Multi-view confocal microscopy enables multiple organ and whole organism live-imaging

The size, shape and organization of organs, tissues or cells are highly regulated during development, homeostasis and repair. As such, capturing cell, tissue and organ growth and morphogenesis over long periods of time is central to understand these biological phenomena and to probe their underlying genetic and biophysical regulations (Collinet and Lecuit, 2021; Goodwin and Nelson, 2021; Hannezo and Heisenberg, 2019). Importantly, cell, tissue or organ dynamics can be driven by long-range mechanical coupling, as well as systemic hormonal regulation (Barresi and Gilbert, 2019; Boulan et al., 2015; Cole et al., 2019; Villedieu et al., 2020), highlighting that the understanding of growth and morphogenesis also necessitates cellular imaging of multiple tissues or organs in whole living organisms.

Fluorescent light-microscopy has emerged as an essential technology to study biological systems by live imaging at spatiotemporal resolutions ranging from subcellular to whole-organism scale (Keller, 2013). This is best illustrated by two distinct imaging approaches: confocal microscopy and light-sheet microscopy (Fig. 1A) (Bayguinov et al., 2018; Keller and Stelzer, 2008; Keller et al., 2008; Oreopoulos et al., 2014; Wan et al., 2019). Confocal fluorescence microscopes are currently the most commonly used imaging systems with optical sectioning capability. In particular, spinning disk confocal microscopy is a mature, commercially available technology for studying development, homeostasis and repair by live imaging. It can resolve biological processes from the subcellular to the whole-tissue scale. Furthermore, spinning disk confocal microscopy has many advantages over laser scanning confocal microscopy for dynamic imaging, particularly in terms of acquisition speed, photobleaching and phototoxicity (Oreopoulos et al., 2014; Wang et al., 2005). Spinning disk confocal microscopy samples are often mounted on glass slides or specific dishes, and observation is only possible from one side, namely perpendicular to the slide. For many applications in biology, and in particular in the field of developmental biology, it is crucial to be able to obtain multiple views of the structure(s) under investigation. Indeed, the observed tissues are often not organized in a uniform layer but present complex 3D structures. In confocal microscopy at the level of individual cells, attempts have been made to better orientate cells by inserting single cells into a glass capillary or fixing them to a glass fiber (Bradl et al., 1994; Bruns et al., 2015; Staier et al., 2011). Whether these approaches would allow a multi-view exploration of tissues or organisms while achieving high-resolution imaging of multiple tissues remains unexplored. Complementing confocal microscopy, light-sheet fluorescence microscopy (LSFM) combines intrinsic optical sectioning with wide-field detection (Keller, 2013). In contrast to epifluorescence microscopy, only a thin slice (usually a few hundred nanometers to a few micrometers) of the sample is illuminated perpendicularly to the direction of observation. This method allows the acquisition of large field of view images in a single, spatially confined illumination step; offering low photobleaching, low phototoxicity, good signal-to-noise ratio and improved acquisition speed (de Medeiros et al., 2015; Guignard et al., 2020; Huisken, 2004; McDole et al., 2018; Mertz, 2011; Pitrone et al., 2013; Siedentopf and Zsigmondy, 1902; Strnad et al., 2016; Voie et al., 1993). In addition, the majority of light-sheet microscopes enable sample imaging from different points of view using a sample holder mounted on a rotating stage and/or by having multiple lenses to collect light from different angles (Chhetri et al., 2015; Krzic et al., 2012; Schmid et al., 2013; Tomer et al., 2012; Wu et al., 2013). So far, LSFM permits live in toto imaging of immersed/embedded samples. However, there are numerous model organisms that cannot be immersed or embedded as their survival depends on oxygen uptake by ventilation. Whether LSFM can be efficiently applied to non-immersed samples has yet to be tested.

