The normal function of the retina is to convey information about natural visual images. It is this visual environment that has driven evolution, and that is clinically relevant. Yet nearly all of our understanding of the neural computations, biological function, and circuit mechanisms of the retina comes in the context of artificially structured stimuli such as flashing spots, moving bars and white noise. It is fundamentally unclear how these artificial stimuli are related to circuit processes engaged under natural stimuli. A key barrier is the lack of methods for analyzing retinal responses to natural images. We addressed both these issues by applying convolutional neural network models (CNNs) to capture retinal responses to natural scenes. We find that CNN models predict natural scene responses with high accuracy, achieving performance close to the fundamental limits of predictability set by intrinsic cellular variability. Furthermore, individual internal units of the model are highly correlated with actual retinal interneuron responses that were recorded separately and never presented to the model during training. Finally, we find that models fit only to natural scenes, but not white noise, reproduce a range of phenomena previously described using distinct artificial stimuli, including frequency doubling, latency encoding, motion anticipation, fast contrast adaptation, synchronized responses to motion reversal and object motion sensitivity. Further examination of the model revealed extremely rapid context dependence of retinal feature sensitivity under natural scenes using an analysis not feasible from direct examination of retinal responses. Overall, these results show that nonlinear retinal processes engaged by artificial stimuli are also engaged in and relevant to natural visual processing, and that CNN models form a powerful and unifying tool to study how sensory circuitry produces computations in a natural context.

A fundamental challenge at the interface of machine learning and neuroscience is to uncover computational principles that are shared between artificial and biological neural networks. In deep learning, normalization methods such as batch normalization, weight normalization, and their many variants help to stabilize hidden unit activity and accelerate network training, and these methods have been called one of the most important recent innovations for optimizing deep networks. In the brain, homeostatic plasticity represents a set of mechanisms that also stabilize and normalize network activity to lie within certain ranges, and these mechanisms are critical for maintaining normal brain function. In this article, we discuss parallels between artificial and biological normalization methods at four spatial scales: normalization of a single neuron's activity, normalization of synaptic weights of a neuron, normalization of a layer of neurons, and normalization of a network of neurons. We argue that both types of methods are functionally equivalent—that is, both push activation patterns of hidden units toward a homeostatic state, where all neurons are equally used—and we argue that such representations can improve coding capacity, discrimination, and regularization. As a proof of concept, we develop an algorithm, inspired by a neural normalization technique called synaptic scaling, and show that this algorithm performs competitively against existing normalization methods on several data sets. Overall, we hope this bidirectional connection will inspire neuroscientists and machine learners in three ways: to uncover new normalization algorithms based on established neurobiological principles; to help quantify the trade-offs of different homeostatic plasticity mechanisms used in the brain; and to offer insights about how stability may not hinder, but may actually promote, plasticity.

The regulation of fundamental aspects of neurobiological function has been linked to the ubiquitin signaling system (USS), which regulates the degradation and activity of proteins and is catalyzed by E1, E2, and E3 enzymes. The Anaphase-Promoting Complex (APC) is a multi-subunit E3 ubiquitin ligase that controls diverse developmental and signaling processes in post-mitotic neurons; however, potential roles for the APC in sensory function have yet to be explored. In this study, we examined the effect of the APC ubiquitin ligase on chemosensation in Caenorhabditis elegans by testing chemotaxis to the volatile odorants, diacetyl, pyrazine, and isoamyl alcohol, to which wild-type worms are attracted. Animals with loss of function mutations in either of two alleles (g48 and ye143) of the gene encoding the APC subunit EMB-27 APC6 showed increased chemotaxis towards diacetyl and pyrazine, odorants sensed by AWA neurons, but exhibited normal chemotaxis to isoamyl alcohol, which is sensed by AWC neurons. The statistically significant increase in chemotaxis in the emb-27 APC6 mutants suggests that the APC inhibits AWA-mediated chemosensation in C. elegans. Increased chemotaxis to pyrazine was also seen with mutants lacking another essential APC subunit, MAT-2 APC1; however, mat-2 APC1 mutants exhibited wild type responses to diacetyl. The difference in responsiveness of these two APC subunit mutants may be due to differential strength of these hypomorphic alleles or may indicate the presence of functional sub-complexes of the APC at work in this process. These findings are the first evidence for APC-mediated regulation of chemosensation and lay the groundwork for further studies aimed at identifying the expression levels, function, and targets of the APC in specific sensory neurons. Because of the similarity between human and C. elegans nervous systems, the role of the APC in sensory neurons may also advance our understanding of human sensory function and disease.

microRNAs (miRNAs) are potent regulators of gene expression that function in a variety of developmental and physiological processes by dampening the expression of their target genes at a post-transcriptional level. In many gene regulatory networks (GRNs), miRNAs function in a switch-like manner whereby their expression and activity elicit a transition from one stable pattern of gene expression to a distinct, equally stable pattern required to define a nascent cell fate. While the importance of miRNAs that function in this capacity are clear, we have less of an understanding of the cellular factors and mechanisms that ensure the robustness of this form of regulatory bistability. In a screen to identify suppressors of temporal patterning phenotypes that result from ineffective miRNA-mediated target repression during  C. elegans  development, we identified  pqn-59,  an ortholog of human UBAP2L, as a novel factor that antagonizes the activities of multiple heterochronic miRNAs.  Specifically, we find that depletion of  pqn-59 can restore normal development in animals with reduced miRNA activity. Importantly, inactivation of  pqn-59  is not sufficient to bypass the requirement of these regulatory RNAs within the heterochronic GRN. The  pqn-59  gene encodes an abundant, cytoplasmically localized and unstructured protein that harbors three essential “prion-like” domains.  These domains exhibit LLPS properties  in vitro  and normally function to limit PQN-59 diffusion in the cytoplasm  in vivo . Like human UBAP2L, PQN-59’s localization becomes highly dynamic during stress conditions where it re-distributes to cytoplasmic stress granules and is important for their formation. Proteomic analysis of PQN-59 complexes from embryonic extracts indicates that PQN-59 and human UBAP2L interact with orthologous cellular components involved in RNA metabolism and promoting protein translation and that PQN-59 additionally interacts with proteins involved in transcription and intracellular transport. Finally, we demonstrate that  pqn-59  depletion results in the stabilization of several mature miRNAs (including those involved in temporal patterning) without altering steady-state pre-miRNAs levels indicating that PQN-59 may ensure the bistability of some GRNs that require miRNA functions by promoting miRNA turnover and, like UBAP2L, enhancing protein translation.