UPCOMING

Biophysical Society Annual Meetings 2024
February 10-24, 2024
Philadelphia, PA

Evaluating rules of intracellular organization in human induced pluripotent stem cells using generative computational models via cellPACK
by Saurabh Mogre
Poster | Saturday, February 11, 2024 | 1:45pm EST

Evaluation and comparison of agent-based model frameworks for characterizing hiPSC shapes
by Jessica Yu
Poster | Saturday, February 11, 2024 | 1:45pm EST

Comparing spatial biophysical simulations across scales and methods
by Karthik Vegesna
Poster | Saturday, February 11, 2024 | 1:45pm EST

Illuminating dynamic cellular states in hiPSCs through endogenous gene tagging
by Brock Roberts
Poster | Saturday, February 11, 2024 | 1:45pm EST

Quantifying biological variability using appropriate data representations of microscopy images
by Ritvik Vasan
Flash Talk | Saturday, February 11, 2024 | 5:45pm EST
Poster | Sunday, February 12, 2024 | 1:45pm EST

Streamlining deep-learning-based segmentation methods for microscopy images
by Gideon Dunster
Poster | Monday, February 13, 2024 | 1:45pm EST

Advancing cell biology by unifying open-source modeling and facilitating broad community collaboration
by ​Megan Riel-Mehan
Talk | Monday, February 13, 2024 | 4:15pm EST

PAST

Cell Bio 2023 - an ASCB conference
December 2-6, 2023
Boston, MA

Evaluating rules of intracellular organization in human induced pluripotent stem cells using generative computational models via cellPACK
by Saurabh Mogre 

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how hiPSCs transition between states during differentiation and disease. To achieve this, we develop a variety of assays and computational tools that are openly available at allencell.org. Here we describe an iterative model development pipeline to test hypotheses for rules underlying intracellular organization in hiPSCs using peroxisomes and endosomes as initial structures of interest. cellPACK is an open-source software that builds 3D spatial models of the biological mesoscale, integrating entities ranging from ~10 to 10,000 nm in size. We leveraged the generative modeling capabilities of cellPACK to simulate hypothetical distributions of subcellular structures within experimentally measured cellular/nuclear shapes of hiPSCs. We applied our previously developed analysis framework for integrated intracellular organization to create parameterized intracellular location representations (PILRs) for peroxisomes and endosomes in hiPSCs as well as their simulated distributions generated by cellPACK. The PILR establishes a shape-independent intracellular coordinate system and enables interpretable quantitative comparisons between observed and simulated distributions of subcellular structures. Starting with peroxisomes, we evaluated a null hypothesis of random distribution due to a lack of any previously established organizational pattern. We compared the average PILRs for peroxisome distribution in hiPSCs with cellPACK simulations and found that a bias towards the nuclear membrane correlated best with observations. In contrast to peroxisomes, observations suggest endosomes may prefer localizing at the apical surface of cells. We used our analysis pipeline to investigate the directionality and strength of a potential apical-basal polarization following a similar approach for peroxisomes. Our unique multimodal data-driven approach lets us go from quantitative descriptors to generative rules of intracellular organization. Ongoing developments of the cellPACK toolkit will extend these analyses to subcellular structures with more complex geometries or interactions to uncover an expansive library of organizational rules.

Streamlining deep-learning based 3D segmentation and application-appropriate validation of quantitative image-based assays
​by ​Gideon Dunste

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. Segmentation is an important step to extract meaningful measurements of the organization of intracellular structures from 3D microscopy images. Segmentations need to be accurate, consistent, and reproducible to ensure biological interpretability across all aspects of a dataset (e.g., data collection days, microscope settings, experimental conditions). To address this challenge, we previously developed the Allen Cell & Structure Segmenter, a Python-based open-source toolkit for 3D segmentation. The first part of the Segmenter, a collection of classic image segmentation workflows together with a lookup table for workflow selection, was developed into a plugin for Napari to allow user-friendly access. The second part of the Segmenter is an iterative deep-learning workflow that begins with the segmentation outputs of the classic workflow and uses two straightforward “human-in-the-loop” curation strategies to convert these outputs into a set of 3D ground truth images for iterative model training without the need for manual painting. The network architectures were designed and tested specifically for 3D fluorescence microscope images. This deep-learning workflow is being developed into another Napari plugin with a user-friendly curation interface implemented using the open-source package CytoDL, which we are developing to simplify deep-learning experiment tracking and code maintenance. Next, the performance of the segmentation, or of other image analysis or deep-learning based image transformation algorithms, must be evaluated within the context of the specific biological application. There is no “one size fits all” workflow or software that, in one shot, covers all the interpretations a researcher may be targeting. Instead, biological interpretability of quantitative image-based assays requires “application-appropriate” validation. We are developing standardized approaches to streamline, and eventually automate this type of validation to guide biology researchers how to identify application-appropriate validation metrics for their specific interpretation goals.

