The content is solely the responsibility of the authors and does not necessarily represent the official 422 views of the National Institutes of Health. This manuscript is the result of funding in whole or in part 423 by the National Institutes of Health (NIH). It is subject to the NIH Public Access Policy. Through 424 acceptance of this federal funding, NIH has been given a right to make this manuscript publicly 425 available in PubMed Central upon the Official Date of Publication, as defined by NIH. 426 427 428 429 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 18

430 1. World Health Organization ed. (2024). Updated recommendations for the treatment of 431 Neisseria gonorrhoeae, Chlamydia trachomatis and Treponema pallidum (syphilis), and new 432 recommendations on syphilis testing and partner services (World Health Organization). 433 2. Unemo, M., Golparian, D., Sánchez-Busó, L., Grad, Y., Jacobsson, S., Ohnishi, M., Lahra, M.M., 434 Limnios, A., Sikora, A.E., Wi, T., et al. (2016). The novel 2016 WHO Neisseria gonorrhoeae 435

strains for global quality assurance of laboratory investigations: phenotypic, genetic 436 and reference genome characterization. J Antimicrob Chemother 71, 3096–3108. 437 https://doi.org/10.1093/jac/dkw288. 438 3. Quillin, S.J., and Seifert, H.S. (2018). Neisseria gonorrhoeae host adaptation and pathogenesis. 439 Nat Rev Microbiol 16, 226–240. https://doi.org/10.1038/nrmicro.2017.169. 440 4. Swanson, J. (1973). Studies on gonococcus infection. IV. Pili: their role in attachment of 441 gonococci to tissue culture cells. J Exp Med 137, 571–589. 442 https://doi.org/10.1084/jem.137.3.571. 443 5. Merz, A.J., So, M., and Sheetz, M.P. (2000). Pilus retraction powers bacterial twitching motility. 444 Nature 407, 98–102. https://doi.org/10.1038/35024105. 445 6. Craig, L., Forest, K.T., and Maier, B. (2019). Type IV pili: dynamics, biophysics and functional 446 consequences. Nat Rev Microbiol 17, 429–440. https://doi.org/10.1038/s41579-019-0195-4. 447 7. Hagblom, P., Segal, E., Billyard, E., and So, M. (1985). Intragenic recombination leads to pilus 448 antigenic variation in Neisseria gonorrhoeae. Nature 315, 156–158. 449 https://doi.org/10.1038/315156a0. 450 8. Meyer, T.F., Mlawer, N., and So, M. (1982). Pilus expression in Neisseria gonorrhoeae involves 451 chromosomal rearrangement. Cell 30, 45–52. https://doi.org/10.1016/0092-8674(82)90010-1. 452 9. Hamrick, T.S., Dempsey, J.A.F., Cohen, M.S., and Cannon, J.G. (2001). Antigenic variation of 453 gonococcal pilin expression in vivo: analysis of the strain FA1090 pilin repertoire and 454 identification of the pilS gene copies recombining with pilE during experimental human 455 infection. Microbiology (Reading) 147, 839–849. https://doi.org/10.1099/00221287-147-4-839. 456 10. Long, C.D., Madraswala, R.N., and Seifert, H.S. (1998). Comparisons between Colony Phase 457 Variation of Neisseria gonorrhoeae FA1090 and Pilus, Pilin, and S-Pilin Expression. Infect 458 Immun 66. https://doi.org/10.1128/IAI.66.5.1918-1927.1998. 459 11. Kellogg, D.S., Douglas S. Kellogg, Kellogg, D.S., Peacock, W.L., William L. Peacock, Deacon, 460 W.E., Deacon, W.E., Brown, L., and Pirkle, C.I. (1963). NEISSERIA GONORRHOEAE I. Virulence 461 Genetically Linked to Clonal Variation. Journal of Bacteriology 85, 1274–1279. 462 https://doi.org/10.1128/jb.85.6.1274-1279.1963. 463 12. Criss, A.K., Kline, K.A., and Seifert, H.S. (2005). The frequency and rate of pilin antigenic 464 variation in Neisseria gonorrhoeae: Neisseria gonorrhoeae pilin variation. Molecular 465 Microbiology 58, 510–519. https://doi.org/10.1111/j.1365-2958.2005.04838.x. 466 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 19 13. Criss, A.K., and Seifert, H.S. (2012). A bacterial siren song: intimate interactions between 467 Neisseria and neutrophils. Nat Rev Microbiol 10, 178–190. 