Anterior Cingulate Pathways May Affect Emotions Through Orbitofrontal Cortex

Abstract

The anterior cingulate cortex (ACC) and posterior orbitofrontal cortex (pOFC) are associated with emotional regulation. These regions are old in phylogeny and have widespread connections with eulaminate neocortices, intricately linking areas associated with emotion and cognition. The ACC and pOFC have distinct cortical and subcortical connections and are also interlinked, but the pattern of their connections—which may be used to infer the flow of information between them—is not well understood. Here we found that pathways from ACC area 32 innervated all pOFC areas with a significant proportion of large and efficient terminals, seen at the level of the system and the synapse. The pathway from area 32 targeted overwhelmingly elements of excitatory neurons in pOFC, with few postsynaptic sites found on presumed inhibitory neurons. Moreover, pathways from area 32 originated mostly in the upper layers and innervated preferentially the middle-deep layers of the least differentiated pOFC areas, in a pattern reminiscent of feedforward communication. Pathway terminations from area 32 overlapped in the deep layers of pOFC with output pathways that project to the thalamus and the amygdala, and may have cascading downstream effects on emotional and cognitive processes and their disruption in psychiatric disorders.

Introduction

The anterior cingulate cortex (ACC) and posterior orbitofrontal cortex (pOFC) are 2 regions of the prefrontal cortex that have a key role in emotions. These regions are old in phylogeny and have widespread connections with eulaminate neocortices, intricately linking areas associated with emotion and cognition [reviewed in Barbas et al. (2011), Barbas (2015)]. However, the ACC and pOFC have distinct sets of connections with cortical, thalamic, and other subcortical structures [reviewed in Barbas et al. (2011)]. The pOFC is distinguished as a major receiver of information from the external environment through high-order sensory association cortices and from the internal environment through the amygdala and other limbic structures (Morecraft et al. 1992; Barbas 1993; Cavada et al. 2000). In contrast, the ACC has significant sensory-related connections only with auditory association areas (Barbas et al. 1999; Medalla and Barbas 2014), but is a chief sender of pathways to the amygdala (Ghashghaei et al. 2007), as well as to a variety of autonomic motor systems (Ongur et al. 1998; Rempel-Clower and Barbas 1998; Barbas et al. 2003). These findings suggest that the pOFC may have a role in assessing the significance of sensory signals. On the other hand, the ACC may be a cortical effector for emotions through widespread and specialized connections with autonomic motor systems.

The pOFC and ACC are also strongly connected with each other (Barbas and Pandya 1989; Barbas 1993; Cavada et al. 2000; Johansen-Berg et al. 2008; Zald et al. 2014), in patterns that are not well understood. In the sensory systems, the laminar pattern of connections has been associated with the flow of signals between cortices (Felleman and Van Essen 1991). A structural model that relates connections to the laminar architecture of linked areas makes predictions about patterns of connections between cortices, in general (Barbas and Rempel-Clower 1997; Barbas 2015). The ACC and pOFC are composed of areas with the simplest laminar structure, characteristic of limbic cortices (Barbas and Pandya 1989). Based on the prediction of the structural model, the 2 regions should have a columnar pattern of connections, with projection neurons found in layers II–III and V–VI and axons terminating in all layers of the respective cortices (Barbas 1986; Barbas and Rempel-Clower 1997). However, ACC pathways terminate to a greater extent in the middle-deep layers than in the upper layers of at least some pOFC areas (Barbas and Rempel-Clower 1997). This evidence suggests that the pattern of connections between ACC and pOFC may be due to subtle differences in the laminar architecture of areas within these cortical regions that cannot be detected with common architectonic methods.

ACC area 32 is of particular interest for its unusually dense pathways distributed to the rest of the prefrontal cortex (Barbas and Pandya 1989; Barbas et al. 1999, 2013). The connections and architectonic features suggest that area 32 may have an additional effector role in emotions by influencing the complex sensory-related integration within pOFC and its output to other cortical and subcortical structures. We investigated this issue by first labeling pathways from ACC area 32 to all pOFC areas in rhesus monkeys at the level of the entire system and the synapse. We then investigated whether sequential pathways from ACC area 32 to pOFC target layers that project to the amygdala and the thalamic mediodorsal (MD) nucleus. Using multiple approaches to label pathways and specific classes of inhibitory neurons, we provide evidence that area 32 has a strong excitatory influence on the output layers of pOFC, which project to the amygdala and the thalamus. The circuits between 2 structurally related regions appear to have complementary roles in the perception and expression of emotions, in processes that are disrupted in psychiatric disorders.

Materials and Methods

Experimental Design

The experimental design is based on placement of tracers in ACC, pOFC, amygdala, and thalamus of rhesus monkeys to label pathways linking these structures. This approach was combined with double and triple labeling to identify the relationship of pathways to postsynaptic excitatory and inhibitory neurons at the light and electron microscope.

Our first goal was to investigate the laminar pattern of ACC pathways to pOFC. We addressed this issue in 2 ways, as depicted in Figure 1A. We first mapped pathways from area 32 to pOFC at the level of the system and the synapse using anterograde tracers injected in area 32 (Fig. 1A, top). Then we mapped these pathways using retrograde or bidirectional tracers injected in pOFC areas (Fig. 1A, bottom). Our second goal was to map pathways in the reverse direction, from pOFC areas to ACC area 32 with injections of retrograde and bidirectional tracers in area 32 (Fig. 1B). Our third goal was to study whether serial pathways connect ACC area 32 with the amygdala and the thalamus via pOFC (Fig. 1C). We mapped these pathways using 2 strategies: first, we used bidirectional tracers injected in pOFC to study retrogradely labeled neurons in area 32 and anterograde label of axon terminals in the amygdala and the thalamic MD nucleus. Second, we injected retrograde tracers in the amygdala and the thalamus in cases that also had injections of anterograde tracers in area 32 on the same side, in order to study potential overlap of area 32 terminals and pOFC neurons directed to the amygdala and the thalamus.

Figure 1.

Experimental design. Injection sites are shown on maps of the medial (top) and the basal (bottom) surface of the rhesus monkey brain. Cortical areas and subcortical structures injected with tracers are shaded in grey: anterior cingulate cortex (ACC, area 32), posterior orbitofrontal cortex (pOFC), the amygdala, and the thalamus. (A) Pathways from area 32 to pOFC. Anterograde tracers were injected in ACC area 32 and retrograde or bidirectional tracers in pOFC areas to label pathways from area 32 to the posterior orbitofrontal cortex (pOFC: areas OPAll, OPro, and 13). (B) Pathways from pOFC to area 32. Retrograde and bidirectional tracers were injected in area 32 to label pathways from pOFC to area 32. (C) Serial pathways connecting area 32 and subcortical structures via pOFC were studied in 2 ways: First, bidirectional tracers in pOFC areas were used to label projection neurons retrogradely in area 32 directed to pOFC, and axon terminations in the amygdala and the thalamus. Second, in cases with anterograde tracer injections in area 32, retrograde tracers were injected in the amygdala and the thalamic MD to label axon terminals from area 32 in pOFC and projection neurons from pOFC to the amygdala and the thalamus. Numbers show prefrontal areas according to the map of Barbas and Pandya (1989). Amy, amygdala; BDA, biotinylated dextran amine; Cg, cingulate sulcus; DY, Diamidino Yellow; FB, Fast Blue; FE, Fluoro Emerald; FR, Fluoro-ruby; HRP-WGA, horseradish peroxidase conjugated to wheat germ agglutinin; LY, Lucifer yellow; MPAll, medial periallocortex (agranular); OLF, primary olfactory cortex; OPAll, orbital periallocortex (agranular); OPro, orbital proisocortex (dysgranular); Thal, thalamus. Injection sites for each case are also shown in Table 1.

Brain Imaging and Surgical Procedures

We conducted tract-tracing experiments on twelve normal young adult (2–5 years) rhesus monkeys (Macaca mulatta) of both sexes (Table 1). We studied the architecture of anterior cingulate and posterior orbital areas in another 6 cases (Table 2). Animals were obtained through Primate Centers and procedures were designed to minimize animal suffering and reduce the number of animals used. Detailed protocols of the procedures were approved by the Institutional Animal Care and Use Committee at Harvard Medical School and Boston University School of Medicine in accordance with NIH guidelines (DHEW Publication no. [NIH] 80–22, revised 1996, Office of Science and Health Reports, DRR/NIH, Bethesda, MD, USA).

Table 1.