Accelerating Cell Biology with Scientific Illustration & User Experience Design
by ​Thao Do

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. We utilize an open science approach to develop a variety of assays, data, and computational tools available at allencell.org. As an in-house Scientific Illustrator and User Experience (UX) Designer, I create visually compelling and accurate presentations of complex scientific concepts, facilitate multidisciplinary collaboration, and optimize the experience of using and accessing our science and tools. I will demonstrate how we use best methods in scientific illustration (SI) to produce effective communication material and to translate abstract mental models into concrete visual aids, which benefits content experts during the early conceptualization stages of writing grants or papers, providing a means to clarify, organize, and articulate their thoughts more effectively. Our UX design process involves similar interactions, but with more diverse groups of researchers, engineers, project managers, and users throughout a project’s lifecycle, which I will compare and contrast with SI. We employ several UX methods to gather valuable insight about the tools and our users to inform our design decisions and guide the development process. I will highlight transferable skills between these two disciplines that can open doors to exciting opportunities. I will also show how each field has its own strengths that can enhance the outcomes of the other, such as making UX interactions more engaging by applying the storytelling and aesthetic skill of an illustrator or approaching the audience from a UX perspective when drafting visualizations. I will use examples from our website throughout the talk to illustrate the importance of deliberate, flexible design and workflows, as well as the necessity of clear guidance and communicative media to ensure accessibility and usability of our data and practices for all users, regardless of their current skill level or resources. I will conclude by sharing upcoming initiatives, such as integrating UX best practices into institute projects beyond our UX team and enhancing the efficiency of our SI process amidst a continuously evolving landscape of tools.

A practical approach to establishing a conceptual framework for holistic cell states and state transitions
by ​Susanne Rafelski

​The cell is distinct from other complex systems in that it is the basic unit of life. In multicellular organisms, the same genome can give rise to many cell phenotypes, and for convenience these have often been classified as distinct cell types. Furthermore, a single cell type can be described as existing in multiple states (e.g., an epithelial cell can be quiescent, motile, apoptotic, etc.). Determining robust definitions for cell states is a long-standing challenge due to the heterogeneity of individual cells within a population and the continuous nature of variation in cell state landscapes. Traditionally, cell state was defined based primarily on what cells looked
like, where they were located, and what function they exhibited. In the molecular era, the observables defining cell state expanded to include specific molecular markers, localizations of proteins within cells, and cellular structure-function relationships. Then in the post-genomic era, large-scale gene expression profiles and gene regulatory networks were folded in, resulting the current, largely molecular-focused, view of cell state. Throughout this history, cell state has always been defined by the cell phenotype (cell-intrinsic observables) and its environment (cell-extrinsic observables). Moving forward, we anticipate that the range of observables used to define cell states will evolve again, as single-cell imaging and analytics are advancing at a breakneck pace via the collection of large-scale, systematic cell image datasets and the application of quantitative image-based data science methods. Now is the time to integrate the spatiotemporal observables of the physical structure and organization of the cell with molecular observables, and to update the concept of a holistic cell state. We propose that cell observables can be conveniently divided into four categories related to the common modalities of data collection: 1) cell molecular census (all of the molecules in a cell), 2) cell organization (the spatio-temporal multi-scale arrangement of these molecules), 3) cell function (the actions and behaviors of a cell), with these three together representing the cell-intrinsic phenotype, and finally 4) the cell-extrinsic environment. In many cases, there is clear bi-directional feedback between all combinations of these four categories. We thus define a stable holistic cell state as one where these four categories of data observables of a cell can fluctuate, but ultimately mutually reinforce one another to maintain a stable state. We hypothesize that, in a cell state transition, at least one type of observable falls out of alignment with the others and then, due to the bi-directional feedback between observables, influences changes in the others to achieve a new mutually reinforcing stable state.

