Observing concurrent subcellular dynamics in large living tissues

30 An outstanding question in eukaryotic biology is the mechanistic connection between events 31 occurring at (sub)cellular levels (time scales of milliseconds to minutes) to those at the tissue levels 32 (tens of minutes to months). Deciphering such mechanisms requires imaging approaches capable 33 of simultaneously achieving high spatial and temporal resolutions for large samples over long 34 periods of time. Here, we demonstrate Airy beam -based light sheet microscopy of organelles in 35 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint tens to hundreds of cells in a few hundred micrometre -wide tissue environments. We achieve a 36 typical resolution of 320 nm over 266 × 266 × 100 μm 3 volumes at a temporal rate of 0.05 Hz, 37 typically with generally used fluorophores such as Green Fluorescent Protein, over extended 38 periods of time that allow tracking of organelle and protein dynamics. We validated our approach 39 across different length and time scales by imaging mitochondria and endosome dynamics in very 40 large fields of view in zebrafish tissue, molecular assemblies of myosin as gastrulation proceeds in 41 Drosophila embryos, 3D mitochondrial streaming in mouse oocytes, pressure -driven motility and 42 protrusions in amoebae, mitochondrial dynamics in cancer spheroids, 5 -colour fast imaging in 43 iBlastoids, and endosomal dynamics in single cells. Through these model systems, we demonstrate 44 the versatility of Airy beam light sheet microscopy to image large tissues at unprecedented high 45 resolution; to capture dynamics in photosensitive, delicate samples; and to screen 3D samples. We 46 anticipate that our Airy beam-based approach will represent a pivotal advance in cellular biology—47 especially developmental biology—as it provides, for the first time, true subcellular resolution over 48 large imaging volumes with high temporal resolution. 49

50 Multiscale measurements are critical to addressing the fundamental question of how molecular and 51 cellular events give rise to emergent tissue-level behaviours. This problem is particularly evident in 52 the context of animal development, which involves processes that span the extremes of biological 53 length and time scales and integrate genetic, biochemical, and mechanical information ( 1). Gene 54 regulation, intra - and intercellular signalling, organellar dynamics, and cell shape changes and 55 movements (taking place over milliseconds to minutes) drive tissue development and sculpting 56 (occurring over tens of minutes to months), which in turn feeds back to modulate lower -level 57 components. This interplay between processes occurring at multiple, hierarchical levels of 58 organisation—linking transcriptional programs to mechanistic execution through signalling and 59 mechanochemical pathways, resulting in macroscopic morphogenetic processes —is thus crucial 60 for coordinating spatial events and generating the temporal patterns required for robust 61 development. 62 However, it is also one of the least explored facets of developmental biology, due largely to 63 measurement challenges. Connecting stochastic components at the molecular and organellar levels 64 (small length scales, fast time scales) to emergent behaviours at higher levels of organisation (large 65 length scales, slow time scales) requires an identification of the cross-scale interactions of patterns 66 and processes that is only obtainable by observing the relevant dynamics within living organisms. 67 Thus, a major bottleneck in understanding mechanisms that operate across multiple levels of 68 organisation has been the ability to make simultaneous measurements across the corresponding 69 spatial and temporal scales. 70 Fluorescence microscopy has yielded tremendous advances in our understanding of biological 71 processes whose details may be captured within the spatial and temporal windows compatible with 72 specific imaging modalities. At the molecular level, specialised approaches, including super -73 resolution microscopy, have proven highly effective in elucidating the dynamics and organisation 74 of interacting molecules within living cells ( 2). At the subcellular level, various imaging 75 technologies have advanced our understanding of biochemical processes that regulate organelle 76 dynamics and their transport, organisation, and regulation (3, 4). At the levels of cells and tissues, 77 in toto imaging approaches typically leverage light sheet -based modalities, which illuminate and 78 image thin planes of light scanned rapidly through a 3D sample. These studies have enabled 79 reconstruction of cell lineages, primarily within the model organisms Drosophila (5), zebrafish (6), 80 mouse embryo (7, 8), arthropod limbs (9) and body-wide circuits of cellular activities and pulsating 81 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint waves of calcium across tissues ( 10). However, the bridge between molecular interactions, 82 organelle dynamics within single cells, and morphogenesis within tissues, remains poorly explored. 83 For example, a challenging question in the field is the connection between rapid intracellular 84 organelle dynamics, such as endosomes and mitochondria, within the context of developing tissue 85 spanning many cells. Light sheet-based approaches can capture fast organelle dynamics at whole -86 cell volumes for single cells ( 3), and more recently, adaptive optics -based lattice light -sheet 87 microscopy (LLSM) has been used to observe organelle dynamics in living tissues and organoids 88 (11). However, given the very limited field of view (FoV), a large volume can be achieved only by 89 tiling multiple sub volumes, which significantly reduces temporal resolution and requires stitching. 90 Within light sheet-based approaches, the size of the lateral FoV is largely determined by the choice 91 of beam. Motivated by the need to study fast developmental processes, including mapping 92 organelle dynamics and morphological cellular and morphogenetic tissue changes simultaneously, 93 and the lack of a technique capable of imaging at the requisite spatial and temporal scales, we 94 explored beams that could offer the versatility and uniform excitation across the FoV needed to 95 permit tiling-free capture of large volumes. 96 Light-sheet approaches using Bessel beams, which have cylindrically symmetric profiles, have 97 successfully imaged subcellular dynamics in living cells ( 12), both using single beams as well as 98 multiple interfering beams as in the case of LLSM ( 13). Like Bessel beams, Airy beams are non -99 diffracting and enable optical sectioning; however, their asymmetric excitation pattern can result 100 in enhanced contrast (14). While low numerical aperture (NA) systems using Airy beams have been 101 demonstrated for large tissues at low resolution ( 14), imaging at subcellular resolution that 102 combines Airy beams with high -NA optics in an optimised fashion to span large FOVs with 103 expanded spatiotemporal coverage in large living tissues has yet to be shown. Key characteristics 104 and benchmarks of existing light -sheet approaches are summarised in Supplementary table T1. 105 Here, we report a light-sheet microscope based on versatile Airy beams that are optimised for the 106 NA of the excitation objective and the FoV relative to the camera chip size, in combination with 107 a high -NA water immersion collection objective. This larger, optimised FoV translates to 108 enhanced temporal resolution (speed) for large tissues while maximising the spatial resolution 109 required to capture organelles. Consequently, our approach achieves 0.05 Hz temporal resolution 110 over 266 × 266 × 100 μm 3 at 320 nm resolution. This is a significant improvement compared to 111 previous imaging solutions (Supplementary table T1). To demonstrate the versatility of this 112 approach for biomedical imaging, we imaged mitochondria and endosomes in zebrafish embryos 113 and larvae, mid-gut invagination in Drosophila embryos, mitochondrial streaming in mouse oocytes, 114 amoeba motility, and mitochondrial and cytoskeletal dynamics in organoids. We also show that 115 rapid imaging allows high-throughput screening in iBlastoids, and that shorter Airy beams can be 116 used to measure organelle dynamics in cells. Through these examples, we demonstrate the 117 capability of Airy beams in enlarging the available spatiotemporal resolution regime, which opens 118 the door to investigate previously inaccessible biological questions requiring ‘across -the-scale’ 119 imaging and measurements, including mapping protein movements, molecular assembly dynamics, 120 organelle dynamics, and cell motilities across large volumes of living tissues. 121

122 Versatility of Airy beams to balance field of view versus resolution 123 To realise an Airy beam light -sheet microscope system, we utilised the previously established 124 geometry of high -NA objective pairings used for Bessel and lattice light -sheet microscopy. We 125 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint utilised a 20× 0.6 NA Thorlabs objective to deliver excitation light and a 25× 1.1 NA Nikon 126