Monkey data and type of tracing analysis

Case	Sex	Hemisphere	Injection site	Tracer injected	Tracing analysis
AE	–	L	32	HRP-WGA	Retrograde	pOFC	LM
AF	–	L	OPro	HRP-WGA	Retrograde	32	LM
					Anterograde	Amy & Thal
AG	F	R	OPAll/OPro	HRP-WGA	Retrograde	32	LM
					Anterograde	Amy & Thal
AJ	F	R	13	FB	Retrograde	32	FM
AK	F	R	32	DY	Retrograde	pOFC	FM
AL	–	R	OPro	DY	Retrograde	32	FM
		R	13	FB	Retrograde	32	FM
BC	M	R	OPro	FB	Retrograde	32	FM
		R	13	FE	Retrograde	32	FM
					Anterograde	Amy & Thal	LM
BG	F	R	32	BDA	Anterograde	pOFC	LM
		R	Thal	FB/DY	Retrograde	pOFC	FM
BI	F	R	32	BDA	Anterograde	pOFC	LM/EM
BL	M	R	32	LY	Anterograde	pOFC	LM
		R	Amy	FR	Retrograde	pOFC	LM
BN	M	L	32/24a	LY	Anterograde	pOFC	LM/EM

13, Orbital area 13; 32, anterior cingulate area 32; 24a, anterior cingulate area 24a; Amy, amygdala; BDA, biotinylated dextran amine; DY, Diamidino Yellow; EM, electron microscopy; F, female; FB, Fast Blue; FE, Fluoro Emerald; FM, fluorescence microscopy, FR, Fluoro-ruby; HRP-WGA, horseradish peroxidase conjugated with wheat germ agglutinin; L, left; LM, light microscopy; LY, Lucifer yellow; M, male; OPAll, orbital periallocortex; OPro, orbital proisocortex; pOFC, posterior orbitofrontal cortex (areas OPAll, OPro, and 13); R, right; Thal, thalamus. Note: Case AY with a small injection in area 32 (Fig. 1A, top) had sparse label in pOFC areas and was not included in the analysis.

Table 2.

Monkey data and type of architectonic analysis

Case	Sex	Hemisphere	Stain
AN	–	R	Nissl & SMI-32
AQ	–	L	Nissl & SMI-32
AR	–	L&R	Nissl
AS	F	L&R	Nissl
AT	F	L&R	Nissl
BB	F	L&R	Nissl & SMI-32

F, female; L, left; R, right; SMI-32, nonphosphorylated neurofilament protein.

In one group of animals we first conducted magnetic resonance imaging (MRI) to guide injection of tracers (cases AY, BG, BI, BL, & BN). After anesthesia with propofol (loading dose, 2.5–5 mg/kg, intravenous; continuous rate infusion, 0.25–0.4 mg/kg⁻¹ min⁻¹) animals were positioned in a nonmetallic stereotactic device for magnetic resonance imaging (MRI) with a 3T-superconducting magnet (Phillips or Siemens). The hollow ear bars of the stereotactic apparatus were filled with betadine salve to render them visible in the scan. For injection of tracers we calculated stereotactic coordinates in 3 dimensions for each injection site using the midline and the interaural line as reference. In another group of monkeys we placed tracers under direct visualization using sulcal landmarks (cases AE, AF, AG, AJ, AK, AL, and BC).

For surgery to inject neural tracers monkeys were sedated with ketamine hydrochloride (10–15 mg/kg, intramuscularly) and deeply anesthetized either with sodium pentobarbital (cumulative dose ~30 mg/kg, intravenous) or with isoflurane until a surgical level of anesthesia was achieved. The monkeys were then placed in the stereotactic apparatus and a small region of the cortex was exposed over each injection site. Surgery was performed under aseptic conditions. Heart rate, muscle tone, respiration, and pupillary dilatation were closely monitored to maintain a surgical level of anesthesia. Neural tracers were injected using a microsyringe (5 or 10 μL, Hamilton, Reno, NV, USA) mounted on a microdrive.

To label pathways linking ACC and pOFC we injected the tracers biotinylated dextran amine (BDA, 10% solution, volume of 6–10 μL, 10 kDa; Invitrogen, Carlsbad, CA, USA), Lucifer yellow (LY, dextran Lucifer yellow, anionic, lysine fixable, 10% solution, volume of 3–4 μL, 10 kDa, Invitrogen), horseradish peroxidase conjugated to wheat germ agglutinin (HRP-WGA, 8% solution, 0.05–0.1 μL, Sigma, St. Louis, MO, USA), Fast Blue (FB, 1–10% solution, volume of 0.25–4 μL; Sigma), Diamidino Yellow (DY, 3% solution, volume of 0.3–1.6 μL; Sigma) and Fluoro Emerald (FE, dextran fluorescein, 10% solution, 3–5 μL; 3 and 10 kDa; Invitrogen). Each tracer was injected in 2–4 penetrations. We left the needle in place for 10–15 min to allow tracer penetration at the injection site and prevent upward diffusion of the dye during retraction of the needle.

We also labeled projection neurons from pOFC directed to subcortical structures to investigate if they overlap with axon boutons from area 32 pathways. To this effect, we injected the tracer Fluoro-ruby (FR, dextran tetramethylrhodamine, 10% solution, volume of 3–4 μL, 3 and 10 kDa, Invitrogen) in the amygdala (case BL) and FB and DY in the thalamus (case BG). Subcortical injections were made on the same hemisphere as the area 32 injections of anterograde tracers.

Dextran amines (LY, FE, FR, and BDA) of 10 kDa molecular weight are optimal for anterograde tracing and label the entire extent of axon terminals and boutons. FE and FR of 3 kDa also act as retrograde tracers labeling cell bodies and proximal dendrites of projection neurons (Veenman et al. 1992; Richmond et al. 1994; Reiner et al. 2000). FB and DY are retrograde tracers (Kuypers et al. 1980; Cavada et al. 1984) and HRP-WGA is a bidirectional tracer (Gonatas et al. 1979; Trojanowski et al. 1981).

After injection of tracers the wound was closed in anatomic layers and the animal was monitored for postoperative recovery and given analgesics.

Perfusion and Tissue Processing

The time of perfusion after injection of tracers varied according to the type of tracer as follows: 48 h for HRP-WGA cases; 10–19 days for FB, DY, and FE cases; 19 days for LY, FR, and BDA cases.

For transcardial perfusion all animals were deeply anesthetized with a lethal dose of sodium pentobarbital (>50 mg/kg, intravenous, to effect). Animals with HRP-WGA injections were perfused with saline followed by 2 L of fixative solution (1.25% glutaraldehyde, 1% paraformaldehyde in PB 0.1 M at pH 7.4) followed by 2 L of cold PB (4 °C, 0.1 M at pH 7.4). The brain was removed from the skull and placed for 1 day in a solution of glycerol phosphate buffer [10% glycerol and 2% dimethyl sulfoxide (DMSO) in PB 0.1 M at pH 7.4] followed by 2 days in 20% glycerol phosphate buffer. Animals with injections of fluorescent tracers were transcardially perfused with 6% paraformaldehyde in PB (0.1 M at pH 7.4; cases AJ, AK, and AL) and postfixed in a solution of 6% paraformaldehyde with 10% glycerol phosphate buffer and 2% DMSO, as described above. In all other cases the fixative solution included 4% paraformaldehyde and 0.2% glutaraldehyde in PB (0.1 M at pH 7.4) for processing for electron microscopy.

After removal from the skull all brains were photographed, cryoprotected in a series of sucrose solutions (10–30% in PBS 0.01 M at pH 7.4) and frozen in −75 °C isopentane (Fisher Scientific, Pittsburg, PA, USA) for rapid and uniform freezing (Rosene et al. 1986). Brains were cut in the coronal plane on a freezing microtome at 40 or 50 μm to produce 10 matched series. For analysis of label with fluorescent tracers, 2 series of sections were mounted onto gelatin-coated slides, dried in the dark and stored at 4 °C. One of these series was coverslipped with Fluoromount (Fischer, Fair Lawn, NJ, USA) or Kristalon (EMD Millipore, Billerica, MA, USA) and stored at 4 °C. To preserve the ultrastructure, the rest of the tissue was stored at −20 °C in antifreeze solution (30% ethylene glycol, 30% glycerol, 40% PB 0.05 M at pH 7.4 with 0.05% azide) until processing.

In cases used for cytoarchitectonic analysis the animals were transcardially perfused with 2 L of 4% paraformaldehyde in cacodylate buffer (0.1 M at pH 7.4; cases AN, AQ, AR, and AS) or 4% paraformaldehyde in PBS (0.1 M at pH 7.4; cases AT and BB), cryoprotected, frozen (as above) and cut at 40 or 50 μm.

Stains to Study Cortical Architecture

Nissl stain: For delineating areal and laminar borders of ACC and pOFC areas we stained a series of sections for Nissl, using either thionin or cresyl violet (Table 2). Sections were mounted on chrome-alum coated slides, dried, defatted in a 1:1 solution of chloroform and 100% ethanol for 1 h, rehydrated through a series of graded alcohols and distilled water, stained with 0.05% thionin (pH 4.5) or 0.1% cresyl violet (pH 3.8) for 3 min and differentiated through graded alcohols and xylenes. Sections were coverslipped with mounting media (Permount, Fisher Scientific; or Entellan, Electron Microscopy Sciences, Hatfield, PA, USA).

SMI-32 stain: In 3 cases (AN, AQ, and BB) we immunolabeled a series of sections for SMI-32, an antibody for a nonphosphorylated intermediate neurofilament protein. In the primate cortex SMI-32 labels a subset of pyramidal projection neurons in layers III and V, and to a lesser extent in layers II and VI (Campbell and Morrison 1989). The density of SMI-32 positive neurons varies across areas and thus can be used as a sensitive architectonic marker.

