Using rotational integration of oblique interferometric scattering (RO-iSCAT) to track axial spatiotemporal responses of membrane protrusions

Despite the crucial importance of dynamic membrane protrusions for understanding 21 phagocytosis, cellular communication and mechanobiology, current imaging modalities struggle 22 to quantitatively track their real -time, 3D spatiotemporal dynamics with sufficient molecular 23 specificity and minimal perturbation. Many membrane protrusions studies still utilize confocal 24 microscopy where its axial resolution and high phototoxicity remains a key limiting factor for live 25 axial imaging. We discovered that multiple rotational oblique interference scattering (RO-iSCA T) 26 leverages off -axis illumination to induce a larger lateral shift in out -of-focus iSCAT signals 27 compared to in -focus signals. This phenomenon provides a foundation to generate speckle -free 28 widefield interferometric signals with a 10-fold signal to noise ratio improvement, eliminating the 29 need for any background subtraction. RO -iSCAT enables real -time, label -free, and minimally 30 invasive imaging of diverse membrane protrusions within complex co -cultures. RO-iSCAT 31 enables nanoscale-sensitive tracking of membrane protrusion dynamics along the axial direction . 32 This allows for the construction of dynamic axial variance maps, facilitating quantitative 33 measurements of membrane protrusion formation at tens to hundreds of nanometer displacements, 34 without requiring 3D volumetric imaging. RO-iSCAT empowers real time quantitatively dissection 35 of the axial spatiotemporal complexities of membrane protrusions and unlock future insights into 36 fundamental processes like cell migration, durotaxis, and intercellular communication. 37 38

filopodia tracking, membrane protrusion, label-free microscopy 39 40 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint Summary Figure 41 42 43 44 Key points 45 • Discovered that multiple integrated rotational oblique interference scattering (RO-iSCAT) 46 generates speckle-free widefield interferometric signals with a 10-fold signal to noise ratio 47 improvement, eliminating the need for any background subtraction. 48 • Removed need for 3D volumetric imaging to quantified axial motion of membrane 49 protrusion forming tethers, trails and bridge with within ~ tens of nanometer accuracy. 50 • Enabled classification of membrane protrusions that, despite possessing identical chemical 51 compositions, are differentiated by their interactions, thus offering a qualitative comparison 52 of membrane protrusions at the nanoscale in living cells. 53 54 55 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint

56 Membrane protrusions (lamellipodia, pseudopodia, filopodia, microvilli, invadopodia, and 57 podosomes) possess dynamic three-dimensional spatiotemporal behaviours because they mediate 58 a wide variety of extracellular interactions between a cell and its three-dimensional 59 microenvironment 1. While these dynamic protrusions are the result of cytoskeletal (e.g. actin, 60 microtubules) rearrangement, their 3D spatiotemporal relationship are initiated by the activation 61 and clustering of membrane receptors 2. In particular, 2D spatiotemporal tracking of filopodia - 62 membrane extensions indicate their role in mechanical and chemical sensing 3, phagocytosis 4, 5 63 and migration 6, 7. Observations of protrusion dynamics on coated substrates, between cells and in 64 tissue 8 have led to discoveries on contact-dependent cell-cell communication 9, twisted tethers 10, 65 forming migrasomes from retraction fibers 11, and tunnelling nanotube 12 as well as gaining closer 66 insight into tissue development 13. 67 Existing tools to track membrane protrusion, extensions and distribution in the spatiotemporal 68 domain relies heavily on standard light microscopy technique (brightfield -phase contrast, 69 fluorescence), have inadequate resolution, are prone to phototoxicity, and lacking specific 70 fluorescence markers, cannot readily classify the transient behaviors of 3D membrane protrusion 71 and extensions in live cell cultures. Whilst the use of volumetric imaging technologies 14 and 72 advanced image processing 15 has made significant advances, the issue of phototoxicity and 73 photobleaching remains a concern for longitudinal imaging 16 that is necessary for quantitative 74 mapping of membrane protrusions. Electron microscopy (EM), on the other hand, has become a 75 routine tool to identify these protrusions that forms membrane bridges because of its ability to 76 measure physical feature membrane protrusion based on physical size, diameter (50 –200 nm 77 diameter), the distance between distant cell and importantly, and their proximity with substrate for 78 classification 17. Unfortunately, EM slices face methodological difficulties because membrane 79 protrusion such as tethers and tunnelling nanotubes are often fragile after chemical fixation, and 80 prone to deform due to sample preparation. Owing to well-defined refractive index difference in 81 actin and lipid in thin membrane protrusion, it is plausible that interference scattering microscopy 82 (iSCAT) 18-20 signals can be more effectively at quantitative tracking the transient movements of 83 different types of membrane protrusions that form, disassemble and maintain at the nanoscale in 84 3D than fluorescence microscopy. However, iSCA T signals often require background subtraction 85 20, which can be challenging to implement and may fail to remove speckles in populated cell 86 cultures, thus hindering the tracking of protrusion spatiotemporal patterns in live cells. 87 This paper examines the optical principles of a Rotational Oblique Interferometric Scattering 88 signals (RO-iSCAT) to achieve speckle free interferometric scattering signals in real time. We then 89 follow on to explain how RO-iSCAT interference patterns are used to track spatiotemporal of 90 membrane protrusion; that can transits into trails (i.e. retraction fibers to migrasomes), membrane 91 tethers or bridges 21. We provided evidence demonstrating that axial variation maps of interference 92 scattering signals are effective in accurately categorizing various membrane protrusions. The axial 93 variation information possesses rich spatial temporal signature even within a single protrusion that 94 surpass standard kymograph in fluorescence or scattering images which only consider rudimentary 95 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint shapes. The paper comprises of 4 main parts – numerical model of RO -iSCAT, quantification of 96 imaging resolution of RO-iSCAT, identification of membrane protrusion using spatial signature in 97 lateral interference pattern alongside with variance in axial displacement analysis of RO-iSCAT 98 images, and longitudinal tracking of membrane protrusions in co-cultured cells lines. 99 100

101 Rotational integration removes out-of-focus interference scattering signals 102 Under off-axis illumination, we observed that out-of-focus iSCA T signal experienced a larger 103 lateral shift than in-focus iSCAT signal (Supplementary Video 1). To explain this effect, we began 104 with a numerical model (Supplementary Note 1 and Methods) and synthesized the interference 105 intensity signals of RO-iSCAT. The scheme of oblique illumination we adopted was at a single 106 angle, where an incoming illumination (blue line and shades) entered the sample at an oblique 107 angle 𝜃 along a single azimuthal orientation 𝜑 (Fig. 1a). The path length difference arises from 108 refractive index difference between glass-water interface (reference field) and scattered light that 109 is necessary to form an iSCA T signal. Whilst the reference field is constant (Fig. 1a, first reflecting 110 surface-blue line), the scattered signal (Fig. 1a, green line) varies along the axial plane 𝑧. Because 111 iSCAT signal is of interferometric nature, the signal is changed by the properties of scattered signal 112 collected by the imaging lens. This phase delay (defocused wavefront) changes with axial distance 113 𝑧 of the imaging lens for each oblique angles 𝜃 and azimuthal direction 𝜑. 114 At each oblique angle, an off-axis phase shift causes iSCAT interference fringe pattern to shift 115 laterally away from the focal plane . We numerically calculated the phase and fringe shifts at 116 different azimuthal angle (Fig. 1b, Supplementary Fig. 1 and Supplementary Video 2). Each 117 off-axis oblique phase delay (Fig. 1b i) from different azimuthal sources (𝜑 = 0°, 90°, 120°, 200°) 118 is convolved with defocusing phase delay that will create nonlinear phase shifts (Fig. 1b ii, iii), 119 which directly translate to the intensity fringes translating laterally (Fig. 1b iv). 120 To confirm th e effect of lateral shifts in intensity fringes in RO-iSCAT, first we simulated 121 lateral shift of intensity fringes across multiple axial planes from -2 µm to 2 µm and compared 122 with the measured experimental results (Supplementary Video 3 left). From both our model and 123 experiment (Fig. 1c), we observed that the lateral fringe shifts increased further away from the 124 focal plane, whereas at the focal plane, the iSCAT signal experience shift almost negligible. Then, 125 we examined the modelled shift in fringes along the transvers plane at a zimuthal rotational 126 direction 𝜑 = 0° and 120° (Fig. 1d i) as well as the final integrated RO-iSCAT images 𝜑 = 0° −127 360° (Fig. 1d ii). The integrated RO-iSCA T shows a significant reduction of side lobe (profiles 128 in Fig. 1d) that indicates an increased visibility of the interference fringes at the focal plane. This 129 rotational oblique configuration reduces out-of-focus signal equivalent to confocal configuration 130 19 (Supplementary Fig. 2 and Supplementary Video 4). Hence, directly integrating multiple 131 oblique illuminated iSCA T signals generates a high contrast iSCAT image at the focal point only, 132 without intensity losses that can occur in background subtraction 20 or pinholes 19. 133 Next, we turned to examine the 3D interference point spread function (iPSF) of on axis (𝜃 =134 0.5° ), iSCAT and off axis ( 𝜃 = 22° ), RO -iSCAT over a 4 48-micron FOV under 1.4 NA 135 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint

