Artificial neural networks (ANNs) have been shown to predict neural responses in primary visual cortex (V1) better than classical models. However, this performance often comes at the expense of simplicity and interpretability. Here we introduce a new class of simplified ANN models that can predict over 70% of the response variance of V1 neurons. To achieve this high performance, we first recorded a new dataset of over 29,000 neurons responding to up to 65,000 natural image presentations in mouse V1. We found that ANN models required only two convolutional layers for good performance, with a relatively small first layer. We further found that we could make the second layer small without loss of performance, by fitting individual "minimodels" to each neuron. Similar simplifications applied for models of monkey V1 neurons. We show that the minimodels can be used to gain insight into how stimulus invariance arises in biological neurons.

Preprint: https://www.biorxiv.org/content/early/2024/07/02/2024.06.30.601394

Often referred to as a 'fight,' survival involves intense competition over resources. Threat displays and high-intensity attacks are just a few of the aggressive actions exhibited during these contests. Certain motor programs are species-specific, like the vibration of a rattlesnake tail. However, conserved behavioral features are found across species, which appear to be mirrored within the brain. Further parallels have been found across sexes between aggression-promoting contexts and the underlying neuronal circuits. Unraveling the complex web of conserved and variable circuit mechanisms has been considerably advanced by the generation of brain-wiring diagrams in adult female and male Drosophila melanogaster. Here, I will summarize current research, primarily in Drosophila, on how contexts, sensory cues, and internal states regulate aggression across sexes.

Sexual dimorphisms are present across brains. Male and female brains contain sets of cell types with differences in cell number, morphology, or synaptic connectivity between the two sexes. These differences are driven by differentially-expressed transcription factors, which set the stage for disparate sexual and social behaviors observed between males and females, such as courtship, aggression, receptivity, and mating. In the Drosophila brain, sexual dimorphisms result from differential expression of two transcription factors, Fruitless (Fru) and Doublesex (Dsx), and genetic reagents driven by enhancers for Fru and Dsx label sexually-dimorphic neurons in both male and female brains. The recent release of the first whole-brain connectome for Drosophila provides a unique opportunity to study the connectivity between these neurons as well as their integration into the larger brain network. Here, we identify 91 putative Fru or Dsx cell types, comprising ~1400 neurons, within the whole-brain connectome, using morphological similarity between electron microscopic (EM) reconstructions and light microscopic (LM) images of known Fru and Dsx neurons. We discover that while Fru and Dsx neurons are highly interconnected, each cell type typically receives more inputs from and sends more outputs to non-Fru/Dsx neurons. We characterize the connectivity in the Fru/Dsx networks to predict the function of cell types not previously characterized, we measure distances to the sensory periphery and uncover multisensory interactions, and we map connections to descending neurons that drive behavior. All Fru and Dsx labels reported here are shared within FlyWire Codex (codex.flywire.ai; gene==Fruitless or Doublesex); this work is a critical first step towards deciphering the neural basis of sexually-dimorphic behaviors and for making comparisons with future connectomes of the male brain.

De novo protein design has emerged as a powerful strategy with the promise to create new tools. The practical performance of designed fluorophore binders, however, has remained far from meeting fluorescence microscopy demands. Here, we design de novo Rhodamine Binder (Rhobin) tags that combine ideal properties including size, brightness, and now adding hyperstability. Rhobin allows live and fixed cell imaging of a wide range of subcellular targets in mammalian cells. Its reversible fluorophore binding further enables live super-resolution STED microscopy with low photobleaching, as well as PAINT-type single-molecule localization microscopy. We showcase Rhobin in the extremophile Sulfolobus acidocaldarius living at 75 degrees Celsius, an application previously inaccessible by existing tags. Rhobin will serve as the basis for a new class of live cell fluorescent tags and biosensors.

