Emerging machine learning approaches to phenotyping cellular motility and morphodynamics.

Cells respond heterogeneously to molecular and environmental perturbations. Phenotypic heterogeneity, wherein multiple phenotypes coexist in the same conditions, presents challenges when interpreting the observed heterogeneity. Advances in live cell microscopy allow researchers to acquire an unprecedented amount of live cell image data at high spatiotemporal resolutions. Phenotyping cellular dynamics, however, is a nontrivial task and requires machine learning (ML) approaches to discern phenotypic heterogeneity from live cell images. In recent years, ML has proven instrumental in biomedical research, allowing scientists to implement sophisticated computation in which computers learn and effectively perform specific analyses with minimal human instruction or intervention. In this review, we discuss how ML has been recently employed in the study of cell motility and morphodynamics to identify phenotypes from computer vision analysis. We focus on new approaches to extract and learn meaningful spatiotemporal features from complex live cell images for cellular and subcellular phenotyping.

Negative regulation of hypoxic responses via induced Reptin methylation.

Lysine methylation within histones is crucial for transcriptional regulation and thus links chromatin states to biological outcomes. Although recent studies have extended lysine methylation to nonhistone proteins, underlying molecular mechanisms such as the upstream signaling cascade that induces lysine methylation and downstream target genes modulated by this modification have not been elucidated. Here, we show that Reptin, a chromatin-remodeling factor, is methylated at lysine 67 in hypoxic conditions by the methyltransferase G9a. Methylated Reptin binds to the promoters of a subset of hypoxia-responsive genes and negatively regulates transcription of these genes to modulate cellular responses to hypoxia.

Sumoylation and Its Contribution to Cancer.

Post-translational modifications play an important role in regulating protein activity by altering their functions. Sumoylation is a highly dynamic process which is tightly regulated by a fine balance between conjugating and deconjugating enzyme activities. It affects intracellular localization and their interaction with their binding partners, thereby changing gene expression. Consequently, these changes in turn affect signaling mechanisms that regulate many cellular functions, such as cell growth, proliferation, apoptosis , DNA repair , and cell survival. It is becoming apparent that deregulation in the SUMO pathway contributes to oncogenic transformation by affecting sumoylation/desumoylation of many oncoproteins and tumor suppressors. Loss of balance between sumoylation and desumoylation has been reported in a number of studies in a variety of disease types including cancer. This chapter summarizes the mechanisms and functions of the deregulated SUMO pathway affecting oncogenes and tumor suppressor genes.

Interpretable Fine-Grained Phenotypes of Subcellular Dynamics via Unsupervised Deep Learning.

Uncovering fine-grained phenotypes of live cell dynamics is pivotal for a comprehensive understanding of the heterogeneity in healthy and diseased biological processes. However, this endeavor poses significant technical challenges for unsupervised machine learning, requiring the extraction of features that not only faithfully preserve this heterogeneity but also effectively discriminate between established biological states, all while remaining interpretable. To tackle these challenges, a self-training deep learning framework designed for fine-grained and interpretable phenotyping is presented. This framework incorporates an unsupervised teacher model with interpretable features to facilitate feature learning in a student deep neural network (DNN). Significantly, an autoencoder-based regularizer is designed to encourage the student DNN to maximize the heterogeneity associated with molecular perturbations. This method enables the acquisition of features with enhanced discriminatory power, while simultaneously preserving the heterogeneity associated with molecular perturbations. This study successfully delineated fine-grained phenotypes within the heterogeneous protrusion dynamics of migrating epithelial cells, revealing specific responses to pharmacological perturbations. Remarkably, this framework adeptly captured a concise set of highly interpretable features uniquely linked to these fine-grained phenotypes, each corresponding to specific temporal intervals crucial for their manifestation. This unique capability establishes it as a valuable tool for investigating diverse cellular dynamics and their heterogeneity.

Roles of sumoylation of a reptin chromatin-remodelling complex in cancer metastasis.

