Neisseria gonorrhoeae Type IV Pili Undergo Multisite, Hierarchical Modifications with Phosphoethanolamine and Phosphocholine Requiring an Enzyme Structurally Related to Lipopolysaccharide Phosphoethanolamine Transferases*

The zwitterionic phospho-forms phosphoethanolamine and phosphocholine are recognized as influential and important substituents of pathogen cell surfaces. PilE, the major pilin subunit protein of the type IV pilus (Tfp) colonization factor of Neisseria gonorrhoeae undergoes unique, post-translational modifications with these moieties. These phospho-form modifications have been shown to be O-linked alternately to a specific, conserved serine residue of PilE. However, the enzymes and precursors involved in their addition are unknown, and the full spectrum of PilE post-translational modifications has yet to be defined. Here, an intact protein-based mass spectrometric approach was integrated with bioinformatics and reverse genetics to address these matters. Specifically we show that a protein limited in its distribution to pathogenic Neisseria species and structurally related to enzymes implicated in phosphoethanolamine modification of lipopolysaccharide is necessary for PilE covalent modification with phosphoethanolamine and phosphocholine. These findings strongly suggest that protein phospho-form modification is mechanistically similar to processes underlying analogous modifications of prokaryotic saccharolipid glycans. We also show that PilE undergoes multisite and hierarchical phospho-form modifications and that the stoichiometries of site occupancy can be influenced by PilE primary structure and the abundance of the pilin-like protein PilV. Together, these findings have important implications for the structure and antigenicity of PilE.

Covalent post-translational modifications of surface proteins in microbial pathogens are now a well established phenomenon. By modifying structure and potentially function, such post-translational modifications are likely to play an important role in the parasite-host interaction. Moreover, post-translational modifications provide effective means to augment the information content of compact genomes, to generate diversity, and to influence antigenicity. By way of example, covalent protein modifications with N-and O-linked carbohydrate appear more and more as common features of proteins of bacterial pathogens (1). Studies have revealed a surprisingly high degree of conservation and relatedness in a number of these systems, and in some instances glycosylation-defective mutants have been shown to be attenuated in virulence-associated properties (2,3) and colonization (4 -7). Nonetheless, the full significance of protein glycosylation has yet to be precisely defined in any prokaryotic system.
Type IV pili (Tfp) 4 are proteinaceous polymeric filaments that serve critical roles in disease pathogenesis and prokaryotic cell biology in many Gram-negative species. Important human pathogens expressing Tfp include Neisseria gonorrhoeae, Neisseria meningitidis, Vibrio cholerae, Pseudomonas aeruginosa, and enteropathogenic strains of Escherichia coli (8). Tfp contribute to colonization by virtue of promoting binding to epithelial cells, motility-generating capacity, and the ability to aggregate organisms into multicellular infectious units. Tfp are also implicated in horizontal gene transfer by acting as phage receptors and being involved in many instances in competence for natural genetic transformation processes. Full comprehension of the functional relationships underlying these processes can only be achieved in the context of a thorough understanding of the structures and chemistry of Tfp, their subunit proteins, and associated molecules. For N. gonorrhoeae and N. meningitidis Tfp, such concerns are complicated by the extensive antigenic variability of the PilE pilin subunit that arises due to gene conversion-like events between a single expression locus and multiple, variable partial gene copies (9). The high degree of PilE antigenic variation likely contributes to both the failure of gonococcal infection to engender protective immunity (10) and the lack of efficacy associated with N. gonorrhoeae Tfp-based vaccines (11).
Post-translational modifications also provide potential sources for PilE structural and functional diversification. The PilE subunit of the N. gonorrhoeae Tfp colonization factor undergoes differential post-translational modifications with phosphoethanolamine (PE) and phosphocholine (PC) at serine 68 (12). This finding is remarkable because these moieties had not previously been observed to be directly O-linked to polypeptide. The presence of PC on pilin was particularly compelling because this moiety is a surface constituent of many microbial pathogens, although in all the other systems it is linked through a carbohydrate moiety. The PC moiety in other bacterial pathogens has been shown to promote epithelial and endothelial cell adherence through binding to the platelet-activating factor (PAF) receptor (13)(14)(15) and to act as an immune recognition target for both C-reactive protein (14,16) and PCrecognizing antibodies (17). In the case of parasitic nematodes, down-modulation of the immune response is attributed to PCcontaining glycoproteins (18,19). In the case of N. gonorrhoeae, the documentation of PC provided a structural basis for the ability of N. gonorrhoeae pilins to react with the TEPC-15 monoclonal antibody recognizing a PC epitope (20). A serinelinked phosphoglycerol has also been documented as being localized at residue 93 on N. meningitidis pilin (21). In this work we use the general term phospho-forms to encompass PE and PC pilin modifications.
