Inferring single-cell spatial gene expression with tissue morphology via explainable deep learning.

The spatial arrangement of cells is vital in developmental processes and organogenesis in multicellular life forms. Deep learning models trained with spatial omics data uncover complex patterns and relationships among cells, genes, and proteins in a high-dimensional space, providing new insights into biological processes and diseases. State-of-the-art in silico spatial multi-cell gene expression methods using histological images of tissue stained with hematoxylin and eosin (H&E) to characterize cellular heterogeneity. These computational techniques offer the advantage of analyzing vast amounts of spatial data in a scalable and automated manner, thereby accelerating scientific discovery and enabling more precise medical diagnostics and treatments. In this work, we developed a vision transformer (ViT) framework to map histological signatures to spatial single-cell transcriptomic signatures, named SPiRiT ( S patial Omics P rediction and R eproducibility integrated T ransformer). Our framework was enhanced by integrating cross validation with model interpretation during hyper-parameter tuning. SPiRiT predicts single-cell spatial gene expression using the matched histopathological image tiles of human breast cancer and whole mouse pup, evaluated by Xenium (10x Genomics) datasets. Furthermore, ViT model interpretation reveals the high-resolution, high attention area (HAR) that the ViT model uses to predict the gene expression, including marker genes for invasive cancer cells ( FASN ), stromal cells ( POSTN ), and lymphocytes ( IL7R ). In an apple-to-apple comparison with the ST-Net Convolutional Neural Network algorithm, SPiRiT improved predictive accuracy by 40% using human breast cancer Visium (10x Genomics) dataset. Cancer biomarker gene prediction and expression level are highly consistent with the tumor region annotation. In summary, our work highlights the feasibility to infer spatial single-cell gene expression using tissue morphology in multiple-species, i.e., human and mouse, and multi-organs, i.e., mouse whole body morphology. Importantly, incorporating model interpretation and vision transformer is expected to serve as a general-purpose framework for spatial transcriptomics.

Feedback in the ß-catenin destruction complex imparts bistability and cellular memory.

Wnt ligands are considered classical morphogens, for which the strength of the cellular response is proportional to the concentration of the ligand. Herein, we show an emergent property of bistability arising from feedback among the Wnt destruction complex proteins that target the key transcriptional co-activator ß-catenin for degradation. Using biochemical reconstitution, we identified positive feedback between the scaffold protein Axin and the kinase glycogen synthase kinase 3 (GSK3). Theoretical modeling of this feedback between Axin and GSK3 suggested that the activity of the destruction complex exhibits bistable behavior. We experimentally confirmed these predictions by demonstrating that cellular cytoplasmic ß-catenin concentrations exhibit an "all-or-none" response with sustained memory (hysteresis) of the signaling input. This bistable behavior was transformed into a graded response and memory was lost through inhibition of GSK3. These findings provide a mechanism for establishing decisive, switch-like cellular response and memory upon Wnt pathway stimulation.

Cellular morphological features are predictive markers of cancer cell state.

Even genetically identical cells have heterogeneous properties because of stochasticity in gene or protein expression. Single cell techniques that assay heterogeneous properties would be valuable for basic science and diseases like cancer, where accurate estimates of tumor properties is critical for accurate diagnosis and grading. Cell morphology is an emergent outcome of many cellular processes, potentially carrying information about cell properties at the single cell level. Here we study whether morphological parameters are sufficient for classification of single cells, using a set of 15 cell lines, representing three processes: (i) the transformation of normal cells using specific genetic mutations; (ii) metastasis in breast cancer and (iii) metastasis in osteosarcomas. Cellular morphology is defined as quantitative measures of the shape of the cell and the structure of the actin. We use a toolbox that calculates quantitative morphological parameters of cell images and apply it to hundreds of images of cells belonging to different cell lines. Using a combination of dimensional reduction and machine learning, we test whether these different processes have specific morphological signatures and whether single cells can be classified based on morphology alone. Using morphological parameters we could accurately classify cells as belonging to the correct class with high accuracy. Morphological signatures could distinguish between cells that were different only because of a different mutation on a parental line. Furthermore, both oncogenesis and metastasis appear to be characterized by stereotypical morphology changes. Thus, cellular morphology is a signature of cell phenotype, or state, at the single cell level.

Using flow cytometry and multistage machine learning to discover label-free signatures of algal lipid accumulation.

Most applications of flow cytometry or cell sorting rely on the conjugation of fluorescent dyes to specific biomarkers. However, labeled biomarkers are not always available, they can be costly, and they may disrupt natural cell behavior. Label-free quantification based upon machine learning approaches could help correct these issues, but label replacement strategies can be very difficult to discover when applied labels or other modifications in measurements inadvertently modify intrinsic cell properties. Here we demonstrate a new, but simple approach based upon feature selection and linear regression analyses to integrate statistical information collected from both labeled and unlabeled cell populations and to identify models for accurate label-free single-cell quantification. We verify the method's accuracy to predict lipid content in algal cells (Picochlorum soloecismus) during a nitrogen starvation and lipid accumulation time course. Our general approach is expected to improve label-free single-cell analysis for other organisms or pathways, where biomarkers are inconvenient, expensive, or disruptive to downstream cellular processes.

