Semantic segmentation of HeLa cells: An objective comparison between one traditional algorithm and four deep-learning architectures.

The quantitative study of cell morphology is of great importance as the structure and condition of cells and their structures can be related to conditions of health or disease. The first step towards that, is the accurate segmentation of cell structures. In this work, we compare five approaches, one traditional and four deep-learning, for the semantic segmentation of the nuclear envelope of cervical cancer cells commonly known as HeLa cells. Images of a HeLa cancer cell were semantically segmented with one traditional image-processing algorithm and four three deep learning architectures: VGG16, ResNet18, Inception-ResNet-v2, and U-Net. Three hundred slices, each 2000 × 2000 pixels, of a HeLa Cell were acquired with Serial Block Face Scanning Electron Microscopy. The first three deep learning architectures were pre-trained with ImageNet and then fine-tuned with transfer learning. The U-Net architecture was trained from scratch with 36, 000 training images and labels of size 128 × 128. The image-processing algorithm followed a pipeline of several traditional steps like edge detection, dilation and morphological operators. The algorithms were compared by measuring pixel-based segmentation accuracy and Jaccard index against a labelled ground truth. The results indicated a superior performance of the traditional algorithm (Accuracy = 99%, Jaccard = 93%) over the deep learning architectures: VGG16 (93%, 90%), ResNet18 (94%, 88%), Inception-ResNet-v2 (94%, 89%), and U-Net (92%, 56%).

CAR-associated vesicular transport of an adenovirus in motor neuron axons.

Axonal transport is responsible for the movement of signals and cargo between nerve termini and cell bodies. Pathogens also exploit this pathway to enter and exit the central nervous system. In this study, we characterised the binding, endocytosis and axonal transport of an adenovirus (CAV-2) that preferentially infects neurons. Using biochemical, cell biology, genetic, ultrastructural and live-cell imaging approaches, we show that interaction with the neuronal membrane correlates with coxsackievirus and adenovirus receptor (CAR) surface expression, followed by endocytosis involving clathrin. In axons, long-range CAV-2 motility was bidirectional with a bias for retrograde transport in nonacidic Rab7-positive organelles. Unexpectedly, we found that CAR was associated with CAV-2 vesicles that also transported cargo as functionally distinct as tetanus toxin, neurotrophins, and their receptors. These results suggest that a single axonal transport carrier is capable of transporting functionally distinct cargoes that target different membrane compartments in the soma. We propose that CAV-2 transport is dictated by an innate trafficking of CAR, suggesting an unsuspected function for this adhesion protein during neuronal homeostasis.

Human spastin has multiple microtubule-related functions.

Hereditary spastic paraplegias (HSPs) are neurodegenerative diseases caused by mutations in more than 20 genes, which lead to progressive spasticity and weakness of the lower limbs. The most frequently mutated gene causing autosomal dominant HSP is SPG4, which encodes spastin, a protein that belongs to the family of ATPases associated with various cellular activities (AAAs). A number of studies have suggested that spastin regulates microtubule dynamics. We have studied the ATPase activity of recombinant human spastin and examined the effect of taxol-stabilized microtubules on this activity. We used spastin translated from the second ATG and provide evidence that this is the physiologically relevant form. We showed that microtubules enhance the ATPase activity of the protein, a property also described for katanin, an AAA of the same spastin subgroup. Furthermore, we demonstrated that human spastin has a microtubule-destabilizing activity and can bundle microtubules in vitro, providing new insights into the molecular pathogenesis of HSP.