An important aspect of PilE phospho-form modifications relates to the enzymatic pathways responsible, and given their novelty, there are no direct precedents for a biosynthetic pathway. In addition, the lic gene family members implicated in PC decoration of LPS and glycoconjugates in other species are absent in both the N. gonorrhoeae or N. meningitidis genomes (23). Taken together with the previous demonstration that phosphoglycerol is a substituent of N. meningitidis pilin, it is noteworthy that all three substituents can be found as the polar head group of eubacterial phospholipids. It is also known that in E. coli, the inner core of LPS is decorated with PE derived from phosphatidylethanolamine in a reaction requiring EptB (24) and that the genomes of most Gram-negative species contain multiple genes whose products are structurally related to EptB. Such gene products directly implicated in PE modification of LPS include Salmonella enterica CptA (25) and PmrC (26) as well as N. meningitidis LptA (27), Lpt3 (28), and Lpt6 (29). Because high frequency frame shifting events within dca (an additional N. meningitidis gene encoding an EptB homologue (30)) was correlated with phase (on-off) variation of the PilE PC epitope, it was renamed pptA as pilin phosphorylcholine transferase A (31). The structural basis for the PC epitope on N. meningitidis pilin remains unknown, although we presume it likely to be the same as described in N. gonorrhoeae pilin. Whatever the case, it remains unclear as to whether the mechanisms for PC and PE modification in N. gonorrhoeae are related to one another.
The structural data for all PilE covalent modifications have been generated by a bottom up approach in which proteolytically derived peptides were examined by electrospray ionization (ESI) mass spectrometry (MS) and tandem MS (MS/MS). Although MS and MS/MS of tryptic peptides are powerful tools for identifying post-translational modifications, they do have limitations. For example, some peptides and post-translational modifications may escape detection due to intrinsic properties relating to ionization efficiency or low stoichiometric abundance, and certain post-translational modifications may exhibit selective instability during digestion and sample preparation (at the level of small peptides). Quantitation using these techniques can also be problematic. As such, the stoichiometries of various modified forms of PilE in Tfp remain unknown, and the full extent of PilE post-translational modifications may have yet to be realized. The latter concern is made more plausible in light of data showing that at least one residue of PilE other than Ser 68 is capable of being modified with PC, although the proportion of molecules with the second site modification(s) is low (12).
Top-down MS approaches, in which post-translational modifications can be detected directly from intact proteins, have the potential to alleviate many of the concerns noted above. For example, tandem MS analysis of intact bacterial flagellin proteins has recently been used in the identification of unusual oligosaccharide structures (32). Here, we empirically developed both pilus/pilin preparation schemes and ESI MS methodologies such that intact PilE protein species could be analyzed. Using these technologies together with defined N. gonorrhoeae mutants, we present conclusive genetic evidence documenting an essential role for a previously characterized gene in the modification of PilE with both PE and PC. We also show data directly demonstrating multisite modification of PilE with PE and PC and address the relative stoichiometries of these modifications as well as identify other factors controlling phosphoform site occupancy.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Mutants-The bacterial strains used in this study are described in Table 1. E. coli and gonococcal strains were grown as described (33). E. coli HB101 was used for plasmid propagation and cloning experiments. To introduce an alanine substitution at residue 68 of PilE, the Bsu36I-StuI fragment from the plasmid pIga::pilE S68A (12) was cloned into pPilE (34) digested with Bsu36I-StuI to give the resulting plasmid pPilE S68A . The S68A substitution creates a new restriction enzyme site (Fnu4HI) that was used to verify the mutation when transformed into gonococcal strains.
Construction of a Gonococcal Strain Carrying an Alanine Substitution at Residue 63-To generate an alanine substitution at residue 63 of PilE, a PCR reaction using the primers pilE5Ј (5Ј-AAATTCCGACCCAATCAACACAC-3Ј) and S63A-3Ј (5Ј-GGGATGCCACGCCGGCAGCAGTGTTGTTTTCCGG-3Ј) (the altered codon is in bold) was used to amplify the N-terminal part of pilE from the plasmid pPilE (34). The Bsu36I-BglI fragment of this PCR product was cloned into pPilE (34) digested with Bsu36I-BglI to give the resulting plasmid pPilE S63A . Direct sequencing of PCR products derived from transformants was done using custom primers at GATC-Biotech AG (Konstanz, Germany) to ensure the correct introduction of the pilE S63A allele and the absence of any other alterations.
Construction of a dca Null Allele-Based on DNA sequence derived from the FA1090 genome project, a 1225-bp fragment of the dca open reading frame was amplified by the primers dca_5Ј-EcoRI (5Ј-CGGAATTCTTGGAAAGCAATCCTGC-CGAGG-3Ј) and dca_3Ј-BamHI (5Ј-GCGGGATCCTCCCCC-AAAAATCATAGCCG-3Ј) (EcoRI and BamHI sites are underlined). The PCR product digested with EcoRI and BamHI was cloned into pUP6 (35) digested with EcoRI and BamHI. The HindIII fragment of plasmid pCM7 (36), containing a chloramphenicol acetyltransferase gene cassette, was treated with Klenow fragment to create blunt ends and then cloned into a blunted BstEII site of the dca clone in pUP6. This BstEII site encompasses the codons for residues 270 -273 of Dca. The resulting plasmid, pdca::cat, was introduced into gonococcal strains by transformation and selection on Gc agar plates con-taining 10 g ml Ϫ1 chloramphenicol. Introduction of the cat cassette into dca was verified by PCR reactions using genomic DNA from representative transformants and the dca specific primers dca_5Ј and dca_3Ј.