As exemplified by numerous past and recent studies, imaging of Drosophila pupae without immersion has been key in the exploration of developmental and repair processes and to characterize core and conserved genetic and biophysical mechanisms driving cell differentiation as well as tissue proliferation, shaping and architecture (Aigouy et al., 2010; Arata et al., 2017; Corson et al., 2017; Curran et al., 2017; Diaz-de-la-Loza et al., 2018; Etournay et al., 2015; Founounou et al., 2013; Franz et al., 2018; Gho et al., 1999; Guirao et al., 2015; Lemke and Schnorrer, 2018; Levayer et al., 2015; Mauri et al., 2014; Michel and Dahmann, 2020; Prat-Rojo et al., 2020; Ray et al., 2015; Sarov et al., 2016). In particular, using spinning disk confocal microscopy, many studies have focused on the dynamics of a single tissue including the pupal wing and dorsal thorax (notum) tissues as well as histoblast nests. Yet, and as in most model systems, it remains crucial to investigate multiple tissue and organ dynamics to define the underlying genetic and biophysical mechanisms coordinating the development of a whole organism. Here, we initially set out to test LSFM as an imaging approach to begin to explore genetic and mechanical coupling between multiple organs during Drosophila pupa development. Although we could use LSFM to image non-embedded pupa, LSFM was not optimal for imaging cell and tissue dynamics with sufficient resolution. Consequently, we developed a Multi-View confocal microScope (MuViScope), which enables confocal multi-view imaging of non-immersed samples over time. We illustrate how its multi-view capability enables visualization of the dynamics of multiple organs, which will be instrumental to address fundamental biological questions related to organismal development, homeostasis and repair.

We set out to image multiple organs during Drosophila pupa development, focusing in particular on the wing and the dorsal thorax. In recent years, LSFM emerged as the technique of choice for the in toto imaging of organisms (Wan et al., 2019). A known drawback of LSFM is the shadowing caused by absorption, scattering or refraction of light inside the sample (Rohrbach, 2009). Although these stripe artifacts can be found in any light microscope, they are more pronounced in light-sheet based microscopes owing to the illumination from the side. Possible solutions to reduce shadowing include scanning illumination and/or multi-view imaging either by sample rotation or by dual-side sample illumination (Baumgart and Kubitscheck, 2012; de Medeiros et al., 2015; Huisken and Stainier, 2007; Keller et al., 2008). However, for samples of more than a couple of hundred cells, these artifacts can undermine the ability to clearly image certain structures (Baumgart and Kubitscheck, 2012; de Medeiros et al., 2015; Huisken and Stainier, 2007; Keller et al., 2008). It remains unknown whether or not these artifacts might hinder the imaging of the Drosophila pupa. In addition, most commercial LSFM systems available require sample immersion in liquid or embedding in agarose. This potentially poses significant problems for organisms that rely on ventilation for gas exchange. To assess the possibility for light-sheet microscopy to image Drosophila pupae, we first opted for the light-sheet Z1 microscope from Carl Zeiss using dual-illumination on the pupal dorsal thorax epithelium labelled by the adherens junction marker Ecad:3×GFP (Fig. 1; Fig. S1). Upon imaging of 54 h after pupa formation (hAPF) old pupae immersed in PBS, the image quality was high on the lateral parts of the tissue and a cellular resolution could be obtained (Fig. 1B,B′; Fig. S1). However, our analysis revealed that immersion of the Drosophila pupa in PBS stopped their development, as exemplified by the mitotic arrest observed as early as 5 min upon immersion in PBS (Movie 1). As liquid immersion of the pupa precludes live imaging, we then assessed LSFM image quality on the Drosophila pupa using the PhaseView ALPHA3 system. This system was chosen because its objective, detection lens and the light sheet alignment can be adapted to non-immersed samples. However, our tests showed that the light sheet did not homogeneously illuminate the tissue when the dorsal thorax was imaged (Fig. 2A). Especially in the medial part of the notum, a very low signal was obtained (Fig. 2A, inset). Likewise, when imaging the lateral side of the pupa the wing, legs and eyes were not homogeneously illuminated (Fig. 2B). In summary, these results illustrate that currently available commercial LSFM systems do not produce high quality images in all areas of the dorsal thorax and wing tissues of the Drosophila pupa to record its development by time-lapse microscopy.

Spinning disk confocal microscopy has often been employed to image individual epithelial tissues in Drosophila such as the wing, notum or the abdomen (Aigouy et al., 2010; Arata et al., 2017; Curran et al., 2017; Diaz-de-la-Loza et al., 2018; Etournay et al., 2015; Founounou et al., 2013; Gho et al., 1999; Guirao et al., 2015; Kanca et al., 2014; Keroles et al., 2014; Levayer et al., 2015; Michel and Dahmann, 2020; Prat-Rojo et al., 2020; Ray et al., 2015). As individual tissues could successfully be imaged using confocal microscopy, we hypothesized that sample rotation along the anterior-posterior (A-P) axis in combination with confocal microscopy could enable the visualization of multiple tissues. Furthermore, we foresaw that increasing the number of optical paths would enable imaging at multiple views at different magnifications. Accordingly, we developed the MuViScope (Figs 2C-F′ and 3; Fig. S2; Movie 2).

