DEVELOPMENT OF MULTISCALE AND MULTIDIMENSIONAL LIGHT SHEET MICROSCOPY FOR IMAGE-BASED MODELLING OF ORGANOGENESIS

Behavioral analysis of individual cells, in the form of structural phenotypes arising in response to mechanical stimuli or genomic alterations as small molecule intermediates of disease, are instrumental for understanding organogenesis. In this regard, dynamic cell tracking or temporal changes in cell shape, area and size are vital for modelling tissue interactions. However, optical diffraction limits imposed by lenses, hinders high resolution volume reconstruction of microscale biological processes and subcellular organelles. Thereby, isotropic intensity is restricted to limited focal volumes, reducing imaging distance to less than 1 mm for high magnification. To combat this, Light Sheet Microscopes (LSM) have emerged as powerful optical sectioning tools, owing to millisecond camera frame rates and individual cell localization for imaging distances beyond 1 mm. However, current LSM resolution and imaging distance are highly dependent on the numerical aperture (NA)/magnification of the objective lens. In AIM 1, we achieved near isotropic, micron-scale 3D resolution across tunable imaging distance (up to 7mm) using a very low magnification (2x), to decouple fundamental relationship of lens NA on resolution and imaging distance. Using a 1x macro detection lens with 2x optical zoom, we refocused 120 micron long optical section across 6500 micrometer distance, in synchrony with modulation of active camera pixels. Further, a 3d pixel resolution algorithm was used to enhance voxel resolution ~3x in lateral perspective and ~5x in axial perspective. Consequently, we quantified 2.5 micrometer resolution in XY plane and 3.65 microns in XZ plane, from native resolution of 8.2 +/ 0.8 micron in XY plane and axial resolution of 17.25 +/-1.75 micron in XZ plane. AS-sLSM was used to reconstruct micro-structures in optically cleared whole mice organs. With regards to multi-dimensional imaging in live animal models, LSM in conjunction with optically transparent zebrafish, has advanced the frontiers of in vivo characterization of evolutionarily conserved developmental signaling pathways. Hence, we have developed a multidimensional (3D, 3D+time) LSM in AIM 2, capable of tunable imaging distance up to 1 millimeter, to characterize organogenesis in smaller animal models at cellular scale. Like AIM 1, we sought to remove dependance of lens NA on resolution and field-of-view for 3D imaging, and hence integrated oblique angle-based sample detection for 3D imaging to apply sub voxel, resolution enhancement algorithm. We successfully quantified isotropic 3D pixel resolution of 2 microns +/- 0.3 microns in the XY plane and 2.5 microns +/- 0.4 microns in the XZ plane using 4X magnification lens, for optical section of 500 micrometer propagation distance. For 4D imaging in AIM 2, traditional orthogonal sample detection was integrated with in silico volume reconstruction algorithm. We quantified XY resolution of 4.25 +/ 0.7 microns and YZ resolution of 9.75 +/ 1.1 microns using 4x magnification lens for optical section of 500-micron propagation length, using single side illumination for 4D mode. Using the 4D imaging pipeline, we reconstructed zebrafish myocardial nuclei (~5-15 microns) across distinct stages of heart maturation. In addition, a feature detector based on Difference of Gaussian (DoG) filter integrated with watershed algorithm, was implemented to localize individual nuclei. AIM 3 is a continuation of AIM 2, wherein we have applied our multi-dimensional LSM to quantify regional anisotropy in zebrafish ventricular nuclei mechanical deformation and morphogenesis. Hessian Difference of Gaussian (HDoG) feature detector from AIM 2 was used to localize and quantify 159 ± 13, 222 ± 17, 260 ± 13, and 284 ± 10 nuclei across 48-120 hours post birth stages respectively. Furthermore, we observed nuclei surface areas ranging from 29 +/- 9 um^2 – 47.45 +/-10 um^2 for spherical nuclei near inflow, in contrast to larger rod-shaped nuclei ranging from 28 +/- 15 um^2 – 57.25 +/- 25 um^2 in the equatorial region near ventricular inflow across 48 – 120 hours post birth. In addition, we quantified dynamic mechanical deformation experienced by individual nuclei across the cardiac cycle, by tracking segmented nuclei centers of mass to calculate area ratio. In this regard, we did not observe any significant differences in deformation across different ventricular regions across 48-120 hours post birth, between the inflow, equatorial and apex regions. Lastly, three-dimensionsional ventricular lumen segmented by HDoG feature detector was used to train a neural net based on semantic segmentation, to automate zebrafish ventricular volume quantification. Consequently, our studies demonstrate the potential of LSM as a user-friendly, multidimensional (3D,3D+time) imaging toolbox, for in vivo reconstruction of cellular phenotypes, across diverse spectrum of preclinical biological specimens.