or a 20× 1.0 NA Olympus (Evident) objective for collection of emission light, providing 127 a 266 × 266 μm2 or a 332 × 332 μm 2 FoV, respectively, for a single plane. We used a spatial light 128 modulator (SLM) at the Fourier plane to the sample plane to modulate the incoming beam with a 129 cubic phase, resulting in an Airy beam after the excitation objective (Materials and methods, 130 Supplementary fig. 1). By applying different scale factors, the non-diffractive propagation distance 131 can be modulated to fill the FoV of the camera to distinct extents. We scan this beam across the 132 entire FoV using a galvo mirror. There is no scanning along the direction of propagation of light 133 after the detection objective, but modulation of the Airy beam to different lengths using the SLM 134 offers a choice between a smaller FoV with a higher resolution, or a larger FoV with a lower 135 resolution. Scanning across multiple planes can be performed in three ways (Fig. 1a). First, an ‘XZ-136 scan’ can be performed where the sample is moved along the principal axis of the collection 137 objective, which is achieved by mounting the stage motor at the same angle as the collection 138

with respect to the base of the optical table. This approach abrogates any need for post-139 acquisition deskew processes. Second, the stage can be moved parallel to the optical table in an 140 ‘X-scan’, permitting imaging of larger areas of the sample by moving them laterally, but requiring 141 deskew. Third, a galvo -electronic tuneable lens (ETL) is used, where the sample is not moved at 142 all, but the beam is scanned across using a galvo mirror, and the focus of the collection objective 143 is shifted in synchrony using the ETL. This approach does not require deskew, but the range of 144 motion is limited compared to the XZ -scan. 145 To characterise the resolution and FoV capabilities of our system, we used 100 nm TetraSpeck 146 beads that produce diffraction -limited images. Different Airy beams cover the FoV to different 147 extents (Fig. 1c–e, Supplementary fig. 2a) and produce distinct PSFs (Fig. 1b, Supplementary fig. 148 2b). As is evident from the PSFs, the compromise for the FoV is the axial resolution: raw images 149 acquired using Airy beams have a characteristic blur in the axial direction, necessitating 150 deconvolution to achieve maximal axial resolution. Here, all images displayed have been 151 deconvolved using PSFs calculated from diffraction -limited images of 100 nm TetraSpeck beads 152 for their respective wavelengths and Airy beams. We then imaged phalloidin -stained zebrafish 153 tissue spanning 266 × 266 × 320 μm3 (Fig. 1f–n). As expected, with longer Airy beams, resolution 154 improved at full volume -in-view (ViV). The XZ -scan, in combination with an appropriate Airy 155 beam length, was also able to capture organelle dynamics in single cells. We next captured 156 endosomal motility in single cells; dual -channel images were simultaneously acquired from HeLa 157 cells expressing Rab-GFP and Rab7-mCherry, spanning 103 × 302 × 25 μm3 and imaged over 17 158 min at 5.7 s per volume (Supplementary fig. 3, Supplementary movie 1), demonstrating the 159 versatility of our approach from large tissues to single cells. 160 We estimated resolution using Fourier ring correlation ( 15) in both single cells with JFX650 -161 labelled mitochondria (Fig. 2a), as well as phalloidin -stained zebrafish embryos (Fig. 2 b,c). While 162 the effective resolution was dependent on the local structure and signal-to-noise, it was on average 163 ~320 nm and did not degrade with depth for zebrafish, a transparent sample. In both cases, for 164 single cells and tissues, we demonstrate that increasing the beam length resulted in improvement 165 of the ViV, albeit with compromised highest achievable spatial resolution. More importantly, while 166 comparisons on resolutions can be made on many static parameters calculated from inanimate 167 samples such as fluorescent beads, we show in the next sections that live imaging of biological 168 processes could be performed, the metric most pertinent to the biological community ( 16). See 169 Supplementary table T2 for a summary of all biological samples, imaging parameters, and 170 benchmarking metrics. 171 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint High spatiotemporal, large -volume, dynamic measurements in zebrafish 172 To demonstrate our approach in measuring dynamic events across the entire volume of imaging 173 (VoI) in living tissues, we imaged GFP-labelled mitochondria and mCherry-labelled endoplasmic 174 reticulum (ER) in the skeletal muscle of transgenic Tg(actc1b:mito-GFP)uom407Tg and Tg(actc1b:ER-175 mCherry)uom408Tg double-positive zebrafish larvae at 3 days post fertilisation (dpf). Maximising the 176 illumination across the entire FoV, and using the XZ-scan, we could image a volume of 266 × 266 177 × 40 μm 3 with a time resolution of 19 s per volume or at a rate of 0.05 Hz for two colours (Fig. 178 3a, Supplementary movie 2). The uniform resolution across the entire volume allows ensemble 179 quantification by precise segmentation (Fig. 3b, Supplementary movie 3) and analysis of the 180 orientation of mitochondria and ER in individual myotomes (Fig. 3c). Furthermore, the high time 181 resolution enabled visualisation of mitochondrial dynamics at multiple locations at the same time 182 (green, magenta, and cyan highlighted regions, Fig. 3d; Supplementary movie 4) revealing fusion 183 and fission events (Fig. 3e; colours correspond to regions highlighted in Fig. 3d). The consistent 184 resolution across the volume also enables morphometric analysis of nuclei or plasma membranes 185 with high accuracy (Supplementary fig. 4,5). 186 To demonstrate concurrent, continuous tracking across a large volume at high temporal resolution, 187 we imaged mKate2 -rab5ab-positive endosomes across 266 × 266 × 60 μm 3 in the tailbud tissue 188 of a 14 -somite stage zebrafish embryo expressing mKate2 -rab5ab with a time resolution of 5.9 s 189 per volume (Fig. 4a –d; Supplementary movies 5,6). Using a custom code for detecting and 190 volumetric tracking of endosomes ( 17), we were able to capture endosomal trajectories within 191 individual cells across the VoI and characterise the motility of endosomes both within and between 192 cells of a large cross -section of developing tissue (Fig. 4e). We also imaged a zebrafish embryo 193 expressing mKate2-rab5ab and membrane-mNeonGreen to visualise both endosomal motility as 194 well as cell boundaries (Fig. 4f –h). 195 196 Cross-scale mapping of macromolecular myosin assemblies in Drosophila 197 development 198 A classical model organism in developmental biology is Drosophila . A key step in Drosophila 199 embryogenesis is cellularisation, which involves the conversion of a single -celled syncytium to a 200 multicellular embryo. Following cellularisation of the embryo, zygotic transcription ensues and 201 distinct cellular behaviours are observable. Within 5 minutes, through the mid -blastula transition, 202 the cephalic furrow and the ventral furrow form, as well as the cellular blastoderm, which expresses 203 a mitotic pattern controlled by string expression that begins at the precephalic region. Gene 204 expression patterns orchestrate cellular force generation based on actomyosin contractility that 205 drives epithelial sculpting. In parallel, cell divisions also contribute to elongation and macroscopic 206 behaviours of the embryo tissue. The divisions begin at the procephalic region and proceed in a 207 successive manner across the embryo (18). Drosophila non-muscle myosin II regulatory light chain 208 (encoded by the spaghetti squash gene, sqh) plays a key role in cellularisation, furrow ingression, and 209 basal closure; it localises to the cleavage furrows during anaphase in dividing cells and is required 210 for cytokinesis ( 19). Following sqh-3x-GFP in Drosophila embryo allows mapping both divisions, 211 as it localises to cytokinetic rings, as well as contractile machinery assemblies that drive tissue-wide 212 movements and folds. We imaged Drosophila embryos expressing sqh-3x-GFP from the stage of 213 syncytial blastoderm through to gastrulation. To demonstrate how our high -resolution imaging 214 approach enables rapid biological processes to be observed across a large area, we focussed on the 215 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint posterior end of the embryo. In this region, dramatic tissue elongation takes place while 216 simultaneously traversing the posterior fold over itself. This region has been difficult to capture in 217 3D at high resolution owing to the fast nature of movements, with the first phase of elongation 218 lasting ~25 minutes. 