An unbroken network of interactions connecting flagellin domains is required for motility in viscous environments

In its simplest form, bacterial flagellar filaments are composed of flagellin proteins with just two helical inner domains, which together comprise the filament core. Although this minimal filament is sufficient to provide motility in many flagellated bacteria, most bacteria produce flagella composed of flagellin proteins with one or more outer domains arranged in a variety of supramolecular architectures radiating from the inner core. Flagellin outer domains are known to be involved in adhesion, proteolysis and immune evasion but have not been thought to be required for motility. Here we show that in the Pseudomonas aeruginosa PAO1 strain, a bacterium that forms a ridged filament with a dimerization of its flagellin outer domains, motility is categorically dependent on these flagellin outer domains. Moreover, a comprehensive network of intermolecular interactions connecting the inner domains to the outer domains, the outer domains to one another, and the outer domains back to the inner domain filament core, is required for motility. This inter-domain connectivity confers PAO1 flagella with increased stability, essential for its motility in viscous environments. Additionally, we find that such ridged flagellar filaments are not unique to Pseudomonas but are, instead, present throughout diverse bacterial phyla.

bacterium that forms a ridged filament on account of the arrangement of the two outer domains 23 of its flagellin protein, motility is categorically dependent on these flagellin outer domains. 24 Moreover, a comprehensive network of intermolecular interactions connecting the inner domains 25 to the outer domains, the outer domains to one another, and the outer domains back to the inner 26 domain filament core, is required for motility. This inter-domain connectivity confers PAO1 27 flagella with increased stability, essential for its motility in viscous environments. Additionally, 28 we find that such ridged flagellar filaments are not unique to Pseudomonas but are, instead, 29 present throughout diverse bacterial phyla. The bacterial flagellum is a complex and dynamic nanomachine that propels bacteria through 33 liquids. In pathogenic species, motility provided by flagella is critical for host colonization and 34 infection, and flagellar filaments are recognized as important virulence factors involved in 35 adherence, toxin delivery, biofilm formation and activation of innate immunity [1][2][3][4][5][6][7][8]. The 36 number of bacterial flagella attached to the cell body varies among species, from one, as in 37 Pseudomonas aeruginosa, to many, as in Escherichia coli and Salmonella enterica serovar 38 Typhimurium [9]. Each flagellum consists of a membrane-embedded basal body that functions 39 as a motor, as well as a hook and a long filament [10,11]. Torque generated by the basal body is 40 transferred by the hook to the filament, rotation of which provides a thrust that propels bacteria 41 through liquid. 42 The flagellar filament, a tubular structure composed of thousands of copies of the protein 43 flagellin (FliC in many bacteria), exhibits helical symmetry [12]. One FliC molecule is copied 11 44 times along the screw axis, making two turns, before it reaches the position directly above this protofilaments determine the twist of the filament and, consequently, swimming speed [13,14]. 49 If all protofilaments are in the same form, they produce a straight filament that does not create 50 thrust, resulting in immotile bacteria.

51
The filament is capped at its distal end by an oligomeric stool-like structure comprised of 52 five or six copies of the protein FliD [15,16]. FliC molecules are synthesized in the cytoplasm 53 and exported through the flagellar type III secretion system at the base of the flagellum [17]. 54 Since the filament extends from its distal end, FliC must be transported through the central 55 channel of the filament, which measures approximately 25 Å in diameter. Thus, FliC must be at 56 least partially unfolded while passing through the channel and, once it reaches the FliD cap, it 57 attains the correct structure and is positioned at the tip of the filament [17,18]. This process is 58 still not fully elucidated, but involves direct interactions between incoming FliC subunits and the 59 oligomeric FliD cap in a species-specific manner [19]. 60 The flagellar system of Salmonella has been studied as the archetypal model of bacterial 61 flagella for decades. The first high-resolution flagellin protein structure was from that of S. 62 Typhimurium [20], revealing a boomerang-shaped molecule with four domainsthe D0 and D1 63 inner domains that are predominantly α-helical and comprise the filament core, and the 64 propeller-shaped D2 and D3 outer domains that protrude from the filament at an angle of 65 approximately 90 degrees to the filament axis [20,21]. In S. Typhimurium  shown to contribute to hand-switching or motility and, instead, are thought to mediate many of 85 the other roles in which flagella are implicated, such as adherence and biofilm formation.

