Posttranslational Modification of Flagellin FlaB in Shewanella oneidensis

Abstract

Shewanella oneidensis is a highly motile organism by virtue of a polar, glycosylated flagellum composed of flagellins FlaA and FlaB. In this study, the functional flagellin FlaB was isolated and analyzed with nano-liquid chromatography-mass spectrometry (MS) and tandem MS. In combination with the mutational analysis, we propose that the FlaB flagellin protein from S. oneidensis is modified at five serine residues with a series of novel O-linked posttranslational modifications (PTMs) that differ from each other by 14 Da. These PTMs are composed in part of a 274-Da sugar residue that bears a resemblance to the nonulosonic acids. The remainder appears to be composed of a second residue whose mass varies by 14 Da depending on the PTM. Further investigation revealed that synthesis of the glycans initiates with PseB and PseC, the first two enzymes of the Pse pathway. In addition, a number of lysine residues are found to be methylated by SO4160, an analogue of the lysine methyltransferase of Salmonella enterica serovar Typhimurium.

INTRODUCTION

Shewanella species, renowned for their respiratory versatility, are motile by a polar flagellum in general, which has been characterized in the model species S. oneidensis recently (1–4). Interestingly, some possess an additional lateral flagellar system whose assembly is condition specific (5, 6). According to the genome annotation data, the polar system in all Shewanella strains sequenced so far contains two flagellins, except for S. baltica OS185 and OS195, whose four flagellins most likely arise from transposition events (4). The polar flagellar systems within the genus are remarkably heterogeneous in terms of flagellin size, ranging from ∼270 amino acids (aa) for the majority to ∼465 aa. Flagellins FlaA (SO3238) and FlaB (SO3237) of S. oneidensis, 273 and 272 aa in length, respectively, are rather small, given that the minimum length for any functional flagellin is ∼250 aa (7). Despite sharing a sequence identity of 89%, FlaA and FlaB differ substantially in their ability to propel cells (4). FlaB is undoubtedly the major flagellin, because a flaB mutant (flaA⁺ ΔflaB) exhibits a significantly reduced swimming/swarming capability. On the contrary, the removal of FlaA (ΔflaA flaB⁺) has little or no effect on filament assembly or motility (4).

An enormous number of proteins that are translated from mRNA undergo modifications before becoming functional. Flagellins have been shown to be posttranslationally modified by methylation at lysine residues and/or glycosylation at serine or threonine residues (8). Although first reported in Salmonella enterica serovar Typhimurium many years ago, the role of the methylation remains unclear, in part due to the finding that loss of methyltransferase has no negative impact on flagellar function (9). In contrast, a wealth of information has been gathered about O-linked glycosylation of bacterial flagellins since its discovery about 2 decades ago (8, 10–12). Presumably, glycosylation of flagellins takes place in the cytoplasm in close proximity to the basal body (10, 13). For the bacterial flagellin characterized to date, the sites of glycosylation of flagellin monomers appear to be specific to some extent, exclusively residing within the two surface-exposed domains (14). It is currently unclear how this specificity is achieved, as no consensus sequences have been identified. In the assembled filament, the glycosylated domains are positioned on the surface and are capable of interacting with either host cells or environmental substrates. In general, glycosylation of flagellins is essential for bacterial flagellar expression, assembly, motility, virulence, and host specificity (8, 10, 11, 15–17). However, exceptions are found in Pseudomonas spp. and Burkholderia spp., in which glycosylation is not required for flagellar assembly (18–20). It is expected that more variations in structure and glycosylation of flagellins will be revealed with the characterization of flagellar systems from other phylogenetically diverse species.

The flagellins of S. oneidensis appear to be glycosylated, as the removal of SO3271, a protein with 66% sequence identity to Campylobacter jejuni PseB, results in a nonmotile phenotype (4). PseB, the first enzyme of the Pse pathway, has been shown to be essential for glycosylation of C. jejuni and Helicobacter pylori flagellins (21). Compared to the 495-aa FliC of S. enterica serovar Typhimurium (domain structure, D₀-D₁-D₂-D₃-D₂-D₁-D₀), both S. oneidensis FlaA and FlaB flagellins retain only four domains that are functionally essential: D₀-D₁-D₁-D₀ (4, 22). The missing D₂-D₃-D₂ segment, replaced by a 23-residue fragment in S. oneidensis, is particularly important for glycosylation, because it is surface exposed. In C. jejuni, all of 22 glycosylation residues reside in this region, of which only one is identified in the S. oneidensis fragment (23–25) (see Fig. S1 in the supplemental material). To date, Pseudomonas syringae has been the only bacterium whose flagellin, similar to S. oneidensis counterparts in size (282 aa), has been characterized in regard to glycosylation (26). In total, six residues within a fragment covering residues from 143 to 201 are subjected to glycosylation in the P. syringae flagellin. However, a sequence alignment reveals that three of these six residues are either missing or replaced by unglycosylatable amino acids in S. oneidensis flagellins (see Fig. S1). Furthermore, it is worth noting that the sequence conservation between the 23-residue fragment of S. oneidensis and D₂-D₃-D₂ segments of other bacterial flagellins is considerably low, making prediction of glycosylation residues unreliable. Thus, pursuing analyses of the sites for glycosylation in S. oneidensis flagellins could reveal intriguing insights.

MATERIALS AND METHODS

Bacterial strains, plasmids, PCR primers, and culture conditions.