The system is based on spinning disk illumination where the light source is composed of a fiber laser bench that is connected to the spinning disk head. The laser beam is focused on the sample by means of mirrors, dichroic mirrors and lenses. The sample is imaged using two dry long working distance objectives located on opposite sides of the sample. Each objective corresponds to an optical path where the excitation and collection of light is carried out via the same objective (Fig. 3A,B). The light emitted from the sample is collected by the excitation lens and passed back to the dichroic mirror located in the spinning disk head, which will reflect the light back to the scientific complementary metal-oxide semiconductor (sCMOS) camera. The collection of the image via the opposite lens is carried out in the same way, by changing the optical path using a motorized flip mirror (Fig. 3A,B; Movie 2). The sample is placed at the end of a capillary or metal rod which is held by a capillary holder (Fig. 3C-F). Samples are mounted by either gluing them to the capillary using dental glue (Fig. 3F) or by sticking them to double-sided tape on a small metal rod with a flattened end (see Materials and Methods and supplementary Materials and Methods). The capillary and metal rod fit into a rotating stage that can be precision controlled along the x-y-z-θ axes (Fig. 3C-F; Fig. S2). Temperature and humidity are controlled within an opaque box, which surrounds the sample, sample holder, x-y-z-θ precision stages and objectives.

First, we assessed the ability of the MuViScope to visualize the dorsal thorax and wing tissues in Drosophila pupae (see Table S1 for acquisition procedure and settings). When imaged from the dorsal side, lateral structures such as the eyes and medial structures of the midline in the notum were well captured (Fig. 2C). In addition, the eye, legs, wing and wing veins were readily visualized by rotating the pupa 90° (Fig. 2D). These results stand in contrast to the images acquired using the Phaseview microscope, for which the notum and wing tissues of living pupae could not be visualized without shadowing effects (Fig. 2A,C insets, E). Furthermore, 180° sample rotation combined with two optical arms of the MuViScope offered the possibility to switch between distinct objectives, enabling the acquisition of the tissue at different magnifications capturing both global and cellular tissue views in the same animal (Fig. 2F,F′). The MuViScope therefore complements LSFM by permitting higher quality live imaging of non-immersed samples.

We next explored the capability of the MuViScope to image samples using fluorescence and autofluorescence at the animal scale. We first imaged a pharate adult Ecad:3×GFP Drosophila pupa over 90° using the 10× objective (Fig. 4A). The resulting images clearly show major structures, such as the body segments of the head and thorax and the legs, as well as smaller structures such as the folds in the wing, ommatidia in the eyes, and the bristles on the thorax (Fig. 4A). A 3D reconstruction of a pupa at 28 hAPF was assembled by merging and deconvoluting eight angles taken over 360° (0°, 45°, 90°, 135°, 190°, 235°, 270° and 315°; see Movie 3). Using 40× magnification, we also acquired a high-resolution dataset of the head and thorax of the pupa over 360°. The 3D fusion and reconstruction of this dataset illustrates the capability of the MuViScope to obtain a 360° view of the anterior part of the pupa with cellular resolution (Movie 4). To test the MuViScope on other model organisms, multiple views of an adult Tribolium and Aranea were obtained. Using only autofluorescence, the legs, body and eyes could be ascertained in an Aranea (Fig. 4B). In the αtub1-LifeAct:GFP, EF1α-nls:GFP transgenic Tribolium adult, the body segments, legs and antennae were clearly distinguishable (Fig. 4C; Movie 5). In addition, the resolution was high enough to discern small dimples present on the surface of the outer chitin layer (Fig. 4C′). The MuViScope is therefore capable of generating high quality 360° views in different species.