A fully annotated, quality controlled single cell 3D image dataset of hiPSCs from the Allen Cell Collection for exploration and reuse
by ​Nathalie Gaudreault

​The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease. We take advantage of 3D microscopy images of the Allen Cell Collection (www.allencell.org), a collection of endogenous fluorescently tagged hiPSC lines. These cell lines, all created using WTC-11 as the parental line, underwent extensive quality control testing to ensure genomic, cell biological, and stem cell integrity. As a first step to assess the cell-to-cell variability in the organization of distinct organelles, we previously collected 3D images of these gene-edited hiPSCs from 25 cell lines, each representing a particular cellular organelle or structure using spinning disk confocal microscopes to create the WTC-11 hiPSC Single-Cell Image Dataset v1. Version 2 of this dataset includes several additional structure-tagged cell lines and experimental metadata to facilitate its reuse and exploration. This annotated dataset now contains nearly 300,000 single-cell images of gene-edited hiPSCs and is available through the Cell Feature Explorer (https://cfe.allencell.org/) for exploration and an open data access platform for download. We provide 2D transmitted light images of the cell culture wells showing the colonies from which the cells were imaged for context on the environment and neighboring cells, 3D high-resolution images of the fields-of-view, and 3D single cell images and corresponding segmentation data for each cell, its nucleus, and one other specific cellular structure. These images can be used as distinct training and mining datasets to explore and answer a range of cell biological questions about hiPSCs. The high-resolution, context-rich information along with pixel data and related metadata for cell line characteristics or cell culture conditions make this dataset ideal for studying the effects of cell cycle (via cell volume and mitotic stage annotations), cell confluency, passage number, and colony formation on cellular organization. This updated dataset also includes a complete description of all annotation fields, guidance on how to use these to interrogate different subsets of the dataset, and case studies of how this dataset can be reused to answer cell biology questions.

Integrating high resolution 3D live cell imaging and agent-based modeling to explore rules driving human induced pluripotent stem cell shape dynamics
by ​Jessica Yu

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. Our dataset (open.quiltdata.com/b/allencell/tree/aics/hipsc_single_cell_image_dataset) comprises over 200,000 3D live hiPSC images (allencell.org) that quantify cell shape variation within colonies. To better understand the underlying mechanisms that drive observed cell shapes, we integrate imaging data with iterative development of modular computational models of cell shape and dynamics. We use agent-based models (ABMs), an intuitive, bottom-up modeling framework that comprises autonomous agents (cells) following rules (hypothesized biological mechanisms) that drive their actions and interactions. We extend an existing ABM framework (github.com/bagherilab/ARCADE) with a 3D Cellular Potts Model (CPM) to represent cell and nuclear shapes as collections of voxels. The model uses simple rules guiding cell proliferation and apoptosis with cell phase durations and size distributions directly parameterized from literature values and image datasets. This voxel-based approach enables us to initialize simulations directly from images and transform outputs into simulated images that can be further analyzed via the same analysis techniques used on imaging data. To support open science and reproducibility, the model framework, simulation, and analysis pipelines are fully cloud-based and open-source. We simulate the ABM with different combinations of rules and compare the simulated size distributions and shape modes to those extracted from the hiPSC image dataset. We observe that the baseline CPM–comprising cell-cell adhesion, volume, and surface area rules–is not sufficient to produce observed cell size and shape distributions. Simulating additional rules–such as height and substrate adhesion–differentially impact emergent distributions. By comparing the similarities and differences between simulated and observed distributions across different rule sets, we gain insight into the mechanisms represented by the ABM rules and how those mechanisms drive observed cell shapes. These insights inform iterative model development and bolster both hypothesis testing and generation.

Integrating spatial biophysical simulations across scales and methods using Simularium and Vivarium
by ​Blair Lyons

​The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease (allencell.org). The Simularium project aims to involve the wider biology community, especially wet lab biologists, in building and analyzing spatial mechanistic simulations at different levels of scale, and connecting them in the context of whole cells. We have composed multiple types of models that simulate objects in 3D space at multiple time and length scales and with different representations, in order to compare their behavior and outputs. To connect these models we have built interfaces using the Vivarium simulation framework, which allows for the flexible composition of many different models, and we are aiming to build multiscale models with adaptive resolution and calculation methods. To demonstrate this approach for actin dynamics, we have built interfaces between the community software engines ReaDDy, Cytosim, and MEDYAN, so that the composed model can be configured to simulate actin fibers either with coarse-grained monomers or filament lines. We have defined simple models that can translate between each of these simulation engines as unit tests for physical properties that are required for more complex models of endocytosis and leading edge dynamics of crawling cells. These test models explore how actin properties like bend and twist coupling, persistence length, and force generation differ when simulated with different methods and in different parameter regimes. We visualize the outputs interactively in 3D using the web-based Simularium viewer to explore differences between models and the implications for reproducible modeling of actin filaments across scales and methods.

