Neuroanatomical Substrates of Circuit-Specific Cholinergic Modulation across the Primate Anterior Cingulate Cortex

Abstract

Acetylcholine is a robust neuromodulator of the limbic system and a critical regulator of arousal and emotions. The anterior cingulate cortex (ACC) and the amygdala (AMY) are key limbic structures that are both densely innervated by cholinergic afferents and interact with each other for emotional regulation. The ACC is composed of functionally distinct dorsal (A24), rostral (A32), and ventral (A25) areas that differ in their connections with the AMY. The structural substrates of cholinergic modulation of distinct ACC microcircuits and outputs to AMY are thought to depend on the laminar and subcellular localization of cholinergic receptors. The present study examines the distribution of muscarinic acetylcholine receptors, m1 and m2, on distinct excitatory and inhibitory neurons and on AMY-targeting projection neurons within ACC areas, via immunohistochemistry and injections of neural tracers into the basolateral AMY in adult rhesus monkeys of both sexes. We found that laminar densities of m1+ and m2+ expressing excitatory and inhibitory neurons depended on area and cell type. Among the ACC areas, ventral subgenual ACC A25 exhibited greater m2+ localization on presynaptic inhibitory axon terminals and greater density of m1+ and m2+ expressing AMY-targeting (tracer+) pyramidal neurons. These patterns suggest robust cholinergic disinhibition and potentiation of amygdalar outputs from the limbic ventral ACC, which may be linked to the hyperexcitability of this subgenual ACC area in depression. These findings reveal the anatomical substrate of diverse cholinergic modulation of specific ACC microcircuits and amygdalar outputs that mediate cognitive–emotional integration and dysfunctions underlying stress and affective disorders.

Significance Statement

Cholinergic neuromodulation of anterior cingulate (ACC) corticolimbic networks are consequential for cognitive–emotional integration. This study explored muscarinic acetylcholine receptor (m1 and m2) expression on distinct excitatory and inhibitory neurons across functionally distinct ACC subregions and their outputs to the amygdala, a key limbic structure for emotional regulation. Our findings reveal that muscarinic receptor expression is most robust in ventral subgenual ACC area 25, in patterns that suggest cholinergic disinhibition and potentiation of area 25 outputs to the amygdala. These neuromodulatory interactions may be linked to the hyperexcitability of subgenual ACC found in depression. Our findings yield novel insight on how cholinergic modulation of diverse ACC circuits can contribute to cognitive–emotional processing and dysfunction in stress and affective disorders.

Introduction

The ability to regulate when and how emotions are used to guide action relies on adaptive circuits that integrate internal limbic–visceral and external sensory–motor information for a given behavioral context. The anterior cingulate cortex (ACC) is a key area for this emotional regulation and integration, acting as a unique connectional hub for the otherwise segregated cognitive, motor, and limbic networks. The functionally distinct dorsal “motor” (A24), rostral “cognitive” (A32), and ventral “limbic” (A25) subdivisions of the ACC each have specialized roles in integrating information from higher order frontal regions and limbic structures, such as the amygdala (AMY) (reviews: Paus, 2001; Phelps and LeDoux, 2005; Etkin et al., 2011). Acetylcholine (ACh) is a robust neuromodulator of corticolimbic interactions, playing a key role in attention, arousal, learning and memory, and stress (reviews: Hasselmo, 2006; Hasselmo and Sarter, 2011). Ascending cholinergic pathways densely innervate the ACC (Mesulam et al., 1984; Ghashghaei and Barbas, 2001), yet circuit specificity of cholinergic modulation depends on the largely unknown laminar and subcellular localization of cholinergic receptors across distinct ACC areas. In primates and rodents, cortical ACh acts mainly through binding to muscarinic and, to a lesser extent nicotinic, cholinergic receptors located on pyramidal and nonpyramidal cell types (Mrzljak et al., 1995; reviews: Sarter et al., 2009; Colangelo et al., 2019). Specifically, ACh predominantly acts via postsynaptic m1 receptors to promote cortical excitation or via presynaptic m2 receptors to suppress neurotransmitter release (McCormick and Prince, 1985; Carr and Surmeier, 2007; review, Hasselmo, 2006). However, the anatomical substrates of cholinergic modulation of microcircuits and corticolimbic pathways within ACC subregions are poorly understood in primates.

The relationship of cortical communication between areas is driven by the coordinated activity of pyramidal neurons in distinct layers (Hilgetag et al., 2016). Previous work found distinct distributions of corticocortical and subcortical pathways within ACC layers and subregions (Ghashghaei et al., 2007; Calderazzo et al., 2021). These laminar patterns of connectivity are highly relevant to cortical functions, since layers have distinct neurochemical microenvironments and participate in different circuits that dictate information flow (review, Barbas, 2015). ACh can modulate distinct cortical pathways depending on layer-specific cholinergic innervation and cholinergic receptor expression (Mesulam et al., 1984; Ghashghaei and Barbas, 2001; Disney et al., 2007; Medalla and Barbas, 2012; Coppola et al., 2016; Zilles and Palomero-Gallagher, 2017). Previous quantitative mapping studies showed laminar patterns of AMY-targeting neurons differed across the ACC (Ghashghaei et al., 2007; Calderazzo et al., 2021). While A24 and A25 had equal distributions of AMY-targeting neurons in superficial (L2–3) and deep (L5–6) layers, those in A32 predominated in deep layers. Our previous work compared m1+ and m2+ receptor expression within the upper layers of ACC area 24 and lateral prefrontal cortical area 46 (Tsolias and Medalla, 2022). What remains unknown is how ACh modulates layer-specific ACC outputs to the amygdala that are dependent on mAChR expression of AMY-targeting pyramidal neurons across distinct ACC areas.

Neurochemically diverse inhibitory neurons in the cortex identified by their expression of calcium-binding proteins, parvalbumin (PV+), calbindin (CB+), or calretinin (CR+), also have distinct laminar distributions (review, DeFelipe, 1997). In monkey prefrontal cortices, PV+ neurons are enriched in layers 3–5, while CB+ and CR+ neurons are enriched in upper layers 2–3a (Meskenaite, 1997; Dombrowski et al., 2001). These interneuron classes target distinct somatodendritic compartments of pyramidal cells and differ in their mode of inhibition. PV+ neurons elicit strong inhibition, mainly targeting pyramidal neuron somata (review, Freund and Katona, 2007). CB+ neurons target distal dendrites and spines of pyramidal neurons, while CR+ neurons mainly inhibit other GABAergic neurons (del Rio and DeFelipe, 1997; Meskenaite, 1997; Melchitzky et al., 2005; review, DeFelipe, 1997).

The current study aims to use immunohistochemistry and neural tract-tracing of AMY outputs to characterize the neuroanatomical substrate of muscarinic modulation of excitatory:inhibitory balance and corticolimbic laminar pathways within ACC. Our data reveal ACC region- and layer-specific distribution and subcellular localization of mAChRs that may confer pathway-specific cholinergic modulation underlying cognitive–emotional integration.

Materials and Methods

Experimental subjects

The brain tissue used in this study was harvested from a total of 12 adult rhesus monkeys (Macaca mulatta) of both sexes that were subjects in larger studies on brain aging and cognition. Among the brains, six were previously used in Tsolias and Medalla (2022) and included 9 ± 1.13 years (two female and four males; Table 1) while the remaining six cases (6.4 ± 0.477; two females and four males; Table 1) received a combination of dextran amine bidirectional tracers fluororuby (FR) or fluoroemerald (FE) into the AMY complex in combination with FE, FR, or Cascade Blue (CBL) into either the dorsal premotor cortex (dPMC), entorhinal cortex (EC) lateral prefrontal cortex (LPFC, A9), or ventral ACC (A25) for experiments that were not a part of this study (Table 1). All monkeys were obtained from either National Primate Centers or private vendors and were individually housed in the Animal Science Center (ASC) at Boston University Chobanian and Avedisian School of Medicine. The ASC is fully accredited by the Association for Assessment and Accreditation of Laboratory Animal Care, with animal research conducted in strict accordance with guidelines of the National Institutes of Health's Guide for the Care and Use of Laboratory Animals and the U.S. Public Health Service Policy on Humane Care. All animal use in this study was fully approved by the Boston University Institutional Animal Care and Use Committee.

Table 1.

Subjects used for tracer and immunolabeling experiments

Case	Sex, age	Tracer type analyzed	AMY nucleus	Other injections tracer type, injection
PIK	M, 7	FR	Basolateral	FE, dPMC
PIL	M, 7	FR	Lateral	FE, dPMC
PIM	F, 6	FR	Basomedial	FE, dPMC; CBL, EC
PIN	F, 6	FR	Cortical	CBL, EC
PIO	M, 6.3	FR	Basolateral	CBL, A9; FE, dPMC
PIP	M, 6.1	FE	Cortical	FR, A9; CBL, A25
AM289	M, 9	No tracer injections and used only for immunohistochemical experiments
AM292	M, 7.6
AM294	F, 9.6
AM295	M, 10.9
AM296	M, 8.8
AM299	F, 8.6

Description of age, sex, and details of bidirectional tracer injections (1:1 cocktail of 10 and 3 kDa isotopes) into the basolateral amygdala complex and other injections not examined in this study. FR, fluororuby; FE, fluoroemerald; CBL, cascade blue; AMY, amygdala; dPMC, dorsal pre-motor cortex; EC, entorhinal cortex.

