Distinct subnetworks of the mouse anterior thalamic nuclei

Abstract

Currently, classification of neuron types in the mouse thalamus remains largely incomplete. The anterior thalamic nuclei (ATN), a Papez circuit component, encompass the anterodorsal (AD), anteroventral (AV), and anteromedial (AM) thalamic nuclei. Structurally, the ATN facilitate communication among the neocortex, hippocampus, amygdala, and hypothalamus. Functionally, they play pivotal roles in learning, memory, spatial navigation, and goal-directed behaviors. Therefore, the ATN provide a promising avenue to investigate the relationship between structural and functional complexity with neuron type diversity. In male mice, comprehensive, systematically collected, pathway tracing data revealed several connectionally unique ATN cell populations, suggesting multiple parallel subnetworks run through each nucleus. Further, we applied genetic sparse labeling, brain clearing, 3D microscopic imaging, and computational informatics to morphologically characterize and catalog ATN neuron types. This study provides insights into how the prefrontal cortex, hippocampus, and amygdala interact through neuron type-specific ATN subnetworks to coordinate cognitive and emotional aspects of goal-directed behavior.

Fig. 1. Approach.

a Tissue sections from tracer injection cases are imported into Outspector, matched to their corresponding ARA atlas template level, and fiducial markers placed within the tissue Nissl channel are aligned with those in the atlas template. Tissue images are then deformably warped (e.g., registered) and (b) the tracer labeling is segmented (e.g., thresholded). The ATN is subdivided into 105 px²  grid cells and the axon (pixels) or cell (cell count) labeling within each grid square is quantified. c The aggregated quantified data from all injections to all the grid cells is then analyzed using a modularity maximization algorithm to group injection sites with terminations in common grid cells. The darker shading indicates stronger connections. d The output is visualized in a color-coded matrix to illustrate the subnetworks within each nucleus. e The generated data was used to create the whole-brain wiring diagram of the AD, AV, and AM and their identified domains. f The overall organization of the ATN nuclei, followed by the unique connections of each nucleus, e.g., SUBd→AV, AD/AV→PRE, IAD→CP, PL↔AM, RSPv→AD/AV, and AM→ACAv. g The laminar organization of thalamo-cortical projections and cortico-thalamic projection neurons. An anterograde and retrograde tracer co-injection in the AM shows labeled thalamo-cortical projections to layers I–V and layer VI corticothalamic cells. The bottom left panel shows a similar laminar organization of thalamocortical projections from the AV. Right panels show laminar organization of hippocampal-thalamic projection neurons and thalamo-hippocampal projection terminals. A CTB retrograde tracer injection in the AV shows layer III labeled POST, PRE, and PAR projection neurons. An anterograde PHAL injection in the AD shows projections to layers I-III of POST, PRE, and PAR. See Supplementary Table 2 for list of structure abbreviations.

Fig. 2. Connectivity of the AD.

a Based on network analysis of their connectivity, the AD was subdivided into two domains: the AD.medial and AD.lateral, across ARA 61 and 63. The matrices represent the percentage of a specified grid covered by axons or cells from each injection site. These data were subject to a modularity maximization algorithm that grouped injection sites (rows) with labels in common grid cells (columns) within the AD. The matrix was reordered to present the identified groups (i.e., domains) along the diagonal. Domain names are listed above the matrices. Finally, each AD pixel grid was recolored according to their community structure to visualize the domains on the atlas AD. A total of 6 cases (6 sections with AD labels at ARA 61 and another 6 sections at ARA 63) were included in the analysis. Injection ROIs are indicated in the rows of the matrix. b CTB retrograde tracer injections in the PRE and PAR show labeled neurons specifically in the AD.medial at ARA levels 61 and 63. c Anterograde (PHAL) and retrograde (FG) injections in the RSPv label mostly the AD.lateral at ARA 61 and 63. d Double retrograde injections in the PRE and RSPv in the same animal clearly demonstrate the AD.medial and AD.lateral distinction. e Numerous anterograde tracer injections placed across the POST, PRE, and PAR did not label axons in the AD. f AD.medial and AV.dorsal project to PRE (AD.medial/AV.dorsal→PRE), but PRE projects back only to AV (AV←PRE). A co-injection of an anterograde and retrograde into the PRE retrogradely labels AD.medial and AV.dorsal neurons, but anterogradely labeled terminals are present only in the AV.dorsal. g No connections were detected between the SUB and the AD. Anterograde tracers in the SUBd and SUBv do not label terminals in the AD. h Anterograde and retrograde injections show no labeling in the AD. i. Left is a canonical schematic of connections among the SUB, AD, RSP, and LM with the AD. Right is an updated version of those connections with the medial and lateral subnetworks running through the AD.