Integrating spatial biophysical simulations across scales and methods
by ​Karthik Vegesna

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease (allencell.org). To enable modeling of complex cellular processes, we have identified and tested use cases for interoperating biophysical simulations of cytoskeletal components across spatiotemporal scales and modeling methods with an initial focus on modeling a role of actin in clathrin-mediated endocytosis. We have developed a framework to compare the behavior of actin filaments modeled by two different simulation engines: Cytosim and ReaDDy. Cytosim specializes in cytoskeleton simulations, representing actin fibers as splines of connected Brownian particles. While effective for modeling large systems involving flexible filaments and associated proteins, Cytosim lacks the ability to capture the impact of filament twist, a feature critical to force production in 3D cellular processes. On the other hand, ReaDDy, a generalized particle-based reaction-diffusion simulator, can be used to model actin filaments as coarse-grained monomers, providing greater detail and granularity in capturing filament twist, albeit at a higher computational cost. In this experiment, we simulated the compression of an actin filament across a range of compression velocities using both simulation engines and developed simulator-independent metrics (including coplanarity and filament twist) to determine which engine is more appropriate to use in different experimental conditions. To support open science and reproducibility, the model frameworks, simulations, and analysis pipelines are fully cloud-based and open-source. To compare approaches, we implemented dimensionality reduction via Principal Component Analysis and PaCMAP on the dynamics of the actin filaments and observed clustering between simulation engines and compression velocities. This allowed us to identify parameter regimes where both simulation engines reflect similar low-dimensional embeddings, implying that the two simulators can be used interchangeably within this range of parameters. We anticipate extending these analyses to compare simulation results with experimental cryo-electron tomography data that captures the behavior of actin filaments in situ, which will more stringently inform appropriate use cases for each simulator. This framework for systematic comparison between modeling frameworks enables researchers to make informed decisions about their choice of simulator in integrative modeling projects and supports the development of robustly designed multi-scale models of cell dynamics.

Quantifying the morphological variability of nuclear structures using conditional variational autoencoders and appropriate data representations
by ​Ritvik Vasan

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. To do this, we take advantage of 3D microscopy images of the Allen Cell Collection (www.allencell.org), a collection of endogenous fluorescently tagged hiPSC lines, each representing a particular cellular organelle or structure. A major challenge in cell biology is being able to fully describe the morphological variability of intracellular structures using a small set of numbers that are interpretable and generative. We devised a strategy to accomplish this by using point clouds and signed distance functions as inputs to a rotation equivariant Variational Autoencoder (VAE), a deep learning framework that uses non-linear dimensionality reduction to explain the sources of variation in a dataset. We applied this framework to DNA replication foci (via PCNA) and nucleoli (via nucleophosmin) and found that we can get compact representations and high fidelity
reconstructions. Importantly, these learned latent representations capture known aspects of variability seen by eye (e.g., changes in total volume and number of nucleoli), and can be used to achieve good cell cycle classifications. We found that point clouds are an appropriate input data representation for punctate spatial distributions that are difficult to segment, whereas signed distance functions are appropriate for irregular shapes with variable numbers of pieces. With this validation, further analysis can more reliably measure the variability in structure organization within and across cell cycle stages. A challenge that emerges in this process is that it is hard to separate changes in nuclear shape from unique changes in spatial distribution of the structure of interest. To address this challenge, we trained a rotation equivariant VAE conditioned on nuclear shape and orientation. We validated the efficacy of this approach using a synthetic dataset with known coupling between structure organization and nuclear shape. Future work will extend these analyses beyond the nucleus, to intracellular structures with other characteristic geometries and interactions.

MegaSeg: A structure agnostic deep learning model for intracellular structure segmentation in 3D fluorescence microscopy images
by ​Suraj Mishra

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSC) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease. A key step towards this goal is the ability to quantify changes in the size and shape of cells and subcellular
structures. We have, therefore, developed image-based assays and segmentation algorithms taking advantage of high-resolution 3D images of cells from the Allen Cell Collection (www.allencell.org), a collection of endogenous fluorescently tagged hiPSC lines, each representing a particular cellular structure. Our primary strategy has been to develop segmentation workflows specific to each cellular structure. Even though effective for individual structures, these workflows typically do not generalize to segmentation targets of different shapes and sizes, cell types, perturbations, microscope modalities or resolutions/magnifications. To overcome such challenges, we took advantage of our extensive 3D live cell image data to develop MegaSeg, a structure agnostic deep learning-based framework for segmentation of intracellular structures in 3D fluorescence microscopy images. In MegaSeg, a single 3D convolutional neural network is trained end-to-end to segment multiple structures for robust target-agnostic deep feature extraction making the network generalizable to many different structures. We selected 14 cell lines with high quality segmentations available to train the current version of MegaSeg and evaluated its performance. We found that MegaSeg accurately segmented all 14 of these structures in unseen control data as well as over 10 additional intracellular structures. We also found MegaSeg to be robust to different cell types as well as changes in fluorescence intensity variations, both between and within images, which arise from changes to microscopy settings and experimental conditions. These results demonstrate the potential of MegaSeg in generating accurate segmentation of known and unknown intracellular structures from 3D fluorescence microscopy images over a wide range of experimental and imaging conditions. Future plans include extending MegaSeg to accurately segment structures captured by different microscope-types, and resolutions/magnifications.