Free-floating sections were rinsed in PBS (0.01 M at pH 7.4), incubated in glycine 0.05 M, preblocked in 10% normal goat serum (NGS) and 5% bovine serum albumin (BSA) with 0.2% Triton-X for 1 h in PBS and incubated overnight in primary antibody against SMI-32 (mouse monoclonal, Sternberger Monoclonals, Lutherville, MD, USA; diluted 1:5000 in PBS with 1% NGS, 1% BSA, 0.1% Triton-X). Sections were then rinsed in PBS and incubated for 3 h in secondary biotinylated goat anti-mouse IgG (Vector Laboratories, Burlingame, CA, USA; diluted 1:200 in PBS with 1% NGS, 1% BSA, 0.1% Triton-X), followed by 1 h in an avidin–biotin horseradish peroxidase complex (AB-HRP; Vectastain PK-6100 ABC Elite kit, Vector Laboratories; diluted 1:100 in PBS with 0.1% Triton X-100). Sections were rinsed and processed for 2–3 min for the peroxidase-catalyzed polymerization of diaminobenzidine (DAB; DAB kit, Vector Laboratories or Zymed Laboratories Inc., South San Francisco, CA, USA; 0.05% DAB, and 0.004% H₂O₂ in PBS).

Pathway Labeling

HRP-WGA: For viewing pathways labeled with HRP-WGA we processed a series of sections as described (Mesulam et al. 1980). Sections were mounted, dried, and counterstained with neutral red.

BDA: For viewing pathways labeled with BDA, free-floating sections were rinsed in PBS (0.01 M at pH 7.4), incubated in 0.05 M glycine, preblocked in 5% NGS and 5% BSA with 0.2% Triton-X in PBS and then processed for AB-HRP and DAB as described above.

LY, FR, and FE: To view pathways labeled with LY, FR, and FE, sections were processed to visualize the fluorescent tracer by peroxidase-catalyzed polymerization of DAB. Sections were incubated overnight in primary antibody against LY, FR, or FE (rabbit polyclonal, Molecular Probes, Eugene, OR, USA; diluted 1:800 in PBS with 1% NGS, 1% BSA, 0.1% Triton-X), followed by secondary biotinylated goat anti-rabbit IgG for 2 h (Vector Laboratories; diluted 1:200 in PBS with 1% NGS, 1% BSA, 0.1% Triton-X), followed by AB-HRP and DAB, as described above. Sections of cases having an injection of BDA were incubated in avidin–biotin blocking reagent (AB blocking kit, Vector Laboratories) before immunobinding to prevent cross-reaction with BDA, as described (Medalla et al. 2007). After BDA, LY, FR, and FE processing, sections were counterstained with thionin for cytoarchitectonic identification and laminar parcellation of ACC and pOFC areas.

Immunohistochemistry for Electron Microscopy

We used pre-embedding immunohistochemistry to view synapses formed by labeled boutons from area 32 pathways in pOFC, as described previously (Medalla and Barbas 2012; García-Cabezas and Barbas 2014). To study area 32 synapses on putative inhibitory neurons in pOFC we labeled sections for expression of the calcium-binding proteins parvalbumin (PV), calretinin (CR), and calbindin (CB), which label nonoverlapping neurochemical classes of inhibitory neurons in the primate cerebral cortex (DeFelipe 1997). Only a minority of excitatory cortical pyramidal neurons expresses CR or CB in the primate brain (DeFelipe et al. 1989; del Rio and DeFelipe 1997), which can be identified in the electron microscope (EM) by their spiny dendritic morphology (Peters et al. 1991; Fiala and Harris 1999).

We employed triple immunohistochemical methods to label the tracer and 2 calcium-binding proteins: BDA and LY were labeled with DAB, and PV, CR, or CB neurons were labeled either with silver-enhanced gold-conjugated secondary antibodies or tetramethylbenzidine (TMB), as described previously (García-Cabezas and Barbas 2014; Timbie and Barbas 2014) resulting in a distinct labeling in the EM: DAB appears as a dark uniform precipitate, silver-enhanced gold particles appear as circular clumps of variable size, and TMB appears as rod-shaped crystals (Gonchar and Burkhalter 2003; Pinto et al. 2003; Moore et al. 2004; Medalla and Barbas 2012; Zikopoulos and Barbas 2012; García-Cabezas and Barbas 2014; Timbie and Barbas 2015).

BDA and LY were processed using DAB as described above for optical microscopy but with reduced amounts of Triton-X (0.025%) to preserve the ultrastructure. Sections were then incubated in AB blocking reagent to prevent cross-reaction with TMB. We incubated BDA-labeled tissue sections overnight in 2 primary antibodies for PV or CR (1:2000, rabbit polyclonal, Swiss Antibodies, Bellinzona, Switzerland) and CB or CR (1:2000, mouse monoclonal, Swiss Antibodies). LY-labeled tissue sections were incubated in 2 mouse monoclonal primary antibodies for PV, CR, and/or CB (1:2000, mouse monoclonal, Swiss Antibodies), processed successively, using the Mouse-on-Mouse blocking kit (M.O.M. basic kit, Vector Laboratories) in between to prevent cross-reaction. After incubation with the primary antibody for each calcium-binding protein, sections were incubated overnight in the appropriate secondary gold-conjugated IgG (1:50, 1 nm gold particle diameter; Aurion, Wageneningen, The Netherlands) or in secondary biotinylated anti-mouse IgG (1:200, Vector Laboratories) followed by AB-HRP. Sections were postfixed in 6% glutaraldehyde with 2% paraformaldehyde using a variable wattage microwave oven (3–6 min at 150 W in the Biowave; Ted Pella, Redding, CA, USA) until the fixative reached 30ºC. Sections were intensified with silver (6–12 min; IntenSE M kit; GE Healthcare Life Sciences, Pittsburgh, PA, USA) and processed for TMB and stabilized with DAB-cobalt chloride solution, as described previously (Medalla and Barbas 2012; García-Cabezas and Barbas 2014). Primary antibodies were omitted in control experiments to test the specificity of secondary antibodies and we used the AB blocking reagent prior to AB binding or the M.O.M. kit prior to secondary antibody binding. In all control experiments there was no immunohistochemical labeling.

We mounted tissue sections on slides and quickly viewed them under the light microscope to capture images with label with a CCD camera. Small blocks of sections containing anterograde and PV-CR-CB label were cut under a dissecting microscope, postfixed in 1% osmium tetroxide with 1.5% potassium ferrocyanide in PB (0.1 M at pH 7.4), washed in PB and water, and dehydrated in an ascending series of alcohols (50–100%). While in 70% alcohol, blocs were stained for 30 min with 1% uranyl acetate (EM Sciences, Hatfield, PA, USA). Subsequently, blocks were infiltrated with propylene oxide and flat embedded in araldite at 60 °C. Pieces of the araldite-embedded sections were cut and re-embedded in resin blocks. We then cut serial ultrathin sections (50 nm) with a diamond knife (Diatome US, Hatfield, PA, USA) using an ultramicrotome (Ultracut UCT, Leica, Vienna, Austria). Ultrathin sections were collected on single slot pioloform-coated grids.

Mapping of Labeled Axons, Boutons, and Neurons

We studied the cytoarchitecture of ACC and pOFC in sections stained for Nissl and SMI-32 and in counterstained sections processed for pathway tracing to place areal and laminar boundaries as described previously (Barbas and Pandya 1989).

To identify laminar patterns of ACC pathways to pOFC we first studied the distribution of anterogradely labeled boutons from area 32 in pOFC areas (tracers BDA and LY) under brightfield and darkfield illumination (Olympus optical microscope, BX 60, and BX 51). We estimated the number of area 32 boutons in pOFC using the unbiased stereological method of the optical fractionator (Gundersen 1986; Howard and Reed 1998) with the aid of a semiautomated commercial system (StereoInvestigator; MicroBrightField, Williston, VT, USA).

We sampled systematically the entire volume of each laminar group in each pOFC area in each case to count labeled boutons. Our goal was to estimate the number of area 32 labeled terminals throughout laminar groups in entire areas that included densely and sparsely labeled sites. For each case and cortical area analyzed, we used systematic random sampling of evenly spaced brain sections throughout the extent of an area (3–6 sections for the smallest area OPAll, 1 section/0.5 mm, and up to 7 sections for the larger areas 13 and OPro, 1 section/mm). The number of sections used depended on the size of each area, which varies across individual animals. We estimated the number of labeled boutons from area 32 axons in superficial layers I–III, and middle-deep layers (IV–VI, here after called “deep”) of all pOFC areas. The stereological data included volume calculation for area and layer, which takes into consideration the sampled area and thickness of each section. The top and bottom of each section (minimum 2 μm for 15 μm sections after shrinkage) were used as guard zones. In BDA cases, boutons were counted using an optical disector restricted to the central fraction of the tissue thickness (11 μm). The penetration of the antibodies in LY cases did not extend to the entire thickness of the section, leaving a central band of tissue unstained; in these cases, the optical disector was restricted to the stained upper portion of the section (5 μm) with a guard zone of 2 μm. The actual mounted section thickness was measured by the program at each counting site. The counting frame/disector size (25 × 25 μm², height = 11 μm for BDA sections; 25 × 25 μm², height = 5 μm for LY sections) and grid spacing (250 × 250 μm²) were set to employ a fraction to yield a coefficient of error of <10% for each marker and each laminar group, as recommended (Gundersen 1986; Howard and Reed 1998). In 2 cases (BG, area 13; BL, area OPro) we doubled the number of sections studied through each area to test whether oversampling affects the findings.

Stereological analysis throughout the entire pOFC region yielded estimates of the total number of area 32 labeled boutons in individual pOFC areas and the laminar distribution within each area. For each area within the pOFC region we computed the proportion of area 32 boutons for each laminar group. We employed analysis of variance (ANOVA, IB SPSS Statistics 20 for Windows) for statistical comparisons of the proportion of labeled boutons by area and laminar group among cases in the pOFC areas. We used the same program to obtain the mean and standard error of each sample.

