Three-dimensional observations of an aperiodic oscillatory gliding motility behaviour in Myxococcus xanthus using confocal interference reflection microscopy

The Delta-proteobacterium, Myxococcus xanthus, has been used as a model for bacterial motility and to provide insights of bacterial swarming behaviours. Fluorescence microscopy techniques have shown that various mechanisms are involved in gliding motility, but these have almost entirely been limited to 2D studies and there is currently no understanding of gliding motility in a 3D context. We present here the first use of confocal interference reflection microscopy (IRM) to study gliding bacteria, and we reveal aperiodic oscillatory behaviour with changes in the position of the basal membrane relative to the coverglass on the order of 90 nm in vitro. Firstly, we use a model plano-convex lens specimen to show how topological information can be obtained from the wavelength-dependent interference pattern in IRM. We then use IRM to observe gliding M. xanthus and show that cells undergo previously unobserved changes in their height as they glide. We compare the wild-type with mutants of reduced motility, which also exhibit the same changes in adhesion profile during gliding. We find that the general gliding behaviour is independent of the proton motive force-generating complex, AglRQS, and suggest that the novel behaviour we present here may be a result of recoil and force transmission along the length of the cell body following firing of the Type IV pili.


Introduction 33
Bacteria use a number of mechanisms to move through their local environment in response 34 to chemotactic signals, to form communities or to invade their host. The most studied mode 35 of bacterial motility is flagellar-mediated movement. Other modes, such as the twitching 36 motility displayed by Pseudomonas aeruginosa, use Type IV pili (T4P) to direct movement 37 based on the extension, adhesion and retraction of polar filaments from the leading pole of 38 the cell 1,2 . However, not all bacteria rely solely on extracellular appendages for motility. The 39 phenomenon of gliding motility has been identified in a diverse range of bacterial species 40 spanning various phyla 3-8 . The Delta-proteobacterium Myxococcus xanthus displays two 41 different modes of gliding motility -adventurous motility and social motility -to seek out 42 nutrients or prey as part of its complex lifecycle 4, 9-15 . 43 There are contrasting models proposed to explain the mechanisms underpinning gliding 44 motility 9,16-19 . The focal adhesion complex (FAC) model proposes that FACs form on the basal 45 surface of the cell and attach to the underlying substrate while coupling to the helical MreB 46 cytoskeleton on the cell's inner-membrane 16,17,20,21 . It has been shown that FACs translocate 47 linearly from the leading pole as the cell moves forwards and is driven by the force generated 48 by the AglRQS gliding complex, which is associated with the FAC 16,22 . The FAC model 49 requires the basal layer of the cell to be firmly attached to the underlying substrate, however, 50 it remains unclear how the complex is able to traverse the peptidoglycan cell wall without 51 compromising the structural integrity of the cell 8,19 . A second model has been suggested where 52 proton motive force (PMF) generated by AglRQS results in a helical rotation of the MreB 53 cytoskeleton in gliding cells which are firmly adhered to a solid substrate 9,22-25 . In the helical 54 rotation model, stationary foci of fluorescently-tagged motor complex subunits have been 55 explained as being a build-up of multiple complexes arrayed in "traffic jams" which result from 56 areas of differing resistance in the underlying substrate 9,24 . Both models converge where the 57 gliding cell is adhered firmly to the surface of the underlying substrate to facilitate gliding. 58 However, our observations show that cells are not in fact firmly adhered during gliding motility, 59 but instead exhibit aperiodic fluctuations in their height as they glide. 60 Bacterial gliding motility has mainly been studied using phase contrast and fluorescence 61 microscopy techniques which do not provide 3D information about cell movement 9,17,22,26 . We 62 hypothesised that axial changes in cell shape during gliding motility may occur due to the 63 complex nature of underlying mechanisms such as FAC translocation and bulk movement of 64 the cytoskeleton. We reasoned that novel 3D behaviours could be visualised using the label-65 free microscopy technique interference reflection microscopy (IRM MreB cytoskeleton are found in distinct foci on the basal side of the cell 9,16 . These membrane-107 associated complexes have been suggested to change the surface topology of the gliding cell 108 depending on the cargo load of the molecular motor 9 . More recently, interferometric scattering 109 microscopy (iSCAT), which detects both the reflected and scattered light, has been used to 110 observe T4P-mediated twitching motility in P. aeruginosa. In this work the authors generated 111 3D illustrations which revealed the role of T4P machinery subunits in extension, attachment 112 and retraction based on the interference pattern in iSCAT images 47 . 113 Background Theory Common assumptions of this IRM model are that no other refractive index boundary exists in 138 the cell specimen (i.e. that the refractive index of the cell is constant) and that incident and 139 reflected light are perpendicular to the coverglass 28 . Also, the impact of the numerical aperture 140 (NA) is neglected which affects the depth of field that is imaged. In IRM, high NA objectives 141 are used to limit detection of reflection signals to those originating from interfaces close to the 142 coverglass, which establishes an experimental condition that is close to the three-layer system 143 established, here, consisting of the coverglass, the cell medium and the cell body 39 . 144