Illuminating dynamic cellular states in hiPSCs through endogenous gene tagging
by ​Gaea Turman

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. We have generated a collection of endogenously tagged human induced pluripotent stem cell (hiPSC) lines to illuminate cell organization in various cell states, including stem cells, migratory cells, cardiomyocytes and endothelial cells. To date, the Allen Cell Collection consists of 56 single- or dual-edited lines that have undergone extensive quality control testing to ensure genomic, cell biological, and stem cell integrity. We have tagged commonly recognized membrane-bound and membrane-less cellular organelles, signaling complexes, phase transition markers, transcription factors, and markers of both the organizational and dynamic identity state of differentiating cells. Here, we highlight our gene-editing and quality control workflows for mono- and biallelic editing of expressed or silent genes that are induced specifically during differentiation and function as reporters of cellular state. Our recent efforts have focused on generating genome edited hiPSC lines in which rather than tag an endogenous protein we install signaling pathway reporters, with the canonical Wnt pathway as one example. We also underscore methods we have employed to improve our endogenous gene tagging workflow such as utilizing Adeno-Associated Virus 6 (AAV6) to deliver donor DNAs and multiplexed transfection strategies for gene tagging at multiple loci, enabling us to generate gene edited cell lines with greater efficiency. We have also increasingly relied on automation workflows for our cell culture and banking processes using the Hamilton Star system. Using this system, we have developed workflows for automating the production of cell line banks for distribution and developed the ability to passage gene edited cell lines including at early stages in the screening and scale-up of candidate clones. Finally, we
make our cell lines, the donors used to generate them, thousands of segmented single cell 3D images of our lines, analysis and visualization tools, integrated cell models and biological findings openly available to the research community for at www.allencell.org.

Re-orientation and subcellular re-organization of human induced pluripotent stem cell-derived endothelial cells in response to shear stress
by ​Becky Zaunbrecher

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. We focus on endothelial cells (ECs) and differentiate endogenously tagged hiPSC lines from the Allen Cell Collection (www.allencell.org) into ECs for our studies. ECs play critical roles in normal vascular development and function and are central to the progression of many cardiovascular diseases. ECs are constantly subject to shear stress as blood flows through the vessels they line, which influences their structure, function, and morphology. hiPSCs-derived endothelial cells (hiPSC-ECs) exhibit sensitivity to shear stress caused by fluid flow similar to that observed in primary and in vivo endothelial cells and have previously been shown to align in the direction of fluid flow. We found that the direction of alignment of hiPSC-ECs was dependent upon the magnitude of fluid shear stress applied to cells within physiological ranges. Low shear stresses (0.8-6 dyn/cm2) elicit alignment parallel to the direction of fluid flow, while high shear stress (15 dyn/cm2) causes the cells to align orthogonally to the direction of fluid flow. hiPSC-ECs exposed to intermediate shear stress levels (9 or 11 dyn/cm2) exhibit a mixed alignment phenotype. To enable studies of changes in morphology and subcellular organization in response to fluid shear stress, we have developed a method to segment the whole cell and nucleus of hiPSC-ECs to extract high quality information from 3D images of cells expressing a mEGFP-tagged protein representing a structure or organelle of interest. We are currently using this method in conjunction with immunofluorescence and live cell imaging of tagged hiPSC-ECs to investigate how the cytoskeleton, focal adhesions, polarity, and adherens junctions of these cells change with applications of different fluid shear stresses. We believe that understanding the relationships between the environmental cue of fluid shear stress, orientation and morphology of whole hiPSC-ECs, and subcellular organization of those cells will enable greater understanding of endothelial cell biology.

