A new twist on bacterial motility – two distinct type IV pili revealed by cryoEM

Many bacteria express flexible protein filaments on their surface that enable a variety of important cellular functions. Type IV pili are examples of such filaments and are comprised of a helical assembly of repeating pilin subunits. Type IV pili are involved in motility (twitching), surface adhesion, biofilm formation and DNA uptake (natural transformation). They are therefore powerful structures that enable bacterial proliferation and genetic adaptation, potentially leading to the development of pathogenicity and antibiotic resistance. They are also targets for drug development. By a complement of experimental approaches, we show that the bacterium Thermus thermophilus produces two different forms of type IV pilus. We have determined the structures of both and built atomic models. The structures answer key unresolved questions regarding the molecular architecture of type IV pili and identify a new type of pilin. We also delineate the roles of the two filaments in promoting twitching and natural transformation.

PilM, PilN and PilO 6 . In Thermus thermophilus, assembly of pilins into a T4P filament depends 55 on the assembly ATPase PilF, which interacts with the inner membrane platform via PilM 7 . 56 Two retraction ATPases, PilT1 and PilT2, are essential for T4P depolymerisation 8,9 . T4P are 57 extruded by the outer membrane secretin PilQ 10-13 . Recently, it has been suggested that 58 expression of the T. thermophilus major pilin PilA4 is temperature dependent, leading to 59 hyperpiliation at suboptimal growth temperatures 14  In this study, we combine different modes of cryoEM: electron cryo-tomography (cryoET) and 76 single-particle cryoEM, with functional data to study the T4P of T. thermophilus, which is a 77 well-established model organism. Surprisingly, we detect two forms of T4P, a wider and a 78 narrower form, assembled through the same machinery. We determine structures of the two 79 filaments at so-far unprecedented resolution (3.2 Å and 3.5 Å, respectively). This has enabled 80 us to visualise near atomic-level detail and build atomic models for each filament ab initio.

81
Our data unambiguously demonstrate that the wider pilus is composed of the major pilin 82  11 . Performing cryoET on cells grown at the optimal growth temperature of 68 91 °C revealed two types of pilus, with differences in their diameter. They emerge from the same 92 assembly machinery (Fig. 1a, b), suggesting that they are both T4P. We have previously shown 93 that transcription of the major pilin gene, pilA4, is upregulated at low temperature 14 . To 94 address the question of whether the growth temperature affects the assembly of the two 95 forms of pilus, we analysed the pili emerging from cells grown at the sub-optimal growth 96 temperature of 58 °C by cryoET. Again, two types of pilus were observed (Fig. 1c,d). Pili 97 emerge from T4P complexes only sporadically 11 , thus filaments were isolated from cells in 98 order to investigate their structure in more detail. Both wide and narrow forms of the filament 99 were detected in these preparations .   Table S1). In order to increase the hyperpiliation phenotype, we further reduced the growth 111 temperature to 55 °C. At this temperature we could identify PilA4 and TT_C1836 as the most 112 abundant proteins in the lower molecular weight bands, likely representing the pilin 113 monomers. In contrast, at the optimal growth temperature of 68 °C, only PilA4 was identified 114 reliably. We questioned whether the two T4P were expressed due to differences in 115 temperature or growth phase. To quantify the abundance of different filaments, cells were

123
To analyse the role of PilA4 and TT_C1836 in pilus assembly, we investigated the number of 124 wide and narrow pili per cell in deletion strains grown in liquid media to exponential phase 125 ( Fig. S3c). PilA4 deficient cells (pilA4::km) were not able to assemble any pili reliably, whereas 126 TT_C1836 deficient cells (TT_C1836::km) were only defective in their ability to assemble 127 narrow pili. These findings suggest that PilA4 has a role in producing both pilus forms, while 128 TT_C1836 appears to be crucial for the formation of narrow pili only. However, these data do 129 not allow us to discriminate whether the proteins have structural roles in comprising the 130 filaments, or have a more functional role in their assembly mechanism. 131 132

133
In order to investigate the architecture and protein composition of T4P at high resolution, 134 both filaments were subjected to analysis by single particle cryoEM and helical reconstruction.