468 https://doi.org/10.1038/nrmicro2713. 469 14. Shafer, W.M., and Rest, R.F. (1989). INTERACTIONS OF GONOCOCCI WITH PHAGOCYTIC 470 CELLS. Annual Review of Microbiology 43, 121–145. 471 https://doi.org/10.1146/annurev.mi.43.100189.001005. 472 15. Howell-Adams, B., Wainwright, L.A., and Seifert, H.S. (1996). The size and position of 473 heterologous insertions in a silent locus differentially affect pilin recombination in Neisseria 474 gonorrhoeae. Molecular Microbiology 22, 509–522. https://doi.org/10.1046/j.1365-475 2958.1996.00128.x. 476 16. Wainwright, L.A., Pritchard, K.H., and Selfert, H.S. (1994). A conserved DNA sequence is 477 required for efficient gonococcal pilin antigenic variation. Molecular Microbiology 13, 75–87. 478 https://doi.org/10.1111/j.1365-2958.1994.tb00403.x. 479 17. Cahoon, L.A., and Seifert, H.S. (2009). An alternative DNA structure is necessary for pilin 480 antigenic variation in Neisseria gonorrhoeae. Science 325, 764–767. 481 https://doi.org/10.1126/science.1175653. 482 18. Cahoon, L.A., and Seifert, H.S. (2013). Transcription of a cis-acting, Noncoding, Small RNA Is 483 Required for Pilin Antigenic Variation in Neisseria gonorrhoeae. PLoS Pathog 9, e1003074. 484 https://doi.org/10.1371/journal.ppat.1003074. 485 19. Prister, L.L., Yin, S., Cahoon, L.A., and Seifert, H.S. (2020). Altering the Neisseria gonorrhoeae 486 pilE Guanine Quadruplex Loop Bases Affects Pilin Antigenic Variation. Biochemistry 59, 1104–487 1112. https://doi.org/10.1021/acs.biochem.9b01038. 488 20. Howell-Adams, B., and Seifert, H.S. (2000). Molecular models accounting for the gene 489 conversion reactions mediating gonococcal pilin antigenic variation. Molecular Microbiology 490 37, 1146–1158. https://doi.org/10.1046/j.1365-2958.2000.02067.x. 491 21. Koomey, M., Gotschlich, E.C., Robbins, K., Bergström, S., and Swanson, J. (1987). Effects of 492 recA mutations on pilus antigenic variation and phase transitions in Neisseria gonorrhoeae. 493 Genetics 117, 391–398. https://doi.org/10.1093/genetics/117.3.391. 494 22. Sechman, E.V., Rohrer, M.S., and Seifert, H.S. (2005). A genetic screen identifies genes and 495 sites involved in pilin antigenic variation in Neisseria gonorrhoeae: Gonococcal mutants with 496 impaired pilin variation. Molecular Microbiology 57, 468–483. https://doi.org/10.1111/j.1365-497 2958.2005.04657.x. 498 23. Skaar, E.P., Lazio, M.P., and Seifert, H.S. (2002). Roles of the recJ and recN Genes in 499 Homologous Recombination and DNA Repair Pathways of Neisseria gonorrhoeae. J Bacteriol 500 184, 919–927. https://doi.org/10.1128/jb.184.4.919-927.2002. 501 24. Mehr, I.J., and Seifert, H.S. (1998). Differential roles of homologous recombination pathways in 502 Neisseria gonorrhoeae pilin antigenic variation, DNA transformation and DNA repair. Mol 503 Microbiol 30, 697–710. https://doi.org/10.1046/j.1365-2958.1998.01089.x. 504 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 20 25. Tesson, F., Hervé, A., Mordret, E., Touchon, M., d’Humières, C., Cury, J., and Bernheim, A. 505 (2022). Systematic and quantitative view of the antiviral arsenal of prokaryotes. Nat Commun 506 13, 2561. https://doi.org/10.1038/s41467-022-30269-9. 507 26. Loenen, W.A.M., Dryden, D.T.F., Raleigh, E.A., Wilson, G.G., and Murray, N.E. (2014). Highlights 508 of the DNA cutters: a short history of the restriction enzymes. Nucleic Acids Res 42, 3–19. 509 https://doi.org/10.1093/nar/gkt990. 