Divergence of Quaternary Structures Among Bacterial Flagellar Filaments

See allHide authors and affiliations

Science  18 Apr 2008:
Vol. 320, Issue 5874, pp. 382-385
DOI: 10.1126/science.1155307


It has been widely assumed that the atomic structure of the flagellar filament from Salmonella typhimurium serves as a model for all bacterial flagellar filaments given the sequence conservation in the coiled-coil regions responsible for polymerization. On the basis of electron microscopic images, we show that the flagellar filaments from Campylobacter jejuni have seven protofilaments rather than the 11 in S. typhimurium. The vertebrate Toll-like receptor 5 (TLR5) recognizes a region of bacterial flagellin that is involved in subunit-subunit assembly in Salmonella and many other pathogenic bacteria, and this short region has diverged in Campylobacter and related bacteria, such as Helicobacter pylori, which are not recognized by TLR5. The driving force in the change of quaternary structure between Salmonella and Campylobacter may have been the evasion of TLR5.

Proteins with almost-unrecognizable sequence similarity can share highly conserved structures, suggesting that sequences can diverge more rapidly than structures over the course of evolution (1, 2). Quaternary structure, on the other hand, may be much more sensitive than tertiary structure to small changes in sequence, and this sensitivity may provide a mechanism for large evolutionary divergence with only small changes in sequence. But relatively few examples exist for dramatic changes in quaternary structure (3, 4). The bacterial flagellar filament has been a model system for understanding protein self-assembly, protein secretion, motility, and immunogenicity (5). The bacterial flagellar filament genes are highly conserved over large evolutionary distances, leading to the reasonable assumption that the atomic structure of the Salmonella flagellar filament (6) is a model for all homologous bacterial flagellar filaments (7). In fact, the assumption that a bacterium will have 11 protofilaments as in Salmonella has been taken as axiomatic (8). We have used electron microscopy (EM) to examine the flagellar filaments from Campylobacter jejuni, and we show that these contain only seven protofilaments.

C. jejuni is one of the major causes of bacterial diarrhea worldwide. The organism, which is a member of the epsilon division of the Proteobacteria, is characterized by a rapid, darting motility mediated by polar flagella. The filament is composed of a minor flagellin, FlaB, and a major flagellin, FlaA. However, full-length filaments can be formed when only FlaA is present, and these have nearly normal motility (9). C. jejuni flagellins, like those of other polar flagellates, are glycosylated, but, unlike other polar flagellates, glycosylation is required for filament formation in Campylobacter and the related organism Helicobacter pylori. Flagellins from the epsilon Proteobacteria also are unique in that they fail to activate vertebrate Toll-like receptor 5 (TLR5). In addition to mediating motility, which is essential for intestinal colonization, Campylobacter flagella, in the absence of type 3 secretion systems, also function to secrete nonflagellar proteins that modulate virulence, and the glycans on flagellin mediate auto-agglutination and microcolony formation (10).

The wild-type flagellar filaments from Campylobacter exist under normal conditions in different supercoiled states (Fig. 1A), consistent with the classical model of polymorphic switching between two discrete states of the component protofilaments (11, 12). The two states are termed R (for right-handed) and L (for left-handed) on the basis of the twist of the filament when all protofilaments are in one state. In Salmonella the structural parameters have been measured to very high precision, and the supercoiling arises from the fact that the R state of the protofilament is ∼1.6% shorter than the L state (13). This nonequivalence of all protofilaments breaks the strict helical symmetry of the flagellin subunits and leads to a curvature of the filament that generates thrust when the flagellar filament is rotated by the flagellar motor. The curvature of these flagellar filaments, as well as the heterogeneity of the protofilaments within them, make these wild-type filaments poor candidates for image analysis. The most detailed picture of the Salmonella flagellar filament (6) has come from the Ala449→Val449 (A449V) mutation (14, 15) in the coiled-coil domain D1 (fig. S1) that locks the protofilaments into the R state, forming straight filaments that cannot generate thrust. The corresponding mutation was constructed in the flaA gene of C. jejuni strain 81-176 and inserted into a chromosomal location in a flagellin-deletion background (16, 17). This mutation (A531V) does not eliminate motility as judged by swarming assays on agar plates, and electron micrographs of these filaments show that they are not straight and have similar supercoiling as wild-type filaments. This observation suggests that details of the packing of flagellin subunits must differ between Salmonella and Campylobacter.

Fig. 1.