Tfp Purification, SDS-PAGE, and Immunoblotting-Pilus purification was carried out as described (37). Procedures for SDS-PAGE and immunoblotting have been described previously (33). PilE was detected by immunoblotting of whole cell lysates using rabbit polyclonal antibodies and alkaline phosphatase-coupled goat anti-rabbit antibodies (Tago, Inc.). PilE-specific sera have been described previously (38). PCdecorated proteins were detected by using a 1:1000 dilution of the monoclonal antibody TEPC-15 (Sigma) and alkaline phosphatase-conjugated goat anti-mouse IgA (Sigma) (12).
CNBr Cleavage of PilE-Isolated PilE proteins were dissolved in 70% formic acid (2 mg/ml) and subjected to CNBr cleavage by adding equal volumes of CNBr solution (50 mg in 100 l of 70% formic acid). After 24 h of incubation at 37°C in the dark, CNBr was removed and trapped in NaOH pellets by repeatedly vacuum-drying the samples. CNBr-free samples were dissolved in 5% formic acid and were subject to mass spectrometric analyses or frozen at Ϫ80°C.
In-gel PilE Digestion-Coomassie-stained protein bands corresponding to PilE were serially rehydrated and washed with 150 l of HPLC grade water, 150 l of acetonitrile/water 1:1 (v/v), and 100% acetonitrile at room temperature. Protein reduction was carried out by the addition of 75 l of 10 mM dithiothreitol, 0.05 M NH 4 HCO 3 (60 min, 56°C) to the dehydrated gel pieces. Then excess reduction buffer was removed, and thiol groups were alkylated by adding 50 mM iodacetamide, 0.05 M NH 4 HCO 3 (45 min at room temperature in the dark). The gel pieces were washed again with 2 ϫ 150 l of acetonitrile/water 1:1 (v/v) and 100% acetonitrile at room temperature. Digestion buffer (5-10 l) containing either 25 ng/l trypsin (pig, modified, sequencing grade) or 125 ng/l endoproteinase Glu-C (from Staphylococcus aureus V8, sequencing grade, both from Sigma-Aldrich) in 0.05 M NH 4 HCO 3 was added, and samples were kept on ice for 30 min to allow rehydration of the gel pieces. To limit autoproteolysis of trypsin, the remaining buffer  This study  GE63D GE63 pilE S63A , dca::cm This study a VD300 is an Opa Ϫ derivative of MS11. b recA6 is an isopropyl 1-thio-␤-D-galactopyranoside-inducible allele of recA. c pilV fs is a null-allele of pilV with a frame-shift mutation at the G Ϫ1 codon (Winther-Larsen et al. (45)).
was removed and replaced with 35-50 l of 0.05 M NH 4 HCO 3 , and digestions were carried out overnight at 37°C. Peptides were extracted successively with 5% formic acid and 5% formic acid, acetonitrile (1:1, v/v) and acetonitrile. The combined supernatants were dried by SpeedVac and then re-dissolved in 5% formic acid. Samples were subjected immediately to mass spectrometric analyses or frozen at Ϫ80°C. Sample Preparation for Intact Protein MS-Isolated PilE proteins were cleaned with a methanol/chloroform precipitation procedure, as described by Wessel and Flügge (39). Briefly, 100 l of the aqueous PilE solution at 2-3 mg/ml protein was diluted 1:3 (v/v) with methanol and mixed briefly. Both 100 l of CHCl 3 and 200 l of water were added consecutively and were followed each time by a mixing step. Phase separation was achieved by centrifugation (4000 ϫ g for 8 min), yielding precipitated PilE at the interface. The upper methanol/water phase was removed, and 400 l methanol was added. After mixing, the pellet was recovered by centrifugation (13,000 ϫ g, 8 min). The pellet was dried for 5 min in the inverted tube before dissolving the sample in 50 l of water, 70% formic acid, acetonitrile 3:1:3 (v/v/v). Samples were subjected immediately to mass spectrometric analyses or frozen at Ϫ80°C.
Infusion MS Analysis of Intact PilE-All data were acquired on a quadrupole time-of-flight mass spectrometer (Q-Tof micro, Micromass, Manchester, UK) equipped with the standard Z-spray ESI source. Sample solutions were infused into the ESI source at a flow rate of 5 l/min using a syringe pump (Cole-Parmer Instrument Co., Model SP 100i, Vernon Hills, IL). The source block temperature was maintained at 80°C. Nitrogen was used as both desolvation and nebulizing gas with flow rates of 300 and 20 liters/h, respectively. Mass spectrometric analyses were performed in the electrospray positive mode with the following parameter settings (parameter names as used in the MassLynx NT software, Version 3.5): capillary voltage, 3000 V; sample cone voltage, 25 V; extraction cone voltage, 4.3 V; ion energy, 3 V; collision energy, 10 eV. Mass spectral resolution was typically 4000. The MS Survey was obtained in a mass range from 150 m/z to 1700 m/z. Mass calibration in a mass range of 100 -2200 m/z was performed using the ES tune mix solution from Agilent (Palo Alto, CA). The MS spectra were analyzed using the MassLynx software (Version 3.5). For deconvolution, spectra were processed with the MaxEnt1 program of the MassLynx software.
