Structure of the bacterial flagellar protofilament and implications for a switch for supercoiling

Article (PDF Available)inNature 410(6826):331-337 · March 2001with48 Reads
DOI: 10.1038/35066504
The bacterial flagellar filament is a helical propeller constructed from 11 protofilaments of a single protein, flagellin. The filament switches between left- and right-handed supercoiled forms when bacteria switch their swimming mode between running and tumbling. Supercoiling is produced by two different packing interactions of flagellin called L and R. In switching from L to R, the intersubunit distance (52 Å) along the protofilament decreases by 0.8 Å. Changes in the number of L and R protofilaments govern supercoiling of the filament. Here we report the 2.0 Å resolution crystal structure of a Salmonella flagellin fragment of relative molecular mass 41,300. The crystal contains pairs of antiparallel straight protofilaments with the R-type repeat. By simulated extension of the protofilament model, we have identified possible switch regions responsible for the bi-stable mechanical switch that generates the 0.8 Å difference in repeat distance.
VOL 410
15 MARCH 2001
| 331
Structure of the bacterial ¯agellar
proto®lament and implications
for a switch for supercoiling
Fadel A. Samatey*, Katsumi Imada*, Shigehiro Nagashima*, Ferenc Vonderviszt
, Takashi Kumasaka
, Masaki Yamamoto
& Keiichi Namba
* Protonic NanoMachine Project, ERATO, JST, 3-4 Hikaridai, Seika, Kyoto 619-0237, Japan
Department of Physics, University of Veszpre
m, Egyetem, u.10 H-8201, Hungary
RIKEN Harima Institute, 1-1-1 Kouto, Mikazuki, Hyogo 679-5198, Japan
§ Advanced Technology Research Laboratories, Matsushita Electric Industrial Co., Ltd, 3-4 Hikaridai, Seika, Kyoto 619-0237, Japan
The bacterial ¯agellar ®lament is a helical propeller constructed from 11 proto®laments of a single protein, ¯agellin. The ®lament
switches between left- and right-handed supercoiled forms when bacteria switch their swimming mode between running and
tumbling. Supercoiling is produced by two different packing interactions of ¯agellin called L and R. In switching from L to R, the
intersubunit distance (,52 A
) along the proto®lament decreases by 0.8 A
. Changes in the number of L and R proto®laments govern
supercoiling of the ®lament. Here we report the 2.0 A
resolution crystal structure of a Salmonella ¯agellin fragment of relative
molecular mass 41,300. The crystal contains pairs of antiparallel straight proto®laments with the R-type repeat. By simulated
extension of the proto®lament model, we have identi®ed possible switch regions responsible for the bi-stable mechanical switch
that generates the 0.8 A
difference in repeat distance.
Bacteria swim by rotating helical ¯agellar ®laments, which are up to
15 mm long, but only 120±250 A
in diameter. The rotary motor at
the base of the ®lament drives the rotation of this helical propeller
at hundreds of revolutions per second
. For chemotaxis and
thermotaxis, the swimming pattern of bacteria such as Salmonella
and Escherichia coli alternates between `run' and `tumble'; a run lasts
for a few seconds and a tumble for a fraction of second. During a
run, the motor rotates counterclockwise (as it is viewed from
outside the cell), and several ¯agellar ®laments with a left-handed
helical shape form a bundle and propel the cell. A tumble is caused
by quick reversal of the motor to clockwise rotation
, which
produces a twisting force that transforms the left-handed helical
form of the ®lament into a right-handed one
, causing the bundle
to fall apart rapidly. The separated ®laments act in an uncoordi-
nated way to generate forces that change the orientation of the
cell. Thus, the structure of the ¯agellar ®lament and its dynamic
properties have an essential role in bacterial taxis.
The ®lament is a helical assembly of a single protein, ¯agellin,
with roughly 11 subunits per 2 turns of the 1-start helix; the ®lament
can also be described as a tubular structure comprising 11 strands of
proto®laments, which are nearly longitudinal helical arrays of
. The helical forms of the ®lament are caused by super-
coiling, which is proposed to occur through a switching of con-
formation or packing interactions of the subunits between two
distinct states and their non- and quasi-equivalent intersubunit
. In the proposed models, each proto®lament exists
in one of two slightly different conformations, which affect the
repeat distance and lateral packing interaction; the regulated
switching of the 11 proto®laments would then produce ten types
of supercoil and two types of straight ®lament.
Many of these forms have actually been observed in various
strains and under various conditions
. The two straight ®laments
are called L-type and R-type according to the left- and right-handed
twist of the proto®laments, respectively
. Accurate measurements
by X-ray ®bre diffraction show that the repeat distances of the L-
and R-type proto®laments are 52.7 A
and 51.9 A
, respectively, the
difference being only 0.8 A
. A simple mechanical simulation of
supercoiled forms with these structural parameters shows a good
agreement with the pitch and diameter of observed supercoils,
indicating that the data and the model are both physically
. This also indicates that the level of precision at which
¯agellin works as a lengthwise mechanical switch is on the order of a
tenth of an a
Electron cryomicroscopy and X-ray ®bre diffraction have
revealed the domain organization of ¯agellin and subunit packing
in the two types of straight ®laments at about 10 A
however, higher resolution is needed to understand the structural
basis of supercoiling in atomic detail. Here we present the crystal
structure of the ¯agellin F41 fragment (relative molecular mass (M
41,300 (41K)) at 2.0 A
resolution, which directly reveals the proto-
®lament structure. A portion of the F41 molecule involved in the
axial subunit packing is associated with conformational switching,
providing insights into a mechanical switch that works with sub-
ngstrom precision.
