JOURNAL OF BACTERIOLOGY, June 2006, p. 4395–4403
Copyright © 2006, American Society for Microbiology. All Rights Reserved.
Vol. 188, No. 12
Glycosylation of b-Type Flagellin of Pseudomonas aeruginosa:
Structural and Genetic Basis
Amrisha Verma,1† Michael Schirm,2† Shiwani K. Arora,1P. Thibault,3Susan M. Logan,4
and Reuben Ramphal1*
Department of Medicine/Infectious Diseases, University of Florida, Gainesville, Florida 326101; Caprion Pharmaceuticals,
Montreal, Quebec, Canada2; Institute for Research in Immunology and Cancer, University of Montreal, Quebec,
Canada3; and NRC Institute for Biological Sciences, Ottawa, Ontario, Canada4
Received 26 October 2005/Accepted 24 March 2006
The flagellin of Pseudomonas aeruginosa can be classified into two major types—a-type or b-type—which can
be distinguished on the basis of molecular weight and reactivity with type-specific antisera. Flagellin from the
a-type strain PAK was shown to be glycosylated with a heterogeneous O-linked glycan attached to Thr189 and
Ser260. Here we show that b-type flagellin from strain PAO1 is also posttranslationally modified with an excess
mass of up to 700 Da, which cannot be explained through phosphorylation. Two serine residues at positions
191 and 195 were found to be modified. Each site had a deoxyhexose to which is linked a unique modification
of 209 Da containing a phosphate moiety. In comparison to strain PAK, which has an extensive flagellar
glycosylation island of 14 genes in its genome, the equivalent locus in PAO1 comprises of only four genes. PCR
analysis and sequence information suggested that there are few or no polymorphisms among the islands of the
b-type strains. Mutations were made in each of the genes, PA1088 to PA1091, and the flagellin from these
isogenic mutants was examined by mass spectrometry to determine whether they were involved in posttrans-
lational modification of the type-b flagellin. While mutation of PA1088, PA1089, and PA1090 genes altered the
composition of the flagellin glycan, only unmodified flagellin was produced by the PA1091 mutant strain. There
were no changes in motility or lipopolysaccharide banding in the mutants, implying a role that is limited to
Protein glycosylation has been long recognized as an impor-
tant posttranslational modification in eukaryotic cells. How-
ever, a number of recent studies have shown glycosylation of
cell surface-located and secreted proteins in prokaryotes.
Pathogenic bacteria such as Neisseria gonorrhoeae (33), Neis-
seria meningitidis (25), and one strain of Pseudomonas aerugi-
nosa (6) are now known to glycosylate pili. Similarly, Chla-
mydia (17) and Escherichia coli (21) glycosylate at least one of
their adhesins, and two species of Ehrlichia (23) have been
reported to glycosylate a surface-exposed immunodominant
protein. Moreover, flagellin, the major subunit of the flagel-
lum, is glycosylated in many bacteria, including P. aeruginosa
(2, 5), Campylobacter coli (22), Campylobacter jejuni (8), Trepo-
nema pallidum (43), Borrelia burgdorferi (10), Helicobacter felis
(12), Caulobacter crescentus (19), Agrobacterium tumefaciens
(7), and Listeria monocytogenes (31). In spite of this explosion
of information concerning the extent of bacterial protein gly-
cosylation, very limited information is available regarding the
actual glycan structures found on these glycoproteins. Re-
cently, the glycan structures on the flagellins of C. jejuni (38)
and H. pylori (29) and the a-type flagellin of P. aeruginosa have
been elucidated (30); the glycans on neisserial pili (25, 33) and
a specific Pseudomonas pilus (6) have also been characterized.
However, the genetic and biochemical basis of the glycosyla-
tion process is still poorly understood.
Some genetic systems responsible for glycosylation of se-
lected bacterial proteins have been described. Both N-linked
and O-linked glycosylation pathways have been described in C.
jejuni (35). Similarly, in N. meningitidis, a polymorphic locus
designated as the pgl locus required for the glycosylation of
class II pili was identified (13). A genomic island consisting of
14 open reading frames (ORFs) involved in the glycosylation
of flagellin in P. aeruginosa was also identified (2). In most
cases, the genes responsible for glycosylation are located in
close vicinity of the gene encoding the target protein. The
biological role(s) associated with bacterial protein glycosyla-
tion is still not precisely understood. However, glycosylation
defective mutants of several bacteria have recently been shown
to be attenuated in virulence attributes, such as adhesion and
invasion (34), colonization (14), and burn wound infection (4),
suggesting a role for protein-associated glycans in bacterial
P. aeruginosa is an opportunistic pathogen which causes se-
rious infections in immunocompromised human hosts. In ad-
dition to possessing a number of known virulence factors such
as exotoxins, proteases, lipases, lipopolysaccharide (LPS), and
pili, its flagellin protein, like others, has been shown to be a
potent stimulator of the inflammatory response via Toll-like
receptor 5 (40, 44) and has been suggested to be an inflam-
matory virulence factor (9). P. aeruginosa flagellins can be
classified into two groups (a and b types) based on their mo-
lecular weights and reactivity with specific sera (1, 18). The
a-type flagellins have more variable molecular masses (45 to 52
* Corresponding author. Mailing address: Department of Medicine/
Infectious Diseases, P.O. Box 100277, JHMHC, University of Florida,
Gainesville, FL 32610. Phone: (352) 392-2932. Fax: (352) 392-6481.