219 As the sqh-3x-GFP signal disappears post-cellularisation, it reappears at cell -cell junctions before 220 the onset of posterior midgut invagination. By imaging sqh-3x-GFP beginning with the movement 221 of primordial germ cells through the posterior midgut (PMG) invagination for 60 min at one 222 volume every 25 s, we could follow changes in cell boundaries, contractile assemblies, and 223 cytokinetic rings (Fig. 5a–c, Supplementary movies 7,8). The ensemble directions of cell divisions 224 could be mapped by following the orientation of the division rings (Fig. 5d, Supplementary fig. 6). 225 Visualisation of the orientation of the cell divisions revealed complex patterns: at a depth of 27 226 μm from the ventral surface, we find divisions that are oriented along the anterior -posterior axis 227 (Fig. 5f), while at 35 μm, cell divisions were symmetrically placed to the midline and oriented along 228 the dorsal-ventral axis (Fig. 5g, Supplementary movie 9). It is interesting to note that the wave of 229 divisions occurs soon after the rapid phase of germ band elongation (GBE) and are programmed 230 to take place in a specific temporal sequence that is constant between embryos ( 18). Consistent 231 with this, we also observe patterns of divisions occurring on the posterior -ventral side of the 232 embryo, as the rapid phase of GBE concludes (Supplementary movie 10). Furthermore, with just 233 two imaging volumes, we captured the entire Drosophila embryo (Supplementary movie 11) at high 234 spatial resolution and with a time resolution of 66 seconds. Therefore, simultaneously occurring 235 events that occur across the entire embryo can be mapped. For example, as precellularisation 236 concludes, sqh -3x-GFP signals immediately localise to the future ventral furrow and execute 237 furrow formation; at the same time, the cells in the precephalic region divide, as visualised by the 238 formation of cytokinetic rings (Supplementary fig. 7). Waves of division in the precephalic region, 239 which occur concurrently with formation of cell boundaries after ventral furrow formation, can 240 also be mapped (Supplementary movie 12). By imaging an entire Drosophila embryo, typically 241 400–500 μm in length and 150 –200 μm in width, we can follow sqh -3x-GFP through 242 cellularisation, the appearance of the ventral and cephalic furrows, germline extension, and the 243 formation of segments, allowing us to capture the evolution of molecular assemblies of sqh -3x-244 GFP that execute force generation (macroscopic, intercellular) and cell division (intracellular, 245 localising in the cytokinetic ring). 246 Organelle dynamics in photosensitive mouse oocytes 247 To demonstrate our approach on a challenging highly photosensitive sample, we imaged 248 mitochondrial streaming in mouse oocytes. In mouse oocytes, maturation is concomitant with 249 cytoplasmic reorganisation. The meiosis II (MII) stage oocytes show a distinct accumulation of 250 mitochondria in the spindle hemisphere, which displays a characteristic streaming (20-22) in which 251 the mitochondria move towards the spindle from the centre of the oocyte. Upon reaching the 252 spindle, they move away from it, along the cortex. This flow pattern is halted at the equator, 253 distinguishing the spindle hemisphere from the non-spindle hemisphere. The flow patterns occur 254 throughout the 3D volume of the oocyte and are difficult to capture since oocytes are extremely 255 sensitive to the light used for excitation, which can inhibit the streaming phenomenon. Therefore, 256 only a few planes had previously been imaged to capture the prominent parts of the streaming 257 patterns. Here, due to the photo-gentle nature of the light-sheet geometry, combined with the long 258 non-diffractive length of the Airy beam that can traverse the diameter of the oocytes (typically 80 259 μm), we were able to capture the movements of mitochondria using mito -Dendra2 ( 23) 260 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint throughout the 3D volume of the oocyte for up to 4 hours at a temporal resolution of 5 minutes 261 (Fig. 6a, Supplementary movie 13). This enabled mapping of complex patterns of mitochondrial 262 streaming within the oocytes (Supplementary fig. 8). This is illustrated by streamlines 263 corresponding to 3D flow at distinct points in time, with sustained flows reaching speeds of ~0.5 264 μm/min within the first 2 hours, which slow down then lose cross -oocyte coherence by 4 hours 265 (Fig. 6b). 266 Fast, large-scale motility of amoebae 267 To capture another example of very large, dynamic single cells, we imaged Amoeba proteus. A. proteus 268 cells form wide, thick pseudopods, referred to as lobopods, that enable amoeboid movement. 269 These protrusions are thought to be driven by the creation of intracellular pressure arising from 270 contractile actomyosin systems; at the site of elongation, the actin cytoskeletal structure collapses, 271 causing cytoplasm to flow into the weakened path ( 24). A single A. proteus cell typically extends 272 250–750 μm, undergoes constant changes in cell shape, and exhibits an average crawling speed of 273 4 μm/s (25). Owing to its extremely large size and rapid dynamics, capturing the entire cell body 274 with a single-volume acquisition is challenging. Taking advantage of the Airy beam’s ability to span 275 the entire FoV (332 μm by 332 μm for a 20× 1.0 NA detection objective) and using large step 276 sizes (1 μm between XZ planes), we were able to image a volume enclosing the entire cell body in 277 6.2 s (Fig. 6c, Supplementary fig. 9, Supplementary movie 14) and thereby capture the rapid 278 protrusion formation and dynamics. We also observed two lobopodial extensions that fused at 279 their distal ends (Fig. 6d, green arrow), resulting in a closed -loop morphology and a lumen within 280 the surface of the amoeba. 281 Organelle dynamics in cancer organoids 282 To demonstrate live imaging in multicellular assemblies, we imaged patient -derived colorectal 283 (CRC) organoids embedded in Matrigel ( 26) (Fig. 7a,b). A longer Airy beam could provide 284 subcellular resolution on single -cell thick hollow organoids but was limited in penetrating solid, 285 non-hollow organoids as the beam quality was compromised by scattering. In the context of 286 cancer, mitochondria are involved in multiple cellular processes that regulate tumour development, 287 including metabolic reprogramming and metastasis. Mitochondrial localisation and morphologies 288 are associated with metastatic potential (27). We captured mitochondrial dynamics including fusion 289 and fission in individual cells, highlighting the ability to observe contemporaneous organelle 290 dynamics across any sub -region within the organoid (Fig. 7c). Combined with screening 291 approaches, using such volumetric imaging to monitor organelle dynamics has the potential to 292 identify patterns that result in spontaneous metastasis. 293 Rapid screening of iBlastoids 294 Blastocysts, the multicellular structures that develop into early embryos, can be modelled in vitro 295 using iBlastoids ( 28). Spatial cell type profiling and localisation to extract patterns are necessary 296 steps to characterise iBlastoids, typically accomplished by immunostaining analysis using markers 297 for distinct cell types including GATA3 (trophectoderm), Nanog (epiblast), and GATA6 (primitive 298 endoderm). iBlastoids are typically 100 μm in diameter and require imaging at sufficient sectioning 299 and resolution to capture all cells, with multiple channels to distinguish individual cell type markers, 300 the nucleus, and cell boundaries. Previously, confocal microscopy was used to acquire such 301 volumes, which is time -consuming and thus prohibitive of high -throughput screening. Here, we 302 could capture each iBlastoid in five different channels in ~10 min, at a step size of 200 nm between 303 each slice. This enabled us to visualise the detailed distribution of transcription factors (Fig. 7d, 304 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Supplementary movie 15) in iBlastoids and to use a morphological feature—the presence of inner 305 cell mass outside a fitted sphere of the iBlastoid —as a parameter for quality assessment (Fig. 7e, 306 f). 307