Structure of F41
Because ¯agellin polymerizes into ®laments, evading crystallization
efforts, we prepared and crystallized a 41K fragment of Salmonella
¯agellin by clipping off peptides from both the amino- and carboxy-
terminal ends. Although the crystals were only several micro-
metres thick, various improvements in the cryocrystallographic
allowed data collection at ESRF and SPring-8, resulting
in a re®ned atomic model at 2.0 A
The Ca backbone trace of F41 is shown in Fig. 1a. The overall
shape of the molecule looks like a boomerang or an aircraft with two
wings and a short body, each wing being about 70 A
long, 25 A
and 20 A
thick. The F41 structure can be divided into three domains,
labelled D1, D2 and D3. Domain D1 comprises an N-terminal
segment from Asn 56 to Gln 176, and a C-terminal segment from
Thr 402 to Arg 450. Domain D2 also comprises two segments:
Lys 177 to Gly 189, and Ala 284 to Ala 401. A central segment
from Tyr 190 to Val 283 makes up domain D3. The domains are
connected by short stretches of two chains in both cases. A cross
b-motif is used to tie up the two ends of domain D1 connecting to
© 2001 Macmillan Magazines Ltd
D2, where two hydrogen bonds are formed between Asn 173 and
Thr 404. The two chains connecting domains D2 and D3, Gly 189 to
Thr 193 and Val 281 to Asn 285, form a short b-strand. The
distribution of domains and secondary structures along the
amino-acid sequence is summarized in Fig. 1c.
Domain D1 is rod shaped, about 70 A
long and 20 A
wide, and
comprises three a-helices and a strand of a unique b-turn/b-
hairpin/a-helix motif. The N-terminal chain forms two a-helices
and a b-hairpin. The ®rst a-helix (residues 57±99) extends through
the long axis of this domain, and the second one is relatively short
(104±129). The b-hairpin (140±160), with a small crossed loop at
its tip (146±153), is ¯anked by two distorted b-turns (131±134 and
135±138) on the N-terminal side, and slightly more than one turn
of a-helix (163±168) on the C-terminal side. This a-helix is
followed by an extended chain (170±176). The C-terminal chain
forms an a-helix (406±447) as long as the ®rst N-terminal a-helix.
This rod-shaped domain forms an extensive hydrophobic core
along its central axis in a similar way to four-helix bundles in
other protein structures. But, because of an irregularity in one of the
four strands involving the b-turn/b-hairpin/a-helix in series, the
packing of hydrophobic side chains is relatively loose in the middle
portion (Fig. 1b).
Domain D2 is made up of mostly b-strands except for one short
helix (285±289) and one short a-helix (288±298). The fold of a
set of these b-strands is unique and best described as a series of
randomly oriented b-hairpins, with some of them also involved in
three-stranded b-sheets. This domain has three distinct hydropho-
bic regions, two of which form enclosed hydrophobic cores; there-
fore, domain D2 can be divided into two compact subdomains, D2a
and D2b (Fig. 1b).
Subdomain D2a consists of the N-terminal stretch (177±189)
and the ®rst half of the C-terminal segment (284±344) of domain
D2. Its hydrophobic core comprising 12 side chains is enclosed in a
slightly irregular and unusual barrel formed by four b-strands and
one a-helix. Subdomain D2b is formed by the last half of the C-
terminal segment (345±401) of domain D2. Its hydrophobic core is
relatively small, consisting of six side chains that are enclosed within
the palm of a right-hand glove made of two b-hairpins correspond-
ing to the thumb and fore®nger and a loop corresponding to the
remaining three ®ngers.
These two subdomains form domain D2 by intimate interactions
between their hydrophobic surfaces. The third hydrophobic region
of domain D2 is formed as part of this interface. These hydrophobic
interactions between the two subdomains might explain why a
proteolytic fragment composed of domains D2 and D3 behaves as
two cooperatively unfolding domains instead of three in calori-
metric analyses
Domain D3 also consists of mostly b-strands with one short
stretch of a helical fold (199±209); however, this helical stretch does
not have the regular hydrogen-bonding pattern of an a-helix or a
-helix. Domain D3 can be regarded as an unusual barrel made of
four b-strands and one helical strand, which encloses 15 hydro-
phobic side chains in its core (Fig. 1a, b). This fold is similar to a part
of D2a described above, and in addition a section of this fold can be
described as a series of randomly oriented b-hairpins. Flagellin can
have domain D3 deleted without losing its ability to form ®laments,
although the ®laments are signi®cantly more unstable than those of
wild-type ¯agellin
. The relatively independent arrangement of this
domain from the rest explains why its deletion has a minimal effect
on the structural integrity of ¯agellin. The reduction in the ®lament
stability is due to the loss of intersubunit interactions, as shown by
an electron microscopic study
In domains D2 and D3, the Ca backbone trace shows a previously
undescribed fold. Three portions with this fold are isolated and
VOL 410
15 MARCH 2001
D2 D1
53 F41 450
Figure 1 The Ca backbone trace, hydrophobic core distribution and structural
information of F41. a, Stereo diagram of the Ca backbone. The chain is colour coded from
blue to red from the N to the C terminus. Prepared with MOLSCRIPT
b, Four distinct hydrophobic cores that de®ne domains D1, D2a, D2b and D3. All the
hydrophobic side-chain atoms are displayed with the Ca backbone. Side-chain atoms are
colour coded: Ala, yellow; Leu, Ile or Val, orange; Phe and Tyr, purple (carbon atoms) and
red (oxygen atoms). Prepared with RASMOL
. c, Position and region of various structural
features in the amino-acid sequence of ¯agellin. Shown are, from top to bottom: the F41
fragment in blue; three b-folium folds in brown; the secondary structure distribution with
a-helix in yellow, b-structure in green, and b-turn in purple; tic mark at every 50th
residue in blue; domains D0, D1, D2 and D3; the axial subunit contact region within the
proto®lament in cyan; the well-conserved amino-acid sequence in red and variable region
in violet; point mutations in F41 that produce the ®laments of different supercoils. Letters
at the bottom indicate the morphology of mutant ®laments: L (D107E, R124A, R124S,
G426A), L-type straight; R (A449V), R-type straight; C (D313Y, A414V, A427V, N433D),
© 2001 Macmillan Magazines Ltd
shown in Fig. 2 (corresponding regions of the amino-acid sequence
are indicated in Fig. 1c). The fold comprises a series of b-hairpins all
pointing away from one another like three thin leaves spread out or
like a curve of cubic polynomial called the `folium of Descartes'. We
therefore named it the b-folium. A common feature found in these
b-foliums is that most of the tips of b-hairpins are either bent or
twisted, and so some of them are better described as a b-®nger with
a claw.