510 27. Sánchez-Busó, L., Golparian, D., Parkhill, J., Unemo, M., and Harris, S.R. (2019). Genetic 511 variation regulates the activation and specificity of Restriction-Modification systems in 512 Neisseria gonorrhoeae. Sci Rep 9, 14685. https://doi.org/10.1038/s41598-019-51102-2. 513 28. Srikhanta, Y.N., Dowideit, S.J., Edwards, J.L., Falsetta, M.L., Wu, H.-J., Harrison, O.B., Fox, K.L., 514 Seib, K.L., Maguire, T.L., Wang, A.H.-J., et al. (2009). Phasevarions Mediate Random Switching 515 of Gene Expression in Pathogenic Neisseria. PLOS Pathogens 5, e1000400. 516 https://doi.org/10.1371/journal.ppat.1000400. 517 29. Adamczyk-Poplawska, M., Lower, M., and Piekarowicz, A. (2009). Characterization of the 518 NgoAXP: phase-variable type III restriction–modification system in Neisseria gonorrhoeae. 519 FEMS Microbiol Lett 300, 25–35. https://doi.org/10.1111/j.1574-6968.2009.01760.x. 520 30. Adamczyk-Poplawska, M., Lower, M., and Piekarowicz, A. (2011). Deletion of One Nucleotide 521 within the Homonucleotide Tract Present in the hsdS Gene Alters the DNA Sequence Specificity 522 of Type I Restriction-Modification System NgoAV. Journal of Bacteriology 193, 6750–6759. 523 https://doi.org/10.1128/jb.05672-11. 524 31. Kwiatek, A., Kobes, M., Olejnik, K., and Piekarowicz, A. (2004). DNA methyltransferases from 525 Neisseria meningitidis and Neisseria gonorrhoeae FA1090 associated with mismatch nicking 526 endonucleases. Microbiology 150, 1713–1722. https://doi.org/10.1099/mic.0.27011-0. 527 32. Pingoud, A., and Jeltsch, A. (2001). Structure and function of type II restriction endonucleases. 528 Nucleic Acids Res 29, 3705–3727. https://doi.org/10.1093/nar/29.18.3705. 529 33. Roberts, R.J., Belfort, M., Bestor, T., Bhagwat, A.S., Bickle, T.A., Bitinaite, J., Blumenthal, R.M., 530 Degtyarev, S.Kh., Dryden, D.T.F., Dybvig, K., et al. (2003). A nomenclature for restriction 531 enzymes, DNA methyltransferases, homing endonucleases and their genes. Nucleic Acids Res 532 31, 1805–1812. https://doi.org/10.1093/nar/gkg274. 533 34. Shaw, L.P., Rocha, E.P.C., and MacLean, R.C. (2023). Restriction-modification systems have 534 shaped the evolution and distribution of plasmids across bacteria. Nucleic Acids Res 51, 6806–535 6818. https://doi.org/10.1093/nar/gkad452. 536 35. Vasu, K., and Nagaraja, V. (2013). Diverse functions of restriction-modification systems in 537 addition to cellular defense. Microbiol Mol Biol Rev 77, 53–72. 538 https://doi.org/10.1128/MMBR.00044-12. 539 36. Swanson, J., Bergström, S., Barrera, O., Robbins, K., and Corwin, D. (1985). Pilus- gonococcal 540 variants. Evidence for multiple forms of piliation control. J Exp Med 162, 729–744. 541 https://doi.org/10.1084/jem.162.2.729. 542 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 21 37. Duckworth, M., Jackson, D., Zak, K., and Heckels, J.E. (1983). Structural variations in pili 543 expressed during gonococcal infection. J Gen Microbiol 129, 1593–1596. 544 https://doi.org/10.1099/00221287-129-5-1593. 545 38. Meyer, T.F., Billyard, E., Haas, R., Storzbach, S., and So, M. (1984). Pilus genes of Neisseria 546 gonorrheae: chromosomal organization and DNA sequence. Proc Natl Acad Sci U S A 81, 6110–547 6114. https://doi.org/10.1073/pnas.81.19.6110. 548 39. Rotman, E., Webber, D.M., and Seifert, H.S. (2016). Analyzing Neisseria gonorrhoeae Pilin 549 Antigenic Variation Using 454 Sequencing Technology. J Bacteriol 198, 2470–2482. 550 https://doi.org/10.1128/JB.00330-16. 551 40. Quillin, S.J., Luo, D., Gavagan, A., Prindle, A., and Seifert, H.S. (2025). A droplet digital PCR 552 assay to measure pilin antigenic variation frequency in Neisseria gonorrhoeae. mSphere 10, 553 e0009425. https://doi.org/10.1128/msphere.00094-25. 