skip to main content


Title: Deep Analysis of Mitochondria and Cell Health Using Machine Learning
Abstract

There is a critical need for better analytical methods to study mitochondria in normal and diseased states. Mitochondrial image analysis is typically done on still images using slow manual methods or automated methods of limited types of features. MitoMo integrated software overcomes these bottlenecks by automating rapid unbiased quantitative analysis of mitochondrial morphology, texture, motion, and morphogenesis and advances machine-learning classification to predict cell health by combining features. Our pixel-based approach for motion analysis evaluates the magnitude and direction of motion of: (1) molecules within mitochondria, (2) individual mitochondria, and (3) distinct morphological classes of mitochondria. MitoMo allows analysis of mitochondrial morphogenesis in time-lapse videos to study early progression of cellular stress. Biological applications are presented including: (1) establishing normal phenotypes of mitochondria in different cell types; (2) quantifying stress-induced mitochondrial hyperfusion in cells treated with an environmental toxicant, (3) tracking morphogenesis in mitochondria undergoing swelling, and (4) evaluating early changes in cell health when morphological abnormalities are not apparent. MitoMo unlocks new information on mitochondrial phenotypes and dynamics by enabling deep analysis of mitochondrial features in any cell type and can be applied to a broad spectrum of research problems in cell biology, drug testing, toxicology, and medicine.

 
more » « less
NSF-PAR ID:
10153338
Author(s) / Creator(s):
; ; ; ; ; ;
Publisher / Repository:
Nature Publishing Group
Date Published:
Journal Name:
Scientific Reports
Volume:
8
Issue:
1
ISSN:
2045-2322
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Plant development requires communication on many levels, including between cells and between organelles within a cell. For example, mitochondria and plastids have been proposed to be sensors of environmental stress and to coordinate their responses. Here we present evidence for communication between mitochondria and chloroplasts during leaf and root development, based on genetic and physical interactions between threeMechanosensitive channel ofSmall conductance‐Like (MSL) proteins fromArabidopsis thaliana.MSLproteins areArabidopsishomologs of the bacterialMechanosensitivechannel ofSmall conductance (MscS), which relieves cellular osmotic pressure to protect against lysis during hypoosmotic shock.MSL1 localizes to the inner mitochondrial membrane, whileMSL2 andMSL3 localize to the inner plastid membrane and are required to maintain plastid osmotic homeostasis during normal growth and development. In this study, we characterized the phenotypic effect of a genetic lesion inMSL1, both in wild type and inmsl2 msl3mutant backgrounds.msl1single mutants appear wild type for all phenotypes examined. The characteristic leaf rumpling inmsl2 msl3double mutants was exacerbated in themsl1 msl2 msl3triple mutant. However, the introduction of themsl1lesion into themsl2 msl3mutant background suppressed othermsl2 msl3mutant phenotypes, including ectopic callus formation, accumulation of superoxide and hydrogen peroxide in the shoot apical meristem, decreased root length, and reduced number of lateral roots. All these phenotypes could be recovered by molecular complementation with a transgene containing a wild type version ofMSL1. In yeast‐based interaction studies,MSL1 interacted with itself, but not withMSL2 orMSL3. These results establish that the abnormalities observed inmsl2 msl3double mutants is partially dependent on the presence of functionalMSL1 and suggest a possible role for communication between plastid and mitochondria in seedling development.