Structure of the proto®lament
The ¯agellar ®lament is made of 11 proto®laments all with the same
polarity. From ®bre-diffraction studies, two distinct subunit repeat
distances along the proto®lament have been identi®ed as L- and
. More accurate measurements have determined that the
L-type repeat is 52.7 A
and the R-type repeat is 51.9 A
, with
experimental errors smaller than 0.1 A
(ref. 17). The repeat distance
along the a axis of the F41 crystal is 51.8 A
(6 0.1 A
), which
encouraged us to look for the proto®lament structure in the crystal
as soon as the electron density map became available. As expected,
we found that the proto®lament structure with F41 molecules lined
up along the a axis in the same orientation. The overall shape of the
proto®lament looks exactly the same as those found in lower
resolution density maps obtained by electron cryomicroscopy
The molecular packing in the crystal is shown in Fig. 3 in two
orthogonal views. Because the crystal has a space group of P2
, two
perfectly straight proto®laments are packed antiparallel and this
pair repeats along the b axis every 36.5 A
, forming a ¯at sheet. This
mode of molecular packing is the same as that found in the zinc
sheet of tubulin, in which the proto®laments of microtubule form
an antiparallel array within the monolayer sheet
. The ¯agellar
proto®laments formed a stack of sheets up to a few hundred layers
with a repeat of 119 A
, producing three-dimensional, plate crystals
that were several micrometres thick.
The atomic model of one proto®lament from the F41 crystal
structure was relatively easily docked onto a low-resolution density
map of the R-type straight ®lament from electron cryomicroscopy
and ®bre diffraction
(Fig. 4). The ®t is nearly perfect for a few
consecutive subunits, clearly showing that the proto®lament struc-
ture in the crystal is almost identical to that in the ®lament.
Although the proto®lament model is straight and the correspond-
ing proto®lament density is gently wound into a right-handed helix,
the twist and curvature of this helix are very small.
Three of the four domains identi®ed previously in the ®lament
density map, D1, D2 and D3 (named in the order of their radial
positions from inside to outside), correspond with the three
domains identi®ed in the F41 structure. Domain D1 of the F41
atomic model is located in the outer-tube region of the ®lament
(Fig. 4, OT), ®lling most of the outer-tube volume, but not the
inner-tube region (Fig. 4, IT). This indicates that a large portion of
the truncated terminal regions from Ala 1 to Arg 52 and from
Ser 451 to Arg 494, which are unfolded in the monomeric form,
occupies the inner-tube volume. Domains D2 and D3 respectively
®ll the two outer domains of the density map. These observations
are consistent with previous amino-acid assignments of the
, with slight corrections to the domain boundaries
(Fig. 1c). The nearly axial alignment of long a-helices in the outer-
tube region matches the original predictions from X-ray ®bre
diffraction data
The outer-tube domain is known to be solely responsible for the
formation of two distinct helical lattices, L-type and R-type
. Even
when the inner-tube structures are removed by truncating the
terminal segments, ¯agellin fragments can polymerize into super-
coiled ®laments if seeded by the wild-type supercoiled ®laments
This means that the outer-tube domain D1 has the ability to switch
between the two states of ¯agellin represented by the L-type and
R-type straight ®lament structures.
The proto®lament of the R-type repeat is held in the crystal by
axial interactions between D1 domains, and it exists in the absence
of lateral proto®lament interactions that construct the ¯agellar
®lament. This indicates that each proto®lament is an independent,
cooperatively switching unit, at least for the lengthwise mechanical
switch that produces the curvatures of supercoiled ®laments by
incorporating the two types of proto®lament into the tubular
structure. This also indicates that the axial intersubunit interactions
VOL 410
15 MARCH 2001
| 333
Figure 2 The b-folium fold. Left, residues 220±260 in domain D3; middle, residues
308±345 in domain D2a; right, residues 345±383 in domain D2b. The chain is colour
coded from blue to green to brown according to the amino-acid sequence. Prepared with
Figure 3 Crystal packing of F41 in two orthogonal views. a, The b±c plane viewed down
the a axis. b, The a±c plane viewed down the b axis. Each array of the F41 molecules
along the a axis, which has been identi®ed as the proto®lament of the ¯agellar ®lament, is
coloured red, yellow and green. The repeat distances are labelled. Antiparallel packing of
the proto®laments by the P2
symmetry of the crystal is clearly shown. Mutation sites that
affect the supercoiling are also labelled with side-chain atoms shown in CPK
representation. Mutations are coloured blue for L-type, red for R-type, and black for curly.
Prepared with MOLSCRIPT
© 2001 Macmillan Magazines Ltd
between D1 domains along the proto®lament are responsible for the
two-state switching to produce the two distinct repeat distances.