Our next aim was to perform time-lapse live microscopy experiments. Viability was assessed by placing undissected pupae in the MuViScope imaging chamber with the temperature set to 25°C and humidity to 70%. Under these conditions high pupal viability was achieved (viability 90%, n=9/10). Furthermore, pupal viability remained high upon removal of the pupal case (puparium) (viability 90%, n=18/20), mounting with dental glue (viability 85%, n=11/13) or on double-sided tape (viability 90%, n=9/10). We then checked whether pupae could be imaged from 0.5 hAPF to eclosion. For this, we used the transgenic hh-DsRed line, which is visible through the puparium and labels the posterior compartment of tissues (Akimoto et al., 2005). When performing 360° imaging using four angles (i.e. 90° apart), hh-DsRed signal could be observed on all sides of the pupa (Fig. 5A; Movie 6). We focused mainly on wing development. Labeling of the posterior compartment of the tissue allowed the observation that the wing posterior compartment underwent a fast and global posterior movement at 11 hAPF and started to undergo extensive twisting at 38 hAPF (Fig. 5A,A′; Movie 6); two morphogenetic events occurring at the time of head eversion and wing folding, respectively (Bainbridge and Bownes, 1981). These results show that the imaging conditions ensure high viability and, accordingly, enables imaging of the entire pupal development.

Mesodermal tissues in the Drosophila pupa develop in multiple locations (Dutta et al., 2004; Jährling et al., 2010; Lemke and Schnorrer, 2018; McGurk et al., 2007). However, some mesodermal tissues in Drosophila have not been imaged live, because they develop inside the organism at positions for which mounting under a coverslip might be challenging. Especially difficult to image are the lateral and ventral muscles in the pupal thorax, because the wing and legs, respectively, prevent placing a coverslip directly on top of the position where the muscles develop. Three-angle (0°, 90°, 180°) imaging was therefore performed to assess the performance of the MuViScope in imaging organs that develop within the Drosophila pupa. The striking pattern of dorsal-longitudinal muscles labeled by Mef2>CD8:GFP in the dorsal thorax could be well captured and imaged during development (Fig. 6; Movie 7). An expansion in the A-P direction was visible between 26 hAPF and 50 hAPF, followed by a phase of muscle definition between 50 hAPF and 98 hAPF (Spletter et al., 2018). Moreover, sample rotation enabled the visualization of muscle development laterally and ventrally (Fig. 6; Movie 7). Muscle expansion in the legs was most notable between 26 hAPF and 50 hAPF. This was followed by a refinement of the pattern of muscles from 50 hAPF and 98 hAPF, similar to muscle development in the dorsal thorax. These results confirm and extend previous findings (Dutta et al., 2004; Jährling et al., 2010; McGurk et al., 2007; Spletter et al., 2018) and show that MuViScope is a suitable tool to explore the coordination of internal tissue development.

In many cases, it might be advantageous to image multiple organs in the same animal to explore whether and how genetic or mechanical developmental interplay exists between organs and how systemic regulation coordinates animal scale development. We therefore set out to explore whether MuViScopy could be well suited to visualize multiple epithelial organs in the same animal. As a test case, we set out to characterize the development of both the wing and the dorsal thorax in both wild-type (wt) and dumpy (dpy) mutant conditions by labelling the tissue using Ecad:3×GFP. The extracellular matrix protein Dpy is known to control morphogenesis in the wing and dorsal thorax by enabling the attachment of the epidermis to the cuticle (Aigouy et al., 2010; Etournay et al., 2015; Metcalfe, 1970; Olguín et al., 2011; Ray et al., 2015). Previous research has shown that loss of Dpy function leads to misshaped wings and the formation of posterior lateral pits and anterior comma-like invaginations in the notum (Aigouy et al., 2010; Etournay et al., 2015; Metcalfe, 1970; Olguín et al., 2011; Ray et al., 2015). Yet, the temporal sequence leading to formation of these defects in different organs has not yet been defined in the same animal using live microscopy. This therefore offers the opportunity to assess the ability of the MuViScope to explore concomitant tissue dynamics in multiple organs.