We also used an alternative method to investigate laminar patterns of connections from ACC to pOFC by study of the distribution of retrogradely labeled neurons in area 32 after injection of retrograde or bidirectional tracers (HRP-WGA, FB, DY, and FE) in pOFC areas. We examined every other section through the entire region in one series (1 in every 800 μm) under brightfield, darkfield, or fluorescence illumination. Plots of labeled neurons ipsilateral to the injection, drawings of cortical areas and layers and the site of blood vessels used as landmarks were transferred from the slides onto paper using a digital plotter (Hewlett Packard 7475 A, Palo Alto, CA, USA) electronically coupled to the stage of a microscope and to a personal computer, as described (Barbas et al. 1999). We followed the same procedure for the description of laminar patterns of pOFC to ACC pathways after injection of retrograde tracers in area 32.

We investigated whether there are potential serial pathways from area 32 to subcortical structures through pOFC using 2 strategies. First, in cases with injections of bidirectional tracers in pOFC areas (cases AF, AG, and BC, tracers HRP-WGA and FE) we mapped retrogradely labeled neurons in area 32 (as described above) and studied the distribution of anterogradely labeled axon terminals in the amygdala and the thalamus. We mapped area 32 pathways to the amygdala in precise register with respect to anatomic landmarks with an encoded microscope stage interfaced to a computer and commercial software (Neurolucida, MicroBrightField). We determined nuclear boundaries as previously described in the amygdala (Barbas and De Olmos 1990) and in the thalamus (Olszewski 1952).

Second, using a different strategy, we studied potential serial pathways through pOFC in cases with injections of anterograde tracers in area 32 and retrograde tracers in subcortical structures (cases BG in thalamus and BL in amygdala). We then mapped labeled neurons in pOFC directed to the amygdala and the thalamus, as well as anterograde terminals from area 32, as described above. FB fluoresces blue and labels the soma of neurons, while DY fluoresces yellow and labels only the nucleus (Kuypers et al. 1980; Cavada et al. 1984).

ACC Bouton Population Analysis

To describe the presynaptic features of ACC to pOFC pathways we acquired image stacks of several focal planes at 1000 × to study the size of boutons from area 32 pathways in pOFC, which is correlated with synaptic efficacy (Stevens 2003; Germuska et al. 2006; Zikopoulos and Barbas 2006). Image stacks were combined to create a composite image using ImageJ (NIH, USA) and scaled as described (Medalla and Barbas 2010). Composite images have high depth of field focused throughout the z-axis. Labeled boutons from area 32 pathways were traced manually using the open software program Reconstruct [www.bu.edu/neural; Fiala (2005)], and data were exported to a database in Excel. More than 30 000 boutons were measured for major diameter. We used k-means cluster analysis of the major bouton diameter to determine a cutoff point and separate labeled boutons into groups of large and small.

Ultrastructural Analysis

To investigate the termination of the area 32 pathway in pOFC at the synaptic level we identified 2 coronal sections per case (BI and BN) with the densest anterograde label through pOFC based on brightfield maps of the label, as described previously (Medalla et al. 2007; Bunce and Barbas 2011; García-Cabezas and Barbas 2014). We processed sections containing all the areas and layers of pOFC from matched levels in adjacent series for immunohistochemistry. Tissue segments with label were cut and processed for EM and flat embedded in resin (as described above). We cut two 250 μm-wide cortical columns including layers I–VI of one pOFC area (OPro) of 2 resin-embedded sections per case (BI and BN). These sections were re-embedded in araldite blocks, trimmed with a diamond trim tool (Diatome US, Hatfield) and then cut (Ultracut UCT, Leica, Vienna, Austria) into serial ultrathin sections (50 nm) with a diamond knife and collected on single-slot pioloform-coated grids.

We viewed labeled boutons and synapses in the electron microscope (100CX, JEOL, Peabody, MA, USA) in layer I, superficial layers II–III, and deep layers IV–VI of pOFC area OPro. We used classical criteria to determine the type of synapse and the postsynaptic profile of labeled boutons (Peters et al. 1991) as described previously for other areas (Medalla et al. 2007; Bunce and Barbas 2011; Medalla and Barbas 2012; García-Cabezas and Barbas 2014). In the cortex boutons form synapses with spines, which are enriched on excitatory pyramidal neuron dendrites, or with aspiny (smooth) or sparsely spiny dendritic shafts, which are characteristic of cortical inhibitory neurons (Peters et al. 1991; Fiala and Harris 1999). For 2D analysis we sampled and photographed exhaustively all labeled boutons from series of 10–30 consecutive sections and followed each bouton through a minimum of 10 adjacent sections for each synapse (BI, n = 159; BN, n = 57). We measured the major diameter of labeled boutons at the level of the synapse using Reconstruct [www.bu.edu/neural; Fiala (2005)]. We also followed labeled boutons in serial sections through synapses with presumed inhibitory postsynaptic targets and reconstructed the synapse in 3 dimensions (case BI, n = 5; case BN, n = 4). Object contours of ACC boutons and their postsynaptic elements were manually traced section-by-section. We calibrated section thickness through measurements of the diameter of mitochondria, which yields the same estimate of section thickness, as the method of minimal folds (Fiala and Harris 2001b; Fiala 2005). Postsynaptic profiles were considered as belonging to presumed inhibitory neurons if they were labeled for any of the calcium-binding proteins (PV, CR, or CB) and/or were part of aspiny or sparsely spiny dendrites. We characterized dendrites by computing a density index for spines (number of spines/μm) and synapses (number of synapses/μm) of reconstructed dendrites as described previously (Fiala and Harris 2001a; Medalla and Barbas 2010; García-Cabezas and Barbas 2014).

Photography

We photographed areas of ACC and pOFC from Nissl and SMI-32 stained sections and labeled boutons, axons, and neurons in ACC, pOFC, the amygdala, and the thalamus using an optical microscope (Olympus BX 51) with a CCD camera (Olympus DP70) connected to a personal computer with a commercial imaging system (DP Controller). We acquired pictures of a large part of coronal sections at low magnification using commercial software (StereoInvestigator, MicroBrightField). At the EM, labeled boutons from area 32 axons and their postsynaptic profiles were captured using a digital camera (ES1000W, Gatan, Pleasanton, CA, USA) at a magnification of 10 000–30 000×. Images were imported into Adobe Illustrator CC software (Adobe Systems Incorporated, San José, CA, USA) to assemble in figures. Minor adjustment of overall brightness and contrast were made but images were not retouched.

Results

ACC and pOFC Areas Show Progressive Laminar Elaboration

To investigate if connection patterns are associated with fine differences in laminar architecture we first identified ACC and pOFC areas and layers in Nissl stained sections. Areas within the pOFC show gradual laminar elaboration from the primary olfactory cortex (Barbas and Pandya 1989). These areas include area OPAll (Fig. 2A,B), which lacks a granular layer IV (agranular), area OPro (Fig. 2C,D) characterized by the appearance of a discontinuous and rudimentary layer IV (dysgranular), and area 13 which shows a slightly better defined layer IV than area OPro (Fig. 2E,F).

Figure 2.

Cytoarchitectonic areas in the posterior orbitofrontal cortex (pOFC). (A) Coronal section through the caudal extent of pOFC shows agranular area OPAll and dysgranular area OPro. (B) Area OPAll, lacks layer IV, as shown at higher magnification. (C) Section anterior to that depicted in (A) shows dysgranular area OPro which has a rudimentary layer IV. (D) Area OPro seen at higher magnification. (E) In the most anterior part of pOFC, area 13 has somewhat better developed layer IV but the area is still dysgranular. (F) area 13 seen at higher magnification. Calibration bar in (E) also applies to (A) and (C). Calibration bar in (F) applies to (B), (D), and (F). All, allocortex; AON, anterior olfactory nucleus; Cd, caudate nucleus; Cl, claustrum; I, layer I; II–III, superficial layers II–III; IV–VI, middle-deep layers in dysgranular areas; a rudimentary layer IV in dysgranular cortex is grouped with layers V and VI; LOT, lateral olfactory tract; MPAll, medial periallocortex; Olf ventricle, olfactory ventricle; OPAll, orbital periallocortex; OPro, orbital proisocortex; V–VI, deep layers V–VI; WM, white matter; 13, area 13; 24a, area 24a; 25, area 25.

ACC areas also show gradual laminar differentiation from an area adjacent to the anterior extension of the hippocampal formation around the corpus callosum (Fig. 3A). The medial periallocortex and the most posterior part of area 24a lack layer IV (Fig. 3B–D). Area 32 lies in front of area 24a (Fig. 3E,F) and is dysgranular with a rudimentary layer IV (Fig. 3G) that is better defined in its most anterior extent (Fig. 3H).

Figure 3.

Cytoarchitectonic areas of the anterior cingulate cortex (ACC). Frames (C, D, G, and H) show the ACC areas of interest at higher magnification (A, B) Coronal sections through the medial prefrontal cortex at the anterior tip of the corpus callosum (A) and anterior to it (B). (C) ACC area MPAll lacks layer IV. (D) At an anterior level, area 24a occupies the central part of the medial prefrontal cortex and shows agranular laminar structure. (E, F) Coronal sections through the medial prefrontal cortex show the posterior (E) and the anterior (F) parts of area 32. (G) The posterior part of area 32 has a rudimentary layer IV. (H) At the anterior part of area 32 layer IV is better delineated. Calibration bar in (F) also applies to (A, B, and E). Calibration bar in H also applies to (C, D, and G). All, allocortex; Cc; corpus callosum; MPAll, medial periallocortex; WM, white matter; 24a, area 24a; 24b, area 24b; 25, area 25; 32, area 32. Roman numerals refer to layers or laminar groups. A rudimentary layer IV in dysgranular cortex is grouped with layers V and VI.