Cell organization and basement membrane dynamics during the epithelial to mesenchymal transition (EMT) from 2D and 3D human induced pluripotent stem cell (hiPSC) cultures
by ​Leigh Harris

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures and how cells transition between states during differentiation and disease. The epithelial to mesenchymal transition (EMT) is a state change that occurs in normal embryo development and in pathological contexts such as cancer metastasis. An EMT reminiscent of primitive streak formation can be induced in a cell culture assay by activating the WNT pathway in hiPSCs, initiating mesodermal differentiation. In this assay, colonies first contract before cells migrate. We found that this early contraction happens because of dilute Matrigel (reconstituted basement membrane), which is included in typical EMT induction conditions. In fact, we found that Matrigel alone induces hiPSC colonies to zipper closed in a purse-string-like mechanism, resulting in an acinar, edgeless epithelium surrounding a central lumen. These “lumenoids” grow as hollow spheres anchored to the glass coverslip, enabling multi-day live imaging in 96-well format. We can therefore study the EMT state transition using a comparative approach, generating hiPSCs in two different geometric states – 2D colonies and 3D lumenoids – and observing EMT from both. By 3D live imaging of fluorescently-tagged structures in lines from our publicly available Allen Cell Collection (allencell.org) and forthcoming lines reporting EMT-related
transcription factor expression, we aim to analyze and quantify the dynamic interplay between cell environment, behavior, organization, and protein expression profile during EMT. Furthermore, to directly monitor basement membrane dynamics, we can label hiPSC-derived collagen IV during live imaging using a fluorescently-labeled human-specific collagen IV antibody. We find that the addition of Matrigel leads to de novo synthesis of an hiPSC-derived basement membrane. Dynamic networks and fibrils of collagen IV form a 3D shell around growing lumenoids, and shell break-down can be monitored during EMT, including the formation and expansion of holes as cells cross the basement membrane. These studies will provide insight into the dynamic interplay between cells and the extracellular microenvironment during EMT, moving us toward a more holistic understanding of cell state transitions.

Characterizing cell states by integrating cell behavior, organization, and molecular census in migratory cells from the epithelial to mesenchymal transition (EMT) of human induced pluripotent stem cells (hiPSCs)
by ​Nivedita Nivedita

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSC) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease. The epithelial to mesenchymal transition (EMT) is a state change that occurs in both normal contexts such as development and pathological contexts such as cancer metastasis. The EMT has been described as a behavioral change from largely non-motile to migratory, a change in organization from apical-basal to front-rear polarity and cell shape from tall to flat and elongated. These changes are closely associated with a shift in cell protein expression profile. However, the dependencies and relationships between these aspects of EMT are not well understood and require a large-scale, multi-modal data integration approach. We have standardized an imaging pipeline to generate 3D live cell time-lapse images of hiPSC from Allen Cell Collection (allencell.org) prior to 4i (Iterative Indirect Immunofluorescence Imaging) 14-plex immunostaining. Based on the comprehensive data obtained from this pipeline, we have developed a generalizable quantification framework to investigate the temporal dynamics and ergodicity of migratory cell behavior. We found that migratory cell behavior during EMT can be grouped into categories based on cell movement, location with respect to the colony, and cell-cell interactions. Additionally, we incorporated an elastic metric of deformation into the displacement metric to account for changes in cell shape. Furthermore, our pipeline allows us to explore the correlation between the different categories of cell behavior and the observed heterogeneous protein expression amongst the migrating cells. We have identified several proteins with heterogeneous expression within the migratory cells (e.g., Sox2, LEF1, BiP , WNT5a, Eomes, DKK1, etc). For example, we only detected Sox2 expression in slow-migrating cells, but not in the fast-moving cells. We are now using this pipeline to determine whether cell behavior is a predictor of protein expression. Overall, we believe that our approach to define the states of hiPSC-derived migrating cells as an integration of cell behavior, cell organization and cell molecular census, could serve as a model to define cell states more generally.

Growth dynamics and size homeostasis of nuclei in growing human induced pluripotent stem cell (hiPSC) colonies
by ​Chantelle Leveille

The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how they transition between states during differentiation and disease. Starting with the nucleus, we asked how hiPSCs control growth of this key cellular structure. We took advantage of the mEGFP-tagged laminB1 line from the Allen Cell Collection of endogenous tagged hiPSC lines (allencell.org). We imaged live, growing colonies in 3D for 48 hours and developed an image analysis workflow providing highly resolved quantitative features for nearly 1000 full cell cycle tracks that can be traced over generations. Examining volume over time, we found that the nucleus undergoes two distinct phases of growth regulation, an early and late growth phase, with a transition point occurring in G1 at a consistent time and size. During the late growth phase, nuclei approximately double their sizes. We tested known models of cell size control on our nuclei and found that the volume added over time is constant. These results and the small variability in nuclear to cytoplasmic volume ratio suggest an adder-like mechanism of growth control. Examining the contribution of growth rate and duration to this mechanism, we observed a volume compensation behavior via tuning of the interphase duration to achieve size homeostasis. Using a linear regression model, we found that the volume at the transition point during early growth and the timing of nuclear breakdown within the overall colony development time in late growth are the features most predictive of interphase duration. These results highlight a balance between the importance of early nuclear size and colony development in determining nuclear growth duration. We also explored the role of lineage and found that sister pairs have highly correlated early volumes and interphase durations compared to unrelated nuclei born at the same time. However, by the end of the nuclear growth trajectory, related volumes fell within the variation of the overall population. This data-driven framework for tracking nuclei throughout the cell cycle and across generations has enabled us to quantitatively
characterize the roles of potential sources of variation in nuclear growth and size control in the dynamic environment of the growing colony.