† A.V. and M.S. contributed equally to this study.
kDa) and are further classified into A1 and A2 subtypes based
on the differences in their amino acid sequences (3). The
b-type flagellins have a more conserved sequence and show an
invariant molecular mass of about 53 kDa (5). The discrepancy
between the predicted molecular mass of the a-type flagellin
(39 kDa) and the observed molecular mass was attributed to a
posttranslational modification (39), which was later shown to
be an addition of glycan chains (5, 30). Precisely, a serine
residue at position 189 and a threonine residue at position 260
were shown to be modified with a heterogeneous glycan com-
prising of up to 11 monosaccharide units that were O linked
through a rhamnose residue to the flagellin backbone. In ad-
dition, two genes, orfA and orfN, belonging to the glycosylation
island (GI) gene cluster considered to be involved in flagellin
glycosylation, were shown to be required for attachment of the
heterogeneous glycan and the proximal rhamnose residue, re-
The b-type flagellin was thought to be nonglycosylated since
any possible modification did not significantly alter the molec-
ular weight of the flagellin and therefore could not be detected
by sodium dodecyl sulfate-polyacrylamide gel electrophoresis
(SDS-PAGE). However, these flagellins were reported to pos-
sess phosphotyrosine (15). In addition, the glycosylation ma-
chinery of the a-type flagellin did not appear to be able to
modify b-type flagellin (2). Nevertheless, the identification of a
possible small GI near the flagellin gene on the PAO1 genome
suggested that the PAO1 flagellin might also be glycosylated.
The aim of the present study therefore was to determine
whether the b-type flagellins are glycosylated and to investigate
the genetic basis for this modification. We provide here evi-
dence that the b-type flagellin produced by P. aeruginosa strain
PAO1 and other similar strains is glycosylated and that four
ORFs constituting the GI are needed for this posttranslational
modification. Compared to the a-type flagellin, which is glyco-
sylated at threonine189 and serine 260, the b-type flagellin is
glycosylated at two serine residues that are located close to
each other (amino acids 191 and 195). Furthermore, the glycan
added to the b-type flagellin is only 709 Da. in molecular mass
and is less heterogeneous compared to that of a-type flagellin
of strain PAK.
MATERIALS AND METHODS
Bacterial strains and antibiotics used. The bacterial strains, plasmid con-
structs, and primers used in the present study are listed in Table 1. For Esche-
richia coli DH5? 100 ?g of ampicillin and 10 ?g of gentamicin/ml was used, and
for P. aeruginosa 50 ?g of gentamicin and 300 ?g of carbenicillin/ml was used.
Construction of mutants. Strain PAO1 is a commonly used strain whose
genome has been sequenced. It is known to carry b-type flagellin. The PA1091
TABLE 1. Bacterial strains and plasmids used in this study
Description, origin, and/or sequencea
Source or reference
F??80lacZ?M15 ?(lacZYA-argF)U169 endA1 recA1 hsdR17 (rK
gyrA96 relA1 phoA; RP4-2-Tc::Mu-Km::Tn7; TprSmrpro res?mod?
?) supE44 thi-1
Wild-type laboratory strain
Gentamicin insertion mutant in the rfbCgene of P. aeruginosa PAO1
Gentamicin insertion mutant in PA1088
In-frame deletion mutant of PA1089
In-frame deletion mutant of PA1090
Apr; oriT?sacB?, gene replacement vector with MCS from pUC18
pBluescript containing 2.2-kb PAO1 DNA with PA1088 and flanking regions
pBS206207 with gentamicin cassette inserted at the StuI site in PA1088
pEX18Gm containing ca. 2-kb PAO1 DNA flanking PA1089 from which aa 2 to 180
pEX18Gm containing ca. 2-kb PAO1 DNA flanking PA1089 from which aa 14 to 208
aTpr, trimethoprim resistance; Smr, streptomycin resistance; Apr, ampicillin resistance. Restriction sites are indicated in boldface. aa, amino acids.
4396VERMA ET AL.J. BACTERIOL.
mutant has been described before (4). A PA1088 mutant was constructed by
inserting a gentamicin resistance cassette into the PA1088 ORF at the unique
StuI site. The primer pair RER206 and RER207 (RER206/207) (Table 1) was
used to amplify a 2.2-kb DNA fragment from P. aeruginosa PAO1. This fragment
contained the complete PA1088 ORF and some flanking DNA. This fragment
was cloned into the vector pBluescript at HindIII and XhoI sites to yield
pBS206207. A gentamicin resistance cassette was then cloned into the unique
StuI site in PA1088 as a blunt-ended DNA fragment, leading to the construction
of pBS206207G. This construct was introduced into P. aeruginosa PAO1 by
electroporation to allow homologous recombination. Double crossovers were
selected that were resistant to 50 ?g of gentamicin/ml and sensitive to 300 ?g of
carbenicillin/ml. The insertion of the gentamicin cassette in the PA1088 ORF
was confirmed by PCR (data not shown).
The two in-frame deletion mutants in ORFs PA1089 and PA1090 were con-
structed by deleting amino acid residues 2 to 180 from PA1089 and amino acid
residues 14 to 208 from PA1090. Primer pairs RER274/275 and RER276/277
(Table 1) were used to amplify approximately 1 kb of DNA flanking the region
to be deleted from PA1089. Similarly, the primer pairs RER278/279 and
RER280/281 were used to amplify approximately 1 kb of DNA flanking the
region to be deleted from PA1090. The two 1-kb DNA pieces were spliced
together as HindIII/XbaI and XbaI/SstI fragments and were cloned as approx-
imately 2-kb HindIII/SstI inserts into the vector pEX18Gm (11), leading to the
construction of two plasmids, pEX?PA1089 and pEX?1090. These two plasmids
were introduced into P. aeruginosa PAO1 by electroporation, and single cross-
overs were selected on plates containing 50 ?g of gentamicin/ml. These clones
were tested for sucrose sensitivity, and sucrose-sensitive clones were resolved to
sucrose-resistant clones as a result of a second recombination event and excision
of the pEX18Gm plasmid backbone. These clones were confirmed by PCR,
which gave smaller PCR products from the PA1089 and PA1090 mutant strains
compared to their wild-type parent strain PAO1 (data not shown).