Defining the functional modules within transcriptional regulatory factors that govern switching between repression and activation events is a central issue in biology. Recently, we have reported the dynamic role of a beta-catenin-reptin chromatin remodelling complex in regulating a metastasis suppressor gene KAI1 (ref.1), which is capable of inhibiting the progression of tumour metastasis. Here, we identify signalling factors that confer repressive function on reptin and hence repress the expression of KAI1. Biochemical purification of a reptin-containing complex has revealed the presence of specific desumoylating enzymes that reverse the sumoylation of reptin that underlies its function as a repressor. Desumoylation of reptin alters the repressive function of reptin and its association with HDAC1. Furthermore, the sumoylation status of reptin modulates the invasive activity of cancer cells with metastatic potential. These data clearly define a functional model and provide a novel link for SUMO modification in cancer metastasis.

Transcriptional regulation of a metastasis suppressor gene by Tip60 and beta-catenin complexes.

Defining the molecular strategies that integrate diverse signalling pathways in the expression of specific gene programmes that are critical in homeostasis and disease remains a central issue in biology. This is particularly pertinent in cancer biology because downregulation of tumour metastasis suppressor genes is a common occurrence, and the underlying molecular mechanisms are not well established. Here we report that the downregulation of a metastasis suppressor gene, KAI1, in prostate cancer cells involves the inhibitory actions of beta-catenin, along with a reptin chromatin remodelling complex. This inhibitory function of beta-catenin-reptin requires both increased beta-catenin expression and recruitment of histone deacetylase activity. The coordinated actions of beta-catenin-reptin components that mediate the repressive state serve to antagonize a Tip60 coactivator complex that is required for activation; the balance of these opposing complexes controls the expression of KAI1 and metastatic potential. The molecular mechanisms underlying the antagonistic regulation of beta-catenin-reptin and the Tip60 coactivator complexes for the metastasis suppressor gene, KAI1, are likely to be prototypic of a selective downregulation strategy for many genes, including a subset of NF-kappaB target genes.

A deep learning-based segmentation pipeline for profiling cellular morphodynamics using multiple types of live cell microscopy.

MOTIVATION: Quantitative studies of cellular morphodynamics rely on extracting leading-edge velocity time series based on accurate cell segmentation from live cell imaging. However, live cell imaging has numerous challenging issues regarding accurate edge localization. Fluorescence live cell imaging produces noisy and low-contrast images due to phototoxicity and photobleaching. While phase contrast microscopy is gentle to live cells, it suffers from the halo and shade-off artifacts that cannot be handled by conventional segmentation algorithms. Here, we present a deep learning-based pipeline, termed MARS-Net (Multiple-microscopy-type-based Accurate and Robust Segmentation Network), that utilizes transfer learning and data from multiple types of microscopy to localize cell edges with high accuracy, allowing quantitative profiling of cellular morphodynamics. SUMMARY: To accurately segment cell edges and quantify cellular morphodynamics from live-cell imaging data, we developed a deep learning-based pipeline termed MARS-Net (multiple-microscopy-type-based accurate and robust segmentation network). MARS-Net utilizes transfer learning and data from multiple types of microscopy to localize cell edges with high accuracy. For effective training on distinct types of live-cell microscopy, MARS-Net comprises a pretrained VGG19 encoder with U-Net decoder and dropout layers. We trained MARS-Net on movies from phase-contrast, spinning-disk confocal, and total internal reflection fluorescence microscopes. MARS-Net produced more accurate edge localization than the neural network models trained with single-microscopy-type datasets. We expect that MARS-Net can accelerate the studies of cellular morphodynamics by providing accurate pixel-level segmentation of complex live-cell datasets.

Bcl3-dependent stabilization of CtBP1 is crucial for the inhibition of apoptosis and tumor progression in breast cancer.

B-cell lymphoma 3 (Bcl3) is a proto-oncogene upregulated in a wide range of cancers, including breast cancer. Although Bcl3 is known to promote cell proliferation and inhibit apoptosis, the molecular mechanisms underlying the proto-oncogenic function of Bcl3 have not been completely elucidated. To gain insight into the oncogenic role of Bcl3, we applied a proteomic approach, which led to the identification of C-terminal binding protein 1 (CtBP1) as a binding partner of Bcl3. A PXDLS/R motif embedded in Bcl3 was found to mediate the interaction between Bcl3 and CtBP1, which caused the stabilization of CtBP1 by blocking proteasome-dependent degradation. Apoptotic stimuli-induced degradation of CtBP1 was significantly abolished by the upregulation of Bcl3, leading to the sustained repression of pro-apoptotic gene expression and subsequent inhibition of apoptosis. Intriguingly, a strong positive correlation between the protein levels of Bcl3 and CtBP1 was detected in breast cancer patient samples. Our study reveals a novel combinatorial role for Bcl3 and CtBP1, providing an explanation for the acquisition of resistance to apoptosis in cancer cells, which is a major requirement for cancer development.