Advancing cell biology by harnessing open-source modeling and facilitating broad community collaboration
by ​Graham Johnson, Jessica Yu, & Blair Lyons

​The Allen Institute for Cell Science aims to understand the principles by which human induced pluripotent stem cells (hiPSCs) establish and maintain robust dynamic localization of cellular structures, and how cells transition between states during differentiation and disease (allencell.org). We also strive to democratize tools that expedite research and foster community-driven biological discovery. Recognizing that the blue-sky vision of modeling dynamic living cells will demand a concerted effort, we are committed to catalyzing this with reusable models, extensible frameworks, usable tools, and broad community collaboration and training. To understand mechanisms underlying cell shape dynamics, we integrate imaging data with iterative model development in a framework that combines an agent-based model (ABM) of agents (cells) following rules (theoretical biological mechanisms) with a 3D Cellular Potts Model (CPM) of cell shape (github.com/bagherilab/ARCADE). Simulating the ABM with varying rule sets, we compare size and shape distributions with distributions calculated directly from our dataset
of over 200,000 3D live hiPSC images. Preliminary simulations reveal that basic CPM rules—cell-cell adhesion, volume, and surface area—don't reproduce observed distributions, while additional rules, like substrate adhesion, impact emergent distributions differently. Comparing simulated and observed distributions across rule sets, we gain insights into mechanisms driving cell shapes to inform model development and hypothesis generation. The Simularium project (simularium.allencell.org) collaborates with model and framework developers and users, to engage the broader biology community in building and analyzing spatial mechanistic simulations. Our user-friendly, open-source viewer allows easy sharing and exploration of 3D biological simulations in a web browser. We have begun prototyping integrations with Vivarium (vivarium-collective.github.io) to create multiscale and multi-method models, with interfaces enabling dynamic transitioning between various open-source simulation packages. Our goal is to develop multiscale models with adaptive resolution and calculation methods, facilitating comparisons between models, enhancing reproducibility, and ultimately fostering collaborations between wet lab biologists and computational modelers. We will outline plans to couple these efforts and resources with training and education, collaborating with the community to shape the next generation of virtual cells.

---

Seattle Cell Symposium 2022

​​SEATTLE CELL SYMPOSIUM - 2022
​
Day 1: January 27, 2022
Featuring: Ru Gunawardane & Susanne Rafelski, Lucas Pelkmans, Chantell Evans, Brian Beliveau, EMT team, Johannes Schöneberg, & Devin Schweppe

​SEATTLE CELL SYMPOSIUM - 2022
​
Day 2: January 28, 2022
Featuring: Graham Johnson, David Goodsell, Sanja Vickovic, Shila Ghazanfar, Sabine Petry, Matthew Akamatsu, & Nuclear Morphogenesis team 

---

Seattle Cell Symposium 2020

​SEATTLE CELL SYMPOSIUM - 2020

Day 1: December 17, 2020
Featuring: Susanne Rafelski, Geeta Narlikar, Alexandra Zidovska, Arjun Raj, Abby Buchwalter Cool, Ting Wu, GW Gant Luxton & Daniel Starr, Megan King & Simon Mochrie, Katharine Ullman & Maho Niwa, & Chris Frick

​SEATTLE CELL SYMPOSIUM - 2020

Day 2: December 18, 2020
Featuring: Ru Gunawardane, Susan Parkhurst, Derek Applewhite, Stem Cell State team, Tools Lightning talks, David Van Valen, Martin Kampmann, Ron Vale

---

Seattle Cell Symposium 2019

​SEATTLE CELL SYMPOSIUM - 2019
​
Julie Theriot, University of Washington
Cooperation, competition & conviction in decision-making for motile cells

​SEATTLE CELL SYMPOSIUM - 2019
​
Alex Paredez, University of Washington
Cytoskeletal innovations for sticking around

​​SEATTLE CELL SYMPOSIUM - 2019

Melissa Hendershott & Rory Donovan-Maiye, Allen Institute for Cell Science
Cell states beyond transcriptomics: integrating structural organization & gene expression in cardiomyocytes

​SEATTLE CELL SYMPOSIUM - 2019
​
Barry Gumbiner, Seattle Children's Research Institute
What cell biological insights into cadherin regulation reveal about disease processes