554 41. Ozer, E.A., Prister, L.L., Yin, S., Ward, B.H., Ivanov, S., and Seifert, H.S. (2019). PacBio Amplicon 555 Sequencing Method To Measure Pilin Antigenic Variation Frequencies of Neisseria 556 gonorrhoeae. mSphere 4, e00562-19. https://doi.org/10.1128/mSphere.00562-19. 557 42. NEB REBASE. https://rebase.neb.com/cgi-bin/onumget?64360. 558 43. Potter, A.D., Baiocco, C.M., Papin, J.A., and Criss, A.K. (2023). Transcriptome-guided metabolic 559 network analysis reveals rearrangements of carbon flux distribution in Neisseria gonorrhoeae 560 during neutrophil co-culture. mSystems 8, e01265-22. 561 https://doi.org/10.1128/msystems.01265-22. 562 44. Seifert, H.S., Wright, C.J., Jerse, A.E., Cohen, M.S., and Cannon, J.G. (1994). Multiple 563 gonococcal pilin antigenic variants are produced during experimental human infections. J. Clin. 564 Invest. 93, 2744–2749. https://doi.org/10.1172/JCI117290. 565 45. nanoporetech/modkit (Oxford Nanopore Technologies). 566 46. Boiko, I., Metaane, S., and Seifert, H.S. (2025). Publicly Available Neisseria Gonorrhoeae 567 Genomes Predominantly Represent In Vitro-Derived Nonpiliated Variants. J Infect Dis, jiaf557. 568 https://doi.org/10.1093/infdis/jiaf557. 569 47. Kline, K.A., Criss, A.K., Wallace, A., and Seifert, H.S. (2007). Transposon Mutagenesis Identifies 570 Sites Upstream of the Neisseria gonorrhoeae pilE Gene That Modulate Pilin Antigenic Variation. 571 J Bacteriol 189, 3462–3470. https://doi.org/10.1128/JB.01911-06. 572 48. Hill, D.J., Griffiths, N.J., Borodina, E., and Virji, M. (2010). Cellular and molecular biology of 573 Neisseria meningitidis colonization and invasive disease. Clin Sci (Lond) 118, 547–564. 574 https://doi.org/10.1042/CS20090513. 575 49. Zhang, J.-R., Hardham, J.M., Barbour, A.G., and Norris, S.J. (1997). Antigenic Variation in Lyme 576 Disease Borreliae by Promiscuous Recombination of VMP-like Sequence Cassettes. Cell 89, 577 275–285. https://doi.org/10.1016/S0092-8674(00)80206-8. 578 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 22 50. Tobiason, D.M., and Seifert, H.S. (2006). The Obligate Human Pathogen, Neisseria 579 gonorrhoeae, Is Polyploid. PLoS Biol 4, e185. https://doi.org/10.1371/journal.pbio.0040185. 580 51. Tinsley, C.R., and Heckels, J.E. (1986). Variation in the expression of pili and outer membrane 581 protein by Neisseria meningitidis during the course of meningococcal infection. J Gen Microbiol 582 132, 2483–2490. https://doi.org/10.1099/00221287-132-9-2483. 583 52. Helm, R.A., and Seifert, H.S. (2010). Frequency and Rate of Pilin Antigenic Variation of Neisseria 584 meningitidis. J Bacteriol 192, 3822–3823. https://doi.org/10.1128/JB.00280-10. 585 53. Takacs, C.N., Wachter, J., Xiang, Y., Ren, Z., Karaboja, X., Scott, M., Stoner, M.R., Irnov, I., 586 Jannetty, N., Rosa, P.A., et al. (2022). Polyploidy, regular patterning of genome copies, and 587 unusual control of DNA partitioning in the Lyme disease spirochete. Nat Commun 13, 7173. 588 https://doi.org/10.1038/s41467-022-34876-4. 589 54. Singh, P., and Bankhead, T. (2025). Breaking a barrier: In trans vlsE recombination and genetic 590 manipulation of the native vlsE gene of the Lyme disease pathogen. PLoS Pathog 21, e1012871. 591 https://doi.org/10.1371/journal.ppat.1012871. 592 55. Liu, Y., McNevin, J., Zhao, H., Tebit, D.M., Troyer, R.M., McSweyn, M., Ghosh, A.K., Shriner, D., 593 Arts, E.J., McElrath, M.J., et al. (2007). Evolution of Human Immunodeficiency Virus Type 1 594 Cytotoxic T-Lymphocyte Epitopes: Fitness-Balanced Escape. J Virol 81, 12179–12188. 595 https://doi.org/10.1128/JVI.01277-07. 