A list of all bacterial strains and plasmids used in this study is given in Table 1. Escherichia coli and S. oneidensis strains were grown in Luria-Bertani (LB; Difco, Detroit, MI) medium at 37 and 30°C, respectively, for genetic manipulation. Where needed, antibiotics were added at the following concentrations: ampicillin at 50 μg/ml, kanamycin at 50 μg/ml, and gentamicin at 15 μg/ml. Information on primers used for generating PCR products in this study is given in Table S1 in the supplemental material.

Table 1.

Strains and plasmids used in this study

Strain or plasmid	Description	Reference or source
E. coli strains
DH5α	Host for regular cloning	Laboratory stock
BL21(DE3)	Expression host	Novagen
WM3064	Donor strain for conjugation; ΔdapA	W. Metcalf, UIUC
S. oneidensis strains
MR-1	Wild type	Laboratory stock
FFM	flaA flaB deletion mutant derived from MR-1; ΔflaA ΔflaB	4
HG3270	pseC deletion mutant derived from MR-1; ΔpseC	This study
HG3271	pseB deletion mutant derived from MR-1; ΔpseB	4
HG4160	fliB deletion mutant derived from MR-1; ΔfliB	This study
Plasmids
pDS3.0	Apr, Gmr; derivative from suicide vector pCVD442	Laboratory stock
pHG101	Promoterless broad-host Kmr vector	4
pHG102	pHG101 containing the S. oneidensis arcA promoter	4
pHG101-flaB	Complementation and template for site-directed mutagenesis	4
pHG101-fliB	Complementation for ΔfliB	This study
pHG102-pseC	Complementation for ΔpseC	This study
pET28a	Apr; expression vector	Novagen
pET28a-pseB	pET28a containing pseB for expression in E. coli	This study
pET28a-pseC	pET28a containing pseC for expression in E. coli	This study

Mutagenesis and complementation of mutation strains.

In-frame deletion strains were constructed using a fusion PCR method essentially as previously described (27). In brief, two fragments flanking the targeted gene were amplified independently first and fused together by the second round of PCR. The resulting fusion fragment for each individual gene was introduced into plasmid pDS3.0. The resulting mutagenesis vector was propagated in E. coli WM3064 and then transferred into S. oneidensis by conjugation. Integration of the mutagenesis construct into the chromosome was selected by gentamicin resistance and confirmed by PCR. Verified transconjugants were grown in LB broth in the absence of NaCl and plated on LB supplemented with 10% sucrose. Gentamicin-sensitive and sucrose-resistant colonies were screened by PCR for deletion of the targeted gene. The deletion mutation was then verified by sequencing of the mutated region. For complementation of genes next to their promoter, a fragment containing the targeted gene and its native promoter was generated by PCR and cloned into pHG101 (4). For other genes, the targeted gene was amplified and inserted into the multiple cloning site (MCS) of pHG102 under the control of the arcA promoter. Introduction of each verified complementation vector into the corresponding mutant was done by mating with E. coli WM3064 containing the vector and confirmed by plasmid extraction and restriction enzyme mapping.

Site-directed mutagenesis.

Plasmid pHG101-flaB was used as the template for site-directed mutagenesis with a QuikChange II XL site-directed mutagenesis kit (Stratagene). Mutated PCR products were generated with primers listed in Table S1 in the supplemental material, subsequently digested by DpnI, and transformed into E. coli WM3064. After sequencing verification, the resulting plasmid was transferred into the S. oneidensis strains by conjugation.

Physiological characterization of mutant strains.

Growth of mutant strains in LB was measured by recording cell densities of cultures at an optical density of 600 nm (OD₆₀₀) under aerobic conditions in triplicate with the parental wild type as the control. Motility testing (swimming) was performed with semisolid LB agar plates (0.25% agar). Briefly, mid-log-phase cultures were adjusted to an equivalent OD₆₀₀ of 0.4 for each strain with fresh LB broth, 5 μl of which was spotted onto a swimming plate by piercing it with a thin pipette tip. Plates were incubated at room temperature. For microscopic analysis, swimming cells were scraped from the leading edges of each swarm, stained for flagellar filaments, and visualized on a glass slide with a Motic BA310 phase-contrast microscope (28). Determination of the swimming speed of the cells was carried out essentially as described elsewhere (6). Micrographs were captured with a Moticam 2306 charged-coupled-device camera and Motic Images Advanced 3.2 software.

Electron microscopy visualization.

For transmission electron microscopy (TEM), cells grown overnight on LB agar plates were suspended in sterile distilled water, spread onto carbon-Formvar copper grids, and then negatively stained with 1% phosphotungstic acid (pH 7.4). Preparations were viewed under a CM12 Philips TEM.

Extraction and purification of flagellin FlaB.

A 250-ml bacterial batch culture was centrifuged at 5,000 × g for 10 min at 4°C. The cell pellet was resuspended in 5 ml phosphate-buffered saline (PBS), pH 7.0, and vortexed for 10 min to shear off flagella. The cells were removed by centrifugation at 10,000 × g for 30 min at 4°C, and the supernatant containing flagella was filtered through a 0.45-μm-pore filter. The filtrate was centrifuged at 100,000 × g for 2 h, and the pellet containing purified flagella was resuspended in double-distilled water (ddH₂O). Sample purity was checked by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), and the protein concentration was determined using a Bradford assay with bovine serum albumin (BSA) as a standard (Bio-Rad).

SDS-PAGE and immunoblotting assay.

