Anthony M. Zador

Elucidating how modifications in neural circuit architecture drive behavioural innovation remains a key challenge in neuroscience and evolutionary biology. In mammals, the neocortex is posited to play a crucial part in facilitating rapid behavioural innovations^1,2,3. Although changes in long-range connectivity have been proposed to underlie such innovations^4,5, these hypotheses remain largely untested quantitatively, which is partly due to the lack of high-throughput neuronal projection data at single-neuron resolution across species. Here we studied the Alston’s singing mouse (Scotinomys teguina), which exhibits a striking vocal behaviour absent in the laboratory mouse (Mus musculus), to quantitatively determine species-specific changes in motor cortical projections throughout the brain. We used bulk tracing, serial two-photon tomography and high-throughput DNA sequencing of more than 76,000 barcoded neurons to discover a specific and substantial expansion of orofacial motor cortical projections to an auditory cortical region and the midbrain periaqueductal grey, regions that are implicated in vocal behaviours^6,7,8,9. Moreover, analyses of projection motifs of individual orofacial motor cortical neurons revealed preferential expansion of exclusive projections to the auditory cortical region in the singing mouse. Our results suggest that selective expansion of ancestral motor cortical projections may lead to behavioural divergence over short timescales. Furthermore, the results facilitate mechanistic investigations of enhanced cortical control over vocalizations—a crucial preadaptation for human language^10,11. This approach of comparing recently diverged species with substantial behavioural divergences can be readily generalized across other model clades to discover quantitative rules of neural circuit evolution.

The collection of articles on the theme Evolution and Development of Neural Circuits explores how brains are built, diversified, and adapted in a variety of species, integrating perspectives from evolutionary biology, developmental neuroscience, and systems neurobiology. Recent advances in molecular genetics, neuroanatomy, physiology, imaging, and computational modeling have enabled unprecedented insights into the mechanisms that shape neural circuits. This collection brings together contributions from leading investigators who examine the architecture and function of neural circuits from multiple angles. Key themes include the evolutionary divergence and convergence of circuit motifs, the conserved molecular and developmental building blocks that underlie connectivity, and the selective pressures that sculpt neural systems to support behavior and cognition. Articles cover topics ranging from retinal mapping and interneuron diversity to thalamocortical connectivity, prefrontal circuit maturation, and the computational modeling of both normal and abnormal circuit development. Collectively, these essays reveal how molecular signaling, cellular variability, and theoretical principles converge to shape the formation and function of circuits across vertebrate and invertebrate brains.

Animals typically learn to solve decision tasks in the laboratory through trial and error, rather than through explicit instruction of decision rules. The decision rule used by the animal may be difficult to read out directly from choice or accuracy data, when multiple decision rules are possible given the task design. Here, we demonstrate that in rats performing a visual decision task, probe stimuli can be used to gain information about decision strategy, and in our task revealed variation in the decision strategies used across rats. Further, we find that in a more general version of this task, rats use varying decision strategies that differ from the optimal ideal observer strategy, but respond to manipulations of the stimulus distribution by adjusting their behavioral strategy. Therefore, we show that informative probe stimuli can be used in both training and testing to confirm and shape behavioral strategy in a perceptual decision task.

The neural representations of prior information about the state of the world are poorly understood¹. Here, to investigate them, we examined brain-wide Neuropixels recordings and widefield calcium imaging collected by the International Brain Laboratory. Mice were trained to indicate the location of a visual grating stimulus, which appeared on the left or right with a prior probability alternating between 0.2 and 0.8 in blocks of variable length. We found that mice estimate this prior probability and thereby improve their decision accuracy. Furthermore, we report that this subjective prior is encoded in at least 20% to 30% of brain regions that, notably, span all levels of processing, from early sensory areas (the lateral geniculate nucleus and primary visual cortex) to motor regions (secondary and primary motor cortex and gigantocellular reticular nucleus) and high-level cortical regions (the dorsal anterior cingulate area and ventrolateral orbitofrontal cortex). This widespread representation of the prior is consistent with a neural model of Bayesian inference involving loops between areas, as opposed to a model in which the prior is incorporated only in decision-making areas. This study offers a brain-wide perspective on prior encoding at cellular resolution, underscoring the importance of using large-scale recordings on a single standardized task.

Neurons in the cortex are heterogeneous, sending diverse axonal projections to multiple brain regions. Unraveling the logic of these projections requires single-neuron resolution. Although a growing number of techniques have enabled high-throughput reconstruction, these techniques are typically limited to dozens or at most hundreds of neurons per brain, requiring that statistical analyses combine data from different specimens. Here we present axonal BARseq, a high-throughput approach based on reading out nucleic acid barcodes using in situ RNA sequencing, which enables analysis of even densely labeled neurons. As a proof of principle, we have mapped the long-range projections of >8000 primary auditory cortex neurons from a single male mouse. We identified major cell types based on projection targets and axonal trajectory. The large sample size enabled us to systematically quantify the projections of intratelencephalic (IT) neurons, and revealed that individual IT neurons project to different layers in an area-dependent fashion. Axonal BARseq is a powerful technique for studying the heterogeneity of single neuronal projections at high throughput within individual brains.