Surgical procedures and injection of neural tracers

Neural tracers were injected based on stereotaxic coordinates calculated from magnetic resonance imaging (MRI) scans taken prior to surgery using the midline and betadine filled ear bar tips as reference points. All MRI and surgical procedures were performed as previously described in Medalla and Barbas (2009) and Calderazzo et al. (2021). For the survival surgery, monkeys were deeply anesthetized with isoflurane (1–3%, to effect) and positioned in the same MRI compatible stereotaxic apparatus (Kopf 1530, David Kopf Instruments), burr holes were made, and tracers were injected into the regions of interest (Table 1). In each hemisphere, the amygdala, as well as other regions of interest not relevant to this study, was injected with either fluoroemerald (FE, dextran fluorescein catalog #D1820 and #D3305), fluororuby (FR, dextran tetra-methylrhodamine, catalog #D1817 and #D3308), or Cascade Blue (CBL, catalog #D7132 and #D1976, Thermo Fisher Scientific) dextran amines using a microsyringe (5 or 10 μl; Hamilton) mounted on a microdrive. Both 3 and 10 kDA isoforms of dextran amines were used to ensure bidirectional retrograde and anterograde transport (Veenman et al., 1992; Reiner et al., 2000). The dyes were diluted to 10 mg/ml in sterile distilled water and delivered in 2–3 µl volumes, and for each case, a single injection was made into the basolateral nuclei complex of the amygdala and left in situ for 10 min to allow for diffusion and to prevent the upward suction during needle retraction.

Perfusion and preparation of tissue

Tissue was harvested using our well–established two-stage perfusion protocol allowing for the harvest of both live tissue and fixed tissue (Amatrudo et al., 2012) for parallel experiments not in the present study. After sedation with ketamine hydrochloride (10 mg/kg), the monkeys were deeply anesthetized with sodium pentobarbital (to effect, 15 mg/kg, i.v.) and then perfused through the ascending aorta with ice-cold Krebs–Henseleit buffer containing (in mM): 6.4 Na₂HPO₄, 1.4 Na₂PO₄, 137 NaCl, 2.7 KCl, 5 glucose, 0.3 CaCl₂, and 1 MgCl₂, pH 7.4 (Sigma-Aldrich). Following the harvest of a fresh tissue block (∼10 mm³) for other parallel studies, the perfusate was switched to 4% paraformaldehyde in 0.1 M phosphate buffer (PB), pH 7.4 (at 37°C) to fix the remaining whole brain. The fixed brain sample was blocked, in situ, in the coronal plane, removed from the skull, cryoprotected in a series of glycerol solutions, and flash frozen in 70°C isopentane (Rosene et al., 1986). The brain was cut on a freezing microtome in the coronal plane at 30 or 60 µm and stored in cryoprotectant (15% glycerol, in 0.1 M PB, pH 7.4) at −80°C (Estrada et al., 2017).

Immunohistochemistry

We performed immunohistochemical (IHC) experiments on serial coronal sections from two sets of tissue: (1) archived tissue from monkeys with no tracer injections that are part of parallel studies on normal aging and (2) tissue from tracer injected monkeys (n = 6; Table 1).

In the first set of IHC experiments in monkeys with no tracer injections, IHC was performed to visualize the distribution and the extent of colocalization of m1+ and m2+ receptors with distinct interneuron subtypes (PV+, CB+, CR+), pyramidal neurons (MAP2+), and inhibitory vesicular transporter protein (VGAT+), as previously described (Tsolias and Medalla, 2022). Briefly, sections were incubated in 0.05 M glycine in 0.01 M phosphate-buffered saline (PBS), pH 7.4, to reduce excessive cross-linking of lipids and proteins during the tissue fixation. Sections were rinsed (0.01 M PBS 2 × 10 min) and antigen retrieval was performed using 10 mM sodium citrate, pH 8.5, in a 60–70°C water bath for 20 min to unmask antigen-binding sites. After that, sections were rinsed in 0.01 M PBS (3 × 10 min, 4°C) and incubated in preblock [0.01 M PBS, 5% bovine serum albumin (BSA), 5% normal donkey serum (NDS), 0.2% Triton X-100] to reduce any nonspecific binding of secondary antibodies. Sections were incubated at 4°C for 48 h in a combination of primary antibodies to label distinct cell types with muscarinic receptors [Table 2; diluted in 0.1 M PB, 0.2% acetylated BSA (BSA-c, Aurion), 1% NDS, 0.1% Triton X-100]. We batched process multiple sections from every case for eight different sets of IHC with distinct marker combinations as summarized in Table 3. To increase the penetration of the antibody, two incubation sessions in low-wattage microwave (2 × 10 min at 150 W) using the Pelco BioWave Pro (Ted Pella) were followed by a 48 h incubation at 4°C with gentle agitation. After rinsing (3 × 10 min) in 0.01 M PBS at 4°C, sections were incubated overnight in secondary antibodies diluted in incubation buffer (Table 3), microwaved 2 × 10 min at 150 W (Ted Pella Pelco BioWave), and placed at 4°C for 24 h with gentle agitation. In some immunolabeling batches, biotinylated secondary antibody and Streptavidin 546 conjugates were used to further amplify m2 labeling.

Table 2.

Antibodies and probes utilized in immunohistochemical experiments

Antibody	Host	Dilution	Vendor, catalog #	RRID
Primary antibodies
Anti-Calbindin D-28k (CB)	Rabbit	1:2,000	Swant, CB38	AB_10000340
Anti-Calretinin (CR)	Rabbit	1:2,000	Swant, 7697	AB_2721226
Anti-Parvalbumin (PV)	Guinea Pig	1:2,000	Swant, GP72	AB_2665495
Anti-Microtubule associated protein-2 (MAP2)	Chicken	1:1,000	Abcam, ab5392	AB_2138153
Anti-Muscarinic Receptor 1 (m1)	Goat	1:500	Abcam, ab77098	AB_1523990
Anti-Muscarinic Receptor 2 (m2)	Rat	1:500	Millipore, MAB367	AB_94952
Anti-Vesicular GABAergic Transporter (VGAT)	Guinea Pig	1:400	Synaptic Systems, 131004	AB_887873
Anti-fluorescein/fluoroemerald (FE)	Mouse	1:800	Jackson, 200-002-037	AB_2313645
Antitetramethylrhodamine/fluororuby (FR)	Rabbit	1:800	Invitrogen, A6397	AB_2536196
Secondary antibodies
Alexa 405 anti-mouse IgG	Donkey	1:200	Thermo Fisher Scientific, A10036	AB_2534012
Alexa 488 anti- Guinea Pig IgG	Donkey	1:200	Jackson, 706-545-148	AB_2340472
Alexa 488 anti-chicken IgG	Donkey	1:200	Jackson, 703-545-155	AB_2340375
Alexa 546 anti- Rabbit IgG	Donkey	1:200	Thermo Fisher Scientific, A10040	AB_2534016
Alexa 647 Donkey anti- Guinea Pig IgG	Donkey	1:200	Jackson, 706-605-148	AB_2340476
Alexa 633 Donkey anti-goat IgG	Donkey	1:200	Thermo Fisher Scientific, A21082	AB_10562400
Alexa 647 Donkey anti-rat IgG	Donkey	1:200	Jackson, 712-606-150	AB_2340695
Biotin-SP-AffiniPure Fab Fragment Donkey Anti-Rat IgG (H + L)	Donkey	1:200	Jackson, 712-067-003	AB_2340650
Streptavidin-546	N/A	1:200	Thermo Fisher Scientific, S11225	AB_2532130
Streptavidin-750	N/A	1:200	Thermo Fisher Scientific, S21384	AB_2336066

Table 3.

Summary of marker combinations for multiple IHC batch processing

Markers and probe combinations for IHC batches	Quantified measures	# sections per area
Batch 1: MAP2 (AF488), VGAT (AF647), m1 (AF405), m2 (biotinylated IgG-Streptavidin AF546)	MAP2 m1 m2 cells VGAT m2 puncta	A24: 1; A32: 1; A25: 1
Batch 2: CB (AF405), SMI-32 (AF488)a, VGAT (AF546), m1 (AF405);	CB m1 cells	A24: 1; A32: 1; A25: 1
Batch 3: CB (AF405), SMI-32 (AF488)a, VGAT (AF546), m2 (AF647)	CB m2 cells	A24: 1; A32: 1; A25: 1
Batch 4 PV (AF488), CR (AF546), m1 (AF405)	PV m1 cells CR m1 cells	A24: 1; A32: 1; A25: 1
Batch 5 PV (AF488), CR (AF546), m2 (AF647);	PV m2 cells CR m2 cells	A24: 1; A32: 1; A25: 1
Batch 6: FR (AF546) FE (AF488) m1 (AF405) m2 (AF647)	FR/FE m1 m2 cells	A24: 2; A32: 2; A25: 2
Batch 7: 5-channel CR (AF405), PV (AF488), CB (AF546), m1 (AF647), m2 (biotinylated IgG-Streptavidin AF750)	CR m1 m2 cells CB m1 m2 cells PV m1 m2 cells	A24: 2; A32: 2; A25: 2
Batch 8: 5-channel VGAT (AF405), MAP2 (AF488), GABRA5 (AF546)a, m1 (AF647), m2 (biotinylated IgG-Streptavidin AF750)	MAP2 m1 m2 cells	A24: 2; A32: 2; A25: 2

^aMarkers used for a different study.

