All Publications


  • Deep learning for cellular image analysis. Nature methods Moen, E., Bannon, D., Kudo, T., Graf, W., Covert, M., Van Valen, D. 2019

    Abstract

    Recent advances in computer vision and machine learning underpin a collection of algorithms with an impressive ability to decipher the content of images. These deep learning algorithms are being applied to biological images and are transforming the analysis and interpretation of imaging data. These advances are positioned to render difficult analyses routine and to enable researchers to carry out new, previously impossible experiments. Here we review the intersection between deep learning and cellular image analysis and provide an overview of both the mathematical mechanics and the programming frameworks of deep learning that are pertinent to life scientists. We survey the field's progress in four key applications: image classification, image segmentation, object tracking, and augmented microscopy. Last, we relay our labs' experience with three key aspects of implementing deep learning in the laboratory: annotating training data, selecting and training a range of neural network architectures, and deploying solutions. We also highlight existing datasets and implementations for each surveyed application.

    View details for DOI 10.1038/s41592-019-0403-1

    View details for PubMedID 31133758

  • NF-kappaB signaling dynamics is controlled by a dose-sensing autoregulatory loop. Science signaling DeFelice, M. M., Clark, H. R., Hughey, J. J., Maayan, I., Kudo, T., Gutschow, M. V., Covert, M. W., Regot, S. 2019; 12 (579)

    Abstract

    Over the last decade, multiple studies have shown that signaling proteins activated in different temporal patterns, such as oscillatory, transient, and sustained, can result in distinct gene expression patterns or cell fates. However, the molecular events that ensure appropriate stimulus- and dose-dependent dynamics are not often understood and are difficult to investigate. Here, we used single-cell analysis to dissect the mechanisms underlying the stimulus- and dose-encoding patterns in the innate immune signaling network. We found that Toll-like receptor (TLR) and interleukin-1 receptor (IL-1R) signaling dynamics relied on a dose-dependent, autoinhibitory loop that rendered cells refractory to further stimulation. Using inducible gene expression and optogenetics to perturb the network at different levels, we identified IL-1R-associated kinase 1 (IRAK1) as the dose-sensing node responsible for limiting signal flow during the innate immune response. Although the kinase activity of IRAK1 was not required for signal propagation, it played a critical role in inhibiting the nucleocytoplasmic oscillations of the transcription factor NF-kappaB. Thus, protein activities that may be "dispensable" from a topological perspective can nevertheless be essential in shaping the dynamic response to the external environment.

    View details for PubMedID 31040261

  • NF-kappa B signaling dynamics is controlled by a dose-sensing autoregulatory loop SCIENCE SIGNALING DeFelice, M. M., Clark, H. R., Hughey, J. J., Maayan, I., Kudo, T., Gutschow, M. V., Covert, M. W., Regot, S. 2019; 12 (579)
  • Techniques for Studying Decoding of Single Cell Dynamics FRONTIERS IN IMMUNOLOGY Jeknic, S., Kudo, T., Covert, M. W. 2019; 10
  • Escalating Threat Levels of Bacterial Infection Can Be Discriminated by Distinct MAPK and NF-kappa B Signaling Dynamics in Single Host Cells CELL SYSTEMS Lane, K., Andres-Terre, M., Kudo, T., Monack, D. M., Covert, M. W. 2019; 8 (3): 183-+
  • Escalating Threat Levels of Bacterial Infection Can Be Discriminated by Distinct MAPK and NF-kappaB Signaling Dynamics in Single Host Cells. Cell systems Lane, K., Andres-Terre, M., Kudo, T., Monack, D. M., Covert, M. W. 2019

    Abstract

    During an infection, immune cells must identify the specific level of threat posed by a given bacterial input in order to generate an appropriate response. Given that they use a general non-self-recognition system, known as Toll-like receptors (TLRs), to detect bacteria, it remains unclear how they transmit information about a particular threat. To determine whether host cells can use signaling dynamics to transmit contextual information about a bacterial stimulus, we use live-cell imaging to make simultaneous quantitative measurements of host MAPK and NF-kappaB signaling, two key pathways downstream of TLRs, and bacterial infection and load. This combined, single-cell approach reveals that NF-kappaB and MAPK signaling dynamics are sufficient to discriminate between (1) pathogen-associated molecular patterns (PAMPs) versus bacteria, (2) extracellular versus intracellular bacteria, (3) pathogenic versus non-pathogenic bacteria, and (4) the presence or absence of features indicating an active intracellular bacterial infection, such as replication and effector secretion.