Axial interactions within the proto®lament
The axial intersubunit interactions are formed between domain D1
of the upper subunit and domain D1 and a small portion of D2a of
the lower subunit (Fig. 5a). The contact surface is relatively small,
and that is probably why, unlike the case of disassembling micro-
tubule tips
, isolated proto®laments have never been observed
under physiological conditions. The proto®lament seems to have
suf®cient stability only in the presence of lateral interactions:
parallel in the ®lament structure and antiparallel in the crystal. A
magni®ed image of the axial interface is shown in Fig. 5b. In the
bottom part of the upper subunit, a short segment of the ®rst N-
terminal a-helix (Asn 56 to Asp 69) and the unique motif of two
consecutive b-turns/b-hairpin (Phe 132 to Asp 151) form a concave
surface. In the top part of the lower subunit, a convex surface is
formed by short segments of two N-terminal a-helices with a short
loop connecting these a-helices (Gln 89 to Asp 107), Leu 408 and
Gln 409 of the C-terminal a-helix, and Asn 315 of subdomain D2a.
These two surfaces have complementary shapes, which produce a
number of van der Waals contacts.
The nature of the interactions is mainly polar±polar or charge±
polar. There are two small patches where the interactions are formed
between polar and hydrophobic groups, but there are no hydro-
phobic±hydrophobic interactions. Charge±polar interactions are
formed between Asp 69 and Asn 100, Asn 132 and Asp 107, the
carbonyl oxygen of Ala 149 and Arg 92, and Asp 151 and Gln 409, of
the upper and lower subunit, respectively. Thus, two of the four
residues, Ala 149 and Asp 151, are in the short loop at the tip of the
b-hairpin of the upper subunit, where Asp 151 is capping the N-
terminal end of the C-terminal a-helix of the lower subunit. A
portion of the b-hairpin also forms a short intersubunit b-strand
with a segment just after the ®rst N-terminal a-helix of the lower
subunit. These indicate the important roles of the b-hairpin in
domain D1 (Asp 140 to Lys 160) in these axial interactions.
Mechanical switching unit
As discussed above, the proto®lament itself is likely to have the
ability to switch between the two mechanically stable states with the
L- and R-type repeat. As the L-type repeat is only 0.8 A
longer than
the R-type, the conformational change would be relatively small. We
VOL 410
15 MARCH 2001
Figure 4 Docking of a single proto®lament model into the density map of the ®lament. Top
panel, end-on view from the top, with a 50 A
thick cross-section of the map. The map is
painted where the density is higher than the contour level. Bottom panel, side view, with a
50 A
thick longitudinal section of the map that spans from the ®lament axis towards
readers. The top points towards the distal end of the ¯agellum. The resolution of the
density map was limited to 20 A
for easy evaluation of the model ®tting. D0, D1, D2 and
D3 indicate the domains of ¯agellin; IT and OT represent the inner- and outer-tube
regions, respectively. Prepared with O
Figure 5 Axial interactions in the proto®lament. a, Space-®lling representation showing
the axial packing of two subunits. Prepared with RASMOL
. b, Stereo close-up view of the
boxed region in a. The two subunits shown are coloured cyan (upper) and dark pink
(lower). Some of the residues are labelled to guide identi®cation. Prepared with
© 2001 Macmillan Magazines Ltd
looked for a local conformational switch in the atomic model of F41
by simulating a gradual extension of the proto®lament by ®xing its
one end and pulling the other end in the axial direction. We used a
three-subunit proto®lament model, and ®xed the top subunit while
pulling down the bottom subunit to see what happened in the
middle subunit. At every step we translated the bottom subunit by
0.1 A
with its Ca backbone as a rigid body and carried out energy
minimization of the model.
The results are shown in Fig. 6. Up to 4.5 A
, every portion of the
middle subunit was elastically stretched gradually, as typically
shown by the elongating pitch of a-helices (Fig. 6a). But immedi-
ately after this point, further translation over the next two steps
(0.2 A
) caused an abrupt change in the conformation of the b-
hairpin in domain D1, which was accompanied by a small move-
ment of the C-terminal a-helix. Further translation caused further
gradual distortion of the molecule as observed before the transition.
The two conformations before and after the transition are super-
imposed for direct comparison in Fig. 6b. The characteristic nature
of this switching is a small but signi®cant movement of the b-
hairpin that slightly pushes out and down the lower subunit at the
axial interface, making the axial subunit dispositions with a longer
repeat distance stable.
The two terminal a-helical segments more or less follow the b-
hairpin movement, which probably maintains the side-chain inter-
actions with the b-hairpin in the hydrophobic core. Because we
carried out the simulated extension with an isolated three-subunit
proto®lament, without having all those restraints from lateral
proto®lament packing that are supposed to be present in the
®lament structure, some of the conformational changes shown
here may be artefacts. A more thorough model simulation is
needed to prove that the b-hairpin is the switch; however,
our current result strongly suggests that the conformational switch-
ing of the b-hairpin is responsible for the two distinct states of
¯agellin packing with two distinct repeat distances at sub-a
Mutations that affect the supercoiling
The amino-acid sequences of ¯agellin from various bacterial strains
show two extremely well conserved regions, which are about 170
residues from the N terminus and about 90 residues from the C
terminus (for example, by ProDom domain comparison
). From
the F41 structure and its position in the ®lament density, it is clear
why these regions are highly conserved. These conserved regions
form the densely packed core of the ®lamentÐthe outer and inner
tube (Fig. 4). About a dozen point mutation sites have been
identi®ed for a wild-type strain of S. typhimurium SJW1103 in
relation to the ®lament morphology
, and all except one are found
within these conserved regions (Fig. 1c). Within F41, mutations
D107E, R124A, R124S, G426A and A449V give rise to straight
®laments; and mutations D313Y, A414V, A427V, N433D and A449T
result in curly ®laments (see Fig. 3). We have attempted to interpret
the effect of these mutations on the basis of the proto®lament
Asp 107 is located at the beginning of the second N-terminal a-
helix in the top portion of the lower subunit, and interacts with
Asn 132 in the ®rst of the two consecutive b-turns before the b-
hairpin in domain D1 of the upper subunit (Fig. 5). The mutant
¯agellin D107E (from strain SJW1663) forms the L-type straight
. In the proto®lament model, the side chain of Asp 107
points up whereas that of Asn 132 points down, and the distance
VOL 410
15 MARCH 2001
| 335
Figure 6 Simulated extension of the proto®lament showing the possible switch region.