In a second set of IHC experiments in tracer monkeys, we assessed expression of m1/m2 cholinergic receptors on tracer-labeled neurons using matched free-floating coronal tissue sections, two sections through A32 and two sections through A24 and A25, for each case. The sections were first incubated in primary antibodies for the FE/FR tracer and corresponding secondary antibodies conjugated to the probe of interest (Table 2). After processing for neural tracers, sections were preblocked [0.01 M PBS, 5% BSA, 5% NDS, 0.2% Triton X-100] to reduce any nonspecific binding of secondary antibodies. Sections were then incubated at 4°C for 48 h in a combination of primary antibodies for m1 and m2, as above [Table 2; diluted in 0.1 M PB, 0.2% acetylated BSA (BSA-c, Aurion), 1% NDS, 0.1% Triton X-100] followed by a secondary fluorescence-conjugated antibody (Table 2). In all experiments, sections had a final rinse (3 × 10 min) in 0.1 M PB and were then mounted onto slides, coverslipped with prolong anti-fade gold mounting medium (Thermo Fisher Scientific) and cured at room temperature in the dark.

Inhibitory and excitatory cell density estimates

As described in our previous work (Tsolias and Medalla, 2022), we quantified the density of immunolabeled somata of m1 and m2 receptor labeled cells and their expression (dual labeling) on either immunolabeled inhibitory (PV, CB, CR) or excitatory (MAP2) neurons or tracer-labeled AMY-targeting neurons using volumetric 3d counting procedures. Immunolabeled sections were imaged at high resolution using laser scanning confocal microscopes (Leica SPE, Zeiss LSM 710, or Zeiss 710 NLO). Image stacks were acquired using a plan apochromat 40×/1.3 NA oil immersion objective at a voxel resolution of 0.268 × 0.268 × 0.5 µm (Leica TCS SPE) or 0.208 × 0.208 × 0.5 µm (Zeiss LSM 710 or Olympus FV3000) voxel size. Using architectonic maps of frontal cortices (Barbas and Pandya, 1989), we identified ACC areas A24, A25, and A32 in three serial sections ∼2,400 µm apart (Table 3). For each area, we identified 1–2 columns per section and systematically imaged laminar ROIs along these columns as follows: L2, L3, L5, and L6. We focused on imaging these layers where somata of pyramidal neurons mainly reside, since layer 1 has very sparse cell bodies, and layer 4 is incipient or lacking in some of the ACC areas (Barbas and Pandya, 1989). The resulting images were deconvolved to improve the signal-to-noise ratio and converted to 8 bit image files for analysis using AutoQuant (Media Cybernetics). For each deconvolved image stack, we performed manual volumetric 3d cell counting, with exclusion and inclusion borders, through an optical stack (Fiala and Harris, 2001), using FIJI software and the “cell counter” plug-in to mark single-labeled somata expressing particular excitatory (MAP2+ pyramidal) and inhibitory (CB+ PV+ CR+) cellular markers, and the subset that were also dual-labeled with mAChRs. Cells dual labeled for m1 or m2 were defined as having cytoplasmic label above a set threshold (signal greater than 3× the background) that outlines the profile of the soma.

Quantification of MAP2+ pyramidal neurons

MAP2+ pyramidal neurons were identified based on the appearance of pyramidal shaped soma and the presence of MAP2+ apical dendrite (Caceres et al., 1984; Peters and Sethares, 1991). A minority of small nonpyramidal neurons lightly labeled with MAP2 (<1% of total MAP2+ cells) were counted separately and excluded from the analyses, as this subset may include inhibitory neurons (de Lima et al., 1997).

Quantification of inhibitory neuron subpopulations

Inhibitory neurons were identified by their nonpyramidal somata and the strong expression of one of the three calcium-binding proteins, CB+, CR+, and PV+. Consistent with our previous work (Tsolias and Medalla, 2022), pyramidal neurons that weakly express CB were noted in all areas and were not analyzed for the current study. Inhibitory neurons were first manually counted in one section per area and case using FIJI and the threshold for classifying cells colabeled with m1 or 2 was established. Then image stacks from the remaining sections were counted using semiautomated classifier tools in QuPath (Bankhead et al., 2017) as follows: (1) For each field, the maximum z-projection of image stacks (25 optical slices) were generated and imported into QuPath. (2) Somata of CB+, PV+, and CR+ were segmented automatically as individual objects. (3) The segmented objects were then classified as m1 and m2 positive cells using the classifier tools in QuPath. The m1 and m2 classifiers were based on a user-defined intensity threshold established with manual counting and was kept consistent within each case. For each marker, the cell density of each subpopulation per area × layer × case was calculated by dividing the total number of cells counted by the total volume of the total image stacks counted, multiplied by a z-shrinkage factor (∼1.5× factor = cut section thickness 60 µm/mounted section thickness 40 µm).

Quantification of Tracer-labeled neurons

Neurons that were tracer labeled were manually and exhaustively counted within either upper L2–L3 and deep L5–L6 of each area and section. We calculated the absolute density of tracer-labeled neurons in each ACC subdivision by dividing the estimated number of neurons by the estimated volume (mm³) of each region of interest (ROI).

Quantification and colocalization of mAChR puncta

Colocalization of mAChR on tracer-labeled pyramidal neuron somatic and dendritic compartments

We used the ROI manager and the colocalization plug-in in FIJI to quantify the colocalization of m1+, m2+, or m1+m2+ receptors on the somatic or proximal apical dendrite of retrograde tracer-labeled, AMY-projecting neurons in A24, A25, and A32 of the ACC. We utilized the segmentation editor to isolate the two compartments (soma, proximal apical dendrite) of neurons located in the superficial (L2–L3) or deep (L5–L6) layers (A24: upper layer n = 15 soma/dendrite, deep layer n = 11 soma/dendrite; A25: upper layer n = 14 soma/dendrite, deep layer n = 14 soma/dendrite; A32: upper layer n = 9 soma, n = 8 dendrite, deep layer n = 9 soma, n = 10 dendrite). The density of mAChR+ within these ROIs was determined by particle analysis to calculate the estimated percent area of label (Schneider et al., 2012).

Colocalization of presynaptic m2+ puncta in the neuropil with VGAT+ terminals

In our previous work, we examined the total optical density of the presynaptic m2+ with vesicular GABAergic transporter (VGAT+) in the neuropil of in the superficial layers (L1–L3) of A24 (Tsolias and Medalla, 2021). Here, we aimed to further investigate the expression within the superficial (L1–3) and deep layers (L5–L6) across all three subdivisions of the ACC using the particle analysis function in FIJI/ImageJ (https://imagej.net/Fiji; 1997–2016; RRID:SCR_002285; Schindelin et al., 2012). We used the same previously established signal thresholds using either the Ostu or Renyi methods for all images per case and calculated the average measure of m2+, VGAT+ (% area labeled; Schneider et al., 2012). To calculate the percent of colocalization, we then utilized the FIJI colocalization plug-in and calculated the Mander's colocalization coefficient, the ratio of percent area colocalized with either channel 1 or channel 2 (Stauffer et al., 2018).

Statistical analyses

All statistical analyses were performed in MATLAB (MathWorks). For laminar and regional comparisons of cell and puncta density and proportion measures, we took the average of all sampling fields in each layer and area, per case (n = 6 cases; Table 1). For receptor density quantification in somatic versus dendritic compartments, measurements from individual superficial and deep layer neurons from each area were taken as replicates (n = 9–15 neurons per group, from 6 cases). An outlier analysis and tests for normality (Z-score calculation and Kolmogorov–Smirnov test) were performed for each outcome measure and group. Between-area (A24, A25, A32) and within-area (between-layer) comparisons of outcomes were conducted using a two-way ANOVA (for layer × area and mAChR × layer within area) with a Fisher's least significant difference (LSD) post hoc analysis. Comparisons between somatodendritic compartments of tracer-labeled neurons by area were performed using a multiple-comparisons one-way ANOVA with a Fisher's LSD post hoc analysis. Between and within groups statistical comparisons of frequency distributions were performed using a Fisher's exact test.