Flagellum purification. Flagella were purified from Pseudomonas strains
grown overnight in L broth. Flagella were sheared from the surface of the
bacteria and collected by ultracentrifugation as previously described (39). In
brief, bacteria were grown overnight in L broth, harvested by centrifugation, and
resuspended in cold phosphate-buffered saline containing 10 mM MgCl2.Fla-
gella were removed from the cells by shearing in a cold Waring blender for 20 s.
The cells were separated from the flagella by centrifugation at 12,000 ? g for 30
min. The supernatant was then filtered through a 0.45-?m-pore-size Millipore
filter and recentrifuged at 12,000 ? g for 30 min. The supernatant thus obtained
was then ultracentrifuged at 100, 000 ? g for 1 h. Flagella, obtained as a pellet,
were then suspended in minimum amount of phosphate-buffered saline contain-
ing 10 mM MgCl2.Purity was assessed by SDS-PAGE analysis. Contaminating
bands approximating the size of pilin (ca. 15.5 kDa) were occasionally seen and
were generally less than 5% of the flagellar preparation. No pilin fragments were
found on peptide analysis and when mass measurements were made on proteins
in the 49- to 50-kDa range.
Mass spectrometry (MS). For intact mass analysis, purified flagellins were
dialyzed in aqueous 0.2% (vol/vol) formic acid by using a Centricon YM-10
membrane filter (Millipore, Mississauga, Ontario, Canada). The solution was
infused into the mass spectrometer at a flow rate of 0.5 ?l/min. Sample digestion
were performed overnight in 50 mM NH4HCO3(5% acetonitrile) using se-
quence-grade trypsin (Promega, Madison, WI) at a enzyme/substrate ratio of
1:50. When required, a part of this tryptic digest was digested with chymotrypsin
(Promega) for 4 h in 100 mM ammonium bicarbonate. All digests were analyzed
with a Waters CapLC liquid chromatograph coupled to nanoelectrospray on a
Q-TOF Ultima instrument (Waters, Milford, Mass). Peptides were separated
and analyzed as described previously (30). The glycosylation site was identified
by using ?-elimination with ammonium hydroxide, which leaves a modified Ser
or Thr residue that could be located by using tandem MS (MS/MS) (26).
For the analysis of intact flagellin, dialyzed proteins were infused into a Waters
Q-TOF Ultima mass spectrometer at a flow rate of 0.5 ?l/min similar to that
described previously (30). MS/MS experiments were performed using argon as
the collision gas with collision energies ranging from 10 to 25 V. Second-gener-
ation fragment ion spectra were obtained by increasing the RF lens1 voltage
from 50 to 125 V, thereby forming fragment ions in the high-pressure region of
the skimmer/cone region of the mass spectrometer that were subsequently se-
lected as precursor ions for MS/MS analyses. MS/MS spectra were searched
against a nonredundant NCBI database using Mascot (Matrix Science, London,
United Kingdom), selecting all known bacteria. Parent ion and fragment ion
mass tolerances were both set at ?0.4 Da. External calibration was performed by
infusing 150 fmol/?l of solution of Glu-fibrinopeptide B (50% aqueous metha-
nol, 0.2% formic acid), providing a mass accuracy of ?50 ppm. Accurate mass
assignments in MS/MS mode were performed using fragment ions of known
composition to bracket the ion of interest and provided a mass accuracy within
10 ppm of the predicted values.
Comparison of the GI of P. aeruginosa strains PAO1 (b type)
and PAK (a type). The GI of P. aeruginosa strain PAK, which
possesses a-type flagellin, has been described previously (2)
and consists of 14 ORFs. This island is located between the
flgL and fliC genes of strain PAK. When the corresponding
chromosomal region of the b-type flagellin strain PAO1 was
analyzed, a much shorter putative GI consisting of only four
ORFs was found. A diagrammatic comparison of the GIs in P.
aeruginosa strains PAK and PAO1 is depicted in Fig. 1. The
homologies of the four ORFs comprising the PAO1 GI with
other proteins in the microbial genome databases are shown in
Table 2. Although we were not able to detect glycosylation of
PAO1 flagellin earlier by Western blot analyses, the new in-
formation about the PAO1 GI and the homologies of these
ORFs to a glycosyl transferase (PA1091) a phosphoserine
phosphatase (PA1089), and a nucleotidyltransferase (PA1090)
suggested that the PAO1 flagellin may also be glycosylated.
FIG. 1. Schematic diagram showing the structure of GIs of a- and b-type P. aeruginosa strains. The diagram shows the region of the P.
aeruginosa chromosome where the flagellar genes and the GIs are located. The a-type P. aeruginosa strain PAK has the GI located on an ?16-kb
fragment of DNA containing 14 ORFs, while the b-type strain PAO1 consists of 4 ORFs located on an ?8-kb piece of DNA.
VOL. 188, 2006 GLYCOSYLATION of b-TYPE FLAGELLIN IN P. AERUGINOSA 4397
It was reported earlier that the GIs in the a-type P. aerugi-
nosa strains are polymorphic, being either long or short (3). In
order to explore whether similar polymorphisms exist in the
b-type strains, 12 b-type strains were chosen from diverse lo-
cations—environmental, blood or urine, and cystic fibrosis
(CF)—and were analyzed by PCR. Two sets of primers,
RER243/258 and RER244/259, were used to amplify two over-
lapping pieces of DNA of 3.685 and 3.575 kb, respectively,
from the wild-type PAO1. All of the 12 strains analyzed yielded
PCR products with sizes identical to those for strain PAO1
(data not shown). The complete GI of one clinical strain, JG3,
was sequenced and was found to be identical to the wild PAO1
sequence, suggesting that the putative GI in the b-type P.
aeruginosa strains is highly conserved (data not shown). Fur-
thermore, the nucleotide sequence of the GI of another b-type
strain, PA14, whose genome has been recently sequenced
(www.tigr.org), was also found to be 99% identical to that of
PAO1, further confirming that the b-type strains have a highly
Intact mass analysis of P. aeruginosa PAO flagellin protein.