Fig. 3. Connectivity of the AV.

a The AV is subdivided into 4 domains, AV.medial tip, AV.medial, AV.dorsal, and AV.lateral, based on connectivity. The matrix presents Louvain-identified AV domains and shows the injection sites that group together based on the similarity of axonal or cell labeling they produce at each AV grid box. Nineteen cases (19 sections at ARA 61) were included in the analysis (injections are listed in the rows). Repeated injections for a single ROI (e.g., SUBd) were included to show they group together. Top: Recolored AV grids visualize the domains, and regions that each domain connects with are listed. b1 CTB in AV backlabels neurons in POST, PRE, PAR deep layers. MMl labeled cells verify the AV injection location. b2 Anterograde tracers in POST and PRE (labeling in LM and LD verify injection location) show POST/PRE/PAR (layer III)→AV.dorsal. Note the absence of labels in AD, suggesting no POST, PRE, PAR inputs to AD. c1 AV PHAL injection labels all layers of POST/PRE/PAR (AV→POST/PRE/PAR). c2 Retrograde injections in POST, PRE, PAR in a single animal reveal this as AV.dorsal→POST/PRE/PAR. d RSPd and RSPagl tracer injections show their connections with AV.dorsal (RSPv/RSPagl↔AV.dorsal). e RSPv injections show their connections with AV.lateral (RSPv↔AV.lateral). f Multiple tracer injections illustrate AV domain distinctions. g Retrograde tracer injection in AV.dorsal/AV.lateral validates POST/PRE/PAR→AV.dorsal and the RSPagl→AV.dorsal connections. A retrograde tracer primarily in AV.medial/medial tip labels SUBv (not POST/PRE/PAR) and RSPv (not RSPagl), corroborating the distinct SUBv/RSPv→AV.medial tip connections. h AV.medial tip domain receives inputs from SUBv (MM label confirms SUBv injection site) and i RSPv deep layers. j The SUBd provides input to AV.lateral. SUBd→MM projection verifies SUB injection location. k Cre-dependent AAV injections in SUBd (red) and SUBv (green; AAVretro-Cre in AV) show their projections to distinct AV subregions (SUBd→AV.lateral; SUBv→AV.medial tip). SUBd and SUBv projections to MM validate their injection locations. l1 PHAL AV injection labels caudal parts of SUBv (AV→caudal SUBv), while l2 a caudal SUBv CTB injection validates the connection and demonstrates it to be with AV.lateral (AV.lateral→caudal SUBv). m Wiring diagram summarizing AV domain connections. n ATN-SUB wiring diagram. See Supplementary Table 2 for abbreviations.

Fig. 4. Connectivity of the AMd.dorsomedial and AMd.dorsal domains.

