Classifying Goliath Grouper (Epinephelus itajara) Behaviors from a Novel, Multi-Sensor Tag.

Inertial measurement unit sensors (IMU; i.e., accelerometer, gyroscope and magnetometer combinations) are frequently fitted to animals to better understand their activity patterns and energy expenditure. Capable of recording hundreds of data points a second, these sensors can quickly produce large datasets that require methods to automate behavioral classification. Here, we describe behaviors derived from a custom-built multi-sensor bio-logging tag attached to Atlantic Goliath grouper (Epinephelus itajara) within a simulated ecosystem. We then compared the performance of two commonly applied machine learning approaches (random forest and support vector machine) to a deep learning approach (convolutional neural network, or CNN) for classifying IMU data from this tag. CNNs are frequently used to recognize activities from IMU data obtained from humans but are less commonly considered for other animals. Thirteen behavioral classes were identified during ethogram development, nine of which were classified. For the conventional machine learning approaches, 187 summary statistics were extracted from the data, including time and frequency domain features. The CNN was fed absolute values obtained from fast Fourier transformations of the raw tri-axial accelerometer, gyroscope and magnetometer channels, with a frequency resolution of 512 data points. Five metrics were used to assess classifier performance; the deep learning approach performed better across all metrics (Sensitivity = 0.962; Specificity = 0.996; F1-score = 0.962; Matthew's Correlation Coefficient = 0.959; Cohen's Kappa = 0.833) than both conventional machine learning approaches. Generally, the random forest performed better than the support vector machine. In some instances, a conventional learning approach yielded a higher performance metric for particular classes (e.g., the random forest had a F1-score of 0.971 for backward swimming compared to 0.955 for the CNN). Deep learning approaches could potentially improve behavioral classification from IMU data, beyond that obtained from conventional machine learning methods.

Immune Sensing of Synthetic, Bacterial, and Protozoan RNA by Toll-like Receptor 8 Requires Coordinated Processing by RNase T2 and RNase 2.

Human toll-like receptor 8 (TLR8) activation induces a potent T helper-1 (Th1) cell response critical for defense against intracellular pathogens, including protozoa. The receptor harbors two distinct binding sites, uridine and di- and/or trinucleotides, but the RNases upstream of TLR8 remain poorly characterized. We identified two endolysosomal endoribonucleases, RNase T2 and RNase 2, that act synergistically to release uridine from oligoribonucleotides. RNase T2 cleaves preferentially before, and RNase 2 after, uridines. Live bacteria, P. falciparum-infected red blood cells, purified pathogen RNA, and synthetic oligoribonucleotides all required RNase 2 and T2 processing to activate TLR8. Uridine supplementation restored RNA recognition in RNASE2-/- or RNASET2-/- but not RNASE2-/-RNASET2-/- cells. Primary immune cells from RNase T2-hypomorphic patients lacked a response to bacterial RNA but responded robustly to small-molecule TLR8 ligands. Our data identify an essential function of RNase T2 and RNase 2 upstream of TLR8 and provide insight into TLR8 activation.

AAV-mediated expression of BAG1 and ROCK2-shRNA promote neuronal survival and axonal sprouting in a rat model of rubrospinal tract injury.

A lesion to the rat rubrospinal tract is a model for traumatic spinal cord lesions and results in atrophy of the red nucleus neurons, axonal dieback, and locomotor deficits. In this study, we used adeno-associated virus (AAV)-mediated over-expression of BAG1 and ROCK2-shRNA in the red nucleus to trace [by co-expression of enhanced green fluorescent protein (EGFP)] and treat the rubrospinal tract after unilateral dorsal hemisection. We investigated the effects of targeted gene therapy on neuronal survival, axonal sprouting of the rubrospinal tract, and motor recovery 12 weeks after unilateral dorsal hemisection at Th8 in rats. In addition to the evaluation of BAG1 and ROCK2 as therapeutic targets in spinal cord injury, we aimed to demonstrate the feasibility and the limits of an AAV-mediated protein over-expression versus AAV.shRNA-mediated down-regulation in this traumatic CNS lesion model. Our results demonstrate that BAG1 and ROCK2-shRNA both promote neuronal survival of red nucleus neurons and enhance axonal sprouting proximal to the lesion.

Hepatocyte growth factor protects retinal ganglion cells by increasing neuronal survival and axonal regeneration in vitro and in vivo.