86
However, the growing catalog of diverse flagellar filament supramolecular architectures suggests 87 that FliC outer domains could, in certain bacteria, play a critical role in motility. The previously 88 reported 4.2 Å cryo-EM structures of L-and R-handed filaments from the P. aeruginosa strain 89 PAO1 revealed that the overall shape of this filament was ridged, unlike the splayed geometry of 90 the S. Typhimurium filament [27]. In the former, the D2 and D3 outer domains of FliC adopt an 91 end-on-end compact fold along the filament axis, resulting in ridges along the filament where 92 they are present and clefts along the filament where they are absent. Due to low local resolution 93 (>10 Å) of the D2 and D3 domains in the cryo-EM structure, the intermolecular interactions 94 maintaining the ridges, as well as connecting the ridges to the inner core, were not resolved. In 95 order to obtain a complete model of the PAO1 filament and to study the functional consequences 96 of the "ridged" structural organization, we turned to X-ray crystallography to resolve the 97 structure of the missing D2-D3 domains and, subsequently, performed a whole-atom 98 reconstruction of the P. aeruginosa PAO1 filament. Based on the resulting model, we identified 99 three interfaces that the D2 and D3 outer domains form with one another as well as with the D1 100 6 inner domain. Combining genetic tools, electron microscopy and swimming motility assays, we 101 showed that these interfaces form a connected network of inter-domain interfaces that are critical 102 not only for the structural integrity of the filament, but also for motility of the bacterium. Using 103 bioinformatic tools, we found that ridged flagellar filaments similar to that in P. aeruginosa 104 PAO1 are common throughout the bacterial kingdom and that such interconnected filament 105 architectures may provide an advantage to bacteria in generating the thrust needed to swim 106 through liquids of high viscosity.

109
High-resolution structure of the P. aeruginosa PAO1 FliC D2 and D3 domains 110 Although the cryo-EM structure of the PAO1 filament achieved near-atomic resolution for the 111 inner core of the filament, the resolution of the outer domains was substantially lower, with two distinct surfaces, the concave surface facing the filament core, and the convex surface 123 exposed to the outside. Starting from the N-terminus, the first β-strand is incorporated into the 124 D3 moiety, before the backbone turns downward, forming one half of D2. The sequence between   Figure 1B).

139
Whole-atom reconstruction of the L-and R-handed filaments in P. aeruginosa PAO1

140
Using the previously determined cryo-EM structure of the L-and R-PAO1 filament, we 141 successfully placed the X-ray crystal structure of the D2 and D3 PAO1 FliC into the residual 142 density and modelled the loops connecting D2-D3 to D1 (Figure 1C and S1C). We  the glycine turn between strands β2 and β3 of D2 +11 is positioned directly above helix 2 of D3 0 , 154 engaging in polar contacts with Q277 and S280 in L-and R-filaments, respectively. An 155 additional hydrogen bond between A266 and V207 is present in the L-filament. Apart from the 156 direct contacts within the outer domains, D3 also forms an interface with the β hairpin on D1 157 domain of the neighboring FliC subunit (D3 0 -D1 +11 interface) of comparable buried surface area. 158 We identified 4 possible polar contacts between D3 0 and D1 +11 in L-handed and none in R-  involving side chains we mutated residues to alanine, while in the cases of hydrogen bonds 174 involving main chain atoms, as well as loops that were part of the buried surface areas of the 175 inter-domain interfaces, but did not form hydrogen bonds, we deleted a variable number of 176 residues. Our mutational studies indicated that each one of the inter-domain interfaces within P. 177 aeruginosa PAO1 filamentsthe D2 0 -D1 0 , D3 0 -D2 +11 and D3 0 -D1 +11 interfacesare critical for 178 motility and that, collectively, they form a network of interactions bridging the outer domains to 179 the inner core of the filament that is required for motility, as follows.