    View details for PubMedID 30904375

  • Techniques for Studying Decoding of Single Cell Dynamics. Frontiers in immunology Jeknić, S. n., Kudo, T. n., Covert, M. W. 2019; 10: 755

    Abstract

    Cells must be able to interpret signals they encounter and reliably generate an appropriate response. It has long been known that the dynamics of transcription factor and kinase activation can play a crucial role in selecting an individual cell's response. The study of cellular dynamics has expanded dramatically in the last few years, with dynamics being discovered in novel pathways, new insights being revealed about the importance of dynamics, and technological improvements increasing the throughput and capabilities of single cell measurements. In this review, we highlight the important developments in this field, with a focus on the methods used to make new discoveries. We also include a discussion on improvements in methods for engineering and measuring single cell dynamics and responses. Finally, we will briefly highlight some of the many challenges and avenues of research that are still open.

    View details for PubMedID 31031756

  • A convex 3D deconvolution algorithm for low photon count fluorescence imaging. Scientific reports Ikoma, H., Broxton, M., Kudo, T., Wetzstein, G. 2018; 8 (1): 11489

    Abstract

    Deconvolution is widely used to improve the contrast and clarity of a 3D focal stack collected using a fluorescence microscope. But despite being extensively studied, deconvolution algorithms can introduce reconstruction artifacts when their underlying noise models or priors are violated, such as when imaging biological specimens at extremely low light levels. In this paper we propose a deconvolution method specifically designed for 3D fluorescence imaging of biological samples in the low-light regime. Our method utilizes a mixed Poisson-Gaussian model of photon shot noise and camera read noise, which are both present in low light imaging. We formulate a convex loss function and solve the resulting optimization problem using the alternating direction method of multipliers algorithm. Among several possible regularization strategies, we show that a Hessian-based regularizer is most effective for describing locally smooth features present in biological specimens. Our algorithm also estimates noise parameters on-the-fly, thereby eliminating a manual calibration step required by most deconvolution software. We demonstrate our algorithm on simulated images and experimentally-captured images with peak intensities of tens of photoelectrons per voxel. We also demonstrate its performance for live cell imaging, showing its applicability as a tool for biological research.

    View details for PubMedID 30065270

  • Live-cell measurements of kinase activity in single cells using translocation reporters NATURE PROTOCOLS Kudo, T., Jeknic, S., Macklin, D. N., Akhter, S., Hughey, J. J., Regot, S., Covert, M. W. 2018; 13 (1): 155–69

    Abstract

    Although kinases are important regulators of many cellular processes, measuring their activity in live cells remains challenging. We have developed kinase translocation reporters (KTRs), which enable multiplexed measurements of the dynamics of kinase activity at a single-cell level. These KTRs are composed of an engineered construct in which a kinase substrate is fused to a bipartite nuclear localization signal (bNLS) and nuclear export signal (NES), as well as to a fluorescent protein for microscopy-based detection of its localization. The negative charge introduced by phosphorylation of the substrate is used to directly modulate nuclear import and export, thereby regulating the reporter's distribution between the cytoplasm and nucleus. The relative cytoplasmic versus nuclear fluorescence of the KTR construct (the C/N ratio) is used as a proxy for the kinase activity in living, single cells. Multiple KTRs can be studied in the same cell by fusing them to different fluorescent proteins. Here, we present a protocol to execute and analyze live-cell microscopy experiments using KTRs. We describe strategies for development of new KTRs and procedures for lentiviral expression of KTRs in a cell line of choice. Cells are then plated in a 96-well plate, from which multichannel fluorescent images are acquired with automated time-lapse microscopy. We provide detailed guidance for a computational analysis and parameterization pipeline. The entire procedure, from virus production to data analysis, can be completed in ∼10 d.

    View details for PubMedID 29266096

  • Competing memories of mitogen and p53 signalling control cell-cycle entry NATURE Yang, H., Chung, M., Kudo, T., Meyer, T. 2017; 549 (7672): 404-+

