A multimodal adaptive optical microscope for in vivo imaging from molecules to organisms

MainLight microscopy has shaped our view of how life works since scientists first began observing cells through a lens, revealing processes from single-molecule kinetics to in toto organismal development in model systems ranging from cultured cell lines to organoids, Caenorhabditis elegans, Drosophila, zebrafish and mice. Historically, this diversity drove the development of distinct microscope modalities, including widefield, confocal, light-sheet and two-photon microscopy, each optimized to meet specific needs1,2. Super-resolution (SR) variants3,4 of these have further expanded the range and power of optical microscopy.However, optimization for one modality or one class of samples invariably comes with constraints that compromise the study of other systems. These constraints encompass not only optical characteristics such as field of view (FOV) and resolution, but also practical considerations such as sample mounting and environmental control. For example, high numerical aperture (NA) objectives optimize spatial resolution, but sacrifice FOV and working distance. Light-sheet microscopes enable low-phototoxicity volumetric imaging, but can require nontraditional mounting arrangements or complex relay optics in inverted systems5. Two-photon point scanning increases imaging depth, but often sacrifices speed and broad multicolor capability6,7. As such, a modern imaging core or well-equipped microscopy laboratory often supports several different instruments tailored to different tasks, multiplying cost and operational complexity. Furthermore, nearly all commercial microscopes that are the backbone of imaging cores ignore the ‘elephant in the room’: sample-induced optical aberrations that quickly compromise performance in multicellular applications.To address these issues, we developed the Multimodal Optical Scope with Adaptive Imaging Correction (MOSAIC), a single microscope that reconfigures on demand to different imaging modalities, each optimized for a different class of specimens. We demonstrate MOSAIC across a broad range of biological systems and spatiotemporal scales, including single-molecule tracking in live adherent cells, organelle remodeling in both isolated cells and developing organisms, quantifying neural activity at single-spine resolution in the cortex of live mice and mapping neuronal ultrastructure across millimeters of cleared tissue.

Collectively, these applications highlight MOSAIC’s ability to switch seamlessly between oblique transmitted light illumination for label-free contrast, widefield epifluorescence imaging, three-dimensional (3D) structured illumination microscopy (SIM), lattice light-sheet microscopy (LLSM), lattice light-sheet (LLS)-SIM, image scanning microscopy (ISM) and two-photon point-scanning microscopy for in vivo imaging. Notably, all modes can apply adaptive optical (AO) direct wavefront sensing8 correction to counter sample-induced aberrations and ensure optimal performance at depth. By reusing the same hardware and software across modes, MOSAIC reduces the overall cost and minimizes the spatial footprint relative to acquiring multiple separate instruments. It also enables quantitative comparisons of different imaging methods on the same biological sample and opens doors for adaptive imaging protocols that demand dynamically switching between modalities.ResultsMicroscope design and characterizationMOSAIC was conceptualized as an evolution of the LLSM design incorporating two-channel AO correction (AO-LLSM)9. The original instrument used a 0.65-NA excitation lens and a 1.1-NA detection lens along with a 0.8-NA epifluorescence objective to view the sample from below10. While this arrangement optimized 3D spatial resolution, it also placed severe constraints on working distance and specimen geometries, typically limiting samples to a 5-mm diameter coverslip positioned between the objectives. For MOSAIC, we opted for a 0.6-NA excitation, a 1.0-NA detection and a coverslip-corrected 1.0-NA inverted epifluorescence objective (Supplementary Fig. 1a). While entailing a modest reduction in resolution11, these choices increase the effective working distance between the light-sheet objectives and the coverslip from near zero to 330 µm (Supplementary Fig. 1a), thereby accommodating 25-mm diameter coverslips for specimen mounting with an unobstructed scan range across its entire area. The two 1.0-NA detection objectives enable high-resolution, millimeter-scale FOV investigations across a variety of imaging modes (Supplementary Figs. 1b–e and 2–4). All three objectives and the specimen are housed within an environmental chamber with full control over temperature, perfusion and CO2.

