3-Dimensional Immunostaining and Automated Deep-Learning Based Analysis of Nerve Degeneration.

Multiple sclerosis (MS) is an autoimmune and neurodegenerative disease driven by inflammation and demyelination in the brain, spinal cord, and optic nerve. Optic neuritis, characterized by inflammation and demyelination of the optic nerve, is a symptom in many patients with MS. The optic nerve is the highway for visual information transmitted from the retina to the brain. It contains axons from the retinal ganglion cells (RGCs) that reside in the retina, myelin forming oligodendrocytes and resident microglia and astrocytes. Inflammation, demyelination, and axonal degeneration are also present in the optic nerve of mice subjected to experimental autoimmune encephalomyelitis (EAE), a preclinical mouse model of MS. Monitoring the optic nerve in EAE is a useful strategy to study the presentation and progression of pathology in the visual system; however, current approaches have relied on sectioning, staining and manual quantification. Further, information regarding the spatial load of lesions and inflammation is dependent on the area of sectioning. To better characterize cellular pathology in the EAE model, we employed a tissue clearing and 3D immunolabelling and imaging protocol to observe patterns of immune cell infiltration and activation throughout the optic nerve. Increased density of TOPRO staining for nuclei captured immune cell infiltration and Iba1 immunostaining was employed to monitor microglia and macrophages. Axonal degeneration was monitored by neurofilament immunolabelling to reveal axonal swellings throughout the optic nerve. In parallel, we developed a convolutional neural network with a UNet architecture (CNN-UNet) called BlebNet for automated identification and quantification of axonal swellings in whole mount optic nerves. Together this constitutes a toolkit for 3-dimensional immunostaining to monitor general optic nerve pathology and fast automated quantification of axonal defects that could also be adapted to monitor axonal degeneration and inflammation in other neurodegenerative disease models.

Understanding Associative Learning Through Higher-Order Conditioning.

Associative learning is often considered to require the physical presence of stimuli in the environment in order for them to be linked. This, however, is not a necessary condition for learning. Indeed, associative relationships can form between events that are never directly paired. That is, associative learning can occur by integrating information across different phases of training. Higher-order conditioning provides evidence for such learning through two deceptively similar designs - sensory preconditioning and second-order conditioning. In this review, we detail the procedures and factors that influence learning in these designs, describe the associative relationships that can be acquired, and argue for the importance of this knowledge in studying brain function.

Dopamine transients do not act as model-free prediction errors during associative learning.

Dopamine neurons are proposed to signal the reward prediction error in model-free reinforcement learning algorithms. This term represents the unpredicted or 'excess' value of the rewarding event, value that is then added to the intrinsic value of any antecedent cues, contexts or events. To support this proposal, proponents cite evidence that artificially-induced dopamine transients cause lasting changes in behavior. Yet these studies do not generally assess learning under conditions where an endogenous prediction error would occur. Here, to address this, we conducted three experiments where we optogenetically activated dopamine neurons while rats were learning associative relationships, both with and without reward. In each experiment, the antecedent cues failed to acquire value and instead entered into associations with the later events, whether valueless cues or valued rewards. These results show that in learning situations appropriate for the appearance of a prediction error, dopamine transients support associative, rather than model-free, learning.

Causal evidence supporting the proposal that dopamine transients function as temporal difference prediction errors.

Reward-evoked dopamine transients are well established as prediction errors. However, the central tenet of temporal difference accounts-that similar transients evoked by reward-predictive cues also function as errors-remains untested. In the present communication we addressed this by showing that optogenetically shunting dopamine activity at the start of a reward-predicting cue prevents second-order conditioning without affecting blocking. These results indicate that cue-evoked transients function as temporal-difference prediction errors rather than reward predictions.