a Tracer injections across cortex and hippocampus label different AMd domains (ARA 61). Manually mapped tracer labels highlight their distinct locations in the AM. b AM grids recolored to visualize the domains identified by the Louvain are shown, and the ROIs that each domain connects with are listed. Note the common connections of AMd.medial and AMv with MPF areas, also evident in the raw data. c The matrix presents the identified Louvain groups (i.e., domains) and shows the injection sites that are grouped together based on the similarity of labeling they produce at each AM grid box. Sixteen cases (16 sections with AM labeling at ARA 61) were included in the analysis. Injection ROIs are indicated in the rows of the matrix. d A PHAL/CTB AM co-injection labels cells (layer VI) and terminals (layers I-V) in rostral (left) and caudal (right) ACAv/ACAd, suggesting rostral/caudal ACA↔AMd connections. Dashed circles denote location of co-injections in (e, f). e A PHAL/CTB co-injection made in rostral ACA labels terminals and cells specifically in the AMd.dorsomedial domain (rostral ACA↔AMd.dorsomedial) and AMv (rostral ACA↔AMv). f A PHAL/CTB co-injection made in caudal ACA labels terminals and cells specifically in the AMd.dorsal domain (rostral ACA↔ AMd.dorsal) and does not label AMv neurons. g A Cre-dependent AAV injection that traces mostly AMd.dorsal (+dorsolateral) neurons (AAVretro-Cre injection made in ACA/RSP) shows projections only in caudal ACA, confirming caudal ACA↔AMd.dorsal connections. h Quantified projections to ACAd, ACAv, PL, and ILA from traced AMd.dorsal neurons in (g), validate their stronger projections to ACAv compared to other ROIs like PL and ILA. i Multiple tracer injections across different ROIs in the same brain showcase the distinct AMd domains. j Summarized connections of the AMd.dorsomedial (left) and AMd.dorsal (right). k Rostral and caudal ACAv injections show that ACA connectivity is selectively with AM, and not AV nor AD. l This selective AM-ACA connection is also shown with tracer injections made in the AV and AMd of the same brain. Neither anterograde nor retrograde AV injections produce labels in ACA, while AMd retrograde injection labels neurons in ACA deep layers.

Fig. 5. Connectivity of the AMd.dorsolateral and AMd.lateral domains.

a A PHAL/CTB AM injection labels only fibers in the RSPv. Dashed circle denotes the location of a PHAL/CTB co-injection in the RSPv in (b). b PHAL/CTB co-injection back labels cells in the AMd.dorsolateral domain (AMd.dorsolateral→RSPv). Note the absence of fibers in the AMv. c A Cre-dependent anterograde injection that tracers AMd.dorsolateral neurons labels the RSPv. d Quantified projections to the PL, ILA, RSPv, PTLp, and MOs from these traced neurons in the AMd.dorsolateral domain show strongest projections to the RSPv compared to other regions. e Wiring diagram of the AMd.dorsolateral domain. f A PHAL/CTB anterograde and retrograde co-injection labels terminals and cells in the PTLp. Dashed circle denotes location of co-injection in (g). g A PHAL/CTB injection in the PTLp (caudal, medial part) labels terminals and cells in the AMd.lateral domain. h Cre-dependent AAV injections made in the AMd.medial tip (left: AAVretro-Cre in PL/ILA) and in the AMd.lateral (right; AAVretro-Cre in PTLp). Only traced neurons in the AMd.lateral label the PTLp, specifically the caudal medial part (AMd.lateral↔PTLp caudal medial). i Bar graph shows the quantification of these projections to the PTLp, which are greater from traced AMd.lateral neurons versus those traced from the AMd.medial. j Connections with the PTLp are exclusively through the AM. k Wiring diagram of the AMd.lateral connections. l An AM CTB injection labels SUBv (caudal) cells. Dashed circle denotes location of an AAV injection made in the same SUBv caudal region that labels terminals in AMd.lateral (SUBv→AMd.lateral). m Multiple tracer injections in a single brain show the distinct AMd.lateral and AMd.dorsolateral domains. See Supplementary Table 2 for structure abbreviations.

Fig. 6. Connectivity of the AMd.medial tip and AMd.ventromedial domains.