     
    more » « less
  2. null (Ed.)
    Introduction: Vaso-occlusive crises (VOCs) are a leading cause of morbidity and early mortality in individuals with sickle cell disease (SCD). These crises are triggered by sickle red blood cell (sRBC) aggregation in blood vessels and are influenced by factors such as enhanced sRBC and white blood cell (WBC) adhesion to inflamed endothelium. Advances in microfluidic biomarker assays (i.e., SCD Biochip systems) have led to clinical studies of blood cell adhesion onto endothelial proteins, including, fibronectin, laminin, P-selectin, ICAM-1, functionalized in microchannels. These microfluidic assays allow mimicking the physiological aspects of human microvasculature and help characterize biomechanical properties of adhered sRBCs under flow. However, analysis of the microfluidic biomarker assay data has so far relied on manual cell counting and exhaustive visual morphological characterization of cells by trained personnel. Integrating deep learning algorithms with microscopic imaging of adhesion protein functionalized microfluidic channels can accelerate and standardize accurate classification of blood cells in microfluidic biomarker assays. Here we present a deep learning approach into a general-purpose analytical tool covering a wide range of conditions: channels functionalized with different proteins (laminin or P-selectin), with varying degrees of adhesion by both sRBCs and WBCs, and in both normoxic and hypoxic environments. Methods: Our neural networks were trained on a repository of manually labeled SCD Biochip microfluidic biomarker assay whole channel images. Each channel contained adhered cells pertaining to clinical whole blood under constant shear stress of 0.1 Pa, mimicking physiological levels in post-capillary venules. The machine learning (ML) framework consists of two phases: Phase I segments pixels belonging to blood cells adhered to the microfluidic channel surface, while Phase II associates pixel clusters with specific cell types (sRBCs or WBCs). Phase I is implemented through an ensemble of seven generative fully convolutional neural networks, and Phase II is an ensemble of five neural networks based on a Resnet50 backbone. Each pixel cluster is given a probability of belonging to one of three classes: adhered sRBC, adhered WBC, or non-adhered / other. Results and Discussion: We applied our trained ML framework to 107 novel whole channel images not used during training and compared the results against counts from human experts. As seen in Fig. 1A, there was excellent agreement in counts across all protein and cell types investigated: sRBCs adhered to laminin, sRBCs adhered to P-selectin, and WBCs adhered to P-selectin. Not only was the approach able to handle surfaces functionalized with different proteins, but it also performed well for high cell density images (up to 5000 cells per image) in both normoxic and hypoxic conditions (Fig. 1B). The average uncertainty for the ML counts, obtained from accuracy metrics on the test dataset, was 3%. This uncertainty is a significant improvement on the 20% average uncertainty of the human counts, estimated from the variance in repeated manual analyses of the images. Moreover, manual classification of each image may take up to 2 hours, versus about 6 minutes per image for the ML analysis. Thus, ML provides greater consistency in the classification at a fraction of the processing time. To assess which features the network used to distinguish adhered cells, we generated class activation maps (Fig. 1C-E). These heat maps indicate the regions of focus for the algorithm in making each classification decision. Intriguingly, the highlighted features were similar to those used by human experts: the dimple in partially sickled RBCs, the sharp endpoints for highly sickled RBCs, and the uniform curvature of the WBCs. Overall the robust performance of the ML approach in our study sets the stage for generalizing it to other endothelial proteins and experimental conditions, a first step toward a universal microfluidic ML framework targeting blood disorders. Such a framework would not only be able to integrate advanced biophysical characterization into fast, point-of-care diagnostic devices, but also provide a standardized and reliable way of monitoring patients undergoing targeted therapies and curative interventions, including, stem cell and gene-based therapies for SCD. Disclosures Gurkan: Dx Now Inc.: Patents & Royalties; Xatek Inc.: Patents & Royalties; BioChip Labs: Patents & Royalties; Hemex Health, Inc.: Consultancy, Current Employment, Patents & Royalties, Research Funding. 
    more » « less