We first visualized wing (lateral view) and dorsal thorax (dorsal view) tissues in Ecad:3×GFP pupa over 90° to assess the wt processes at cell and tissue scale using both 10× and 40× imaging (Fig. 7). In line with published results independently described in the wing and the notum (Aigouy et al., 2010; Etournay et al., 2015; Guirao et al., 2015; Ray et al., 2015), the lateral view enables the observations of the hinge contraction at 17 hAPF and the subsequent elongation of the wing blade, whereas in the dorsal view we could observe the flows of thorax cell towards the anterior as well as the morphogenesis of the posterior lateral region of the notum starting at 21.5 hAPF (Fig. 7A,B; Movie 8). We also checked whether 40× MuViScope imaging enables quantification of tissue flows using particle image velocimetry (PIV; Fig. S3; Movie 9) as well as performing cell segmentation, tracking and morphometric quantification of cell dynamics in the wing and notum tissues (Figs S4 and S5). We then performed similar movies in dpy^OV1 animals. In dpy^OV1 before 17 hAPF, before hinge contraction, the distal end of the wing and the dorsal thorax were similarly shaped compared with wt (Fig. 7A-D, green arrowheads). The distal part of the wing was round and no invaginations could be observed in the notum. Starting from 17 hAPF, when the hinge started to contract, the distal part of the dpy^OV1 wing started to change its shape while moving towards the hinge. Accordingly, the wing blade failed to elongate as observed in wt tissues (Movie 8; Fig. 7A-D). Using both the lateral and dorsal view we could observe that, like the wt tissue, the hinge contraction was followed in the notum by cell flow towards the anterior at 21 hAPF (Movie 8). However, in dpy^OV1 conditions the anterior comma like invaginations started to appear at 25 hAPF and became progressively more visible until 34 hAPF (Movie 8; Fig. 7C, yellow arrowhead). The posterior lateral pits in the notum appeared at 26.5 hAPF (Fig. 7C, red arrowhead). These pits could not be observed in wt tissue (Fig. 7A, yellow and red arrowheads). Together, these results illustrate that the MuViScope can be used to investigate the morphogenetic processes in both wt and mutant animals in multiple organs within the same animal.

Here, we implemented an advanced microscopy methodology named MuViScopy based on spinning disk confocal illumination and multidirectional imaging by sample rotation and double-sided illumination/detection. We showed that this methodology can be used to image arthropods such as Tribolium, Aranea and Drosophila. In particular, we explored the ability to visualize organs and multiple organ development in Drosophila. MuViScopy enables high quality automated 360-degree imaging of living biological samples without sample embedding/emersion. Its conventional sample illumination and fluorescence collection by a single objective enables homogeneous sample illumination without the need for post-acquisition image processing to reconstruct one homogeneously illuminated field of view. External organs such as the epidermis are well captured throughout the entire thorax and can be visualized with cellular resolution. Furthermore, MuViScopy permits live imaging of the morphogenesis of internal organs that are challenging to image due to their location inside the organism. In addition, multidirectional imaging enables the simultaneous visualization of the development of distinct organs with different magnifications of the same sample in wt or mutant animals. MuViScopy therefore opens up a new range of experimental opportunities to better understand how multiple organs develop as well as to explore the genetic, biophysical and systemic regulations that underlie the organ dynamics. We therefore believe that the MuViScope will be advantageous for the study of fundamental biological questions in organisms that rely on ventilation for gas exchange. Accordingly, we have provided the MuViScope blueprint and detailed technical information to support the dissemination of MuViScope microscopy.

We foresee that MuViScopy and LSFM will be highly complementary to analyze tissue and organ dynamics (Table S2). Basic optical principles explain some of the complementarity to image immersed versus non-immersed samples. The advantage of LSFM imaging is that the acquisition orthogonal to the illumination plane optimizes photon collection from a given illuminated tissue section (Huisken, 2004), providing a benefit in terms of photobleaching and acquisition time compared with the MuViScope (Bayguinov et al., 2018; Keller and Stelzer, 2008; Keller et al., 2008; Oreopoulos et al., 2014; Wan et al., 2019). However, the orthogonal acquisition of the LSFM is less favorable when imaging in dry conditions. When the sample is not immerged the mismatch between the air (n₀≃1) and the sample index (n within the range of 1.38-1.42) (Tuchin, 2007) is in the order of 0.4. In both LSFM and MuViScopy, this increases the deflection of the light beam as it hits the sample. In the case of LSFM with acquisition orthogonal to the illumination, this generates a z-offset in focal plane that increases as dtan (Δθ), whereas the z-offset only scales with d[1−cos (Δθ)] in the case of MuViScopy (with d the sample depth, Δθ=θ_i−θ_r the difference between the angles of incidence θ_i and refraction θ_r with given by the Snell-Descartes law; Fig. S6A). For example, with n≃1.4 (the mean value of the refractive index for the tissue), θ_i of 20° and a depth of 50 µm, the z-offset will be in the order of 5.13 µm for the LSFM and 0.26 µm for the MuViScope; the LSFM one being above the Nikon 20×/0.45 and 40×/0.6 air objective axial resolution (respectively 5.09 µm and 2.86 µm for the GFP emission wavelength of 515 nm; Fig. S6B), and the MuViScope one being in the correct range. In brief, this partly explains why, in dry conditions, LSFM can still resolve the lateral regions relative to the acquisition direction but not the medial regions, whereas the MuViScope achieves good cellular resolution in the two regions. In both LSFM and MuViScopy, sample rotation enhances the microscope capabilities by increasing resolution for LSFM and enables the reconstruction of sample 360° views for both the LSFM and MuViScope.