308 Light-sheet fluorescence microscopy enables imaging of cells and tissues across a wide range of 309 length scales. However, current methodologies require a trade -off between volume of imaging, 310 spatial resolution, and temporal resolution, where only two of any of these factors may be 311 optimised within a single acquisition. Thus, a key challenge is to achieve simultaneous imaging 312 from near -diffraction-limited structures up to tissue -level features, whilst retaining sufficient 313 temporal resolution to capture subcellular dynamics. Here, we report the first use of Airy beam 314 light-sheet microscopy in a high-NA configuration to capture large volumes of biological samples 315 without significant loss of spatial or temporal resolution. To demonstrate the broad utility of this 316 approach, we present applications across a range of systems and processes. 317 First, the ability to capture rapid dynamics at high resolution over a large volume is beneficial for 318 the study of biological processes coordinated across scales, such as animal development. We 319 illustrate simultaneous capture of organelle dynamics from large tissue samples from zebrafish 320 embryos and larvae, as well as coordination of cellular division events with rapid, large -scale 321 changes in tissue morphology in Drosophila embryos. Second, large 3D samples that require 322 exceptionally photo-gentle imaging, such as oocytes, benefit from the ability to capture complete 323 volumes without the need for tiling. Third, this technique permits the capture of extremely rapid 324 events occurring at or above the cellular scale, as illustrated by movies of pressure -driven 325 morphological changes or stochastic lobopodial extensions across the entire surface of an amoeba. 326 Finally, this approach is suitable for high -throughput screening of large, multicolour samples 327 including cell lines, cancer organoids, and iBlastoids, which is often time-prohibitive using standard 328 methodologies. 329 One limitation of Airy beam imaging is the highly asymmetric PSF, which necessitates that all data 330 be deconvolved prior to visualisation and analysis. We offset this disadvantage through the design 331 of an XZ-scan that does not require the additional deskew step typical for similar setups. We also 332 perform deconvolution with PSFs obtained for specific beams on a high-performance computing 333 cluster to minimise the resultant time delay. Another limitation is that single photon excitation 334 restricts imaging in deeper parts of samples that are highly scattering, such as solid cancer 335 organoids. This can be overcome using multiphoton excitation, which has been demonstrated with 336 Airy beam light -sheet microscopes ( 29, 30 ), albeit with low -NA objective systems. In addition, 337 while the use of Airy beams offers extended FoVs, it does limit depth-dependent correction using 338 adaptive optics as has been demonstrated for lattice light -sheet microscopy (11). However, using 339 shorter Airy beams provides the versatility to implement similar approaches. The most significant 340 post-acquisition challenge lies in the size of the datasets generated, which can reach >1 TB for a 341 single volume. As such, adequate computational infrastructure must be in place to transfer, store, 342 visualise, and analyse these very large multidimensional datasets. New approaches such as 343 hierarchical image visualisation are needed to efficiently load images and allow users to interact in 344 3D with datasets where a single volume can easily exceed available memory. PetaKit5D, for 345 example, is a tool that mitigates many such issues associated with large datasets ( 31). This is 346 required both to qualitatively interpret the phenomena captured, as well as to guide rigorous 347 quantitative analysis. 348 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint The ability to image complex biology across a wide range of spatial and temporal scales provides 349 a significant advancement toward deciphering emergent phenomena that span multiple levels of 350 organisation. Instead of building a description of dynamic events from datasets acquired using 351 distinct modalities for different samples under varying conditions, Airy beam light -sheet 352 microscopy permits simultaneous acquisition of processes ranging from organelle dynamics to 353 tissue-level rearrangements within a single live biological specimen. This opens new opportunities 354 to link mechanisms operating at the molecular and subcellular levels to concomitant higher -level 355 processes, such as active growth and patterning in developing tissues. Large imaging volumes also 356 directly entail acquisition of large numbers of individual events operating at small, rapid scales, 357 such as protein movements and organelle interactions, which is needed to establish mechanism in 358 highly stochastic processes. This approach carries significant potential, especially in the field of 359 developmental biology, to accelerate efforts to understand how biochemical events drive robust 360 processes such as cell differentiation and tissue patterning ( 32-34). 361

362 Optics Configuration 363 Light-sheet imaging was performed using a custom -designed Aurora Airy Beam Light -Sheet 364 Imaging System (build outsourced to M-Squared). This upright light -sheet microscope utilises an 365 asymmetric orthogonal high-NA objective configuration with a 20× 0.6 NA (Thorlabs) excitation 366

and a 25× 1.1 NA (Nikon) or 20× 1.0 NA (Olympus) detection objective. Six laser lines 367 with wavelengths in the visible spectrum (405, 445, 488, 561, 647 and 685 nm, respectively) are 368 available for excitation. The Airy beam is generated by applying a cubic phase mask on the 369 collimated beam (9 mm diameter) using a reflective spatial light modulator (SLM, Meadowlark 370 MSP1920 400 -800). The SLM modulates the wave front at the back aperture with a cubic 371 polynomial function: 𝛼(𝑢𝑦3 + 𝑢𝑧3)𝜆, where 𝑢𝑦 and 𝑢𝑧 are the normalised Cartesian pupil 372 coordinates aligned with the y- and z-axes respectively, and λ is the excitation wavelength. The 373 dimensionless parameter α dictates the propagation invariance of the Airy beam. The beam is 374 demagnified by a factor of 0.375, and the conjugate SLM plane is projected onto the galvanometer 375 mirror. The beam is scanned using the mirror corresponding to the x-direction, located after the 376 excitation objective, to create a scanned light sheet. The beam is then magnified by a factor of 377 1.125 compared to the SLM. After Fourier transform, the propagation length is measured with 378 respect to the camera field of view (FoV) to define the Airy beams. For example, Airy 5 379 corresponds to 5% of 266 µm, which is the length of the side of the camera FoV for a 25× 1.1 380 NA objective. A second galvanometer mirror is used to position the beam in the plane of the focus 381 of the collection objective. In the detection path a dichroic mirror is used to split emission light of 382 different wavelengths to two sCMOS cameras (Hamamatsu Orca Flash 4.0 V3) enabling sequential 383 multi- or simultaneous dual-channel acquisitions. All parameters of imaging of various samples are 384 summarised in Supplementary table T2. 385 Zebrafish Husbandry 386 Zebrafish (Danio rerio ) husbandry and breeding was conducted in the AquaCore facility at Monash 387 University according to standard procedures (34). All experimental procedures were approved by 388 the Monash University Animal Ethics Committee under ethics approval numbers ERM22161 and 389 ERM41803. Embryos were reared at 28 °C or 22 °C in E3 media (5 mM NaCl, 0.17 mM KCl, 0.33 390 mM CaCl2, 0.33 mM MgSO4) until they reached the appropriate developmental stage. Larvae older 391 than 24 hours post fertilisation (hpf) were treated with 75 µM PTU (1 -phenyl-2-thiourea) to 392 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint prevent pigmentation. Ethyl -m-aminobenzoate methanesulfonate (Tricaine) was used to 393 anaesthetise embryos and larvae (0.168 mg/mL) and adult fish (0.3 mg/mL), where necessary. 394 Zebrafish wild type strains Tuebingen (TU) and AB as well as the transgenic line Tg(actc1b:mCherry-395 CAAX)pc22Tg (35) were used in this project. To maintain genetic diversity, transgenic lines were 396 outcrossed to wild type strains every second generation. 397 Generation of plasmids and transgenic strains 398 Transgenic constructs were assembled using the multisite gateway cloning kit ( 36). The muscle -399 specific mito-GFP construct (actc1b:mito-GFP) was generated as per (37). The muscle-specific ER-400 mCherry construct (actc1b:ER-mCherry) was generated using p5E-actc1b (36, 37), pME-mCherry-401 ER-3, which was subcloned from a plasmid gifted by Michael Davidson (Addgene plasmid # 402 55041; http://n2t.net/addgene:55041; RRID:Addgene_55041), p3E -pA and pDEST-Tol2-pA2 403 (36). Plasmids were injected at 30 ng/μL into 1 -cell-stage embryos along with transposase RNA 404 (25 ng/μL) that was synthesized from the pcs2FA -transposase vector using the mMessage 405 machine Sp6 kit (Ambion, AM1340). The final transgenic lines created were: Tg(actc1b:mito-406 GFP)uom407Tg and Tg(actc1b:ER-mCherry)uom408Tg. 407 mRNA Microinjections 408 Prior to injection, plates containing 3% agarose imprinted with grooves were prepared by placing 409 a Tu-1 microinjection mould into liquid agarose. Microinjection needles were prepared using a P-410 2000 micropipette puller to pull needles from glass capillaries (1 mm outer diameter, 0.78 mm 411 inner diameter). A standard microinjection apparatus was used to inject 1 nL of 50 –100 ng/µL 412 capped mRNA in nuclease free water, supplemented with 10% phenol red as injection guide, into 413 zebrafish embryos at the single -cell stage. The relevant mRNA was prepared from linearised 414 plasmid DNA containing a SP6 promoter site by in -vitro transcription using the mMessage 415 mMachine SP6 Transcription Kit (Invitrogen). The plasmid for Membrane -mNeonGreen was a 416 gift from Amro Hamdoun (Addgene plasmid # 198057; http://n2t.net/addgene:198057; 417 RRID:Addgene_198057). The PC2+ -mKate2-rab5ab was generated by GenScript by cloning the 418 mKate2-rab5ab sequence from a reference plasmid into the PC2+ backbone. The actc1b-mKate2-419 rab5ab reference plasmid was a gift from Rob Parton (Addgene plasmid # 109649; 420 http://n2t.net/addgene:109649; RRID:Addgene_109649). 