    Abstract

    Regulation of cell proliferation is necessary for immune responses, tissue repair, and upkeep of organ function to maintain human health. When proliferating cells complete mitosis, a fraction of newly born daughter cells immediately enter the next cell cycle, while the remaining cells in the same population exit to a transient or persistent quiescent state. Whether this choice between two cell-cycle pathways is due to natural variability in mitogen signalling or other underlying causes is unknown. Here we show that human cells make this fundamental cell-cycle entry or exit decision based on competing memories of variable mitogen and stress signals. Rather than erasing their signalling history at cell-cycle checkpoints before mitosis, mother cells transmit DNA damage-induced p53 protein and mitogen-induced cyclin D1 (CCND1) mRNA to newly born daughter cells. After mitosis, the transferred CCND1 mRNA and p53 protein induce variable expression of cyclin D1 and the CDK inhibitor p21 that almost exclusively determines cell-cycle commitment in daughter cells. We find that stoichiometric inhibition of cyclin D1-CDK4 activity by p21 controls the retinoblastoma (Rb) and E2F transcription program in an ultrasensitive manner. Thus, daughter cells control the proliferation-quiescence decision by converting the memories of variable mitogen and stress signals into a competition between cyclin D1 and p21 expression. We propose a cell-cycle control principle based on natural variation, memory and competition that maximizes the health of growing cell populations.

    View details for PubMedID 28869970

  • Measuring Signaling and RNA-Seq in the Same Cell Links Gene Expression to Dynamic Patterns of NF-?B Activation. Cell systems Lane, K., Van Valen, D., DeFelice, M. M., Macklin, D. N., Kudo, T., Jaimovich, A., Carr, A., Meyer, T., Pe'er, D., Boutet, S. C., Covert, M. W. 2017; 4 (4): 458-469 e5

    Abstract

    Signaling proteins display remarkable cell-to-cell heterogeneity in their dynamic responses to stimuli, but the consequences of this heterogeneity remain largely unknown. For instance, the contribution of the dynamics of the innate immune transcription factor nuclear factor κB (NF-κB) to gene expression output is disputed. Here we explore these questions by integrating live-cell imaging approaches with single-cell sequencing technologies. We used this approach to measure both the dynamics of lipopolysaccharide-induced NF-κB activation and the global transcriptional response in the same individual cell. Our results identify multiple, distinct cytokine expression patterns that are correlated with NF-κB activation dynamics, establishing a functional role for NF-κB dynamics in determining cellular phenotypes. Applications of this approach to other model systems and single-cell sequencing technologies have significant potential for discovery, as it is now possible to trace cellular behavior from the initial stimulus, through the signaling pathways, down to genome-wide changes in gene expression, all inside of a single cell.

    View details for DOI 10.1016/j.cels.2017.03.010

    View details for PubMedID 28396000

  • Deep Learning Automates the Quantitative Analysis of Individual Cells in Live-Cell Imaging Experiments. PLoS computational biology Van Valen, D. A., Kudo, T., Lane, K. M., Macklin, D. N., Quach, N. T., DeFelice, M. M., Maayan, I., Tanouchi, Y., Ashley, E. A., Covert, M. W. 2016; 12 (11)

    Abstract

    Live-cell imaging has opened an exciting window into the role cellular heterogeneity plays in dynamic, living systems. A major critical challenge for this class of experiments is the problem of image segmentation, or determining which parts of a microscope image correspond to which individual cells. Current approaches require many hours of manual curation and depend on approaches that are difficult to share between labs. They are also unable to robustly segment the cytoplasms of mammalian cells. Here, we show that deep convolutional neural networks, a supervised machine learning method, can solve this challenge for multiple cell types across the domains of life. We demonstrate that this approach can robustly segment fluorescent images of cell nuclei as well as phase images of the cytoplasms of individual bacterial and mammalian cells from phase contrast images without the need for a fluorescent cytoplasmic marker. These networks also enable the simultaneous segmentation and identification of different mammalian cell types grown in co-culture. A quantitative comparison with prior methods demonstrates that convolutional neural networks have improved accuracy and lead to a significant reduction in curation time. We relay our experience in designing and optimizing deep convolutional neural networks for this task and outline several design rules that we found led to robust performance. We conclude that deep convolutional neural networks are an accurate method that require less curation time, are generalizable to a multiplicity of cell types, from bacteria to mammalian cells, and expand live-cell imaging capabilities to include multi-cell type systems.

    View details for DOI 10.1371/journal.pcbi.1005177

    View details for PubMedID 27814364

    View details for PubMedCentralID PMC5096676

  • Laguerre Filter Analysis with Partial Least Square Regression Reveals a Priming Effect of ERK and CREB on c-FOS Induction PLOS ONE Kudo, T., Uda, S., Tsuchiya, T., Wada, T., Karasawa, Y., Fujii, M., Saito, T. H., Kuroda, S. 2016; 11 (8)