Nanoflow On-line Liquid Chromatographic MS Analysis of Proteolytic and CNBr Peptides-Reverse phase (C18) nano online liquid chromatographic MS/MS analyses of proteolytic or CNBr peptides were performed using a HPLC system consist-FIGURE 1. Intact mass analyses of N. gonorrhoeae PilE pilin and its CNBrderived peptide-3. Infusion-positive ESI analysis was done using the Q-Tof mass spectrometer. A, m/z spectrum over a range of 150 -1700 m/z showed a charge-state distribution from ϩ11 to ϩ24. Oxonium ions for hexose, hexose-DATDH, and acetylated hexoseDATDH are seen. Note that the site of acetyl linkage to the disaccharide has yet to be defined. B, a deconvolution algorithm was applied to the multiple charged spectrum of A, and a deconvoluted molecular weight spectrum was generated. Multiple forms reflect species differing in glycosylation and PE modification. A complete list of m/z forms and corresponding species of PilE is found in supplemental Table 1. C, a deconvoluted MW spectrum of the CNBr-derived peptide-3 of wild type PilE showing the presence of both unmodified and single-PE-modified forms (minor peaks in the MS spectra toward the higher mass region of m/z 7330 and 7453 are caused by metal-ion adducts (Na ϩ , ϩ23 Da; K ϩ , ϩ39 Da) or by peptide formylation (ϩ28 Da). SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38

JOURNAL OF BIOLOGICAL CHEMISTRY 27715
ing of a Waters CapLC binary pump with corresponding autosampler (Waters Corp., Milford, MA). This LC system was coupled to the quadrupole time-of-flight mass spectrometer equipped with the nanoflow ESI source or, alternatively, with an ion trap mass spectrometer (Bruker Esquire HCT, Bruker Daltonics, Bremen, Germany). For the analyses, 10 l of peptide solution was injected onto the 5 ϫ 0.3-mm extraction column filled with Zorbax 300 SB-C18 of 5-m particle size (Agilent). Samples were washed with a mobile phase of 100% 0.1% formic acid. The flow rate was 20 l/min. After 6.5 min, the switching valve of the stream select module (Waters) was activated, and the peptides were eluted in the back-flush mode from the extraction column onto the 150 ϫ 0.3-mm C 18 , 3.0-m capillary column (ProntoSIL C18 -300, G&T SEP-TECH AS, Kolbotn, Norway). The mobile phase consisted of acetonitrile and HPLC grade water, both containing 0.1% formic acid. Chromatographic separation was achieved using a binary gradient from 10 to 90% of acetonitrile in 85 min. The flow rate was 0.5 l min Ϫ1 . Fifteen minutes after column switching, the positive ESI mass spectrometric analysis was started. In MS/MS experiments, the automatic selection of the precursor ion was based on the three most intensive peaks (double-, triple-, or quadruple-charged) observed in the spectra obtained from the MS Survey. Alternatively, ion selection for MS/MS experiments was done manually. Fragments of mass selected peptide ions were induced by collision-induced dissociation in the collision cell using argon as the collision gas. The collision-induced dissociation gas cell pressure was 1.5 ϫ 10 Ϫ3 millibar. Scan duration in the survey mode as well as in the MS/MS mode was set to 2.5 s. Data were acquired and processed by the MassLynx NT software (Version 3.5).
Nano on-line liquid chromatographic MS/MS experiments conducted on the Q-Tof were run under the same settings as described for infusion whole protein MS with the exception that the collision energy in MS/MS experiments were optimized for the corresponding peptides and ranged from 14 to 45 eV. Alternative LC-MS n experiments (n ϭ 2, 3) were conducted on the Bruker Esquire quadrupole ion trap instrument equipped with a manufacturer's online nanospray interface that was fitted with a distal metal-coated fused-silica PicoTip needle (PicoTip, FS360-20-10-D5-C7; New Objective, Woburn, MA). The ionizer and ion transfer optics parameters were as follows: capillary voltage, Ϫ1500 V; end plate, Ϫ1000 V; capillary exit, 238 V; skimmer, 40 V; octopole I offset, 7 V; octopole II offset, 2 V; octopole radio frequency, 200 V; peak-topeak, Ϫ5 V (lens I) and Ϫ60 V (lens II). The drying gas flow rate was 3 liters/min, the drying gas temperature was 220°C, and the trap drive was set to 61. Using ESI in positive ion mode, mass spectra were acquired from m/z ϭ 250 to 1600. All MS experiments were performed using the ion charge control facility with an accumulation time of 300 ms and with an ion target of 100,000. In MS n experiments (n ϭ 2,3), precursor ion selection was performed manually, and the ion trap was operated in "smart fragmentation mode," whereby the fragmentation amplitude of 0.8 V was ramped from 30 to 200% over a period of 40 ms. Data were acquired by the Compass software (Version 1.0) and processed by the BioTools software (Version 1.1).

