Neural networks of the mouse neocortex

Abstract

Numerous studies have examined the neuronal inputs and outputs of many areas within the mammalian cerebral cortex, but how these areas are organized into neural networks that communicate across the entire cortex is unclear. Over 600 labeled neuronal pathways acquired from tracer injections placed across the entire mouse neocortex enabled us to generate a cortical connectivity atlas. A total of 240 intracortical connections were manually reconstructed within a common neuroanatomic framework, forming a cortico-cortical connectivity map that facilitates comparison of connections from different cortical targets. Connectivity matrices were generated to provide an overview of all intracortical connections and subnetwork clusterings. The connectivity matrices and cortical map revealed that the entire cortex is organized into four somatic sensorimotor, two medial, and two lateral subnetworks that display unique topologies and can interact through select cortical areas. Together, these data provide a resource that can be used to further investigate cortical networks and their corresponding functions.

Figure 1. Strategy for generating the cortical connectivity atlas

A. Schematic illustrating a PHAL/CTb and BDA/FG double co-injection in two different structures labeling both input to, and output from each injection site. Reciprocal interactions between brain regions and circuit interactions between each injection site may also be revealed. B. A coronal section showing co-injections made into the MOs and ACAv viewed with Nissl background to reveal cytoarchitecture. Scale bar: 1 mm. C. Intermixed anterogradely labeled axons (green, PHAL) and retrogradely labeled neurons (pink, CTb) in the MOs following a co-injection in the contralateral hemisphere (first two panels, arrows). Note: the PHAL/CTb co-injection is the same as pictured in B. Image histogram was adjusted differently for the two hemispheres so that PHAL/CTb labeling on the left side can be viewed properly without over exposing the injection site on the right side. Last panel, comparison of retrogradely labeled neurons from injection in ACAv (arrow, yellow, Fluorogold (FG)) and fibers and cells from injection in MOs (PHAL/CTb). Fluorescent Nissl in blue, scale bar 200μm. D. Strategy for mapping fluorescent labeling from a raw image (left, scale bar: 1mm) onto the corresponding level of the ARA (middle) to generate a comprehensive map of projection pathways for all injection sites (right). Note: anterogradely labeled pathways were rendered as layer and regional-specific shading, while retrogradely labeled neurons were represented by individual dots. The large circle on the right hemisphere represents an injection site (see corresponding region on raw image). See Figures S1, S2, S4, and Table S1 for more information.

Figure 2. Weighted and directed cortico-cortical connectivity matrices

Connectivity matrices were constructed based on either anterograde (PHAL, A) or retrograde (FG/CTb, B) tract tracing data. In both matrices, connection origin is listed along the row while targets are listed across the columns (sorted alphabetically). The weighting of each connection is indicated by red (strong), orange (moderate), and yellow (light) coloring. In C and D, the anatomical data in A and B has been re-ordered illustrating a total of 12 distinct modules in different cortical subnetworks. Combining retrograde and anterograde tracing methods formed the composite matrix (E), a consensus perspective of cortico-cortical subnetwork connectivity. See Supplemental Experimental Procedures for details regarding construction of the matrices, Figure S3 for injection cases, and Table S2 for list of abbreviations.

Figure 3. The somatic sensorimotor subnetworks

A. Overview of the four major components of somatic sensorimotor areas (SSp, SSs, MOp, MOs). Each region is extensively interconnected with all others. Parcellation of cortical areas in map based on ARA and drawn to scale. Diamond shape on midline indicates bregma. B. Projections from representative injection sites (colored dots) in each of four basic body representations in primary somatosensory cortex: orofaciopharyngeal (orf, blue), upper limb (ul, green), lower limb and trunk (ll/tr, red), and whisker-related caudomedial barrel field (bfd.cm, yellow). A cartoon (inset) shows approximate size and location of the four body areas defined here (inspired from Brecht et al. 2004). Top-down view (left) shows topographic organization of projections from each area to corresponding primary and secondary motor areas (MOp, MOs). ARA defined boundary between MOs and MOp added for reference. Projection data in top-down view were drawn to scale using coronal sections (right) and shaded regions represent the areal extent of the most dense projections from each of the injected regions. Representative coronal sections also show projection trends to supplemental somatosensory area (SSs). Numbers indicate position of sections relative to bregma (mm). C. Projections from each of the somatosensory sub-regions define presumably functionally related MOs and MOp sub-regions, which are tightly reciprocated, as indicated by closely overlapped axonal fibers and retrogradely labeled cell bodies following co-injection. Here, co-injections of PHAL (green)/CTb (pink) in either SSp-ll (top panel, right) or a corresponding MOp region (middle panel, middle) reveal intermixed labeling in other corresponding domains of the somatic sensorimotor region, confirming their strong reciprocal connectivity. Both co-injections reveal intermixed labeling in the same MOs domain (left images on the top and middle panels). Retrograde injection in the same MOs domain (left image on the lower panel) confirms the specificity of this interaction, showing retrogradely labeled neurons in the former two areas (middle and right images). Their anatomical locations and interactions are summarized in corresponding atlas levels in the bottom panel. Scale bars 500μm and 100μm (inset). D. Building on these observations, four network graphs were created using each of the defined somatosensory regions as starting points. Each subnetwork is distinct and all components within it share a high degree of interconnection. Each are composed of several somatic sensorimotor “nodes” (color coded to match anatomically defined functional domains in B) that are reciprocally connected (as indicated with red arrows). Each of these subnetworks also includes other non-somatic “peripheral” nodes (gray circles) and their connections are shown with gray arrows. See also Figure S5. For abbreviations of nomenclatures, please see Figure 2 and Table S2.