Results

We examined the distribution of m1 or m2 expression in upper (L2, L3) and deep (L5, L6) layers across the subdivisions of the ACC. Two-way ANOVA (layer × area) comparisons revealed significant within- and between-area differences among m1+, m2+, and coexpressing m1+m2+ neurons for distinct excitatory or inhibitory neuronal subtypes and circuits.

Laminar density and proportion of m1+ and m2+ expressing MAP2+ pyramidal neurons within distinct ACC areas

Pyramidal neuron cell bodies within upper (L2, L3) and deep (L5, L6) layers of ACC areas were identified based on morphology and labeling of a cytoskeletal protein, MAP2, which is strongly expressed in cortical pyramidal neuron dendrites (Caceres et al., 1984; Peters and Sethares, 1991; Fig. 1). Identified MAP2+ pyramidal neurons were further characterized by their coexpression of m1+ (m1+m2−), m2+ (m1−m2+), or m1+m2+ (double positive) receptors (Fig. 1A). Within-area comparisons showed the density of total MAP2+ pyramidal neurons were equivalent across upper versus deep layers in A24, A25, and A32 (Fig. 1B). Between-area comparisons showed significant regional differences in the density of MAP2+ pyramidal neurons in the deep and upper layers. Specifically, the density of total MAP2+ neurons in the upper layers of A24 was significantly lower than A25 (p < 0.01 for L2 and L3; two-way ANOVA layer × area; Fisher's LSD post hoc) and both upper layers and deep layer L5 of A32 (p < 0.01 for L2, L3, and L5; Fig. 1B, Extended Data Table 1-1).

Figure 1.

Regional and laminar distribution of m1+ and m2+ expressing MAP2+ pyramidal neurons in the ACC. A, Representative confocal multichannel images at a single optical slice showing colabeling of MAP2+ pyramidal neurons (green) with m1 (blue) and m2 (magenta) in L3 and L5 of ACC areas A24, A25, and A32. Scale bar, 10 µm. B–F, Box and whisker plots of cell densities of MAP2+ pyramidal neuron subpopulations within distinct layers (L2, L3, L4, L5) of ACC A24 (pink, top), A25 (purple, middle), A32 (green, bottom): total MAP2+ (B), MAP2+ with m1+ (C), MAP2+ with m2+ (D), MAP2+ with m1+ and m2+ (E), or MAP2+ only, m1− and m2− (F). G, Medial surface view of the rhesus monkey brain showing ACC regions of interest: areas 24, 32, and 25. Cg, cingulate sulcus. H–J, Pie charts showing the relative proportion of MAP2+ pyramidal neurons coexpressing either (m1+ only, m2+ only), both (m1+ and m2+), or neither (single MAP2+ only/m1− and m2−) mAChRs in each layer of A24 (H), A25 (I), and A32 (J). Data from n = 6 monkeys; box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, **p < 0.01, ***p < 0.001. See Extended Data Table 1-1 for details on the statistical analyses.

Table 1-1

Two-Way ANOVA (area x layer) p-values for within and between area comparisons of MAP2+ pyramidal neuron distributions. Download Table 1-1, DOCX file ^{(27.7KB, docx)} .

Among the MAP2+ m1+ or m2+ expressing subpopulations, within-area analyses showed that the ACC subdivisions had equivalent densities of MAP2+m1+ pyramidal neurons in both upper and deep layers (Fig. 1C). For MAP2+m2+ subpopulation of pyramidal neurons, we found within-area differences only in A25 where L6 had a significantly greater density than L2 and L5 (p < 0.05 for all comparisons; Fig. 1D, Extended Data Table 1-1). Interestingly, for the MAP2+m1+ m2+ pyramidal neuron subpopulation, A24 had a significantly greater density in deep layer L5 compared with upper layers L2, L3 (p < 0.05 for all comparisons; Fig. 1E, Extended Data Table 1-1), while laminar differences were not found within both A25 and A32. The single MAP2+ (m1−m2−) neuron subpopulation also showed laminar density differences in A24, which had significantly higher densities in the upper (L2 and L3) compared with the deep layers (L5 or L6; p < 0.01 for all comparisons; Fig. 1F, Extended Data Table 1-1). In contrast, A25 and A32 had equivalent densities of single MAP2+ (m1−m2−; Fig. 1F, Extended Data Table 1-1).

Between-area comparisons revealed that the density of MAP2+m1+ pyramidal neurons was equivalent between the ACC subdivisions. The density of MAP2+m2+ neuronal subpopulation was significantly greater in A25 compared with A24 in L3 and L6 and compared with A32 in L6 (p < 0.05 for all comparisons; Fig. 1D, Extended Data Table 1-1). Both A25 and A32 had significantly greater densities of the MAP2+m1+m2+ pyramidal neurons in the upper layers compared with A24 (p < 0.01 for all comparisons; Extended Data Table 1-1), while in the deep layers only A25 L6 had a significantly greater density compared A24 (p < 0.05; Fig. 1E, Extended Data Table 1-1). For single MAP2+ (m2−m1−) neurons, there were equivalent densities across ACC areas within the upper layers and significant differences only between A24 and A25 in L5, where A25 had a significantly greater density than A24 (p < 0.01 for all comparisons; Fig. 1F, Extended Data Table 1-1).

Comparisons of the relative proportions of MAP2+ pyramidal neuron subpopulations revealed regional and laminar patterns consistent with density distributions (Fig. 1H–J). We observed that in all areas and layers, the majority of MAP2+ pyramidal neurons coexpressed m1+ and m2+ receptors (55–83%), while 11–26% expressed m1+ alone and 1–6% express m2+ alone (Fig. 1H–J, open fill). Within A24, we observed a significantly greater proportion of MAP2+m1+m2+ pyramidal neurons in the deep layers (82–83%) compared with the upper layers (55–59%; p < 0.001 for all comparisons; Fig. 1H, Extended Data Table 1-1). Conversely, A24 had a significantly greater proportion of MAP2+ only (m1−m2−) pyramidal neurons in the upper layers compared with the deep layers (p < 0.001 for all comparisons) and equivalent proportions of both MAP2+m1+ and MAP2+m2+ neurons across layers (Fig. 1H, Extended Data Table 1-1). In contrast, A32 and A25 exhibited largely similar proportions of MAP2 m1+/m2+ subpopulations between upper and deep layers (Fig. 1I,J, Extended Data Table 1-1). In summary, laminar differences in distribution patterns of subpopulations of mAChR m1+ and m2+ expressing MAP2+ pyramidal neurons were most pronounced in ACC A24 compared with A25 and A32, which exhibited similar laminar distributions.

Differential distribution of m1+ and m2+ amygdala targeting pyramidal neurons across distinct ACC areas

We then examined whether subpopulations of target-specific pyramidal neurons in these distinct ACC areas show differences in m1+/m2+ expression. Given that ACh is a potent modulator of corticolimbic and prefrontal-medial temporal lobe pathways (Mesulam et al., 1984; Scanziani et al., 1995; Ghashghaei and Barbas, 2001; Kremin and Hasselmo, 2007; Fernandez de Sevilla et al., 2021; reviews, Sarter and Bruno, 1997; Picciotto et al., 2012), we focused on the robust output neurons from the ACC to the AMY. Previous work from our group showed that all ACC areas 24, 25, and 32 provided widespread projections to the AMY, with regional differences in laminar density (Calderazzo et al., 2021). Our goal here was to examine whether there are differences in the localization and coexpression of m1+ and m2+ receptors on subpopulations of AMY-projecting neurons in either the upper layers (L2, L3) or deep layers (L5, L6) of these distinct ACC areas. To accomplish this, we utilized tissue from a total n = 6 cases (Table 1; six fields per layer), where injections of retrograde neural tract tracers were placed in the basolateral amygdala complex to label AMY projection neuron cell bodies in ACC (Fig. 2A,B). We quantified the laminar densities of tracer-labeled AMY projection neurons with m1 and/or m2 colabeling within A24, A25, and A32 (Fig. 2C–H).

Figure 2.

Distribution of m1+ and m2+ AMY-targeting pyramidal neurons in the ACC. A, Schematic map and (B) representative images of coronal sections through the AMY (from cases PIO, FR tracer in cortical nucleus; PIP, FE tracer in BLA) and showing locations of injection sites in AMY. Note that some cases were also used in previous studies (Calderrazzo et al., 2021; Medalla et al., 2022). Abbreviations of AMY nuclei: AAA, anterior; ACo, anterior cortical; Ce, central; Co, cortical; BM, basomedial; BL, basolateral; L, lateral; M, medial. C, Representative confocal image of triple labeling of FR tracer (red), m1 (blue), and m2 (magenta) to show tracer-labeled pyramidal neurons that are m1+ m2+ (yellow arrow), m1−m2+ (white arrow), or m1+ m2− (blue arrow). Scale bar, 25 µm. D–H, Box and whisker plots of the density of AMY-targeting neurons in ACC A24 (pink, top), A25 (purple, middle), and A32 (green, bottom): total AMY-targeting neurons (D), m1+ (m2−) AMY-targeting neurons (E), m2+ (m1−) AMY+ targeting neurons (F), m1+ m2+ AMY-targeting neurons (G), and m1−m2− AMY-targeting neurons (H) in the upper cortical layer 2–3 (light shade) or deep layers 5–6 (dark shade). I, J, Pie charts of the relative proportion of AMY-targeting neurons in either the upper layers (F) or deep layers (G) that expressed either, both, or neither mAChRs. Data from n = 6 monkeys; box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, **p < 0.01, ***p < 0.001. See Extended Data Table 2-1 for details on the statistical analyses.