  3. INTRODUCTION Eukaryotes contain a highly conserved signaling pathway that becomes rapidly activated when adenosine triphosphate (ATP) levels decrease, as happens during conditions of nutrient shortage or mitochondrial dysfunction. The adenosine monophosphate (AMP)–activated protein kinase (AMPK) is activated within minutes of energetic stress and phosphorylates a limited number of substrates to biochemically rewire metabolism from an anabolic state to a catabolic state to restore metabolic homeostasis. AMPK also promotes prolonged metabolic adaptation through transcriptional changes, decreasing biosynthetic genes while increasing expression of genes promoting lysosomal and mitochondrial biogenesis. The transcription factor EB (TFEB) is a well-appreciated effector of AMPK-dependent signals, but many of the molecular details of how AMPK controls these processes remain unknown. RATIONALE The requirement of AMPK and its specific downstream targets that control aspects of the transcriptional adaptation of metabolism remain largely undefined. We performed time courses examining gene expression changes after various mitochondrial stresses in wild-type (WT) or AMPK knockout cells. We hypothesized that a previously described interacting protein of AMPK, folliculin-interacting protein 1 (FNIP1), may be involved in how AMPK promotes increases in gene expression after metabolic stress. FNIP1 forms a complex with the protein folliculin (FLCN), together acting as a guanosine triphosphate (GTP)–activating protein (GAP) for RagC. The FNIP1-FLCN complex has emerged as an amino acid sensor to the mechanistic target of rapamycin complex 1 (mTORC1), involved in how amino acids control TFEB activation. We therefore examined whether AMPK may regulate FNIP1 to dominantly control TFEB independently of amino acids. RESULTS AMPK was found to govern expression of a core set of genes after various mitochondrial stresses. Hallmark features of this response were activation of TFEB and increases in the transcription of genes specifying lysosomal and mitochondrial biogenesis. AMPK directly phosphorylated five conserved serine residues in FNIP1, suppressing the function of the FLCN-FNIP1 GAP complex, which resulted in dissociation of RagC and mTOR from the lysosome, promoting nuclear translocation of TFEB even in the presence of amino acids. FNIP1 phosphorylation was required for AMPK to activate TFEB and for subsequent increases in peroxisome proliferation–activated receptor gamma coactivator 1-alpha (PGC1α) and estrogen-related receptor alpha (ERRα) mRNAs. Cells in which the five serines in FNIP1 were mutated to alanine were unable to increase lysosomal and mitochondrial gene expression programs after treatment with mitochondrial poisons or AMPK activators despite the presence and normal regulation of all other substrates of AMPK. By contrast, neither AMPK nor its control of FNIP1 were needed for activation of TFEB after amino acid withdrawal, illustrating the specificity to energy-limited conditions. CONCLUSION Our data establish FNIP1 as the long-sought substrate of AMPK that controls TFEB translocation to the nucleus, defining AMPK phosphorylation of FNIP1 as a singular event required for increased lysosomal and mitochondrial gene expression programs after metabolic stresses. This study also illuminates the larger biological question of how mitochondrial damage triggers a temporal response of repair and replacement of damaged mitochondria: Within early hours, AMPK-FNIP1–activated TFEB induces a wave of lysosome and autophagy genes to promote degradation of damaged mitochondria, and a few hours later, TFEB–up-regulated PGC1⍺ and ERR⍺ promote expression of a second wave of genes specifying mitochondrial biogenesis. These insights open therapeutic avenues for several common diseases associated with mitochondrial dysfunction, ranging from neurodegeneration to type 2 diabetes to cancer. Mitochondrial damage activates AMPK to phosphorylate FNIP1, stimulating TFEB translocation to the nucleus and sequential waves of lysosomal and mitochondrial biogenesis. After mitochondrial damage, activated AMPK phosphorylates FNIP1 (1), causing inhibition of FLCN-FNIP1 GAP activity (2). This leads to accumulation of RagC in its GTP-bound form, causing dissociation of RagC, mTORC1, and TFEB from the lysosome (3). TFEB is therefore not phosphorylated and translocates to the nucleus, inducing transcription of lysosomal or autophagy genes, with parallel increases in NT-PGC1α mRNA (4), which, in concert with ERRα (5), subsequently induces mitochondrial biogenesis (6). CCCP, carbonyl cyanide m-chlorophenylhydrazone; CLEAR, coordinated lysosomal expression and regulation; GDP, guanosine diphosphate; P, phosphorylation. [Figure created using BioRender] 
    more » « less
  4. Abstract Background