lens ( Supplementary Fig. 3a) through a Boundary -Element-Method platform 22. RO-136 iSCAT possess a higher lateral signal to noise ratio and narrower expansion than conventional 137 iSCAT (Supplementary Fig. 3b and 3c). To quantity the improvement on noise rejection and SNR, 138 we synthetically generated fringe by 1) the pure signal from a single particle at focal plane 139 alongside with 2) different levels of speckle noise from an out -of-focus plane (Fig. 1 e i and 140 Methods). RO-iSCA T effectively rejected most of the speckle artifacts from background (Fig. 1e 141 ii) thus improved the SNR from 0.49 (Fig. 1e ii left) significantly to 5.65 (Fig. 1e ii right). Even 142 under increasing noise conditions, RO-iSCA T consistently achieve high SNR when compared to 143

substraction (Fig. 1e iii, light blue versus light red scatters) and a ten-fold higher image 144 SNR (Fig. 1e iii, blue versus red curve. Supplementary Fig. 4). 145 146 RO-iSCAT imaging achieve speckle-free iSCAT without needing background subtraction 147 The role of background subtraction in majority of iSCAT methods 19, 20, 23-26 is to remove all 148 extrinsic factors (i.e. non-uniform illumination, and unwanted coherent noise and interference) so 149 as to increase SNR and reach single protein sensitivity. Background subtraction uses iSCAT images 150 recorded with no focal drift and ultra clean glass slides that is free of any sample feature 20. In this 151 section, we compare d SNR of RO-iSCA T versus standard background subtraction (iSCA T). The 152 sample was a glass coverslip dish containing surface -bound sub -diffraction limited gold 153 nanoparticles (37.0 nm - 43.0 nm diameter) and cancer-associated fibroblast cells (CAFs) cultured 154 over 7 days (Methods). 155 Our home -built RO-iSCA T system 18 involved a pair of galvanometer scanner to achi eve 156 customized off-axis oblique and azimuth (Fig. 2a). Two raw iSCAT image were recorded, one with 157 (Fig. 2b) and one without gold nanoparticles (Fig. 2c, as background), and formed the final image 158 iSCAT image after background is manually subtracted (Fig. 2d). Each raw image was recorded by 159 positioning the galvanometer mirrors at a single azimuthal position. On the other hand, the RO-160 iSCAT image was captured after turning the galvanometer mirror at azimuthal angles 𝜑 from 0 to 161 360o continuously. To capture a full RO-iSCA T image, the camera exposure rate was synchronised 162 to integrate over a series of oblique RO-iSCAT images over a single cycle of rotation, finally get 163 Σ𝐼scat(𝜑) (acquisition speed up to 40 fps in our system) . Here we demonstrated four RO -iSCAT 164 images at each azimuth 0°, 90°, 180° and 270° (taken without rotational integration, Fig. 2e) and 165 the full-integrated image (Fig. 2f). Considering that the final iSCA T and RO-iSCA T images were 166 of the same field of view and taken over the same exposure time, the full RO-iSCA T images 167 outputted significantly lower background noise and speckle than iSCAT background subtraction, 168 as well as an alignment with our numerical simulation (inset of Fig. 2f). 169 To quantitate the improvement of RO-iSCAT over the background subtraction in iSCAT, we 170 chose a smaller field of view (cyan dotted box in Fig. 2d and 2f) and adopted the metric to 171 determine if two closely spaced sub-diffraction limited gold nanoparticles can be resolved . From 172 the line plot of intensity variation between adjacent nanoparticles (Fig. 2g and 2h), it appeared that 173

subtraction and integration both possessed almost the same signal to noise ratio. 174 However, rotational integration was able to fully resolve adjacent 40 nm particles , where 175 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint separation between the 40 nm particles was approximately 170 nm that was close to Abbe 176 diffraction limit of 163.7 nm, which was not possible using background subtraction for the same 177 acquisition time (insets in Fig. 2h and 2g). RO -iSCAT removes of the speckle noise in the 178