135
In our micrographs, the wide pili appeared very straight while the narrow pili showed a much 136 higher degree of curvature, up to 2 µm radius (arrowhead in Fig. S4A, B). Based on 2D classes 137 (Fig. S4c,d) we determined the helical parameters for the wide and the narrow pilus ( Fig. S5a-138 d, see methods section for details). The rise and twist of the wide pilus measured 9.33 Å and 139 92.5° respectively, and the narrow pilus had a rise of 11.30 Å and a twist of 84.3°. Helical 140 reconstruction 28 resulted in maps at 3.22 Å (wide) and 3.49 Å (narrow) (Fig. 2a, b, S5e, f). The 141 diameter of the wide fibre is 70 Å (Fig. 2a) and is roughly cylindrical. In contrast, the narrow 142 filament has a zigzag-like appearance in projection, owing to a 15 Å-wide groove that winds 143 through the fibre. The diameter at any position along the long axis of the fibre axis is therefore 144 only 45 Å (Fig. 2b). Both structures were in good agreement with the data obtained in situ by 6 cryoET (Fig. 1). A low-resolution structure of the T4P from T. thermophilus was previously 146 determined by cryoET and sub-tomogram averaging, with a diameter of ~3.5 nm 11 . It seems 147 likely that this conformation represents the narrower form of the pilus.

170
The resolution and quality of both maps allowed us to unambiguously build an atomic model 171 for each filament ab initio (Fig. 3a -f). Guided by our mass spectrometry results, the position 172 of large side chains and clear differences in the length of the polypeptide backbones, we were 173 able to identify PilA4 as the building block for the wide pilus and the previously 174 uncharacterised protein TT_C1836 as the subunit for the narrow filament. We now propose 175 that TT_C1836 be named PilA5, in keeping with Thermus nomenclature.

177
The N-terminal a-helix, including the unfolded stretch, is comprised of the first 54 (PilA4) or 178 53 amino acids (PilA5). In both proteins the helix is disrupted by an unfolded stretch around 179 the conserved Pro22 (Fig. 3c, d, S6). The stretch in PilA4 is 4 amino acids long as opposed to 180 10 amino acids long in PilA5. The region between the N-terminal a-helix and the C-terminal 7 b-sheet, the so-called glycosylation loop, ranges in PilA4 from amino acids 55 to 77, with a 182 two-turn a-helix comprising amino acids 61-67. The C-terminal region is an antiparallel four-183 stranded b-sheet with the last strand facing towards the N-terminus followed by a loop that 184 ends on the b-sheet. A disulphide bond between Cys89, which is located in the second strand, 185 and the penultimate amino acid Cys124, likely stabilises the C-terminus (Fig. 3e) 198 199 A network of cooperative interactions between pilin subunits holds the fibres together. Each 8 movements between the subunits when the filaments are stretched. This is in accordance 218 with the observation that pili can stretch up to threefold upon force 31 . In addition, PilA4 219 contains salt bridges between Asp53 in subunit A and Arg30 in subunit D, and between Glu48 220 in subunit A and Arg28 in subunit E. In contrast, PilA5 contains a single salt bridge between 221 Glu68 in subunit A and Arg23 in subunit D (Fig. S7c). The conserved Glu5 is likely required to 222 neutralise the positive charge of the N-terminus within the hydrophobic core of the filament 223 4,32,33 . A salt bridge is also found between Glu5 and the N-terminus of the neighbouring subunit 224 in other T4P 23,24 . For both T. thermophilus fibres the distance between Glu5 and adjacent N-225 termini is too far to form a salt bridge. Instead, Glu5 forms an intramolecular salt bridge to 226 the N-terminus in the same subunit (Fig. S7c). This was also modelled for the related Klebsiella Densities were observed in both EM maps that protrude into the solvent and cannot be 231 attributed to the polypeptide backbone and were too large to account for an amino acid side 232 chain (Fig. 4). Interestingly, these densities co-localised with serine residues and were similar 233 in appearance to previously published densities attributed to glycosylation sites 29,35,36 . We We analysed cellular motility by twitching assays at 68 °C and 55 °C. Wild-type cells formed 253 characteristic twitching zones of ~2 cm and ~1.2 cm in diameter, respectively. The mutants 254 pilA4::km and pilA5::km did not exhibit any twitching motility (Fig. 5a). Since the immotile 255 pilA5::km cells could still produce wide pili comprised of PilA4, we deduce that PilA5 is 256 required to promote cell movement. Cells lacking both types of pili in the pilA4::km mutant 257 were completely defective in natural transformation, in agreement with our former finding 40 .