a, Superimposition of 12 sampled stages of the simulated extension. Domains D1 and D2
of three F41 molecules are shown (bottom half of the top, whole of the middle, and top half
of the bottom) and the 12 stages are colour coded from blue to red. Each step of the
extension was 0.1 A
, but every ®ve steps is sampled and displayed here; therefore, the
bottom subunits in different stages are equally separated by 0.5 A
. The whole b-hairpin at
the bottom of domain D1 of the middle subunit shows a discrete jump in its conformation
at an extension from 4.5 to 5.0 A
, whereas other parts show only gradual deformation of
an elastic nature. Prepared with O
. b, Close-up view of the axial contact region in stereo,
including the b-hairpin showing the conformational switch. The structures at two sampled
stages, at an extension of 4.5 A
in cyan and of 4.7 A
in dark pink, are superimposed to
show the conformational switch of the b-hairpin, of which the start and the end are
marked by yellow arrowheads. Prepared with MOLSCRIPT
© 2001 Macmillan Magazines Ltd
between them is slightly longer than an average hydrogen bond.
Mutation from aspartate to glutamate, which has a slightly longer
side chain, would strengthen this hydrogen bond, which may
stabilize the conformation of the switch region in the L-state.
Asp 313 is the only mutation located away from the conserved
regions, and was previously puzzling. The involvement of this
residue is now clearly shown in the structure. Asp 313 is close to
Asn 315 in a b-hairpin of subdomain D2a of the lower subunit,
where Asn315 makes a hydrogen bond with the main-chain oxygen
of Gly 133 of the upper subunit, which is also part of the b-turn to
which Asn 132 belongs (Fig. 5). Although not directly, Asp 313 is
involved in the axial packing interactions by three portions of the
molecule: the b-turn of the upper subunit; the loop between the two
N-terminal a-helices; and the tip of the b-hairpin of subdomain
D2a of the lower subunit. The mutation D313Y gives rise to curly
®lamentsÐone of the right-handed supercoils produced by
increasing the number of the R-type proto®laments from two to
four or ®ve (a normal ®lament of wild-type ¯agellin, a left-handed
supercoil, consists of two R-type and nine L-type proto®laments).
Because tyrosine has a signi®cantly bulkier side chain than has
aspartic acid, this may further stabilize the packing of the three
portions in the R-type conformation.
Ala 414 is involved in the hydrophobic core of domain D1 and is
close to the short a-helix after the b-hairpin. Its replacement with
valine also results in curly ®laments, probably through stabilization
of the hydrophobic core, which in turn stabilizes the R-type
conformation of the switch region.
The other mutated residues do not have any bonding partners
within the F41 proto®lament; Arg 124 and Ala 427 probably face the
lateral neighbours, and Gly 426, Asn 433 and Ala 449 appear to
point toward the inner-tube domain, which is missing in the current
model. This indicates that the speci®c form of supercoiling is
determined mostly by the lateral interactions between the proto-
®laments and/or the interactions between the outer- and inner-tube
Conformation of ¯agellin during transport
For ®lament growth, ¯agellin molecules are transported to the distal
end of the ¯agellum through its central channel, which has a
diameter of about 30 A
(refs 18, 19). Although there is no data on
the conformation of ¯agellin during the transport, ¯agellin
obviously either has to be unfolded or at least have its domain
largely rearranged.
The structure of F41 suggests that, even with intact terminal
regions, ¯agellin is made of independent domains that are con-
nected linearly, and that each domain except for domain D2 is thin
enough to ®t inside the channel. Therefore, it may be suf®cient to
have partially unfolded domain D2 and ¯exible interdomain con-
nections during the transport. If ¯agellin comes out of the ¯agellum
through the opening under the cap at the distal end
in the order of
D3 ®rst followed by D2 and D1, it would be ideal for a rapid self-
assembly process. This question remains to be answered by experi-
mental data.
Towards construction of the ®lament model
The structure of the ¯agellar proto®lament has provided many
insights into the structural mechanism of ¯agellin polymerization
and the possible switch region responsible for the two-state
mechanical switch of sub-a
ngstrom precision, which de®nes the
curvatures of the supercoiled ¯agellar ®laments. However, under-
standing the mechanisms that determine a particular twist of a
supercoil and its dynamic switching between left- and right-handed
ones, for example by the twisting force produced by rapid motor
reversal, requires detailed analyses of the lateral packing interactions
between the proto®laments in the ®lament structure.
The approximate models of the L- and R-type straight ®laments
produced by docking the proto®lament model into low-resolution
density maps with slight deformations are not suf®cient, because
the switching in the lateral interactions is a mutual sliding over a few
. To identify correctly the atomic interactions involved
in the switching, we are now incorporating experimental data
obtained from electron cryomicroscopy and X-ray ®bre diffraction
with those obtained from X-ray crystallography to construct
accurate models of the L- and R-type straight ®laments.