TISMorph: A tool to quantify texture, irregularity and spreading of single cells.

A number of recent studies have shown that cell shape and cytoskeletal texture can be used as sensitive readouts of the physiological state of the cell. However, utilization of this information requires the development of quantitative measures that can describe relevant aspects of cell shape. In this paper we develop a toolbox, TISMorph, that calculates a set of quantitative measures to address this need. Some of the measures introduced here have been used previously, while others are new and have desirable properties for shape and texture quantification of cells. These measures, broadly classifiable into the categories of textural, irregularity and spreading measures, are tested by using them to discriminate between osteosarcoma cell lines treated with different cytoskeletal drugs. We find that even though specific classification tasks often rely on a few measures, these are not the same between all classification tasks, thus requiring the use of the entire suite of measures for classification and discrimination. We provide detailed descriptions of the measures, as well as the TISMorph package to implement them. Quantitative morphological measures that capture different aspects of cell morphology will help enhance large-scale image-based quantitative analysis, which is emerging as a new field of biological data.

Cell Form and Function: Interpreting and Controlling the Shape of Adherent Cells.

Beautiful images of animal cells cultured on surfaces are ubiquitous in biological research, but these shapes also carry valuable information about the cells and the organism that they came from. Cell morphology is an emergent property of the cellular phenotype as well as of the physiological and signaling state of the cell. Many functional changes in cells cause stereotypical changes in cellular morphology, and some changes in shape can also cause characteristic changes in cellular phenotype. Thus, controlling cell shape through substrate engineering may emerge as another mechanism to modulate cell function for human health. This review summarizes current understanding of the morphology-phenotype connection, and surveys progress in the effort to interpret and control cell morphology.

Force Spectrum Microscopy Using Mitochondrial Fluctuations of Control and ATP-Depleted Cells.

A single-cell assay of active and passive intracellular mechanical properties of mammalian cells could give significant insight into cellular processes. Force spectrum microscopy (FSM) is one such technique, which combines the spontaneous motion of probe particles and the mechanical properties of the cytoskeleton measured by active microrheology using optical tweezers to determine the force spectrum of the cytoskeleton. A simpler and noninvasive method to perform FSM would be very useful, enabling its widespread adoption. Here, we develop an alternative method of FSM using measurement of the fluctuating motion of mitochondria. Mitochondria of the C3H-10T1/2 cell line were labeled and tracked using confocal microscopy. Mitochondrial probes were selected based on morphological characteristics, and their mean-square displacement, creep compliance, and distributions of directional change were measured. We found that the creep compliance of mitochondria resembles that of particles in viscoelastic media. However, comparisons of creep compliance between controls and cells treated with pharmacological agents showed that perturbations to the actomysoin network had surprisingly small effects on mitochondrial fluctuations, whereas microtubule disruption and ATP depletion led to a significantly decreased creep compliance. We used properties of the distribution of directional change to identify a regime of thermally dominated fluctuations in ATP-depleted cells, allowing us to estimate the viscoelastic parameters for a range of timescales. We then determined the force spectrum by combining these viscoelastic properties with measurements of spontaneous fluctuations tracked in control cells. Comparisons with previous measurements made using FSM revealed an excellent match.

Measuring systematic changes in invasive cancer cell shape using Zernike moments.

We study the shape characteristics of osteosarcoma cancer cell lines on surfaces of differing hydrophobicity using Zernike moments to represent cell shape. We compare the shape characteristics of four invasive cell lines with a corresponding less-invasive parental line on three substrates. Cell shapes of each pair of cell lines are quite close and display overlapping characteristics. To quantitatively study shape changes in high-dimensional parameter space we project down to principal component space and define a vector that summarizes average shape differences. Using this vector we find that three of the four pairs of cell lines show similar changes in shape, while the fourth pair shows a very different pattern of changes. We find that shape differences are sufficient to enable a neural network to classify cells accurately as belonging to the highly invasive or the less invasive phenotype. The patterns of shape changes were also reproducible for repetitions of the experiment. We also find that shape changes on different substrates show similarities between the eight cells studied, but the differences were typically not enough to permit classification. Our paper strongly suggests that shape may provide a means to read out the phenotypic state of some cell types, and shape analysis can be usefully performed using a Zernike moment representation.

Changes in cell shape are correlated with metastatic potential in murine and human osteosarcomas.

Metastatic cancer cells for many cancers are known to have altered cytoskeletal properties, in particular to be more deformable and contractile. Consequently, shape characteristics of more metastatic cancer cells may be expected to have diverged from those of their parental cells. To examine this hypothesis we study shape characteristics of paired osteosarcoma cell lines, each consisting of a less metastatic parental line and a more metastatic line, derived from the former by in vivo selection. Two-dimensional images of four pairs of lines were processed. Statistical analysis of morphometric characteristics shows that shape characteristics of the metastatic cell line are partly overlapping and partly diverged from the parental line. Significantly, the shape changes fall into two categories, with three paired cell lines displaying a more mesenchymal-like morphology, while the fourth displaying a change towards a more rounded morphology. A neural network algorithm could distinguish between samples of the less metastatic cells from the more metastatic cells with near perfect accuracy. Thus, subtle changes in shape carry information about the genetic changes that lead to invasiveness and metastasis of osteosarcoma cancer cells.