​SEATTLE CELL SYMPOSIUM - 2019
​
Barbara Wakimoto, University of Washington
Bending the rules of organelle biogenesis: views from a highly specialized cell

​SEATTLE CELL SYMPOSIUM - 2019
​
Claudia Moreno, University of Washington
Better together: Insights into the clustering of ion channels

​SEATTLE CELL SYMPOSIUM - 2019

Steve Henikoff, Fred Hutchinson Cancer Research Center
Genome-wide mapping of protein-DNA interaction dynamics

 ​SEATTLE CELL SYMPOSIUM - 2019

Megan Riel-Mehan & Chris Frick, Allen Institute for Cell Science
Integrated cells: moving through mitosis and into the nucleus 

---

Seattle Cell Symposium 2018

​SEATTLE CELL SYMPOSIUM - 2018

Rick Horwitz, Allen Institute for Cell Science
Welcome & opening remarks

​SEATTLE CELL SYMPOSIUM - 2018

Keiko Torii, University of Washington
Cellular decision making during stomatal patterning & differentiation in plants

​SEATTLE CELL SYMPOSIUM - 2018

Clemens Cabernard, University of Washington
The molecular cell biology and mechanics of asymmetric cell division

​SEATTLE CELL SYMPOSIUM - 2018
​
Kami Ahmad, Fred Hutchinson Cancer Research Center
High-throughput chromatin profiling in mutants, tissues, and cells

​SEATTLE CELL SYMPOSIUM - 2018
​
​Lisa Maves, Seattle Children's Research Institute
Using zebrafish to understand how to build a heart

​SEATTLE CELL SYMPOSIUM - 2018
​
Kaytlyn Gerbin, Allen Institute for Cell Science Team Talk
Mapping cell state from pluripotency to differentiation

​SEATTLE CELL SYMPOSIUM - 2018
​
Jeff Ranish, Institute for Systems Biology
Proteomic approaches for studying the architecture of macromolecular machines and gene regulatory networks

​SEATTLE CELL SYMPOSIUM - 2018
​
Yasemin Sancak, University of Washington
Function and regulation of mitochondrial calcium uptake

​SEATTLE CELL SYMPOSIUM - 2018
​
Jesse Zalatan, University of Washington
Biochemical mechanisms for kinase signaling in the Wnt pathway: What are the scaffold proteins really doing?

​SEATTLE CELL SYMPOSIUM - 2018
​
Jianxu Chen, Allen Institute for Cell Science
A new open source toolkit for segmenting 3D intracellular structure in microscopy images

​SEATTLE CELL SYMPOSIUM - 2018
​
David Russell, University of Washington
Universal donor stem cells

---

Seattle Cell Symposium 2017

​SEATTLE CELL SYMPOSIUM - 2017

Rick Horwitz, Allen Institute for Cell Science
Welcome and opening remarks

​SEATTLE CELL SYMPOSIUM - 2017

​David Baker, University of Washington
The coming of age of de novo protein design

​SEATTLE CELL SYMPOSIUM - 2017

​Bill Noble, University of Washington
Using Hi-C to interrogate and model dynamic nuclear 3D architecture

​SEATTLE CELL SYMPOSIUM - 2017

​Chuck Murry, UW Medicine
Stem cells as building blocks for tissue repair

​SEATTLE CELL SYMPOSIUM - 2017

​John Scott, University of Washington
Exploring and exploiting the spatial constraints of local signaling

​SEATTLE CELL SYMPOSIUM - 2017

​Rhishikesh Bargaje, Institute for Systems Biology
Single-cell analysis of cell fate branching: from computational pattern recognition to cell state dynamics

​SEATTLE CELL SYMPOSIUM - 2017

​Jennifer Nemhauser, University of Washington
Plant logic

​SEATTLE CELL SYMPOSIUM - 2017

​Amanda Haupt, Allen Institute for Cell Science
Endogenous gene tagging illuminates cell organization

​SEATTLE CELL SYMPOSIUM - 2017

​David Rawlings, Seattle Children's Research Institute
Engineering human primary cells for novel clinical applications

​SEATTLE CELL SYMPOSIUM - 2017

​Irina Mueller, Allen Institute for Cell Science
Cell organization during human iPS cell mitosis

​SEATTLE CELL SYMPOSIUM - 2017

​David William, Allen Institute for Cell Science
Modeling at the Allen Institute for Cell Science: I want to see a bit more of that

​SEATTLE CELL SYMPOSIUM - 2017

Tom Daniel, University of Washington
Computational models of flows inside muscle cells: do sarcomeres breathe?