Biophysical basis of phage liquid crystalline droplet-mediated antibiotic tolerance in pathogenic bacteria

Inoviruses are abundant filamentous phages infecting numerous prokaryotic phyla, where they can symbiotically promote host fitness and increase bacterial virulence. Due to their unique properties, inoviruses have also been utilised in biotechnology for phage display and as models for studying phase behaviour of colloidal rods. Inoviral phages secreted by bacteria can self-assemble into liquid crystalline droplets that protect bacterial cells in biofilms from antibiotics, however, factors governing the formation of such droplets and the mechanism of antibiotic protection are poorly understood. Here, we investigate the structural, biophysical, and protective properties of liquid crystalline droplets formed by Pseudomonas aeruginosa and Escherichia coli inoviral phages. We report a cryo-EM structure of the capsid from the highly studied E. coli fd phage, revealing distinct biochemical properties of fd compared to Pf4 phage from P. aeruginosa. We show that fd and Pf4 form liquid crystalline droplets with diverse morphologies governed by the underlying phage particle geometry and biophysics, rather than their surface biochemical properties. Finally, we show that these morphologically diverse droplets made of either phage can protect rod-shaped bacteria from antibiotic treatment, despite differing modes of association with cells. This study advances our understanding of phage assembly into liquid crystalline droplets, and provides insights into how filamentous molecules protect bacteria from extraneous molecules under crowding conditions, which are found in biofilms or on infected host tissues.


Summary 26
Inoviruses are abundant filamentous phages infecting numerous prokaryotic phyla, 27 where they can symbiotically promote host fitness and increase bacterial virulence. 28 Due to their unique properties, inoviruses have also been utilised in biotechnology for 29 phage display and as models for studying phase behaviour of colloidal rods. Inoviral 30 phages secreted by bacteria can self-assemble into liquid crystalline droplets that 31 protect bacterial cells in biofilms from antibiotics, however, factors governing the 32 formation of such droplets and the mechanism of antibiotic protection are poorly 33 understood. Here, we investigate the structural, biophysical, and protective properties 34 of liquid crystalline droplets formed by Pseudomonas aeruginosa and Escherichia coli 35 inoviral phages. We report a cryo-EM structure of the capsid from the highly studied 36 E. coli fd phage, revealing distinct biochemical properties of fd compared to Pf4 phage 37 from P. aeruginosa. We show that fd and Pf4 form liquid crystalline droplets with 38 diverse morphologies governed by the underlying phage particle geometry and 39

