A multimodal cell census and atlas of the mammalian primary motor cortex
Title: A multimodal cell census and atlas of the mammalian primary motor cortex 1
2
Authors: BRAIN Initiative Cell Census Network (BICCN) 3
4
5
ABSTRACT 6
7
We report the generation of a multimodal cell census and atlas of the mammalian primary motor 8
cortex (MOp or M1) as the initial product of the BRAIN Initiative Cell Census Network 9
(BICCN). This was achieved by coordinated large-scale analyses of single-cell transcriptomes, 10
chromatin accessibility, DNA methylomes, spatially resolved single-cell transcriptomes, 11
morphological and electrophysiological properties, and cellular resolution input-output mapping, 12
integrated through cross-modal computational analysis. Together, our results advance the 13
collective knowledge and understanding of brain cell type organization: First, our study reveals a 14
unified molecular genetic landscape of cortical cell types that congruently integrates their 15
transcriptome, open chromatin and DNA methylation maps. Second, cross-species analysis 16
achieves a unified taxonomy of transcriptomic types and their hierarchical organization that are 17
conserved from mouse to marmoset and human. Third, cross-modal analysis provides compelling 18
evidence for the epigenomic, transcriptomic, and gene regulatory basis of neuronal phenotypes 19
such as their physiological and anatomical properties, demonstrating the biological validity and 20
genomic underpinning of neuron types and subtypes. Fourth, in situ single-cell transcriptomics 21
provides a spatially-resolved cell type atlas of the motor cortex. Fifth, integrated transcriptomic, 22
epigenomic and anatomical analyses reveal the correspondence between neural circuits and 23
transcriptomic cell types. We further present an extensive genetic toolset for targeting and fate 24
mapping glutamatergic projection neuron types toward linking their developmental trajectory to 25
their circuit function. Together, our results establish a unified and mechanistic framework of 26
neuronal cell type organization that integrates multi-layered molecular genetic and spatial 27
information with multi-faceted phenotypic properties. 28
29
30
INTRODUCTION 31
32
Unique among body organs, the human brain is a vast network of information processing units, 33
comprising billions of neurons interconnected through trillions of synapses. Across the brain, 34
diverse neuronal and non-neuronal cells display a wide range of molecular, anatomical, and 35
physiological properties that together shape the network dynamics and computations underlying 36
mental activities and behavior. A remarkable feature of brain networks is their self-assembly 37
through the developmental process, which leverages genomic information shaped by evolution to 38
build a set of stereotyped network scaffolds largely identical among individuals of the same 39
species; life experiences then sculpt neural circuits customized to each individual. An essential 40
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.343129doi: bioRxiv preprint
A multimodal cell census and atlas of the mammalian primary motor cortex
step toward understanding the architecture, development, function and neuropsychiatric diseases 41
of the brain is to discover and map its constituent neuronal elements together with the many 42
other cell types that comprise the full organ system. 43
44
The notion of “neuron types”, cells with similar properties as the basic units of brain circuits, has 45
been an important concept since the discovery of stereotyped neuronal morphology over a 46
century ago
1,2
. However, a rigorous and quantitative definition of neuron types has remained 47
surprisingly elusive
3–7
. Neurons are remarkably complex and heterogeneous, both locally and in 48
their long-range axonal projections that can span the entire brain and connect to many target 49
regions. Many conventional technologies analyze one neuron at a time, and often study only one 50
or two cellular phenotypes in an incomplete way (e.g. missing axonal arbors in distant targets). 51
As a result, despite major advances in past decades, until recently phenotypic analyses of neuron 52
types remained severely limited in resolution, robustness, comprehensiveness, and throughput. 53
Besides technical challenges, complexities in the relationship among different cellular 54
phenotypes (multi-modal correspondence) have fueled long-standing debates on how neuron 55
types should be defined
8
. These debates reflect the lack of a biological framework of cell type 56
organization for understanding brain architecture and function. 57
58
In the past decade, single-cell genomics technologies have rapidly swept across many areas of 59
biology including neuroscience, promising to catalyze a transformation from phenotypic 60