SUMOylation of pontin chromatin-remodeling complex reveals a signal integration code in prostate cancer cells.

Posttranslational modification by small ubiquitin-like modifier (SUMO) controls diverse cellular functions of transcription factors and coregulators and participates in various cellular processes including signal transduction and transcriptional regulation. Here, we report that pontin, a component of chromatin-remodeling complexes, is SUMO-modified, and that SUMOylation of pontin is an active control mechanism for the transcriptional regulation of pontin on androgen-receptor target genes in prostate cancer cells. Biochemical purification of pontin-containing complexes revealed the presence of the Ubc9 SUMO-conjugating enzyme that underlies its function as an activator. Intriguingly, 5alpha-dihydroxytestosterone treatments significantly increased the SUMOylation of pontin, and SUMOylated pontin showed further activation of a subset of nuclear receptor-dependent transcription and led to an increase in proliferation and growth of prostate cancer cells. These data clearly define a functional model and provide a link between SUMO modification and prostate cancer progression.

Co-option of neurotransmitter signaling for inter-organismal communication in C. elegans.

Biogenic amine neurotransmitters play a central role in metazoan biology, and both their chemical structures and cognate receptors are evolutionarily conserved. Their primary roles are in cell-to-cell signaling, as biogenic amines are not normally recruited for communication between separate individuals. Here, we show that in the nematode C. elegans, a neurotransmitter-sensing G protein-coupled receptor, TYRA-2, is required for avoidance responses to osas#9, an ascaroside pheromone that incorporates the neurotransmitter, octopamine. Neuronal ablation, cell-specific genetic rescue, and calcium imaging show that tyra-2 expression in the nociceptive neuron, ASH, is necessary and sufficient to induce osas#9 avoidance. Ectopic expression in the AWA neuron, which is generally associated with attractive responses, reverses the response to osas#9, resulting in attraction instead of avoidance behavior, confirming that TYRA-2 partakes in the sensing of osas#9. The TYRA-2/osas#9 signaling system represents an inter-organismal communication channel that evolved via co-option of a neurotransmitter and its cognate receptor.

Deconvolution of subcellular protrusion heterogeneity and the underlying actin regulator dynamics from live cell imaging.

Cell protrusion is morphodynamically heterogeneous at the subcellular level. However, the mechanism of cell protrusion has been understood based on the ensemble average of actin regulator dynamics. Here, we establish a computational framework called HACKS (deconvolution of heterogeneous activity in coordination of cytoskeleton at the subcellular level) to deconvolve the subcellular heterogeneity of lamellipodial protrusion from live cell imaging. HACKS identifies distinct subcellular protrusion phenotypes based on machine-learning algorithms and reveals their underlying actin regulator dynamics at the leading edge. Using our method, we discover "accelerating protrusion", which is driven by the temporally ordered coordination of Arp2/3 and VASP activities. We validate our finding by pharmacological perturbations and further identify the fine regulation of Arp2/3 and VASP recruitment associated with accelerating protrusion. Our study suggests HACKS can identify specific subcellular protrusion phenotypes susceptible to pharmacological perturbation and reveal how actin regulator dynamics are changed by the perturbation.

Deep transfer learning-based hologram classification for molecular diagnostics.

Lens-free digital in-line holography (LDIH) is a promising microscopic tool that overcomes several drawbacks (e.g., limited field of view) of traditional lens-based microcopy. However, extensive computation is required to reconstruct object images from the complex diffraction patterns produced by LDIH. This limits LDIH utility for point-of-care applications, particularly in resource limited settings. We describe a deep transfer learning (DTL) based approach to process LDIH images in the context of cellular analyses. Specifically, we captured holograms of cells labeled with molecular-specific microbeads and trained neural networks to classify these holograms without reconstruction. Using raw holograms as input, the trained networks were able to classify individual cells according to the number of cell-bound microbeads. The DTL-based approach including a VGG19 pretrained network showed robust performance with experimental data. Combined with the developed DTL approach, LDIH could be realized as a low-cost, portable tool for point-of-care diagnostics.