The architectonic differences in pOFC areas appear subtle in Nissl stained sections, but are evident in tissue labeled for SMI-32, revealing that most labeled neurons in pOFC areas OPAll and OPro are found in the deep layers V–VI (Fig. 4A). Labeled neurons in the superficial layers (II–III) were sparse in area OPAll (Fig. 4B). In area OPro, SMI-32 neurons were more numerous in the deep layers (V–VI) than in the superficial layers, where a rudimentary layer IV could be delineated (Fig. 4C, light zone). In area 13, SMI-32 neurons were more abundant than in areas OPAll and OPro and delineated a thin but continuous layer IV (Fig. 4D,E).

Figure 4.

Laminar changes in areas of the posterior orbitofrontal cortex (pOFC) labeled with SMI-32. (A) Coronal section through the caudal extent of pOFC shows agranular area OPAll and dysgranular area OPro. (B) SMI-32 immunhistochemistry labels pyramidal neurons in layers V–VI in area OPAll, as shown at higher magnification. (C) In area OPro, SMI-32 labels pyramidal neurons in the deep layers (V–VI); some pyramids are also labeled in layer III. (D) Section anterior to that depicted in (A) shows dysgranular area 13. (E) At higher magnification, SMI-32 label in layers V and III of area 13 leaves a band of unstained tissue (white arrowheads in D and E) corresponding to a better developed layer IV compared with OPro. Calibration bar in (D) also applies to (A). Calibration bar in (E) applies to B, C, and E). All, allocortex; AON, anterior olfactory nucleus; Cd, caudate nucleus; Cl, claustrum; LOT, lateral olfactory tract; MPAll, medial periallocortex; OPAll, orbital periallocortex; OPro, orbital proisocortex; WM, white matter; 13, area 13; 24a, area 24a; 25, area 25. Roman numerals refer to layers or laminar groups. A rudimentary layer IV in dysgranular cortex is grouped with layers V and VI.

Injection Sites Encompassed Several Regions of ACC, pOFC, Amygdala, and Thalamus

The experimental design and tracer injection sites are shown in Figure 1. Figure 1A shows 2 ways used to study pathways from area 32 to pOFC: first, by tracing the pathway from area 32 (Fig. 1A, top) to their termination in pOFC, and second, by mapping projection neurons in area 32 after injection of retrograde tracers in pOFC (Fig. 1A, bottom). In cases depicted in Figure 1A, top (to map pathways from area 32 to pOFC), an anterograde tracer was restricted to area 32 in 2 cases (BI and BL), in one case (BG) the tracer impinged on the medial part of adjacent area 9 and in another case (BN) the tracer was within the posterior part of area 32 and spread posteriorly into area 24a. In a fifth case (AY) a small injection of tracer was in an anterior part of area 32. In 6 cases with injections in pOFC areas (to map labeled neurons in ACC) the retrograde or bidirectional tracer was restricted to the orbital proisocortex (area OPro) or to area 13 (Fig. 1A, bottom; cases AF and AJ; cases AL and BC, 2 distinct tracers in each) and in  1 case the tracer involved areas OPAll and OPro (case AG, bidirectional tracer). All injection sites involved superficial and deep layers of area 32 and pOFC areas, with the exception of one injection in area 13 in which a bidirectional tracer was restricted to the deep layers (V–VI, case BC).

We also studied the pathway in the reverse direction, from pOFC to area 32 after injection of retrograde tracers in area 32, as depicted in Figure 1B. In one case, a retrograde tracer was restricted to area 32 (AK) and a bidirectional tracer in another case impinged on the dorsal part of area 14 (case AE). Both injection sites involved superficial and deep layers of area 32.

We used 2 distinct approaches to investigate whether there are serial pathways from area 32 to pOFC and from pOFC to the amygdala and the thalamic MD, as depicted in Figure 1C. In the first approach we used bidirectional tracers injected in pOFC areas, as described above (cases AF, AG, and BC, Fig. 1C, bottom) to label neurons retrogradely in area 32, and anterograde pathways in the amygdala and MD. In a second approach we injected anterograde tracers in area 32, as described above (cases BG and BL) and bidirectional or retrograde tracers in the amygdala and the thalamic MD of the same hemisphere (Fig. 1C, top). This approach made it possible to study whether pathways from area 32 overlapped in pOFC areas with projection neurons directed to the amygdala and the thalamus. In one of these cases (BL) the bidirectional tracer (FR) was in the posterior part of the cortical nucleus of the amygdala. In another case (BG), retrograde tracers were in the central portion of the parvicellular part of the thalamic MD nucleus (tracer FB), in magnocellular MD and in the adjacent parafascicular nucleus (tracer DY).

Laminar Patterns of ACC Pathways Varied Across Distinct pOFC Areas

We first analyzed quantitatively the laminar pattern of pathways from ACC area 32 that terminate in pOFC (Fig. 1A) with the use of unbiased stereologic methods (cases BG, BI, BL, and BN). In one case (AY) labeling in pOFC areas was sparse and not sufficient for quantitative analyses. Analyses of anterograde label were based on the 4 other cases that showed robust labeling in pOFC areas. The rudimentary and discontinuous layer IV in area OPro and area 13 was included with the deep layers V–VI. In each pOFC area, we expressed the laminar pattern of area 32 pathways as the percentage of labeled boutons in superficial (I–III) and deep layers (IV–VI). As shown in Figure 5A, axon boutons from area 32 were found mostly in the middle-deep layers of area OPAll (77%, se ± 3) and of area OPro (61%, SE ± 6). In area 13, axon boutons were distributed equally in the superficial (52%, standard error [SE] ± 7) and deep layers (48%, SE ± 7; Fig. 5A), revealing statistically significant differences among these areas (ANOVA, P < 0.05). These patterns are illustrated in photomicrographs in Figure 5D–F. In posterior levels of the pOFC (agranular area OPAll and dysgranular area OPro), light or moderate axon terminations from area 32 were found mostly in the middle-deep layers (Fig. 5D). There was also label in layer I of area OPro (Fig. 5D). At more anterior levels of dysgranular area OPro there were dense patches of terminals in the deep layers and moderate terminations in the upper layers, but overall the density of terminals was lower in the superficial than in the deep layers (Fig. 5A,E). In the more anteriorly situated pOFC area 13, area 32 pathways terminated in a columnar arrangement, but patches of axon terminals were denser in the superficial layers I–III than in the more caudally situated pOFC areas (Fig. 5A,F). Overall, the quantitative distribution of label was comparable in the upper and deep layers of area 13 (Fig. 5A).

Figure 5.

Laminar patterns of pathways connecting area 32 and pOFC. (A) Terminations of area 32 pathways in pOFC: normalized data show proportion of area 32 axon boutons in laminar groups of pOFC areas estimated using unbiased stereological methods. The proportion of boutons in the upper layers (I–III) is progressively higher in dysgranular areas OPro and 13, which have a rudimentary layer IV, than in agranular area OPAll. The reverse pattern is seen for the middle-deep layers (IV–VI). Note that in area 13, the proportion of area 32 axon terminals in the superficial and middle-deep layers shows a columnar pattern. (B) Area 32 pathways to pOFC areas originate mostly from neurons in the superficial layers (II–III). Note the higher proportion of neurons in the superficial layers II–III in area 32 directed to agranular area OPAll than to dysgranular areas OPro and 13. (C) pOFC pathways to area 32 originate mostly in the deep layers (V–VI). These findings are consistent with the structural model. (D–F) Examples of laminar patterns of ACC pathways to pOFC; white dotted lines delineate laminar boundaries; solid white lines mark areal boundaries. (D) Darkfield photomicrograph through the posterior pOFC region shows light anterograde (BDA) label in the middle-deep layers of areas OPAll and moderate label in area OPro (white arrowheads); there is also labeling in layer I of OPro (white arrow; case BG). (E) Darkfield photomicrograph shows columnar anterograde BDA label in area OPro with dense patches of labeled axons in the middle-deep layers IV–VI (white arrowheads) and moderate label in superficial layers I–III (white arrows; case BI). (F) Darkfield photomicrograph through anterior area 13 shows dense patches of BDA label in superficial layers I–III (white arrows) and moderate labeling in 2 patches in the deep layers (white arrowheads; case BI). Calibration bar in (F) applies to (D–F). Cd, caudate nucleus; Cl, claustrum; OPAll, orbital periallocortex (agranular); OPro, orbital proisocortex (dysgranular); 13, area 13 (dysgranular). Roman numerals refer to layers or laminar groups. A rudimentary layer IV in dysgranular cortex is grouped with layers V and VI. Vertical lines on bars show the standard error.

In 2 cases (BG and BL) in which we sampled 5 sections in one large pOFC area and repeated the analysis with 9 sections, the estimates of labeled boutons in the laminar groups were comparable (case BG, area 13, % in upper layers: sampling from 5 sections, interval of 1 section/mm, 55%, Gundersen error, 0.06; sampling from 9 sections, interval of 1 section/0.5 mm: 56%, Gundersen error, 0.04; case BL, area OPro, % in upper layers: sampling from 5 sections, interval of 1 section/mm, 33%, Gundersen error, 0.06; sampling from 9 sections, interval of 1 section/0.5 mm, 38%, Gundersen error, 0.04). These findings provide evidence that sampling additional sections or sampling sites after an error below 10% is achieved does not change the results significantly, consistent with the principles of stereology (Howard and Reed 1998).