description and classification to a mechanistic and explanatory molecular genetic framework for 61
the cellular basis of brain organization
9–12
. These technologies provide unprecedented resolution 62
and throughput to measure the molecular profiles of individual cells, including the complete sets 63
of actively transcribed genes (the transcriptome) and genome-wide epigenetic landscape (the 64
epigenome). Application of single cell RNA-sequencing (scRNA-seq) to the neocortex, 65
hippocampus, hypothalamus and other brain regions has revealed a complex but tractable 66
hierarchical organization of transcriptomic cell types that are consistent overall with knowledge 67
from decades of anatomical, physiological and developmental studies but with an unmatched 68
level of granularity
13–19
. Similarly, single-cell DNA methylation and chromatin accessibility 69
studies have begun to reveal cell type-specific genome-wide epigenetic landscapes and gene 70
regulatory networks in the brain
20–25
. Importantly, the scalability and high information content 71
of these methods allow comprehensive quantitative analysis and classification of cell types, both 72
neuronal and non-neuronal, revealing the molecular basis of cellular phenotypes and properties. 73
Further, these methods are readily applicable to brain tissues across species including humans, 74
providing a quantitative means for comparative analysis that has revealed compelling 75
conservation of cellular architecture as well as specialization of cell types across mammalian 76
species. 77
78
Other recent technological advances have crossed key thresholds to provide the resolution and 79
throughput to tackle brain complexity as well, for example for whole-brain neuronal morphology 80
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.343129doi: bioRxiv preprint
A multimodal cell census and atlas of the mammalian primary motor cortex
and comprehensive projection mapping
26,27
. Furthermore, powerful new methods, including 81
imaging-based single-cell transcriptomics, the combination of single-cell transcriptome imaging 82
and functional imaging, and the integration of electrophysiological recording and single-cell 83
sequencing, allow mapping of the spatial organization, function, and electrophysiological, 84
morphological and connectional properties of molecularly defined cell types
28–32
. Finally, the 85
molecular classification of cell types allows the generation of models for genetic access to 86
specific cell types using transgenic mice and, more recently, short enhancer sequences
33–39
. All 87
of these methods have been applied to brain tissues in independent studies, but not yet in a 88
coordinated fashion to establish how different modalities correspond with one another, and how 89
explanatory a molecular genetic framework is for other functionally important cellular 90
phenotypes. 91
92
Recognizing the unprecedented opportunity to tackle brain complexity brought by these 93
technological advances, the overarching goal of the BRAIN Initiative Cell Census Network 94
(BICCN) is to generate an open-access reference brain cell atlas that integrates molecular, 95
spatial, morphological, connectional, and functional data for describing cell types in mouse, 96
human, and non-human primate brains
40
. A key concept is the Brain Cell Census, similar 97
conceptually to a population census, which accounts for the population of constituent neuronal 98
and non-neuronal cell types, along with their spatial locations and defining phenotypic 99
characteristics that can be aggregated as cellular populations that make up each brain region. 100
This cell type classification scheme, organized as a taxonomy, should aim for a consensus across 101
modalities and across mammalian species (for conserved types). Beyond the cell census, a Brain 102
Cell Atlas would be embedded in a 3D Common Coordinate Framework (CCF) of the brain
41
, in 103
which the precise location and distribution of all cell types and their multi-modal features are 104
registered and displayed. Such a cell-type resolution spatial framework will greatly facilitate 105
integration, interpretation and navigation of various types of information for understanding brain 106
network organization and function. 107
108
Here we present the cell census and atlas of cell types in one region of the mammalian brain, the 109
primary motor cortex (MOp or M1) of mouse, marmoset and human, through an analysis with 110
unprecedented scope, depth and range of approaches (Fig. 1, Table 1). MOp is important in the 111
control of complex movement and is well conserved across species. Decades of accumulated 112
anatomical, physiological, and functional studies have provided a rich knowledge base for the 113
integration and interpretation of cell type information in MOp
42,43
. This manuscript describes a 114
synthesis of results and findings derived from eleven core companion papers through a multi-115