a The AMd.medial can be divided into the AMd.medial tip and AMd.ventromedial domains, each of which shows distinct connections. b An AM CTB injection labels cell in the SUBv (top), PL (middle), and ILA (bottom). Dashed circles denote locations of PHAL injections in (c). c PHAL injections in the SUBv (top), PL (middle), and ILA (bottom) show their distinct connections with the AMd.medial tip and AMv (SUBv/PL/ILA→AMd.medial tip; PL/ILA→AMv). d An AM AAV injection labels fibers in the PL (top), ILA (middle), and BLA (bottom). Dashed circles denote location of retrograde tracer injections in (e). e Retrograde tracer injections in the PL (top), ILA (middle), and BLA (bottom) show their connections with the AMd.medial tip, AMd.ventromedial, and the AMv (AMd.medial tip/AMv→PL; AMd.medial tip/ventromedial/AMv→ILA; AMd.medial tip→BLA). f Cre-dependent AAV injections traced AMd.medial tip neurons (green; AAVretro-Cre in MPF) or avoided those neurons and traced only those in the AMd.core (red; AAVretro-Cre in ACA/MOs). Only the AMd.medial tip traced neurons labeled terminals in the PL, ILA, and BLA validating those connections showing the specificity of the domain-level connections. g A CTB AM injection (left) labels neurons in the ORBm. Dashed circle denotes location of the PHAL and CTB ORBm injections in the right, which show labeled fibers and cells in the AMd.ventromedial domain (AMd.ventromedial↔ORBm). h A Cre-dependent AAV injection that traces primarily AMd.ventromedial neurons (AAVretro-Cre in the MPF) labels terminals in the ORBm and ILA validating those projections. i Double retrograde tracer injections in the PL and ILA in the same animal show (1) the distinct AMd.medial tip and AMd.ventromedial domains, (2) the select AMd.ventromedial/AMv→PL connections, and (3) less selective AMd.medial tip/AMd.ventromedial/AMv→ILA projections. j Quantification of projections to the PL and ILA from the Cre-dependent AAV injections that traced neurons in either the AMd.medial tip or the AMd.ventromedial validates the stronger AMd.medial tip→PL and the AMd.ventromedial→ILA connections. k Wiring diagrams summarizing the connections of the AMd.ventromedial (top), AMd.medial tip (middle), and AMv (bottom) domains. See Supplementary Table 2 for structure abbreviations.

Fig. 7. Connectivity of the AMd.core domain, domain networks, and single neuron validations.

a Multiple tracer injections display the AMd periphery/core organization. b Connections of AMd.core. c Labeled terminals and cells in MOs-fef following AAV (left) and CTB (right) AM injections (AMd↔MOs-fef). d PHAL/CTB MOs-fef co-injection confirms AMd↔MOs-fef reciprocity and reveals specificity of the connection with AMd.core. Note the periphery/core structure. e Cre-dependent AAV tracing of AMd.core neurons validates projections to MOs-fef (left) (AAVretro-Cre in ACA), with quantified data showing that the strongest projections to MOs stem from AMd.core, compared to AMd.ventromedial, AMd.medial tip/dorsomedial, and AMv (right). f PHAL AM injection labels terminals in ECT, PERI, ENTl layer  V (L5). A retrograde tracer ENTl L5 injection (dashed circle at ARA 97) labels neurons in AMd.core, showing that projections to ENTl originate from AMd.core. g Cre-dependent AAV tracing of AMd.core neurons (AAVretro-Cre in ACA) shows consistent projections to ECT, PERI, and ENTl L5. h Cre-dependent AAV injection not tracing AMd.core neurons shows no labels in ECT, PERI, or ENTl. i Quantification of projections from neurons traced from AMd.core, AMv, and AMd.dorsal/dorsolateral reveals that the strongest projections to ENTl, PERI, and ECT arise from AMd.core, with no projections to ENTm. j Hierarchical clustering of Cre-dependent AAV tracing from AMd domains to medial prefrontal and higher-order cortical areas. Output validates the unique or preferential connectivity of the proposed domains indicated by the asterisks. k Hierarchical clustering of projections from injections made in medial versus lateral AMd (n = 1 each) validate their distinct connections. Medial injection shows strongest projections to ILA and PL, while the lateral targets RSPv, RSPd, and PTLp. l Schematic of AM connections with ROIs that process visual/spatial information (top). Schematic of AMd.medial connections with MPF areas shown in Supplementary Fig. 4k. Dashed gray line represents the approximate division between ventral-medial domains that connect with limbic regions like MPF, HPF, amygdala (bottom left), and the dorsal-lateral AMd domains that connect with visual/spatial processing areas like ACA, RSP, PTLp (bottom right). m Normalized projections from axon reconstructions of traced neurons in AMd.medial, AMd.dorsal/dorsolateral, and AMd.core correlate with the mesoscale connections, showing that the AMd.medial neuron projects mostly to PL and ILA, the AMd.dorsal/dorsolateral to ACAv and RSPv, and the AMd.core to MOs. See Supplementary Table 2 for structure abbreviations.