We see room to improve the MuViScope in multiple ways. First, a sample finder module as is present on the commercial Carl Zeiss Z1 system could reduce the time spent finding the optimal focus. Second, the design of the sample holder was inspired by light-sheet microscopy. This entails that the sample holder currently only fits one pupa. Enlarging the mounting area would increase sample throughput at the cost of reducing the number of angle position that can be imaged. Third, photon collection could be optimized by using a higher quantum efficiency sCMOS camera and a field homogenizer module. Lastly, the addition of new electrically tunable lenses would increase the acquisition speed of the z-stacks by enabling successive z-plane imaging without moving the sample. We foresee that these future changes will further enhance the range of applications of the MuViScope in the study of fundamental biological questions.

A schematic outline of the MuViScope microscope (MuViScope) that was created using the 3D modeling software Solidworks (Dassault Systèmes SolidWorks Corporation) is shown in Fig. 3. Technical drawings of the MuViScope and custom optical components are shown in Fig. S2. A comprehensive list of MuViScope components is also provided in Fig. S2B. The MuViScope consists of a multi-laser unit (GATACA Systems), one spinning disk confocal scan head (Yokogawa CSU-W1, Andor Technology) equipped with sCMOS camera (Orca Flash 4.0 v2, Hamamatsu) and a home-made specimen holder, which is magnetically attached to a four-axis specimen-positioning system (x-y-z-θ). The system was built on an optical anti-vibration table (TMC). The multi-laser bench is equipped with four laser lines 405 nm, 488 nm, 561 nm and 642 nm. The laser beam is guided through a single-mode fiber into the confocal spinning disk head. The optical path between the spinning disk head and the sample consists of two optical arms that direct light into two opposing dry objective lenses (Fig. 3A,B,D). Each arm is composed of mirrors, a tube lens (TI-E 1×, Nikon Instruments Inc.) and an objective (CFI Plan Apo Lambda 10×/0.45 or CFI S Plan Fluor ELWD 40×/0.6, Nikon Instruments Inc.).

The collimated beam at the exit of the spinning-disk head is passing through a relay lens system and then diverted into the desired optical arm using a motorized flip mirror (custom-made by Errol Laser). The collected light follows the same optical path and is reflected from this flip mirror back onto the spinning disk head. A dichroic mirror and an internal detection-filter wheel located in the spinning disk head reflect the fluorescence to the camera. The detection-filter wheel contains various detection filters (BrightLine long-pass and band-pass filters, Semrock) for imaging GFP and DsRed fluorophores. The capillaries (or rods) are inserted into a custom specimen holder made of stainless steel and held in place by the use of Teflon glands. The sample holder is then fitted into the axis of the rotating stage using a magnetic adapter. The whole assembly is held tightly and accurately to the four-axis specimen-positioning system (Fig. 3; Fig. S1). This four-axis (x-y-z-θ) specimen-positioning system consists of three motorized linear stages (2× M-404.1DG and 1×M-111.1DG, Physik Instrument) and one rotary stage (M-660.55, Physik Instrument) controlled by a four-axes motion controller (C-884.4DC, Physik Instrument). With this system, the specimen holder can be translated along three axes and rotated around its main axis. A thermostatically controlled black plexiglass enclosure covers the whole unit. The temperature is controlled by the Cube temperature module (Life Imaging Services), and the humidity by a commercial humidifier (Eva, Stadler Form). These modules include sensors that are placed near the sample and maintain the user-defined temperature and humidity level. A cover at the top of the chamber provides access to the sample holder. A cold light source is available to help for sample positioning using brightfield illumination.

The optomechanical devices of the MuViScope are all operated through Metamorph software (Molecular Devices, version 7.8.13). The motorized flip mirror is controlled by a custom-made controller (Errol Laser) via a digital-to-analog controller (DAC).