145
Characterisation of model specimen 146 We characterised the axial height intensity profile of a specimen of known structure to compare 147 with IRM theory by acquiring IRM images at different wavelengths of a plano-convex lens 148 (focal length = 72 mm) that was placed on a coverglass. A composite of IRM images of the 149 lens specimen acquired at 488 nm, 514 nm and 543 nm is shown in Figure 1a. In Figure 1b, 150 a cross-sectional schematic of a plano-convex lens specimen is shown, outlining the axial 151 position of the intensity maxima that are caused by constructive interference for different order 152 and wavelength. We analysed the intensity of the interference fringes by comparing the radial 153 intensity profile with the theory of IRM regarding fringe separation 27 . We calculated that the 154 theoretical spacing (l/2n) between the intensity maxima caused by constructive interference 155 for the different wavelengths are 244 nm, 258 nm and 272 nm (n = 1). With the experimental 156 data we obtained slightly different spacings for the intensity maxima, with 249 ± 1 nm, 262 ± 157 1 nm and 277 ± 1 nm ( Figure 2b). Thus, experimental and expected values deviated by 2.07%, 158 1.63% and 1.91% for the different wavelengths. Additionally, the overlap of the intensity 159 maxima of different acquisition wavelengths decreased with lens-to-coverslip distance ( Figure  160 2). These observations provided a sense of directionality regarding specimen topology, where 161 the curvature of the plano-convex lens specimen was clear from the acquired images and 162 allowed for 3D reconstruction of the lens specimen ( Figure 2c). This means that assumptions 163 outlined in Background Theory are appropriate to reconstruct the morphology of simple model 164 systems. Here a multi-wavelength IRM approach provides important additional morphological 165 information over single wavelength IRM. 166

167
Axial changes along the cell body during gliding motility is independent of AglQ 168 To demonstrate the benefit of using confocal IRM over widefield IRM we first imaged wild-type 169 M. xanthus on a commercial widefield system. We demonstrated that low contrast images are 170 generated in widefield IRM where interference fringes along the cell cannot be resolved 171 clearly. The raw widefield data is presented with a magnified region showing gliding cells 172 (Supplementary Figure 2a) accompanied by the background-corrected data (Supplementary 173 Figure 2b). Interference fringes cannot be clearly resolved in either the raw or corrected data, 174 meaning that a widefield approach is not suitable for studying the changing adhesion profile 175 of gliding bacterial cells. 176 Using confocal IRM we observed previously undocumented changes in the axial position of 177 M. xanthus cells as they glide. We reason that force transmission via T4P extension, attachment and retraction could be 201 responsible for the height fluctuations we report. 202 203 Using multi-wavelength IRM for extracting 3D-directionality 204 Figure 4 illustrates how multi-wavelength IRM data can be assessed qualitatively to 205 understand the geometry of a single gliding M. xanthus cell (additional gliding morphologies 206 are presented in Supplementary Figure 3). A representative wild-type gliding cell is presented 207 where multi-colour fringes can be observed along the cell body (Figure 4a). In the IRM image, 208 lifting of the cell body is clearly indicated by an alternating fringe pattern along the body. By  209 interpreting the intensity plot profile along the cell body (Figure 4b) we can extract qualitative 210 topological information about the cell morphology. Figure 4b  Using IRM to measure the velocity of gliding cells 225 We used IRM to determine the mean velocity of motile cells and showed that deletion of aglQ 226 decreased the length of time which cells remain adhered to the glass substrate compared to 227 the wild-type (Figure 5a). This implies that AglQ is responsible for maintaining adherence to 228 the coverglass during gliding. The mean velocity of gliding cells was determined by measuring 229 the displacement of the cells over time as they glide, selecting cells with an approximately 230 linear trajectory. We found a 41.2% decrease in mean velocity of DK1622-ΔaglQ (mean 231 velocity = 9.26 ± 0.72 µm/min) when compared with the wild-type (mean velocity = 15.76 ± 232 0.89 µm/min), which concurs with their altered motility phenotype (Figure 5b). 233