421 Zebrafish Embryo and Larva Preparation for Live Imaging 422 Before mounting, zebrafish embryos/larvae were manually dechorionated and anaesthetised using 423 168 mg/L Tricaine in methylene blue-free E3, where necessary. All further steps were carried out 424 using methylene blue -free E3 media. 425 The embryos and larvae were mounted in a volcano-shaped mount placed inside the microscope’s 426 imaging dish. To produce a volcano -shaped mount, 1.2% low melting point agarose in E3 was 427 shaped using a mould (based on ( 11)). Live anaesthetised embryos/larvae were transferred to 428 melted 0.8% low melting point agarose in E3 at 42 °C and then transferred into the volcano mount, 429 adjusting their orientation before the agarose solidified. After the agarose solidified, the imaging 430 dish was filled with E3 media containing 168 mg/L Tricaine for imaging. 431 Drosophila 432 D. melanogaster were raised at room temperature (22 –23 °C) or 18 °C on food made with yeast, 433 glucose, agar and polenta. Animals were fed in excess food availability to ensure that nutritional 434 availability was not limiting. All experiments were carried out at 25 °C. Males and females were 435 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint used for all experiments. Mounting was performed on 5 mm coverslips coated with glue extract 436 prepared by leaving double-side sticky tape in hexane overnight. The following genotype was used: 437 sqh-3x-GFP. 438 Drosophila crosses were established at 25°C on apple juice plates with yeast paste in embryo 439 collection cages. Embryos of the appropriate stage were collected from apple juice plates using a 440 wet paintbrush, washed in an embryo strainer with deionised water and manually dechorionated 441 using Dumont no. 5 forceps on double sided tape. Dechorionated embryos were adhered to the 442 imaging sample holder using folded double -sided tape to create an angled surface, allowing 443 appropriate orientation of embryos for image acquisition. Embryos were submerged in deionised 444 water for the duration of the imaging session. 445 Mouse Oocytes 446 All animal experiments in this study were approved by the Monash University Animal Ethics 447 Committee and conducted in accordance with the Australian National Health and Medical 448 Research Council (NHMRC) Guidelines on Ethics in Animal Experimentation. 449 7-week-old PhAM (photo -activatable mitochondria) mice ( 23) were superovulated by 450 intraperitoneal injection of 5 IU of pregnant mare’s serum gonadotropin (Prospec) followed 44 –451 48 h later by intraperitoneal injection of 5 IU of human chorionic gonadotropin (hCG) (MSD 452 Animal Health). 12 –13 h after hCG injection, oviductal cumulus masses were released into pre -453 warmed M2 medium (Sigma -Aldrich) supplemented with 300 µg/mL hyaluronidase (Sigma -454 Aldrich) to remove cumulus cells. Oocytes displaying a first polar body (indicating metaphase II 455 arrest) were washed and transferred to drops of M2 medium under mineral oil. 456 These oocytes were microinjected with mRNA using an electrophysiology -based picopump 457 (PV820, World Precision Instruments) and a micromanipulator (MMN -1, Narishige). Following 458 microinjection, oocytes were incubated in M2 medium under mineral oil (RT, 10 min) before being 459 transferred to a heat block (37 °C, 10 min) to facilitate oocyte retrieval. Oocytes were cultured in 460 M2 medium for at least 3 h for transgene expression. 461 Amoeba Sample Preparation and Mounting 462 Amoebae were cultured on 5 mm glass coverslips submerged in protist culture medium (Southern 463 Biological) supplemented with five grains of rice. The cultures were maintained in the dark at room 464 temperature for 5–7 days prior to imaging. For live imaging, coverslips were mounted onto a raised 465 platform sample holder inside the imaging dish, which was filled with protist culture medium 466 containing the lipophilic dye Fast DiI Solid (Thermo Fisher) at a final concentration of 2.5 µM. 467 Imaging commenced 10 min post-staining without subsequent media replacement. 468 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Cancer organoids 469 Patient-derived colorectal cancer (CRC) organoids were established as previously ( 26) described 470 and in accordance with the Declaration of Helsinki, and the protocol was approved by the Cabrini 471 Research Governance Office (CRGO04 -19-01-15) and the Monash Human Research Ethics 472 Committee (MHREC ID 2518). TUBB::TagGFP2 CRC organoids lines were then generated using 473 CRISPR-HOT(39) with CRISPaint gene tagging Kit (Addgene #1000000086) (38) and an sgRNA 474 (5’-gaggccgaagaggaggccta -3’) plasmid (Genscript). 10 days post-passage organoids were mixed at a 475 1:1 (v/v) ratio with TrypLE-passaged organoids and resuspended in Matrigel (Corning) containing 476 1:50,000 fluorescent beads. A 10 µL droplet of the suspension was seeded onto a UV -sterile 477 Parafilm on ice and immediately covered with a 5 mm round coverslip. After the Matrigel solidified 478 for 10 min in 37 ºC, the coverslip with the Matrigel disc was carefully lifted from the Parafilm and 479 placed into a 6 -well tissue culture plate (Nunc) with the Matrigel layer facing upward. 2 mL of 480 phenol-red-reduced complete CRC organoids culture medium ( 26) supplemented with 10 μM Y-481 27632 dihydrochloride kinase inhibitor (Tocris Bioscience) were added to each well. The following 482 day, coverslips were incubated with PKmitoDeepRed 10 nM for 20 min. For imaging the coverslip 483 was mounted on a raised platform sample holder inside the imaging dish, which was filled with 484 pre-warmed culture media. 485 iBlastoids 486 iBlastoids were generated according to established protocols (28). iBlastoids were collected into a 487 protein low binding tube under a dissecting microscope. After washing once with PBS by 488 centrifuging for 1 min at 10–20 × g, iBlastoids were fixed in 4% PFA for 40 min, washed with 489 PBS and permeabilised with 0.1% Triton X -100 (Sigma) in PBS for 20 min, then blocked with 490 10% donkey serum (Thermo Fisher). Primary antibodies used were rabbit anti -NANOG 491 polyclonal (1:100, Abcam, ab21624), mouse anti -GATA3 (1:100, BD Biosciences, 558686) and 492 goat anti-GATA6 (R&D AF1700). Primary antibody incubation was conducted overnight at 4 °C 493 on shakers followed by incubation with secondary antibodies (donkey anti rabbit 488, donkey anti 494 mouse 555, donkey anti goat 647, 1:500, Thermo Fisher). After labelling, iBlastoids were stained 495 with 4′,6-diamidino-2-phenylindole, dihydrochloride (DAPI) (1:1000, Thermo Fisher) for 10 min. 496 iBlastoids were stained with phalloidin (A22286, Thermo Fisher) for 1 h before imaging. Stained 497 iBlastoids were transferred into an FEP tube (FT0.8X1.0 FEP UTW, Adtech). Tubes were sealed 498 with grease and mounted in the imaging chamber, which was filled with PBS. 499 Cell Culture and Mounting 500 HeLa Rab5-GFP Rab7-mCherry and RPE1 ER -StayGold HaLo-Mito cells were incubated at 37 501 °C in 5% CO 2 in high glucose Dulbecco’s modified Eagle’s medium (DMEM) (Life 502 Technologies), supplemented with 10% foetal bovine serum (FBS) and 1% penicillin and 503 streptomycin (Life Technologies). Cells were seeded at a density of 200,000 cells per well 504 in a six-well plate containing 5 mm glass coverslips. RPE1 ER-StayGold HaLo-Mito cells 505 were incubated in 50nM JFX650 for 30 min, followed by a PBS wash and media 506 replacement prior to imaging. For imaging the coverslip was mounted on a raised platform 507 sample holder inside the imaging dish. All cells were imaged in phenol red -free DMEM 508 heated to 37 °C. 509 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Resolution Maps 510 Resolution maps were calculated for each plane of an image volume using single image Fourier 511 ring correlation ( 15). Images corresponding to single cells consisted of a thin strip near the glass 512 coverslip, scanned across the width of the FoV (using an acquisition mode where each variation 513 in depth corresponds to lateral movement through the sample). The location of glass–cell interface 514 was identified by segmentation, and at each image depth, a resolution map was generated by 515 calculating the resolution within subregions of 256 × 256 pixels using a rolling window over a 516 column spanning 256 pixels in width that covers the cells without incorporating blank regions far 517 from the glass interface. The full resolution map was collected by aggregating each local value of 518 the resolution at each depth (corresponding in this case to specific positions along the X axis 519 within the final image volume). Resolution maps for zebrafish tissue were calculated by tiling 520 subregions spanning 256 × 256 pixels (i.e., 64 discrete subregions at each depth). Resolution maps 521 were calculated using the original 1FRC MATLAB implementation (15), supplemented by custom 522 MATLAB code for tiling and rolling window analyses. For visualisation, final resolution maps were 523 smoothed using a Gaussian filter and grayscale data were assigned RGB values corresponding to 524 the local value of the resolution using custom Python code, prior to visualisation using napari (39). 