Across mammals, brain regions can duplicate, expand, and diversify, requiring long-range connectivity to accommodate species-specific specializations while preserving globally ordered wiring. This challenge is especially pronounced in primate thalamocortical circuits, where select cortical fields and their thalamic partners have expanded disproportionately. Although gene expression in the thalamus follows broad and conserved gradients, how thalamocortical projections are organized at single-neuron resolution, and how this organization is reshaped by expansion, remain unknown. Here, we investigate thalamocortical projection organization by in situ sequencing and BARseq projection mapping in marmosets and mice. We profiled the gene expression of 1.5 million marmoset neurons and jointly measured gene expression and cortical projections in 708 marmoset and 1,518 mouse neurons that spanned multiple thalamic nuclei. In both species, projections of individual neurons targeted diverse areas that together spanned a large fraction of the cortex. Comparing projections at the single-neuron level and local neighborhood level revealed that marmoset thalamocortical projections were more spatially segregated, producing a point-to-point architecture. Strikingly, this local specialization coexisted with a conserved gradient that predominated over discrete anatomical borders: In the higher-order sensory thalamus of both species, gene expression and projections varied continuously across nucleus borders, and borders had only a small effect on projections. Furthermore, in both marmoset and mouse, gene expression gradients were associated with the anteroposterior locations of cortical targets. These results reconcile discrete nucleus and gradient-based models of thalamic organization and suggest that primate circuit specialization is superimposed on a conserved molecular-projection gradient.

Norepinephrine (NE) is released throughout most of the central nervous system by neurons in the locus coeruleus (LC). We found a relationship between the morphologies, gene expression, and activity of LC-NE neurons in mice. Axonal projections of individual neurons were extensive but largely confined to subsets of brain regions. Axonal projections and graded gene expression correlated with locations of cell bodies in LC. In a behavioral task requiring ongoing learning from actions, neurons in dorsal LC projecting to the cerebral cortex were excited when mice made a different choice from the previous one and by reward prediction errors, a signal driving learning. Background activity of neurons in ventral LC was higher when mice ignored stimuli indicating potential reward availability. These observations reveal a topographically organized structure and function of a neurotransmitter system and show that it contains learning signals for flexible behavior.

Abstract Parallel visual streams segregate information into pathways specialized for distinct computations, but how this segregation is achieved by anatomical segregation of primary visual cortex output remains unclear. This problem is complicated because individual neurons frequently send broadcasting projections to multiple cortical areas, and the target choice depends on both topographical location and molecular identity. Here we developed axonal BARseq2 to jointly map gene expression and high-resolution axonal projections from 1,448 neurons spanning the mouse primary visual cortex (VISp). Axonal BARseq2 recapitulated projection patterns observed by bulk tracing and single-neuron reconstruction, and recovered transcriptomic identities consistent with reference snRNA-seq datasets. Retinotopy strongly predicted projections to individual cortical targets, particularly for areas proximal to VISp, but explained little of which areas are frequently co-innervated. Instead, co-innervation patterns defined three preferential output pathways that largely corresponded to the ventral stream and two subdivisions of the dorsal stream. These pathways were associated with fine-grained transcriptional identities of L4/5 intra-telencephalic neurons, which were further validated with an external MERFISH dataset. Thus, VISp output is organized by two distinct rules: retinotopy constrains where neurons project, whereas cell-type-associated collateralization constrains which targets are co-innervated. This selective broadcasting, in which single neurons reach many higher visual areas in cell-type-specific combinations, could provide an anatomical substrate for visual-stream segregation at the level of VISp output in mice.

Spatial transcriptomics promises to transform our understanding of tissue biology by molecularly profiling individual cells in situ. A fundamental question they allow us to ask is how nearby cells orchestrate their gene expression. Rather than focus on how these cells (samples) communicate with each other, we reframe the problem to investigate how genes (features) coordinate their expression between neighboring cells. To study these phenomena - called cross-expression - we compare all genes to find pairs that coordinate their expression between adjacent cells, thereby avoiding curating gene lists or annotating cell types. Our end-to-end method recovers ligand-receptor pairs as cross-expressing genes and finds gene combinations that mark anatomical regions, complementing marker gene-based region annotation. Leveraging the overlapping genes across different panels, we use multiple atlas-scale adult mouse brain datasets (~25 million cells, 695 samples, 8 technologies) to create an integrated, meta-analytic cross-expression network, whose communities are enriched in spatial processes such as synaptic signaling and G protein coupled receptor activity. Highlighting cross-expression's biological utility, our network shows that genes Drd1 and Gpr6, which are individually implicated in Parkinson's disease (PD) and are being pursued as therapeutic targets, are cross-expressed within the striatum, hinting at their joint role in PD pathophysiology. We provide an efficient R package (https://github.com/gillislab/CrossExpression/) to computationally analyze and visually explore cross-expression patterns, which allow us to better understand how genes coordinate their expression in space to perform tissue-level functions.