Protein samples were loaded onto SDS-PAGE (10%) and either stained with Coomassie brilliant blue or electrophoretically transferred to a polyvinylidene difluoride (PVDF) membrane according to the manufacturer's instructions (Bio-Rad). The gels were blotted for 2 h at 60 V using a Criterion blotter (Bio-Rad). The blotting membrane was probed with antibodies that recognize both FlaA and FlaB flagellins of S. oneidensis as used previously (4). The goat anti-rabbit IgG-HRP (horseradish peroxidase) (Roche Diagnostics) was used as the secondary antibody (1:5,000) and the signal was detected using a chemiluminescence Western blotting kit (Roche Diagnostics) in accordance with the manufacturer's instructions. Images were visualized with a UVP imaging system.

His₆-tagged protein expression and purification.

PCR was used to amplify S. oneidensis SO3271 and SO3270 with the genomic DNA as the template. The resulting products were cloned into pET28a, generating fusion proteins with a His₆ tag at the N terminus, and transformed into E. coli BL21(DE3). Expression of the SO3271 and SO3270 proteins in E. coli BL21(DE3) cells was similar to previously described methods (27, 29). Briefly, cells were induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) from the mid-log phase (OD₆₀₀ of 0.5 to 0.6) at 30°C. After being grown to saturation, the cells were collected by centrifugation, resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 1 mM MgCl₂, 10 mM β-mercaptoethanol, 1 mM phenylmethylsulfonyl fluoride, 5 μg/ml DNase I), and broken by passage twice through a French press (10,000 lb/in²). The soluble proteins were purified using a talon resin column according to the manufacturer's instructions. Fractions containing the purified protein of interest, as determined by SDS-PAGE (12.5%) and Coomassie staining, were pooled and dialyzed against dialysis buffer (25 mM sodium phosphate, pH 7.3, 50 mM NaCl) overnight at 4°C. Protein concentration was determined spectrophotometrically and using a Bradford assay with BSA as a standard (Bio-Rad).

Enzymatic reactions.

Enzymatic reactions were performed in HEPES [4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid] buffer at 37°C with chemicals from Sigma. The purified His₆SO3271 (6 μg/ml) was first used in the single-step reaction with 0.5 mM UDP-GlcNAc, and the reaction mixture was then examined by mass spectrometry (MS) analysis. The stepwise enzymatic synthesis of intermediates or products was accomplished by mixing 0.5 mM UDP-GlcNAc, 0.1 mM pyridoxal-5′-phosphate (PLP), 10 mM l-glutamate (Glu), 6 μg/ml His₆SO3271, and 40 μg/ml His₆SO3270. The results were examined using a Waters Micromass Quattro II triple quadrupole mass spectrometer (Waters, Milford, MA).

Mass spectrometry analysis.

Purified flagellar filaments were subjected to MS as previously reported (30). Briefly, flagellin isolates (50 μg) were digested overnight with trypsin (Promega, Madison, WI) at a protein/enzyme ratio of 30:1 (vol/vol) in 50 mM ammonium bicarbonate at 37°C. The flagellin digests were analyzed by nano-liquid chromatography-tandem MS (nano-LC-MS/MS) using a Q-Tof Ultima hybrid quadrupole time-of-flight mass spectrometer coupled to a NanoAquity ultra performance liquid chromatography (UPLC) system (Waters). The digests were injected onto a 5-mm- by 300-μm-inner diameter (i.d.) Acclaim PepMap100 C₁₈ μ-precolumn (Dionex/Thermo Scientific, Sunnyvale, CA) and separated on a 100-μm- by 100-mm-i.d., 1.7-μm BEH130 C₁₈ column (Waters) using the following gradient conditions: 5 to 45% acetonitrile (ACN) in 0.2% formic acid for 35 min and 45 to 95% ACN for 5 min at 450 nl/min. MS/MS spectra were acquired on doubly, triply, and quadruply charged ions and searched against the NCBInr database using the Mascot search engine (Matrix Science Ltd., London, United Kingdom). Glycopeptide MS/MS spectra were interpreted manually.

Comprehensive glycopeptide mapping was achieved using a stepped-collision offset to generate diagnostic glycan ions as well as intact peptide and glycopeptide ions. This was performed using the same nano-LC-MS setup as that described above and by the collision offset between alternate scans: 10 V (no fragmentation) and 35 V (fragmentation). Glycopeptide ions were identified by aligning the glycan fragment ions produced using the higher collision offset with the intact ions of the low-collision offset conditions (i.e., intact ions with exactly the same retention time profile as the glycan oxonium ions). Front-end collision-induced dissociation-MS/MS (FeCID-MS/MS) was performed on the glycan modifications using the Q-Tof Ultima. The RF lens 1 voltage was increased from 40 to 100 V so that the glycopeptides fragmented as they entered the mass spectrometer. Standard MS/MS was then performed on the glycan oxonium ions.

Nano-LC-electron-transfer dissociation (ETD)-MS/MS analysis was performed on a NanoAquity UPLC (Waters) coupled to an LTQ XL ion trap mass spectrometer (Thermo Fisher Scientific). The column and gradient conditions were the same as those described above. The reagent gas was fluoranthene, the ETD dwell time was 200 ms, and supplementary activation was enabled.

Electrospray MS analysis of recombinant S. oneidensis His₆SO3271 and His₆SO3270 was performed on a Waters Micromass Quattro II triple quadrupole MS.

Statistical analysis.

Values are presented as means ± SD (standard deviations). Student's t test was performed with statistical significance set at the 0.05 confidence level.

RESULTS

Mass spectrometry analysis of flagellin FlaB trypsin digests.