Discussion 234
We report aperiodic changes in the adhesion profile of gliding myxobacteria using the label-235 free technique, IRM. We began with the hypothesis that previous studies had failed to identify 236 any axial behaviours in gliding cells due to the drawbacks with conventional imaging 237 techniques, and that changes in cell height may arise due to the complex mechanisms which 238 govern gliding motility. The data presented here shows new behaviour in gliding myxobacteria 239 which do not fit the current gliding motility models. The behaviours we report suggest that 240 there are additional factors which mediate gliding motility and show the benefit of using IRM 241 to extract 3D information from bacterial specimens by using an easily-implemented 242 microscopy technique. 243 The current consensus is that gliding cells must be firmly attached to a solid surface to facilitate 244 gliding, however these data show that this is not the case. Our results show that throughout 245 gliding, cells undergo changes in the axial position of their basal surface on the order of 90-246 180 nm. This new information indicates that the helical rotation and FAC motility models do 247 not fully explain the mechanisms of bacterial gliding. One previous study has used RICM to 248 investigate the adhesion profile of myxobacteria during detachment 17 . However, they did not 249 report any fringes along the cell body, or any aperiodic changes in height in gliding cells. This 250 is likely due to the poor contrast of widefield RICM compared to confocal IRM that prohibits 251 the detection of higher interference fringe orders. It is important to also note that some 252 researchers have claimed that the ability of IRM to detect regions of close contact is 253 questionable due to the inhomogeneous refractive index of the cytosol proximal to the plasma 254 membrane and from self-interference from higher fringe orders within the cell 42 We therefore used a manual tracking method to measure the mean velocity of gliding cells. 300 We concluded that the DK1622-ΔaglQ moved on average 30% slower than the wild-type. Also, 301 the wild-type remains adhered to the surface for longer periods of time, yet the dynamic 302 fluctuations in height we report remain. 303 This work has provided new insights into the 3D motility of bacteria and identified novel motility 304 behaviours in M. xanthus which suggest there may be unknown mechanisms which do not 305 agree with the current FAC and helical rotation models. This study does not establish if the 306 novel 3D motility behaviour is only limited to M. xanthus, or aperiodic oscillations in height are 307 a behaviour common to other gliding bacteria. We suggest that the fluctuations in height we 308 observe are mediated by T4P and occur due to recoil following firing of pili from the leading 309 pole. In this work we attempted to image T4P mutants to confirm this hypothesis 310 ( Supplementary Movies 3 and 4), however these mutants were unable to attach to the glass 311 substrate and therefore unable to glide. Therefore, this study does not resolve the precise 312 mechanism(s) which underpin this behaviour, and further investigation is required to identify 313 factors which play a role in the gliding motility of these organisms. Moving forward it would be 314 interesting to assess the role of T4P in the changing adhesion profile of gliding cells. One way 315 to achieve this would be by investigating the spatiotemporal dynamics of T4P firing in 316 association with the changing adhesion profile of gliding cells using a correlative TIRF-IRM 317 approach. This would allow for simultaneous imaging of the adhesion profile of gliding cells 318 and fluorescently-tagged T4P proximal to the coverglass. The development of an image 319 processing workflow to extract 3D information from confocal IRM images of myxobacteria 320 would also provide users with a method to quantify the aperiodic oscillatory behaviour. 321