We then used an alternative method to study the laminar pattern of pathways from area 32 to pOFC with the aid of retrograde tracers injected in pOFC areas (Fig. 1A, bottom). We expressed the laminar pattern of area 32 labeled neurons directed to pOFC as the percentage of labeled neurons in the superficial (II–III) and deep layers (V–VI). As shown in Figure 5B, projection neurons in area 32 directed to agranular area OPAll and dysgranular area OPro were found mostly in the superficial layers (II–III; OPAll/OPro: 91%; OPro: 77%, SE ± 3). In contrast, projection neurons in area 32 directed to area 13 were equally distributed in superficial (51%, SE ± 3) and deep layers (49%, SE ± 3; Fig. 5B). The pattern of label in Figure 5B mirrors the laminar pattern of anterograde label seen in Figure 5A, as predicted by the structural model for corticocortical connections.

We then studied the reverse pathway, from pOFC to area 32 (Fig. 1B) and confirmed that projection neurons in area OPAll were found mostly in the deep layers V–VI (83%, SE ± 8, Fig. 5C). Projection neurons in areas OPro and 13 directed to area 32 were also found in the deep layers (V–VI), but in lower proportions than in agranular area OPAll (OPro: 64%, SE ± 7; area 13: 63%, SE ± 5; Fig. 5C). The difference in projection neurons between OPro and adjacent area 13 was very small, possibly due to the small number of cases in the study of this pathway (n = 2), or biological variation. In aggregate, these findings show a consistent relationship in the laminar pattern of connections between area 32 and pOFC based on the differences in their laminar structure.

Size of Area 32 Terminals in pOFC: Segregation Into Clusters of Large and Small Boutons

We then measured at the light microscope the size of labeled boutons from area 32 axons in all layers of pOFC (areas OPAll, OPro, and 13; Fig. 6A). The rationale is based on evidence that bouton size is strongly correlated with synaptic efficacy (Stevens 2003). Labeled axons from area 32 formed both terminaux and en passant boutons. From columns with anterograde label, we traced exhaustively a minimum of 1000 boutons per case and for each pOFC area (n = 4 cases; n = >30 000 boutons). The mean major diameter of labeled boutons from area 32 across cases was ~1 μm (OPAll = 0.99 μm, SE ± 0.07; OPro = 0.96 μm, SE ± 0.08; area 13 = 0.98 μm, SE ± 0.06). Figure 6B shows the distribution of labeled boutons by size across cases in all pOFC areas (BG, BI, BL, and BN; ± SE). Labeled boutons in pOFC were segregated into clusters of small and large in each case by k-means cluster analysis. The mean cutoff value across cases was ~1.1 μm and the mean small and large cluster center across cases were respectively ~0.8 μm and ~1.4 μm. The proportion of large boutons across cases in the 3 pOFC areas was ~30% and did not differ significantly among areas (ANOVA, P > 0.05). These findings show that significant proportions of boutons from area 32 were large and potentially synaptically efficient, as investigated at the synapse below.

Figure 6.

Presynaptic features of area 32 terminals in pOFC. (A) Labeled boutons were traced manually using a semiautomated system. (B) Distribution of labeled boutons from area 32 in pOFC plotted by size of the major diameter. Bars show the mean frequency in 4 cases analyzed (BG, BI, BL, and BN). Vertical lines on bars show the standard error.

Boutons From ACC Axons in pOFC Formed Synapses Mostly With Putative Excitatory Neurons

We then studied the synaptic relationship of axon terminals from area 32 in pOFC. For this analysis we used columns of tissue spanning all layers of area OPro (cases BI and BN: 2 sections per case, 2 columns per section). We followed boutons from area 32 through a minimum of 10 adjacent sections in an uninterrupted series to identify the type of synapse and postsynaptic profile (n = 216; case BI, n = 159 boutons; case BN, n = 57).

The major diameter of area 32 boutons in area OPro was 0.91 μm in one case (BI) and 0.82 μm in another case (BN). There was a trend for area 32 boutons to be smaller in the superficial layers (layer I: case BI = 0.81 μm, case BN = 0.80 μm; layers II–III: case BI = 0.89 μm, case BN = 0.80 μm) than in the deep layers (layers IV–VI: case BI = 0.94 μm, case BN = 0.87 μm), but it was not statistically significant (ANOVA, P > 0.05). Using the cutoff value calculated for light microscopy analysis (1.1 μm), the proportion of small boutons from area 32 was 78% (case BI) and 84% (case BN) and the rest were large.

All boutons formed asymmetric synapses (Fig. 7), characteristic of excitatory neurons that use glutamate as neurotransmitter (Peters et al. 1991). Some area 32 boutons formed perforated synapses (case BI, n = 47, 30%; case BN, n = 9, 16%, Fig. 7A,C) which have segmented postsynaptic densities and are more efficacious than non-perforated synapses (Greenough et al. 1978; Sirevaag and Greenough 1985; Geinisman et al. 1987; Ganeshina et al. 2004). There were more boutons with perforated synapses in the deep layers IV–VI (case BI, n = 31, 33%; case BN, n = 4, 24%) than in layer I (case BI, n = 3, 19%; case BN, n = 3, 10%) and superficial layers II–III (case BI, n = 13, 26%; case BN, n = 2, 22%). Boutons with perforated synapses were significantly larger (case BI: 1.18 μm; case BN: 1.08 μm) than boutons with non-perforated synapses (case BI: 0.8 μm; case BN: 0.77 μm; two-tailed t-test, P < 0.05).

Figure 7.

The fine structure of synapses from area 32 boutons with postsynaptic elements in pOFC area OPro. (A) Most boutons from area 32 axons (At) in area OPro formed synapses with 1 spine (sp); the synapse is perforated. (B) Synapses of 1 bouton with 2 spines. (C) Synapse of area 32 bouton with 1 spine and 1 dendritic shaft (Den-CB) which is labeled with TMB for CB (rod-shaped crystals, black arrow); the synapse on the spine is perforated. (D) Synapse with 1 dendritic shaft (Den-CR) labeled with gold for CR (black dots). Arrowheads point to synapses. Calibration bar in (D) also applies to (A–C). At: axon terminal; Den-CB: dendrite labeled for calbindin; Den-CR: dendrite labeled for calretinin; sp: spine.

Area 32 boutons formed synapses with 1 spine (Fig. 7A), 2 or 3 spines (Fig. 7B), 1 spine and 1 dendritic shaft (Fig. 7C), or 1 dendritic shaft (Fig. 7D). In the cerebral cortex spines are found on dendrites of pyramidal and stellate neurons, which are excitatory (Peters et al. 1991). In area OPro spines were the postsynaptic profile in >90% of area 32 labeled boutons (Fig. 8A). Single spines were the major target of boutons from area 32 across layers (76–100%; Fig. 8B–D), while a minority formed synapses with 2 or more spines in all layers (2–12%, Fig. 8B–D). This evidence suggests that area 32 boutons in pOFC form synapses overwhelmingly with putative excitatory neurons.

Figure 8.

Proportion of area 32 labeled boutons in area OPro with distinct postsynaptic targets. Cases: BI and BN. (A–D) Boutons formed synapses mostly with 1 spine in all layers of area OPro. Boutons from area 32 axons formed synapses on dendritic shafts of presumed inhibitory neurons in fewer than 5% of the total.

A Minority of Boutons From ACC Axons in pOFC Formed Synapses With Inhibitory Neurons

Only a few boutons from area 32 formed synapses with dendritic shafts of putative inhibitory neurons in area OPro (n = 9 total; case BI, n = 5, ~3%; case BN, n = 4, 7%; Fig. 8A). These included boutons that formed synapses with 1 spine and 1 dendritic shaft in layer I (case BI, n = 2, 13%; case BN, n = 1, 3%; Fig. 8B) and in the deep layers IV–VI (case BI, n=2, 2%; Fig. 8D), or only on 1 dendritic shaft in layer I (case BN, n = 1, 3%; Fig. 8B), in layers II–III (case BI, n = 1, 2%; Fig. 8C), and in the deep layers IV–VI (case BN, n = 2, 12%; Fig. 8D). In the cerebral cortex synapses on dendritic shafts are most common on inhibitory neurons (Peters et al. 1991) that are either nonspiny or sparsely spiny (Peters et al. 1991; Fiala and Harris 1999).