FlaB was chosen for mass spectrometry analysis because it is the functional flagellin. It has been shown that both the wild type and flagellin-free mutant (referred to as FFM [ΔflaA ΔflaB] here to simplify the description) strains had a hypermotile phenotype when they contained plasmid pHG101-flaB (4). This is due to overproduction of FlaB flagellins, as the plasmid is present in 5 to 7 copies per cell (4, 31). To prevent interference from FlaA, FlaB was isolated from the ΔflaA ΔflaB strain carrying pHG101-flaB (FFM/FlaB^WT) according to the method described in Materials and Methods, and purified flagellins were of high purity (>80%) as judged by SDS-PAGE (see Fig. S2 in the supplemental material). The purified flagellins were digested with trypsin, and the resulting digests were analyzed in duplicate by LC-MS.

LC-MS/MS data acquired on the flagellin tryptic digest resulted in an approximately 89% coverage of the expected FlaB amino acid sequence (Fig. 1), and the majority of the peptides identified were unmodified. However, groups of modified peptide ions separated from each other by 14 Da were observed throughout the LC-MS chromatogram. An example is provided in Fig. 2A. MS/MS analysis of the triply charged ion at m/z 851.3 (Fig. 2B) generated abundant b and y fragment ions that identified the T^94-113 flagellin peptide, DLTVQSENGANSSADLSALK. In addition, the low m/z region of the MS/MS spectrum was dominated by ions (underlined peaks in Fig. 2) stemming from a unique posttranslational modification (PTM) with a mass of 538.3 Da. MS/MS analysis of the triply charged ions at m/z 848.8 and m/z 844.1, the other modified peptides in this group, revealed that they were also derived from the same peptide sequence but were modified with a 524.3- and 510.3-Da moiety, respectively (Fig. 2C and D). Interestingly, the fragment ion at m/z 275.1 was observed in all three MS/MS spectra, suggesting that these PTMs consist of two residues: a constant one giving rise to the 275.1 peak, and a second residue which can vary in mass by multiples of 14 Da, possibly due to varying degrees of methylation. Further analysis of the 538.3-Da PTM by FeCID-MS/MS (Fig. 3A and B) supported the hypothesis that this modification is composed of two residues of roughly equal mass. Interestingly, many of the fragment ions in the MS/MS spectrum of the m/z 275.2 fragment ion (underlined peaks in Fig. 3A) were also observed in the MS/MS spectra of pseudaminic acid (23, 32) and other nonulosonic acids. These sialic acid-like sugars are known to decorate the flagellin of many Gram-negative and Gram-positive bacteria, so it is possible that the constant residue is a nonulosonic acid derivative. The T^94-113 flagellin peptide contains 5 potential sites for O-glycosylation, 4 serines and 1 threonine. ETD-MS/MS identified Ser¹⁰⁶ and Ser¹⁴³ as the glycosylation sites (Fig. 4A).

Fig 1.

FlaB amino acid sequence coverage obtained from the nano-LC-MS/MS analysis. The peptide (in black) and glycopeptide (underlined) coverage of the FlaB protein amino sequence is indicated. Peptides (in gray) were missed from the analysis. The glycosylation and methylated lysine residues (italicized) that have been confirmed by MS are shown here also. Approximately 89% of the FlaB amino acid sequence has been mapped in this MS study.

Fig 2.

Nano-LC-MS and -MS/MS of the modified FlaB tryptic glycopeptide, ⁹⁴DLTVQSENGANSSADLSALK¹¹³. (A) Nano-LC-MS mass spectrum showing a group of modified peptide ions separated by 14 Da, a pattern observed frequently throughout this analysis. Expanded views of the mass spectrum showing the doubly (separated by 7 m/z in the spectrum) and triply (separated by 4.7 m/z in the spectrum) protonated ions are presented in the insets. Rel. Int., relative intensity. (B) CID-MS/MS spectrum of the MH₃³⁺ ion at m/z 852.4. The b and y fragment ions identify the T^94-113 peptide, while the underlined ions in the low m/z region arise from a 538.3-Da modification. (C and D) CID-MS/MS of the other triply charged ions in this grouping (m/z 848.8 and m/z 844.1, respectively), indicating that they are composed of the same peptide sequence but modified with a 524.3- and 510.3-Da modification, respectively. The fragment ion at m/z 275.1 was observed in all three MS/MS spectra, indicating that a portion of these modifications is constant and common to all. Rel. Int., relative intensity.

Fig 3.

FeCID-MS/MS of unknown glycan modifications. FeCID/MS/MS spectra of the glycan fragment ions arising from the PTM, namely, the ion at m/z 521.3 corresponding to the dehydrated modification (A), as well as the constant fragment ion at m/z 275.2 (B). The lack of abundant ions in the upper half of the MS/MS spectrum in panel A indicates that the m/z 521.3 glycan is composed of two residues of roughly equal mass. Many of the fragment ions (underlined) observed in the MS/MS spectrum of the m/z 275.2 ion (B) are also found in the fragment ion spectra of pseudaminic acid (23, 32) and other nonulosonic acids, suggesting that this residue is a similar glycan.

Fig 4.

Glycopeptide detection using ETD-MS/MS and stepped-collision offset LC-MS. (A) ETD-MS/MS analysis of the ion at m/z 852.7, corresponding to the MH₃³⁺ precursor ion of the modified T^94-113 peptide. The fragmentation observed identified the linkage site as Ser¹⁰⁶. Therefore, even though this peptide contains a consensus sequon for N-glycosylation, it is actually O-glycosylated. (B) Glycopeptide detection using stepped-collision offset LC-MS. (a) Total ion LC-MS chromatogram acquired for the FlaB tryptic digest using low-collision offset (10 V). (b) Extracted ion chromatogram (EIC) acquired for the m/z 275.1 fragment ion using high-collision offset (35 V). This fragment ion is common to all glycan modifications. (c to e) EICs of the intact glycan fragment ions at m/z 539.3, 525.3, and 511.3, respectively. Glycopeptides are to be found where the peaks in the intact glycan fragment ion EICs align with those in the m/z 275.1 trace. Most glycopeptides elute between 26.5 and 32.0 min (hashed lines). TIC, total ion current.

