Bo-Jui Chang

Somatic oncogenic mutations are typically defined by their molecular alterations, yet how they reorganize cellular architecture within intact tissues remain largely unknown. Here, we demonstrate that distinct oncogenic drivers produce unique multiscale architectural phenotypes that can be quantitatively resolved in intact liver tissue. Using iterative expansion microscopy, multiscale light-sheet imaging, and three-dimensional morphometric analysis, we systematically mapped structural remodeling from single-cell morphology to mitochondrial architecture in mosaic mouse models of hepatocellular oncogene activation. NRAS and CTNNB1 induced fundamentally different morphological programs. NRAS activation drove extensive remodeling of cell morphology, membrane curvature, and surface irregularity, whereas CTNNB1 activation largely preserved global cell morphology while selectively altering mitochondrial organization and shape. We first established a zonation-aware reference state for interpreting oncogene-associated organelle remodeling by resolving mitochondrial differences between periportal and pericentral hepatocytes. Within this framework, CTNNB1 activation shifted mitochondrial features toward a pericentral-like state, consistent with the role of Wnt/β-catenin signaling in hepatic zonation and metabolic identity. Furthermore, integrating cellular morphology, membrane geometry, and mitochondrial architecture improved discrimination of oncogenic states beyond any individual structural feature, demonstrating that mutation-specific phenotypes arise through coordinated remodeling across multiple biological scales. Together, these findings establish multiscale structural phenotyping as a framework for linking oncogenic genotype to three-dimensional cellular organization and reveal that distinct oncogenic drivers remodel different architectural compartments during liver oncogene activation.

For over three decades, we have known that oncogenic RAS alters the actin cytoskeleton organization and cell surface morphology. RAS activates the GTPase RAC1, which triggers the growth of branched actin networks to promote cell membrane protrusions. In melanoma, the hyperactive RAC1 mutant, Rac1P29S, was recently shown to drive extended lamellipodia, which then empower cell proliferation through sequestration and localized inhibition of the merlin tumor suppressor. This discovery illustrates cell morphological programs not only as outputs but also as regulators of human oncogenic signals. Hence, we wondered whether the pronounced branched actin-driven membrane protrusions (BAMPs) downstream of oncogenic RAS are not mere outputs of RAS signaling but rather an active component in mediating the oncogenic penetrance of RAS mutants. We used volumetric light sheet microscopy and biochemical approaches to investigate the role of BAMPs in regulating the molecular signaling of oncogenic KRAS in pancreatic and lung cancer models. We found that elevated BAMP formation regulated the interaction of oncogenic KRAS with downstream effectors, specifically with the RAC1 GEF TIAM1. This implies that BAMPs amplify their own upstream regulators in a positive feedback. This meritorious cycle upregulates cyclin D1 expression by inactivating the merlin tumor suppressor, independently of the mitogen activated protein kinase pathway (MAPK). In the absence of BAMPs, cells carrying oncogenic KRAS mutations are unable to attain their full penetrance in proliferation. Overall, this work unveils the long-overlooked role of branched actin-driven cell morphology in the functionalization of KRAS mutants as potent oncogenes.

Despite advances in fluorescence microscopy, spectral overlap and limited resolution hinder the dense mapping of the cellular ultrastructure. To overcome these challenges, we developed Cy-ExM, a high-plex imaging strategy that integrates optimized cryo-fixation for antigen preservation, expansion microscopy, and iterative immunofluorescence labeling. Using oblique plane microscopy, we perform three-dimensional super-resolution imaging of 20 biological targets encompassing the full cellular volume of individual mammalian cells.

Cell segmentation is the foundation of a wide range of microscopy-based biological studies. Deep learning has revolutionized two-dimensional (2D) cell segmentation, enabling generalized solutions across cell types and imaging modalities. This has been driven by the ease of scaling up image acquisition, annotation and computation. However, three-dimensional (3D) cell segmentation, requiring dense annotation of 2D slices, still poses substantial challenges. Manual labeling of 3D cells to train broadly applicable segmentation models is prohibitive. Even in high-contrast images annotation is ambiguous and time-consuming. Here we develop a theory and toolbox, u-Segment3D, for 2D-to-3D segmentation, compatible with any 2D method generating pixel-based instance cell masks. u-Segment3D translates and enhances 2D instance segmentations to a 3D consensus instance segmentation without training data, as demonstrated on 11 real-life datasets, comprising >70,000 cells, spanning single cells, cell aggregates and tissue. Moreover, u-Segment3D is competitive with native 3D segmentation, even exceeding when cells are crowded and have complex morphologies.