without any loss to intensity. Because no subtraction is made, the full dynamic range 179 of the camera is preserved. Also we observed that there were spatially varying speckles intensities 180 in the images taken from 𝐼scat(90°), 𝐼scat(270°) to 𝐼scat(0°) and 𝐼scat(180°) profiles which were 181 uncorrelated and so were removed in RO-iSCAT Σ𝐼scat(𝜑) because of rotational integration (Fig. 182 2e). 183 To further validate our numerical model that rotational integration improves the imaging 184 resolution along the axial direction (Fig. 1e and Supplementary Fig. 3), we capture d the 185 interferometric signal using a nano-stage that was moved along a fixed step interval of 10 nm. 186 When comparing the axial intens ity profiles (Fig. 2i), the axial intensity variation of RO-iSCA T 187 shown higher contrast along the axial plane and matched well with Boundary-Element-Method 188 simulation (Supplementary Fig. 3). Then, we evaluated the imaging performance of rotational 189 integration with RO-iSCAT on membrane protrusion from adherent cells alongside with fixed 40 190 nm gold nanoparticles (Fig. 2j). While 40 nm gold nanoparticles (Fig. 2j, cyan boxed inset) were 191 marginally visible, it was only RO-iSCAT that the fine membrane protrusions can be detected (Fig. 192 2j, red boxed inset). Moreover, the rotational integration effect can be more readily observed and 193 quantified by discretizing the integration process (Supplementary Video 5) which illustrated that 194 the increasing number of azimuthal scanning angles 𝜑 for integration will form a higher final SNR 195 (Supplementary Fig. 5). 196 197 Differing spatiotemporal dynamics between membrane trails, tethers and bridges 198 Benefit from the high SNR fringes, we put our focus on the protrusion growth and external 199 connections. First, we captured a time -lapse dataset of endothelial cells with high dynamics 200 (Supplementary Fig. 6, Supplementary Video 6). On the smooth cell membrane, multiple 201 protrusion emerged in random directions, then converging toward another cell, with lamellipodia 202 driving them aggregating toward the target location, ultimately connecting with the target cell. 203 During the growth, the interference pattern on protrusion were varying, mainly the bright -dark 204 periods (yellow arrows in Supplementary Fig. 6). Here we ask if RO-iSCAT interference patterns 205 can be used to identify membrane types, particular cell -substrate versus membrane bridges 206 between cells. 207 To answer this question, we examine d RO-iSCAT images in a single culture of CAFs cells 208 that is known to form tight networks i.e. fibrotic tissue. RO-iSCAT provided the clear FOV where 209 a CAF cell adhered to glass coverslip and multiple CAF cells forming extensive membrane bridges 210 over 20-30 µm long (Fig. 3a i and ii). We then selected three different membrane protrusion to 211 characterise which were chosen based on their distinct types of interference patterns and their 212 assumed spatial locations. Combining the lateral morphology (Fig. 3b) and relative axial position 213 (Supplementary Fig. 7 and Supplementary Video 7), we formed a biological sketch (Fig. 3c) 214 and shown the membrane protrusions far from the cell body exhibit uniform fringe intensity, 215 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint indicating a flat height and membrane has adhered to the glass-bottom surface (Trail, Fig. 3b top). 216 In contrast, protrusions between cell bod ies (basal surface) display interference pattern s with 217 alternating bright and dark fringes, spaced approximately 0.5 micrometers apart. This suggested 218 that the protrusion grow s from the apical side slanting downward to the bottom surface with a 219 rapid height gradient (Tether, Fig. 3b middle). Additionally, fringes with membrane bridges tend 220 display interference pattern s with alternating bright and dark fringes, spaced approximately 3 221 micrometers apart, forming connections at a similar height between two cells (Bridge, Fig. 3b 222 bottom). 223 Besides the morphology characteristic, we examined the motion modalities of these 224 membrane protrusions from the entire time -lapse dataset by mapping from interference signal 225 intensity to relative depth based on calibration data (Sup. Fig. 9, Methods). Kymograph is a classic 226 tool for recording motion along one line over time (Supplementary Fig. 8), however, to capture 227 the spatiotemporal changes across the whole imaging field, we applied a new axial-variation map 228 (Fig. 3d i, ii, pixel-level standard deviation on the entire 2D image relative depth over the time 229 period) to measure the effective range of axial displacement ( Fig. 3d iii, iv ). Because axial 230 variation was applied across the whole imaging field, we can directly quantitate whole membrane 231 protrusion dynamics directly from the intensity mapped to the magnitude of the axial variation 232 (brighter intensity indicates larger axial fluctuations). To prove that this axial variation information 233 was only retrieved using RO -iSCAT imaging, we also applied axial variation treatment to 234 scattering-only images (Supplementary Fig. 8). It shown the distinct spatial temporal intensity 235 changes were only observed under RO-iSCAT but not under scattering-only imaging (only up to 236 87 grayscale), indicating that this was a direct consequence of interferometry ( Supplementary 237 Video 8). The axial-variation map directly determined highly motile membrane bridges between 238 CAF cells. The large range of intensity variation in RO-iSCAT images occurs in up to 300 nm z-239 axis movement along membrane bridge. This observation suggest ed that membrane bridges 240 possess a taut behavior where axial movements are greater than lateral movements. Axial variation 241 responses of the RO-iSCAT images (Fig. 3e i and ii) and the statistics of the values for all the three 242 types of membrane protrusion (Fig. 3f) indicated a clear difference between the spatial temporal 243 behavior among them. The mean values of the three grouped distributions histograms further 244 illustrated that the suspended cell -cell bridges showed a 2 -fold more axial movement (averaged 245 range of 142.60 nm) than tethers (77.01 nm) and 4-fold more than trails (31.08 nm), even though 246 they may physically appear tight and straight along. 247 248 RO-iSCAT’s performance in tracking membrane protrusion 249 CAF cells can form extensive cellular networks that will be filled with various types of 250 membrane protrusions particularly cell -cell membrane bridge 27, 28. Two separate cell co-culture 251 experiments using CAF and KPC cells were adopted (Methods) here to track the formation and 252 transition of protrusion and connections between identical and different cell types. 253 First, we examined whether the interferometric images from RO -iSCAT can track different 254 membrane protrusions more effectively than that based on intensity-only morphological tracks in 255 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint fluorescence imaging. We co-cultured two cell populations of CAF cells: wild type (WT) CAFs 256 and CAF transfected with Lck10 -GFP to label the plasma membrane (Fig. 4a i). Two separate 257 cultures of WT and Lck10-GFP transfected CAF cells were grown over 3 days before both cultures 258 were seeded into the same culture dish. The culture was imaged four hours after seeding under 259 scattering-only (Fig. 4a ii), fluorescence (Fig. 4a iii), alongside with RO-iSCAT all under oblique 260 highly inclined thin illumination (HiLO). A sequence of membrane protrusions (red dotted box in 261 Fig. 4a iii) over ~ 7.9 minutes, under HiLO fluorescence imaging (Fig. 4b) and RO-iSCA T (Fig. 262 4c), shown a transfected Lck10-GFP CAF cell protrusion its membrane towards neighboring WT-263 CAFs (not visible under fluorescence) imaging. Under RO-iSCAT imaging, we can readily observe 264 both transfected-CAF and WT-CAF beginning to form membrane bridge ~5µm length. 265 While fluorescence imaging permitted the identification of 2D morphology of the membrane 266 protrusion, in contrast under RO-iSCAT, we observed a highly complex interference scattering 267 patterns within the same protrusion. The intensity stripes changed from high to low spatial cycles 268 (all arrows in Fig. 4c i, ii compared with Fig. 4c iii, iv). At time point 349 and 584 seconds, an 269 adjacent stationary protrusion appeared to display a uniform dark stripe ( yellow arrow in Fig. 4c 270 ii and red arrow in Fig. 4c iii), suggesting the complete disassociation from the target cell 271 individually. The variation of t hese intensity fringes indicated physical axial movements of the 272 protrusion 18. Over 10 mins, these interferometric intensity patterns exhibit regular axial movement 273 when associating and disassociating with neighboring cells (Supplementary Video 9). All these 274

suggest that RO-iSCA T overcomes the limit of fluorescence modality for the purpose of 275 quantifying spatial temporal dynamics in 3D membrane protrusions. Besides these CAF-CAF 276 intercellular connections, d irect communication between CAFs and cancer cells mediated by 277 surface receptors or adhesion molecules can play an much more important role in tumour 278 progression 29. Hence, next we looked at whether CAF can form membrane bridges with pancreatic 279 ductal adenocarcinoma (PDA) cells. 280 The second co-culture involved CAF and PDA cells from the murine KPC model, where we 281 seeded the two cell groups separately at opposite side of the glass bottom dish where each will 282 migrate towards the center over 7 day -long culture in incubator ( Fig. 4d i). On day 7 and 8, we 283 imaged the proximity of the cell ( Fig. 4d ii and iii) under brightfield to observe the proximity of 284 the 2 cells population indicated by yellow and red dotted line. On day 8, we used scattering-only 285 images to identify the border between a CAF and PDA cell using cell morphology (Fig. 4e). Within 286 the chosen field of view, we monitored the space between CAF and PDA cell (Fig. 4e, blue dotted 287 box). RO-iSCAT imaging was conducted over the same field of view over 60 mins at 10 mins 288 intervals (Fig. 4f and Supplementary Video 10). Using RO-iSCAT and interferometric signals, 289 we identified active transition between cell -substrate connections to membrane bridges. T wo 290 individual connection appeared to gradually merge to one single protrusion through a twisting 291 protrusion morphology 10. We can see that a t the initial time point (0 min), the protrusion from a 292 single CAF and PDA cell were first individually separated (yellow and red dashed box). Then at 293 17 min, the protrusion gradually merged to form direct membrane bridge. Starting from 31 min, 294 the two separated membrane bridge merge through high degree of axial variation that maybe 295 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint indicate large strain akin to twisting to form a single membrane connection at 56 min with a 296 straighter morphology. 297 Whilst the interference intensity show n highly dynamics activities (Supplementary Video 298 10), it failed to reflect spatial temporal information visually. To measure the spatial temporal events, 299 we applied axial-variation treatment to the RO -iSCAT images to measure fluctuation from the 300 interference intensity signals (Fig. 4g), where higher value of axial variation indicates increasing 301 protrusion motility. It directly indicated the increasing level of axial motion along each protrusion 302 across each time point. Quantitatively, a violin plot (Fig. 4h) shows the time-variance distribution 303 along the protrusion at each time point that indicates distinguishable mean values between 304 membrane bridges forming between the PDA and CAF cells. Particularly, the spatiotemporal 305 dynamics of the protrusion from tethers to formation of bridge between cells. N ewly connected 306 tether was observed at 17 min and 31 min utes displayed increasing axial motion up to 140 nm 307 which was 2 times compared to initial protrusion tethers (0 min), and at the last 56 minutes, we 308 observed a bridge was formed between the cells indicated by 210 nm range of axial motion (3-fold 309 increase). This result indicates that RO -iSCAT interference signal s may be used to identify 310 spatiotemporal behavior of membrane bridges that is otherwise missed by classical 2D intensity-311 imaging (fluorescence or scattering-only) techniques. 312 313