A comprehensive search for glycopeptides in the flagellin tryptic digest was achieved using the stepped-collision offset method, described in Materials and Methods, to generate fragment ions unique to the glycan PTMs. The LC-MS total ion chromatogram (low-collision offset) for the flagellin tryptic digest as well as the extracted ion chromatograms for the key glycan oxonium ions (high-collision offset) are presented in Fig. 4B. A glycopeptide would be expected to produce both the m/z 275.1 ion as well as one of the intact glycan fragment ions (m/z 511.3, 525.3, or 539.3). Most glycopeptides were detected between 26.5 and 32 min, and nano-LC-MS/MS confirmed the identity of the majority of the glycopeptides. MS/MS analysis also revealed that some of the glycopeptides were further modified with methyl-lysines. Furthermore, a second search of the nano-LC-MS/MS data identified a number of methylated but nonglycosylated peptides. An example is provided in Fig. 5 for the tryptic peptide, T^147-163. The LC-MS survey spectrum (Fig. 5A) reveals a series of doubly and triply charged ions for peptide ions separated from one another by 14 Da. MS/MS analysis of the doubly charged ion at m/z 932.0 identified the T^147-163 peptide (Fig. 5B) and confirmed that residue 159 is a methyl-lysine [K(Me)¹⁵⁹]. The MS/MS spectrum of the MH₂²⁺ ion at m/z 939.0 confirmed it to be the same peptide with two methyl-lysine residues [K(Me)¹⁵⁹ and K(Me)¹⁶³] (Fig. 5C).

Fig 5.

Nano-LC-MS and -MS/MS of the FlaB tryptic peptide, T^147-163, containing methylated lysines. (A) Mass spectrum showing a series of doubly and triply protonated T^147-163 ions, each separated by 14 Da. (B) MS/MS analysis of the MH₂²⁺ ion at m/z 931.5 corresponding to methylated T^147-163. The peptide fragment ions confirm the peptide contains a single methylated lysine residue, K(Me)¹⁵⁹. (C) MS/MS analysis of the MH₂²⁺ ion at m/z 939.0, corresponding to T^147-163 containing two methylated lysine residues, K(Me)¹⁵⁹ and K(Me)¹⁶². These peptides are not glycosylated, although others are both glycosylated and methylated (data not shown).

All modified peptides, both O-glycosylated and methylated, observed in the FlaB flagellin tryptic digest are listed in Table 2. The glycopeptides as well as the sites of O-linkage and methylation are indicated in Fig. 1.

Table 2.

Modified peptides identified in the tryptic digest of FlaB from S. oneidensis

Peptide	Mass (Da)	Mass of modification (Da)a	Residues for modificationb
94-113	2,557.3	538.31	DLTVQSENGANSSADLSALK
	2,543.3	524.31
	2,529.3	510.31
137-146	1,457.8	538.31	LLAGGFSAGK
	1,443.8	524.31
	1,429.8	510.31
147–159	1,433.7	141	NFQVGAQDGEDIK(Me)
147–163	1,874.9	142	NFQVGAQDGEDIK(Me)VTVK(Me)
168–183	2,045.1	538.31	SSLSVGSLGNTTSAAR
	2,031.1	524.31
	2,017.1	510.31
	2,583.4	538.32
	2,569.4	538.31 + 524.31
	2,555.4	524.32 or (538.31 + 510.31)
	2,541.4	524.31 + 510.31
	2,527.4	510.32
164–183	2,997.6	538.32 + 141	ASNK(Me)SSLSVGSLGNTTSAAR
	2,983.6	538.31 + 524.32 + 141
	2,969.6	524.32 or (538.31 + 510.31) + 141
	2,955.6	524.31 + 510.31 + 141
	2,941.6	510.32 + 141
184–188	1,042.6	538.31	ASSLK
	1,028.6	524.31
184–189	1,170.1	538.31	ASSLKK
184–195	1,796.1	538.31 + 14e1	ASSLKK(Me)IDAAIK or ASSLK(Me)KIDAAIK
	1,782.1	524.31 + 141
	1,768.1	510.31 + 141
	1,810.1	538.31 + 142	ASSLK(Me)K(Me)IDAAIK
	1,796.1	524.31 + 142
	1,782.1	510.31 + 142

^aSubscript number represents the number of modification.

^bUnderlined residues are those glycosylated.

Mutational analysis of potential glycosylation residues by cysteine scanning.

The MS/MS analysis of S. oneidensis FlaB identifies two residues (Ser¹⁰⁶ and Ser¹⁴³) that are glycosylated. To evaluate the effect of glycosylation at Ser¹⁰⁶ and Ser¹⁴³ on motility and to screen for other glycosylation residues, we utilized the cysteine-scanning analysis protocol because cysteine is the most similar amino acid in structure to both serine and threonine. Assuming that the basic structure of S. oneidensis flagellin is similar to that of Salmonella, a fragment covering residues 95 to 205 is likely to contain all glycosylation sites because of its surface exposure in the assembled filament (see Fig. S1 in the supplemental material). In total, 22 serine and threonine residues were chosen for the site-directed mutagenesis analysis of FlaB with pHG101-flaB as the template. After sequencing verification, each resulting vector carrying mutated flaB was introduced into the FFM strain for the motility assay.