F41 preparation and crystallization
The F41 fragment of ¯agellin from Salmonella typhimurium was prepared and crystallized
as described
. Brie¯y, crystals were grown in solutions containing 6% PEG-6000, 12%
glycerol, 3±6% isopropanol, 50±75 mM NaCl, 20 mM Tris-HCl, pH 7.8, at 16 8C. Glycerol
and isopropanol were important in reducing the otherwise very high nucleation rate,
which resulted in many micrometre-sized needle crystals
Data collection
The space group of the crystal was P2
, with the cell dimensions a = 51.8, b = 36.5,
c = 118.7 A
, b = 90.88. The solvent content was about 55% and the crystal contained
one molecule per asymmetric unit. The size of the F41 crystal was typically 0.5 mm ´
0.2 mm ´ ,0.01 mm, which made data collection dif®cult. The crystals were frozen in
liquid propane and routinely annealed in the cryostream to improve the diffraction at high
. All the data sets were collected at temperatures around 100 K. Because of the
long exposure time (5 h per frame) required to record high-resolution spots, even with the
high-brilliance X-ray beam of our in-house Rigaku X-ray oscillation camera (®ne focus
X-ray generator, FR-D; Yale-type optics; image plate detector, RAXIS-IV), only well-
diffracting crystals were selected in the laboratory, and full data sets were collected at
synchrotron beam lines. Data sets used for the structure analysis are listed in the
Supplementary Information.
A set of multi-wavelength anomalous diffraction (MAD) data, which produced a high-
quality electron density map at 2.0 A
resolution, were collected under Trichromatic
at RIKEN beamline I (BL45XU) at the 8 GeV Super Photon ring (SPring-8) in
. The data were reduced by DENZO and SCALEPACK
. The atomic model was built as shown in the Supplementary Information. The
model was re®ned at 2.0 A
resolution including 354 water molecules (Table 1).
Structure analysis
Because the F41 crystals prepared from three different ¯agellins (wild type, L-type, and
G365C mutant of the R-type) all showed the same cell dimensions within experimental
errors, we treated them as isomorphous crystals. An electron density map was ®rst
obtained at 2.5 A
resolution by multiple isomorphous replacement using two derivative
data sets (L57 and C77; see Supplementary Information). We traced the main chain in
domain D1 using the graphics program O
, but it was dif®cult to complete the model
building. Then, we obtained new phases to 2.0 A
using a set of MAD data (G19; see
Supplementary Information) and program SOLVE
, which enabled us to complete the
model building. The N-terminal three residues are not included in the model because that
part of electron density was obscure. We then re®ned the atomic model, comprising 395
amino-acid residues and 354 water molecules, against the data collected at ESRF ID14-3 by
using X-PLOR
(see Table 1). We also re®ned the model against three data sets obtained
from the F41 crystals prepared from three different ¯agellins, but could not ®nd any
signi®cant difference between the three, which we will describe and discuss in more detail
Simulated extension of the proto®lament
Simulation of the proto®lament extension was carried at every 0.1 A
step by energy
minimization of the three-subunit proto®lament model with Ca atoms of both top and
VOL 410
15 MARCH 2001
Table 1 Summary of re®nement statistics
Resolution range No. of re¯ections
with F . 2j
No. of residues No. of protein atoms No. of water
* R
² r.m.s. bond length r.m.s. bond angle
10±2.0 A
395 2,880 354 23.4%
0.009 A
* R
hkl 2 F
, but calculated on 8% of data set aside for re®nement. Numbers in parentheses refer to those in the highest resolution shell.
© 2001 Macmillan Magazines Ltd
bottom subunits treated as rigid bodies. We used a routine in X-PLOR
for energy
Received 2 November 2000; accepted 11 January 2001.
1. Berg, H. C. & Anderson, R. A. Bacteria swim by rotating their ¯agellar ®laments. Nature 245, 380±382
2. Silverman, M. & Simon, M. Flagellar rotation and the mechanism of bacterial motility. Nature 249,
73±74 (1974).
3. Kudo, S., Magariyama, Y. & Aizawa, S.-I. Abrupt changes in ¯agellar rotation observed by laser dark-
®eld microscopy. Nature 346, 677±680 (1990).
4. Ryu, W. S., Berry. R. M. & Berg, H. C. Torque-generating units of the ¯agellar motor of Escherichia coli
have a high duty ratio. Nature 403, 444±447 (2000).
5. Larsen, S. H., Reader, R. W., Kort, E. N., Tso, W. W. & Adler, J. Change in direction of ¯agellar rotation
is the basis of the chemotactic response in Escherichia coli. Nature 249, 74±77 (1974).
6. Macnab, R. M. & Ornston, M. K. Normal-to-curly ¯agellar transitions and their role in bacterial
tumbling. Stabilization of an alternative quaternary structure by mechanical force. J. Mol. Biol. 112,
1±30 (1977).
7. Turner, L., Ryu, W. S. & Berg, H. C. Real-time imaging of ¯uorescent ¯agellar ®laments. J. Bacteriol.
182, 2793±2801 (2000).
8. O'Brien, E. J. & Bennett, P. M. Structure of straight ¯agella from a mutant Salmonella. J. Mol. Biol. 70,
133±152 (1972).
9. Asakura, S. Polymerization of ¯agellin and polymorphism of ¯agella. Adv. Biophys. (Japan) 1, 99±155
10. Calladine, C. R. Construction of bacterial ¯agella. Nature 225, 121±124 (1975).
11. Calladine, C. R. Design requirements for the construction of bacterial ¯agella. J. Theor. Biol. 57, 469±
489 (1976).