To ascertain whether the predicted mass of this flagellin was
equal to the measured mass, nanoelectrospray MS analysis of
purified flagellin from strain PAO was performed. This re-
vealed three well-defined peaks at 49,820 Da and two minor
peaks at 49,402 Da and 49,611 Da, corresponding to excess
masses of 707, 289, and 498 Da, respectively, from the pre-
dicted protein mass (Mtheo? 49,111 Da). The peak at 49,402
Da is consistent with the addition of two deoxyhexose residues
to the flagellin protein. This is similar to what was observed
previously for a-type flagellin protein from P. aeruginosa strain
JJ692 (30). The other two peaks (49,611 and 49,802 Da) cor-
respond to flagellin modified with two deoxyhexose residues to
which is added either one (49,611 Da) or two (49,820 Da)
residues of mass 209 Da (Fig. 2A).
Liquid chromatography-nanoelectrospray MS/MS analysis
of chymotryptic or tryptic digest of PAO flagellin protein. To
more precisely determine the type and location of glycosyla-
tion, the flagellin protein was digested with trypsin and ana-
lyzed by capillary liquid chromatography-nanoelectrospray
MS. A database search of the acquired MS/MS spectra was
performed and confirmed the identification of the flagellin
protein with 85% sequence coverage (Fig. 3). Only a few small
tryptic peptides of less than 800 Da and a large peptide in the
central region of the protein were not identified (T136-207,
theoretical mass of 6,809.3 Da). In general, peptides of this size
are difficult to separate by reversed-phase C18 chromatogra-
phy and additional enzymatic degradation using chymotrypsin
was required to obtain a proteolytic fragment of adequate size.
Analysis of the chymotryptic or tryptic digest identified a pep-
tide (sequence segment 186-207) modified with either a deoxy-
hexose as found in strain JJ692 or with an abundant oxonium
ion at m/z 356 as shown in Fig. 4a. The second-generation
fragment ion spectrum of the oxonium ion at m/z 356 is shown
in Fig. 4B and provides information on the composition of the
glycan moiety. Most notably, a second-generation fragment ion
of m/z 130 is obvious in the spectrum (Fig. 4b). The loss of 226
Da from the m/z 356 ion, which results in the appearance of
this m/z 130 fragment ion, likely corresponds to loss of a de-
oxyhexose plus a phosphate group from the glycan moiety. It is
noteworthy that no loss of H3PO4is observed, suggesting that
the phosphate group could be linked to the carbohydrate and
the unknown 129 Da residue via a phosphoester bond. An
exact mass measurement by top-down analysis (30) of the ox-
onium ion at m/z 356 was determined to be 356.112 ? 10 ppm.
The second-generation fragmentation pattern suggests that the
O-linked glycan comprises a deoxyhexose to which is attached
a phosphate group and an unknown residue of 129.085 Da.
The finding of a phosphate group confirms a previous report
that b-type flagellin was phosphorylated. However, in that
study the phosphate was suggested to be attached to a tyrosine
residue (15). The occurrence of an O-linked glycan containing
an amino acid is a potential assignment since glutamine and
TABLE 2. Similarities of P. aeruginosa PAO1 ORFs to proteins in the microbial genome databasea
PA ORF (identity) Protein or productLocalization
PA1088 (NP_249779) Hypothetical protein
Cytoplasmic Methylase involved in ubiquinone/
(coenzyme Q metabolism)
Hydrolase, haloacid dehalogenase-like
53% identical to SAM-dependent
56% similar to Phosphoserine
phosphatase (serB) (Methanococcus
jannaschii); amino acid transport and
49% similar to UDP-N-acetylglucosamine
PA1089 (NP_249780) Hypothetical protein
PA1090 (NP_249781) Hypothetical protein
Cytoplasmic Predicted sugar
envelope biogenesis, outer
Putative glycosyl/ glycerophosphate
transferases involved in teichoic
acid biosynthesis (cell envelope
biogenesis outer membrane)
PA1091 (NP_249782) Hypothetical protein
45% similar to O- antigen biosynthesis
protein RfbC (Myxococcus xanthus);
53% similar to C-terminal fragment of
RfbC (Riftia pachyptila endosymbiont);
42% similar to putative
glycosyltransferase in a locus involved
in synthesis of a carbohydrate antigen
aClass 4 genes encode hypothetical proteins. These genes are defined in the Pseudomonas genome database version 2 as homologs of previously reported genes of
unknown function or with no similarity to any previously reported sequences.
4398 VERMA ET AL.J. BACTERIOL.
glycine substituents were previously reported on pseudaminic
acid of the flagellar glycan of Campylobacter sp. (32) and on the
core oligosaccharide of H. influenzae LPS (20). In addition, the
N-linked flagellar glycan of the archaeon Methanococcus voltae
was recently shown to contain a threonine addition to the
glycan trisaccharide (41). However, structural elucidation by
NMR analysis of the purified glycan will be required for the
unequivocal assignment of this novel moiety.
Determination of glycan attachment site(s). To precisely
determine the site of attachment of this novel glycan, the
flagellin peptide 173-207 (for amino acids 173 to 207) was
subjected to ?-elimination using NH4OH. This analysis iden-
tified serine residues 191 and 195 as the sites of modification.