Table 2-1

Two-Way ANOVA (area x layer) p-values for within and between area comparisons of tracer labeled neurons. Download Table 2-1, DOCX file ^{(28.1KB, docx)} .

Within-area comparisons showed equivalent densities of tracer-labeled AMY-targeting neurons in the upper and deeper layers. However, between-area comparisons showed that A25 had a significantly greater mean density of AMY-targeting neurons in both upper and deep layers compared with either A24 or A32 (p < 0.05 for all comparisons; two-way ANOVA area × layer; main effect area; Fisher’s LSD post hoc; Fig. 2D; Extended Data Table 2-1), consistent with our previous stereological whole brain mapping study (Calderazzo et al., 2021). We then examined the subpopulation of AMY-targeting neurons that express m1 or m2 receptors. We found that majority of the AMY-targeting neurons in all ACC areas express either m1, m2, or both receptors (Fig. 2D–J). In both upper and deep layers of all ACC areas, we found a greater proportion of AMY-projecting neurons that were mAChR+, compared with the subpopulation that were mAChR− (Fig. 2I,J). Within-area comparisons revealed that A24 had a significantly greater proportion of m2+ AMY-targeting neurons in deep compared with upper layers (p = 0.002), while both A25 and A32 had similar laminar patterns of expression of mAChRs expressing AMY projection neurons. Between-area comparison revealed that A32 had a significantly greater proportion of AMY-targeting neurons coexpressing m1 and m2 in the upper layers compared with other areas (A32 vs A25 p = 0.018; A32 vs A24 p = 0.058). In the deep layers, A25 had a significantly greater proportion of AMY-targeting neurons expressing m1+ alone (m2−) compared with A24 (p = 0.016; Fig. 2I). Conversely, deep layers of A24 had a significantly greater proportion of AMY-targeting neurons that express m2+ alone (m1−; 42%) than in A25 or A32 (p < 0.001 for all comparisons; Fig. 2J, Extended Data Table 2-1). Approximately 42% of AMY-projecting neurons in A24 were m2+ (m1−), which is significantly greater than those in A25 and A32 (∼2%). These data show diverse relative distributions of m1 versus m2 expressing subpopulations of AMY-targeting neurons specifically in the deep layers; A25 had a relatively greater frequency of m1+ AMY projection neurons, while A24 exhibited relatively a greater frequency of m2+ AMY projection neurons.

Subcellular localization of mAChR along individual tracer-labeled AMY-targeting neurons

We next examined and quantified the subcellular localization of m1+ and m2+ receptors along the somatic and proximal apical dendritic compartments of individual AMY-projecting neurons in the upper (L2–L3) and deep layers (L5–L6; Fig. 3A,B). After performing a segmentation of somatic versus apical dendritic trunk ROIs (Fig. 3B), we calculated the density (% area) of m1+ and m2+ receptor labeling within each subcellular compartment of AMY-projecting neurons in A24 (upper layer n = 15 soma/dendrite, deep layer n = 11 soma/dendrite), A25 (upper layer n = 14 soma/dendrite, deep layer n = 14 soma/dendrite), A32 (upper layer n = 9 soma, n = 8 dendrite, deep layer n = 9 soma, n = 10 dendrite). For both the somatic and dendritic compartments of AMY-targeting projection neurons, our findings show a greater degree of laminar and regional differences with regard to m2+ receptor density (% area labeled) compared with m1.

Figure 3.

Somatodendritic m1+ and m2+ receptor localization on AMY-targeting pyramidal neurons in ACC. A, Representative confocal image of tracer-labeled (red) pyramidal neurons that are m1+ m2+ (yellow arrow), m1−m2+ (white arrow), or m1+ m2− (blue arrow). Scale bar, 25 μm. B, Representative confocal image showing the somatic and proximal apical dendritic regions of interest (white dotted lines) on a retrogradely labeled AMY-targeting pyramidal neurons (red) with m1 (blue), m2 (magenta), and merged composite image (maximum projection of z-stack). Scale bar, 10 µm. C, D, Box and whisker plots of receptor density (% area label) within somatic ROI and (E, F) receptor density (% area label) within proximal apical dendritic ROI. Data from n = 6 monkeys: A24: upper layer n = 15 soma/dendrite, deep layer n = 11 soma/dendrite; A25: upper layer n = 14 soma/dendrite, deep layer n = 14 soma/dendrite; A32: upper layer n = 9 soma, n = 8 dendrite, deep layer n = 9 soma, n = 10 dendrite; box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, **p < 0.01, ***p < 0.001. See Extended Data Table 2-1 for details on the statistical analyses.

Between-area comparisons of m1+ label density in the somatic compartment revealed that A32 and A25 upper layer neurons had a significantly greater density of somatic m1+ than in A24 (two-way ANOVA area × layer; main effect area; p = 0.004; Fig. 3C, Extended Data Table 2-1). A24 had the lowest density of m1+ label in the soma compared with those in either A25 (Fisher’s LSD post hoc; p = 0.041) or A32 (p = 0.001; Fig. 3C). Similarly, we observed that upper layer A32 neurons had a significantly greater density of somatic m2+ label compared with those in both A24 (p = 0.010) and A25 (p = 7.51 × 10⁻⁴); however, the deep layer neurons in both A24 and 25 had a significantly greater density of somatic m2+ compared with those in A32 (A24 vs A32: p = 0.001, A25 vs A32 p = 0.004; Fig. 3D). Within area, we observed that A32 upper layer neurons had a greater density of somatic m2+ label compared with deep layer neurons (p = 7.24 × 10⁻⁵), while A25 deep layer neurons had a greater somatic m2+ density (p = 0.035; Fig. 3D).

The density of m1+ within the dendritic apical trunk was similar between upper and deep layer AMY-targeting neurons within each area. However, between-area comparisons showed that A25 upper layer neurons had a significantly greater density of dendritic m1+ compared with neurons in A24 (p = 0.046; two-way ANOVA; Fisher’s LSD post hoc) or A32 (p = 0.010; Fig. 3E). Within-area comparisons of dendritic m2+ density showed similar densities between upper and deep neurons in A24 and A25, but a greater density of dendritic m2+ in A32 in upper compared with deep layer neurons (p = 1.95 × 10⁻⁴; Fig. 3F). Further, between-area comparisons showed that A32 upper layer neurons had a greater density of dendritic m2+ compared with neurons in A24 (p = 2.88 × 10⁻⁵) or A25 (p = 0.006). In contrast, L5 A32 neurons exhibited a significantly lower density of m2+ within dendrites when compared with A25 L5 neurons (p = 0.006; Fig. 3F).

In summary these data reveal regional difference in m1 and m2 receptor densities within specific subcellular compartments of AMY-targeting neurons that are dependent on layer. Strong laminar differences were evident in A32 AMY-targeting neurons, with greater m1 and m2 receptor densities within somato-dendritic compartments of upper compared with deep neurons. Between-area differences were more prominent in the upper layer neurons, with AMY-targeting neurons from A25 and A32 having relatively greater m1 densities within somato-dendritic domains compared with A24.

Density and proportion of m1+ and m2+ expressing CB, PV, and CR inhibitory neurons in distinct ACC areas

We then compared across ACC areas the expression of m1+ and m2+ receptors on the three nonoverlapping neurochemical classes of inhibitory neurons within the rhesus monkey, identified by their expression of the calcium-binding proteins parvalbumin (PV; Fig. 4), calbindin (CB; Fig. 5), and calretinin (CR; Fig. 6). Using tissue labeled with CB, PV, CR, and either m1 or m2, we assessed the laminar distribution and relative proportion of each interneuron subtype coexpressing either m1 and m2 receptors within and between each region of the ACC. In accordance with previously published work (Gabbott and Bacon, 1996; Dombrowski et al., 2001), we observed within-area laminar density differences among inhibitory neurons that were consistent in pattern across ACC areas. However, compared with our previous study (Tsolias and Medalla, 2022), the new set of tissue, which had different location and number of sampling sites, yielded, in general, a greater proportion of interneurons that were immuno-negative for m1 receptors. For all ACC areas and interneuron subtypes, we found that the proportion of m1+ or m2+ expressing subpopulations were equivalent. However, between-area differences were found that were layer and marker dependent.

Figure 4.