Introduction 48
Viruses that infect bacteria and archaea, known as phages, are among the most 49 common biological entities on Earth (Edwards and Rohwer, 2005 The first inovirus identified, named fd, was reported in 1963 as a DNA-containing 63 phage with an unusual filamentous morphology infecting Escherichia coli (Marvin et 64 al., 2014, Marvin andHoffmann-Berling, 1963). Fd has since become a model for 65 filamentous phage structure, infection, and assembly (Marvin, 1998, Marvin et al., 66 2014), as well as having found use in numerous biotechnological applications, 67 including the first described instance of phage display (Smith, 1985), and as a model 68 for examining colloidal-rod phase transitions (Dogic and Fraden, 2001). Despite its 69 importance in multiple areas of research, past structural studies have reported 70 markedly differing atomic structures of the fd phage capsid (protein pVIII in fd) using 71 fibre diffraction (Marvin et al., 2006, Marvin et al., 1994, solid-state nuclear magnetic 72 resonance (ssNMR, (Zeri et al., 2003, Marvin et al., 2006) and 8 Å-resolution electron 73 cryomicroscopy (cryo-EM, (Wang et al., 2006)), meaning that the structure of the fd 74 virion remains controversial. Thus far, only two high-resolution cryo-EM structures of 75 filamentous phage capsids have been solved; the first from Ike, which like fd, is a 76 member of the class I type of invoviral bacteriophages with pentameric (C5) capsid 77 symmetry (Xu et al., 2019); and second from Pf4, a class II inoviral bacteriophage with 78 C1 symmetry, which is expressed as a prophage in P. aeruginosa biofilms (Tarafder 79 et al., 2020). 80 81 A unique property of inoviral phages including fd and Pf4 is that they can self-assemble 82 to form ordered but dynamic liquid crystalline droplets under crowding conditions 83 outside the cell. Within droplets, laterally associated phages are orientationally 84 aligned, but not regularly ordered as in a crystal (Lekkerkerker and Tuinier, 2011). Due 85 to their monodisperse nature, filamentous phages have been used to study phase 86 behaviour of rod-like molecules, with fd having been employed for this purpose for 87 more than half a century (Lapointe andMarvin, 1973, Nakamura andOkano, 1983, 88 Tang and Fraden, 1995, Tomar  In this study, to determine the mechanism of protection conferred by phage liquid 102 crystalline droplets, we studied the biochemical and biophysical parameters 103 influencing liquid crystalline droplet formation by inoviral phages. We compared the 104 atomic structures, droplet-forming characteristics, and bacterial encapsulation 105 properties of fd and Pf4 phages on E. coli and P. aeruginosa cells. Using cryo-EM, we 106 report a 3.2 Å-resolution structure of the intact, native fd capsid, settling a long-107 standing debate about its structure. Using optical microscopy and electron 108 cryotomography (cryo-ET), we then show that fd and Pf4 form droplets with diverse 109 morphologies, governed by the underlying phage geometry rather than on the 110 biochemical traits of the phages. We show that even though droplets of fd and Pf4 are 111 Atomic structure of the fd phage capsid from cryo-EM 123 To study the biochemical properties of the fd phage capsid, we generated phage 124 preparations using previously established procedures (Methods). We observed ~64 Å 125 wide phage particles on cryo-EM grids ( Figure S1A). Using helical reconstruction, a 126 3.2 Å-resolution map of the fd phage capsid was obtained, which allowed derivation 127 of an atomic model of the capsid (Figures 1A-C and S1B-E, Movie S1 and Table S1). 128 In agreement with previous predictions (Marvin, 1966), the atomic model reveals a 129 pentameric (C5) subunit arrangement of the pVIII capsid protein, which forms a single 130 a-helix containing 50 residues. The helix is terminated at the N-terminus of the mature 131 protein by a proline residue (P6), and residues 1-5 of the capsid protein are disordered 132 ( Figure 1C). Due to its helical symmetry, capsid protein monomers stack almost 133 vertically to form a highly symmetrical arrangement along the helical axis ( Figure 1B). 134 The capsid proteins interact predominantly via a hydrophobic interaction network 135 ( Figure 1D), which include the residue Y21 that was shown previously to induce 136 structural instability (Tan et al., 1999) and hence mutated in several studies to 137 methionine (Zeri et al., 2003, Marvin et al., 2006, Blanco et al., 2011. In our specimen 138 of wild-type fd phage, despite retaining Y21, there was no significantly lowered 139 resolution detected at this location in the structure, even though local resolution varied 140 slightly within other parts of the capsid array ( Figure S1E). 141

142
The C-terminus of the capsid protein, which faces the inner lumen of the cylindrical 143 phage, interacting with the genomic DNA, has four positively charged lysine residues 144 exposed to the lumen ( Figure 1E). These lysines presumably bind to the negatively-145 charged DNA phosphates, as also seen in other phages (Xu et al., 2019, Tarafder et 146 al., 2020. Clear density for the DNA was not resolved in our fd cryo-EM map ( Figure  147 S1F), in agreement with previous studies on class I inoviral bacteriophages ( Figure  148 S2), where an unfeatured central DNA density was reported (Xu et al., 2019). 149 Comparison of our fd capsid structure with previous structural models of fd shows that 150 while the structure of the a-helical capsid subunit pVIII has been well-approximated in 151 some structural models, the arrangement of subunits into the overall capsid 152 architecture was imprecise in all ( Figure S3). properties of the droplet and its physical properties: 216 where R is the long (major) axis of the droplet, r is the short (minor) axis of the droplet, 219 K is the Frank elastic constant, γ is the surface tension, and V is the droplet volume. 220 The relation, which is valid for ≪ , predicts a decrease in the aspect ratio of the liquid crystalline droplet with an increase in the size (volume) of the droplet. As 222 predicted, we observed that the aspect ratios and volumes of both the fd and Pf4 223 droplets displayed this relationship ( Figure 4F). Interestingly, we found the aspect 224 ratios of Pf4 liquid crystalline droplets to be larger than the aspect ratios of the fd 225 droplets at similar volumes ( Figure 4F). 226