Drosophila stocks were maintained on standard food at 25°C. Stocks used were: Ecad:3×GFP (Pinheiro et al., 2017), Ecad:GFP (Huang et al., 2009), dpy^OV1 (Bloomington Drosophila Stock Center, #276), Mef2:GAL4>UAS-CD8:GFP (Mef2>CD8:GFP; gift from F. Schnorrer, Institut de Biologie du Développement de Marseille, France), hh-DsRed [hh-Pyr215, Kyoto Stock Center (DGRC), 109137].

Drosophila pupae were staged by collecting white pupae (0 hAPF). The puparium was removed for imaging except in the hh-DsRed experiment. Mounting and dissection was carried under a binocular (Carl Zeiss). The posterior part of the pupal case was glued either: (1) at the end of a capillary as done for other specimens in Laroche et al. (2019) using dental glue (Protemp II RF A3, 3 M) for full rotation imaging (Fig. 3F); or (2) the entire pupa was placed on double-sided tape (D6661933, 3M) on a flat platform at the end of a metal rod. The pupal case was then delicately dissected from the head to the beginning of the abdomen, either completely or only dorsal-laterally (Jauffred and Bellaiche, 2012; Zitserman and Roegiers, 2011). The capillary was then installed on its holder and inserted into the rotating stage. Dental glue mounting enabled unconstrained 360° rotation around the A-P axis of the pupa. A detailed protocol of the sample preparation is available in the supplementary Materials and Methods.

Tribolium transgenic αtub1-LifeAct:GFP, EF1α-nls:GFP (van Drongelen et al., 2018) adults were grown at 29°C in a Tupperware container containing flour and yeast. The Tribolium was euthanized by placing it at −20°C for 1 h. It was then fixed using dental glue to the end of the capillary as detailed above.

Aranea were collected in the wild in Paris (France) and stored in ethanol before imaging. They were fixed to the end of the capillary by using dental glue. All imaging with MuViScope was performed at 25°C.

Staged Drosophila pupae (30 hAPF) were lined up in a Petri dish with the basal part on the plastic side. Four conditions were tested: pupae without glue, pupae glued on the ventral side with double-sided tape (3M), pupae glued at their end with dental glue (Protemp II RF A3, 3M), pupa glued at their end with UV-activated glue (UHU, EAN:4026700481501). The Petri dishes were placed in a box containing wet paper and placed in an incubator at 25°C for 7 days. The number of empty pupae was counted to assess the survival rate. To assess pupal viability upon immersion, staged Ecad:GFP pupae (12 hAPF) were mounted under a coverslip (Bosveld et al., 2012; Gho et al., 1999) and imaged with a 40× oil objective every 5 min on a Nikon CSU-W1 inverted spinning disk microscope fitted with a Hamamatsu Orca Flash 4.0 sCMOS camera and a Borealis module from Andor (Oxford Instruments) for better illumination homogeneity. After 50 min 200-400 µL of PBS was added with a pipette between the slide and the coverslip and images were acquired every 5 min to assess cell and tissue dynamics upon immersion.

The Metamorph acquisition software allows total control of the MuViScope via the Multi-acquisition mode. The control of the axes is done in the ‘Stage’ tab of the ‘Multi Dimension Acquisition’ menu (MDA) of the software. We use the Z2 stage function to control the rotating stage. The acquisition mode includes the acquisition of z-stack time-lapse images at multiple angles and the creation of mosaics enabling the acquisition of the entire sample. All of these modalities can be combined depending on the type of acquisition required (e.g. multi-view tiling with time-lapse). The following parameters of acquisitions are available in the MuViScope control software (Metamorph) in a sequential order: xy coordinate of the image (xy), multi-channel imaging (ch), z-stack (z), multi-positions allowing tiling (P), multi-angles (θ) and two different magnifications (M). In brief, the order of acquisition is as follows: xy→ch→z→P→θ→M→t. For most of the acquisitions, the exposure time per image was 400 ms in binning 2 and a laser power of 0.6 mW at 488 nm and 1.5 mW at 561 nm measured at the back focal plane of the objectives. For imaging, depending on the sample, a 100-500 µm z-stack was acquired each time-point at 4 µm (10× objective) or 2 µm (40× objective) z-spacing. The rotation speed of the stage is 212°/sec in our experimental conditions. This represents the speed at which the stage can turn from one angle to the next and does not include the time necessary for image acquisition. The time between two acquisitions was either 15 min or 30 min. The detailed acquisition parameters for each figure are provided in Table S1.