Distribution of m1+ and m2+ expressing PV+ interneurons. A, Representative confocal images of PV+ interneurons (green) coexpressing either m1+ (top) or m2+ (bottom) mAChRs (magenta). Scale bar, 20 µm. B–D, Box and whisker cell density plots of m1+ (filled) or m1− (open) PV+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). E–G, Box and whisker cell density plots of m2+ (filled) or m2− (open) PV+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). H–M, The relative proportion of PV+ m1+ (H–J) or PV+ m2+ (K–M) neurons across the three subdivisions of the ACC in L2, L3, L5, and L6. Data from n = 6 monkeys; box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, **p < 0.01, ***p < 0.001. See Extended Data Tables 4-1 (within area), 4-2 (between area), and 4-3 (Fisher's exact tests for proportions) for details on the statistical analyses.

Figure 5.

Distribution of m1+ and m2+ expressing CB+ interneurons. A, Representative confocal images of CB+ interneurons (green) coexpressing either m1+ (top) or m2+ (bottom) mAChRs (magenta). Scale bar, 20 µm. B–D, Box and whisker cell density plots of m1+ (filled) or m1− (open) CB+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). E–G, Box and whisker cell density plots of m2+ (filled) or m2− (open) CB+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). H–M, The relative proportion of CB+ m1+ (H–J) or CB+ m2+ (K–M) neurons across the three subdivisions of the ACC in L2, L3, L5, and L6. Data from n = 6 monkeys; Box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, ***p < 0.001. See Extended Data Tables 4-1 (within area), 4-2 (between area), and 4-3 (Fisher's exact tests for proportions) for details on the statistical analyses.

Figure 6.

Distribution of m1+ and m2+ expressing CR+ interneurons. A, Representative confocal images of CR+ interneurons (green) coexpressing either m1+ (top) or m2+ (bottom) mAChRs (magenta). Scale bar, 20 µm. B–D, Box and whisker cell density plots of m1+ (filled) or m1− (open) CR+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). E–G, Box and whisker cell density plots of m2+ (filled) or m2− (open) CR+ neurons in L2, L3, L5, and L6 of ACC A24 (pink), A25 (purple), and A32 (green). H–M, The relative proportion of CR+ m1+ (H–J) or CR+ m2+ (K–M) neurons across the three subdivisions of the ACC in L2, L3, L5, and L6. Data from n = 6 monkeys; Box and whisker plots: bars show interquartile range and median (horizontal line) with mean = x and error bars = 95% confidence interval; *p < 0.05, **p < 0.01, ***p < 0.001. See Extended Data Tables 4-1 (within area), 4-2 (between area), and 4-3 (Fisher's exact tests for proportions) for details on the statistical analyses.

Table 4-1

Two-Way ANOVA (area x layer) p-values for within-area comparisons of interneuron distributions. Download Table 4-1, DOCX file ^{(40.5KB, docx)} .

Table 4-2

Two-Way ANOVA (area x layer) p-values for between-area comparisons of interneuron distributions. Download Table 4-2, DOCX file ^{(36.4KB, docx)} .

Table 4-3

Fisher's exact tests for comparing proportions within interneuron subpopulations. Download Table 4-3, DOCX file ^{(27.2KB, docx)} .

PV+m1+/m2+ interneurons

Within-area comparisons revealed that, in general, PV m1+/m2+ neurons were more abundant in the middle to deep layers. We found that A24 had a significantly greater density of PV+m1+ in L3 compared with L2 (p = 0.03; two-way ANOVA layer × area for each marker; Fisher's LSD post hoc; Fig. 4B, Extended Data Table 4-1), and A25 and A32 had a significantly greater density of PV+m1+ in L5 compared with L2 and L3 (p < 0.05 for all comparisons; Fig. 4C,D, Extended Data Table 4-1). Similar to PV+m1+, the density of PV+m2+ was significantly greater in L3 and L5 compared with L2 in A24 (Fig. 4E), was greater in L5 compared with L2, L3, and L6 in A25 (Fig. 4F), and was greater in L5 compared with L6 in A32 (p < 0.05 for all comparisons; Fig. 4G, Extended Data Table 4-1).

Between-area comparisons showed that A25 had a greater mean density of PV+m1+ neurons than A24 in L5 (p = 0.01; Fig. 4B,C; Extended Data Table 4-2). Further, we observed that PV+m2+ neurons had a greater mean density in A32 L2 compared with A25 (p = 0.03; Fig. 4F,G). In L3, we observed that A24 had a greater mean density of PV+m2+ neurons than A25 (p = 0.02; Fig. 4E,F).

The proportion of either m1+ or m2+ expressing PV+ neurons in A24 was significantly greater in all layers compared with L2 (p < 0.01 for all comparisons; Fisher’s exact test; Fig. 4H,K; Extended Data Table 4-3). Additionally, A24 L3 (89%) and L6 (92%) had significantly greater proportions of PV+m1+ compared with L5 (73%; p < 0.01 for all comparisons; Fig. 4H, Extended Data Table 4-3). In A25, the proportion of PV+m1+ neurons was significantly greater in L5 (89%) compared with L2 (68%; p = 4.8 × 10⁻⁴; Fig. 4I), while both L3 (77%) and L6 (67%) had significantly greater proportions than L5 (89%; p < 0.05; Fig. 4I, Extended Data Table 4-3). However, there were no significant laminar differences in the proportions of PV+m2+ neurons (Fig. 4L, Extended Data Table 4-3). In A32, the proportion of PV+m1+ neurons was greater within L3 (87%), L6 (93%), and L5 (82%), compared with L2 (49%; p < 0.001 for all comparisons; Fig. 4J, Extended Data Table 4-3). The highest proportion of PV+m1+ neurons in A32 was found in L6 (93%; Fig. 4J, Extended Data Table 4-3). In contrast, PV+m2+ in A32 were present in higher proportions within L5 (91%) compared with L3 (76%) and L6 (78%; p < 0.05 for all comparisons; Fig. 4M, Extended Data Table 4-3).

Between-area comparisons showed a significantly greater proportion of PV+m1+ neurons in A25 L2 (68%) and L5 (89%) compared with A24 (L2, 53%; L5, 73%) and A32 (L2, 49%; p < 0.05 for all comparisons; Fisher’s exact test; Fig. 4H–J; Extended Data Table 4-3). In L3 we found that A24 (89%) had a greater proportion of PV+m1+ neurons compared with A25 (77%), while in L6 both A24 (92%) and A32 (93%) had greater proportions compared with A25 (67%; p < 0.05 for all comparisons; Fig. 4H–J; Extended Data Tables 4-2, 4-3). PV+m2+ neurons had the lowest proportion in A25 in L3 and the deep layers (L5, L6) compared with other areas (p < 0.01 for all comparisons; Fig. 4K–M, Extended Data Table 4-3).

CB+m1+/m2+ interneurons

Quantification of strongly labeled inhibitory CB+ interneurons revealed a similar laminar distribution of CB+m1+ and CB+m2+ neurons across the distinct subdivisions of the ACC (Fig. 5A–G). Within all ACC areas, significantly greater densities of CB+m1+ and CB+m2+ neurons were found in L2 compared with L3, L5, and L6 (p < 0.051 for all comparisons; two-way ANOVA area × layer; Fisher’s LSD post hoc; Fig. 5B–G, Extended Data Table 4-1). Between-area comparisons revealed that A24 and A32 L2 had a significantly greater density of CB+m1+ compared with A25 (p < 0.001 for all comparisons; Fig. 5B–D; Extended Data Table 4-2). Conversely, A25 L6 had a significantly greater density of CB+m2+ compared with either A24 or A32 (p < 0.05 for all comparisons; Fig. 5E–G; Extended Data Table 4-2).

The proportion of CB+m1+ neurons was significantly greater in A24 L3 (89%) and L6 (88%) compared with L2 (76%; p < 0.05 for all comparisons; Fisher’s exact test; Fig. 5H, Extended Data Table 4-3). Similarly, in A5 L3 (81%) and L5 (87%) had significantly greater proportions compared with L2 (68%; p < 0.05 for all comparisons; Fig. 5I, Extended Data Table 4-3). In A32, L6 (98%) had a significantly greater density compared with all other layers (L2, 89%; L3, 83%; L5, 78%; p < 0.05 for all comparisons; Fig. 5J, Extended Data Table 4-3). Conversely, the proportion of CB+m2+ neurons was only significantly different between the deep layers of A24, where L6 (86%) was significantly greater than L5 (73%; p = 0.03; Fig. 5K, Extended Data Table 4-3). Between-area differences that are dependent on layer were found only for the proportion of CB+m1+, while CB+m2+ was equivalent. In both L2 and L6, A32 had a significantly greater proportion of CB+m1+ neurons compared with either A24 or A25 (p < 0.05 for all comparisons; Fisher’s exact test; Fig. 5H–J, Extended Data Table 4-3).

CR+m1+/m2+ interneurons

Within-area comparisons revealed that both CR+ m1+ and CR+ m2+ had a significantly greater mean densities in L2 compared with L3, L5, and L6 within all ACC areas (p < 0.01 for all comparisons; Fig. 6B–G; Extended Data Table 4-1). Between areas, A25 had a significantly greater mean density of CR+m1+ neurons compared with both A24 and A25 in L2 and compared with A24 in L6 (p < 0.05; for all comparisons; Fig. 6B,C; Extended Data Table 4-2).