The most ubiquitous form of aberration correction for microscopy is deconvolution; however, deconvolution relies on the assumption that the system’s point spread function is the same across the entire field of view. This assumption is often inadequate, but space-variant deblurring techniques generally require impractical amounts of calibration and computation. We present an imaging pipeline that leverages symmetry to provide simple and fast spatially varying deblurring. Our ring deconvolution microscopy method utilizes the rotational symmetry of most microscopes and cameras, and naturally extends to sheet deconvolution in the case of lateral symmetry. We derive theory and algorithms for ring deconvolution microscopy and propose a neural network based on Seidel aberration coefficients as a fast alternative. We demonstrate improvements in speed and image quality as compared to standard deconvolution and existing spatially varying deblurring across a diverse range of microscope modalities, including miniature microscopy, multicolor fluorescence microscopy, multimode fiber micro-endoscopy and light-sheet fluorescence microscopy. Our approach enables near-isotropic, subcellular resolution in each of these applications.

Cell segmentation is the foundation of a wide range of microscopy-based biological studies. Deep learning has revolutionized 2D cell segmentation, enabling generalized solutions across cell types and imaging modalities. This has been driven by the ease of scaling up image acquisition, annotation, and computation. However, 3D cell segmentation, requiring dense annotation of 2D slices still poses significant challenges. Manual labeling of 3D cells to train broadly applicable segmentation models is prohibitive. Even in high-contrast images annotation is ambiguous and time-consuming. Here we develop a theory and toolbox, u-Segment3D, for 2D-to-3D segmentation, compatible with any 2D method generating pixel-based instance cell masks. u-Segment3D translates and enhances 2D instance segmentations to a 3D consensus instance segmentation without training data, as demonstrated on 11 real-life datasets, >70,000 cells, spanning single cells, cell aggregates, and tissue. Moreover, u-Segment3D is competitive with native 3D segmentation, even exceeding when cells are crowded and have complex morphologies.

Projection imaging accelerates volumetric interrogation in fluorescence microscopy, but for multi-cellular samples, the resulting images may lack contrast, as many structures and haze are summed up. Here, we demonstrate rapid projective light-sheet imaging with parameter selection (props) of imaging depth, position and viewing angle. This allows us to selectively image different sub-volumes of a sample, rapidly switch between them and exclude background fluorescence. Here we demonstrate the power of props by functional imaging within distinct regions of the zebrafish brain, monitoring calcium firing inside muscle cells of moving Drosophila larvae, super-resolution imaging of selected cell layers, and by optically unwrapping the curved surface of a Drosophila embryo. We anticipate that props will accelerate volumetric interrogation, ranging from subcellular to mesoscopic scales.

Understanding the intricate interplay and inter-connectivity of biological processes across an entire organism is important in various fields of biology, including cardiovascular research, neuroscience, and developmental biology. Here, we present a mesoscopic oblique plane microscope (OPM) that enables whole organism imaging with high speed while revealing fine details such as endothelial nuclei. A microprism underneath the sample enhances the axial resolution and optical sectioning through total internal reflection of the light sheet. Through rapid refocusing of the light sheet, the imaging depth is extended up to threefold while keeping the axial resolution constant. Using low magnification objectives with a large field of view, we realize mesoscopic imaging over a volume of 3.7mm×1.5mm×1mm with ∼2.3µm lateral and ∼9.2µm axial resolution. Applying the mesoscopic OPM, we demonstrate in vivo and in toto whole organism imaging of the zebrafish vasculature and its endothelial nuclei, and blood flow dynamics at 12 Hz acquisition rate, resulting in a quantitative map of blood flow across the entire organism.

Most human cells require anchorage for survival. Cell–substrate adhesion activates diverse signalling pathways, without which cells undergo anoikis—a form of programmed cell death¹. Acquisition of anoikis resistance is a pivotal step in cancer disease progression, as metastasizing cells often lose firm attachment to surrounding tissue^2,3. In these poorly attached states, cells adopt rounded morphologies and form small hemispherical plasma membrane protrusions called blebs^{4,5,6,7,8,9,10,11}. Bleb function has been thoroughly investigated in the context of amoeboid migration, but it has been examined far less in other scenarios¹². Here we show by three-dimensional imaging and manipulation of cell morphological states that blebbing triggers the formation of plasma membrane-proximal signalling hubs that confer anoikis resistance. Specifically, in melanoma cells, blebbing generates plasma membrane contours that recruit curvature-sensing septin proteins as scaffolds for constitutively active mutant NRAS and effectors. These signalling hubs activate ERK and PI3K—well-established promoters of pro-survival pathways. Inhibition of blebs or septins has little effect on the survival of well-adhered cells, but in detached cells it causes NRAS mislocalization, reduced MAPK and PI3K activity, and ultimately, death. This unveils a morphological requirement for mutant NRAS to operate as an effective oncoprotein. Furthermore, whereas some BRAF-mutated melanoma cells do not rely on this survival pathway in a basal state, inhibition of BRAF and MEK strongly sensitizes them to both bleb and septin inhibition. Moreover, fibroblasts engineered to sustain blebbing acquire the same anoikis resistance as cancer cells even without harbouring oncogenic mutations. Thus, blebs are potent signalling organelles capable of integrating myriad cellular information flows into concerted cellular responses, in this case granting robust anoikis resistance.