laboratory coordinated data generation within BICCN. We derive a cross-species consensus 116
transcriptomic taxonomy of cell types and identify conserved and divergent gene expression and 117
epigenomic regulatory signatures from a large and comprehensive set of single-cell/nucleus 118
RNA-sequencing, DNA methylation and chromatin accessibility data. Focusing on mouse MOp, 119
we map the spatial organization of transcriptomic cell types by multiplexed error-robust 120
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.343129doi: bioRxiv preprint
A multimodal cell census and atlas of the mammalian primary motor cortex
fluorescence in situ hybridization (MERFISH) and their laminar, morphological and 121
electrophysiological properties by Patch-seq; we report the cell-type resolution input-output 122
wiring diagram of this region by anterograde and retrograde tracing and investigate how axon 123
projection patterns of glutamatergic excitatory neurons correlate with molecularly-defined cell 124
types by Epi-Retro-Seq, Retro-MERFISH (the combination of MERFISH and retrograde 125
labeling), and single-neuron full morphology reconstruction; we describe transgenic driver lines 126
systematically targeting glutamatergic cell types based on marker genes and lineages. Finally, we 127
integrate this vastly diverse array of information into a cohesive depiction of cell types in the 128
MOp region with correlated molecular genetic, spatial, morphological, connectional, and 129
physiological properties and relating them to traditionally described cell types. Such integration 130
is illustrated in detail in example cell types with unique features in MOp: the layer 4 131
intratelencephalic-projecting (L4 IT) cells and layer 5 extratelencephalic-projecting (L5 ET) 132
cells. This multitude of datasets are organized by the BRAIN Cell Data Center (BCDC) and 133
made public through the BICCN web portal www.biccn.org. Key concepts and terms are 134
described in Table 2, including anatomical terms for input and output brain regions for MOp, 135
and hierarchical cell class, subclass and type definitions. 136
137
Table 1. Experimental and computational techniques used in this study and associated 138
datasets 139
Feature
Experimental or
analytic
technique(s) Abbreviations References
Samples (e.g. # of
cells or nuclei) in
MOp/M1
Total samples in
flagship and
companion
papers
Transcription
Single-cell mRNA
sequencing
scRNA-Seq:
SMART-Seq
v4, 10x
Chromium v2,
v3
Background:
15,44
Companion:
45
SMART-seq v4:
6,288 cells (mouse)
10x Chromium v2,
v3:
193,824 cells
(mouse) 1,163,727 cells
Single nucleus
mRNA sequencing
snRNA-Seq:
SMART-Seq
v4, 10x
Chromium v2,
v3
Background:
18,46,47
Companion:
45,48
SMART-seq v4:
6,171 nuclei (mouse)
10,534 nuclei
(human)
10x Chromium v2,
v3:
294,717 nuclei
(mouse)
69,279 nuclei
(marmoset)
15,842 nuclei
(macaque)
76,533 nuclei
(human) 1,100,168 nuclei
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.343129doi: bioRxiv preprint
A multimodal cell census and atlas of the mammalian primary motor cortex
DNA
methylation
Single-nucleus
methylcytosine
sequencing 2 snmC-Seq2
Background:
49
Companion:
45,48,50
9,941 nuclei (mouse)
5,324 nuclei
(marmoset)
5,222 nuclei (human) 110,294 nuclei
Open
chromatin
Single nucleus Assay
for Transposase-
Accessible
Chromatin
snATAC-Seq
Background:
21,51
Companion:
45,52
79,625 nuclei
(mouse) 813,799 nuclei
Combined
transcription/
Open
chromatin
Single-nucleus
chromatin
accessibility and
mRNA expression
sequencing
SNARE-seq2
Background:
53
Companion:
48
9,946 nuclei
(marmoset)
84,178 nuclei
(human)
94,124 nuclei
Spatially
resolved
single-cell
transcriptomi
cs
Multiplexed error-
robust fluorescence
in situ hybridization
MERFISH
Background:
28,29
Companion:
54
~300,000 cells
(mouse)
~300,000 cells
Clustering
and data
integration
methods
Clustering -
Hierarchical iterative
clustering scrattch.hicat
Background:
15,44
Companion:
45,48
Clustering - Metacell
hierarchical
clustering with
dynamic tree pruning
tree-based
method
Companion:
48
Clustering of
snATAC-seq data SnapATAC
Background:
55
Companion:
52
Clustering - Leiden
clustering
Background:
56
Companion:
48
Multimodality and
cross-species
integration
LIGER, Seurat,
SingleCellFusi
on
(SCF),scrattch.
hicat
Background:
44,47,57–60
Companion:
45,48
Statistical
validation
Cross-dataset
replicability analysis MetaNeighbor
Background:
61
Companion:
45,48
Electrophysio
logy, cellular
morphology
and
Combined in vitro
slice physiology,
biocytin cell filling,
cytoplasm extraction
Patch-Seq,
Smart-seq v2
Background:
30,62,63
Companion:
48,64,65
1,237 cells (mouse)
6 cells (macaque)
6 cells (human)
133 cells (mouse)
6 cells (macaque)
391 cells
(human)
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprintthis version posted October 21, 2020. ; https://doi.org/10.1101/2020.10.19.343129doi: bioRxiv preprint