The MS/MS spectra of the ?-eliminated chymotryptic or tryp-
tic peptide 172-207 is shown in Fig. 5. The y fragments extend-
ing beyond y9all showed an increase of ?1 Da resulting from
the deamidation of asparagine to aspartic acid. The y14and y17
ions showed, respectively, a mass shift of ?1 Da from the
predicted mass, indicating that serines 191 and 195 are modi-
fied with O-linked glycans.
Functional characterization of PA1088, PA1089, PA1090,
and PA1091 from the GI of PAO1. In contrast to strain PAK,
which has an extended GI containing 14 genes, the equivalent
region in PAO displays four ORFs: PA1088, PA1089, PA1090,
FIG. 2. Intact mass analysis of P. aeruginosa flagellin. Reconstructed molecular mass profiles of flagellin from PAO1 (A), PAO1 rfbC (B),
PA1088 (C), and PA1089 (D).
FIG. 3. Assignment map of PAO flagellin. The sites of O-linked glycosylation are enclosed in boxes, and assigned peptides are in boldface,
whereas peptides that have not been assigned are in normal typeface.
VOL. 188, 2006 GLYCOSYLATION of b-TYPE FLAGELLIN IN P. AERUGINOSA 4399
and PA1091. Purified flagellin from PA1088, PA1089, PA1090,
and PA1091 mutant strains were analyzed by infusion MS.
Intact mass analysis of flagellin from PA1091 revealed a single
peak (49,109 Da) corresponding to flagellin that is nonglyco-
sylated (Mrpredicted 49,111 Da; Fig. 2B). As we had shown
with PAK flagellin (30), inactivation of this ORF in PAO also
results in an inability to attach the glycan/deoxyhexose moiety
at the site of glycosylation, and the flagellin protein remains
unmodified. Flagellin from PA1088 displayed a mass of 49,400
Da corresponding to flagellin protein modified with only two
deoxyhexose residues, indicating a role for the PA1088 gene
product either in the biosynthetic pathway or in the subsequent
addition of the novel 209 Da moiety to each O-linked deoxy-
hexose monosaccharide (Fig. 2C). In contrast, analysis of
flagellin from PA1089 revealed a more heterogeneous pattern
with O-linked glycans bearing either a single deoxyhexose res-
idue (49,400 Da) or the 356 Da residue at the two glycosylation
sites (49,819 Da). Compared to native PAO flagellin, where a
single major glycosylated form of mass 49,820 Da was obvious,
the flagellin from PA1089 displayed a much greater heteroge-
neity in size distribution (Fig. 2D), which is indicative of an
inability to efficiently glycosylate the flagellin protein. We were
unsuccessful in obtaining an intact mass spectrum for flagellin
purified from PA1090 (data not shown). Although we are cur-
rently unable to explain this result, the failure of this flagellin
protein to efficiently ionize under the conditions used indicates
a potential change in glycan composition.
The four mutants were examined for phenotypic changes
such as motility and LPS banding pattern, which are known to
occur when mutations are made in the glycosylation genes of
other organisms. None of the mutants were affected in these
phenotypes, similar to what was seen previously with mutations
in the GI genes of strain PAK.
Complementation of PA1091 mutant of PAO1 with the ho-
mologous PAK orfN gene. Structural analysis of flagellin from
PAO1 had revealed the presence of a deoxyhexose monosac-
charide as a component of the 335-Da glycan moiety. The orfN
gene from PAK is 41% identical to PA1091 and has been
shown to be responsible for the addition of rhamnose to the
PAK flagellin protein. To determine whether the product of
PA1091 is indeed a functional homolog, a plasmid containing
the orfN gene from PAK was inserted into the PA1091 mutant
and flagellin from the resulting transconjugant was examined
by top-down MS. Intact mass analysis revealed a single species
of flagellin of the predicted mass lacking glycosylation and no
evidence for restoration of the wild-type glycosylation profile
(data not shown). In contrast, complementation of PAK orfN
mutant restored a wild-type glycosylation profile. These pre-
liminary data are suggestive of a unique specificity for the
enzymatic products of the PA1091 gene and orfN from the
PAK glycosylation locus. As a consequence of this observation,
in addition to the structural analysis of flagellin from a PA1091
mutant which revealed that glycosylation had been abolished,
we now propose that the gene be renamed to fgtA (for flagellar
Flagellar glycan structures of P. aeruginosa. It appears that
the flagellar glycan produced by PAO is unique in structure
and quite distinct from the glycan described on PAK flagellin
(Fig. 6). As with the flagellar glycan of PAK, the linking sugar
appears to be a deoxyhexose that appears to be transferred to
the peptide backbone by the glycosyltransferase encoded by
PA1091 (fgtA). In contrast to PAK, the remaining glycan struc-
ture appears to contain a phosphate group to which is attached
in phosphoester linkage a terminal modification of 192 Da.
Mutational analysis of the first gene in the PAO GI PA1088,
which shows homology to a phosphoethanolamine N-methyl
transferase, is responsible for the addition of this 209-Da mod-
ification to the deoxyhexose monosaccharide. The remaining
two genes in the GI PA1089 and PA1090 clearly have an effect
on the ability to produce the wild-type structure, although we
have been unable to determine the precise role each plays in
the biosynthetic pathway.
This study characterizes the unique posttranslational modi-
fications found on the b-type PAO1 flagellin. Prior to this
work, it had been suggested based on migration in SDS-PAGE
and glycan staining that type b strains did not produce glyco-
sylated flagellin (2). However, the presence of four ORFs in
the corresponding region, two of which had homology to genes
involved in the synthesis of glycan chains, suggested that these
may be the functional equivalents of a GI. Intact mass analysis
of the PAO1 b-type flagellin revealed a significant increase in
mass of the protein monomer of ca. 1.4% from the predicted.
