Quantitative proteomics of the 2016 WHO Neisseria gonorrhoeae reference strains surveys vaccine candidates and antimicrobial resistance determinants

The sexually transmitted disease gonorrhea (causative agent: Neisseria gonorrhoeae) remains an urgent public health threat globally due to the repercussions on reproductive health, high incidence, widespread antimicrobial resistance (AMR), and absence of a vaccine. To mine gonorrhea antigens and enhance our understanding of gonococcal AMR at the proteome level, we performed the first large-scale proteomic profiling of a diverse panel (n=15) of gonococcal strains, including the 2016 World Health Organization (WHO) reference strains. These strains show all existing AMR profiles, previously described in regard to phenotypic and reference genome characteristics, and are intended for quality assurance in laboratory investigations. Herein, these isolates were subjected to subcellular fractionation and labeling with tandem mass tags coupled to mass spectrometry and multi-combinatorial bioinformatics. Our analyses detected 901 and 723 common proteins in cell envelope and cytoplasmic subproteomes, respectively. We identified nine novel gonorrhea vaccine candidates. Expression and conservation of new and previously selected antigens were investigated. In addition, established gonococcal AMR determinants were evaluated for the first time using quantitative proteomics. Six new proteins, WHO_F_00238, WHO_F_00635, WHO_F_00745, WHO_F_01139, WHO_F_01144, and WHO_F_01226, were differentially expressed in all strains, suggesting that they represent global proteomic AMR markers, indicate a predisposition toward developing or compensating gonococcal AMR, and/or act as new antimicrobial targets. Finally, phenotypic clustering based on the isolates’ defined antibiograms and common differentially expressed proteins yielded seven matching clusters between established and proteome-derived AMR signatures. Together, our investigations provide a reference proteomics databank for gonococcal vaccine and AMR research endeavors, which enables microbiological, clinical, or epidemiological projects and enhances the utility of the WHO reference strains.

AMR markers, indicate a predisposition toward developing or compensating gonococcal 48 AMR, and/or act as new antimicrobial targets. Finally, phenotypic clustering based on the 49 isolates' defined antibiograms and common differentially expressed proteins yielded seven 50 matching clusters between established and proteome-derived AMR signatures. Together, 51 our investigations provide a reference proteomics databank for gonococcal vaccine and 52 AMR research endeavors, which enables microbiological, clinical, or epidemiological 53 projects and enhances the utility of the WHO reference strains. 54

INTRODUCTION 74
Neisseria gonorrhoeae is an obligate human pathogen and the causative agent of 75 the sexually transmitted disease gonorrhea. Gonorrhea is a global public health concern. In 76 2012 the World Health Organization (WHO) estimated over 78 million new urogenital cases 77 per year in adults (15-49 years of age) worldwide (1, 2). The spread of gonorrhea is 78 facilitated by the high prevalence of asymptomatic infections. Urogenital gonorrhea is 79 asymptomatic in up to 10-15% of infected men and up to 50% of infected women. 80 Pharyngeal and rectal infections, which have increased in prevalence in both sexes and are 81 predominant among men who have sex with men, are primarily asymptomatic (3, 4). 82 Untreated or inappropriately treated gonorrhea can result in serious consequences on 83 reproductive and neonatal health. Women, in particular, are disproportionately affected, as 84 gonococcal infection can ascend from the cervix to the uterus, Fallopian tubes, ovaries, and 85 surrounding tissue, causing pelvic inflammatory disease. Long-term sequelae include 86 ectopic pregnancy, chronic pelvic pain, and infertility. Furthermore, gonorrhea is strongly 87 associated with an increased risk of both the acquisition and transmission of HIV (5). 