12. Calladine, C. R. Change of waveform in bacterial ¯agella: The role of mechanics at the molecular level.
J. Mol. Biol. 118, 457±479 (1978).
13. Kamiya, R. & Asakura, S. Helical transformations of Salmonella ¯agella in vitro. J. Mol. Biol. 106, 167±
186 (1976).
14. Kamiya, R. & Asakura, S. Flagellar transformations at alkaline pH. J. Mol. Biol. 108, 513±518 (1977).
15. Hotani, H. Micro-video study of moving bacterial ¯agellar ®laments III. Cyclic transformation
induced by mechanical force. J. Mol. Biol. 156, 791±806 (1982).
16. Kamiya, R., Asakura, S., Wakabayashi, K. & Namba, K. Transition of bacterial ¯agella from helical to
straight forms with different subunit arrangements. J. Mol. Biol. 131, 725±742 (1979).
17. Yamashita, I. et al. Structure and switching of bacterial ¯agellar ®lament studied by X-ray ®ber
diffraction. Nature Struct. Biol. 5, 125±132 (1998).
18. Mimori, Y. et al. The structure of the R-type straight ¯agellar ®lament of Salmonella at 9 A
by electron cryomicroscopy. J. Mol. Biol. 249, 69±87 (1995).
19. Morgan, D. G., Owen, C., Melanson, L. A. & DeRosier, D. J. Structure of bacterial ¯agellar ®laments at
11 A
resolution: Packing of the a-helices. J. Mol. Biol. 249, 88±110 (1995).
20. Mimori-Kiyosue, Y., Vonderviszt, F., Yamashita, I., Fujiyoshi, Y. & Namba, K. Direct interaction of
¯agellin termini essential for polymorphic ability of ¯agellar ®lament. Proc. Natl Acad. Sci. USA 93,
15108±15113 (1996).
21. Mimori-Kiyosue, Y., Vonderviszt, F. & Namba, K. Locations of terminal segments of ¯agellin in the
®lament structure and their roles in polymerization and polymorphism. J. Mol. Biol. 270, 222±237
22. Mimori-Kiyosue, Y., Yamashita, I., Fujiyoshi, Y., Yamaguchi, S. & Namba, K. Role of the outermost
subdomain of Salmonella ¯agellin in the ®lament structure revealed by electron cryomicroscopy.
J. Mol. Biol. 284, 521±530 (1998).
23. Samatey, F. A., Imada, K., Vonderviszt, F., Shirakihara, Y. & Namba, K. Crystallization of the F41
fragment of ¯agellin and data collection from extremely thin crystals. J. Struct. Biol. 132, 106±111
24. Vonderviszt, F., Uedaira, H., Kidokoro, S. -I. & Namba, K. Structural organization of ¯agellin. J. Mol.
Biol. 214, 97±104 (1990).
25. Honda, S., Uedaira, H., Vonderviszt, F., Kidokoro, S. -I. & Namba, K. Folding energetics of a
multidomain protein, ¯agellin. J. Mol. Biol. 293, 719±732 (1999).
26. Yoshioka, K., Aizawa, S. -I. & Yamaguchi, S. Flagellar ®lament structure and cell motility of Salmonella
typhimurium mutants lacking part of the outer domain of ¯agellin. J. Bacteriol. 177, 1090±1093 (1995).
27. Nogales, E., Sharon, G. W., & Downing, K. H. Structure of the ab tubulin dimer by electron
crystallography. Nature 391, 199±203 (1998).
28. Yamashita, I. et al. Radial mass analysis of the ¯agellar ®lament of Salmonella: Implications for subunit
folding. J. Mol. Biol. 253, 547±558 (1995).
29. Namba, K., Yamashita, I. & Vonderviszt, F. Structure of the core and central channel of bacterial
¯agella. Nature 342, 648±654 (1989).
30. Vonderviszt, F., Aizawa, S. -I. & Namba, K. Role of the disordered terminal regions of ¯agellin in
®lament formation and stability. J. Mol. Biol. 221, 1461±1474 (1991).
31. Mandelkow, E. -M., Mandelkow, E. & Milligan, R. A. Microtubule dynamics and microtubule caps: A
time-resolved cryo-electron microscopy study. J Cell Biol. 114, 977±991 (1991).
32. Corpet, F., Gouzy, J. & Kahn, D. The ProDom database of protein domain families. Nucleic Acids Res.
26, 323±326 (1998).
33. Kanto, S, Okino, H., Aizawa, S. -I. & Yamaguchi, S. Amino acids responsible for ¯agellar shape are
distributed in terminal regions of ¯agellin. J. Mol. Biol. 219, 471±480 (1991).
34. Kamiya, R., Asakura, S. & Yamaguchi, S. Formation of helical ®laments by copolymerization of two
types of `straight' ¯agellins. Nature 286, 628±630 (1980).
35. Yonekura, K. et al. The bacterial ¯agellar cap as the rotary promoter of ¯agellin self-assembly. Science
290, 2148±2152 (2000).
36. Yamamoto, M., Kumasaka, T., Fujisawa, T. & Ueki, T. Trichromatic concept at Spring-8 RIKEN
beamline I. J. Synchrotron Rad. 5, 222±225 (1998).
37. Ueki, T. & Yamamoto, M. The start of a new generation: the present status of the Spring-8 synchrotron
and its use in structural biology. Structure 7, R183±R187 (1999).
38. Otwinowski, Z. & Minor, W. Processing of X-ray Diffraction Data Collected in Oscillation Mode
(Academic, New York, 1997).
39. Leslie, A. G. W. CCP4/ESF-EACMB. Newslett. Protein Crystallogr. Vol. 26 (Daresbury Laboratory,
Warrington, UK, 1992).