We also used calcium-binding proteins to identify the neurochemical type of inhibitory dendrites (DeFelipe 1997). Of the total number of boutons that formed synapses on dendritic shafts (n = 9) 5 were labeled for 1 of 3 calcium-binding proteins (PV: n = 1, deep layers; CR: n = 1, superficial layers; CB: n = 3, deep layers and layer I) and 4 were not labeled (n = 2, layer I; n = 2, deep layers). To further characterize these dendritic shafts we computed a density index for spines (number of spines/μm) and synapses (number of synapses/μm) of reconstructed dendrites as described previously (Fiala and Harris 2001a; García-Cabezas and Barbas 2014). PV-, CR-, and nonlabeled dendritic shafts were smooth and had ~2 synapses/μm. CB positive shafts were sparsely spiny (0.8–2 spines/μm) and the dendritic segments had fewer overall synapses (0.5–1.3 synapses/μm). In the cerebral cortex sparsely spiny dendritic segments may be part of the proximal dendritic segments of excitatory neurons. The spine density of sparsely spiny dendrites overlaps with the density of spiny dendrites, but inhibitory neurons have a significantly higher density of synapses on shafts, which are virtually absent on spiny dendrites of excitatory neurons [Medalla and Barbas (2009, 2010), for a review see Peters et al. (1991), Fiala and Harris (1999)]. We concluded that the PV-, CR-, and nonlabeled dendritic shafts in pOFC that were innervated by area 32 boutons (n = 6) were elements of putative inhibitory neurons. In contrast, CB positive dendritic shafts (n = 3) were sparsely spiny and had low synapse indices, suggesting that they may be elements of putative excitatory neurons.

Serial Pathways From ACC are Connected With pOFC, the Amygdala, and the Thalamus

The above analyses showed that strong pathways from area 32 project to the deep layers of pOFC, which are the major output layers to subcortical structures, including the amygdala and the thalamus (Giguere and Goldman-Rakic 1988; Ghashghaei et al. 2007). We found evidence that these pathways are serial after injections of bidirectional tracers in pOFC areas (cases AF, AG, and BC, Fig. 1C, bottom). In one case (BC) in which the tracer injection was restricted to the deep (output) layers (V–VI) of area 13 (Fig. 9A, arrows) labeled neurons were found throughout area 32 (Fig. 9B). In the same case, axons from the deep layers of area 13 terminated in several nuclei of the amygdala and were densest in the basolateral nucleus (Fig. 9C,D). In the same case there were dense patches of axon terminals in the thalamic magnocellular MD intermingled with labeled neurons directed to area 13 (Fig. 9E,F). Serial pathways were also evident in the other 2 cases with injections of bidirectional tracers in areas OPAll/OPro and OPro involving both superficial and deep layers (cases AG and AF; Fig. 10A). This analysis revealed that area 32 projects to pOFC (Fig. 10B), and pOFC projects to the amygdala (Fig. 10C) and to the thalamic MD (Fig. 10D). In these cases, area 32 neurons directed to areas OPAll and OPro were found mostly in the superficial layers II–III (Fig. 10B). Patches of axon terminals in the amygdala were found in the inhibitory intercalated masses (IM) and the basolateral (BL) nucleus (Fig. 10C). In the thalamic magnocellular MD nucleus axon terminals were intermingled with neurons projecting to pOFC (Fig. 10D).

Figure 9.

Serial pathways connect area 32 with the amygdala and the thalamus through the deep layers of pOFC. (A) Fluorescent photomicrograph of a coronal section through the posterior orbitofrontal cortex shows an injection site (tracer FE) in area 13 (white arrows) restricted to the deep layers (case BC); dashed line marks the boundary of the cortex with the white matter. (B) Fluorescent photomicrograph through area 32 shows labeled neurons in layers II–III (white horizontal arrows) and in the deep layers (V–VI, white vertical arrows) that project to the deep layers of area 13. (C) Pathways originating in the deep layers (V–VI) of area 13 were distributed throughout several amygdalar nuclei and were denser in the basolateral nucleus (blue lines); there were also retrogradely labeled neurons in the amygdala (yellow dots) directed to area 13. (D) Darkfield photomicrograph at higher magnification of an area plotted in C shows dense labeling in the basolateral (BL) nucleus of the amygdala. Horizontal arrows point to pathways from area 13; vertical arrows point to projection neurons in the amygdala directed to area 13. (E) Darkfield photomicrograph of a coronal section through the thalamus shows dense pathways from area 13 as they terminate in the magnocellular part of the thalamic mediodorsal nucleus (MDmc). (F) Darkfield photomicrograph at higher magnification shows dense axon terminals from area 13 (white horizontal arrows) and thalamic neurons directed to area 13 (white vertical arrows). AON, anterior olfactory nucleus; BM, basomedial nucleus; BL, basolateral nucleus; Cc, corpus callosum; Ce, central nucleus; FE, Fluoro Emerald; L, lateral nucleus; LOT, lateral olfactory tract; MDmc, mediodorsal nucleus magnocellular part, Me, medial nucleus; PCo, posterior cortical nucleus; Pv, paraventricular nucleus; Sm, stria medullaris; WM, white matter; 13, area 13; 32, area 32. Roman numerals refer to layers or laminar groups.

Figure 10.

Serial pathways connect area 32 with the amygdala and the thalamus through pOFC. (A) Darkfield photomicrograph of a coronal section through the pOFC shows an injection site in areas OPAll and OPro involving the deep and superficial layers (case AG, tracer HRP-WGA); dashed line marks the boundary of the cortex with the white matter. (B) Darkfield photomicrograph through area 32 shows that projection neurons directed to areas OPAll and OPro originate mostly in superficial layers (II–III, white horizontal arrows) with fewer neurons found in the deep layers (V–VI, white vertical arrows). (C) Darkfield photomicrograph of a coronal section through the amygdala shows dense pathways from the same injection site in area OPAll/OPro that terminate in the intercalated masses (IM) and the basolateral (BL) nucleus of the amygdala (white vertical arrows); white horizontal arrows point to amygdalar neurons directed to OPAll and OPro which originate in the basomedial (BM, also known as accessory basal) and the lateral (L) nuclei. (D) Darkfield photomicrograph of a coronal section through the thalamus shows dense pathways from the same injection site in area OPAll/OPro terminating in the magnocellular part of the mediodorsal nucleus (MDmc, white horizontal arrows). White vertical arrows point to thalamic neurons directed to OPAll/OPro. Calibration bar in (C) applies to (A and C). AON, anterior olfactory nucleus; BM, basomedial nucleus; BL, basolateral nucleus; Cd, caudate nucleus; Ce, central nucleus; Cl, claustrum; IM, intercalated masses; L, lateral nucleus; LOT, lateral olfactory tract; MDmc, mediodorsal nucleus magnocellular part, Me, medial nucleus; OPAll, orbital periallocortex; OPro, orbital proisocortex; PCo, posterior cortical nucleus; Pv, paraventricular nucleus; Sm, stria medullaris; WM, white matter; 32, area 32. Roman numerals refer to layers or laminar groups.

We also used a different approach to map serial pathways from area 32 to pOFC and from pOFC to the amygdala and the thalamus, as depicted in Figure 1C. This was accomplished after injection of a retrograde tracer in the amygdala and an anterograde tracer in area 32 (case BL), and after injection of 2 retrograde tracers in the thalamus (case BG) and an anterograde tracer in area 32 on the same side. In these cases we mapped labeled neurons in pOFC that project to the amygdala or to the thalamus, and the termination of axons from area 32 in pOFC areas in adjacent sections. Maps of retrogradely labeled neurons projecting to subcortical structures were superimposed on photomicrographs (case BG) or drawings (case BL) of labeled terminations from area 32 (Fig. 11). Many of the pOFC neurons projecting to the amygdala (Fig. 11A–C) and to thalamic nuclei (Fig. 11D) in the deep layers of areas OPAll, OPro, and 13 were intermingled with axon terminals from area 32 pathways distributed in the same layers.

Figure 11.

Overlap of area 32 axon terminals with projection neurons in pOFC directed to subcortical structures. (A–C) Neurons labeled with FR in pOFC areas that project to the amygdala (red dots; case BL) overlapped with LY-labeled terminals from area 32 axons (green). (D) Neurons labeled with FB (blue squares) or DY (yellow squares) in the deep layers of pOFC that project to the thalamus overlap with axon terminations from area 32 (case BG). White arrows in area OPro show sites of overlap of neurons projecting to the thalamus, superimposed on a photomicrograph showing moderate to dense label of axon terminals from area 32; white arrowheads show overlap in area OPAll, where the anterograde label is light. Calibration bar in (D) applies to all panels. AON, anterior olfactory nucleus; BDA, biotinylated dextran amine; Cd, caudate nucleus; Cl, claustrum; DY, Diamidino Yellow; FB, Fast Blue; LOT, lateral olfactory tract; LY, Lucifer yellow; Olf ventricle, olfactory ventricle; OPAll, orbital periallocortex (agranular); OPro, orbital proisocortex (dysgranular); 13, area 13, 25: area 25. Roman numerals refer to layers or laminar groups. A rudimentary layer IV is grouped with layers V and VI.

Discussion

The posterior medial (ACC) and orbital (pOFC) regions of the prefrontal cortex have complementary roles in emotion and sensory processing [reviewed in Barbas (2000)]. We found that the interconnections between them vary in a systematic way that can be predicted by the laminar structure of the respective cortices. Area 32 pathways to pOFC were distinguished for their strength, large size of axon terminals, strong bias for innervating excitatory neurons, and preferential targeting of the deep (output) layers of pOFC which project to subcortical structures, including the amygdala and thalamus.