Negatively stained and cryo-EMs of Campylobacter flagellar filaments. Negatively stained wild-type filaments (A) are curly, whereas the G508A mutant filaments (B) are rather straight. Nevertheless, some filaments can be seen [(B) arrows] that appear to be supercoiled. (C) A cryo-EM image of the G508A mutant shows long straight filaments. An averaged cryo-EM image of the filaments (Dwas generated from 10,557 overlapping segments (each 200 pixels long, 2.4 Å per pixel) by using the IHRSR filrecon procedure. An averaged power spectrum (E), generated from 16,904 negatively stained segments each 460 Å long, shows two layer lines (marked with arrows). The higher layer line is at ∼1/(26 Å), whereas the lower one is at ∼1/(50 Å). A third layer line is so close to the equator that it cannot be resolved with segments of this length.

We found that the Gly508 → Ala508 (G508A) mutation in Campylobacter, corresponding (fig. S1) to the G426A mutation in domain D1 that locks Salmonella flagellar filaments into the L state (14), produces largely straight filaments (Fig. 1, B and C) as judged by EM. Bacteria with the G508A mutation display no motility in solution by light-microscopic observation but do display partial motility in swarming assays using agar plates. The partial motility is not explained by revertants, because sequencing of the flagellin gene from these apparently motile bacteria shows that the G508A mutation still exists and no other mutations (second site revertants) are present in the flagellin gene. Given the greater molecular weight of Campylobacter flagellin (59.5 kD and ∼65 kD with glycosylation) compared with Salmonella flagellin (52.2 kD) and the fact that the sequence insertions in Campylobacter are in the D3 domain that is on the outside of the filament, it is surprising that the Campylobacter filaments are only ∼180 Å in diameter (Fig. 1D), in contrast to the ∼220 Å diameter of the Salmonella filaments. This is explained, as we show, by the different symmetry of these two filaments.

The straight filaments of Salmonella have a nearly crystalline order, which has allowed for structure determination by EM at near-atomic resolution (6). In contrast, images of the Campylobacter G508A filaments display limited diffraction in either negative stain or by cryo-EM, and we can show that this results in part from structural variability (movie S1). An averaged power spectrum (Fig. 1E) computed from these filaments shows only three layer lines.

We used quick freeze and deep etch (18) EM to visualize the surface features of the Campylobacter flagellar filaments (Fig. 2, A and C). Under the conditions used, all filaments examined had left-handed long-pitch helical stripes (Fig. 2D). However, the pitch of the long-handed stripes was variable. In contrast to the left-handed protofilaments in the quick freeze–deep etch images, cryo-EM images were sorted into segments having either left-handed or right-handed protofilaments, and different twist states of either hand could be found (movie S1). Because the cryo-EM samples were prepared under different conditions (diluted in a low-salt buffer), we speculated that the low ionic strength and low salt used for cryo-EM imaging might introduce structural heterogeneity because changes in salt, pH, and ionic strength can change the polymorphic state of flagellar filaments (8, 19, 20). Under low ionic strength and low salt conditions, protofilaments could be seen by quick freeze–deep etch EM in a number of different states (Fig. 2, F to H), from L to nearly vertical to R. Interestingly, the transitions between these states appeared to be discrete and associated with “kinks” or bends (Fig. 2, F to H, arrows), which is probably similar to the highly cooperative transitions that have been observed in flagellar filaments at a macroscopic scale (21). Furthermore, many of the filaments were fragmented into short lengths because of the large forces that would result from a change in twist of a long filament adsorbed to the mica. For example, one end of a 1-μm-long segment would need to rotate by one or two full turns in the transition from the L state to the vertical state or the R state, respectively.

Fig. 2.

Quick-freeze/deep etch EMs of the Campylobacter G508A flagellar filaments. (A and C) Filaments can be seen that contain protofilaments with a left-handed coil under normal buffer conditions. An averaged power spectrum (B) generated from 48 non-overlapping segments (each 1024 pixels long, with a sampling of 5.4 Å per pixel) shows a one-sided layer line at ∼1/(50 Å) that arises from a left-handed three-start helix. A two-dimensional average (D) was made from 3806 overlapping segments (each 100 pixels long) that shows the long-pitch left-handed protofilaments. Although the metal shadowed images do not correspond to projections of a helical structure, the IHRSR technique (21) can be used to generate helical reconstructions from such images (4, 27), and a cross section from a reconstruction is shown (E). When these filaments are extensively washed with pure water after adsorption to the mica, structural heterogeneity can be seen (F to H). For all images, right-handed RecA-DNA filaments [labeled in (F)] were added as a control for the helical hand. In (F), the arrow indicates a transition from nearly vertical protofilaments (top of filament) to protofilaments with a left-handed coil (middle of filament). In (G), the arrow indicates a transition from nearly vertical protofilaments (left portion of filament) to a right-handed coil. In (H), the arrow indicates a transition from left-handed coiling (left portion of filament) to nearly vertical protofilaments. The scale bar in (F) indicates 500 Å.