40. Collaborative Computational Project Number 4. The CCP4 suite: Programs for protein
crystallography. Acta Crystallogr. D 50, 760±763 (1994).
41. Jones, T. A., Zhou, J. Y., Cowan, S. W. & Kjeldgaard, M. Improved methods for building protein
models in electron density maps and the location of errors in these models. Acta Crystallogr. A 47,
110±119 (1991).
42. Terwilliger, T. C. & Berendzen, J. Automated structure solution for MIR and MAD. Acta Crystallogr. D
55, 849±861 (1999).
43. Bru
nger, A. T., Kuriyan, J. & Karplus, M. Crystallography R factor re®nement by molecular dynamics.
Science 235, 458±460 (1987).
44. Kraulis, P. J. MOLSCRIPT: a program to produce both detailed and schematic plots of protein
structures. J. Appl. Crystallogr. 24, 946±950 (1991).
45. Merritt, E. A. & Bacon, D. J. Raster3D: Photorealistic molecular graphics. Methods Enzymol. 277, 505±
524. (1997).
46. Sayle, R. A. & Milner-White, E. J. RasMol: Biomolecular graphics for all. Trends Biochem. Sci. 20, 374±
376 (1995).
Supplementary information is available on Nature's World-Wide Web site
( or as paper copy from the London editorial of®ce of Nature.
We thank T. Tomizaki, L. Dumon, W. Burmeister, S. Arzt and S. Wakatsuki at ESRF, and
M. Kawamoto, N. Kamiya and K. Miura at SPring-8 for technical help with beamlines. We
also thank I. Yamashita and K. Hasegawa for a mutant strain of Salmonella that produces
SJW1655-derived site-directed mutant ¯agellin (G365C), which forms the R-type straight
¯agellar ®lament, and helpful information of heavy-atom binding to the ®lament; J. Tame
for critically reading the manuscript; and S. Asakura, T. Nitta and F. Oosawa for
continuous support and encouragement.
Correspondence and requests for materials should be addressed to K.N.
(e-mail: Atomic coordinates have been deposited in the Protein
Data Bank under accession code 1IO1.
VOL 410
15 MARCH 2001
| 337
© 2001 Macmillan Magazines Ltd
    • "Monomers of flagellin self-assemble through intermolecular contacts between the conserved D1 and D0 domains into flagellar filaments that enable bacterial motility [21,30] . The conserved amino acid residues within these regions represent an appropriate conserved target for the innate immune receptor TLR5 because structural requirements restrict extensive modifications , which would allow immune evasion, in this area of flagellin. "
    [Show abstract] [Hide abstract] ABSTRACT: Toll-like receptor 5 (TLR5) is a receptor of the innate immune system that recognizes flagellin from certain bacterial species and triggers an inflammatory response. The Salmonella dublin flagellin in complex with zebrafish TLR5 has been crystallized previously. In the present study, we extrapolate the structure of this complex using structure-guided mutagenesis to determine the recognition modes of human and mouse TLR5 receptors and demonstrate species-specific differences in flagellin recognition. In general, the recognition mode of the mouse receptor can be said to be more robust in comparison to that of the human receptor. All-atom molecular dynamics simulation showed differences between the two receptors within the primary binding region. Using a functional motility assay, we show that although the highly conserved area of the flagellin analyzed in this study encompasses key structural requirements for flagella formation, a direct correlation between immune recognition and structure on the level of amino acid residues is not observed.
    Full-text · Article · Jul 2016
    • "In order to accomplish functional insertion of an oligopeptide into the variable central portion of flagellin, it is essential to use appropriate linkers which ensure proper conformational flexibility and orientation. According to the high-resolution structural model of Salmonella flagellin [21,47] the resulting inner ends of the molecule upon removal of D3 are close to each other, separated by only 6 Å. We concluded that short linkers at both ends of the inserted motif may suffice. "
    [Show abstract] [Hide abstract] ABSTRACT: Statement of significance: In the present work, we show for the first time that.
    Full-text · Article · Jul 2016
    • "Following our analysis of the FliC Bp crystal structure (see below) and comparison with FliC from S. typhimurium, it became apparent that F96-111 is located in a region of the protein that mediates monomer-monomer interactions during assembly of FliC into its 11-subunit ring protofilament arrangement [49]. Therefore, when FliC Bp is assembled in its native form in the flagella, this region would not be solvent accessible, in agreement with its lack of recognition by human antibody. "
    [Show abstract] [Hide abstract] ABSTRACT: Author Summary Melioidosis is an infectious disease caused by Burkolderia pseudomallei that poses a major public health problem in Southeast Asia and northern Australia. This bacterium is difficult to treat due to its intrinsic resistance to antibiotics, poor diagnosis, and the lack of a licensed vaccine. Vaccine safety is a prime concern, therefore recombinant protein subunit and/or peptide vaccine components, may represent safer alternatives. In this context, we targeted one of the main subunit vaccine candidates tested to date, flagellin from B. pseudomallei (FliCBp) that comprises the flagellar filament that mediates bacterial motility. Based on the knowledge that activation of both cell-mediated and antibody-mediated responses must be addressed in a melioidosis vaccine, we identified B and T cell immunoreactive peptides from FliCBp, using both sequence-based and structure-based computational prediction programs, for further in vitro immunological testing. Our data confirm the accuracy of sequence-based epitope prediction tools, and two structure-based methods applied to the FliCBp crystal structure (here-described), in predicting both T- and B-cell epitopes. Moreover, we identified two epitope peptides with significant joint T-cell and B-cell activities for further development as melioidosis vaccine components.
    Full-text · Article · Jul 2015
Show more