525 Endosome Tracking and Analysis 526 Endosomes were detected and tracked as previously reported ( 17) using custom Python code. 527 Briefly, a Laplacian of Gaussian filter was used to detect individual endosomes within each frame, 528 and complete trajectories were constructed using Trackpy. To analyse endosomal motility, a rolling 529 window mean-squared displacement (MSD) was conducted using custom Python code. Briefly, all 530 trajectories were dedrifted, then the MSD was calculated for segments of each trajectory spanning 531 at least 10 frames (60 s), and the anomalous diffusion exponent ( α) extracted for all segments for 532 which the lag time and MSD fit well to a line on a log–log plot. The resulting time-dependent value 533 of α was then smoothed using a Savitzky -Golay filter. 534 Oocyte Flow Maps 535 An image stack containing a single oocyte was first dedrifted by using phase cross -correlation to 536 account for translational drift due to temperature fluctuations across the movie, as well as 537 histogram equalisation to account for variations in image intensity over time. The optical flow was 538 then calculated for each pair of subsequent frames using the iterative Lucas -Kanade (iLK) 539 algorithm as implemented in scikit -image. To smooth out fluctuations, a time -average flow map 540 was constructed by averaging four contiguous 3D flow maps (corresponding to 60 min of 541 imaging), prior to calculating 3D streamlines for visualisation. All analyses were conducted using 542 custom Python code. 543 Amoeba Visualisation 544 Raw volumetric image data of DiI -stained amoebae underwent initial pre -processing using a 545 custom semi-automatic pipeline in Fiji to remove coverslip signal. Segmentation was achieved by 546 training a custom machine learning-based pixel classifier in Labkit. Downstream analysis employed 547 custom scripts (MATLAB 2023b) for post -processing the segmentation output, specifically to 548 isolate the largest object and thus exclude smaller, non -target microorganisms present in the 549 culture. Volumetric rendering of the segmented amoebae was performed using the 3DScript plugin 550 in Fiji. Finally, custom MATLAB scripts were used to temporally colour code the segmented 551 amoeba volumes for visualisation. Final visualisations were generated by projecting the temporally 552 coloured volumes onto raw data extracted regions. 553 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Zebrafish Myotome Analysis 554 Segmentation masks for myotomes and nuclei were generated in MATLAB utilizing the Medical 555 Imaging Toolbox Interface for Cellpose Library support package ( 40). Prior to segmentation, the 556 volumetric dataset of the endoplasmic reticulum (ER) channel was scaled down using a custom 557 MATLAB code such that the final mean equivalent diameter of the object of interest was 50 pixels. 558 Nuclei segmentation employed a custom Cellpose model trained using ten image slices, evenly 559 distributed throughout the volume, that were manually annotated in Labkit. Nuclear boundaries 560 were specifically defined by annotating the signal -devoid spaces within the myotomes. 561 Segmentation of mitochondria and ER was achieved using a machine learning-based pixel classifier 562 trained in Labkit for each channel respectively. The resulting confidence maps were subsequently 563 post-processed via a custom MATLAB script to threshold the signal and remove false positives 564 arising from noise, based on object size criteria. For ER and mitochondrial orientation analysis, 565 the segmentation output of the tissue’s centre slice was skeletonised. Orientations of individual 566 branches were computed using the Image Processing Toolbox in MATLAB, and the orientation 567 distribution was displayed using representative semi -polar plots generated by custom -written 568 MATLAB code. 569 iBlastoid Nuclei Extraction and Analysis 570 iBlastoid volumes were first pre -processed with channel registration using Fiji plugin Fast4DReg 571 to ensure precise signal localisation of all channels. Nuclei were segmented using a custom-trained 572 model built upon the Cellpose -SAM model (40). The model training involved manual annotation 573 on randomly generated XY, YZ, and XZ slices. Segmentation outputs were validated for accuracy 574 by visualising overlays onto the raw data using syGlass with an Oculus VR headset. Inner and 575 outer cell masses were manually annotated in VR. To quantify nuclei distribution of the inner cell 576 mass, custom MATLAB code was written. Least squares fitting was utilised to generate the best -577 fit sphere which encapsulated the outer cell mass. The proportion of inner cell mass cells inside 578 and outside of this fitted sphere was then quantified. 579 Drosophila Embryo Division Orientation Analysis 580 Drosophila embryo volumes were segmented using a machine learning-based pixel classifier trained 581 in Labkit. Division rings at relevant timepoints were manually identified and annotated using 582 syGlass and an Oculus VR headset. The resulting division ring masks were exported and processed 583 using custom MATLAB code. The ring’s orientation was calculated by applying Principal 584 Component Analysis (PCA) to the mask voxels where each division ring’s orientation was defined 585 by the normal vector (N), corresponding to the third principal component. The anterior-posterior 586 (AP) axis endpoints of the embryo were manually defined using VR, which was then used to 587 establish an orthogonal 3D coordinate system relative to the embryonic axes: anterior -posterior 588 (VAP), lateral ( VL), and dorsal-ventral (VDV). N vectors extracted across several timepoints were 589 mapped onto this coordinate system and colour-coded using a continuous Cyan, Magenta, Yellow 590 (CMY) colormap. The final colour of a vector was derived from the normalized absolute dot 591 product of the individual N vector with the VAP, VL, and VDV axes, respectively. This determined 592 the weighting of the respective colours, with Cyan indicating complete AP alignment, Magenta 593 indicating complete L alignment, and Yellow indicating complete DV alignment. A single 594 representative timepoint was extracted and segmented using Labkit to finally generate a surface 595 rendered visualization in MATLAB, which was overlaid onto the same 3D coordinate system as 596 the orientation vectors. 597 NVIDIA IndeX Visualisation 598 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint NVIDIA IndeX (https://developer.nvidia.com/index ) was deployed to visualise movies of 0.5–3 599 TB in real time on 4× H100 Nvidia GPUs, Intel 72 Core CPU, and 1 TB RAM with 1 PB network 600 storage. 601

602 S.A. is supported by The EMBL Australia Partnership Laboratory (EMBL Australia) under the 603 National Collaborative Research Infrastructure Strategy of the Australian Government. The 604 authors thank Monash BDI Advanced Bioimaging. The Australian Regenerative Medicine Institute 605 is supported by grants from the State Government of Victoria and the Australian Government. 606 S.A. acknowledges NVIDIA for sharing NVIDIA IndeX to visualise our large datasets. The 607 computational analysis and visualisation with NVIDIA IndeX were supported by Monash 608 eResearch capabilities, including M3 High Performance Computing. S.A. acknowledges Wellcome 609 Trust Team Science Grant. K.F.H. was supported by grants from the NHMRC (APP1194467) 610 and ARC (DP230101406 and DP 250103072). A.A.R. was supported by a grant from ARC 611 (DP240102721), P.D.C. was supported by grants from the NHMRC (GNT2016338) and ARC 612 (DP240101647 and DP240102156). J.K. is funded by NHMRC Ideas grants GNT2037953, ARC 613 Discovery Project Grant DP210103501, ARMI Accelerator Fellowship, Research Council of 614 Finland, and the Sigrid Juselius Foundation and Biocenter Finland. I -W.L. and J.C. are funded by 615 the ARC DP160104892 and NHMRC 1165627 and 200112. This work was additionally supported 616 by NHMRC project grants APP1104560 and APP2004627 to J.M.P. The authors acknowledge 617 Prof Paul McMurrick and the colorectal surgeons at the Cabrini Monash Department of Surgery 618 for their contributions to specimen collection. S.A. would like to thank Srigokul Upadhyayula 619 (University of California, Berkeley) for discussions and advice on data handling and visualisation. 