Intact Mass Analysis of PilE, the N. gonorrhoeae Tfp Pilin
Protein-Sample preparation conditions allowing electrospray ionization of intact pilin protein were empirically established. Analysis of purified Tfp from N. gonorrhoeae N400 revealed a broad, heterogeneous envelope of multiply charged ions extending over a range of 500 -1700 m/z as well as the presence of oxonium ions for DATDH and HexDATDH at m/z 229.1 and 391.2, respectively (Fig. 1A). A prominent signal was also observed at m/z 433.2, and further analysis has shown that this species is likely HexDATDH modified by acetylation (acHex-DATDH, 391.2 ϩ 42). 5 The reconstructed molecular mass profile of the corresponding spectrum indicated a remarkably broad distribution of modified forms (Fig. 1B). The most abundant species observed in the mass spectrum of intact PilE correspond to the protein modified with a single acHexDATDH moiety and PE at m/z 17733, consistent with earlier observations (12). Additional modified PilE forms were also observed that could be assigned to pilin protein differentially modified with PE alone and in combination with DATDH or Hex-DATDH moieties of m/z 17301, 17529, and 17691. Surprisingly, additional signals were observed that are consistent with the presence of an additional PE modification on PilE. The mass of these forms observed at m/z 17425, 17652, 17814, and 17857 are consistent with the addition of a further PE mass increment. Thus, PilE in purified Tfp from this wild type background is detected as a mixture of single-and double-PE modified forms.

Identification of Ser 156 as a Second Site for Covalent Modification of PilE with PE-LC-MS analyses of tryptic digests of
PilE failed to reveal any peptides other than 57 WPENNT-SAGVASPPTDIK 74 (encompassing Ser 68 ) with a characteristic shift in mass of 123 Da. Therefore, PilE was digested with Glu-C, and the resulting proteolytic fragments were analyzed by LC-MS. Here, two hydrophilic peptides were observed that eluted close to one another with retention times of 35.6 and 35.7 min. Fig. 2 shows a total ion chromatogram of Glu-C digestion fragments as well as two selected ion chromatograms, chosen based on the predicted mass of the Glu-C-derived proteolytic peptide 141 IDTKHLPSTCRDKASDAK 159 either unmodified or modified with PE (both peptides contain also carboxyamidomethyl cysteine instead of cysteine). Both eluting peptides showed four major signals in their mass spectra corresponding to multiply charged molecular ions: 35 5ϩ . The mass difference between the two sets of ion series derived from the eluted peptides is consistent with PE. The identified candidates were then subjected to further MS/MS and MS 3 experiments to confirm the identity of the peptide and the exact site of modification.
The modified peptide candidate lost predominantly 141 Da when subjected to low collision energy MS/MS experiments in a Q-Tof mass spectrometer (Fig. 3) or the ion trap (Fig. 4B), suggesting its modification with PE. The mass difference of 141 Da observed between m/z 2165.9 and 2025.0 corresponds to the loss of PE, concomitantly with 1 molecule of water resulting in the formation of a dehydroalanine residue in the peptide chain. Only small signals within the MS spectra corresponded to the loss of either ethanolamine or PE without additional water. The dominant loss of 141 Da was charge state independent in both instruments. Among all charge states subjected to collision- induced dissociation MS/MS analyses, the doubly charged ions of the modified peptide generated spectra with the best quality, showing more peptide backbone fragmentation than other spectra in variously charged states (valid for both instruments). Although a complete series of y-ions was not observed, partial backbone fragmentation during Q-Tof analysis allowed unambiguous identification of the peptide as the C-terminal Glu-C cleaved residue of PilE (Fig. 3). In addition, localization of PE to Ser 156 is derived from the observation of fragment ions y 3 and y 4 , representing a mass difference of 69 Da, corresponding to a dehydroalanine residue formed by the loss of PE from Ser 156 . Furthermore, the signal for the y 4 -ion modified with PE (543.2 m/z) was detectable (Fig. 3). Further evidence for the PE modification at Ser 156 , was obtained by ion trap MS/MS experiments on both the unmodified and modified peptides (Fig. 4, A  and B). The signals 473.2 and 601.3 m/z, representing the dehydro y 5 and y 6 ions of the modified peptide, are replaced by signals at 491.2 and 619.3 m/z in the spectrum of the unmodified peptide, indicating an intact serine residue at Ser 156 .