270
We have determined the first cryoEM structures of T4P that have allowed atomic models to 271 be built ab initio. Moreover, we have discovered two distinct T4P filaments, which are 272 composed of different proteins. Our data provide compelling evidence that PilA5 is essential 273 for twitching motility and confirm the previous finding that PilA4 is involved in natural 274 transformation 40 . In addition, we find that PilA4 is essential for the assembly of both wide and 275 narrow pili. PilA4 may therefore play a crucial regulatory role, could initiate pilus formation, 276 or even form a capping structure. In many bacteria, minor pilins are thought to prime pilus 277 assembly by reducing the energy barrier to the extraction of pilins from the membrane 41 . In

353
Magnifications varied depending on microscope; pixel sizes were within the range 3.8 -4.2 Å.

384
The microscope was fitted with K2 Summit direct electron detector and Quantum energy filter 385 (both Gatan, Pleasanton, USA). Dose-fractionated data were collected at 1.5 -4 µm defocus 386 using EPU (Thermo Fisher). 3138 micrographs containing both forms of pili were collected as 387 40-frame movies, corresponding to 8 seconds at a frame rate of 1 frame for every 0.2 seconds.

388
The total dose was 48 electrons/Å 2 at a magnification of 130,000 x, corresponding to a pixel 389 size of 1.048 Å.

391
Image processing, symmetry determination and helical reconstruction 392 Drift correction was performed using UNBLUR 51 . Straight sections of thin and wide fibres were 393 boxed separately from the drift corrected images using the helixboxer function of EMAN2, 394 such that the filaments were centred in each rectangular box. Helical reconstruction was 395 performed using the boxed filaments and SPRING as follows 52 . Contrast transfer function 396 13 (CTF) correction was performed using CTFFIND 53 . In order to determine the helical 397 parameters of the wide filaments a subset of the boxed filaments were cut into small 398 segments of 373 Å (a multiple of the helical rise) with an 80 % overlap, yielding a total of 9576 399 segments, which were classed in 2D (Fig. S4a). Close examination of the segments indicated a 400 filament diameter of 75 Å. The calculated power spectrum from the total segments indicated 401 clear layer lines that could be indexed. A meridional reflection at approximately 9 Å and a 402 layer line of order 1 at approximately 36 Å indicated that there are approximately 4 subunits 403 per turn. The ninth layer was found to be of order 1, suggesting that the helix repeats exactly 404 after nine turns, with a non-integer number of subunits per turn. The 405 SEGMENTCLASSRECONSTRUCT module in SPRING on class averages was used to determine 406 the accurate helical symmetry (Fig. S5a). The suggested output was determined to be either 407 4.10 or 3.89 subunits in a helical pitch of 36.3 Å. In order to determine the helical parameters 408 for the narrow filaments a subset of the boxed filaments were cut into segments of 800 Å with 409 a step size of 330 Å and classed in 2D (Fig. S4b). The helical pitch could be determined directly 410 as 48.1 Å. A meridional reflection and thus a helical rise at 11.3 Å could be identified. These 411 parameters allow calculation of a helical rotation of 84.6° and 4.26 subunits per turn. The 412 SEGMENTCLASSRECONSTRUCT module in SPRING was again used to determine the accurate 413 helical symmetry (Fig. S5b). The suggested output was determined to be either 4.11, 4.14,