Automatic Dendritic Length Quantification for High Throughput Screening of Mature Neurons

High-content image-based analysis and proteomic profiling identifies Tau phosphorylation inhibitors in a human iPSC-derived glutamatergic neuronal model of tauopathy

Mutations in MAPT (microtubule-associated protein tau) cause frontotemporal dementia (FTD). MAPT mutations are associated with abnormal tau phosphorylation levels and accumulation of misfolded tau protein that can propagate between neurons ultimately leading to cell death (tauopathy). Recently, a p.A152T tau variant was identified as a risk factor for FTD, Alzheimer's disease, and synucleinopathies. Here we used induced pluripotent stem cells (iPSC) from a patient carrying this p.A152T variant to create a robust, functional cellular assay system for probing pathophysiological tau accumulation and phosphorylation. Using stably transduced iPSC-derived neural progenitor cells engineered to enable inducible expression of the pro-neural transcription factor Neurogenin 2 (Ngn2), we generated disease-relevant, cortical-like glutamatergic neurons in a scalable, high-throughput screening compatible format. Utilizing automated confocal microscopy, and an advanced image-processing pipeline optimized for analysis of morphologically complex human neuronal cultures, we report quantitative, subcellular localization-specific effects of multiple kinase inhibitors on tau, including ones under clinical investigation not previously reported to affect tau phosphorylation. These results demonstrate the potential for using patient iPSC-derived ex vivo models of tauopathy as genetically accurate, disease-relevant systems to probe tau biochemistry and support the discovery of novel therapeutics for tauopathies.

Comparing Automated Morphology Quantification Software on Dendrites of Uninjured and Injured Drosophila Neurons

Dendrites shape inputs and integration of depolarization that controls neuronal activity in the nervous system. Neuron pathologies can damage dendrite architecture and cause abnormalities in morphologies after injury. Dendrite regeneration can be quantified by various parameters, including total dendrite length and number of dendrite branches using manual or automated image analysis approaches. However, manual quantification is tedious and time consuming and automated approaches are often trained using wildtype neurons, making them poorly suited for analysis of genetically manipulated or injured dendrite arbors. In this study, we tested how well automated image analysis software performed on class IV Drosophila neurons, which have several hundred individual dendrite branches. We applied each software to automatically quantify features of uninjured neurons and neurons that regenerated new dendrites after injury. Regenerated arbors exhibit defects across multiple features of dendrite morphology, which makes them challenging for automated pipelines to analyze. We compared the performances of three automated pipelines against manual quantification using Simple Neurite Tracer in ImageJ: one that is commercially available (Imaris) and two developed by independent research groups (DeTerm and Tireless Tracing Genie). Out of the three software tested, we determined that Imaris is the most efficient at reconstructing dendrite architecture, but does not accurately measure total dendrite length even after intensive manual editing. Imaris outperforms both DeTerm and Tireless Tracing Genie for counting dendrite branches, and is better able to recreate previous conclusions from this same dataset. This thorough comparison of strengths and weaknesses of each software demonstrates their utility for analyzing regenerated neuron phenotypes in future studies.

TDP-43 Regulates Coupled Dendritic mRNA Transport-Translation Processes in Co-operation with FMRP and Staufen1

Tightly regulated transport of messenger ribonucleoprotein (mRNP) granules to diverse locations of dendrites and axons is essential for appropriately timed protein synthesis within distinct sub-neuronal compartments. Perturbations of this regulation lead to various neurological disorders. Using imaging and molecular approaches, we demonstrate how TDP-43 co-operates with two other RNA-binding proteins, FMRP and Staufen1, to regulate the anterograde and retrograde transport, respectively, of Rac1 mRNPs in mouse neuronal dendrites. We also analyze the mechanisms by which TDP-43 mediates coupled mRNA transport-translation processes in dendritic sub-compartments by following in real-time the co-movement of RNA and endogenous fluorescence-tagged protein in neurons and by simultaneous examination of transport/translation dynamics by using an RNA biosensor. This study establishes the pivotal roles of TDP-43 in transporting mRNP granules in dendrites, inhibiting translation inside those granules, and reactivating it once the granules reach the dendritic spines.