The final MS 3 spectrum (1083.9 Ͼ 1012.9 m/z) of the modified peptide generated by the ion trap exhibits only the fragment ions 473.2 and 601.3 m/z, which correspond to the insertion of dehydroalanine (Fig. 4C). The spectrum confirms that the investigated peptide peak is not a mixture of peptides modified at different serine or threonine residues within the peptide chain. In summary, these results showed that Ser 156 is a second site for PilE covalent modification with PE.
Influence of an Alanine Substitution at Residue 68 on PilE Phospho-form Modification-Given the evidence for PE modifications at both Ser 68 and Ser 156 , it was of interest to determine the relative stoichiometries of modifications at these sites. For example, in the case of the single-modified PilE forms seen in the intact protein spectra, it is not clear what fraction of these species was accounted for by those with modification at residue 68 versus those modified at residue 156. It is also unknown if such species might be composed of molecules with PE modification at as yet unidentified residues. To begin to address these concerns, PilE protein from wild type Tfp was digested with cyanogen bromide, and the C terminal fragment (termed CNBr-derived peptide-3 and corresponding to residues 93-159) was isolated and characterized. As seen in Fig. 1C, the two most prominent species from wild type Tfp represented CNBr-derived peptide-3 in its unmodified and single-PE-modified forms. As another means to assessing the stoichiometries of PE modifications, intact mass analysis of PilE from the strain expressing PilE with an alanine substitution at residue 68 (carrying the pilE S68A allele) was performed. The reconstructed molecular mass profile here was devoid of any species repre-  senting PilE bearing two PE modifications, consistent with Ser 68 being the predominant site for PE modification (Fig. 5). Furthermore, the profile of CNBr-derived peptide-3 species from this background was nearly identical to that seen for the corresponding wild type preparation save for a slight reduction in the relative level of the single PE-modified form at m/z 7453. Taken together, these data are consistent with wild type Tfp being composed of a population of PilE species bearing PE modifications at both Ser 68 and Ser 156 together with a second population bearing only the modification at Ser 68 .
Influence of the Pilin-like Protein PilV on PilE Phospho-form Modification-Earlier observations of the PilE tryptic peptide 57 WPENNTSAGVASPPTDIK 74 showed that the pilin-like protein PilV influences phospho-form modification at Ser 68 (12). In pilV null mutants a mixture of this peptide modified with either PE or PC was observed, and when PilV was overexpressed, the peptide devoid of modification was found. Intact mass analysis of PilE from a pilV background revealed a dramatic shift in which species with 2 PE modifications predominated (m/z 17856) (Fig. 5B) together with the appearance of a new PilE form modified with both PE and PC (m/z 17898). Data for CNBr-derived peptide-3 from this background revealed a corresponding relative increase in a PE-modified form (m/z 7453), although signals corresponding to species modified with PC were not seen (Fig. 5B, inset). Thus, in addition to promoting PilE PC modification, the pilV null mutation is associated with increased PE modification of a site in CNBr-derived peptide-3 presumably occurring at Ser 156 . These changes were specific to the absence of PilV because they were rectified by expression of pilV from an ectopic site in this background (data not shown). The presence of the pilE S68A allele in the pilV background has been shown previously to lead to a 15-fold reduction in PilE reactivity with a PC-recognizing monoclonal antibody (12), a finding inferred to reflect reduced PC modification. This presumption was confirmed in the analysis of intact PilE from this strain that showed no signals corresponding to PCmodified forms along with relatively high levels of PilE species devoid of phospho-form modification and reduced levels of single-PE modified species (Fig. 5C). Surprisingly, analysis of CNBr-derived peptide-3 did not show the increase in the PEmodified form seen for wild type PilE in a pilV background. Therefore, phospho-form hypomodification associated with the S68A mutant in a pilV background is not attributable solely to the lack of modification at residue 68. Finally, analysis of intact PilE and CNBr-derived peptide-3 from a strain overexpressing PilV (carrying the pilV oe allele) revealed a spectrum consistent with reduced phospho-form modification at both Ser 68 and at a residue within CNBr-derived peptide-3 (Fig. 5D).
Dca (PptA) Is Required for PilE Phospho-form Modification-Given the uniqueness of PilE phospho-forms and the absence of lic-related genes in N. gonorrhoeae, there were no direct precedents for the biosynthetic pathway involved. The N. gonorrhoeae genome does contain four genes whose products are structurally related to EptB and other members of the so-called YhjW/YjdB/YijP protein family implicated in PE modification of LPS (27,28,29). We disrupted each of these genes in a pilV background and found that only dca (pptA) null mutations abolished PilE reactivity with monoclonal antibody TEPC-15,  Table 1. Note that minor peaks in the CNBr-derived peptide-3 MS spectra toward the higher mass region of m/z 7330 and 7453 are caused by metal-ion adducts (Na ϩ , ϩ23 Da; K ϩ , ϩ39 Da) or by peptide formylation (ϩ28 Da).