180
In the D2 0 -D1 0 interface, we mutated seven residues that engage in hydrogen bonds and 181 reside on secondary structural elements in our model into alanine. Although also a part of the 182 D1 0 -D1 +11 interface, Y154 of FliC +11 participates in the hydrogen bond network formed by 183 Q326, Q356, D396 and N400, which are located in the D2 0 -D1 0 interface, and we mutated this 184 residue as well. As observed by electron microscopy, all of these mutations resulted in flagellar 185 filament formation ( Figure 3A). The majority of these mutations statistically significantly 186 impaired swimming motility compared to wild type. FliC mutants Y154A and N358A exhibited 187 the most substantial decreases in motility, with reductions in motile spread of 35% and 45%, 188 respectively.

189
In the D3 0 -D2 +11 interface ( Figure 3B), single site alanine mutations had either modest or 190 no effect on motility, however, alanine mutation of Q277, which forms a hydrogen bond with 191 G205 +11 in the L-handed filament, decreased the motile spread by one-third of that of the wild 192 type FliC. Conversely, shortening of the loops by truncation led to significant decreases in motile 193 spread in all but one case. As for the D2 0 -D1 0 interface, all FliC modifications in the D3 0 -D2 +11

194
interface resulted in the formation of flagellar filaments, as observed by electron microscopy.

195
Even in the most extreme case of Δ202Δ207-208, which was nearly entirely immotile, filaments 196 formed. However, the morphology of the filaments for some mutants differed compared to the 197 wild type. While the filaments in Δ249-250 were shorter than the wild type, those of Δ204-205, 198 Δ202Δ207-208 and Δ267 were primarily characterized by the loss of the wavy form observed for 199 wild type filaments, potentially implicating the area around the glycine turn in handedness and/or 200 hand-switching in these filaments. It should be noted that any filament visible in these 201 micrographs, however short they appear, are composed of at least thousands of subunits of FliC 202 extending well beyond the flagellar hook.

203
While mutations of individual residues in the D3 0 -D1 +11 interface did not markedly affect 204 PAO1 motility, we observed a profound effect on swimming motility, when either of the two 205 loop regions on D3 0 or D1 +11 that engage in hydrogen bonds were truncated in order to abrogate 206 these hydrogen bonding networks (Δ291-293 on D3 and Δ141-144 on D1 +11 ); again, the 207 formation of filaments was not affected ( Figure 3C). However, filaments in Δ141-144 and 208 Δ291-293 were significantly shorter compared to wild type, which could be due to lack of 209 optimal packing in the D3 0 -D1 +11 interface and a resulting loss of structural integrity of the 210 filament.

211
Regardless of how the above-described mutations affected motility, all resulted in the 212 formation of flagellar filaments, suggesting that each FliC mutant was expressed, exported and 213 properly folded by the bacterium. In order to validate that these mutant FliC proteins were 214 similar to wild type, we performed additional in silico and biochemical analyses. First, we 215 confirmed that AlphaFold [32,33] was able to properly predict our crystal structure of PAO1 216 D2/D3, which it did with an RMSD of 0.573 Å, suggesting that AlphaFold can predict PAO1-217 like flagellins with high confidence. Then, we used AlphaFold to predict structures of each 218 mutated FliC that we employed in the complementation studies above; each mutant protein was 219 highly structurally similar to our D2/D3 crystal structure ( Figure S4). We recombinantly 220 expressed and purified all of the full-length (i.e., inclusive of all domains D0 through D3) FliC 221 deletion mutants, and tested them for proper folding and thermal stability ( Figure S5). As 222 assessed by size exclusion chromatography, all of them were soluble and monomeric, except for 223 Δ204-205, which was predominantly dimeric. In addition, we subjected each of these deletion 224 mutants to differential scanning fluorimetry and determined that their melting temperatures were 225 all similar to that of the wild type. These data indicate that the changes in swimming motility 226 related to mutations or truncations in the complemented fliC genes were not a result of 227 substantial FliC protein structural differences, aggregation or instability. showed that while the swimming speeds of the PAK and PAO1 strains in liquid are comparable, 239 the ability of the PAK strain to swim in the semi-solid agar was significantly lower than that of 240 PAO1, with a motile spread of only 20% relative to PAO1, even though both strains have fully 241 formed flagella (Figures 4B and S2U). 242 We next sought to determine whether it was possible to replace type B FliC in PAO1 243 with type A FliC from PAK. We complemented the PAO1-ΔfliC strain with a plasmid-borne 244 type A FliC and tested for motility using the agar-based swimming assay and the presence of  Conversely, the FliC inner domains, D0 and D1, exhibit 54% identity and 75% similarity.  We first searched for flagellin structures from Pseudomonas species other than P. 314 aeruginosa and found that at least 10 of these have an outer domain that corresponds to type B 315 FliC (Figure 6B and S6). Next, we expanded our search for flagellin sequences from the entire  throughout the bacteria kingdom than previously appreciated. Using P. aeruginosa PAO1 as a 367 model system of ridged flagellar filaments, we dissected interactions involving its D2 and D3 368 outer domains and their effects on motility. As a consequence of the architecture of these ridged 369 filaments, these outer domains create a network of interfaces along the protofilament from the 370 inner core, through the outer domains and back to the inner core. By measuring motility of P. 371 aeruginosa PAO1 mutant filaments, we found that each one these interfacesbetween D1 0 and 372 D2 0 , connecting the inner core to the outer domains; between D3 0 and D2 +11 , bridging the outer 373 domains along the protofilament; and between D3 0 and D2 +11 , reconnecting the outer domains to 374 the inner coreis required for motility. Together, these interfaces create a network of 375 interactions upon which motility depends. Breaking this network at any point results in immotile 376 bacteria.