322
Bacterial cell culture 323 Myxococcus xanthus cultures (see Table 1) were maintained on double casitone yeast extract 324 (DCYE) medium (20 g/L casein hydrolysate, 2 g/L yeast extract, 8 mM MgSO4, 10 mM Tris-325 HCl, (with 20 g/L agar for solid medium)). For imaging, cells were inoculated at high cell 326 densities in liquid DCYE and grown for 48 hours at 30°C while shaking at 250 rpm. Prior to 327 imaging an 800 µL sample of the exponentially-growing culture was removed from liquid 328 culture and placed in a 35 mm optical-bottom petri dish with a coverslip thickness of 180 µm 329 (cat. no. 80136; ibidi GmbH, Germany) and incubated at 30°C for 20 minutes to allow cells to 330 adhere. 331 332 The refractive index of liquid DCYE medium was measured to be 1.33 using an Abbe 333 Refractometer (Billingham & Stanley Ltd., U.K.). 334 335

336
For the characterisation of a model lens specimen, the specimen was placed convex-side 337 down on a 170 µm-thick coverglass measuring 50 x 24 mm which bridged the microscope 338 stage insert. An Olympus IX81 inverted microscope coupled to an Fluoview FV1000 confocal 339 scanning unit (Olympus, Japan) was used to image the lens specimen. The microscope was 340 configured for IRM by replacing the emission dichroic by an 80/20 beam splitter. Images were 341 acquired using a 10x/0.3 N.A. UPlanFl objective lens (Olympus, Japan) and reflection signals 342 were detected using a photomultiplier tube (PMT) for each wavelength, with spectral detection 343 limited to a 10 nm bandwidth over the wavelength of incident light used. A 488 nm line from 344 an Argon laser source (GLG3135; Showa Optronics, Japan) was used for single wavelength 345 acquisition. For multi-wavelength acquisition, 488 nm and 514 nm lines were provided by an 346 Argon laser and 543 nm was provided by a Helium-Neon-Green laser source (GLG3135;  347 Showa Optronics, Japan The length of the neighbourhood, k, was selected depending on the specimen that was 372 imaged. For the lens specimen, a large k-value (k = 1000) was chosen to prevent lowering the 373 contrast of the lens signal. For IRM images of M. xanthus, where the frequency of the observed 374 interference fringes relative to the pixel density is high, a relatively small k-value (k = 30) 375 proved suitable. Line intensity profiles from raw and corrected IRM images were checked to 376 verify that the position and intensity succession of the interference fringes were not altered 377 due to the correction method (Supplementary Figures 4, 5 and 6). 378 379 Lens analysis and reconstruction 380 In this work, we verified that multi-wavelength IRM can be used to study the change in cell 381 topology during gliding by imaging a lens specimen of known geometry and comparing the 382 results to the theoretical model (refer to Background Theory). The one-dimensional height 383 profile of a lens zlens follows: 384 with R as the radius of the lens and x as the distance to the centre point of the lens that 386 touches the surface/coverslip (x = 0, z = 0). 387 Images acquired at multiple wavelengths of the lens specimen were linearly contrast adjusted 388 and cropped to create a square image using FIJI 57 . A composite RGB image was created by 389 merging the channels. The data was imported into MATLAB 2018b, and using the same radial 390 analysis method as presented by Tinning et al., we calculated the axial height of interference 391 fringes from the composite RGB image of the lens specimen 58 . We used a findpeaks function 392 to determine the spacing between the experimental intensity maxima of the interference 393 fringes. Each subsequent constructive interference fringe was subtracted from its 394 neighbouring fringe to calculate the experimental spacing. 395 Lens reconstruction was performed using MATLAB. Firstly, the radial distance for each pixel 396 was extracted from the centre of the RGB IRM image of the lens specimen. By applying a 397 priori knowledge of the lens specimen geometry to the radial distance of each pixel (refer to  398 Background Theory), we were able to assign each pixel to an axial height value based on the 399 experimental fringe separation calculated previously. The x, y and z coordinates along with the 400 image intensities were plotted to create a 3D reconstruction of the RGB IRM image. 401  shows the maxima and minima of the interference fringes acquired using both 488 nm (cyan) 581 and 635 nm light (magenta). The spectral separation of the two interference patterns can also 582 be observed. Axial directionality of the cell can be determined by interpreting the colour-583 ordering of the fringes, where fringes arising from the longer wavelength appear after those 584 from the shorter wavelength when the cell in inclined, and the opposite for declining slopes. 585 The plot was acquired by averaging the signal over a line width of 3 pixels. DK1622 has an increased mean path length when compared to DK1622-ΔaglQ, which shows 596 that DK1622-ΔaglQ has a lower adhesion profile than the wild type.