recognizing the PC epitope (Fig. 6). Intact mass analyses of PilE from two strains bearing different dca (pptA) null alleles were virtually identical to one another, with both devoid of any signals characteristic of phospho-form-modified species (Fig. 6A and supplemental Fig. S1). Signals for PilE derived from a dca (pptA), pilV double mutant were identical to those seen for the dca (pptA) mutants (Fig. 6B). To ensure that the lack of phospho-form modification was due to the absence of Dca (PptA) and not polar effects of the mutations used, PilE from a dca(pptA) mutant strain, which carried a wild type allele at an ectopic site, was examined. Here only modest signals reflecting single PE-modified forms were found (Fig. 6C) but when complementation was examined in a pilV background, these signals were remarkably increased (Fig. 6D). Although no signals characteristic of PC modification were seen, the PC-recognizing monoclonal antibody reacted with PilE in this background (Fig. 7). Together, these results demonstrated unequivocally that Dca (PptA) is required for PilE post-translational modification with both PE and PC and that pilV is genetically downstream of dca (pptA) in influencing phospho-form modification.
Influence of an Alanine Substitution at Residue 63 on PilE Phospho-form Modification-Both MS data on tryptic peptides generated from strain N400-derived PilE used here (12) and x-ray crystallographic data of a variant strain PilE (40) demonstrated the presence of a disaccharide linked to Ser 63 . To assess if glycan occupancy at this site might influence phospho-form modifications, intact mass analyses was carried out on pilin from a strain expressing PilE with an alanine substitution at residue 63 (carrying the pilE S63A allele). Here, the reconstructed molecular mass spectrum indicated a limited distribution of PilE modified forms at m/z 17285 and 17408, which can be accounted for solely by the presence of one or two PE modifications with a minor signal at m/z 17162 reflecting unmodified PilE (Fig. 8A). Interestingly, the results also revealed a shift in which species with two PE modifications predominated that was confirmed by a concurrent increase in the levels of PEmodified CNBr-derived peptide-3 (Fig. 8A, inset). Finally, intact mass analysis of Tfp-derived PilE from a strain expressing the pilE S63A allele in a dca/pptA background revealed a profile corresponding to nonglycosylated PilE devoid of phospho-form modifications (Fig. 8B). Because levels of Tfp were undiminished in this as well as the other backgrounds tested (data not shown), abolition of glycosylation and/or phospho-form modifications do not grossly perturb PilE assembly proficiency.

DISCUSSION
PE and PC are increasingly being seen as substituents of surface glycoconjugates in a variety of Gram-negative and Grampositive pathogens. Although the importance of YhjW/YjdB/ YijP protein family members in the addition of PE to the core and lipid A of LPS is well established, the enzyme(s) that generates phospho-form-modified N. gonorrhoeae Tfp has remained unknown. Here, we show that N. gonorrhoeae pptA, nee dca, is necessary for PilE phospho-form modification and propose accordingly that PilE phospho-form modification is mechanistically and evolutionarily related to LPS PE modification systems. Such a model is remarkable given that in one case covalent linkage occurs through a hydroxyl group at serine residues, whereas in the other it occurs at hydroxyl groups on sugars within the core and lipid A domains of LPS. It is also extremely interesting that although both pathogenic and commensal Neisseria species have found it advantageous to display PC at their surfaces, the former does it in the context of Tfp and involves an enzyme related to LPS PE transferases, whereas the latter does it in the context of LPS and utilizes the lic gene pathway. Because pptA and the lic genes are mutually exclusive, the crux of this dichotomy ultimately lies in the gene gain/loss   Table 1. Note that minor peaks in the CNBr-derived peptide-3 MS spectra toward the higher mass region of m/z 7330 and 7453 are caused by metal-ion adducts (Na ϩ , ϩ23 Da; K ϩ , ϩ39 Da) or by peptide formylation (ϩ28 Da). events occurring during Neisseria evolution and the relative benefits that accrued from them.
Although it is attractive to assume that PptA acts as a PE transferase as proposed for EptB (formerly YhjW), some caution may be warranted. This assumption is based in most instances solely on genetic evidence; that is to say that null mutations at corresponding loci are associated with the absence of PE in target molecules. The exception is eptB, which when overexpressed yielded high levels of membrane-associated PE transferase activity for Kdo 2 -lipid A (24). Here in this work reintroduction of a pptA gene copy at an ectopic locus in the pptA null mutant restored PE and PC modification, albeit not to levels seen in wild type cells. This result is attributable to reduced level of pptA expression from the ectopic site since it lacks its normal promoter. Nonetheless, this finding formerly rules out polarity on distal gene expression as a contributor to the phenotype seen in N. gonorrhoeae pptA mutants.