    Cytoplasmic male sterility (CMS) is a maternally inherited failure to produce functional pollen that most commonly results from expression of novel, chimeric mitochondrial genes. InZea mays, cytoplasmic male sterility type S (CMS-S) is characterized by the collapse of immature, bi-cellular pollen. Molecular and cellular features of developing CMS-S and normal (N) cytoplasm pollen were compared to determine the role of mitochondria in these differing developmental fates.

    Results

    Terminal deoxynucleotidyl transferase dUTP nick end labeling revealed both chromatin and nuclear fragmentation in the collapsed CMS-S pollen, demonstrating a programmed cell death (PCD) event sharing morphological features with mitochondria-signaled apoptosis in animals. Maize plants expressing mitochondria-targeted green fluorescent protein (GFP) demonstrated dynamic changes in mitochondrial morphology and association with actin filaments through the course of N-cytoplasm pollen development, whereas mitochondrial targeting of GFP was lost and actin filaments were disorganized in developing CMS-S pollen. Immunoblotting revealed significant developmental regulation of mitochondrial biogenesis in both CMS-S and N mito-types. Nuclear and mitochondrial genome encoded components of the cytochrome respiratory pathway and ATP synthase were of low abundance at the microspore stage, but microspores accumulated abundant nuclear-encoded alternative oxidase (AOX). Cytochrome pathway and ATP synthase components accumulated whereas AOX levels declined during the maturation of N bi-cellular pollen. Increased abundance of cytochrome pathway components and declining AOX also characterized collapsed CMS-S pollen. The accumulation and robust RNA editing of mitochondrial transcripts implicated translational or post-translational control for the developmentally regulated accumulation of mitochondria-encoded proteins in both mito-types.

    Conclusions

    CMS-S pollen collapse is a PCD event coincident with developmentally programmed mitochondrial events including the accumulation of mitochondrial respiratory proteins and declining protection against mitochondrial generation of reactive oxygen species.

     
    more » « less
  5. Summary

    Malonyl‐CoA is a key intermediate in a number of metabolic processes associated with its role as a substrate in acylation and condensation reactions. These types of reactions occur in plastids, the cytosol and mitochondria, and although carboxylation of acetyl‐CoA is the known mechanism for generating the distinct plastidial and cytosolic pools, the metabolic origin of the mitochondrial malonyl‐CoA pool is still unclear. In this study we demonstrate that malonyl‐CoA synthetase encoded by the ArabidopsisAAE13(AT3G16170) gene is localized in both the cytosol and the mitochondria. These isoforms are translated from two types of transcripts, one that contains and one that does not contain a mitochondrial‐targeting pre‐sequence. Whereas the cytosolicAAE13 protein is not essential, due to the presence of a redundant malonyl‐CoA generating system provided by a cytosolic acetyl‐CoA carboxylase, the mitochondrialAAE13 protein is essential for plant growth. Phenotypes of theaae13‐1mutant are transgenically reversed only if the mitochondrial pre‐sequence is present in the ectopically expressedAAE13 proteins. Theaae13‐1mutant exhibits typical metabolic phenotypes associated with a deficiency in the mitochondrial fatty acid synthase system, namely depleted lipoylation of the H subunit of the photorespiratory enzyme glycine decarboxylase, increased accumulation of glycine and glycolate and reduced levels of sucrose. Most of these metabolic alterations, and associated morphological changes, are reversed when theaae13‐1mutant is grown in a non‐photorespiratory condition (i.e. a 1%CO2atmosphere), demonstrating that they are a consequence of the deficiency in photorespiration due to the inability to generate lipoic acid from mitochondrially synthesized fatty acids.

     
    more » « less