Figure 4. The medial subnetworks

A. Major components of the medial subnetworks, which mediate transduction of information between sensory areas (VIS, AUD, and caudal-most SSp) and higher order association areas along the medial bank of the neocortex, such as the retrosplenial (RSP), parietal (PTLp), anterior cingulate (ACA), and orbital (ORB) areas. B. Connectivity pathways of the medial subnetwork revealed by co-injections in the ORB (note: these are aggregated pathways for three different cases, see three co-injection sites in ORB, colored pink, light brown, and dark brown, medial to lateral). C. Representative raw images from an ORBvl co-injection. PHAL labeled axons and CTb labeled neurons are found in other medial network components such as the ACAd and adjacent MOs-fef, PTLp, RSPd, and primary and secondary VIS areas (VISp, VISal and VISam). Scale bars 500μm (first panel) and 200μm. D. Laminar specific differences in axonal projections to primary visual cortex (VISp) arising from either ORBvl (red, BDA labeling) or ACAd (green, PHAL). Injection sites in the same brain, left panels, scale bars 500μm (left) and 1mm. Projections to different layers of the same section of VISp (right panels). Underlying fluorescent Nissl was inverted to aid visualization of layers (right-most panel). Scale bar 200μm. E. Four different retrograde tracers were injected into the VISp (two), VISam, and VISpm within the same brain, resulting in distinct, topographically arranged clusters of neurons in the ACA and adjacent MOs-fef. In the ORBvl, these retrogradely labeled neurons are intermixed, but mostly not colocalized (bottom right, 5% colocalization for any combination of tracers in ORBvl, 436 cells counted, 21 had two or more tracers present). Scale bars: 1mm (top left), 500μm (bottom left), 200μm (top right), 100μm (bottom right). F. Summary of interactions among the medial subnetworks. Left, interaction between sensory and association areas. Dashed lines indicate sparse connection. Claustrum (CLA) is included due to high degree of interconnection with medial network. Middle, connections between the association areas. Thicker arrows indicate dense projection patterns between regions. Dashed line separates a direct pathway to medial prefrontal region along the ventro-medial bank of the cortex (second medial subnetwork). Asterisks indicate a unidirectional connection between CLA and RSP. Right panel, overview of medial network interactions including TEa and parahippocampal structures (i.e. SUBd, ENTm), which project to RSP (red arrows). Reciprocal connections of visual (blue) and auditory (green) areas with all major medial network components shown. Caudal-most somatosensory areas (SSp-ll/tr; SSp-bfd.cm) are included as well (gray arrows). See also Figure S6.

Figure 5. The lateral subnetworks

A. Sagittal view of the major components of two lateral subnetworks: the anterolateral insular (including the AId, AIv, AIp, VISC, GU) and posterior temporal (including TEa, ECT, PERI). These two interconnected subnetworks are also connected with olfactory (e.g. PIR) and medial prefrontal (mPFC) areas and with the ENTl. TEa in particular forms extensive connections with much of the rest of the neocortex (gray arrows). B. Distinct projection patterns of the anterior agranular areas (PHAL injections involved both AId and AIv, left panel) and AIp (right panel) with the mPFC areas (PL, ILA, DP), posterior temporal areas (TEa, ECT, PERI), and ENTl. The AIp targets more ventral structures in mPFC and more heavily innervates the central nucleus of the amygdala (CEA). Scale bars: 500μm. C. Map of neuronal inputs to (left panel) and output from (right panel) the TEa, which are arranged topographically along the rostrocaudal direction. Note: these pathways are aggregated from 6 co-injections made into different parts of the TEa from rostral (red) to caudal (blue) direction (top most sagittal image, numbers relative to bregma in mm). Retrogradely labeled neurons are indicated as colored dots and demonstrate the layer specific origin of cortical projections to TEa. Axonal pathways arising from TEa (outputs) are rendered as shaded areas of color. D. Raw image of retrograde labeling (FG, yellow) following injection in TEa. Cells are distributed extensively across numerous cortical regions following a single, small injection, suggesting a high level of convergence. Bottom left panel shows close up of layer specificity in somatosensory barrel field, with most cell bodies residing in upper layer 2/3, 5a, and some layer 6. Fluorescent Nissl inverted (right) to aid in discriminating layers. Layer 4 “barrels” indicated with arrow. Scale bars: 500μm(top) and 200μm (bottom). E. Raw image of co-injection in TEa. Fibers are predominately ipsilateral, but retrogradely labeled inputs are evenly distributed across both hemispheres. See right panel, middle, comparing labeling in contralateral and ipsilateral MOs. Scale bars: 1mm (top right), 200μm (middle), 500μm (bottom). See also Figure S7.