377
To demonstrate that FliC outer domains conferred a motility advantage to bacteria in 378 viscous environments, we complemented P. aeruginosa PAO1 with PAO1/PAK chimeric 379 filaments. Unlike the ridged filaments in P. aeruginosa PAO1, P. aeruginosa PAK exhibits 380 splayed filaments in which the outer domains do not make contacts with one another along the 381 protofilament. When we measured motility in liquid, wild type PAO1 and PAK, as well as all 382 chimeras that produced flagella, were similarly motile. Conversely, when challenged with the 383 relatively more viscous soft agar environment, wild type PAO1 was substantially more motile 384 than wild type PAK. Of those chimeras that produced flagella, only that which included the D1, 385 D2 and D3 domains from PAO1thereby reconstituting the entire network of interactions 386 between the inner core and outer domains in wild type PAO1was similarly motile to wild type 387 PAO1. 388 We found that ridged flagellar filaments, and consequently the networked interactions 389 between the inner core and outer domains, are not unique to P. aeruginosa PAO1. Our undercount; ridged flagellar filaments may be even more widespread throughout bacteria than we 399 have found here. It is also reasonable to believe that the total number of different outer domain 400 structures is smaller than suggested by sequence diversity of outer domains and that filaments 401 could be classified into just a handful of architectures, minimally including: naked (e.g., B. 402 subtilis), splayed (e.g., S. Typhimurium), ridged (e.g., P. aeruginosa PAO1), screw-like (e.g., S.  Although the maximum swimming speed of P. aeruginosa of 50 µm/s is half the speed of 423 C. jejuni, it is still twice that of S. Typhimurium [40]. Both P. aeruginosa and C. jejuni have 424 polar flagella. Due to the higher force and load to which their filament is subjected, it is 425 reasonable to expect a much tighter packing to preserve its structural integrity. As such, outer  Protein purification and differential scanning fluorimetry (DSF). Genes of the wild type FliC 501 and deletion mutants were subcloned into pET28-a vector with C-terminal polyhistidine tag.

502
Plasmids were transformed into E. coli BL21(DE3) cells, grown in LB medium at 37⁰ C until 503 OD600=0.6. After induction with 1 mM IPTG, cells were grown for another 4 hours at 37⁰ C.

504
Clarified lysate was applied to NiNTA resin (Qiagen) and eluted with 500 mM imidazole in PBS 505 buffer, followed by size-exclusion chromatography using Superdex 200 Increase 10/300 GL 506 column (Cytiva). For DSF analysis, 5 µM of protein in PBS was mixed with SYPRO Orange dye 507 (ThermoFisher) to a final concentration of 25X. The experiment was performed on a real-time 508 thermal cycler for a temperature range between 25⁰ and 90⁰ C.