A fourth objective, also of 1.0 NA, resides in an upright station within MOSAIC for two-photon imaging of larger specimens such as live mice (Supplementary Fig. 1f).To maximize the excitation power needed for large FOV and fast light-sheet imaging, MOSAIC uses a Powell/cylindrical lens combination (Supplementary Fig. 2a), concentrating laser power onto the long, narrow region of the spatial light modulator (SLM) to which the LLS pattern is written. Any combination of seven visible lasers can be used simultaneously, as we first split their post-Powell lens overlapping light sheets with a stack of parallel dichroic mirrors into seven independent parallel stripes on the SLM (Supplementary Fig. 5a). This facilitates independent phase pattern control of light-sheet tip, tilt and AO correction for each laser. After reflection off the SLM, the seven colors return through the same dichroic stack, recombining into an overlapped, collinear multicolor beam (Supplementary Fig. 5a,b), which reflects off a sample-conjugate resonant galvo (to reduce sample-induced shadowing artifacts9) and two pupil-conjugate galvos (for LLS axial alignment and lateral dither) before entering the excitation objective (Supplementary Fig. 2a). The resulting volumetric FOV (without tiling) for LLSM increases from 100 × ~30 × 100 µm3 in the previous implementation to 200 × ~30 × 20,000 µm3 in MOSAIC, and the imaging speed is more than doubled by the focused excitation and the use of two cameras (Supplementary Figs. 2b and 6b) for simultaneous two-color imaging.AO-LLSM requires a substantial investment in hardware (for example, objectives, lasers, galvos, SLM, deformable mirror (DM), control electronics, sample stages and cameras). MOSAIC extracts maximal value from this investment by repurposing these components as needed for multiple imaging modes, all with AO correction: LLSM and LLS-SIM (Supplementary Figs. 2a,b and 6b), oblique transmitted light illumination (Supplementary Figs. 2c and 6c), widefield and 3D-SIM (Supplementary Figs. 2d and 6d), ISM (Supplementary Figs.

3a and 6e), two-photon Bessel beam light-sheet (Supplementary Figs. 3c and 7a,b), two-photon point-scanning microscopy (Supplementary Figs. 4a and 7d, in both inverted and upright configurations) and point-scanning photostimulation (Supplementary Figs. 4b and 7f). Optical toggles (Supplementary Fig. 5c) reconfigure the light path between modes in 2–5 s or reconfigure the imaging cameras as wavefront sensors during wavefront measurement. The entire instrument, including lasers, fits within a 1 × 1 × 1 m3 volume, promoting alignment stability (Supplementary Video 1). Pre-centered mirrors and lenses with no adjustment (Supplementary Fig. 5d) simplify assembly and alignment of MOSAIC by eliminating unnecessary degrees of freedom, further contributing to stability. Characterizations of the optical path as well as the experimentally measured point-spread function (PSF) and optical transfer function (OTF) for each mode are provided in Supplementary Figs. 2–4 and 8.Long-term multimodal imaging of cultured cell growth over vast fields of viewEven clonal populations of cultured cells exhibit considerable structural and functional heterogeneity12. A more comprehensive understanding of such cells requires mapping their phenotypic diversity by imaging many cells over multiple rounds of division within a single experiment. Using MOSAIC and the noninvasiveness of LLSM, we volumetrically imaged a ~1 × 0.75 mm2 field of LLC-PK1 cells stably expressing markers for the endoplasmic reticulum (ER) and nucleus at ~260 × 260 × 431 nm3 nominal resolution every 90 s over ~24 h (984 time points; Fig. 1a,b and Supplementary Video 2). The entire 49-TB dataset, which contains ~1.5 million nuclear image volumes (Supplementary Fig. 9a), was acquired at a peak rate of 4 TB h−1 and processed by PetaKit5D13.Fig. 1: Large field-of-view dynamic imaging with 3D lattice light-sheet or 2D label-free oblique illumination microscopy.

The alternative text for this image may have been generated using AI.Full size imagea, LLSM image from Supplementary Video 2 of LLC-PK1 cells expressing Calnexin–mEmerald (ER) and H2B–mCherry (nuclei). Maximum-intensity-projections (MIPs) show xy (top) and xz (bottom) views of the 1,000 × 750 × 10 µm3 volume. Scale bar, 50 µm. b, Zoomed-in views of cell division events from a. Magenta panel (left), volume rendering of a metaphase cell having a large ER protrusion; blue panel (middle), four points in nominal cell division from metaphase to telophase; yellow panel (right), six points during a rare tripolar mitotic event. Scale bar, 10 µm. c, Label-free OI imaging at 1 Hz (Supplementary Video 3) captures HeLa cell lamellipodial ruffling and replication of contaminating bacteria. Scale bar, 10 µm. Right, Magnified view (magenta box) of three bacterial division events (arrows). Scale bar, 2 µm. d, Tiled label-free imaging over a 1,218 × 975 µm2 field of live U2OS cells at 1 Hz. Scale bar, 200 µm. Right, zoomed-in view (white box) of a single dividing U2OS cell at four points, showing condensed chromosomes (arrows). Scale bar, 10 µm (right).The cell population plateaued upon reaching confluency ~12 h from initial imaging (Supplementary Fig. 9b and Supplementary Video 2), after which patches of cells exhibited occasional tissue-like behaviors such as epithelial detachment (Fig. 1a) and reattachment. While the great majority of mitotic events followed the archetypal pattern (Fig. 1b), we also identified atypical events such as a dividing cell with a large apical ER extrusion (Fig. 1b), and an example of tripolar mitosis. Such divisions have been observed previously13,14, but MOSAIC offers a more detailed four-dimensional (4D) view of this process, and its noninvasiveness enabled us to study the daughter cells through to their ultimate apoptosis (Fig.