A comparison of the proportions (Fig. 6H–M) found differences within- and between- area in both m1+ and m2+ expressing CR+ interneurons. We observed both A24 and A32 had significantly lower proportions of CR+m1+ in L2 compared with all other layers (p < 0.05; Fisher's exact test; Fig. 6H,J; Extended Data Table 4-3). Within A25, we found equivalent proportions of CR+m1+ neurons across layers (Fig. 6I). In examining the proportion of CR+m2+ neurons, we found that within A24, L2 (62%) had a significantly lower proportion compared with all other layers (81–92%; p < 0.05; Fig. 6K; Extended Data Table 4-3), and L6 (92%) had a greater proportion compared with L3 (81%; p = 0.04; Fig. 6K; Extended Data Table 4-3). Similarly, in A25, we found L2 (44%) had a significantly lower proportion of CR+m2+ neurons compared with all other layers (60–81%; p < 0.05; Fig. 6L, Extended Data Table 4-3). Lastly, A32 also had a lower proportion of CR+m2+ neurons in L2 (63%) compared with either L3 (77%) or L5 (80%; p < 0.05; Fig. 6M, Extended Data Table 4-3). Between-area comparisons revealed a significantly greater proportion of m2+ expressing neurons in A24 L2 and L6 compared with A25 (A24 vs A25: L2; p = 0.02 and L6 p = 1.20 × 10⁻⁷) and A32 L6 (A24 vs A32 L6 p = 6.45 × 10⁻⁴; Fig. 6K–M; Extended Data Table 4-3). Additionally, A32 had a significantly greater proportion of m2+CR+ neurons compared with A25 (p = 0.01; Fig. 6L,M; Extended Data Table 4-3).

Differences in relative laminar distribution of distinct interneuron subpopulations expressing m1+ or m2+ in ACC areas

Within each area, we compared the relative laminar predominance of distinct inhibitory neuron subtypes expressing m1+ and m2+. We normalized the density distribution of each subpopulation to the total density of all interneurons (PV, CB, CR) in each area (Fig. 7). Within A24 and A32, we observed that CR+ and CB+m1+/m2+ neurons showed similar laminar distribution patterns, with a strong predominance in the upper compared with the deep layers (p < 0.05, for all comparisons; two-way ANOVA area × layer; Fisher’s LSD post hoc; Extended Data Table 4-1; Fig. 7A,B). The PV+ subpopulations showed an opposite laminar pattern, with greater density in the deep layers compared with upper layers (p < 0.05 for all comparisons; Extended Data Table 4-1; Fig. 7A,B). Interestingly, for A24 and A32, there is a predominance of CB+m1+ and CR+m1+ interneurons in the L2 compared with PV+m1+ (p < 0.05; for all comparisons; Fig. 7B,D,F). However, this upper layer bias is less pronounced in A25 CB+m1+ neurons (Fig. 7D), and all the m1− subpopulations (Fig. 7B,D,F, right). Indeed there is a significant main effect of layer (p = 0.00016) and interactive effect of layer × area (p = 0.024; Extended Data Table 4-2) on the upper layer predominance of CB+m1+ neurons; the relative frequency of CB+m1+ in upper layers was significantly less in A25 compared with A32 and A24 (Fischer's LSD post hoc; p < 0.001; Fig. 7B,D,F; Extended Data Table 4-2). In contrast, the upper layers A25 had relatively high concentrations of CR+m1+ neurons, which strongly predominated in L2 and L3 compared with the other cell subclasses and layers (p < 0.001; for all comparisons; Fig. 7D, Extended Data Table 4-1). An upper layer predominance of m2+ and m2− subpopulations CB+ and CR+ interneurons was also found in A24 and A25, with significantly greater densities in L2 compared with the other layers (p < 0.05; Fig. 7C,E,G; Extended Data Table 4-1). In A32 however, this upper layer increase in L2 relative to other layers was only significant for CB+m2+ and CR+m2− subpopulations (p < 0.0001; Fig. 7G, Extended Data Table 4-2). In summary, while the general laminar distributions of interneurons subpopulations were similar across areas, the degree of upper layer predominance of CB and CR subpopulations based on m1 and m2 expression differed.

Figure 7.

Relative laminar distribution of mAChR expressing CB+PV+ and CR+ inhibitory interneuron subpopulations in ACC areas. A, Representative low magnification confocal column images of coronal sections (pial surface to white matter) of m1/m2 (magenta), PV (green), CB (blue), and CR (red). Scale bar, 100 µm. B–F, Mean normalized laminar density of interneuron cell types PV (green), CB (blue), and CR (red) that are m1+ or m1− (B,D,F) or m2+ or m2− (C,E,G) as a proportion of total summed interneuron densities in each area. Data are means from n = 6 monkeys; error bars represent SEM.

Presynaptic colocalization of m2+ receptors on inhibitory VGAT+ axon terminals

The m2+ receptor subtype has been shown to localize with presynaptic GABAergic axon terminals to suppress neurotransmitter release (Mrzljak et al., 1993; Hajos et al., 1998; Salgado et al., 2007; Takacs et al., 2018). Our previous work examined the colocalization of m2+ with vesicular GABA transporter (VGAT), a protein expressed in GABAergic terminals (Chaudhry et al., 1998), in the neuropil of the superficial layers of the A24 (L1–L3) and observed no laminar differences in either the density of m2+ or VGAT+m2+ colocalized puncta (Tsolias and Medalla, 2022). Here, we extended our work to examine and compare the laminar density of m2+VGAT+ terminals within the neuropil of A24, A25, and A32 (Fig. 8A–D).

Figure 8.

Presynaptic localization of m2+ with inhibitory axon terminals in ACC. A, Representative confocal image of m2+ (magenta), VGAT (green), and m2+ VGAT+ colocalized puncta (white). Scale bar, 10 µm. B, C, The density (% area label) of m2+ (B), VGAT (C), and VGAT+ with m2+ (D) in the superficial and deep layers of ACC areas A24, A25, and A32. Data are means from n = 6 monkeys; error bars represent SEM. See Extended Data Table 8-1 for details on the statistical analyses.

Table 8-1

Two-Way ANOVA p-values for within and between-area comparisons of particle densities. Download Table 8-1, DOCX file ^{(23.9KB, docx)} .

Within-area comparisons revealed significant laminar differences in m2+ density (% area) within A24, with L5 having a significantly greater mean density of m2+ compared with all other layers examined (p < 0.01; for all comparisons; two-way ANOVA area × layer; Fisher LSD post hoc; Fig. 8B, Extended Data Table 8-1). A32 similarly revealed a significantly greater density of m2 label in L5 compared with L1 (p = 0.009; Fig. 8B, Extended Data Table 8-1). However, A25 revealed a more uniform density of m2 across the neuropil of both the superficial and deep layers (Fig. 8B). While the density of VGAT+ puncta did not significantly differ within or between the ACC areas (Fig. 8C), there were differences in the density of VGAT+m2+ colocalized puncta (Fig. 8D). Within each area, we found that L5 had a significantly greater density of colocalized VGAT+m2+ compared with L1 (p < 0.05 for all comparisons; Fig. 8D, Extended Data Table 8-1). This L5 peak in VGAT+m2+ puncta density was most pronounced in ACC A24 (Fig. 8D).

Between-area comparisons showed that within the superficial layers, A25 had a significantly greater density of m2+ puncta compared with both A24 and A32 (L2 and L3 A25 vs A24; and L3 A25 vs A32 p < 0.05 for all comparisons; two-way ANOVA area × layer; Fisher LSD post hoc; Fig. 8B, Extended Data Table 8-1). Further, the density of VGAT+m2+ colocalized puncta was significantly greater in A25 compared with A24 and A32, especially in L2, L3, and L6 (p < 0.03 for all comparisons; Fig. 8D, Extended Data Table 8-1). Interestingly, A24 and A32 had a similar density of VGAT+m2+ colocalized puncta in all layers except in L5 where A24 had a significantly greater density (p = 0.038; Fig. 8D, Extended Data Table 8-1). These data suggest that m2+ mediated suppression of inhibitory synapses is greater in the upper layers of A25 compared with A24 and A32.

Discussion

Our findings revealed laminar, cell-type, and circuit-specific expression of m1 and m2 receptors across distinct ACC areas of the rhesus monkey. The data show a general pattern of greater mAChR-expressing excitatory subpopulations and presynaptic m2+ localization on inhibitory terminals within ACC layers and areas that provide robust corticolimbic outputs (Aggleton et al., 1980; Barbas and Rempel-Clower, 1997; Ghashghaei et al., 2007).