Systematic variation in the architecture of ACC and pOFC predicts their laminar connection patterns

A theoretical model—the structural model for connections—predicts that the relationship in laminar structure between areas determines the laminar pattern of their interconnections (Barbas 1986; Barbas and Rempel-Clower 1997). The structural model is based on classical studies that show systematic variation in the laminar structure across the cortical landscape [von Economo (1927/2009); Sanides (1970), reviewed in Barbas (2015)]. Specifically, the structural model predicts that projections from areas that are more stratified communicate with areas that are less stratified according to a predominant feedforward pattern: projection neurons originate mostly in the upper layers (II–III) and their axons terminate in the middle-deep layers (IIIb–Va). Pathways proceeding in the reverse direction—from areas that are less stratified to areas that are more stratified—have a feedback-like pattern: they originate in neurons from the deep layers (V–VI) and their axons terminate in the upper layers (I–IIIa) of the target area. When 2 areas have comparable structure, the model predicts that the pattern of connections between them will be columnar, with projection neurons found in layers II–III and V–VI and axons terminating in all layers in the respective areas (Barbas and Rempel-Clower 1997).

The ACC and the pOFC of the macaque cortex are composed of areas with the simplest laminar structure within the prefrontal cortex but still vary somewhat (Barbas and Pandya 1989; Morecraft et al. 1992), in architectonic gradations seen also in humans (Beck 1949; Sanides 1964; Hof et al. 1995; Semendeferi et al. 1998; Ongur et al. 2003; Mackey and Petrides 2014). The degree of laminar elaboration can be approximated by neuron density for most (though not all) areas of the cerebral cortex (Dombrowski et al. 2001; Collins et al. 2010; Hilgetag et al. 2016). The variation in neuronal density is especially prominent in the upper layers, which are less dense in agranular and dysgranular (limbic) than in eulaminate areas (Barbas and Pandya 1989; Dombrowski et al. 2001). The present study showed that even subtle architectonic differences are evident with SMI-32, which labels a subset of pyramidal projection neurons in layers III and V (Campbell and Morrison 1989) and clearly demarcates an incipient layer IV in dysgranular areas.

The present study revealed that the laminar distribution of connections is sensitive to even small differences in laminar structure, providing strong support for the structural model. Thus, pathways from area 32 showed a bias for a feedforward pattern of connections with the comparatively lesser differentiated agranular pOFC (area OPAll) and the adjacent dysgranular area (OPro), and a columnar pattern in the more rostrally situated area 13, which appears to be comparable in overall structure to area 32. These connectional relationships were consistently replicated by labeling projection neurons in area 32 directed to pOFC areas and in experiments that labeled pathways in the reverse direction, from pOFC areas to area 32. This evidence attests to the sensitivity of the structural model to explain patterns of connections. These findings add to growing empirical support about the generality of the model to help explain the laminar pattern, strength, presence, and even absence of connections in a variety of species and cortical areas [Rempel-Clower and Barbas (2000); Barbas et al. (2005); Medalla and Barbas (2006); Hoistad and Barbas (2008); Hilgetag and Grant (2010); Goulas et al. (2014); Hilgetag et al. (2016), reviewed in Barbas (2015); Barbas and García-Cabezas (2016)].

Area 32 Pathways to pOFC are Strong and Innervate Preferentially Excitatory Neurons

Our findings also provided novel evidence that pathways from area 32 to pOFC were not only robust as found in previous studies (Barbas and Pandya 1989; Barbas 1993; Cavada et al. 2000), but also included a significant population of large terminals, seen at the level of the entire region quantitatively and confirmed at the synaptic level. The significance of large terminals is based on evidence that they have more synaptic vesicles (Germuska et al. 2006; Zikopoulos and Barbas 2006, 2007) associated in physiologic studies with efficient signal transmission (Murthy et al. 1997; Walmsley et al. 1998; Stevens 2003). Novel evidence in this context emerged from our findings showing that a significant proportion of area 32 boutons in pOFC formed perforated synapses with interrupted postsynaptic density, which are more efficacious than simple synapses (Greenough et al. 1978; Sirevaag and Greenough 1985; Geinisman et al. 1987; Ganeshina et al. 2004).

The pathway from area 32 to pOFC was further distinguished for its strong bias to form synapses on spines (~95%), which are enriched on excitatory neurons in the cerebral cortex (Peters et al. 1991). Fewer area 32 boutons formed synapses with shafts of smooth dendrites, some of which were labeled with PV or CR and likely belong to inhibitory neurons (Peters et al. 1991; DeFelipe 1997; Fiala and Harris 1999). Dendritic shafts labeled with CB were sparsely spiny and had fewer other shaft synapses suggesting that they could have been on putative excitatory neurons (DeFelipe 1997).

These findings are in sharp contrast to the synaptic innervation in other cortices, where area 32 boutons form synapses with a higher proportion of elements from inhibitory neurons, as shown for lateral prefrontal areas 9, 10, and 46 [~18–33%; Medalla and Barbas (2009, 2010)], and in the memory-related parahippocampal and rhinal cortices (Bunce and Barbas 2011; Bunce et al. 2013). The only region in which area 32 axons innervate mostly spines of presumed excitatory neurons is the primary olfactory cortex (~90%), which is also situated in the posterior orbitofrontal region (García-Cabezas and Barbas 2014).

Functional Implications

The connectional relationship between ACC and pOFC suggests functional specialization in these 2 regions, which are operationally defined as limbic by virtue of their simple laminar structure [reviewed in Barbas (1997)]. The pOFC appears to be a specialized receiver of signals from the external environment through high-order sensory association cortices, and from the internal environment through connections with other limbic areas (Barbas 1993). The tendency of pOFC to be a dominant receiver of sensory-related information extends to the strong input it receives from the amygdala (Ghashghaei et al. 2007). These pathways may allow the pOFC to integrate sensory information from the external and internal environments and add affective significance to stimuli and events (Timbie and Barbas 2014, 2015; Wilson et al. 2014; John et al. 2016). By comparison, the ACC appears to be a specialized effector system with strong projections to hypothalamic, brainstem and spinal autonomic motor structures (An et al. 1998; Ongur et al. 1998; Rempel-Clower and Barbas 1998; Barbas et al. 2003) that are activated during emotional arousal. The ACC also innervates the output nuclei of the amygdala that project downstream to autonomic structures (Ghashghaei et al. 2007).

Our findings showed that the efferent bias of pathways from area 32 extends to its projection to a “sister” limbic area, the pOFC. Area 32 innervates preferentially the deep layers of pOFC which—like other limbic cortices—have a higher density of neurons than the superficial layers (Dombrowski et al. 2001), receive the bulk of pathways from other cortices (Barbas and Rempel-Clower 1997) and give rise to pathways to the amygdala, thalamus, and other subcortical structures (Haber et al. 1995; Barbas and Rempel-Clower 1997; Ghashghaei and Barbas 2002; Barbas et al. 2003; Xiao and Barbas 2004; Zikopoulos and Barbas 2006, 2012; Ghashghaei et al. 2007; Zikopoulos and Barbas 2007; Xiao et al. 2009). Among these subcortical pathways, the pOFC sends a uniquely strong projection to the IM of the amygdala (Ghashghaei and Barbas 2002). Neurons in IM are inhibitory (Zikopoulos et al. 2016) and project to the central nucleus of the amygdala, which is also inhibitory and innervates autonomic structures (Pare and Smith 1993). The above circuit places the ACC in a key position to influence emotional expression in complementary ways: through direct projections to autonomic structures (An et al. 1998; Ongur et al. 1998; Rempel-Clower and Barbas 1998; John et al. 2013) and through serial pathways via the deep layers of pOFC that project to subcortical structures, as shown here.

Strong efferent pathways from ACC also innervate lateral prefrontal cortices, consistent with the presumed role of ACC in directing attention under high cognitive demand [Botvinick (2007), Pessoa (2008), reviewed in Fuster (2008)]. However, in the lateral prefrontal cortices—which have 6 well differentiated layers—the ACC innervates preferentially the upper layers (Medalla and Barbas 2009, 2010, 2012), consistent with the rules of the structural model for connections (Barbas and Rempel-Clower 1997).

In conclusion, our findings showed that even small differences in laminar structure successfully predict the laminar pattern of connections between 2 phylogenetically ancient prefrontal regions. Differences in the timing of neurogenesis and neuron migration in some ACC and orbitofrontal areas (Rakic 2002) may underlie the differences in the architecture of these areas, but the mechanism of forming connections is largely unknown in primates. By analogy with the sensory areas (Felleman and Van Essen 1991), the patterns of connections from ACC to the middle-deep layers of pOFC suggest a feedforward flow of information. The significance of these findings is 2-fold. First, the linkage of connections to the systematic structural variation of the cortex suggests a global organizing principle that can be used to predict patterns of connections in humans by studying cortical architecture from postmortem brains. Second, the pattern of innervation between ACC and pOFC has distinct implications for psychiatric diseases, such as obsessive compulsive disorder and phobias that affect the pOFC region (Zald and Kim 1996; Milad and Rauch 2012; John et al. 2013). Heightened activity in ACC in a variety of psychiatric disorders, including autism and depression (Holtzheimer et al. 2012), may also increase excitatory drive in pOFC, with cascading effects that reach subcortical structures, which feed back to the cortex and influence areas associated with cognition and emotions.

Funding

This work was supported by the National Institute of Neurological Disorders and Stroke at the National Institutes of Health (Grant number R01NS024760); the National Institute of Mental Health at the National Institutes of Health (Grant number R01MH057414); and by Center of Excellence for Learning in Education, Science and Technology (CELEST), a National Science Foundation Science of Learning Center (Grant number NSF SBE-0354378). M.Á. García-Cabezas was the recipient in 2011 and 2012 of a Grant from Fundación Alfonso Martín Escudero (Spain) and a 2014 NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation (Grant number 22777, P&S Fund Investigator).