Deciphering how cortical architecture evolves to drive behavioral innovations is a long-standing challenge in neuroscience and evolutionary biology. Here, we leverage a striking behavioral novelty in the Alston’s singing mouse ( Scotinomys teguina ), compared to the laboratory mouse ( Mus musculus ), to quantitatively test models of motor cortical evolution. We used bulk tracing, serial two-photon tomography, and high-throughput DNA sequencing of over 76,000 barcoded neurons to discover a specific and substantial expansion ( ∼ 200%) of orofacial motor cortical (OMC) projections to the auditory cortical region (AudR) and the midbrain periaqueductal gray (PAG), both implicated in vocal behaviors. Moreover, analysis of individual OMC neurons’ projection motifs revealed preferential expansion of exclusive projections to AudR. Our results imply that selective expansion of ancestral motor cortical projections can underlie behavioral divergence over short evolutionary timescales, suggesting potential mechanisms for the evolution of enhanced cortical control over vocalizations—a crucial preadaptation for human language.

Animals are born with extensive innate behavioral capabilities, which arise from neural circuits encoded in the genome. However, the information capacity of the genome is orders of magnitude smaller than that needed to specify the connectivity of an arbitrary brain circuit, indicating that the rules encoding circuit formation must fit through a “genomic bottleneck” as they pass from one generation to the next. Here, we formulate the problem of innate behavioral capacity in the context of artificial neural networks in terms of lossy compression of the weight matrix. We find that several standard network architectures can be compressed by several orders of magnitude, yielding pretraining performance that can approach that of the fully trained network. Interestingly, for complex but not for simple test problems, the genomic bottleneck algorithm also captures essential features of the circuit, leading to enhanced transfer learning to novel tasks and datasets. Our results suggest that compressing a neural circuit through the genomic bottleneck serves as a regularizer, enabling evolution to select simple circuits that can be readily adapted to important real-world tasks. The genomic bottleneck also suggests how innate priors can complement conventional approaches to learning in designing algorithms for AI.

The development of neural connectivity is a crucial biological process that gives rise to diverse brain circuits and behaviors. Neural development is a stochastic process, but this stochasticity is often treated as a nuisance to overcome rather than as a functional advantage. Here we use a computational model, in which connection probabilities between discrete cell types are genetically specified, to investigate the benefits of stochasticity in the development of neural wiring. We show that this model can be viewed as a generalization of a powerful class of artificial neural networks—Bayesian neural networks—where each network parameter is a sample from a distribution. Our results reveal that stochasticity confers a greater benefit in large networks and variable environments, which may explain its role in organisms with larger brains. Surprisingly, we find that the average fitness over a population of agents is higher than a single agent defined by the average connection probability. Our model reveals how developmental stochasticity, by inducing a form of non-heritable phenotypic variability, can increase the probability that at least some individuals will survive in rapidly changing, unpredictable environments. Our results suggest how stochasticity may be an important feature rather than a bug in neural development.

The cerebral cortex is composed of neuronal types with diverse gene expression that are organized into specialized cortical areas. These areas, each with characteristic cytoarchitecture ^1,2 , connectivity ^3,4 and neuronal activity ^5,6 , are wired into modular networks ^3,4,7 . However, it remains unclear whether these spatial organizations are reflected in neuronal transcriptomic signatures and how such signatures are established in development. Here we used BARseq, a high-throughput in situ sequencing technique, to interrogate the expression of 104 cell-type marker genes in 10.3 million cells, including 4,194,658 cortical neurons over nine mouse forebrain hemispheres, at cellular resolution. De novo clustering of gene expression in single neurons revealed transcriptomic types consistent with previous single-cell RNA sequencing studies ^8,9 . The composition of transcriptomic types is highly predictive of cortical area identity. Moreover, areas with similar compositions of transcriptomic types, which we defined as cortical modules, overlap with areas that are highly connected, suggesting that the same modular organization is reflected in both transcriptomic signatures and connectivity. To explore how the transcriptomic profiles of cortical neurons depend on development, we assessed cell-type distributions after neonatal binocular enucleation. Notably, binocular enucleation caused the shifting of the cell-type compositional profiles of visual areas towards neighbouring cortical areas within the same module, suggesting that peripheral inputs sharpen the distinct transcriptomic identities of areas within cortical modules. Enabled by the high throughput, low cost and reproducibility of BARseq, our study provides a proof of principle for the use of large-scale in situ sequencing to both reveal brain-wide molecular architecture and understand its development.