Figure 6. The CLA and ENTl

A. Axonal projections arising from the CLA are distributed throughout the entire neocortex and ENTl on the ipsilateral hemisphere. Note: these axons display different regional and laminar distribution specificity (on right panel) Scale bars: 1mm (left), 500μm (top right), 200μm (bottom). Panels B and C show asymmetric connections of the CLA with other cortical areas in the two hemispheres. Cortical inputs to CLA project to both sides with equal densities (B, labels: dorsal and ventral claustrum (CLAd, CLAv), and endopiriform nucleus (EPd)), while outputs from CLA to other cortical areas indicated by retrograde tracers are almost exclusively ipsilateral (C). Moreover, the CLA has a dorsal to ventral topography in its projections to the cortex, with cells in CLAv (yellow, C, right panel) preferentially targeting ventral cingulate (C, left panel, injection site in ACAv). Very little co-labeling was observed among cells labeled from a neighboring, more dorsal injection (pink). Scale bars: 1mm (left) and 200μm (right). D. Neural inputs to the CLA from almost all cortical areas in medial, somatic, and lateral subnetworks. The somatomotor inputs preferentially target the dorsal-most aspect of CLA. E. Representative images of PHAL labeled axons in layer 1 of a wide range of neocortical areas arising from the rostrodorsal ENTl. F. Laminar specificity of PHAL labeled axons and CTb labeled neurons in the ENTl after co-injections made into the AI or ILA. Both cortical regions provide strong, direct input to ENTl, further supported by retrograde data in G. These data also confirm the CLA is a specific source of input to ENTl (G, middle panel). Scale bars: 1mm (F, top left and G, left), 500μm (F, bottom and G, middle), 100μm (F, top right). See also Figure S1B-E.

Figure 7. Interactions with prefrontal cortex

Cumulative projections from components of the somatic sensorimotor, lateral, and medial networks in two representative coronal sections of the prefrontal cortex (PFC) (A-B). Collectively these represent inputs from the entire neocortex to the PFC. Inputs were color coded based on the location of the injection sites in different components of the network (A). For example all primary motor projections arising from multiple injections along the length of this structure were colored green and all somatosensory projections were colored blue (A, top). ACAv was colored red to separate it as a component of the second medial subnetwork (A, bottom, see Fig S6E). Note that RSP has very little interaction with the PFC. B. All inputs from three somatic sensorimotor subnetworks (as shown in A) converge onto three distinct zones, dorsolateral (dl), dorsal (d), and dorsal medial (dm), in the dorsolateral half of the prefrontal cortex (PFCdl, green and blue). In contrast, the medial and lateral subnetworks converge onto the ventromedial half of the prefrontal cortex with distinctive patterns. Note that caudal-most somatosensory and motor regions make some contribution to lateral-most, and caudal aspects of ORB (green and blue shading). C. A schematic view of cortico-cortical network information flow as seen in a top-down view of the cortex (left, lateral edge on left, PFC at the top). All subnetworks are colored according to the scheme used in A and B. Right, a more detailed overview of these interactions (lateral edge of cortex on the right, PFC at the top). Somatic sensorimotor boxes are meant to include both the sensory area and its corresponding primary motor area with which it is strongly interconnected. All functionally distinctive subnetworks are organized along the longitudinal axis of the cerebrum. Information processed in the medial and lateral subnetworks is integrated within the ventromedial half of the prefrontal cortex (PFCvm) and the ENTl. The claustrum (CLA) may also provide an additional means of direct interaction between each of the subnetworks. For abbreviations, please see Figure 2 and Table S2. Additional abbreviations: AMY, amygdala; AH, Ammon’s Horn; HPF, hippocampal formation.