The temperature of the thermostatically controlled chamber was set to 25°C with a humidity level of 70%. Samples were located using transmission light (white LED, Lumiled) and then finely focused using spinning disk confocal imaging by moving the motors in x, y, z, θ via the Metamorph software.

LSFM was performed using the Lightsheet Z1 microscope (Carl Zeiss; illumination lens 10×/0.2 Carl Zeiss, detection lens 20×/1.0 W Carl Zeiss) for pupae immersed in PBS and the alpha3 light-sheet microscope system (PhaseView; illumination lens: EPIPLAN 10×/0.23 Carl Zeiss, detection lens MPLN 10×/0.25 Olympus) for pupae without immersion. Pupae were attached to the capillary with dental glue under the same conditions as for the MuViScope (see supplementary Materials and Methods).

Time-lapse images and multi-view z-stack acquisitions were processed using Fiji software (Schindelin et al., 2012). Briefly, for each angle and at each time point, a maximum projection (MIP) of z-stack images was made and then, if necessary, the resulting multi-view MIP images were stitched with the Pairwise Stitching plugin (Preibisch et al., 2009). This was done for every angle and every time point. A movie was then made from the sequential acquisitions and a montage showing the angles side by side was created by using the Stack Combiner plugin. For Fig. 7, the images were denoised using ND-Safir software (Boulanger et al., 2010) and Movie 8 was realigned using the ImageJ plugin StackReg (Thévenaz et al., 1998). 3D reconstruction by fusion/deconvolution of multi-angle light-sheet or MuViScopy data was performed using Huygens Fuser software (SVI). The alignment of the different views was first carried out manually using the interactive Fuser & Decon alignment module then an automatic registration in ‘full optimization on initial guess’ was performed followed by a fusion deconvolution (Huygens Fuser, SVI).

Local tissue flows were quantified using movies of pupae expressing Ecad:3×GFP. For each movie, the flow field was obtained by image cross-correlation (IC) velocimetry along sequential images using customized MATLAB routines (Bosveld et al., 2012; Raffel et al., 2018) based on the particle image velocimetry (PIV) toolbox ‘matpiv’ (http://folk.uio.no/jks/matpiv). PIV was performed within boxes of 40 square pixels (∼13×13 μm², pixel size 0.32 μm) with 50% overlap for spatial averaging and integrated over 2 h. Averaging PIV flows from different pupae was carried out by generating a tissue archetype in order to register and rescale each time-lapse movie. Registration and rescaling were performed using the four dorsocentral macrochaetae as landmarks (Guirao et al., 2015). Each PIV box was then positioned according to the barycenter of the four macrochaetae and rescaled. Together, this enables improved comparison of the MuViScope and spinning disk microscope conditions by reducing the spatial dispersion of the landmarks and averaging of PIV measurements.

Segmentation was performed using Imaris software (Oxford Instruments) within the wing or using previously published MATLAB (Mathworks) routines based on standard watershed algorithm within the scutellum (Bosveld et al., 2012; Guirao et al., 2015). Upon segmentation and multiple rounds of manual and automated correction, cells were tracked and cell areas analyzed using previously published MATLAB and C++ routines (Bosveld et al., 2012; Guirao et al., 2015).

We thank Frank Schnorrer (Institut de Biologie du Développement de Marseille), the Bloomington Drosophila Stock Center and Kyoto Stock Center for Drosophila stocks, Maurijn van der Zee and Kees Koops (both Leiden University) for Tribolium stocks, France Lam (Institute of Biology Paris-Seine) for access to and help with the Phaseview Alpha3 system, the cellular imaging platform at SFR Necker for access to the Zeiss Z1 light-sheet microscope and Sébastien Dupichaud for his help in operating it, Maïté Coppey (Institut Jacques Monod) for help with writing funding applications, Christine Rollard (Muséum national d'Histoire naturelle) for help with spider identification, François Graner and Isabelle Bonnet for the fruitful discussions and preliminary tests at the beginning of the project, Aude Maugarny-Calès for her advice and help with Drosophila sample mounting and Cyril Kana Tepakbong for help with image analysis.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199760.