The motility data are provided in Fig. 6A. In total, substitutions at seven sites introduced a difference in motility that was statistically significant. These include 105^SàC, 106^SàC, 134^TàC, 135^TàC, 143^SàC, 171^SàC, and 180^SàC. The impacts of most mutations were rather modest, resulting in at least 70% motility remaining, suggesting that neither substitution by cysteine nor glycosylation at the majority of sites has profound influence on flagellar assembly and function. FFM/FlaB^S143C displayed a decrease in motility to approximately 37% relative to the FFM/FlaB^WT level (Fig. 6B). As serine 143 is a confirmed glycosylation site, these data indicate that glycosylation at this residue is critical for flagellar function. To examine whether the flagellar assembly is affected by this mutation, cells were subjected to flagellar staining and transmission electron microscopy (TEM) (see Fig. S3 in the supplemental material). As expected, none of the FFM cells were flagellated (Table 3). In contrast, 56% ± 9% of FFM cells possessed a flagellum when pHG101-flaB was present, comparable to that of the wild type/FlaB^WT (53% ± 7%). Similar results were obtained when we expressed in the FFM strain one of seven mutant flagellins that caused a difference in motility, including FlaB^S105C, FlaB^S106C, FlaB^T134C, FlaB^T135C, FlaB^S143C, FlaB^S171C, and FlaB^S180C. However, expression of FlaB^S143C in FFM substantially reduced the percentage of cells that were flagellated (31% ± 5%), suggesting that loss of glycan at this site hampers the assembly of flagellar filaments. Although it is impossible to accurately measure the length of S. oneidensis flagellar filaments, because they are extremely long and are of various lengths in any given sample, we managed to assess the swimming speed of individual cells to determine whether the flagellum per se is functionally impaired. In LB medium, cells of wild-type/FlaB^WT, FFM/FlaB^WT, and FFM/FlaB^S143C were monitored to swim at speeds of 53 ± 8, 52 ± 9, and 29 ± 7 µm per second, respectively. As the data of cell swimming speed carry rather large variations, FFM containing one of the other mutant FlaB proteins failed to show a difference that was statistically significant compared to FFM/FlaB^WT (Table 3). Nevertheless, reduction in motility caused by substitution at residue 143 indicates that glycosylation at this site is important for flagellar assembly.

Fig 6.

Characteristics of S. oneidensis cells expressing mutant flagellins. (A) Relative motility of S. oneidensis cells expressing each of the mutant flagellins. Relative motility was obtained as the ratio of the diameter of the area of motility from mutant flagellin to that from wild-type flagellin. (B) Mutant flagellins that introduced significant reduction in motility or that carried the mutation at the glycosylation sites. FFM with FlaB^WT (top) and FFM with each of the mutated flagellins (bottom) were compared to each other in a single assay. FFM carrying the empty vector pHG101 was used as the negative control (FFM/V^e). (C) Western blot analyses of flagellins carrying mutations at glycosylation sites. Flagellar filaments from FFM containing one of the mutant flagellins were purified and applied to SDS-PAGE for detecting the migration shift. Those migrating differently on SDS-PAGE were examined using antibodies against S. oneidensis flagellins and are presented here.

Table 3.

Characteristics of S. oneidensis cells expressing mutant flagellins^a

Strain	Motility (%)	Flagellated cells (%)	Swimming speed (μm s−1)
WT/FlaBWT	100	53 ± 7	53 ± 8
FFM (ΔflaA ΔflaB)	0	0	0
FFM/FlaBWT	100	56 ± 9	52 ± 9
FFM/FlaBS105C	77 ± 6	52 ± 9	49 ± 11
FFM/FlaBS106C	73 ± 4	53 ± 4	50 ± 13
FFM/FlaBT134C	82 ± 6	52 ± 9	47 ± 12
FFM/FlaBT135C	78 ± 4	49 ± 6	45 ± 13
FFM/FlaBS143C	37 ± 6	31 ± 5	29 ± 7
FFM/FlaBS171C	79 ± 7	53 ± 11	50 ± 10
FFM/FlaBS180C	84 ± 5	57 ± 3	51 ± 8
FFM/FlaBS185C	89 ± 9	55 ± 5	47 ± 9

^aFor swimming speed, a minimum of 10 cells were measured.

The cysteine-scanning analysis not only succeeded in confirming the important role of residue 143 for flagellar function but also revealed that glycosylation at the other putative glycosylation sites had a lower impact on motility. In an attempt to identify the other glycosylation sites, we performed SDS-PAGE and Western blotting of the mutant flagellins. Glycosylation can alter the migration of flagellins on SDS-PAGE; indeed, among these mutant flagellins, five moved faster than wild-type FlaB on SDS-PAGE (Fig. 6C). These included FlaB^S106C, FlaB^S143C, FlaB^S171C, FlaB^S180C, and FlaB^S185C. Interestingly, all five of these residues are serines. As cysteine differs from serine in a single atom, the replacement is unlikely to introduce significant changes in protein conformation other than annulling glycosylation. As a consequence, it is very likely that these five residues are glycosylated.

Methylation is not required for full motility and flagellar assembly.