88 Antimicrobial therapy is the only mainstay in the effective management and control 89 of gonorrhea. However, N. gonorrhoeae exhibits an extraordinary capacity to develop 90 antimicrobial resistance (AMR) through mutations and acquisition of AMR genes. The 91 evolution of AMR in N. gonorrhoeae has overcome every therapeutic option since the 92 "miracle drug" penicillin was introduced for gonorrhea treatment. Currently, a dual 93 antimicrobial therapy (mainly ceftriaxone and azithromycin) is recommended for treatment 94 of uncomplicated infections (6). Of grave concern over the past decade is the proliferation 95 of resistance or decreased susceptibility to ceftriaxone worldwide. Azithromycin resistance 96 has also emerged in most settings (7). The first failure of one of the recommended dual 97 (phenotypic and genetic) for all antimicrobials currently and previously used for gonorrhea 146 treatment, in addition to novel antimicrobials considered for future interventions. This panel 147 includes WHO X, the first extensively drug-resistant gonococcal strain identified with high-148 level resistance to ceftriaxone, as well as additional strains with different levels of resistance 149 to ceftriaxone, azithromycin and any additional therapeutic antimicrobials. Complete The AMR profiles of all isolates were described previously (50). Gonococcal strains were 180 cultured from frozen stocks (−80°C) onto gonococcal base agar (GCB) medium (Difco) with 181 Kellogg's supplements I and II, diluted 1:100 and 1:1,000, respectively (57). After incubation 182 at 37°C in a 5% CO2-enriched atmosphere for 18-20 h, nonpiliated and transparent colonies 183 were subcultured onto GCB and incubated as described above. To initiate growth in liquid 184 medium, nonpiliated colonies were collected from GCB and suspended to an OD600 of 0.1 185 in pre-warmed GCB liquid (GCBL) medium supplemented as described above with the 186 addition of 0.042% sodium bicarbonate. Suspensions were incubated at 37°C with shaking 187 at 220 rpm. were separated using a 2.1 mm x 100 mm Polysulfoethyl A column (PolyLC) over 60 min at 220 a flow rate of 200 µL/min. The separation profile was as follows: hold 2% B for 5 min, 2% to 221 8% B in 0.1 min, 8% to 18% B in 14.9 min, 18% to 34% B in 12 min, 34% to 60% B in 18 222 min, 60% to 98% B in 0.1 min and hold for 10 min. Fractions were collected in 96-well 223 microtiter plates at 1 min/fraction. Sixty fractions were pooled into 12 and dried using a 224 Proteomic data analysis. Data  Differential protein expression in four proteomics data sets (CE and C fractions in two 269 biological replicates) was designated by fold changes ≥1.5 or ≤0.667 in reference to strain 270 WHO F. Due to the variable nature of protein expression in N. gonorrhoeae, we took a 271 conservative approach to designate protein expression and a protein was categorized as 272 "up-regulated" or "down-regulated" solely when the fold change abundance was higher than 273 1.5 or lower than 0.667, respectively, to that of WHO F consistently in two biological 274 experiments. A protein was designated as "ubiquitous" when its abundance was between 275 0.667-1.5-fold compared to WHO F in both experiments, or "variable" when its protein levels 276 were not consistent between experiments. algorithms and a majority voting strategy. Furthermore, for proteins whose subcellular 281 localization was not predicted using the aforementioned algorithms, we relied on the 282 difference between their unique peptide counts in the CE and C fractions as follows: 283 where "i " is the sequential number assigned for samples and " " is the total number of 285 peptides detected in each fraction. Cytoplasmic proteins had more of their unique peptides 286 detected in the C fraction (UPCD < 0), while membrane proteins had unique peptides 287 enriched in the CE fraction (UPCD > 0). Proteins with UPCD=0 were excluded from analysis 288 using this UPCD formula. Proteins were categorized as follows: outer membrane, 289 periplasmic, inner membrane, C proteins, and proteins with unknown localization. 290 The phenotypic and proteotypic clusters of all strains were constructed using as 291 variables both their AMR (50) and proteomic profiles obtained in this study. These clusters 292 were designed based on the Hamming distance between tested strains, which counts how 293 many elements differ between two vectors, and is equivalent to Manhattan distance on 294 binary data. Average linkage was used to determine distances between clusters. 295 Graphs were generated with GraphPad Prism version 7 for Mac (GraphPad 296 Software). The proteotypes of strains that belong to the same phenotypic cluster were 297 compared, highlighting proteins that are significantly up-or down-regulated with respect to 298 those proteins of WHO F. 299 Data Availability. The raw mass spectrometry data have been deposited to the 300 ProteomeXchange Consortium via the PRIDE (66) partner repository with the data set 301 identifier PXD008412. 302

RESULTS and DISCUSSION 304