We have used the iterative helical real space reconstruction (IHRSR) method (22) to generate reconstructions from the different twist states of the Campylobacter filaments imaged by cryo-EM. Resolution is limited by the heterogeneity of structural states, and we have not been able to do better than ∼15 Å resolution. For comparison with the atomic model of Salmonella derived from right-handed filaments (Fig. 3, A and D), we show the R state of Campylobacter flagellar filaments (Fig. 3, B and E).

Fig. 3.

Three-dimensional reconstructions of flagellar filaments. In (A and D) the atomic model of the Salmonella flagellar filament (6) has been filtered to 15 Å resolution for comparison with the cryo-EM reconstruction of the Campylobacter flagellar filament (B to E). The reconstruction has been generated from 10,557 overlapping segments (each 200 pixels long, 2.4 Å per pixel). (C) The core of the reconstruction (reconstruction truncated at a radius of 33 Å) shows vertical rods of density that are consistent with coiled-coils. The atomic model of the Salmonella D0 coiled-coil (red ribbons) has been fit to the Campylobacter reconstruction. Cross sections of the model (D) and the Campylobacter reconstruction (E) show the similar central lumen, surrounded by 11 protofilaments in Salmonella (D) and seven protofilaments in Campylobacter (E). The region surrounding the lumen, arising from D0 in Salmonella, has been colored red in (D) and (E). The scale bar for (D) and (E) indicates 50 Å.

Despite the large difference in symmetry between Salmonella (∼5.5 subunits per 26 Å pitch helical turn) and Campylobacter (∼3.5 subunits per 26 Å pitch helical turn) flagellar filaments, there are some conserved parameters. In Salmonella the axial rise per subunit (13) within each of the 11 protofilaments is 52.7 Å in the L state and 51.9 Å in the R state, whereas in Campylobacter the axial rise per subunit within each of the seven protofilaments is ∼55.5 Å in the L state and ∼52.5 Å in the R state. It should be noted that these values in Salmonella were determined to very high precision by using x-ray diffraction (13), whereas there is much greater error in our values using EM. In Campylobacter there is a ring of density surrounding the central lumen that contains seven rodlike features, in contrast to the 11 rodlike densities in Salmonella. The length of these rodlike densities is very similar to the length of the coiled-coil D0 domain in Salmonella (Fig. 3C). The packing of these coiled-coils is looser in Campylobacter than in Salmonella, which is consistent with the nearly crystalline order in Salmonella and the structural heterogeneity in Campylobacter. Despite the fact that there are only seven coiled-coils surrounding the lumen in Campylobacter and 11 in Salmonella, the sizes of the lumens are comparable.

The resolution of our reconstruction is not sufficient to suggest where D1 is located. Residues 89 to 96 in Salmonella are located in D1, and these residues are recognized by TLR5 (23). Because mutation of these residues in Salmonella leads to a loss of bacterial motility (24), presumably by either interfering with flagellar assembly or preventing polymorphic switching, and these residues are not conserved in the comparison with Campylobacter (a bacterium that does not activate TLR5), it has been suggested that the evolution of a flagellin that is both functional and evades TLR5 would require a complex sequence of mutations (23). Our results suggest that a substantially different packing of D1 domains between Salmonella and Campylobacter may have served this purpose.

Are Campylobacter flagellar filaments unique, or could other bacteria also have seven rather than 11 protofilaments? Previous observations in the literature may be explained by filament symmetries other than that found in Salmonella. It was shown (8) that the flagellar filaments of both Idiomarina loihiensis and Pseudomonas aeruginosa have different macroscopic forms than that found in Salmonella. In addition, fewer polymorphic states were observed for these filaments compared with those of Salmonella. It was shown by model building that a change of twist and curvature of the component protofilaments from the parameters found for Salmonella was needed to fit the observed macroscopic states of I. loihiensis, but the existence of 11 protofilaments in Idiomarina was taken as an axiom. An obvious prediction of the models for polymorphic switching (11, 12) is that the number of possible states is related to the number of protofilaments. With fewer protofilaments, we expect that Campylobacter would also have fewer polymorphic states than found in Salmonella.

One of the main mechanisms in evolutionary divergence is the accumulation of point mutations. Sequence divergence in proteins does not always map into corresponding structural divergence of these proteins because some protein folds have remained largely invariant while their sequences have diverged so greatly as to have no recognizable homology. For example, the core of the bacterial RecA recombination protein (25) can be superimposed quite well on the core of the mammalian F1 adenosine triphosphatase (ATPase) (26). However, RecA polymerizes to form a helical polymer, whereas this core is part of a hexameric ring in the F1 ATPase. We do not know at what point in the divergence of these two proteins that the change in quaternary structure took place. The results presented here show that, within a family of highly conserved proteins (flagellin), a different quaternary structure can be adopted with only limited sequence divergence of the D1 domain largely responsible for the helical symmetry. This may emerge as an important mechanism in the divergence of entire organisms.

Supporting Online Material

Materials and Methods

Fig. S1


Movie S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article