596 56. Villa, M., and Lässig, M. (2017). Fitness cost of reassortment in human influenza. PLoS Pathog 597 13, e1006685. https://doi.org/10.1371/journal.ppat.1006685. 598 57. Waterhouse, A.M., Procter, J.B., Martin, D.M.A., Clamp, M., and Barton, G.J. (2009). Jalview 599 Version 2—a multiple sequence alignment editor and analysis workbench. Bioinformatics 25, 600 1189–1191. https://doi.org/10.1093/bioinformatics/btp033. 601 58. Kim, D., Paggi, J.M., Park, C., Bennett, C., and Salzberg, S.L. (2019). Graph-based genome 602 alignment and genotyping with HISAT2 and HISAT-genotype. Nat Biotechnol 37, 907–915. 603 https://doi.org/10.1038/s41587-019-0201-4. 604 59. Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., 605 Durbin, R., and 1000 Genome Project Data Processing Subgroup (2009). The Sequence 606 Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079. 607 https://doi.org/10.1093/bioinformatics/btp352. 608 60. Heger, A., Marshall, J., and Jacobs, K. (2009). Pysam: Htslib interface for python. (pysam-609 developers). 610 61. Interleukine/Two-paralogous-restriction-modification-systems-are-required-for-Neisseria-611 gonorrhoeae-pilin-antigen https://github.com/Interleukine/Two-paralogous-restriction-612 modification-systems-are-required-for-Neisseria-gonorrhoeae-pilin-antigen. 613 62. Seifert, H.S. (1997). Insertionally inactivated and inducible recA alleles for use in Neisseria. 614 Gene 188, 215–220. https://doi.org/10.1016/S0378-1119(96)00810-4. 615 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 23 63. Hu, L.I., Yin, S., Ozer, E.A., Sewell, L., Rehman, S., Garnett, J.A., and Seifert, H.S. (2020). 616 Discovery of a New Neisseria gonorrhoeae Type IV Pilus Assembly Factor, TfpC. mBio 11, 617 10.1128/mbio.02528-20. https://doi.org/10.1128/mbio.02528-20. 618 619 620 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 24 EXTENDED DATA 621 Table S1. List of the predicted Type II RM with their predicted recognition motif and the number of 622 occurrences in the Gc strain FA1090 N-1-60. 623 624 Predicted Type II RM system Motif (5’ to 3’) Number of occurrences in FA1090 N-1-60 Number of non-overlapping occurrences in FA1090 N-1-60 NgoFVII GCSGC 16173 13952 NgoAXIV and NgoAXIII CCGG 11103 7884 NgoAII GGCC 4590 2018 NgoAXIP GATC 2259 2112 NgoAXV GGNNCC 2033 818 NgoAX CCACC 1575 1380 NgoAIV GCCGGC 1572 0 NgoAI RGCGCY 622 224 NgoAVIII GACNNNNNTGA 407 338 NgoAIII CCGCGG 218 45 NgoAXVII GAGNNNNNTAC 118 89 625 Table S2 – Number of occurrences of each of the known motifs targeted by type II restriction-626 modification systems in Gc strain FA1090 627 628 Motif (5’ to 3’) # in pilE # in 19 pilS copies GCSGC 2 41 CCGG 4 87 GGCC 3 48 GATC 1 1 GGNNCC 0 14 CCACC 1 12 GCCGGC 1 32 RGCGCY 0 8 GACNNNNNTGA 0 1 CCGCGG 0 0 GAGNNNNNTAC 0 0 629 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 25 630 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 26 631 Fig. S1- Sequence alignment of pilE amplicon from the SM598 mutant variants. The variants were reisolated from blebs on GCB IPTG 632 after 30h of growth. The reference sequence is indicated in green, the pilS inserts are indicated in purple and the 5’-CCGG sites are 633 indicated in orange. 634 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 27 635 Fig. S2- Genomic organization of the RM1 RM2 and RM3 operons in Gc FA1090. Rectangles 636 follow Rebase nomenclature. ORFs are labeled with Ngo_ identifiers, while NEIS numbers 637 corresponds to PubMLST nomenclature. Red indicates restriction genes; blue, methylase genes. 