Study rationale. In our study design (Fig. 1), all 15 strains were cultured concurrently to 305 mid-logarithmic growth, harvested, and subjected to subcellular fractionation to separate CE 306 (outer membrane, periplasmic, inner membrane) and C proteins. We utilized TMT reagent 307 technology for protein identification and quantitation as it provides a highly sensitive method 308 for peptide labeling (56) and allows up to 10 biological samples to be analyzed in a single 309 experiment (67). TMT-labeling, two-dimensional liquid chromatography fractionation, and 310 subsequent MS/MS analyses were conducted on every 6-plex and 10-plex experiment 311 pertaining to the CE and C fractions derived from each strain (Fig. 1). We selected WHO F 312 as the reference strain for protein identification and quantitation because it has the largest 313 genome (2,292,467 bp) and proteome (2,450 ORFs) among the 2016 WHO reference 314 strains (50) and FA6140 (68), and it is susceptible to most antimicrobials currently or 315 historically used for gonorrhea treatment. 316 Sub-cellular fractionation experiments coupled with proteomics repeatedly show 317 cytoplasmic proteins associated with the membranes, which are commonly regarded as 318 "contaminating" or "moonlighting" proteins (29, 30, 69, 70). Therefore, to focus solely on the 319 enriched proteins in individual subproteomes, we first eliminated C and CE proteins that 320 were detected in the CE and C protein fractions, respectively, from further analyses. 321 Complete lists of all identified proteins are in Supplemental Tables S1-S2. Subsequently, 322 we performed two-armed proteomic data analyses: 1) for vaccine antigen mining, we 323 focused on common proteins identified in all strains in the CE fraction with the overarching 324 goal to discover omnipresent N. gonorrhoeae proteins; 2) to profile AMR signatures, we 325 performed a pairwise comparison of individual strains to WHO F to enhance the discovery 326 of strain-specific feature(s). 327 Overview of cell envelope and cytoplasmic proteomes. The 10-plex biological replicate 328 experiments identified a total of 1150 proteins in the CE fraction of all ten strains, of which 329 1010 were common in both sets (Fig. 2 A). In the two 6-plex experiments, 1194 proteins 330 were identified; 975 were shared in all six isolates (Fig. 2 A). Taken together, the 10-plex 331 and the 6-plex experiments resulted in identification of 1084 proteins in the CE fractions, of 332 which 901 were common among all examined N. gonorrhoeae strains (Fig. 2 A). The 333 proteome coverage per strain ranged from 41.22% (981 proteins) for WHO Y to 45.32% 334 (1042 proteins) for WHO G (Supplemental Table S3). 335 Proteomics of the C fraction in the 10-plex set conducted in biological replicates 336 yielded 904 proteins that were shared among all 10 strains, of which 747 were common in 337 both experiments (Fig. 2 B). The two 6-plex experiments identified 1023 shared proteins, 338 with 852 common among the two replicates (Fig. 2 B). Cumulatively, C fraction profiling 339 resulted in identification of 876 proteins with 723 common in all 15 N. gonorrhoeae strains 340 ( Fig. 2 B). Proteome coverage ranged from 31.37% (746 proteins) in WHO U to 38.43% 341 (852 proteins) in FA6140 (Supplemental Table S3). 342 Subsequently, we allocated common proteins that were identified in all 15 N. 343 gonorrhoeae strains to outer membrane, inner membrane, periplasm, cytoplasm, or 344 unknown localization categories based on PSORTb 3.0.2 (62), SOSUIGramN (63), and 345 CELLO (61) predictions and the majority-voting strategy. We used these software packages 346 to take advantage of their different algorithms and statistical approaches for the prediction 347 of protein subcellular localization. As expected from our subcellular fractionation approach 348 (49, 69, 71), the CE fraction was enriched in membrane proteins in comparison to the C 349 sample, with outer membrane (26 vs. 8), periplasmic (51 vs. 38), and inner membrane 350 proteins (145 vs. 6) that were also identified with considerably higher peptide counts (Fig. 3  351  A-C, and Supplemental Tables S1-S2,  were found in the majority of examined strains (Supplemental Table S6). In addition,  Table S4), whereas down-regulated outer 377 membrane (3.85 -26.9%), periplasmic (0 -23.53%) and inner membrane (1.38 -24.8%) 378 proteins were moderately prevalent (Supplemental Figure S4). Further analysis of the CE 379 fraction detected 121 common proteins with unknown localization (Fig. 3 E, Supplemental 380 Table S4). Ubiquitous expression was the dominant pattern for these proteins in WHO G,  Table S5). The percentage of variable 388 proteins ranged from 32.9% to 82.6% for WHO G and WHO Y, respectively (Supplemental 389   Table S5). Ubiquitous proteins were the next most common category and oscillated from 390 15.5% in WHO Y to 56.4% in WHO G. The third group contained up-regulated proteins (0 -391 21.28%), and down-regulated proteins ranged from 0.5 -3.88% (Supplemental Table S5). 