Laminar distribution of mAChR+ pyramidal neurons suggest pathway-specific cholinergic modulation

Our data revealed that the proportion of MAP2+ neurons expressing m1+ and m2+ in A24 was greater in deep compared with superficial layers, while A25 and A32 had more uniform laminar distributions. Further, between-area differences were dependent on layer, with a greater proportion of MAP2+mAChR+ neurons in A24 compared with A32 and A25 in the deep layers, while the opposite pattern was found for upper layers. These data suggest that muscarinic modulation of pyramidal neurons is more robust in the deep layers of A24, but in the upper layers of A32 and A25. Pyramidal projection neurons that arise in distinct layers are thought to be significant for the temporal flow and dynamics of signal representations (Barbas, 2015). There is evidence from sensory cortices that upper layer neurons provide feedforward “driving” signals, while deep layer neurons form feedback pathways, as well as provide major outputs to subcortical structures (Felleman and Van Essen, 1991; Hilgetag et al., 2016). Further, data from monkey prefrontal and premotor cortices show that cognitive task-related oscillatory dynamics in the upper layers are consistent with rapid signal processing for selection, while deep layer signals represent slower dynamics associated with response outputs (Ninomiya et al., 2015; Chandrasekaran et al., 2017; Bastos et al., 2018, Mendoza-Halliday et al., 2024).

The upper layers of A32 and A25 are thought to provide “feedforward” input to the amygdala and other limbic orbitofrontal, cingulate, and rhinal cortices (Ghashghaei et al., 2007; Bunce et al., 2013; Joyce and Barbas, 2018; Calderazzo et al., 2021). Thus, the predominance of mAChR+ upper layer MAP2+ pyramidal neurons in A32 and A25 suggest greater cholinergic modulation of these “feedforward” corticolimbic pathways. The greater densities of mAChR+ deep layer MAP2+ pyramidal neurons of A24 suggests greater cholinergic influence on the ACC→AMY and ACC→premotor “feedback” projections that both emanate from L5 (Calderazzo et al., 2021). Further, our previous single-cell biophysical study suggest that compared with premotor-targeting neurons, AMY-targeting neurons in L5 of A24 are more excitable and have a greater dynamic range of action potential firing to support low and high frequency synchrony (Medalla et al., 2022). Given that these biophysically distinct pools of L5 neurons in A24 will likely be differentially modulated by ACh (McCormick and Prince, 1985), our data provide an anatomical substrate for cholinergic modulation of coordinated motivational and motor signals in A24 (Kennerley et al., 2006).

Regional diversity in m1 versus m2 expression of amygdalar projection neurons from ACC: implications on region-specific cholinergic modulation of ACC→amygdala pathways

In primates, AMY-targeting neurons emanate from both upper and deep layers of the ACC depending on the specific subareas (Ghashghaei et al., 2007; Calderazzo et al., 2021). Our data here show that the density of AMY-targeting neurons expressing m1+/m2+ receptors is highest in A25 compared with A24 and A32. Further, in deep layers, there is regional diversity in the distributions of m1+ versus m2+ AMY-targeting neuronal subpopulations. AMY-targeting neurons from the deep layers of A25 and A32 predominantly coexpress m1+ m2+ (∼82–83%) or m1+ (∼11–13%), and a very small proportion express m2+ alone (∼2%). In contrast, within the deep layers of A24, AMY-targeting neurons mainly express m2+ only (44%), consistent with previous autoradiographic studies (Zilles and Palomero-Gallagher, 2017). Based on their predominant presynaptic localization, m2+ expressed within AMY-targeting neuron somata are likely trafficked to axon terminals within the amygdala, to suppress glutamatergic release in the presence of ACh (Hasselmo and Bower, 1992; Levey, 1996; Salgado et al., 2007; Thiele, 2013). While future physiological studies are needed, our data suggest that A32 and A25 projections to AMY are more likely to be subjected to m1+ mediated cholinergic activation, while deep layer A24 projections to AMY are more subjected to m2+ mediated cholinergic suppression.

Since m1 receptor activation have postsynaptic depolarizing effects (Haj-Dahmane and Andrade, 1996), the m1+ AMY-targeting neurons in A25 and A32 are likely stimulated by ACh. Further, we found that AMY-projecting neurons in A32 and A25 exhibited a greater expression of m1+ compared with m2+ receptors within somata. Thus, A25 and A32 not only have a higher frequency of m1+ AMY-targeting neurons, but these neurons also have m1+ enriched somatic compartments where they can elicit stronger activating effects. These m1+ A25 and A32 projections can activate the basolateral (BLA) AMY nucleus to promote emotional salience, autonomic arousal, and mnemonic processing (Jhang et al., 2018; Kellis et al., 2020; Tryon et al., 2021; reviews, Gold, 2003; Etkin et al., 2011; Orsini and Maren, 2012). Indeed, ACh has direct activating effects on rodent medial frontal areas analogous to the primate ACC, to enhance signal processing for cognitive tasks and facilitate reward and fear learning (McGaughy et al., 1996; Parikh et al., 2007; Howe et al., 2017; Major et al., 2018; reviews, Gold, 2003; Etkin et al., 2011; Orsini and Maren, 2012). Further, BLA neuronal firing in neurons is directly enhanced by ACh stimulation to mediate fear learning (Jiang et al., 2016). While muscarinic modulation of working memory-related activity has been demonstrated in lateral prefrontal cortex of rhesus monkeys (Major et al., 2018; Vijayraghavan and Everling, 2021), the physiological and behavioral effects of mAChR activation within the ACC→AMY circuits in primates are unknown. In vivo experiments in monkeys show that AMY lesions result in diminished value coding in ventral/orbital prefrontal areas (which include A25), but not in dorsal ACC A24 (Rudebeck et al., 2013). The capability for circuit-specific m1 and m2 modulation of ACC→AMY pathways is likely an important molecular-anatomical substrate of synaptic plasticity for learning and memory (Hasselmo and McGaughy, 2004; Disney et al., 2007). Subgenual A25 is thought to be important for sustaining arousal during anticipated rewards (Rudebeck et al., 2014), a process that can be facilitated by cholinergic modulation of A25-AMY interconnections.

Cholinergic modulation of inhibitory microcircuits in ACC areas

Together with neuromodulatory influences, inhibitory neurons can control distinct network states and facilitate communication between regions (Markram et al., 2004; DeFelipe et al., 2013; Cardin, 2018). We found that the distribution of m1+/m2+ expressing PV, CB, and CR inhibitory interneuron populations differed across ACC areas, in patterns that suggest stronger net excitatory effects of ACh especially in A25 and A32. Consistent with the known laminar cortical distribution of these inhibitory neurons (DeFelipe et al., 2013), CB+ and CR+mAChR+ were concentrated in upper layers, while PV+mAChR+ predominated in deep layers. The ACC areas mainly differed in the predominance of m1+ versus m2+ expressing CB+ and CR+ neurons in the upper layers. A32 and A24 exhibit a higher proportion of CB+m1+ neurons in the upper layers compared with A25, suggesting that ACh can engage CB mediated dendritic inhibition. In contrast, in A25, we found that CR+m1+ neurons predominate over PV+m1+ and CB+ m1+ subpopulations, suggesting that ACh can mainly enhance activity of CR+ interneurons, which are thought to engage disinhibition in cortical upper layers (del Rio and DeFelipe, 1997; Meskenaite, 1997; Melchitzky et al., 2005). Further, we found a greater density of VGAT+m2+ axon terminals, suggesting enhanced presynaptic suppression of GABA release (Hajos et al., 1998; Salgado et al., 2007), in the upper layers of A25 compared with the other ACC areas. Thus, this cholinergic suppression of inhibition (via m2+ on inhibitory terminals) together with enhancement of disinhibition (via CR+m1+ activation) in A25 can facilitate the potential cholinergic activation of upper layer corticolimbic output neurons.

Summary

The present findings characterize the anatomical substrates of local and long-range circuit-specific cholinergic modulation within the primate ACC. The mAChR expression patterns shown here suggest how ACh can particularly disinhibit and enhance corticolimbic outputs from ventral ACC A25 to AMY. While cholinergic modulation can enhance attentional and motivational components of cognitive tasks, excessively high levels of ACh, such us during stress, can lead to hyperexcitability in ACC A25 (Wang et al., 2004; Dasilva et al., 2019; Williams and Fletcher, 2019; reviews, Devinsky et al., 1995; Hasselmo and Sarter, 2011; Picciotto et al., 2012). Thus, with neurochemical imbalance, A25 outputs to AMY could be subject to a higher risk of hyperexcitability and excitotoxicity, consistent with pathology seen in affective disorders such as depression (reviews, Drevets et al., 2008; Hamani et al., 2011; Arnsten et al., 2023). Indeed, the cholinergic system and ACC dysfunction are both linked to mood and anxiety disorders (Janowsky et al., 1974; Phillips et al., 2003; Milad et al., 2007; Ressler and Mayberg, 2007). Our findings provide insight on how disruption of ACC circuits in conjunction with cholinergic dysfunction can lead to misattribution of motivational context and emotional salience (Nemeth et al., 1988; Rauch et al., 2006; Peters et al., 2009), as seen in depression, bipolar disorder, and post-traumatic stress disorder.