Mass spectrometry analysis of S. oneidensis flagellin revealed that at least five lysine residues are methylated. To identify the gene encoding the corresponding methylase, we performed a BLAST search against the S. oneidensis proteome with S. enterica serovar Typhimurium lysine-N-methylase FliB (401 aa). The analysis returned only one hit with a sequence identity of 24%, SO4160. The protein is 392 aa in length and is not well conserved in Shewanella, as its encoding gene was revealed by BLAST in 13 (out of 25) sequenced strains, suggesting that methylation is not universal within the genus. In species/strains containing a homolog, the sequence identities to S. oneidensis SO4160 range from 65% (S. putrefaciens CN-32) to 41% (S. piezotolerans WP3).

To determine whether SO4160 is the lysine-N-methylase and whether this type of PTM has an impact on flagellar assembly in S. oneidensis, an SO4160 in-frame deletion strain was created and characterized. The deletion of SO4160 failed to introduce any noticeable difference in either motility of cells or migration of flagellins on SDS-PAGE compared to the wild type (data not shown). This is not surprising, because loss of fliB does not affect the motility of S. enterica serovar Typhimurium (9). Despite this, the SO4160 mutant produced flagellins whose lysine residues were not methylated (Fig. 7). These lysine residues became methylated in the SO4160 mutant when SO4160 was expressed in trans. However, the glycan modifications were the same as those of the wild type (data not shown). These data indicate that SO4160 is the enzyme that catalyzes lysine methylation but is not accountable for methylation of glycans. We therefore named the gene SO4160 as fliB.

Fig 7.

Nano-LC-MS and -MS/MS of the FlaB tryptic peptide, T^147-163, containing methylated lysines. (A) MS/MS analysis of the MH₂²⁺ ion at m/z 717.9 corresponding to T^147-159 containing the single methylated lysine residue K(Me)¹⁵⁹. (B) MS/MS analysis of the MH₃³⁺ ion at m/z 626.1 corresponding to T^147-163 containing two methylated lysine residues, K(Me)¹⁵⁹ and K(Me)¹⁶². These peptides are not methylated in the SO4160 mutant.

Glycosylation of S. oneidensis flagellins requires PseB and PseC.

Previously, we have shown that SO3271 is essential for motility of S. oneidensis, and flagellins in an SO3271 null mutant migrate faster than the wild-type ones (4). SO3271 shares a sequence identity of 66% with PseB of H. pylori, the first enzyme of the Pse glycosylation pathway converting UDP-α-d-GlcNAc to UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose (33). In addition, SO3270, predicted to be transcribed with SO3271 into a single cistron, encodes a protein sharing 36% sequence identity with H. pylori PseC, which converts the product of PseB to UDP-4-amino-4,6-dideoxy-β-l-AltNAc. To test whether SO3270 is involved in glycosylation of flagellins, we constructed an SO3270 in-frame deletion mutant. A swimming assay on 0.25% agar plates revealed that ΔSO3270 was nonmotile, and its motility was fully restored with a copy of SO3270 expressed in trans (Fig. 8). As glycosylation is required for flagellar filament assembly, ΔSO3270 cells grown to exponential phase were examined by TEM. As expected, the ΔSO3270 strain, like the ΔSO3271 strain, was aflagellated (Fig. 8).

Fig 8.

SO3270 is required for flagellar assembly in S. oneidensis. Motility and TEM of the indicated strains. ΔSO3270^C indicates that the mutant is complemented in trans.

Although sequence homology suggests that S. oneidensis employs SO3271 and SO3270 to catalyze the conversion of UDP-α-d-GlcNAc to UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose and then to UDP-4-amino-4,6-dideoxy-β-l-AltNAc, direct evidence is lacking. Therefore, we cloned SO3271 and SO3270 into pET28a and expressed and purified the N-terminal His-tagged products in E. coli (see Fig. S4A in the supplemental material). MS analysis of reaction mixtures containing UDP-α-d-GlcNAc and His₆SO3271 revealed a reaction product that is 18 Da smaller than UDP-α-d-GlcNAc, which was not observed in the protein-free control reaction. As PseB removes a molecule of H₂O from UDP-α-d-GlcNAc, the 18-Da loss in mass supports the hypothesis that SO3271 is an analogue of PseB (see Fig. S4B). We utilized triple quadrupole MS to assess the reaction mixture containing a variety of combinations of UDP-α-d-GlcNAc, pyridoxal-5′-phosphate (PLP), l-glutamate (Glu), His₆SO3271, and His₆SO3270. As shown in Fig. 9, we observed three negative ions at m/z 606, 588, and 589. The substrate that was a negative ion at m/z 606 corresponded to UDP-α-d-GlcNAc. From reaction A, it was clear that a novel negative ion at m/z 588 was observed when UDP-α-d-GlcNAc was mixed with His₆SO3271, losing a mass of 18 Da that was consistent with H₂O. This product did not appear in the absence of His₆SO3271 (reactions B and C), eliminating the possibility that SO3270 could act on UDP-α-d-GlcNAc directly. In the coupled assay (reaction D) containing His₆SO3271 and His₆SO3270 with l-glutamate as the amino donor and PLP as the transaminase cofactor, the product of His₆SO3271 (at m/z 588) was depleted, and the intensity of m/z 589 was enhanced significantly, indicating that chemicals at m/z 588 and m/z 589 were the substrate and product of His₆SO3270, respectively. The mass (m/z 589) of the His₆SO3270 product coincides with that of the transamination product of PseC, UDP-4-amino-4,6-dideoxy-GlcNAc. These results, together with the synteny feature and sequence conservation, strongly suggest that SO3271 and SO3270 are the analogues of H. pylori and C. jejuni PseB and PseC, respectively. We therefore name SO3271 and SO3270 as pseB and pseC, respectively, in S. oneidensis.