Fig. 8. Medial prefrontal, hippocampal, and amygdalar synaptic circuits through AMd.medial tip and AMv.

a Injection strategy used in MORF3 mice to validate the SUBv→AMd.medial tip→BLA disynaptic circuit. b A synaptophysin tagged anterograde AAV injection made in the SUBv labels terminals in the AMd.medial tip, while an AAVretro-Cre injection in the BLA back labels neurons in the AMd.medial tip. c 60x magnification of a BLA projecting AMd.medial tip neuron. d The same AMd.medial tip→BLA neuron in (c) merged with labeled synaptic terminals from the SUBv. e Note the close apposition of the neuron processes and terminals suggesting putative synaptic contacts of SUBv terminals onto the BLA projecting AMd.medial tip neuron. f, g Putative contacts were reconstructed and validated via Neurolucida (see “Methods” for details). h Injection strategy used in MORF3 mice to validate the SUBv→AMd.medial tip→PL disynaptic circuit. i A synaptophysin-tagged anterograde AAV injection made in the SUBv labels terminals in the AMd.medial tip, while an AAVretro-Cre injection in the PL back labels neurons in the AMd.medial tip. j 60x magnification of a PL projecting AMd.medial tip neuron. k The same AMd.medial tip→PL neuron in (j) merged with labeled synaptic terminals from the SUBv. l Close apposition of neuron processes and terminals suggests putative synaptic contacts of SUBv terminals onto the PL projecting AMd.medial tip neuron. m Putative contacts were validated via Neurolucida same as in (f–g). n The same injection strategy in a and h was applied in MORF3 mice to validate the ILA→AMv→PL circuit. o Synaptic contacts of ILA projections onto PL projecting AMv neurons were reconstructed and validated. See Supplementary Table 2 for structure abbreviations.

Fig. 9. ATN connections with hypothalamus, midbrain, and hindbrain.

a Projections from mammillary bodies to ATN. a1 A retrograde CTB injection in IAD labels neurons in MM median (MMme; MMme→IAD) (top). A PHAL anterograde tracer injection in the MMme (middle) labels the IAD (bottom) validating the connection. a2 CTB injection in AD labels LM neurons (MMl neurons labeled from injection leakage into AV). The LM→AD is validated with a PHAL LM injection that labels the AD. a3 A CTB AV injection labels MMl neurons. This MMl→AV connection is validated by an MMl AAV injection. a4 A CTB AM injection labels MMm neurons (MMm→AM), which is validated by an AAV MMm injection. b Schematic summarizing ATN, MM, and LM connectivity. c PHAL and AAV tracers in MMm and LM in the same brain show the bilateral LM→AD projections compared to the unilateral MMm→AM projection. d MMme PHAL injection shows bilateral MMme→IAD projections. e Cre-dependent TVA receptor mediated rabies tracing of AMv and AMd.medial neurons retrogradely label MMm validating MMm→AMd/AMv connections. Bar graph shows the quantification of retrogradely labeled cells in the MM versus LM following the injections. f AAV injection in the DTN (top), confirmed by strong projections to LM (middle), labels terminals in AD and AV (bottom). g AAV in the VTN (top), confirmed by the strong projections to MM (middle), labels terminals in AD and AV (bottom). h Validation of the DTN/VTN→AD/AV connections. CTB injection in AM/AD labels cells in VTN, while a FG AM/AV injection, confirmed by MM labeled cells, shows labels in DTN and LDT. i Double retrograde injections in AV (CTB, pink) and MM (CTB, green) show labeled cells in VTN (VTN→AV/MM) and in SUBv (SUBv→AV/MM). Note the labeled cells in MMl, which confirm AV injection location. Also, note the laminar specific arrangement of SUBv→AV and SUBv→MM projecting neurons. j Schematic of connections among ATN, DTN, and VTN. k Schematic showing generally segregated networks of head directionality/visual-spatial processing (peach) involving AD←LM↔DTN and theta rhythmicity (blue) involving AV←MM↔VTN. Direct connections shown in (f, g) suggest a more interconnected network rather than segregated circuits. See Supplementary Table 2 for structure abbreviations.