638 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 28 0.0 0.2 0.4 0.6 0.8 1.0 PDCMC Score <0.0001 0.0007 parental strain ΔR1 ΔR1 iga-trpB::R1 RM1 0.0 0.2 0.4 0.6 0.8 1.0 PDCMC Score 0.0086 0.0499 0.0016 parental strain ΔR2 ΔR2 iga-trpB::R2 RM2 639 Fig. S3 – PDCMC score measured in the strains ΔR1.RM1 (SM323) and ΔR.RM2 (SM327) and 640 their complements. The gene of interest RMR1 on the left, RM2R on the right was reinserted in the 641 iga-trpB site using pMR69. Fischer’s LSD test was used, only the p-values <0.05 were plotted. 642 643 644 645 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 29 646 Fig. S4 – Measure of the growth of mutants carrying single gene deletions of the three operons. 647 CFU/colony of the strains of the individual mutants for the operons A. RM1, B. RM3, and C. RM2, 648 after 30h of growth on GCB with and without IPTG. Multiple comparisons were performed using 649 Fischer’s LSD test, only the p-values <0.05 were plotted. 650 651 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 30 652 653 Fig. S5 – Ligation of the adapter within the genomic DNA. Migration of the PCR-amplified 654 fragments using ligated genomic DNA as a template and several pairs of primers where one of them 655 is internal to the ligated adapter (246 or 245). The bands expected if the adapter has ligated into a 656 5’-CCGG cut are boxed in reindicated with red arrows. The ligA gene is used as a control fragment 657 that contains only one 5’-CCGG site. 658 659 660 Fig. S6. Detecting CCGG cuts within the pilE gene. Detection of the adapter within pilE. The 661 schematic representation of the pilE shows its different regions with the 5’-CCGG sites are 662 represented in orange. The panel below the pilE gene shows the locations of the detected CCGG 663 cuts in, from top to bottom: the parental strain recA6, it’s HpaII-pretreated control, followed with 664 the RM1RM2 double mutant (SM500), its HpaII control, the CCGG-disrupted strain (SM598) and its 665 HpaII control. 666 667 668 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 31 669 670 Fig. S7. Measure of the growth of mutants carrying pilE mutations. CFU/colony of the strains 671 ΔpilE, pilE promoter mutant, pilE CCGG-disrupted mutant and a mutant bearing both mutations at 672 16h of growth on GCB with IPTGMultiple comparisons were performed using Fischer’s LSD test. 673 674 Table S3. Strains used in this study 675 Bacterial strains FA1090 recA6 62 recA6 FA1090 recA6 CCGG disrupted pilE This study SM598 FA1090 recA6 ΔR1.RM1 This study SM323 FA1090 recA6 ΔR2.RM1 This study SM325 FA1090 recA6 ΔM.RM1 This study SM356 FA1090 recA6 ΔRM1 This study SM310 FA1090 recA6 ΔM1.RM2 This study SM362 FA1090 recA6 ΔM2.RM2 This study SM360 FA1090 recA6 ΔR.RM2 This study SM327 FA1090 recA6 ΔRM2 This study SM364 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint 32 FA1090 recA6 ΔRM1ΔRM2 This study SM500 FA1090 recA6 ΔR.RM3 This study SM600 N-1-60 CRISPRi ngo0873 (M.RM3) This study SM602 FA1090 recA6 ΔRM3 This study SM604 FA1090 recA6 ΔpilS123678 hexadeleted mutant Shaohui Yin Q409 FA1090 recA6 ΔpilS123678 hexadeleted mutant KanR This study SM564 FA1090 recA6 hexadeleted mutant::pilS3C1 KanR This study SM568 FA1090 recA6 hexadeleted mutant::CCGG-disrupted pilS3C1 KanR This study SM570 FA1090 recA6 hexadeleted mutant pilE CCGG-disrupted KanR This study SM582 FA1090 recA6 hexadeleted mutant::pilS3C1 pilE CCGG-disrupted KanR This study SM584 FA1090 recA6 hexadeleted mutant::CCGG-disrupted pilS3C1 CCGG-disrupted pilE KanR This study SM585 FA1090 recA6 ΔR1.RM1 R1.RM1 at iga-trpB This study SM527 FA1090 recA6 ΔR.RM2 R.RM2 at iga-trpB This study SM560 FA1090 recA6 ΔpilE 63 SM648 FA1090 recA6 pilE-10::NheI 47 SM650 FA1090 recA6 CCGG-disrupted pilE-10::NheI This study SM655 676 677 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 14, 2026. ; https://doi.org/10.64898/2026.01.10.698803doi: bioRxiv preprint