Fig 9.

MS analysis of reactions catalyzed by SO3271 and SO3270. The MS graphs are lined with the corresponding chemicals shown on the left (upper, UDP-GlcNac; middle, UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose; lower, UDP-4-amino-4,6-dideoxy-β-l-AltNAc). In reactions A and B, UDP-d-GlcNAc is possibly converted to UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose by SO3271; in reaction C, no conversion; in reaction D, UDP-d-GlcNAc is possibly converted to UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose by SO3271, followed by UDP-2-acetamido-2,6-dideoxy-β-l-arabino-4-hexulose possibly being converted to UDP-4-amino-4,6-dideoxy-β-l-AltNAc by SO3271. y axis represents relative intensity.

DISCUSSION

S. oneidensis produces a single polar flagellum composed of two flagellins, which have been implicated to be posttranslationally modified (4). In this article, we present evidence to suggest that the major flagellin FlaB contains five variable-mass O-linked PTMs, differing from one another by 14 Da, presumably due to varying degrees of methylation. Although the exact composition of these PTMs is unknown, each contains a constant 274-Da residue and a residue whose mass varies by 14 Da depending on the PTM. The constant residue bears a resemblance to nonulosonic acids, a family of negatively charged nine-carbon backbone α-keto sugars that include neuraminic, legionaminic, and pseudaminic acids (34). In addition, the flagellin is subjected to methylation at multiple lysine residues, which has no detectable impact on flagellar function, as in S. enterica serovar Typhimurium (9).

Full-size flagellins, as in S. enterica serovar Typhimurium and C. jejuni, are composed of multiple domains, D₀-D₁-D₂-D₃-D₂-D₁-D₀ (22–25). In the well-characterized C. jejuni flagellin, all glycosylated residues are located in two D₂ and D₃ domains that are thought to be surface exposed. Like P. syringae, S. oneidensis possesses rather small flagellins, replacing the amino acid sequence for the D₂ and D₃ domains of the S. enterica serovar Typhimurium and C. jejuni flagellins with a short variable fragment (residues 168 to 191) (22, 26). Given the surface exposure feature of this fragment, it conceivably contains most of the residues (Ser¹⁷¹, Ser¹⁸⁰, and Ser¹⁸⁵) predicted for glycosylation in addition to two (Ser¹⁰⁶ and Ser¹⁴³) in the distal region of N-terminal D₁. While glycosylation on S. oneidensis flagellin is O-linked as expected, all proposed residues for such PTM identified in this study are serines. This resonates with similar observations in other bacteria (24–26, 35), suggesting that O-linked PTM on flagellins prefers serine to threonine.

Studies on bacterial flagellin glycosylation have focused on two sets of mutants, defective either in the glycan biosynthesis pathway or flagellin glycosylation by replacing individual glycosylation sites. Except for Burkholderia pseudomallei, the former results in aflagellated cells, as in S. oneidensis of this work, indicating that glycosylation of flagellin in cytoplasm is necessary for assembly of the flagellar filament (10, 17, 20, 36). The latter usually have little impact on flagellar assembly but may cause motility defects or other defects, such as virulence and adherence without compromising motility (24, 26, 35, 37). A good example is P. syringae, in which loss of glycan at each of its six glycosylation sites introduces motility defects that are comparable to one another (35). In line with this, removal of PTM at any of the proposed glycosylation sites on the S. oneidensis flagellin also results in reduction in motility. However, two differences between S. oneidensis and P. syringae are apparent. First, glycosylation at residue 143 in S. oneidensis is particularly critical for flagellar function. Interestingly, Ser¹⁴³ is the only residue that is glycosylated in these two microorganisms; even their flagellins are similar in size (see Fig. S1 in the supplemental material). Second, the decrease in motility caused by the removal of PTM at the other putative glycosylation sites on the S. oneidensis flagellin is much less significant than that for P. syringae.

To date, only a few flagellin glycosylation pathways have been elucidated, including the Pse pathway of H. pylori and C. jejuni (33), the Leg pathway of C. jejuni (38), the Wbp pathway of Pseudomonas aeruginosa (39), and the Vio pathway of P. syringae (40). In the best-studied Gram-negative bacteria examples, C. jejuni and H. pylori, both of which possess multiple flagellin glycosylation pathways, monosaccharides found in the glycans include derivatives from pseudaminic acid and legionaminic acid (33, 36). In the Gram-positive bacteria Listeria monocytogenes and Clostridium botulinum, flagellins are glycosylated with O-linked N-acetylglucosamine and the sialic acid-like nonulosonate sugar at up to 6 or 7 sites (41, 42). Flagellins are also found to be modified with a disaccharide of 540 Da in P. syringae and oligosaccharide glycans in P. aeruginosa and Azospirillum brasilense (26, 41, 43).

In S. oneidensis, synthesis of the flagellin glycans starts with PseB and PseC, the first two enzymes of the Pse pathway (33). However, it appears that S. oneidensis does not utilize a complete Pse pathway for glycosylation of flagellins, because the genome does not encode homologues of any other components of the Pse pathway by the BLAST analysis. This idea is further supported by the finding that neither residue of PTMs on S. oneidensis flagellins is Pse (316 Da), although the modification could be a modified Pse. Nevertheless, flagellar glycosylation in S. oneidensis is likely to be carried out by a novel pathway with components which may have evolved independently as promiscuous enzymes that work in multiple pathways. Additional work is in progress to identify other genes in the pathway and to dissect residues of PTMs biochemically and structurally in S. oneidensis.