Advancements in imaging and molecular techniques enable the collection of subcellular-scale data. Diversity in measured features, resolution, and physical scope of capture across technologies and experimental protocols pose numerous challenges to integrating data with reference coordinate systems and across scales. This paper describes a collection of technologies that we have developed for mapping data across scales and modalities, such as genes to tissues, specifically in a 3D setting. Our collection of technologies include (i) an explicit censored data representation for the partial matching problem mapping whole brains to subsampled subvolumes, (ii) a multi, scale-space optimization technology for generating resampling grids optimized to represent spatial geometry at fixed complexities, and (iii) mutual-information based functional feature selection. We integrate these technologies with our cross-modality mapping algorithm through the use of image-varifold measure norms to represent universally data across scales and imaging modalities. Collectively, these methods afford efficient representations of peta-scale imagery providing the algorithms for mapping from the nano to millimeter scales, which we term cross-modality image-varifold LDDMM (xIV-LDDMM).

Spatial transcriptomics allow us to ask a fundamental question: how do nearby cells orchestrate their gene expression? Rather than focus on how these cells (samples) communicate with each other, we reframe the problem to investigate how genes (features) coordinate their expression between neighboring cells. To this end, we introduce “cross-expression,” which models the degree to which genes coordinate their expression across spatially adjacent cells, avoiding the use of curated databases and cell type labels while controlling for cell-intrinsic processes. We use multiple atlas-scale adult mouse brain datasets ( 25 million cells, 695 slices from 52 brains, 8 technologies) to create an integrated, meta-analytic cross-expression network, whose communities are enriched in spatial processes such as synaptic signaling and G protein coupled receptor activity. Highlighting cross-expression’s biological utility, our network shows that genes Drd1 and Gpr6, which are individually implicated in Parkinson’s disease (PD), are cross-expressed within the striatum, hinting at their joint role in PD pathophysiology. It also recovers ligand-receptor pairs as cross-expressing genes and finds gene combinations that mark anatomical regions, thus complementing cell–cell communication approaches and marker gene-based region annotation, respectively. We offer a gene-centric perspective to understand spatially coordinated expression between neighboring cells. Our method only requires the gene expression and cell location matrices to find cross-expressing gene pairs. The R package is available at https://github.com/gillislab/CrossExpression .

Advancements in imaging and molecular techniques enable the collection of subcellular-scale data. Diversity in measured features, resolution, and physical scope of capture across technologies and experimental protocols pose numerous challenges to integrating data with reference coordinate systems and across scales. This paper describes a collection of technologies that we have developed for mapping data across scales and modalities, such as genes to tissues, specifically in a 3D setting. Our collection of technologies include (i) an explicit censored data representation for the partial matching problem mapping whole brains to subsampled subvolumes, (ii) a multi, scale-space optimization technology for generating resampling grids optimized to represent spatial geometry at fixed complexities, and (iii) mutual-information based functional feature selection. We integrate these technologies with our cross-modality mapping algorithm through the use of image-varifold measure norms to represent universally data across scales and imaging modalities. Collectively, these methods afford efficient representations of peta-scale imagery providing the algorithms for mapping from the nano to millimeter scales, which we term cross-modality image-varifold LDDMM (xIV-LDDMM).

Neurodevelopmental disorders that cause cognitive, behavioural or motor impairments affect around 15% of children and adolescents worldwide1, with diagnoses of profound autism and attention deficit hyperactivity disorder increasing in the USA and contributing to a major economic burden2,3. Yet the origins and mechanisms of these conditions remain poorly understood, limiting progress in therapies. Comprehensive cell atlases of the developing human brain, alongside those of model organisms such as mice and non-human primates, are now providing high-resolution measures of gene expression, cell-type abundance and spatial distribution. In this Perspective, we highlight recent studies that have identified novel developmental cell populations, revealed conserved and divergent patterns of cell genesis, migration and maturation across species, and begun testing hypotheses that link them to processes ranging from transcriptional control of cell fate specification to the emergence of complex behaviours. We present remaining conceptual and technical challenges and provide an outlook on how further studies of human and mammalian brain development can empower a deeper understanding of neurodevelopmental and neuropsychiatric disorders. Future efforts expanding to additional developmental stages, including adolescence, as well as whole-brain, multimodal and cross-species integration, will yield new insights into how development shapes the brain. These atlases promise to serve as essential references for unravelling mechanisms of brain function and disease vulnerability, and for advancing precision medicine.

