Faculty Bibliography

Gonadal hormones act throughout the brain ¹ , and neuropsychiatric disorders vary in symptom severity over the reproductive cycle, pregnancy, and perimenopause ^2-4 . Yet how hormones influence cognitive processes is unclear. Exogenous 17 β -estradiol modulates dopamine signaling in the nucleus accumbens core (NAcc) ^5,6 , which instantiates reward prediction errors (RPEs) for reinforcement learning ^7-16 . Here we show that endogenous 17 β -estradiol enhances RPEs and sensitivity to previous rewards by reducing dopamine reuptake proteins in the NAcc. Rats performed a task with different reward states; they adjusted how quickly they initiated trials across states, balancing effort against expected rewards. NAcc dopamine reflected RPEs that predicted and causally influenced initiation times. Elevated endogenous 17 β -estradiol increased sensitivity to reward states by enhancing dopaminergic RPEs in the NAcc. Proteomics revealed reduced dopamine transporter expression. Finally, knockdown of midbrain estrogen receptors suppressed reinforcement learning. 17 β -estradiol therefore controls RPEs via dopamine reuptake, mechanistically revealing how hormones influence neural dynamics for motivation and learning.

Biological accounts of reinforcement learning posit that dopamine encodes reward prediction errors (RPEs), which are multiplied by a learning rate to update state or action values. These values are thought to be represented in synaptic weights in the striatum, and updated by dopamine-dependent plasticity, suggesting that dopamine release might reflect the product of the learning rate and RPE. Here, we leveraged the fact that animals learn faster in volatile environments to characterize dopamine encoding of learning rates in the nucleus accumbens core (NAcc). We trained rats on a task with semi-observable states offering different rewards, and rats adjusted how quickly they initiated trials across states using RPEs. Computational modeling and behavioral analyses showed that learning rates were higher following state transitions, and scaled with trial-by-trial changes in beliefs about hidden states, approximating normative Bayesian strategies. Notably, dopamine release in the NAcc encoded RPEs independent of learning rates, suggesting that dopamine-independent mechanisms instantiate dynamic learning rates.

A key aspect of nutrient absorption is the exquisite division of labour across the length of the small intestine, with individual nutrients taken up at different proximal:distal positions. For millennia, the small intestine was thought to comprise three segments with indefinite borders: the duodenum, jejunum and ileum. By examining the fine-scale longitudinal transcriptional patterns that span the mouse and human small intestine, we instead identified five domains of nutrient absorption that mount distinct responses to dietary changes, and three regional stem cell populations. Molecular domain identity can be detected with machine learning, which provides a systematic method to computationally identify intestinal domains in mice. We generated a predictive model of transcriptional control of domain identity and validated the roles of Ppar-δ and Cdx1 in patterning lipid metabolism-associated genes. These findings represent a foundational framework for the zonation of absorption across the mammalian small intestine.

Action potentials trigger neurotransmitter release at the presynaptic active zone with spatiotemporal precision. This is supported by protein machinery that mediates synaptic vesicle priming and clustering of Ca_V2 Ca²⁺ channels nearby. One model posits that scaffolding proteins directly tether vesicles to Ca_V2s; however, here we find that at mouse hippocampal synapses, Ca_V2 clustering and vesicle priming are executed by separate machineries. Ca_V2 nanoclusters are positioned at variable distances from those of the priming protein Munc13. The active zone organizer RIM anchors both proteins but distinct interaction motifs independently execute these functions. In transfected cells, Liprin-α and RIM form co-assemblies that are separate from Ca_V2-organizing complexes. At synapses, Liprin-α1-Liprin-α4 knockout impairs vesicle priming but not Ca_V2 clustering. The cell adhesion protein PTPσ recruits Liprin-α, RIM and Munc13 into priming complexes without co-clustering Ca_V2s. We conclude that active zones consist of distinct machineries to organize Ca_V2s and prime vesicles, and Liprin-α and PTPσ specifically support priming site assembly.

1Recurrent neural networks (RNN) are ubiquitously used in neuroscience to capture both neural dynamics and behaviors of living systems. However, when it comes to complex cognitive tasks, training RNNs with traditional methods can prove difficult and fall short of capturing crucial aspects of animal behavior. Here we propose a principled approach for identifying and incorporating compositional tasks as part of RNN training. Taking as target a temporal wagering task previously studied in rats, we design a pretraining curriculum of simpler cognitive tasks that reflect relevant sub-computations. We show that this pretraining substantially improves learning efficacy and is critical for RNNs to adopt similar strategies as rats, including long-timescale inference of latent states, which conventional pretraining approaches fail to capture. Mechanistically, our pretraining supports the development of slow dynamical systems features needed for implementing both inference and value-based decision making. Overall, our approach is an important step for endowing RNNs with relevant inductive biases, which is important when modeling complex behaviors that rely on multiple cognitive computations.