314 In this study, we performed two investigations: 1) eliminate out-of-focus speckle background 315 noise in RO -iSCAT interferometric signatures u sing rotational integration. 2) preliminary 316 evidence that axial variation map of label-free RO-iSCAT images can measure and quantify spatial 317 temporal membrane protrusion types that surpass conventional kymograph. 318 319 Can RO-iSCAT operate with incoherent sources? 320 SNR of RO-iSCAT is defined by fringe visibility. Along the transverse and axial planes (x, y 321 and z), fringe visibility is generally affected by the coherence of the light source and relative 322 difference of both the optical path length and intensities between reflected and scattered intensities 323 21. Interferometric scattering (iSCAT) 19, 20, 24, 30 -32 is gaining traction for label -free sub-cellular 324 imaging due to its ability to detect nanoscale scatterers. RO-iSCA T could be treated as a partially 325 coherent detection (temporal) along with the integration time of the camera. Out-of-focus 326 interference fringes are suppressed based on the difference in path length . In supplementary, we 327 have done some modeling of partially coherent sources indicating coherence plays a minimal role. 328 We suspect that this rotational integral configuration can be of use for most interferometric 329 microscopies. 330 What are the key drawbacks of RO-iSCAT modality? 331 A common problem in interferometric imaging is the repeating interference fringes due to 332 wrapped phase. This meant that direct 3D quantitative measurement of interference fringes in the 333 axial plane becomes a challenge. Secondly, the axial iSCAT sections are taken with a moving 334

lens or the first reflective surface of a glass coverslip which can incur additional phase 335 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint shifts. This is a similar issue with common path interferometry. It will refresh background artifacts 336 from optical components, and the moving length of the objective lens will not convert to the same 337 change of interference intensity as described in the initial intensity-height relationship. 338 339 To combine with optical tweezers to study membrane tension 340 Using interferometric spatiotemporal images of RO-iSCAT and in combination with optical 341 tweezers 33, we expect to potentially quantify membrane tension of the twisting nascent filopodia 342 10 (Fig. 4). More recently, Belly et al 34 showed using optogenetics that 2D actin-driven protrusions 343 can elicit rapid global membrane tension propagation resulting in long-range membrane flows . 344 Using RO-iSCAT along with calibrated tweezers, one can study 3D axial membrane protrusion 345 when dynamics forces (optical forces) are applied to the actin cortex. 346 347 Incomplete abscission , phagocytosis and cell adhesion were quantified by measuring 348 cytoplasmic bridge properties, 3D filopodial dynamics on bacteria and nanotopologies. 349 The physical properties of membrane bridges provide insight into abscission completion 350 during the final stages of cell division 35, and 3D filopodia extensions define the distinct stages of 351 phagocytosis as immune cells clearing bacteria 4and recognition of nanotopologies that guides cell 352 migration 7. While many studies utilize confocal microscopy, its axial resolution and high 353 phototoxicity remains a key limiting factor for live axial imaging and tracking of filopodia. 354 ROiSCAT's value lies in its axial sensitivity and low phototoxicity, which operates below the 355 diffraction limit and require minimal power (~microwatts). Consequently, we anticipate that 356 ROiSCAT will be highly valuable in quantifying incomplete abscission , cell migrating on nano-357 scale surfaces and phagocytosis. 358 359

360 We showed that RO-iSCAT interference patterns generate highly distinctive spatial-temporal 361 interference intensity patterns between different cell membrane protrusion, i.e. membrane that are 362 tethered onto substrates, trails (e.g. migrasomes) and membrane bridges between adherent cells 363 across large physical gaps 27, 36 Unlike scattered only or fluorescence signals used in kymograph, 364 the spatial temporal interference patterns create d unique axial variation plots for image -based 365 classification. These characteristics were shown to be applicable across a range of adherent cell 366 types, including endothelial, CAF and PDA cells. This pilot study has highlighted the potential of 367 our method in extracting membrane specific interferometric patterns that eludes fluorescent 368 imaging. This study enables the classification of membrane protrusions that, despite possessing 369 identical chemical compositions, are differentiated by their interactions, thus offering a qualitative 370 comparison of cell-cell communication at the nanoscale in living cells 371 372 Author Contributions 373 Conceived project and directed research: W.M.L. 374 Prepared samples: Y .J.L., P.T., W.M.L., D.H., T.G.P, P.T. 375 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint Built model and programed simulations: J.L., M.G., H.L. 376 Designed and performed experiments: J.L., Y .J.L., W.M.L. 377 Examined and analyzed data: J.L., W.M.L. 378 Wrote manuscript: J.L., W.M.L., with advice from all authors. 379 Provided biological insight and advice: Y .J.L., W.M.L. 380 Supervised research: M.G., H.L., W.M.L. 381 382 Acknowledgments 383 We thank Alpha Yap (IMB, UQ) and Melanie White (IMB, UQ) on membrane protrusion and 384 filopodia discussions; and Hari Shroff (Janelia, HHMI) on the RO -iSCAT. We acknowledge the 385 Australian Research Council (DE160100843, DP190100039, and DP200100364) and NHMRC 386 (APP2000485) for their support. H.L. acknowledges funding from the National Natural Science 387 Foundation of China (62427807) and the Talent Program of Zhejiang Province (2021R51004) . 388 M.G. acknowledges funding from the National Natural Science Foundation of China (62475232). 389 J.L. acknowledges the support from Zhejiang University Global Partnership Fund. 390 391 392

393 Experiment setup 394 RO-iSCAT/Scattering data acquisition 395 Our rotating optical coherent scattering platform (ROCS) is equipped with 60  1.49NA oil 396 immersion objective lens (Olympus) and an sCMOS camera (PCO edge 4.2) for wide field 397 illumination and detection resulting in a pixel size of 20 nm and a full field -of-view over 41 398 microns  41 microns. 399 A 488 nm laser beam is directed onto a two-axis galvanometer and conjugated onto the back 400 focal plane of the objective lens to generate an oblique angle and rotational azimuth. The camera 401 performs capture under a preset framerate (up to 100 fps) and duty cycle while the incident beam 402 separately rotates at a fixed speed of 200 rounds per second. ROCS sets oblique illumination angles 403 to switch between interferometric (22 ) and scattering imaging (60 ) modes for simultaneous 404 multimodal imaging. In addition, RO -iSCA T modality requires a glass bottom culture dish or 405 coverglass for generating reference reflection light, so in the other scattering mode , the reflection 406 will be rejected by an electronic amplitude filter (diaphragm) placed at the imaging back focal 407 plane. 408 409 iSCAT data acquisition and post-processing 410 The iSCAT raw images were acquired from ROCS system by fixing the galvanometer with 411 𝜃 = 22° and 𝜙 = 0° under RO-iSCAT mode. Because ROCS platform adopts the strategy that 412 moving the objective lens or sample container to adjust the focal plane, the reference field is 413 different at different focal position s thus caused varying background images . A series of 414 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint

images were acquired from an empty glass dish or coverglass under different focal 415 positions and exposure times. 416 For isolating pure background artifacts arising from optical components, we meticulously 417 maintained the system’s condition, particularly the stage position, and replaced the petri dish with 418 another one only with PBS as a control, then captured the background image (Fig. 2c). Noted that 419 the apparent fringes in background will blur or focus and undergo overall intensity changes during 420 the stage moving along the z-axis. Thus, background subtraction necessitates the 1) pre-collection 421 of a series of background pattern s at each z position and after each biological acquisition, 2) a 422 manual selection for matched background due to the limited repeatability of the translation stage. 423 After background subtraction ( Fig. 2d), the interference image marginally excluded some 424 ambiguous artifacts, but generally, no new fringes of gold particles emerged from the background. 425 426 Calibration protocol 427 We used 40nm AU nanoparticles in a cell culture dish full of DMEM solution to illustrate the 428 sinusoidal relationship between interferometric intensity and depth gradient (Supplementary Fig. 429 9a, 9b) in RO-iSCAT. The exposure time of the camera was set to 85ms. 430 The dish was placed on a piezo nanostage that adopted axial sweeping of the sample across 431 the focus at 10 nm steps. The axial intensity map of a single nanoparticle is plotted as the orange 432 scattered plot, and the black line indicates the moving average of 8 points (Supplementary Fig. 433 9c). The cursor s represent the linear region that can be used to map from intensity (ΔI) to axial 434 displacement (Δz). 435 436 RO-iSCAT model 437 Field model from physical optics theory 438 In RO -iSCAT, particles of a sample are illuminated by incident coherent laser filed 𝐄inc 439 propagating from the objective lens and create a scattering field 𝐄scat. Meanwhile, the glass bottom 440 of petri dish or coverglass reflects part of the incident light and form a weaker reflection field 𝐄refl. 441 The reflection field 𝐄refl and scattering field 𝐄scat return to the imaging plane and jointly form the 442 interference pattern. 37, 38 443 The final fields reaching the camera are the convolution of the initial field in object space and 444 the intrinsic coherent transfer function 𝐂 of the optical system, and we get the interference pattern 445 as 446 𝐼 = |𝐄refl ⋆ 𝐂 + 𝐄scat ⋆ 𝐂|2 1) 447 We built the Cartesian coordinate system xyz where z-axis is fully aligned with the optical 448 axis of the objective lens, the coverglass and the focal plane of the objective lens are respectively 449 set as 𝑧 = 0 and 𝑧 = 𝑧𝑓. Considering that nano-scale system usually adopts high -NA design, we 450 assume that 𝐂 is a vector with an impulse response function as amplitude for simplicity, 451 correspondingly, the interference pattern can be treated as formed by the initial reflection and 452 scattering field reaching the focal plane 𝑧 = 𝑧𝑓 453 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 𝐼(𝑥, 𝑦) = 𝐸refl 2 (𝑥, 𝑦, 𝑧𝑓)+ 𝐸scat 2 (𝑥, 𝑦, 𝑧𝑓)+ 2𝐸refl(𝑥, 𝑦, 𝑧𝑓)𝐸scat(𝑥, 𝑦, 𝑧𝑓)cos[𝜙scat(𝑥, 𝑦, 𝑧𝑓)− 𝜙refl(𝑥, 𝑦, 𝑧𝑓)] 2) 454 We assumed the laser beam with constant intensity in section instead of Gaussian distribution, 455 and an incident at the oblique angle 𝜃 to the z axis and rotational azimuth 𝜑 around it. Ignoring 456 the 𝜔𝑡 item that describes how the wave evolves over time and supposing the initial phase at 457 (0,0,0) as zero 458 𝐄inc(𝑥, 𝑦, 𝑧) = 𝐸inc𝑒𝑖𝑛𝑘(𝑥 sin 𝜃 cos𝜑+𝑦 sin 𝜃 sin 𝜑+𝑧 cos𝜃) 3) 459 where 𝑛 is the refractive index of the air and 𝑘 = 2𝜋 𝜆⁄ is the vacuum wavevector. As for the 460 reflection field from oblique incidence, the amplitude and phase changes are complexly 461 determined by the oblique angle according to the Fresnel formula, so we simply noted reflective 462 index 𝜏𝜃 and phase shift 𝜙𝜃 as function of 𝜃 in reflection field 463 𝐄refl(𝑥, 𝑦, 𝑧) = 𝐸inc𝜏𝜃𝑒𝑖𝑛𝑘(𝑥 sin 𝜃 cos𝜑+𝑦 sin 𝜃 sin 𝜑−𝑧 cos𝜃)+𝜙𝜃 4) 464 We consider a single nano-scale particle with sub-wavelength size and 𝑛𝑝 density located at 465 (𝑥𝑝, 𝑦𝑝, 𝑧𝑝). The laser is scattered by the particle with the initial phase 466 𝜙scat(𝑥𝑝, 𝑦𝑝, 𝑧𝑝) = 𝑛𝑚𝑘(𝑥𝑝 sin𝜃cos𝜑 + 𝑦𝑝 sin𝜃sin𝜑 + 𝑧𝑝 cos𝜃) 5) 467 and amplitude variation to 468 𝐸scat(𝑥𝑝, 𝑦𝑝, 𝑧𝑝) = 𝐸inc(1 − 𝜏𝜃) 2√2𝜋2 (𝜆 𝑛𝑚⁄ )2 𝑎3 ( 𝑛𝑝2 − 𝑛𝑚2 𝑛𝑝 + 2𝑛𝑚2 ) √1 + cos2 𝜃scat 6) 469 in which 𝑎 is the particle radius and 𝜃scatis the scattering angle for representing anisotropic 470 scattering efficiency. By combining the initial phase and amplitude, we can achieve the scattering 471 field with spherical wavefront 472 𝐄scat(𝑥, 𝑦, 𝑧)= 𝐸scat(𝑥𝑝, 𝑦𝑝, 𝑧𝑝) 𝑟(𝑥, 𝑦, 𝑧) 𝑒𝜙scat(𝑥𝑝,𝑦𝑝,𝑧𝑝)+𝑖𝑛𝑚𝑘𝑟(𝑥,𝑦,𝑧) 7) 473 where 𝑟 is the propagation length 𝑟(𝑥, 𝑦, 𝑧) = √(𝑥 − 𝑥𝑝) 2 + (𝑦 − 𝑦𝑝) 2 + (𝑧 − 𝑧𝑝) 2 474 and cos𝜃scat = (𝑧𝑝 − 𝑧𝑓) 𝑟(𝑥, 𝑦, 𝑧)⁄ . 475 And finally, we obtain the constant, amplitude , and phase items of the interference field at 476 the focal plane 477 𝐸const(𝑥, 𝑦)= (𝐸inc(1 − 𝜏𝜃) 2√2𝜋2 (𝜆 𝑛𝑚⁄ )2 𝑎3 ( 𝑛𝑝 2 − 𝑛𝑚 2 𝑛𝑝 + 2𝑛𝑚2 ) √1 + cos2 𝜃scat 𝑟(𝑥, 𝑦, 𝑧𝑓) ) 2 + (𝐸inc𝜏𝜃)2 8) 478 𝐸intef(𝑥, 𝑦)= 2𝐸inc 2 𝜏𝜃(1 − 𝜏𝜃) 2√2𝜋2 (𝜆 𝑛𝑚⁄ )2 𝑎3 ( 𝑛𝑝2 − 𝑛𝑚2 𝑛𝑝 + 2𝑛𝑚2 ) √1 + cos2 𝜃scat 𝑟(𝑥, 𝑦, 𝑧𝑓) 9) 479 𝜙intef(𝑥, 𝑦) = 𝑛𝑘𝑚{sin𝜃[(𝑥 − 𝑥𝑝)cos𝜑 + (𝑦 − 𝑦𝑝)sin𝜑]} −𝑛𝑘𝑚{𝑟(𝑥, 𝑦, 𝑧𝑓)+ (𝑧𝑝 − 𝑧𝑓)cos𝜃} + 𝜙𝜃 10) 480 The equation of the interference phase relates to two key series of var iables, (𝜃, 𝜑) for 481 describing the incidence off principle optical axis and 𝑧𝑝 − 𝑧𝑓 for measuring the defocused length. 482 𝑧𝑓 is typically maintained as we usually fix the relative position between objective lens and 483 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint container during imaging process. iPSF can be generated by changing 𝑧𝑝 − 𝑧𝑓 and 𝑧𝑓, respectively 484 corresponding to the situation axial movement of 1) sample-only and 2) focal plane by adjusting 485 the objective lens or the whole sample container (Supplementary Fig. 6). To visualize their roles 486 in phase difference, we correspondingly split it into one off-axis and one defocused item 487 𝜙off−axis(𝑥, 𝑦)= 𝑛𝑘𝑚 sin𝜃[(𝑥 − 𝑥𝑝)cos𝜑 + (𝑦 − 𝑦𝑝)sin𝜑] 11) 488 𝜙defocus(𝑥, 𝑦)= −𝑛𝑘𝑚[𝑟(𝑥, 𝑦, 𝑧𝑓)+ (𝑧𝑝 − 𝑧𝑓)cos𝜃] 12) 489 490 Lateral shift of out-of-focus pattern 491 In simulation, we set 𝜃 to 22° as a constant while 𝜑 in the range of 0 °~360° as a variable 492 considering the configuration of our RO -iSCAT and place the phantom particle at the center of 493 FOV (0,0, 𝑧𝑝). The container position was fixed for a constant 𝑧𝑓 and we investigated the field 494 created by defocused particle at 𝑧𝑝 with ∆𝑧 = 𝑧𝑝 − 𝑧𝑓 defocused length. 495 To mathematically quantify the lateral shift of pattern, we took the partial derivative of the 496 total phase difference respectively to 𝑥, 𝑦 497 𝜕𝜙intef(𝑥, 𝑦, 𝑧𝑓) 𝜕𝑥 = 𝑛𝑘𝑚 (sin𝜃cos𝜑 − 𝑥 √𝑥2 + 𝑦2 + ∆𝑧2 ) 13) 498 𝜕𝜙intef(𝑥, 𝑦, 𝑧𝑓) 𝜕𝑦 = 𝑛𝑘𝑚 (sin𝜃sin𝜑 − 𝑦 √𝑥2 + 𝑦2 + ∆𝑧2 ) 14) 499 From which we can achieve the specific extremum point (𝑥𝑒, 𝑦𝑒, 𝑧𝑝)by setting the gradient 500 simultaneously to 0 501 𝑥𝑒 = tan𝜃cos𝜙 ∆𝑧 15) 502 𝑦𝑒 = tan𝜃sin𝜙 ∆𝑧 16) 503 and the distance of its biased to centre point is 504 𝐿 = tan𝜃∆𝑧 17) 505 506 Quantitative image quality analysis 507 Measuring the radius of lateral shifting 508 The focused position of 2 microns markers was set as zero axial position and gradually moved 509 the along z-axis from negative 2000 nm to positive 2000 nm in 10 nm step using a high-dynamics 510 Z nano-positioning stage (Physikinstrumente P-736.ZR1). 