Fig. 10. Dendritic morphology analysis of ATN neurons.

Representative images of (a) AM, (b) AD, and (c) AV neurons and their corresponding representative digital reconstructions. d MORF expressing ATN neurons. e Representative dendritic arbors from the ATN, where edge (left panels) and non-edge (right panels) neurons are displayed for (f) AM, (g) AD, and (h) AV. i Location and orientation of AMd edge (green) and non-edge neurons (teal). j Graphical definition of a compartment’s angular orientation relative to the local AMd center vector: dashed straight lines that connect a compartment’s origin point to AMd center. 1 radian = 57.3°. k Compartment deviation angle from the AMd center as a function of the somatic distance from AMd center. The greater the distance of a neuron’s soma from the AMd center, the lesser its average compartment deviation (R = −0.63, p = 6.1821e⁻¹⁰). l Edge neurons (n = 27) show reduced angular deviation from the AMd center compared to non-edge neurons (n = 45) (two-sided t-test with FDR correction; t(70) = −5.4403, p = 7.378e⁻⁰⁷). Edge neurons have a higher proportion of dendritic compartments oriented toward AMd center. Red colored dendrites have an acute angular deviation and are oriented towards AMd center as opposed to the blue dendrites that have an obtuse angular deviation and are oriented away from AMd center (right panel). The x-axis increases from medial to lateral, while y-axis increases from dorsal to ventral (in microns). m Edge neurons have a higher proportion of total dendrites closer to the center than their individual cell bodies [t(70) = 4.353, p = 4.4886e⁻⁰⁵]. Pink designates closer dendrites, while blue dendrites are further away (right panel). Traced AM, AD, and AV neurons were collected across independent cases and aggregated for analysis. The line inside of each boxplot is the sample median. Top and bottom edges are upper and lower quartiles, respectively. The distance between top and bottom edges is the interquartile range (IQR). Upper quartile corresponds to the 0.75 quantile and lower quartile corresponds to the 0.25 quantile. Outliers (“o”) are values that are more than 1.5 IQR away from the top or bottom of the box. Top whisker connects the upper quartile to the nonoutlier maximum, and the other connects the lower quartile to the nonoutlier minimum. ***p ≤ 0.001.

Fig. 11. Functional ATN networks.

a Depiction of the classic Papez circuit. b Network of structures involved in drug seeking behavior centered around the AMd.medial domains. Projections from the AM→BLAa are to a specific domain, BLA.ac, which is also connected with regions shown to be involved in drug seeking behavior (see “Discussion”). c Functional neural network of the ATN. An extended ATN wiring diagram showing how each of the identified ATN domains connects with different areas within the medial cortical subnetwork, which controls movements and the orientation of the eyes, head, and neck for attention, navigation, and exploratory behavior via the cortico-basal ganglia (SNr serves as the output portal) or through its cortico-tectal projections (SC serves as the output portal). Meanwhile, several ATN subnuclei (especially AMv, AMd.m, AMd.l, AV.l, AV.m, and AV.mt) connect with the MPF (ILA, PL, DP), BLAa, and SUBv. These three cortical areas are highly interconnected and generate massive descending projections to the medial amygdalar nucleus (MEA), central amygdalar nucleus (CEA), anterior (BSTa) and posterior (BSTp) bed nuclei of the stria terminalis, hypothalamus, and PAG to regulate neuroendocrine, autonomic, and goal-directed behavior associated with emotion (e.g., hunting or attacking). Note that the dorsal premammillary nucleus (PMd) generates dense projections back to the AMv. Finally, these three cortical areas also generate dense projections to the ACB, which, together with the VTA, are essential for reward and addiction. The structures within the dashed box are included in the classic Papez circuit. As illustrated here, the ATN serves as a critical network hub, bridging communication between the medial cortical subnetwork and the emotional network to control goal-directed behavior. See Supplementary Table 2 for structure abbreviations.