Comparisons of PptA and EptB as PE transferases also suggest that common precursors for PE may be involved. In the case of EptB, this is undoubtedly phosphatidylethanolamine, as it is required for transferase activity, and stoichiometric levels of diacylglycerol are generated during the reaction (24). In pathogenic Neisseria phosphatidylethanolamine is the major membrane phospholipid species, but phosphatidylcholine has only been detected once, and there it was reported to constitute only 3% of the phospholipid profile (41). A more recent study of N. gonorrhoeae phospholipids using fast atom bombardment-MS and gas liquid chromatography-MS technologies did not note the presence of phosphatidylcholine (42). In light of these data, one cannot rule out the possibility that the role of PptA in PilE modification with PC is indirect. In such a scenario PptA would decorate PilE with PE that would subsequently be modified into PC by the action of PE methyltransferase(s). The designation of PptA as a pilin PC transferase is, therefore, not substantiated by current data. Nonetheless, we support the retention of this nomenclature with the modification that PptA corresponds to pilin phospho-form transferase A.
Our findings also demonstrate that PilE undergoes multisite modifications requiring PptA. As documented by MS and reactivity with the PC epitope recognizing monoclonal antibody TEPC-15, both Ser 68 and Ser 156 are capable of being modified with either PE or PC. It is also important to note that Ser 156 , in contrast to Ser 68 , maps in a variable domain of pilin subject to amino acid substitutions. Thus, changes in PilE primary structure can directly impact phospho-form modification by creating and eradicating occupancy sites. Likewise, it was possible to assess the relative degree of occupancy at these sites by PE and PC by quantitation of ion currents of various intact PilE forms standardized against one another. Thus, in a wild type background, PilE in purified Tfp is composed of a mixture of singleand double-PE-modified forms in which modification at Ser 68 predominates. Remarkably, the pilin-like protein PilV profoundly influenced modification at both sites. Given the hypermodification seen in the pilV background and hypomodification seen for the pilV oe background, PilV appears to impact on the system by inhibiting PptA-mediated modification. It remains to be determined whether PilV exerts its effects by acting directly on PptA, influencing accessibility of PilE to PptA or through some other controlling mechanism.
Changes in PilE primary structure were also associated with alterations in phospho-form modifications. In addition to the profiles anticipated from the absence of a major occupancy site, substitution of alanine for serine at residue 68 led to reduced occupancy at Ser 156 , and a similar effect was seen for the Ser 68 substitution mutant in a pilV background. These findings are consistent with a hierarchical model in which occupancy at Ser 68 increases the likelihood for modification at Ser 156 . In the case of the Ser 63 substitution mutant that precluded glycosylation, an opposite effect was seen in which levels of PilE with two PE were increased. In the data derived for both the Ser 63 and Ser 68 substitutions from intact protein analyses, one cannot rule out that the substitutions themselves influence the profiles indirectly by altering ion signal intensities or other aspects of detection efficiencies. Nonetheless, the data for the CNBr-derived peptide-3 (whose primary structures are unchanged in these backgrounds) nicely document altered site occupancy in these backgrounds. It is also noteworthy that the N. gonorrhoeae PilE PC epitope has been reported to be phase-variable (20). In N. meningitidis, this phenomenon correlated with high frequency, frame-shifting events within a monotonous stretch of G residues in pptA, but N. gonorrhoeae pptA lacks this stretch of residues (and, thus, presumably does not undergo high frequency, frame-shifting). Therefore, there must be other factors responsible for this phenomenon. We have observed that in some instances the PC epitope can be detected despite the expression of PilV and, conversely, that phospho-form hypermodification does not always ensue from pilV null mutations. In all these cases the altered behavior is associated with expression of a novel PilE variant arising by gene conversion. Like single substitutions within a conserved domain (see above), it appears that changes in the variable domains of PilE arising by combinatorial diversification can also influence phospho-form modification patterns.
Another observation here relates the increased signals corresponding to intact, non-glycosylated PilE forms seen in the pilE Ser 68, pilV oe , and pptA backgrounds, each of which is associated with decreased or abolished phospho-form modification at Ser 68 . We consider two explanations for these correlations, one in which phospho-form-site occupancy at Ser 68 in some way promotes glycosylation and a second in which Ser 68 -site occupancy stabilizes the covalent carbohydrate linkage to withstand in source fragmentation. We currently favor the latter scenario because the levels of glyco-free PilE forms seen for the pilE Ser 68, pilV oe , and pptA pilus samples can be considerably diminished by altering ionization conditions (data not shown). Whatever the basis for these correlations, the data provide the first evidence for interplay between PilE glycosylation and phospho-form modification and further suggests that the proximity of modifications at Ser 63 and Ser 68 is not entirely fortuitous.
In summary, N. gonorrhoeae PilE structure is more complex and dynamic than originally conceived based solely on combinatorial gene diversification and earlier structural and MS studies. This situation is likely to have significance for how pilin subunits are assembled and organized into helical filaments, particularly if one assumes that phospho-form sites are exposed on the pilus surface. It follows then that PC and PE and reagents recognizing them may serve as valuable probes to assess PilE topology within Tfp. Other important facets remaining to be addressed concern the catalytic mechanisms of PptA and the identification of the precursors involved. Here, insights into the properties and structure of PptA will be pivotal to characterizing the pathway and its elements. Finally, the ability to produce structurally defined and altered PilE molecules should help elucidate the potential contributions of post-translational modifications to Tfp-associated functions and phenotypes.