The growing channel count of silicon probes has substantially increased the number of neurons recorded in electrophysiology (ephys) experiments, rendering traditional manual spike sorting impractical. Instead, modern ephys recordings are processed with automated methods that use waveform template matching to isolate putative single neurons. While scalable, automated methods are subject to assumptions that often fail to account for biophysical changes in action potential waveforms, leading to systematic errors. Consequently, manual curation of these errors, which is both time-consuming and lacks reproducibility, remains necessary. To improve efficiency and reproducibility in the spike-sorting pipeline, we introduce here the Spike-sorting Lapse Amelioration System (SLAy), an algorithm that automatically merges oversplit spike clusters. SLAy employs two novel metrics: (1) a waveform similarity metric that uses a neural network to obtain spatially informed, time-shift invariant low-dimensional waveform representations, and (2) a cross-correlogram significance metric based on the earth-movers distance between the observed and null cross-correlograms. We demonstrate that SLAy achieves 85% agreement with human curators across a diverse set of animal models, brain regions, and probe geometries. To illustrate the impact of spike sorting errors on downstream analyses, we develop a new burst-detection algorithm and show that SLAy fixes spike sorting errors that preclude the accurate detection of bursts in neural data. SLAy leverages GPU parallelization and multithreading for computational efficiency, and is compatible with Phy and NeuroData Without Borders, making it a practical and flexible solution for large-scale ephys data analysis.

High-density silicon probes have transformed neuroscience by enabling large-scale neural recordings at single-cell resolution. However, existing technologies have provided limited functionality in nonhuman primates (NHPs) such as macaques. In the present report, we describe the design, fabrication and performance of Neuropixels 1.0 NHP, a high-channel electrode array designed to enable large-scale acute recording throughout large animal brains. The probe features 4,416 recording sites distributed along a 45-mm shank. Experimenters can programmably select 384 recording channels, enabling simultaneous multi-area recording from thousands of neurons with single or multiple probes. This technology substantially increases scalability and recording access relative to existing technologies and enables new classes of experiments that involve electrophysiological mapping of brain areas at single-neuron and single-spike resolution, measurement of spike-spike correlations between cells and simultaneous brain-wide recordings at scale.

Preprint: https://doi.org/10.1101/2023.02.01.526664 

The brain is a disproportionately large consumer of fuel, estimated to expend \~20% of the whole-body energy budget, and therefore it is critical to adequately control brain fuel expenditures while satisfying its on-demand needs for continued function. The brain is also metabolically vulnerable as the inability to adequately fuel cellular processes that support information transfer between cells leads to rapid neurological impairment. We show here that a genetic driver of early onset epileptic encephalopathy (EOEE), SLC13A5, a Na+/citrate cotransporter (NaCT), is critical for gating the activation of local presynaptic glycolysis. We show that SLC13A5 is in part localized to a presynaptic pool of membrane-bound organelles and acts to transiently clear axonal citrate during electrical activity, in turn activating phosphofructokinase 1. We show that loss of SLC13A5 or mistargeting to the plasma membrane results in suppressed glycolytic gating, activity dependent presynaptic bioenergetic deficits and synapse dysfunction.

Prolonged wakefulness leads to persistent, deep recovery sleep (RS). However, the neuronal circuits that mediate this process remain elusive. From a circuit screen in mice, we identified a group of thalamic nucleus reuniens (RE) neurons activated during sleep deprivation (SD) and required for sleep homeostasis. Optogenetic activation of RE neurons leads to an unusual phenotype: presleep behaviors (grooming and nest organizing) followed by prolonged, intense sleep that resembles RS. Inhibiting RE activity during SD impairs subsequent RS, which suggests that these neurons signal sleep need. RE neurons act upstream of sleep-promoting zona incerta cells, and SD triggers plasticity of this circuit to strengthen their connectivity. These findings reveal a circuit mechanism by which sleep need transforms the functional coupling of a sleep circuit to promote persistent, deep sleep.

Representation learning in neural networks may be implemented with supervised or unsupervised algorithms, distinguished by the availability of instruction. In the sensory cortex, perceptual learning drives neural plasticity1-13, but it is not known whether this is due to supervised or unsupervised learning. Here we recorded populations of up to 90,000 neurons simultaneously from the primary visual cortex (V1) and higher visual areas (HVAs) while mice learned multiple tasks, as well as during unrewarded exposure to the same stimuli. Similar to previous studies, we found that neural changes in task mice were correlated with their behavioural learning. However, the neural changes were mostly replicated in mice with unrewarded exposure, suggesting that the changes were in fact due to unsupervised learning. The neural plasticity was highest in the medial HVAs and obeyed visual, rather than spatial, learning rules. In task mice only, we found a ramping reward-prediction signal in anterior HVAs, potentially involved in supervised learning. Our neural results predict that unsupervised learning may accelerate subsequent task learning, a prediction that we validated with behavioural experiments.

Preprint: https://www.biorxiv.org/content/early/2024/02/27/2024.02.25.581990