392 For proteins with no assigned localization, variable expression was the most prevalent 393 pattern in WHO K, L, O, P, U, W, X, Y, Z, and FA6140 ( Fig. 5 B), ranging from 79.75% (n= 394 63, WHO Y) to 46.83% (n= 37, WHO L). Ubiquitous proteins were the dominating group in 395 WHO G, M, N, and V (Fig. 5 B). Finally, up-and down-regulated proteins constituted up to 396 11.39% and 7.59%, respectively, of the total proteins with unknown localization in the C 397

fraction. 398
Together, the first quantitative proteomic profiling of the 15 N. gonorrhoeae strains 399 demonstrated distinct differences in their proteomes and showed that a pattern of ubiquitous 400 protein expression was prevalent in the CE fraction, whereas variably expressed proteins 401 were the dominant group in the C subproteome.  Table S1). The 407 latter group of proteins was subjected to signal sequence and transmembrane motif 408 analyses to increase the coverage of potential vaccine candidates. Together, these 409 investigations yielded nine novel antigens including NGO0282, NGO0425, NGO0439, 410 NGO0778, NGO1251, NGO1688, NGO1889, NGO1911a, and NGO2105 in addition to 411 previously discovered proteomics-derived antigens [(29, 30); outer membrane protein that is positively regulated by the oxygen-sensing transcription 435 factor, FNR (88). NGO1911a is a predicted pilus assembly protein that is associated with 436 the adhesin PilY (89). Finally, NGO0778, NGO1251 (a putative lipoprotein), and NGO1889 437 are hypothetical proteins. NGO1889 belongs to the LprI family (PFO7007) that comprises 438 bacterial proteins of ~120 amino acids in length that contain four conserved cysteine 439 residues. LprI from Mycobacterium tuberculosis acts as a lysozyme inhibitor (90), providing 440 the exciting possibility that N. gonorrhoeae LprI contributes to residual lysozyme resistance 441 observed in gonococci deprived of surface-exposed lysozyme inhibitors SliC and ACP (73, 442 91). Lastly, NGO2105 contains peptidase S6 (residues 43-310) and autotransporter 443 (residues 1215-1468) domains potentially involved in proteolytic activity and auto-444 translocation, respectively, suggesting that this is a newly identified autotransporter protein 445 in N. gonorrhoeae. In support of this notion, the NGO2105 locus, also known as adhesion 446 and penetration protein or "NEIS1959 (iga2)" in the PubMLST database, encodes IgA2 447 protease (AidA) and has homologs in other Neisseria (Table 1)  and PorB were up-and down-regulated, respectively, in 12 isolates (Table 2). Compared to 457 WHO F, PorB was present at similar levels only in WHO G and N. PilQ (98) was ubiquitously 458 expressed in 10 strains, whereas expression of Opa proteins was widely variable, as 459 expected (106, 107). The TbpA level was similar in 8 strains; however, we did not detect 460 TbpB (108). Nor did we detect ACP (109, 110) or OpcA (111, 112) under the standard 461 growth conditions used in our studies, which suggested that they might be specifically 462 regulated. AniA was present at different levels in 7 strains, ubiquitous in five, and up-463 regulated in two isolates. Immunoblotting experiments with anti-AniA antisera corroborated 464 these findings (105). The cellular pool of NspA (113) varied in ten isolates, while lactoferrin 465 binding protein LbpA (114) was variable in five strains and was ubiquitous in WHO L and G 466 (Table 2). 467 Strikingly, most of the proteome-derived vaccine candidates showed ubiquitous 468 expression among numerous strains (Table 1). In particular, SliC, PldA, BamE, BamA, and 469 BamG were ubiquitous in all 15 isolates. Similar results for these proteins were obtained by 470 iTRAQ-MS/MS applied to the proteomic profiling of cell envelopes and outer membrane 471 vesicles (OMVs) isolated from four different strains of N. gonorrhoeae (29). Further, LolB, 472 Ng-MIP, NGO1559, and NGO2054 were unvaryingly expressed in at least 12 isolates. 