Unsupervised learning on spontaneous retinal activity leads to efficient neural representation geometry.

Prior to the onset of vision, neurons in the developing mammalian retina spontaneously fire in correlated activity patterns known as retinal waves. Experimental evidence suggests that retinal waves strongly influence the emergence of sensory representations before visual experience. We aim to model this early stage of functional development by using movies of neurally active developing retinas as pre-training data for neural networks. Specifically, we pre-train a ResNet-18 with an unsupervised contrastive learning objective (SimCLR) on both simulated and experimentally-obtained movies of retinal waves, then evaluate its performance on image classification tasks. We find that pre-training on retinal waves significantly improves performance on tasks that test object invariance to spatial translation, while slightly improving performance on more complex tasks like image classification. Notably, these performance boosts are realized on held-out natural images even though the pre-training procedure does not include any natural image data. We then propose a geometrical explanation for the increase in network performance, namely that the spatiotemporal characteristics of retinal waves facilitate the formation of separable feature representations. In particular, we demonstrate that networks pre-trained on retinal waves are more effective at separating image manifolds than randomly initialized networks, especially for manifolds defined by sets of spatial translations. These findings indicate that the broad spatiotemporal properties of retinal waves prepare networks for higher order feature extraction.

Perineuronal Nets in the Dorsomedial Striatum Contribute to Behavioral Dysfunction in Mouse Models of Excessive Repetitive Behavior.

Background: Excessive repetitive behavior is a debilitating symptom of several neuropsychiatric disorders. Parvalbumin-positive inhibitory interneurons in the dorsal striatum have been linked to repetitive behavior, and a sizable portion of these cells are surrounded by perineuronal nets (PNNs), specialized extracellular matrix structures. Although PNNs have been associated with plasticity and neuropsychiatric disease, no previous studies have investigated their involvement in excessive repetitive behavior. Methods: We used histochemistry and confocal imaging to investigate PNNs surrounding parvalbumin-positive cells in the dorsal striatum of 4 mouse models of excessive repetitive behavior (BTBR, Cntnap2, Shank3, prenatal valproate treatment). We then investigated one of these models, the BTBR mouse, in detail, with DiI labeling, in vivo and in vitro recordings, and behavioral analyses. We next degraded PNNs in the dorsomedial striatum (DMS) using the enzyme chondroitinase ABC and assessed dendritic spine density, electrophysiology, and repetitive behavior. Results: We found a greater percentage of parvalbumin-positive interneurons with PNNs in the DMS of all 4 mouse models of excessive repetitive behavior compared with control mice. In BTBR mice, we found fewer dendritic spines on medium spiny neurons (targets of parvalbumin-positive interneurons) and differences in neuronal oscillations as well as inhibitory postsynaptic potentials compared with control mice. Reduction of DMS PNNs in BTBR mice altered dendritic spine density and inhibitory responses and normalized repetitive behavior. Conclusions: These findings suggest that cellular abnormalities in the DMS are associated with maladaptive repetitive behaviors and that manipulating PNNs can restore normal levels of repetitive behavior while altering DMS dendritic spines and inhibitory signaling.

The influence of spontaneous and visual activity on the development of direction selectivity maps in mouse retina.

In mice, retinal direction selectivity is organized in a map that aligns to the body and gravitational axes of optic flow, and little is known about how this map develops. We find direction selectivity maps are largely present at eye opening and develop normally in the absence of visual experience. Remarkably, in mice lacking the beta2 subunit of neuronal nicotinic acetylcholine receptors (ß2-nAChR-KO), which exhibit drastically reduced cholinergic retinal waves in the first postnatal week, selectivity to horizontal motion is absent while selectivity to vertical motion remains. We tested several possible mechanisms that could explain the loss of horizontal direction selectivity in ß2-nAChR-KO mice (wave propagation bias, FRMD7 expression, starburst amacrine cell morphology), but all were found to be intact when compared with WT mice. This work establishes a role for retinal waves in the development of asymmetric circuitry that mediates retinal direction selectivity via an unknown mechanism.

Adult-born granule cell mossy fibers preferentially target parvalbumin-positive interneurons surrounded by perineuronal nets.

Adult-born granule cells (abGCs) integrate into the hippocampus and form connections with dentate gyrus parvalbumin-positive (PV+) interneurons, a circuit important for modulating plasticity. Many of these interneurons are surrounded by perineuronal nets (PNNs), extracellular matrix structures known to participate in plasticity. We compared abGC projections to PV+ interneurons with negative-to-low intensity PNNs to those with high intensity PNNs using retroviral and 3R-Tau labeling in adult mice, and found that abGC mossy fibers and boutons are more frequently located near PV+ interneurons with high intensity PNNs. These results suggest that axons of new neurons preferentially stabilize near target cells with intense PNNs. Next, we asked whether the number of abGCs influences PNN formation around PV+ interneurons, and found that near complete ablation of abGCs produced a decrease in the intensity and number of PV+ neurons with PNNs, suggesting that new neuron innervation may enhance PNN formation. Experience-driven changes in adult neurogenesis did not produce consistent effects, perhaps due to widespread effects on plasticity. Our study identifies abGC projections to PV+ interneurons with PNNs, with more presumed abGC mossy fiber boutons found near the cell body of PV+ interneurons with strong PNNs.