    Abstract

    Signaling networks are made up of limited numbers of molecules and yet can code information that controls different cellular states through temporal patterns and a combination of signaling molecules. In this study, we used a data-driven modeling approach, the Laguerre filter with partial least square regression, to describe how temporal and combinatorial patterns of signaling molecules are decoded by their downstream targets. The Laguerre filter is a time series model used to represent a nonlinear system based on Volterra series expansion. Furthermore, with this approach, each component of the Volterra series expansion is expanded by Laguerre basis functions. We combined two approaches, application of a Laguerre filter and partial least squares (PLS) regression, and applied the combined approach to analysis of a signal transduction network. We applied the Laguerre filter with PLS regression to identify input and output (IO) relationships between MAP kinases and the products of immediate early genes (IEGs). We found that Laguerre filter with PLS regression performs better than Laguerre filter with ordinary regression for the reproduction of a time series of IEGs. Analysis of the nonlinear characteristics extracted using the Laguerre filter revealed a priming effect of ERK and CREB on c-FOS induction. Specifically, we found that the effects of a first pulse of ERK enhance the subsequent effects on c-FOS induction of treatment with a second pulse of ERK, a finding consistent with prior molecular biological knowledge. The variable importance of projections and output loadings in PLS regression predicted the upstream dependency of each IEG. Thus, a Laguerre filter with partial least square regression approach appears to be a powerful method to find the processing mechanism of temporal patterns and combination of signaling molecules by their downstream gene expression.

    View details for DOI 10.1371/journal.pone.0160548

    View details for Web of Science ID 000381381100043

    View details for PubMedID 27513954

    View details for PubMedCentralID PMC4981404

  • A method to rapidly create protein aggregates in living cells NATURE COMMUNICATIONS Miyazaki, Y., Mizumoto, K., Dey, G., Kudo, T., Perrino, J., Chen, L., Meyer, T., Wandless, T. J. 2016; 7

    Abstract

    The accumulation of protein aggregates is a common pathological hallmark of many neurodegenerative diseases. However, we do not fully understand how aggregates are formed or the complex network of chaperones, proteasomes and other regulatory factors involved in their clearance. Here, we report a chemically controllable fluorescent protein that enables us to rapidly produce small aggregates inside living cells on the order of seconds, as well as monitor the movement and coalescence of individual aggregates into larger structures. This method can be applied to diverse experimental systems, including live animals, and may prove valuable for understanding cellular responses and diseases associated with protein aggregates.

    View details for DOI 10.1038/ncomms11689

    View details for Web of Science ID 000376669800001

    View details for PubMedID 27229621

    View details for PubMedCentralID PMC4894968

  • Controlling low rates of cell differentiation through noise and ultrahigh feedback. Science Ahrends, R., Ota, A., Kovary, K. M., Kudo, T., Park, B. O., Teruel, M. N. 2014; 344 (6190): 1384-1389

    Abstract

    Mammalian tissue size is maintained by slow replacement of de-differentiating and dying cells. For adipocytes, key regulators of glucose and lipid metabolism, the renewal rate is only 10% per year. We used computational modeling, quantitative mass spectrometry, and single-cell microscopy to show that cell-to-cell variability, or noise, in protein abundance acts within a network of more than six positive feedbacks to permit pre-adipocytes to differentiate at very low rates. This reconciles two fundamental opposing requirements: High cell-to-cell signal variability is needed to generate very low differentiation rates, whereas low signal variability is needed to prevent differentiated cells from de-differentiating. Higher eukaryotes can thus control low rates of near irreversible cell fate decisions through a balancing act between noise and ultrahigh feedback connectivity.

    View details for DOI 10.1126/science.1252079

    View details for PubMedID 24948735

    View details for PubMedCentralID PMC4733388

  • Controlling low rates of cell differentiation through noise and ultrahigh feedback SCIENCE Ahrends, R., Ota, A., Kovary, K. M., Kudo, T., Park, B. O., Teruel, M. N. 2014; 344 (6190): 1384-1389

    Abstract

    Mammalian tissue size is maintained by slow replacement of de-differentiating and dying cells. For adipocytes, key regulators of glucose and lipid metabolism, the renewal rate is only 10% per year. We used computational modeling, quantitative mass spectrometry, and single-cell microscopy to show that cell-to-cell variability, or noise, in protein abundance acts within a network of more than six positive feedbacks to permit pre-adipocytes to differentiate at very low rates. This reconciles two fundamental opposing requirements: High cell-to-cell signal variability is needed to generate very low differentiation rates, whereas low signal variability is needed to prevent differentiated cells from de-differentiating. Higher eukaryotes can thus control low rates of near irreversible cell fate decisions through a balancing act between noise and ultrahigh feedback connectivity.

    View details for DOI 10.1126/science.1252079

    View details for Web of Science ID 000337531700037

    View details for PubMedCentralID PMC4733388