FIG. 4. Characterization of PAO glycopeptide 186-207 from chy-
motryptic digestion. (a) MS/MS spectrum of m/z 1,328.5 corresponding
to the doubly protonated peptide ion GTATASGIASGTVNLVG
GQVK modified with two neutral 355-Da residues. In the low mass
region an oxonium ion at m/z 356 from the O-linked glycan is observed.
(b) MS/MS spectrum of oxonium ion m/z 356 formed from collisional
activation in the ion source.
4400 VERMA ET AL.J. BACTERIOL.
The reconstructed mass profile of flagellin from PAO1 indi-
cated that the majority of flagellin was glycosylated with an
additional mass of 709 Da, while minor amounts of alternate
glycoforms of additional masses of 291 and 500 Da were also
apparent. In contrast to PAK a-type flagellin, PAO1 b-type
flagellin produces a shorter, less heterogeneous glycan moiety.
Consistent with this finding, while the a-type strains show sig-
nificant heterogeneity of their GIs (3), the b-type strains show
marked conservation of their genetic content.
While the glycan characterized on PAK flagellin was a het-
erogeneous oligosaccharide comprising up to 11 monosaccha-
ride residues, the glycan present on PAO1 flagellin is of mass
356.112 Da and appears to be composed of a single deoxyhex-
ose monosaccharide linked to a unique modification of mass
209 Da which contains a phosphate moiety. Although the
structural configuration of this novel substituent is currently
unknown, accurate mass analysis and fragmentation pattern
have provided preliminary information on the composition. As
with the a-type flagellin, it appears that the glycan of b-type
flagellin is linked to the protein backbone through a deoxyhex-
ose moiety. In contrast to the remaining structure, this deoxy-
hexose linkage appears to be a common feature of flagellar
glycosylation in Pseudomonas. The identification of nonglyco-
sylated flagellin from the PA1091 mutant was a result identical
to that observed in previous studies of the orfN mutant of PAK.
FIG. 6. Comparison of P. aeruginosa flagellar glycan structure.
(A) PAK flagellin; (B) PAO flagellin. Rha, rhamnose; Pen, pentose;
Hex, hexose; dHex, deoxyhexose; HexA, hexuronic acid; dHexN, de-
oxyhexosamine, PO4, phosphate.
FIG. 5. Determination of glycosylation sites on peptide 173-207. MS/MS spectrum of m/z 1,048.3 after ?-elimination. The MS/MS spectrum of
the peptide QVGSNGAGTVASVAGTATASGIASGTVNLVGGGQVK after ?-elimination is shown. Fragment ions showing a downward shift
of either 1 or 2 m/z units (indicated by an asterisk) enabled the identification of the linkage site of the glycan. Note that the asparagine is
deamidated to the corresponding isoaspartate residue (y9? 1).
VOL. 188, 2006 GLYCOSYLATION of b-TYPE FLAGELLIN IN P. AERUGINOSA4401
However, the two genes do not appear to be functional ho-
mologs, and this indicates either specificity for a unique de-
oxyhexose monosaccharide or, alternatively, a unique substrate
specificity in terms of the respective flagellin monomers.
Mapping of the glycosylation sites revealed a different pat-
tern from that found on PAK flagellin, where the two sites on
the primary sequence were a considerable distance apart (T189
and S260). In contrast, the two sites on PAO flagellin are
situated in close proximity to each other on the primary se-
quence (S191 and S195). Unlike N-linked glycosylation of pro-
teins, where a defined consensus sequence has been identified
in both eukaryotes (16) and prokaryotes (24, 42), the basis of
site specificity for O linkage of glycans to proteins is much
more poorly defined, especially with respect to prokaryotic
glycoproteins. The site of glycosylation may be related to local
hydrophobicity, whereby particular Ser/Thr residues within this
local hydrophobic environment project outward and so are
accessible to glycosyltransferases. It seems feasible that the
location of these sites in the respective PAK and PAO1 folded
flagellin protein or chaperone/flagellin complex may indeed be
in a similar hydrophobic pocket even though they do not align
on the primary sequences. Nonetheless, the localization of
glycosylation sites in both PAK and PAO to a central region of
the flagellin monomer most likely results in an exposed surface
location on the assembled filament rather than buried within
the central core. The crystal structure of the flagellin protein
from Salmonella enterica serovar Typhimurium had indicated
that the less well conserved central primary sequence of flagel-
lin comprises the D2 and D3 domains, which are clearly sur-
face exposed in the flagellar filament (28).
Other species of pseudomonads have also been shown to
possess similar genomic GIs as part of their flagellar regulons.
For example, the flagellar regulon of Pseudomonas syringae
carries a small island in a similar location and was also shown
to possess glycosylated flagellin (36). Glycosylation of flagellin
in this plant pathogen appears to be involved in specific host
cell recognition of distinct pathovars (37). Currently, the struc-
ture of the glycan is unknown but may be related to that found
on a-type flagellin in the present study considering the simi-
larity in GI composition.
Although the function(s) of the modification on P. aerugi-
nosa flagellin is unclear at present, there is mounting evidence
of a biological role in virulence. Glycosylation has been shown
to play a role in virulence in the burn mouse model of P.
aeruginosa infection (4) and is also involved in triggering an
inflammatory response (40). Future studies will be directed
toward determining the precise structure and biosynthetic
pathway of the novel glycans synthesized by both a-type and
b-type flagellin-bearing strains, the immunological response to
these glycans, and the characterization of any precise interac-
tions with host cell molecules during infection.
This study was supported by NIH grant AI 47852 (R.R.) and by the
NRC Genomics and Health Initiative of Canada (S.M.L.).