Hepatocyte growth factor (HGF) is known to promote the survival and foster neuritic outgrowth of different subpopulations of CNS neurons during development. Together with its corresponding receptor c-mesenchymal-epithelial transition factor (Met), it is expressed in the developing and the adult murine, rat and human CNS. We have studied the role of HGF in paradigms of retinal ganglion cell (RGC) regeneration and cell death in vitro and in vivo. After application of recombinant HGF in vitro, survival of serum-deprived RGC-5 cells and of growth factor-deprived primary RGC was significantly increased. This was shown to be correlated to the phosphorylation of c-Met and subsequent activation of serine/threonine protein kinase Akt and MAPK downstream signalling pathways involved in neuronal survival. Furthermore, neurite outgrowth of primary RGC was stimulated by HGF. In vivo, c-Met expression in RGC was up-regulated after optic nerve axotomy lesion. Here, treatment with HGF significantly improved survival of axotomized RGC and enhanced axonal regeneration after optic nerve crush. Our data demonstrates that exogenously applied HGF has a neuroprotective and regeneration-promoting function for lesioned CNS neurons. We provide strong evidence that HGF may represent a trophic factor for adult CNS neurons, which may play a role as therapeutic target in the treatment of neurotraumatic and neurodegenerative CNS disorders.

Acute axonal degeneration in vivo is attenuated by inhibition of autophagy in a calcium-dependent manner.

Axonal degeneration is a pathological hallmark of many traumatic and neurodegenerative neurological disorders. Although the underlying mechanisms remain largely unclear, increased autophagy and the influx of extracellular calcium have been implicated in the pathogenesis of axonal degeneration based on in vitro data. Using in vivo imaging of the rat optic nerve after crush lesion we could now show that both mechanisms are linked and play an important role in acute axonal degeneration in vivo. Our data suggest that crush lesion of the optic nerve induces a rapid calcium influx through calcium channels, which results in a secondary induction of autophagy that participates actively in axonal degradation. Therapeutic manipulation of both events could significantly alter the time course of acute axonal degeneration in vivo and may thus represent promising therapeutic targets for the future.

Mechanisms of acute axonal degeneration in the optic nerve in vivo.

Axonal degeneration is an initial key step in traumatic and neurodegenerative CNS disorders. We established a unique in vivo epifluorescence imaging paradigm to characterize very early events in axonal degeneration in the rat optic nerve. Single retinal ganglion cell axons were visualized by AAV-mediated expression of dsRed and this allowed the quantification of postlesional acute axonal degeneration (AAD). EM analysis revealed severe structural alterations of the cytoskeleton, cytoplasmatic vacuolization, and the appearance of autophagosomes within the first hours after lesion. Inhibition of autophagy resulted in an attenuation of acute axonal degeneration. Furthermore, a rapid increase of intraaxonal calcium levels following crush lesion could be visualized using a calcium-sensitive dye. Application of calcium channel inhibitors prevented crush-induced calcium increase and markedly attenuated axonal degeneration, whereas application of a calcium ionophore aggravated the degenerative phenotype. We finally demonstrate that increased postlesional autophagy is calcium dependent and thus mechanistically link autophagy and intraaxonal calcium levels. Both processes are proposed to be major targets for the manipulation of axonal degeneration in future therapeutic settings.

BAG1 promotes axonal outgrowth and regeneration in vivo via Raf-1 and reduction of ROCK activity.

Improved survival of injured neurons and the inhibition of repulsive environmental signalling are prerequisites for functional regeneration. BAG1 (Bcl-2-associated athanogene-1) is an Hsp70/Hsc70-binding protein, which has been shown to suppress apoptosis and enhance neuronal differentiation. We investigated BAG1 as a therapeutic molecule in the lesioned visual system in vivo. Using an adeno-associated viral vector, BAG1 (AAV.BAG1) was expressed in retinal ganglion cells (RGC) and then tested in models of optic nerve axotomy and optic nerve crush. BAG1 significantly increased RGC survival as compared to adeno-associated viral vector enhanced green fluorescent protein (AAV.EGFP) treated controls and this was independently confirmed in transgenic mice over-expressing BAG1 in neurons. The numbers and lengths of regenerating axons after optic nerve crush were also significantly increased in the AAV.BAG1 group. In pRGC cultures, BAG1-over-expression resulted in a approximately 3-fold increase in neurite length and growth cone surface. Interestingly, BAG1 induced an intracellular translocation of Raf-1 and ROCK2 and ROCK activity was decreased in a Raf-1-dependent manner by BAG1-over-expression. In summary, we show that BAG1 acts in a dual role by inhibition of lesion-induced apoptosis and interaction with the inhibitory ROCK signalling cascade. BAG1 is therefore a promising molecule to be further examined as a putative therapeutic tool in neurorestorative strategies.