227
To explain the observed difference between the Pf4 and fd curves in Figure 4F, we 228 performed a second scaling calculation to link the physical droplet properties in the 229 prefactor of Eq. (1) to the phage geometry (see Methods). By approximating the elastic 230 constant K and the surface tension γ, we derived the following scaling relationship: 231 where b is the length (major axis) of the phage (fd 0.9 µm and Pf4 3.8 µm), and a is 234 the width (minor axis) of the phage (fd 64 Å and Pf4 62 Å). The above relation, which 235 contains only geometrical properties of the phages and droplets, predicts that for 236 droplets of similar volume, a larger phage length b should correspond to a larger 237 droplet aspect ratio. Both fd and Pf4 phages have similar widths, but Pf4 phages are 238 longer; as predicted, this is in line with the trend for larger aspect ratios of Pf4 droplets 239 than fd droplets at similar volumes ( Figure 4F). Biochemical phage interactions could 240 affect the physical prefactors in Eq. (1), and thus could enter Eq. (2), but do not seem 241 to underlie the qualitative differences in our measurements of droplet aspect ratios. 242 This suggests that our biophysical model of tactoids containing hard rods is sufficient 243 to explain the observed differences between Pf4 and fd liquid droplets in Figure 4F.  In this study, we have solved the capsid structure of an archetypal class I inoviral 299 bacteriophage, fd, which has been intensely studied, being the first discovered 300 inovirus, used in biotechnology for phage display (Smith, 1985)  stranded DNA genome (Marvin, 1966), and this is supported by our data as the linear 325 ssDNA in Pf4 would require less positively charged residues for encapsidation. Given 326 that fd has a comparable number of capsid proteins per length unit, the fact that the fd 327 capsid can compensate twice as many negative charges as Pf4 suggests that its DNA 328 genome indeed is circular. weight cut-off) snakeskin dialysis membranes (ThermoFisher). Pf4 was isolated from 385 static PAO1 biofilms as described previously (Tarafder et al., 2020). For further 386 amplification of Pf4, 1 x 10 3 pfu (plaque forming units)/ml of the Pf4 isolated from PAO1 387 biofilms were incubated with 1 ml of PAO1 culture at 0.5 OD600 for 15 minutes and 388 mixed with hand-hot 0.8% (w/v) agar. The mix was plated onto 10 cm 2 LB-agar plates 389 and incubated overnight at 37 °C. Each plate was covered with 5 ml PBS and 390 incubated for 6 hours before the PBS was collected, centrifuged (12,000 g, 30 minutes, 391 4°C) and the supernatant adjusted to 0.5 M NaCl and phage precipitated with 10% 392 (w/v) PEG 6000. Precipitated phage was harvested by centrifugation (12,000 g, 30  393 minutes, 4 °C). The phage containing pellet was resuspended in PBS and dialysed 394 against PBS overnight using 10 kDa MWCO snakeskin dialysis membranes 395 (ThermoFisher). Yield of both fd and Pf4 phage preparations was estimated using 396 Nanodrop (Thermo Scientific). 397 398

Cryo-EM and cryo-ET data collection 415
Cryo-EM data for screening specimens were collected using a Talos Arctica 416 (ThermoFisher) operated at 200 kV. High-throughput data was collected on a Titan 417 Krios microscope operated at 300kV fitted with a Quantum energy filter (slit width 20 418 eV) and a K3 direct electron director (Gatan) operating in counting mode at an 419 unbinned, calibrated pixel size of 1.1 Å using the EPU software. A combined total dose 420 of approximately 53.9 e -/A 2 per exposure was applied with each exposure lasting 3.6 421 s and 40 frames were recorded per movie. In total 4259 movies were collected 422 between -1 to -3 µm defocus. Tilt series data for cryo-ET was collected on a Titan 423 Krios using the Quantum energy filter and K2 direct electron director with SerialEM 424 software (Mastronarde, 2005). Tilt series were collected in two directions starting from 425 0° between ±60° with a 1° tilt increment, acquired with defoci ranging from -4 to -5 µm, 2020) and the fd capsid cryo-EM structure described in this paper were cropped using 543 VMD (Humphrey et al., 1996) and PyMOL (Schrodinger, 2015) to prepare simulations 544 of a smaller and computationally judicious system. The Pf4 and fd capsid systems 545 were centred in a 127.3 Å 3 and 126.2 Å 3 box respectively. Both systems were solvated 546 using the TIP3P water model (Jorgensen et al., 1983) and neutralised in 0.15 M NaCl. 547 One round of steepest descent energy minimisation was performed on each system 548 for 100 ps, followed by one 5 ns NVT and one 5 ns NPT equilibration, with restraints 549 applied to the heavy backbone atoms.