620 Author contributions 621 Project supervision: SA 622 Biological reagents, and sample preparation: CSW, SU, LZK, HRG, AP, HMY, SS, SAM,GS, I-WL, 623 WHC, EB, SH, SC, HEA, JK, PC, KFH, JMP, JC 624 Analysis software: CSW, AP, SU, SA 625 Formal analysis: CSW, SU, SA 626 Data visualisation, figure and movie preparation: SA, AP, SU, CSW 627 Manuscript preparation: SA, CSW with input from all coauthors 628 629 630 631 632 633 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figures 634 635 636 637 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figure 1. Airy beam light -sheet microscope acquisition modes and beam types. (a) 638 Schematic illustrating main acquisition types. The XZ -scan moves along the same axis as the 639 detection objective and is suitable for most samples. The X-scan moves along the same axis as the 640 optical table and is suitable for scans through wide but relatively thinner samples. (b) 641 Experimentally measured PSFs of Airy beams corresponding to phase masks of increasing strength 642 (Airy 5, 30, and 90, respectively). Top: XY projections. Bottom: YZ projections. Images were 643 obtained with 561 nm excitation wavelength and 100 nm diameter TetraSpeck beads. (c–e) Top: 644 Experimentally measured beam profiles corresponding to the same phase masks as in b. Bottom: 645 Section through a volume of 100 nm diameter TetraSpeck beads illustrating horizontal variation 646 in intensity and resolution accompanying different Airy beams as in b. (f–n) Alexa Fluor 647 -647 labelled phalloidin in zebrafish tissue measured with the same Airy beams as in b. (f–h)The sample 648 spans 266 × 266 × 320 µm3. (i–k) Each block of insets shows medium zoom cross-section views 649 taken from regions near the centre ( left), representing the dorsal tail tissue, and the edge ( right), 650 representing the yolk tissue, of the corresponding images in f–h. A further zoom of the marked 651 regions in i–k is displayed in l–n. Scale bars: (c–h): 100 μm, (i–k): 25 μm, (l–n): 5 μm. 652 653 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 654 655 656 657 658 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figure 2. Resolution maps calculated from live samples. (a) Scan through a thin sample of 659 confluent HeLa cells with JFX650 -labelled mitochondria illustrating the variation in lateral 660 resolution as a function of Airy beam strength across a wide (332 μm) region. The image has been 661 colour-coded according to the calculated resolution. Rows are arranged from top to bottom in 662 order of increasing Airy beam strength. Images were acquired with a 20× 1.0 NA collection 663 objective. (b–c) Section of a thick (266 × 266 × 320 µm3) volume of fixed zebrafish tissue labelled 664 with phalloidin-Alexa Fluor 647 illustrating the trade-offs between maximum achievable resolution 665 in the centre of the FoV (lower beam strength) and maximal effective resolution across the full 666 FoV (higher beam strength). Images were acquired with a 25× 1.1 NA collection objective. (b) 667 Resolution maps, with columns arranged from left to right in order of increasing Airy beam 668 strength. (c) Section along the length of the Airy beam, where images have been colour -coded 669 according to the calculated resolution. Note that the difference in the observed limits of the 670 resolution between a and b–c is largely due to the collection objective. Scale bars: (a,c): 50 μm. 671 672 673 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 674 675 676 677 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 678 Figure 3. Organelle dynamics in large volumes of zebrafish tissue. (a) Raw images of skeletal 679 muscle tissue of a 3 dpf transgenic Tg(actc1b:mito-GFP)uom407Tg and Tg(actc1b:ER-mCherry)uom408Tg 680 double positive larva at progressively higher zooms, showcasing different scales of organisation 681 from the whole tissue to individual myotomes and mitochondria. (b) Schematic of imaged region 682 of zebrafish larva ( left). Corresponding 3D segmentations of the regions displayed in a, depicting 683 the whole tissue, individual myotomes, and mitochondria. (c) Cross-sectional slice extracted from 684 the middle of the volume of a representative tissue section, displaying both the endoplasmic 685 reticulum (ER) and mitochondrial channels, alongside their respective segmentations. Two 36 μm 686 × 36 μm zoomed -in regions of interest (ROIs), indicated by blue and orange boxes, highlight 687 distinct sub-regions within the segmented volume. Semi-polar plots illustrating the orientation of 688 ER banding (magenta) and mitochondrial networks (green). The left plot presents data from 689 myotomes 1 and 3. The right plot displays data from myotomes 2 and 4. (d) Segmentation of an 690 entire muscle tissue volume with three independently selected ROIs, colour -coded orange, 691 magenta, and cyan. (e) A montage of deconvolved raw time -series data from the corresponding 692 colour-coded ROIs in d displaying mitochondrial dynamics. Mitochondrial fission (grey arrows) 693 and fusion (yellow arrows) events are highlighted. Scale bars: (a,d): 50 μm (whole tissue), 30 μm 694 (myotomes), and 10 μm (mitochondria) ; (d): 50 μm; (e): 5 μm. 695 696 697 698 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 699 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 700 Figure 4. Endosomal motility in zebrafish embryo tail tissue. (a–d) Data were collected by 701 imaging the tailbud region of a 14 -somite stage from a zebrafish embryo expressing mKate2 -702 rab5ab after mRNA injection. mKate2 -rab5ab-labelled endosomes were detected in a volume 703 spanning approximately 266 × 266 × 60 µm 3 and imaged over 12 min at 5.9 s per volume. Max 704 intensity projections taken from (top) XY plane and (bottom) YZ plane are shown. (b) Overlays of 705 endosome tracks onto raw images as in a, showing all endosomes tracked for at least 10 frames 706 (60 s). Track segments are coloured according to the anomalous diffusion exponent (α) calculated 707 by fitting the mean squared displacement (MSD) for each track segment. (c–d) Selected volumes 708 corresponding to dotted ( c) and dashed (d) insets in a and b. Panels are as described in a. Each 709 volume spans 20 × 40 × 15 µm 3. (e) Distribution of observed values of α, illustrated by ( top) a 710 representative trajectory that exhibits a range of motion from directed to confined, coloured 711 according to α, and (bottom) the probability density of α calculated for each track segment in a. (f) 712 Volume projection of data collected by imaging the lateral tail region of a zebrafish embryo 32 713 hours post fertilisation expressing mKate2 -rab5ab (orange) and membrane-mNeonGreen (cyan) 714 after mRNA injection. (g) Zoom of the region highlighted in f. (h) 3D view of the tail region of 715 the zebrafish embryo expressing mKate2 -rab5ab (orange) and membrane -mNeonGreen (cyan), 716 displaying ridges and complex cellular morphologies. Scale bars: (a, b): 25 μm, (f): 50 μm, (g, h): 717 10 μm. 718 719 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 720 721 722 723 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figure 5. Correlation of rapid cellular dynamics and tissue -wide rearrangements in a 724 Drosophila embryo expressing sqh -3x-GFP. (a) Timelapse imaging of posterior midgut 725 (PMG) invagination shown every 1 min after initiation. (*) indicates primordial germ cells (PGC) 726 (b) Side view of the PMG invagination at the posterior end of the embryo shown every 3 min. 727 Cells appear to turn and elongate with the progress of invagination, subsequently undergoing cell 728 divisions, inferred by formation of cytokinetic rings. (c) Divisions are also observed on the dorsal 729 side of the extending tissue; arrows in the lower panel follow divisions (closing cytokinetic 730 furrows). (d) Left: Transparent surface rendering of a Drosophila embryo during midgut 731 invagination (orange arc) overlaid with division ring orientation vectors. Vectors are colour-coded 732 by alignment with the lateral (L, magenta), anterior–posterior (AP, cyan), and dorsal–ventral (DV, 733 yellow) axes. The white box indicates the magnified inset. Right: Representative division rings for 734 each axis and the corresponding 3D orientation colormap. (e) Schematic describing the views 735 presented in the images. (f) Cross-section at a depth of 27 µm from the ventral surface, displaying 736 a cell division (in rectangle) oriented in the anterior-posterior axis. (g) Cross-section at a depth of 737 35 µm highlighting two cell divisions occurring symmetrically. In this case, the orientation is dorsal-738 ventral. Scale bars: (a): 100 μm, (b): 25 μm, (c): 10 μm, (f, g): 20 μm. 739 740 741 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 742 743 744 745 746 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figure 6. Quantification of cellular dynamics in large, photosensitive single cells. (a) 747 Mouse meiosis II oocyte with labelled mitochondria (Dendra2) with diameter ~80 μm was imaged 748 every 5 min for 4 h. Max -intensity projections through the oocyte show the distribution of 749 mitochondria at four equally spaced time points. (b) Maps of mitochondrial motion were 750 calculated from optical flow. 3D flow maps of mitochondria throughout the entire oocyte volume 751 were time-averaged over 60 min centred at the indicated timepoint. Time -average velocity at each 752 voxel within the oocyte is colour -coded as indicated. (c) Multi-view temporal projections of an 753 amoeba captured at 6.2 s per volume. Morphology changes over 160 s are colour-coded according 754 to the temporal colour bar. The top -left orientation guide indicates the viewing angles (coloured 755 crosses) corresponding to the perspectives shown in the magenta, cyan, and orange boxed panels. 