473 Among the novel vaccine candidates identified in our study, LptE, LolB, IgA2, and NGO1251 474 were found ubiquitous in at least 13 strains. In addition, LprI and NGO0778 were similarly 475 expressed in 12 and 9 isolates, respectively. In support of our proteomics data,  (Table 1). In addition, Ng-MIP-like proteins exist in C. trachomatis, 492 G. vaginalis, and P. ruminicola; BamA and NGO1559 homologs were found in C. 493 trachomatis and P. ruminicola. MetQ, a methionine transporter (74), was the only 494 proteomics-derived vaccine candidate with homologs across all examined bacteria with the 495 exception of C. trachomatis and P. ruminicola. Further, we detected protein homologs of 496 PilQ in two C. trachomatis strains and MtrE and ZnuD in P. ruminicola; these three proteins 497 were absent in commensal species. 498 Together, our investigations provide pioneering information into newly identified and 499 existing gonorrhea vaccine candidates. We have established each candidate's expression 500 pattern in diverse N. gonorrhoeae isolates and identified homologs among other pathogenic 501 and/or commensal bacteria that share the same ecological niche. Stable expression in the 502 WHO gonococcal panel coupled to presence in N. meningitidis and co-infecting agents -503 but rarely in urogenital commensalsfurther highlights the importance of including these 504 antigens in gonorrhea vaccine(s). 505

Proteomics profiling of N. gonorrhoeae antimicrobial resistance. Various genome-506
based AMR determinants have been deciphered in the gonococcus over the past decades 507 (51, 122-127). However, many AMR determinants remain to be identified and characterized, 508 e.g. the chromosomally-encoded penicillin and cephalosporin resistance determinant "factor 509 X" (128-130) and the AMR mechanisms that contribute to a large proportion of azithromycin 510 resistance (131). The uncertainty behind these AMR determinants illustrates the need for 511 alternative approaches to enhance our understanding of gonococcal AMR complexity. At 512 the proteomic level, only two studies have attempted to address this challenge, both of which 513 used 2D-SDS PAGE exclusively (47, 132). Therefore, we focused on identifying proteomic 514 AMR signatures that exist in the absence of antimicrobial pressure during standard in vitro 515 growth conditions by performing a pairwise comparison of all identified proteins in each 516 individual strain to the WHO F reference strain (Supplemental Tables S1-S2). As expected, 517 we identified different numbers of proteins in the CE and C fractions in each comparison set 518 due to differences in the number of open reading frames (ORFs) between the gonococcal 519 strains (Supplemental Table S3). Similarly to our previous analysis, we excluded typical 520 cytoplasmic proteins from the CE subproteome and cell envelope proteins from the C 521 fraction. We solely focused on proteins with significantly different expression in the 522 examined strains compared to the fully antimicrobial-susceptible strain WHO F with the 523 rationale that these proteins may provide clues about the proteomic basis of AMR. For 524 instance, we identified MtrE as up-regulated in many strains with increased resistance to 525 numerous antimicrobials even in the absence of antimicrobial exposure, which represents 526 an up-regulation of the multidrug MtrCDE efflux pump and possibly additional efflux pumps 527 for which MtrE acts as the outer membrane channel (29, 94, 133-135). Overall, we identified 528 162 (including 21 known AMR determinants) and 95 proteins with known and unpredicted 529 subcellular locations, respectively ( Figure 6). Peptide counting performed for the latter group 530 of proteins yielded 55 and 36 proteins that are likely localized to the CE and C, respectively, 531 and four proteins with ambiguous localization. Next, we separated proteins that have been 532 previously verified as N. gonorrhoeae AMR determinants (Table 3) (Table 3). WHO L is also the only 552 examined strain that contains an mtr120 mutation, which generates a novel promoter for 553 mtrCDE transcription and further enhances the expression of the MtrCDE efflux pump (50, 554 138). The second efflux system that showed differential expression was the MacAB efflux 555 pump, which can decrease macrolide susceptibility (139). MacA expression varied across 556 the isolates. Expression of the inner membrane component, MacB, was enhanced in the 557 azithromycin resistant strains WHO P and V, but also in the azithromycin susceptible strains 558 WHO N and K, as well as WHO L, which is intermediately susceptible to azithromycin. Among other established AMR determinants that were differentially expressed was 570 the major porin of N. gonorrhoeae, PorB (143, 144), which was down-regulated in all strains 