We thank T. Devesceri for assistance with figure preparation.
1. Allison, J. S., M. Dawson, D. Drake, and T. C. Montie. 1985. Electrophoretic
separation and molecular weight characterization of Pseudomonas aerugi-
nosa H-antigen flagellins. Infect. Immun. 49:770–774.
2. Arora, S., M. Bangera, S. Lory, and R. Ramphal. 2001. A genomic island in
Pseudomonas aeruginosa carries the determinants of flagellin glycosylation.
Proc. Natl. Acad. Sci. USA 98:9342–9347.
3. Arora, S. K., M. C. Wolfgang, S. Lory, and R. Ramphal. 2004. Sequence
polymorphism in the glycosylation island and flagellins of Pseudomonas
aeruginosa. J. Bacteriol. 186:2115–2122.
4. Arora, S. K., A. N. Neely, B. Blair, S. Lory, and R. Ramphal. 2005. Role of
motility and flagellin glycosylation in the pathogenesis of Pseudomonas
aeruginosa burn wound infections. Infect. Immun. 73:4395–4398.
5. Brimer, C. D., and T. C. Montie. 1998. Cloning and comparison of fliC genes
and identification of glycosylation in the flagellin of Pseudomonas aeruginosa
a-type strains. J. Bacteriol. 180:3209–3217.
6. Castric, P., F. J. Cassels, and R. W. Carlson. 2001. Structural characteriza-
tion of the Pseudomonas aeruginosa 1244 pilin glycan. J. Biol. Chem. 276:
7. Deakin, W. J., V. E. Parker, E. L. Wright, K. J. Ashcroft, G. J. Loake, and
C. H. Shaw. 1999. Agrobacterium tumefaciens possesses a fourth flagellin
gene located in a large cluster concerned with flagellar structure, assembly,
and motility. Microbiology 145:1397–1407.
8. Doig, P., N. Kinsella, P. Guerry, and T. J. Trust. 1996. Characterization of
a posttranslational modification of Campylobacter flagellin: identification of
a sero-specific glycosyl moiety. Mol. Microbiol. 19:379–387.
9. Feldman, M., R. Bryan, S. Rajan, L. Scheffler, S. Brunnert, H. Tang, and A.
Prince. 1998. Role of flagella in pathogenesis of Pseudomonas aeruginosa
pulmonary infection. Infect. Immun. 66:43–51.
10. Ge, Y., C. Li, L. Corum, C. A. Slaughter, and N. W. Charon. 1998. Structure
and expression of the FlaA periplasmic flagellar protein of Borrelia burgdor-
feri. J. Bacteriol. 180:2418–2425.
11. Hoang, T. T., R. R. Karkhoff-Schweizer, A. J. Kutchma, and H. P. Schweizer.
1998. A broad-host-range Flp-FRT recombination system for site-specific
excision of chromosomally-located DNA sequences: application for isolation
of unmarked Pseudomonas aeruginosa mutants. Gene 212:77–86.
12. Josenhans, C., R. L. Ferrero, A. Labigne, and S. Suerbaum. 1999. Cloning
and allelic exchange mutagenesis of two flagellin genes of Helicobacter felis.
Mol. Microbiol. 33:350–362.
13. Kahler, C. M., L. E. Martin, Y.-L. Tzeng, Y. K. Miller, K. Sharkey, D. S.
Stephens, and J. K. Davies. 2001. Polymorphism in pilin glycosylation locus
of Neisseria meningitidis expressing class II pili. Infect. Immun. 69:3597–3604.
14. Karlyshev, A. V., P. Everest, D. Linton, S. Cawthraw, D. G. Newell, and B. W.
Wren. 2004. The Campylobacter jejuni general glycosylation system is impor-
tant for attachment to human epithelial cells and in the colonization of
chicks. Microbiology 150:1957–1966.
15. Kelly-Wintenberg, K., S. L. South, and T. C. Montie. 1993. Tyrosine phos-
phate in a- and b-type flagellins of Pseudomonas aeruginosa. J. Bacteriol.
16. Kornfeld, R., and S. Kornfeld. 1985. Assembly of asparagine-linked oligo-
saccharides. Annu. Rev. Biochem. 54:631–664.
17. Kuo, C., N. Takahashi, A. F. Swanson, Y. Ozeki, and S. Hakomori. 1996. An
N-linked high-mannose type oligosaccharide, expressed at the major outer
membrane protein of Chlamydia trachomatis, mediates attachment and in-
fectivity of the microorganism to HeLa cells. J. Clin. Investig. 98:2813–2818.
18. Lanyi, B. 1970. Serological properties of Pseudomonas aeruginosa. II. Type-
specific thermolabile (flagellar) antigens. Acta Microbiol. Acad. Sci. Hung.
19. LeClerc, G., S. P. Wang, and B. Ely. 1998. A new class of Caulobacter
crescentus flagellar genes. J. Bacteriol. 180:5010–5019.
20. Li, J., S. H. Bauer, M. Mansson, E. R. Moxon, J. C. Richards, and E. K.
Schweda. 2001. Glycine is a common substituent of the inner core in Hae-
mophilus influenzae lipopolysaccharide. Glycobiology 11:1009–1015.
21. Lindenthal, C., and E. A. Elsinghorst. 1999. Identification of a glycoprotein
produced by enterotoxigenic Escherichia coli. Infect. Immun. 67:4084–4091.
22. Logan, S. M., J. F. Kelly, P. Thibault, C. P. Ewing, and P. Guerry. 2002.
Heterogeneity of carbohydrate modifications affects serospecificity of
Campylobacter flagellins. Mol. Microbiol. 46:587–597.