This paper explicates a solution to building correspondences between molecular-scale transcriptomics and tissue-scale atlases. This problem arises in atlas construction and cross-specimen/technology alignment where specimens per emerging technology remain sparse and conventional image representations cannot efficiently model the high dimensions from subcellular detection of thousands of genes. We address these challenges by representing spatial transcriptomics data as generalized functions encoding position and high-dimensional feature (gene, cell type) identity. We map onto low-dimensional atlas ontologies by modeling regions as homogeneous random fields with unknown transcriptomic feature distribution. We solve simultaneously for the minimizing geodesic diffeomorphism of coordinates through LDDMM and for these latent feature densities. We map tissue-scale mouse brain atlases to gene-based and cell-based transcriptomics data from MERFISH and BARseq technologies and to histopathology and cross-species atlases to illustrate integration of diverse molecular and cellular datasets into a single coordinate system as a means of comparison and further atlas construction.

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.

Comprehensive maps of neuronal connectivity provide a foundation for understanding the structure of neural circuits. In a circuit, neurons are diverse in morphology, electrophysiology, gene expression, activity, and other neuronal properties. Thus, constructing a comprehensive connectivity map requires associating various properties of neurons, including their connectivity, at cellular resolution. A commonly used approach is to use the gene expression profiles as an anchor to which all other neuronal properties are associated. Recent advances in genomics and anatomical techniques dramatically improved the ability to determine and associate the long-range projections of neurons with their gene expression profiles. These studies revealed unprecedented details of the gene-projection relationship, but also highlighted conceptual challenges in understanding this relationship. In this article, I delve into the findings and the challenges revealed by recent studies using state-of-the-art neuroanatomical and transcriptomic techniques. Building upon these insights, I propose an approach that focuses on understanding the gene-projection relationship through basic features in gene expression profiles and projections, respectively, that associate with underlying cellular processes. I then discuss how the developmental trajectories of projections and gene expression profiles create additional challenges and necessitate interrogating the gene-projection relationship across time. Finally, I explore complementary strategies that, together, can provide a comprehensive view of the gene-projection relationship.

Mapping the connectivity of diverse neuronal types provides the foundation for understanding the structure and function of neural circuits. High-throughput and low-cost neuroanatomical techniques based on RNA barcode sequencing have the potential to map circuits at cellular resolution and a brain-wide scale, but existing Sindbis virus-based techniques can only map long-range projections using anterograde tracing approaches. Rabies virus can complement anterograde tracing approaches by enabling either retrograde labeling of projection neurons or monosynaptic tracing of direct inputs to genetically targeted postsynaptic neurons. However, barcoded rabies virus has so far been only used to map non-neuronal cellular interactions in vivo and synaptic connectivity of cultured neurons. Here we combine barcoded rabies virus with single-cell and in situ sequencing to perform retrograde labeling and transsynaptic labeling in the mouse brain. We sequenced 96 retrogradely labeled cells and 295 transsynaptically labeled cells using single-cell RNA-seq, and 4,130 retrogradely labeled cells and 2,914 transsynaptically labeled cells in situ. We found that the transcriptomic identities of rabies virus-infected cells can be robustly identified using both single-cell RNA-seq and in situ sequencing. By associating gene expression with connectivity inferred from barcode sequencing, we distinguished long-range projecting cortical cell types from multiple cortical areas and identified cell types with converging or diverging synaptic connectivity. Combining in situ sequencing with barcoded rabies virus complements existing sequencing-based neuroanatomical techniques and provides a potential path for mapping synaptic connectivity of neuronal types at scale.

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. BARseq interrogates the expression of 104 cell-type marker genes in 10.3 million cells over nine mouse forebrain hemispheres to reveal the role of peripheral inputs on cortical area development.

Mapping the connectivity of diverse neuronal types provides the foundation for understanding the structure and function of neural circuits. High-throughput and low-cost neuroanatomical techniques based on RNA barcode sequencing have the potential to map circuits at cellular resolution and a brain-wide scale, but existing Sindbis virus-based techniques can only map long-range projections using anterograde tracing approaches. Rabies virus can complement anterograde tracing approaches by enabling either retrograde labeling of projection neurons or monosynaptic tracing of direct inputs to genetically targeted postsynaptic neurons. However, barcoded rabies virus has so far been only used to map non-neuronal cellular interactions in vivo and synaptic connectivity of cultured neurons. Here we combine barcoded rabies virus with single-cell and in situ sequencing to perform retrograde labeling and transsynaptic labeling in the mouse brain. We sequenced 96 retrogradely labeled cells and 295 transsynaptically labeled cells using single-cell RNA-seq, and 4130 retrogradely labeled cells and 2914 transsynaptically labeled cells in situ. We found that the transcriptomic identities of rabies virus-infected cells can be robustly identified using both single-cell RNA-seq and in situ sequencing. By associating gene expression with connectivity inferred from barcode sequencing, we distinguished long-range projecting cortical cell types from multiple cortical areas and identified cell types with converging or diverging synaptic connectivity. Combining in situ sequencing with barcoded rabies virus complements existing sequencing-based neuroanatomical techniques and provides a potential path for mapping synaptic connectivity of neuronal types at scale.