571 with the exception of WHO G and N (Table 4). This down-regulation suggests reduced 572 import of antimicrobials such as penicillins, cephalosporins and tetracyclines, which can 573 contribute to a decreased antimicrobial susceptibility. Furthermore, the WHO F, G, and N 574 express PorB1a, which is associated with a lack of the AMR determinant penB and 575 consequently high-level chromosomally-mediated resistance to penicillins and 576 cephalosporins (127), while all other strains express PorB1b. All WHO strains with PorB1b 577 (n=11), except WHO U, contained the AMR determinant penB. Consequently, our proteomic 578 data suggest that penB may be associated with also a decreased expression of PorB1b in 579 addition to the previously documented decreased penetration through PorB1b, resulting in 580 a decreased susceptibility to several antimicrobials. The expression of penicillin-binding 581 protein 1 (PBP1) was significantly down-regulated in nine out of the twelve WHO strains that 582 possess the ponA1 resistance determinant, which encodes a L421P amino acid substitution 583 in PBP1 that contributes to high-level chromosomally-mediated penicillin resistance (50). 584 Accordingly, our proteomic data indicate that the PBP1 L421P amino acid alteration, in 585 addition to decreased expression of PBP1, might contribute to high-level chromosomally 586 mediated penicillin resistance. In contrast, PBP2 (the main lethal target for penicillins and 587  (Table 3). 594 New potential proteomic-derived antimicrobial resistance signatures. In the CE 595 fraction, two hypothetical proteins predicted to localize to the inner membrane, 596 WHO_F_00238c and WHO_F_01226, were down-regulated in all examined strains 597 compared to the antimicrobial-susceptible WHO F strain (Table 4). WHO_F_00238c, which 598 corresponds to NGO0222 in the FA1090 genome, is a small protein with a predicted 599 molecular weight of 8.32 kDa that contains two predicted transmembrane domains but no 600 signal peptide. WHO_F_01226 lacks a homologous protein in FA1090. This is also a small 601 protein (5.39 kDa) with no peptides predicted to be recognized by signal peptidase I or II. In anti-virulence or antimicrobial target. Finally, we detected two proteins with undefined 612 subcellular localization displaying global differential expression. WHO_F_01139 and 613 WHO_F_01144, which have no homologs in the FA1090 genome, were down-regulated in 614 all strains. Our use of UPCD predicted WHO_F_01139 to localize to the cell envelope. 615 WHO_F_01139 is a putative lipoprotein (16.9 KDa) with a predicted signal peptide II domain. 616 Based on UPCD, in addition to the lack of a predicted signal peptide and the absence of 617 transmembrane domains, we predict the hypothetical protein WHO_F_01144 (7.4 kDa) is 618 cytoplasmic. The impact of these six proteins on AMR is yet to be elucidated; however, our 619 data suggest that they may represent general proteomic markers for gonococcal AMR, a 620 predisposition toward developing or compensating for gonococcal AMR, and/or new 621 antimicrobial targets. Tables 4-6). 630 Cluster I strains, WHO P and U, exhibit resistance to azithromycin and the majority 631 of up-regulated proteins identified were involved in ribosomal biogenesis: 30S ribosomal 632 proteins S15 and S19; 50S ribosomal proteins L1, L2, and L22; the small GTPase EngA; 633 pseudouridine synthase; RNA helicase; and ribonuclease E. In contrast, proteins involved 634 in cell envelope biogenesis -PilE, LolA, and PglBwere down-regulated in both strains, 635 which may be associated with the strains' decreased susceptibility to penicillin G.  (Table 6). 662 Finally, the cluster VII strains (WHO O and FA6140), displaying resistance to penicillin 663 G and tetracycline, had nine common differentially expressed proteins (Table 6). Among 664 these proteins, two metabolic coenzymes (NGO0360 and NGO2056) and a putative 665 cytochrome b561 involved in energy production were down-regulated. This cluster 666 possessed a similar expression profile to strains in Cluster I that are intermediately 667 susceptibility to penicillin G and tetracycline. Finally NGO2017, a putative integral inner 668 membrane protein; NGO0452, a potassium proton/antiporter; PilW; and PilE were also 669 down-regulated in the cluster VII strains (Table 6).      1217  1218  1219  1220  1221  1222  1223  1224  1225  1226  1227  1228  1229  1230  1231  1232  1233  1234  1235  1236  1237  1238  1239  1240  1241  1242  1243  1244  1245  1246  1247 1248 1249   I  II  III  IV  V  VI  VII  I  II  III  IV  V  VI