23. McBride, J. W., X. J. Yu, and D. H. Walker. 2000. Glycosylation of homol-
ogous immunodominant proteins of Ehrlichia chaffeensis and Ehrlichia canis.
Infect. Immun. 68:13–18.
24. Nita-Lazar, M., M. Wacker, B. Schegg, S. Amber, and M. Aebi. 2005. The
N-X-S/T consensus sequence is required but not sufficient for bacterial
N-linked protein glycosylation. Glycobiology 15:361–367.
25. Parge, H. E., K. T. Forest, M. J. Hickey, D. A. Christensen, E. D. Getzoff, and
J. A. Tainer. 1995. Structure of the fibre-forming protein pilin at 2.6 Å
resolution. Nature 378:32–38.
26. Rademaker, G. J., S. A. Pergantis, L. Blok-Tip, J. I. Langridge, A. Kleen, and
J. Thomas-Oates. 1998. Mass spectrometric determination of the sites of
O-glycan attachment with low picomolar sensitivity. Anal. Biochem. 257:
27. Reference deleted.
28. Samatey, F. A., K. Imada, F. Vonderviszt, Y. Shirakihara, and K. Namba.
2000. Crystallization of the F41 fragment of flagellin and data collection from
extremely thin crystals. J. Struct. Biol. 132:106–111.
29. Schirm, M., E. C. Soo, A. J. Aubry, J. Austin, P. Thibault, and S. M. Logan.
4402 VERMA ET AL.J. BACTERIOL.
2003. Structural, genetic and functional characterization of the flagellin gly- Download full-text
cosylation process in Helicobacter pylori. Mol. Microbiol. 48:1579–1592.
30. Schirm, M., S. K. Arora, A. Verma, E. Vinogradov, P. Thibault, R. Ramphal,
and S. M. Logan. 2004. Structural and genetic characterization of glyco-
sylation of type a flagellin in Pseudomonas aeruginosa. J. Bacteriol. 186:2523–
31. Schirm, M., M. Kalmokoff, A. Aubry, P. Thibault, M. Sandoz, and S. M.
Logan. 2004. Flagellin from Listeria monocytogenes is glycosylated with ?-O-
linked N-acetylglucosamine. J. Bacteriol. 186:6721–6727.
32. Schirm, M., I. Schoenhofen, S. M. Logan, K. C. Waldron, and P. Thibault.
2005. Identification of unusual bacterial glycosylation by tandem mass spec-
trometry analyses of intact proteins. Anal. Chem., 77:7774–7782.
33. Stimson, E., M. Virji, K. Makepeace, A. Dell, H. R. Morris, G. Payne, J. R.
Saunders, et al. 1995. Meningococcal pilin: a glycoprotein substituted with
digalactosyl 2,4-diacetamido-2,4,6-trideoxyhexose. Mol. Microbiol. 17:1201–
34. Szymanski, C. M., D. H. Burr, and P. Guerry. 2002. Campylobacter protein
glycosylation affects host cell interactions. Infect. Immun. 70:2242–2244.
35. Szymanski, C. M., S. M. Logan, S. Linton, and B. W. Wren. 2003. Campy-
lobacter: a tale of two protein glycosylation systems. Trends Microbiol. 11:
36. Taguchi, F., R. Shimizu, Y. Inagaki, K. Toyoda, T. Shiraishi, and Y. Ichinose.
2003. Post-translational modification of flagellin determines the specificity of
HR induction. Plant Cell Physiol. 44:342–349.
37. Takeuchi, K., F. Taguchi, Y. Inagaki, K. Toyoda, T. Shiraishi, and Y. Ichinose.
2003. Flagellin glycosylation island in Pseudomonas syringae pv. glycinea and
its role in host specificity. J. Bacteriol. 185:6658–6665.
38. Thibault, P., S. M. Logan, J. F. Kelly, J. R. Brisson, C. P. Ewing, T. J. Trust,
and P. Guerry. 2001. Identification and characterization of the carbohydrate
moieties and glycosylation motifs in Campylobacter flagellin. J. Biol. Chem.
39. Totten, P. A., and S. Lory. 1990. Characterization of the a-type flagellin gene
from Pseudomonas aeruginosa PAK. J. Bacteriol. 172:7188–7199.
40. Verma, A., S. K. Arora, S. K. Kuravi, and R. Ramphal. 2005. Roles of
specific amino acids in the N terminus of Pseudomonas aeruginosa flagellin
and of flagellin glycosylation in the innate immune response. Infect. Immun.
41. Voisin, S., R. S. Houliston, J. Kelly, J. R. Brisson, D. Watson, S. L. Bardy,
K. F. Jarrell, and S. M. Logan. 2005. Identification and characterization of
the unique N-linked glycan common to the flagellins and S-layer glycopro-
tein of Methanococcus voltae. J. Biol. Chem. 280:16586–16593.
42. Wacker, M., D. Linton, P. G. Hitchen, M. Nita-Lazar, S. M. Haslam, S. J.
North, M. Panico, H. R. Morris, A. Dell, B. W. Wren, and M. Aebi. 2002.
N-linked glycosylation in Campylobacter jejuni and its functional transfer into
Escherichia coli. Science 298:1790–1793.
43. Wyss, C. 1998. Flagellins, but not endoflagellar sheath proteins, of Trepo-
nema pallidum and of pathogen-related oral spirochetes are glycosylated.
Infect. Immun. 66:5751–5754.
44. Zhang, J., K. Xu, B. Ambati, and F. S. Yu. 2003. Toll-like receptor 5-medi-
ated corneal epithelial inflammatory responses to Pseudomonas aeruginosa
flagellin. Investig. Ophthalmol. Vis. Sci. 44:4247–4254.
VOL. 188, 2006GLYCOSYLATION of b-TYPE FLAGELLIN IN P. AERUGINOSA4403