511 At each axial position, the camera got images from 12 azimuth points . The 12 centers of the 512 marker were labeled and recorded by Manual-tracking plugin in Fiji (ImageJ2 core) , then fitted 513 the circle from the shifted centers to get the radius. We use d negative and positive signs to 514 distinguish from clockwise and counterclockwise rotational shifting. 515 516 Interference fringe contrast for azimuth-sampling dataset 517 The RO -iSCAT images were used as ground truth, i.e. the pure signal, while the middle 518 outputs as the polluted overlay with noise and artifacts. During azimuth down sampling, due to 519 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint lateral shifting direction varying at different azimuth point, the asymmetric azimuth combination 520 (i.e., odd number sampling points) will retain some of this property. This results in a slightly 521 distorted morphology that does not perfectly match the ground truth. Therefore, instead of 522 subtracting the pure signal to determine the noise level, a common approach in SNR calculations, 523 we directly use the down azimuthal sampling image to measure the level of noise. 524 We respectively selected 8 rois from the empty background and signal-intensive cell regions, 525 then valued the pollution of messy noise to smooth structures by 526 𝐶𝑜𝑛𝑡𝑟𝑎𝑠𝑡𝑛 = 𝑉𝑎𝑟(𝐼𝑅𝑂−𝑖𝑆𝐶𝐴𝑇) 𝑉𝑎𝑟(𝐼𝑛) 18) 527 528 Simulation and SNR-measurement for speckle-noise 529 According to experimental observation (Supplementary Video 1), speckle noise is frame-530 uncorrelated and reserves the property of lateral shift same as physical particles. Hence, we model 531 them as random particles with varying reflectivity value at one out-of-focus plane that generate 532 unexpected fringes onto focal plane. For Fig. 1e , in a 200 ×200 pixels (4μm×4μm) FOV, we 533 uniformly set one single particle on focal plane alongside with 600 particles at 1 μm depth (speckle 534 noise from out-of-focus plane), then overlap all the fringes simulated from our model. Specifically, 535 while the reflectivity of in -focus particle is set to 1 as reference, the relative reflectivity among 536 those 600 speckles follows the Gaussian distribution but with an absolute operation to avoid 537 negative values |𝒩(0, 𝜎2)| 538 𝑓(𝑥)= | 1 √2𝜋𝜎2 𝑒− 𝑥2 2𝜎2| 539 where we use 𝜎 to measure the speckle noise level. 540 To calculate the SNR of the synthetic fringes under iSCAT and RO-iSCAT modality, we 541 selected the central region 𝐼𝑠 (5×5 pixels) from the on-focus particle fringe to calculate the variance 542 of pure si gnal. As to the variance of speckle noise, we measure the variance of the overlapped 543 fringe (on-focus particles and out-of-focus speckle) excluding the central region, 𝐼𝑛𝑜𝑖𝑠𝑒. 544 𝑆𝑁𝑅 = 𝑉𝑎𝑟(𝐼𝑠) 𝑉𝑎𝑟(𝐼𝑛𝑜𝑖𝑠𝑒) 19) 545 546 Sample preparation 547 Cell culture 548 All reagents for cell culture were sourced from Thermo Fisher Scientific (Waltham, MA, 549 USA). CAF and PDA cells (Passage 30) were maintained in T75 flasks with DMEM supplemented 550 with high glucose (4.5 g/L), 10% fetal bovine serum, L -glutamine (4 mM) and pyruvate (1 mM) 551 at 37°C and 5% CO2. Cells were split 1:20 at 80% confluence. Primary human lung microvascular 552 cells (HMVEC, Lonza) were cultured in EGM2-MV2 Bulletkit (Lonza) at 37°C and 5% CO2 and 553 split at a 1:6 ratio at 80% confluence. 554 555 Live cell imaging 556 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint Cells were first grown to 80% confluence and then 100,000 cells seeded to 29 mm glass 557 bottom dishes ( #1 coverslip, CellVis). Cells were incubated at 37°C and 5% CO2 for at least 1 558 hour prior to imaging. 559 560 CAF cell transfection 561 For fluorescent labeling of the CAF cell membrane, we used an Lck10 -GFP plasmid 562 (generous gift from the late Katharina Gaus). The Lck10-GFP consists of the first 10 amino acids 563 of the membrane protein Lck and eGFP linked to the C -terminus. Plasmids were propagated in 564 E.coli in LB Medium and purified using a miniprep kit (Genejet, Thermo Scientific). C AF cells 565 were transfected with polyethyleneimine (40 kDA, linear, Polysciences Inc) using 9 ug PEI and 3 566 µg DNA per 29 mm dish over 48 hours prior to imaging. 567 568 Nanoparticles 569 Gold nanoparticles ( 40 nm ) were diluted 1000 times in distilled water . 1 mL of diluted 570 nanoparticles were then added to a dry 29 mm glass bottom dishes (#1 coverslip thickness, CellVis) 571 and allowed to dry for 1 hour at room temperature. Prior to imaging, 1 mL of 1X PBS was carefully 572 pipetted along the walls of the dish. The dish was then mounted onto a Z nano-positioning stage 573 (Physikinstrumente P-736.ZR1) for imaging. 574 575 CAF cell with nanoparticle 576 40 nm gold nanoparticles were dried onto a glass bottom dishes (#1 coverslip thickness, 577 CellVis) for 1 hour as described above. The dried particles were then immersed in high glucose 578 DMEM prewarmed to 37°C. 1 mL of CAF cells (100,000 cells/mL) was then added dropwise 579 onto the dish and incubated for 1 hour at 37°C and 5% CO2 and then mounted onto a heated 580 stage for imaging. 581 582 CAF cell co-culture (WT and Lck10) 583 We seeded 100,000 cells into two separate 29 mm glass bottom dishes (#1 coverslip, CellVis). 584 After 24 hours, one dish was transfected with Lck10 -GFP plasmid and incubated for another 48 585 hours. Cells in WT and transfected dishes were then detatched with trypsin (0.25% (w/v)) and 586 EDTA (1 mM) and 50,000 cells each seeded and mixed into a new 29 mm glass bottom dish. Cells 587 were incubated for 4 hours at 37°C and 5% CO2 prior to imaging. 588 589 CAF and PDA co-culture 590 To create separate CAF and PDA cell populations on a single dish, cells were concentrated to 591 1 million cells/mL and 100 μL were pipetted to s eparate corners of a 29 mm glass bottom dish. 592 Cells were incubated for 1 hour 37°C and 5% CO2 to allow the cells to attach to the glass dish. 593 Cell attachment was monitored under a bright field microscope. Cells were then supplementing 594 with 1 mL of DMEM and incubated for 7 days, with media changed every 2 days. 595 596 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 597 Data availability 598 The data that support the findings of this study are included in Figs. 1–4, Supplementary Figs. 599 1–10 and Supplementary Videos 1 –10. All experimental data from figures ( Figs. 2 , 3, 600 Supplementary Fig s. 5-9) are publicly available at https://doi.org/10.5281/zenodo.14960905. 601 Other time-lapse co-culture datasets (Fig. 4) are available from the corresponding author W.M.L 602 upon request due to their large file size. 603 604 605 Code availability 606 All numerical modelling and analysis were achieved using Python 3.11.0. Generation of iPSF by 607 Boundary-Element-Method was performed in MA TLAB (Mathworks, R2022b). Customized RO-608 iSCAT model and analysis codes are available at https://github.com/ejunyuliu/RO-iSCA T. Initial 609 numerical iSCAT model was installed from https://github.com/manoharan-lab/applied-optics-610 iscat-code. iSCAT software based on Boundary-Element-Method platform was downloaded from 611 https://pubs.acs.org/doi/suppl/10.1021/acsphotonics.4c00621/suppl_file/ph4c00621_si_001.zip. 612 613 614 615 616 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint

617 1. Bruce, A. Molecular biology of the cell. (Second edition. New York : Garland Pub., [1989] 618 ©1989, 1989). 619 2. Jaqaman, K. & Grinstein, S. Regulation from within: the cytoskeleton in transmembrane 620 signaling. Trends in Cell Biology 22, 515-526 (2012). 621 3. Wood, W. & Martin, P. Structures in focus —filopodia. The International Journal of 622 Biochemistry & Cell Biology 34, 726-730 (2002). 623 4. Möller, J., Lühmann, T., Chabria, M., Hall, H. & V ogel, V . Macrophages lift off surface-624 bound bacteria using a filopodium-lamellipodium hook-and-shovel mechanism. Scientific 625 Reports 3, 2884 (2013). 626 5. Kress, H. et al. Filopodia act as phagocytic tentacles and pull with discrete steps and a load-627 dependent velocity. Proceedings of the National Academy of Sciences 104, 11633-11638 628 (2007). 629 6. Bischoff, M.C., Lieb, S., Renkawitz -Pohl, R. & Bogdan, S. Filopodia -based contact 630 stimulation of cell migration drives tissue morphogenesis. Nature Communications 12, 791 631 (2021). 632 7. Albuschies, J. & V ogel, V . The role of filopodia in the recognition of nanotopographies. 633 Scientific Reports 3, 1658 (2013). 634 8. Mogilner, A. & Rubinstein, B. The Physics of Filopodial Protrusion. Biophysical Journal 635 89, 782-795 (2005). 636 9. Sherer, N.M. et al. Retroviruses can establish filopodial bridges for efficient cell -to-cell 637 transmission. Nature Cell Biology 9, 310-315 (2007). 638 10. Leijnse, N. et al. Filopodia rotate and coil by actively generating twist in their actin shaft. 639 Nature Communications 13, 1636 (2022). 640 11. Jiang, D., He, J. & Yu, L. The migrasome, an organelle for cell–cell communication. Trends 641 in Cell Biology (2024). 642 12. Hu, H.T. et al. The cellular protrusions for inter -cellular material transfer: similarities 643 between filopodia, cytonemes, tunneling nanotubes, viruses, and extracellular vesicles. 644 Frontiers in Cell and Developmental Biology 12 (2024). 645 13. Sanders, T.A., Llagostera, E. & Barna, M. Specialized filopodia direct long-range transport 646 of SHH during vertebrate tissue patterning. Nature 497, 628-632 (2013). 647 14. Gan, Y . et al. Enhanced axial resolution of lattice light sheet microscopy by fluorescence 648 differential detection. Opt. Express 30, 27381-27394 (2022). 649 15. Urbančič, V . et al. Filopodyan: An open -source pipeline for the analysis of filopodia. 650 Journal of Cell Biology 216, 3405-3422 (2017). 651 16. Reiche, M.A. et al. When light meets biology – how the specimen affects quantitative 652 microscopy. Journal of Cell Science 135 (2022). 653 17. Abounit, S., Delage, E. & Zurzolo, C. Identification and Characterization of Tunneling 654 Nanotubes for Intercellular Trafficking. Current Protocols in Cell Biology 67, 12.10.11-655 12.10.21 (2015). 656 18. Zheng, Y . et al. Combined Scattering, Interferometric, and Fluorescence Oblique 657 Illumination for Live Cell Nanoscale Imaging. ACS Photonics 9, 3876-3887 (2022). 658 19. Küppers, M., Albrecht, D., Kashkanova, A.D., Lühr, J. & Sandoghdar, V . Confocal 659 interferometric scattering microscopy reveals 3D nanoscopic structure and dynamics in 660 live cells. Nature Communications 14, 1962 (2023). 661 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 20. Ortega Arroyo, J., Cole, D. & Kukura, P . Interferometric scattering microscopy and its 662 combination with single -molecule fluorescence imaging. Nature Protocols 11, 617 -633 663 (2016). 664 21. Biswas, T., García Díaz, M. & Winter, A. Interferometric visibility and coherence. 665 Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 473, 666 20170170 (2017). 667 22. Hitzelhammer, F. et al. Unified Simulation Platform for Interference Microscopy. ACS 668 Photonics 11, 2745-2756 (2024). 669 23. Hsiao, Y .-T., Tsai, C.-N., Chen, T. -H. & Hsieh, C. -L. Label -Free Dynamic Imaging of 670 Chromatin in Live Cell Nuclei by High-Speed Scattering-Based Interference Microscopy. 671 ACS Nano 16, 2774-2788 (2022). 672 24. Cole, D., Young, G., Weigel, A., Sebesta, A. & Kukura, P. Label -Free Single-Molecule 673 Imaging with Numerical -Aperture-Shaped Interferometric Scattering Microscopy. ACS 674 Photonics 4, 211-216 (2017). 675 25. Park, J.-S. et al. Label-free and live cell imaging by interferometric scattering microscopy. 676 Chemical Science 9, 2690-2697 (2018). 677 26. Kukura, P. et al. High-speed nanoscopic tracking of the position and orientation of a single 678 virus. Nature Methods 6, 923-927 (2009). 679 27. Zani, B.G. & Edelman, E.R. Cellular bridges. Communicative & Integrative Biology 3, 680 215-220 (2010). 681 28. Goliwas, K.F. et al. Mitochondrial transfer from cancer -associated fibroblasts increases 682 migration in aggressive breast cancer. Journal of Cell Science 136 (2023). 683 29. Dhungel, N. & Dragoi, A.-M. Exploring the multifaceted role of direct interaction between 684 cancer cells and fibroblasts in cancer progression. Frontiers in Molecular Biosciences 11 685 (2024). 686 30. Hecht, E. Optics. (Pearson Education, Incorporated, 2017). 687 31. Dahmardeh, M., Mirzaalian Dastjerdi, H., Mazal, H., Köstler, H. & Sandoghdar, V . Self -688 supervised machine learning pushes the sensitivity limit in label -free detection of single 689 proteins below 10 kDa. Nature Methods 20, 442-447 (2023). 690 32. Sheppard, C.J.R., Cogswell, C.J. & Gu, M. Signal strength and noise in confocal 691 microscopy: Factors influencing selection of an optimum detector aperture. Scanning 13, 692 233-240 (1991). 693 33. Pompeu, P. et al. Protocol to measure the membrane tension and bending modulus of cells 694 using optical tweezers and scanning electron microscopy. STAR Protocols 2, 100283 (2021). 695 34. De Belly, H. et al. Cell protrusions and contractions generate long-range membrane tension 696 propagation. Cell 186, 3049-3061.e3015 (2023). 697 35. Chaigne, A. & Brunet, T. Incomplete abscission and cytoplasmic bridges in the evolution 698 of eukaryotic multicellularity. Current Biology 32, R385-R397 (2022). 699 36. Korenkova, O., Pepe, A. & Zurzolo, C. Fine intercellular connections in development: 700 TNTs, cytonemes, or intercellular bridges? Cell Stress 4, 30-43 (2020). 701 37. de Wit, X.M. et al. Precise characterization of nanometer -scale systems using 702 interferometric scattering microscopy and Bayesian analysis. Appl. Opt. 62, 7205 -7215 703 (2023). 704 38. Gholami Mahmoodabadi, R. et al. Point spread function in interferometric scattering 705 microscopy (iSCA T). Part I: aberrations in defocusing and axial localization. Opt. Express 706 28, 25969-25988 (2020). 707 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 708 709 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 710 Fig. 1 Modelling and simulation of R O-iSCAT. a) Schematic diagram of incident, reflection, 711 scattering fields, and several reference planes in azimuthal iSCA T. b) Numerical simulation with 712 14 microns FOV and 70 nm step based on the modelling. Including i, off-axis oblique phase and 713 ii, oblique convolved with defocused phase, respectively at 0 /90120 with a 10 μm 714 defocused length and 22 oblique angle. iii, Profiles of off-axis oblique, defocused, and total phase 715 difference along the horizontal cent ral axis of FOV. iv, Interference pattern at corresponding 716 azimuth angle. c) The radius of lateral shifting under defocused length ranging from -2.5 μm to 717 2.5 μm in ~290 nm step extracted from simulation and experimental data. The experiments were 718 repeated five times independently, as indicated by error bars (mean +/- SD). d) i, Circumferential 719 lateral-shifting related to illumination azimuth and defocused length . ii, Azimuthally integrated 720 interference pattern. Also attached the corresponding intensity profile along the horizontal central 721 axis of FOV. e) i, Sketch presentation to show the strategy of simulating different speckle noise 722 level. One single particle at in-focus plane as signal source, while a series of speckles are placed 723 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint at 10-micron depth with their reflectivity value following the Gaussian distribution . ii, In-focus 724 fringes from iSCAT versus RO-iSCAT based on the object in i. Inserted ground truths are pure 725 interference fringe only from the in-focus particle. iii, Noise variance and fringe SNR curve with 726 varying reflectivity levels of speckle sources, both in iSCAT versus RO-iSCAT. Scale bars: b), d) 727 3 μm, e) 1 μm. 728 729 730 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 731 Fig. 2 Reduction of speckle background through rotational integration . a) Diagrammatic 732 sketch of conventional iSCAT and RO-iSCAT imaging method. Bottom right is the sketch map of 733 orthogonal galvos and the transition from flipping at the galvo to lateral circling at the back focal 734 plane, and finally to illumination with specific tilting angle and varying azimuth emitted from the 735 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint

lens. b) Raw iSCAT image of 40nm gold particles with 500 ms exposure time. Gold 736 particles were dried thus adhered to the inner glass bottom, then PBS for 1:20 dilution to the initial 737 storage liquid of particles. c) Manually selected background pattern from PBS -only control dish 738 under the same exposure time. d) Final iSCA T image after post -processing of background 739 subtraction. The d ashed rectangle region highlights the blurring pattern in iSCAT but well -740 distinguished by RO -iSCAT. e) RO-iSCA T images from different incoming azimuths without 741 integration. f) Final RO -iSCA T image with time -integrating during rotational scanning. g) 742 Intensity profile of raw image, background and the final result after subtraction, along the dashed 743 line in magnified image cropped from d), the orange and background scatters are labeled by left y 744 axis while the blue curve is by right y axis. h) Intensity profile of image at azimuth position of 0, 745 90, 180, 270 and final integrated result along the f) region that corresponds to the dashed line, 746 the blue curve is by right y axis while the scatters in other colours are labelled by left y axis. i) Z 747 section and the corresponding profile along z axis from interference PSF captured from 40 nm 748 particles, individually without rotational integration and with rotational integration configuration. 749 j) Without rotational integration (top) against with rotational integration (bottom) modality of CAF 750 cells seeded with 40 nm particles. Scale bars: b)-d), f) 250 nm, e) 500 nm, g)-i) 400 nm, j) 5 μm. 751 752 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 753 Fig. 3 Identifying types of membrane trails and connections with RO -iSCAT. a) i and ii are 754 raw RO-iSCAT of CAF cell membrane trails and connections from RO-iSCAT captured at 5 fps 755 (50% duty cycle) over 10 seconds. b), Magnified images of the cyan rectangle regions in a). Top, 756 cell membrane trails observed directly on substrates . Middle, cell membrane tethers mixed with 757 membrane trails on substrates. Bottom, direct cell-cell tethers without any membrane trails on the 758 surface. c) Proposed concept of biological diagrammatic sketch of neighbouring cells cultured on 759 glass bottom dish, in which each main membrane protrusion type will create different scattering 760 field. d) Procedure of calculating axial-variation map. i, RO-iSCAT fringe along one single 761 protrusion at 0 s, 3 s and 6 s time points. ii, Axial distribution of this protrusion at several time 762 points. Profiles were fitted from raw curve only for better presentation here (8th degree polynomial 763 with R2=0.94, 0.96, 0.92, 0.82 separately). iii, Standard deviation measuring the effective axial 764 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint displacement at each pixel. iv. Pixel-level standard deviation on the entire 2D image relative depth 765 over the time period . e) Calculated axial-variation maps of respective raw RO -iSCAT images in 766 a). f) Histograms counted from the trails and connections regions in b), with mean values noted 767 for membrane trail, tether and bridge groups. Scale bars: a), e) 3 μm, b), d) 500 nm. 768 769 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint 770 Fig. 4 Tracking and quantifying protrusion between trails and connections with RO-iSCAT. 771 a) i, Experiment time sequence of co-culturing Lck10-GFP transfected and non -transfected WT 772 CAF cells. On day 1, WT CAF cells were seeded into 2 dishes and only one of them was transfected 773 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint by Lck10-GFP dye. On day 3, we transferred the transfected cells into the WT CAF cell dish . 4 774 hours later, the mixed dish was uploaded for imaging. ii, Scattering-only image captured at 4th hour 775 on day 3 containing only one transfected cell (marked by light blue circle) . iii , Magnified 776 fluorescence image of the Lck10 -GFP transfected cell in FOV (dashed light blue circle in ii. b) 777 Time-lapse fluorescent images of the red rectangle region in a). c) The comparative RO -iSCAT 778 images where the two different arrows consistently track the same protrusion across different time 779 points. d) i, Experiment time sequence of PDA and CAF cells co-culture. On day 1, we seeded 780 CAF and PDA cells separately at opposite side of one dish. During day 1 to 8, each will migrate 781 towards the center over 7 day-long culture in incubator. On day 8, imaging was performed where 782 the two cell populations intersected . ii, iii, Bright field images under stereo microscope of PDA-783 CAF co-culture dish at day 7 and 8 after seeding. e) Scattering-only image of a FOV containing 784 one PDA and CAF cell. f) Snapshots in time -lapse stack of the cyan rectangle region in e), 785 individually displaying the filopodia interacting, connecting, merging and the final merged 786 dynamics. g) Axial variation map of f) counting from 50 frames (5 fps) around each time point. h) 787 Violin plot of axial variation distribution along membrane protrusion in each time point. Cyan and 788 red lines represent median and quartiles, respectively. Counted protrusions have been pointed out 789 by yellow arrows. Scale bars, a), e), 5 μm, b), f) 1 μm. 790 791 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 25, 2025. ; https://doi.org/10.1101/2025.03.23.644841doi: bioRxiv preprint