756 (d) Volumetric rendering of a segmented amoeba captured at 5.8 s per volume. The white box 757 indicates the region shown in the time -lapse sequence below. White arrowheads mark extending 758 protrusions and the green arrowhead highlights a fusion event, over a 24 -s interval. Scale bars: 759 (a,b): 50 μm, (c): 100 μm, (d): 50 μm. 760 761 762 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 763 764 765 766 767 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint Figure 7. (a) PKmitoDeepRed-labelled patient -derived colorectal cancer organoid rendered in 768 NVIDIA IndeX (b) Cross-section at the diameter of the cancer organoid showing single-cell-thick 769 wall of the organoid. (c) Time-lapse montage of mitochondrial dynamics acquired every 64.6 770 seconds. (d) 3D raw rendering (left) and cross-sectional view (middle) of a representative iBlastoid 771 stained for GATA3 (magenta), Nanog (orange), and GATA6 (cyan). Right: Corresponding instance 772 segmentation of individual nuclei. (e) Cross-sectional schematic of iBlastoid analysis (top left): a 773 sphere is fitted to the outer cell mass (OCM) nuclei (grey) to classify inner cell mass (ICM) nuclei 774 as inside (pink) or outside (blue) the boundary. The remaining panels display representative 3D 775 colour-coded segmentations of three blastoids analysed using this method. Coordinate units are in 776 µm. (f) Quantification of the proportion of ICM nuclei located inside (pink) versus outside (blue) 777 the fitted sphere for the three representative blastoids shown in e. Scale bars: (a,b): 50 μm, zoom: 778 10 μm; (c): 10 μm; (d): 50 μm. 779 780 781 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint

782 783 1. Y. Chen et al., Bridging single cells to organs: Mesoscale modules as fundamental units of tissue 784 function. Cell 188, 6393-6410 (2025). 785 2. A. N. Kapanidis et al., From sequence to function: Bridging single-molecule kinetics and molecular 786 diversity. Science 391, 458-465 (2026). 787 3. L. Z. Kreplin, S. Arumugam, High -resolution light-sheet microscopy for whole-cell sub-cellular 788 dynamics. Current Opinion in Cell Biology 85, 102272 (2023). 789 4. W. C. Lemon, K. McDole, Live-cell imaging in the era of too many microscopes. Current Opinion 790 in Cell Biology 66, 34-42 (2020). 791 5. P. J. Keller et al., Fast, high-contrast imaging of animal development with scanned light sheet-based 792 structured-illumination microscopy. Nat Methods 7, 637-642 (2010). 793 6. M. Lange et al., A multimodal zebrafish developmental atlas reveals the state-transition dynamics 794 of late-vertebrate pluripotent axial progenitors. Cell 187, 6742-6759.e6717 (2024). 795 7. P. Strnad et al., Inverted light-sheet microscope for imaging mouse pre-implantation development. 796 Nat Methods 13, 139-142 (2016). 797 8. K. McDole et al., In Toto Imaging and Reconstruction of Post-Implantation Mouse Development 798 at the Single-Cell Level. Cell 175, 859-876.e833 (2018). 799 9. C. Wolff et al., Multi-view light-sheet imaging and tracking with the MaMuT software reveals the 800 cell lineage of a direct developing arthropod limb. Elife 7, (2018). 801 10. V. M. S. Ruetten et al., Imaging cellular activity simultaneously across all organs of a vertebrate 802 reveals body-wide circuits. bioRxiv, 2025.2008.2020.670374 (2025). 803 11. T. L. Liu et al., Observing the cell in its native state: Imaging subcellular dynamics in multicellular 804 organisms. Science 360, (2018). 805 12. T. A. Planchon et al., Rapid three-dimensional isotropic imaging of living cells using Bessel beam 806 plane illumination. Nat Methods 8, 417-423 (2011). 807 13. B. C. Chen et al. , Lattice light -sheet microscopy: imaging molecules to embryos at high 808 spatiotemporal resolution. Science 346, 1257998 (2014). 809 14. T. Vettenburg et al., Light-sheet microscopy using an Airy beam. Nature Methods 11, 541-544 (2014). 810 15. B. Rieger, I. Droste, F. Gerritsma, T. Ten Brink, S. Stallinga, Single image Fourier ring correlation. 811 Opt Express 32, 21767-21782 (2024). 812 16. D. Li, E. Betzig, Response to Comment on “Extended-resolution structured illumination imaging 813 of endocytic and cytoskeletal dynamics”. Science 352, 527-527 (2016). 814 17. H. M. York et al., Deterministic early endosomal maturations emerge from a stochastic trigger-and-815 convert mechanism. Nature Communications 14, 4652 (2023). 816 18. V. E. Foe, Mitotic domains reveal early commitment of cells in Drosophila embryos. Development 817 107, 1-22 (1989). 818 19. R. E. Karess et al., The regulatory light chain of nonmuscle myosin is encoded by spaghetti-squash, 819 a gene required for cytokinesis in Drosophila. Cell 65, 1177-1189 (1991). 820 20. K. Yi et al., Dynamic maintenance of asymmetric meiotic spindle position through Arp2/3 -821 complex-driven cytoplasmic streaming in mouse oocytes. Nature Cell Biology 13, 1252-1258 (2011). 822 21. J. Van Blerkom, M. N. Runner, Mitochondrial reorganization during resumption of arrested 823 meiosis in the mouse oocyte. Am J Anat 171, 335-355 (1984). 824 22. C. M. Dalton, J. Carroll, Biased inheritance of mitochondria during asymmetric cell division in the 825 mouse oocyte. J Cell Sci 126, 2955-2964 (2013). 826 23. A. H. Pham, J. M. McCaffery, D. C. Chan, Mouse lines with photo -activatable mitochondria to 827 study mitochondrial dynamics. Genesis 50, 833-843 (2012). 828 24. M. Dembo, Mechanics and control of the cytoskeleton in Amoeba proteus. Biophysical Journal 55, 829 1053-1080 (1989). 830 25. P. Soneji, E. J. Challita, S. Bhamla, Trackoscope: A low -cost, open, autonomous tracking 831 microscope for long-term observations of microscale organisms. PLoS One 19, e0306700 (2024). 832 26. R. M. Engel et al., Patient-Derived Colorectal Cancer Organoids Upregulate Revival Stem Cell 833 Marker Genes following Chemotherapeutic Treatment. J Clin Med 9, (2020). 834 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint 27. L. Minarrieta et al., Mitochondrial elongation impairs breast cancer metastasis. Science Advances 10, 835 eadm8212. 836 28. X. Liu et al., Modelling human blastocysts by reprogramming fibroblasts into iBlastoids. Nature 591, 837 627-632 (2021). 838 29. X.-J. Tan et al., Volumetric two-photon microscopy with a non-diffracting Airy beam. Opt. Lett. 44, 839 391-394 (2019). 840 30. N. A. Hosny et al., Planar Airy beam light-sheet for two-photon microscopy. Biomed Opt Express 11, 841 3927-3935 (2020). 842 31. X. Ruan et al., Image processing tools for petabyte-scale light sheet microscopy data. Nature Methods 843 21, 2342-2352 (2024). 844 32. M. Kirschner, J. Gerhart, T. Mitchison, Molecular “Vitalism”. Cell 100, 79-88 (2000). 845 33. A. M. Turing, The chemical basis of morphogenesis. Philosophical Transactions of the Royal Society of 846 London. B, Biological Sciences 237, 37-72 (1952). 847 34. P. Liberali, A. F. Schier, The evolution of developmental biology through conceptual and 848 technological revolutions. Cell 187, 3461-3495 (2024). 849 35. J. Berger et al., Loss of Tropomodulin4 in the zebrafish mutant trage causes cytoplasmic rod 850 formation and muscle weakness reminiscent of nemaline myopathy. Dis Model Mech 7, 1407-1415 851 (2014). 852 36. K. M. Kwan et al., The Tol2kit: a multisite gateway -based construction kit for Tol2 transposon 853 transgenesis constructs. Dev Dyn 236, 3088-3099 (2007). 854 37. A. A. Ruparelia et al., Atrogin-1 promotes muscle homeostasis by regulating levels of endoplasmic 855 reticulum chaperone BiP. JCI Insight 9, (2024). 856 38. J. L. Schmid-Burgk, K. Höning, T. S. Ebert, V. Hornung, CRISPaint allows modular base-specific 857 gene tagging using a ligase-4-dependent mechanism. Nature Communications 7, 12338 (2016). 858 39. n. c. (2019), napari: a multi-dimensional image viewer for python. 859 40. C. Stringer, T. Wang, M. Michaelos, M. Pachitariu, Cellpose: a generalist algorithm for cellular 860 segmentation. Nat Methods 18, 100-106 (2021). 861 862 .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint .CC-BY-NC-ND 4.0 International licenseperpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for thisthis version posted February 25, 2026. ; https://doi.org/10.64898/2026.02.25.707857doi: bioRxiv preprint