This protocol describes the application of BARseq-style in situ sequencing adapted for barcoded rabies virus. Similar procedures for both trans-synaptic tracing and retrograde tracing experiments.

Mapping and molecularly annotating mammalian neural circuits is challenging due to the inability to uniquely label cells while also resolving subcellular features such as synaptic proteins or fine cellular processes. We argue that an ideal technology for connectomics would have the following characteristics: the capacity for robust distance-independent labeling, synaptic resolution, molecular interrogation, and scalable computational methods . The recent development of high-diversity cellular barcoding with RNA has provided a way to overcome the labeling limitations associated with spectral dyes, however performing all-optical circuit mapping has not been demonstrated because no method exists to image barcodes throughout cells at synaptic-resolution. Here we show ExBarSeq, an integrated method combining in situ sequencing of RNA barcodes, immunostaining, and Expansion Microscopy coupled with an end-to-end software pipeline that automatically extracts barcode identities from large imaging datasets without data processing bottlenecks. As a proof of concept, we applied ExBarSeq to thick tissue sections from mice virally infected with MAPseq viral vectors and demonstrated the extraction of 50 barcoded cells in the visual cortex as well as cell morphologies uncovered via immunostaining. The current work demonstrates high resolution multiplexing of exogenous barcodes and endogenous synaptic proteins and outlines a roadmap for molecularly annotated connectomics at a brain-wide scale.

In most sensory modalities, neuronal connectivity reflects behaviorally relevant stimulus features, such as spatial location, orientation, and sound frequency. By contrast, the prevailing view in the olfactory cortex, based on the reconstruction of dozens of neurons, is that connectivity is random. Here, we used high-throughput sequencing-based neuroanatomical techniques to analyze the projections of 5,309 mouse olfactory bulb and 30,433 piriform cortex output neurons at single-cell resolution. Surprisingly, statistical analysis of this much larger dataset revealed that the olfactory cortex connectivity is spatially structured. Single olfactory bulb neurons targeting a particular location along the anterior-posterior axis of piriform cortex also project to matched, functionally distinct, extra-piriform targets. Moreover, single neurons from the targeted piriform locus also project to the same matched extra-piriform targets, forming triadic circuit motifs. Thus, as in other sensory modalities, olfactory information is routed at early stages of processing to functionally diverse targets in a coordinated manner.

This protocol describes the application of BARseq as a standalone in situ sequencing method to achieve multiplexed interrogation of endogenous genes. In this variation, BARseq is similar to in situ sequencing (ISS), but uses Illuimna SBS for sequencing readout.

1 The cortex is composed of neuronal types with diverse gene expression that are organized into specialized 2 cortical areas. These areas, each with characteristic cytoarchitecture 1-3 , connectivity 4,5 , and neuronal 3 activity 6-10 , are wired into modular networks 4,5,11 . However, it remains unclear whether cortical areas and 4 their modular organization can be similarly defined by their transcriptomic signatures. Here we used 5 BARseq, a high-throughput in situ sequencing technique, to interrogate the expression of 107 cell type 6 marker genes in 1.2 million cells over a mouse forebrain hemisphere at cellular resolution. De novo 7 clustering of gene expression in single neurons revealed transcriptomic types that were consistent with 8 previous single-cell RNAseq studies 12,13 . Within medium-grained cell types that are shared across all 9 cortical areas, gene expression and the distribution of fine-grained cell types vary along the contours of 10 cortical areas. The compositions of transcriptomic types are highly predictive of cortical area identity. We 11 grouped cortical areas into modules so that areas within a module, but not across modules, had similar 12 compositions of transcriptomic types. Strikingly, these modules match cortical subnetworks that are highly 13 interconnected 4,5,11 , suggesting that cortical areas that are similar in cell types are also wired together. This 14 “wire-by-similarity” rule reflects a novel organizing principle for the connectivity of cortical areas. Our 15 BARseq-based strategy is high-throughput and low-cost, and scaling up this approach to many animals can 16 potentially reveal the brain-wide molecular architecture across individuals, developmental times, and 17 disease models. 18