Automated Live-Cell Imaging of Synapses in Rat and Human Neuronal Cultures

Synapse loss and dendritic damage correlate with cognitive decline in many neurodegenerative diseases, underlie neurodevelopmental disorders, and are associated with environmental and drug-induced CNS toxicities. However, screening assays designed to measure loss of synaptic connections between live cells are lacking. Here, we describe the design and validation of automated synaptic imaging assay (ASIA), an efficient approach to label, image, and analyze synapses between live neurons. Using viral transduction to express fluorescent proteins that label synapses and an automated computer-controlled microscope, we developed a method to identify agents that regulate synapse number. ASIA is compatible with both confocal and wide-field microscopy; wide-field image acquisition is faster but requires a deconvolution step in the analysis. Both types of images feed into batch processing analysis software that can be run on ImageJ, CellProfiler, and MetaMorph platforms. Primary analysis endpoints are the number of structural synapses and cell viability. Thus, overt cell death is differentiated from subtle changes in synapse density, an important distinction when studying neurodegenerative processes. In rat hippocampal cultures treated for 24 h with 100 μM 2-bromopalmitic acid (2-BP), a compound that prevents clustering of postsynaptic density 95 (PSD95), ASIA reliably detected loss of postsynaptic density 95-enhanced green fluorescent protein (PSD95-eGFP)-labeled synapses in the absence of cell death. In contrast, treatment with 100 μM glutamate produced synapse loss and significant cell death, determined from morphological changes in a binary image created from co-expressed mCherry. Treatment with 3 mM lithium for 24 h significantly increased the number of fluorescent puncta, showing that ASIA also detects synaptogenesis. Proof of concept studies show that cell-specific promoters enable the selective study of inhibitory or principal neurons and that alternative reporter constructs enable quantification of GABAergic or glutamatergic synapses. ASIA can also be used to study synapse loss between human induced pluripotent stem cell (iPSC)-derived cortical neurons. Significant synapse loss in the absence of cell death was detected in the iPSC-derived neuronal cultures treated with either 100 μM 2-BP or 100 μM glutamate for 24 h, while 300 μM glutamate produced synapse loss and cell death. ASIA shows promise for identifying agents that evoke synaptic toxicities and screening for compounds that prevent or reverse synapse loss.

Automated Neuron Detection in High-Content Fluorescence Microscopy Images Using Machine Learning

The study of neuronal morphology in relation to function, and the development of effective medicines to positively impact this relationship in patients suffering from neurodegenerative diseases, increasingly involves image-based high-content screening and analysis. The first critical step toward fully automated high-content image analyses in such studies is to detect all neuronal cells and distinguish them from possible non-neuronal cells or artifacts in the images. Here we investigate the performance of well-established machine learning techniques for this purpose. These include support vector machines, random forests, k-nearest neighbors, and generalized linear model classifiers, operating on an extensive set of image features extracted using the compound hierarchy of algorithms representing morphology, and the scale-invariant feature transform. We present experiments on a dataset of rat hippocampal neurons from our own studies to find the most suitable classifier(s) and subset(s) of features in the common practical setting where there is very limited annotated data for training. The results indicate that a random forests classifier using the right feature subset ranks best for the considered task, although its performance is not statistically significantly better than some support vector machine based classification models.

HyphaTracker: An ImageJ toolbox for time-resolved analysis of spore germination in filamentous fungi

The dynamics of early fungal development and its interference with physiological signals and environmental factors is yet poorly understood. Especially computational analysis tools for the evaluation of the process of early spore germination and germ tube formation are still lacking. For the time-resolved analysis of conidia germination of the filamentous ascomycete Fusarium fujikuroi we developed a straightforward toolbox implemented in ImageJ. It allows for processing of microscopic acquisitions (movies) of conidial germination starting with drift correction and data reduction prior to germling analysis. From the image time series germling related region of interests (ROIs) are extracted, which are analysed for their area, circularity, and timing. ROIs originating from germlings crossing other hyphae or the image boundaries are omitted during analysis. Each conidium/hypha is identified and related to its origin, thus allowing subsequent categorization. The efficiency of HyphaTracker was proofed and the accuracy was tested on simulated germlings at different signal-to-noise ratios. Bright-field microscopic images of conidial germination of rhodopsin-deficient F. fujikuroi mutants and their respective control strains were analysed with HyphaTracker. Consistent with our observation in earlier studies the CarO deficient mutant germinated earlier and grew faster than other, CarO expressing strains.

Deep Learning Segmentation of Optical Microscopy Images Improves 3-D Neuron Reconstruction

Digital reconstruction, or tracing, of 3-D neuron structure from microscopy images is a critical step toward reversing engineering the wiring and anatomy of a brain. Despite a number of prior attempts, this task remains very challenging, especially when images are contaminated by noises or have discontinued segments of neurite patterns. An approach for addressing such problems is to identify the locations of neuronal voxels using image segmentation methods, prior to applying tracing or reconstruction techniques. This preprocessing step is expected to remove noises in the data, thereby leading to improved reconstruction results. In this paper, we proposed to use 3-D convolutional neural networks (CNNs) for segmenting the neuronal microscopy images. Specifically, we designed a novel CNN architecture, that takes volumetric images as the inputs and their voxel-wise segmentation maps as the outputs. The developed architecture allows us to train and predict using large microscopy images in an end-to-end manner. We evaluated the performance of our model on a variety of challenging 3-D microscopy images from different organisms. Results showed that the proposed methods improved the tracing performance significantly when combined with different reconstruction algorithms.