The human genome is highly dynamic across all scales. At the gene level, chromatin is persistently remodeled and rearranged during active processes such as transcription, replication and DNA repair. At the genome level, chromatin moves in micron-scale domains that break up and re-form over seconds, but the origin of these coherent motions is unknown. Here, we investigate the connection between genomic motions and gene-level activity. Simultaneous mapping of single-gene and genome-wide motions shows that the coupling of gene transcriptional activity to flows of the nearby genome is modulated by chromatin compaction. A motion correlation analysis suggests that a single active gene drives larger-scale motions in low-compaction regions, but high-compaction chromatin drives gene motion regardless of its activity state. By revealing unexpected connections among gene activity, spatial heterogeneities of chromatin and its emergent genome-wide motions, these findings uncover aspects of the genome's spatiotemporal organization that directly impact gene regulation and expression.

Coarsening is a ubiquitous phenomenon in droplet systems near thermodynamic equilibrium-as an increase in droplet size lowers the system's free energy-however, coarsening of droplets in nonequilibrium systems, such as the cell nucleus, is far from understood. Liquid condensates in the cell nucleus, like nucleoli, form by liquid-liquid phase separation and play a key role in the nuclear organization. In human cells, nucleolar droplets are nucleated at the beginning of the cell cycle and coarsen with time by coalescing with each other. Upon coarsening, human nucleoli exhibit an anomalous volume distribution P(V)∼V^-1, which cannot be explained by any existing theory. In this work, we investigate physical mechanisms behind the anomalous coarsening of human nucleoli. Using spinning disk confocal microscopy, we simultaneously record dynamic behavior of nucleoli and their surrounding chromatin before their coalescence in live human cells. We find that nucleolar anomalous coarsening persists during the entire cell cycle. We measure chromatin flows and density between and around nucleoli, as well as relative motion of two nucleoli before they coalesce. We find that, before nucleolar coalescence, chromatin concentration decreases in the space between nucleoli and the nucleoli move faster toward each other, resembling an effective depletion attraction between the coalescing nucleoli. Indeed, our computational simulations of nucleolar dynamics show that short-ranged attraction is sufficient to explain the observed anomalous volume distribution of human nucleoli. Overall, our results reveal a potential physical mechanism contributing to coarsening of human nucleoli. Such knowledge expands our picture of the physical behavior of liquid condensates inside the cell nucleus and our understanding of the dynamic nuclear organization.

Higher brain functions depend on experience-dependent representations of relevant information that may be organized by attractor dynamics or by geometrical modifications of continuous "neural manifolds". To explore these scenarios we analyzed odor-evoked activity in telencephalic area pDp of juvenile and adult zebrafish, the homolog of piriform cortex. No obvious signatures of attractor dynamics were detected. Rather, olfactory discrimination training selectively enhanced the separation of neural manifolds representing task-relevant odors from other representations, consistent with predictions of autoassociative network models endowed with precise synaptic balance. Analytical approaches using the framework of manifold capacity revealed multiple geometrical modifications of representational manifolds that supported the classification of task-relevant sensory information. Manifold capacity predicted odor discrimination across individuals, indicating a close link between manifold geometry and behavior. Hence, pDp and possibly related recurrent networks store information in the geometry of representational manifolds, resulting in joint sensory and semantic maps that may support distributed learning processes.

Models trained with self-supervised learning objectives have recently matched or surpassed models trained with traditional supervised object recognition in their ability to predict neural responses of object-selective neurons in the primate visual system. A self-supervised learning objective is arguably a more biologically plausible organizing principle, as the optimization does not require a large number of labeled examples. However, typical self-supervised objectives may result in network representations that are overly invariant to changes in the input. Here, we show that a representation with structured variability to input transformations is better aligned with known features of visual perception and neural computation. We introduce a novel framework for converting standard invariant SSL losses into "contrastive-equivariant" versions that encourage preservation of input transformations without supervised access to the transformation parameters. We demonstrate that our proposed method systematically increases the ability of models to predict responses in macaque inferior temporal cortex. Our results demonstrate the promise of incorporating known features of neural computation into task-optimization for building better models of visual cortex.

Corticospinal neurons (CSNs) synapse directly on spinal neurons, a diverse assortment of cells with unique structural and functional properties necessary for body movements. CSNs modulating forelimb behavior fractionate into caudal forelimb area (CFA) and rostral forelimb area (RFA) motor cortical populations. Despite their prominence, the full diversity of spinal neurons targeted by CFA and RFA CSNs is uncharted. Here, we use anatomical and RNA sequencing methods to show that CSNs synapse onto a remarkably selective group of spinal cell types, favoring inhibitory populations that regulate motoneuron activity and gate sensory feedback. CFA and RFA CSNs target similar spinal neuron types, with notable exceptions that suggest that these populations differ in how they influence behavior. Finally, axon collaterals of CFA and RFA CSNs target similar brain regions yet receive highly divergent inputs. These results detail the rules of CSN connectivity throughout the brain and spinal cord for two regions critical for forelimb behavior.