|Home | About | Journals | Submit | Contact Us | Français|
The central nucleus of the amygdala (CEA) and lateral bed nucleus of stria terminalis (BST) are highly interconnected limbic forebrain regions that share similar connectivity with other brain regions that coordinate behavioral and physiological responses to internal and environmental stressors. Their similar connectivity is frequently referred to when describing the CEA and lateral BST together as a unified “central extended amygdala”. However, the CEA and BST reportedly play distinct roles in behavioral and physiological responses associated with fear, anxiety, and social defeat, presumably due to differences in connectivity. To identify common and unique sources of input to the CEA and lateral BST, we performed dual retrograde tracing. Fluorogold and cholera toxin β were iontophoresed into the medial CEA (CEAm) and the anterior ventrolateral BST (BSTvl) of adult male rats. The anatomical distribution of tracer-labeled neurons was mapped throughout the brain. Regions with overlapping populations of CEAm- and BSTvl-projecting neurons were further examined for the presence of double-labeled neurons. Although most regions with input to the mCEA also projected to the BSTvl, and vice versa, cortical and sensory system-related regions projected more robustly to the CEAm, while motor system-related regions primarily innervated the BSTvl. The incidence of double-labeled neurons with collateralized axonal inputs to the CEAm and BSTvl was relatively small (~2 to 13%) and varied across regions, suggesting regional differences in the degree of coordinated CEAm and BSTvl input. The demonstrated similarities and differences in inputs to CEAm and BSTvl provide new anatomical insights into the functional organization of these limbic forebrain regions.
The central nucleus of the amygdala (CEA) and lateral bed nucleus of stria terminalis (BST) are highly interconnected limbic forebrain regions that form the crux of the integrative functional unit often referred to as the central extended amygdala (de Olmos and Heimer 1999). The original concept of the extended amygdala as proposed by de Olmos and Heimer was based upon observations that extended amygdala structures share similar cytoarchitectural and histochemical characteristics (Alheid and Heimer 1988; de Olmos and Heimer 1999). According to their proposal, the central division of the extended amygdala includes the CEA and the lateral BST, in addition to interposed regions such as the substantia innominata (SI) that bridge the anatomical gap between CEA and lateral BST. De Olmos and Heimer further speculated that “all or most of the central extended amygdala would share similar inputs” (de Olmos and Heimer 1999). Indeed, in addition to robustly innervating each other, the CEA and lateral BST receive inputs from broadly similar brain regions implicated in visceral and somatosensory functions, and send axonal projections to broadly similar regions that maintain homeostasis by modifying behavioral, autonomic, and endocrine outflow (Dong et al. 2001b; Dong and Swanson 2003, 2004). Results from many studies indicate that CEA lesions often produce experimental results that are quite similar to results obtained after lateral BST lesions (Tanimoto et al. 2003; Deyama et al. 2007; Nakagawa et al. 2005; Zardetto-Smith et al. 1994). On the other hand, the concept of the central extended amygdala has been challenged by results from behavioral studies that suggest a dissociation of CEA and BST functions (Walker and Davis 1997; Walker et al. 2009; Walker et al. 2003; Jasnow et al. 2004; Funk et al. 2006; Fendt et al. 2003). For example, the CEA and lateral BST have been reported to play unique roles in mediating behavioral processes associated with fear, anxiety (Walker et al. 2003; Walker and Davis 1997; Fendt et al. 2003), social defeat (Jasnow et al. 2004), social interaction (Cecchi et al. 2002) and ethanol self-administration (Funk et al. 2006).
The apparent similarities and differences in CEA- and lateral BST-mediated functions may be due, at least in part, to similarities and differences in the connectivity of subregions within each structure. Evidence from retrograde tracing studies indicates that major subdivisions of the lateral BST are interconnected with specific CEA subdivisions that share similar inputs from defined subsets of diencephalic, pontine, and medullary regions (Sun et al. 1991). For example, dopaminergic inputs are most heavily concentrated in the lateral CEA (CEAl) and the dorsolateral BST (BSTdl), while noradrenergic (NA) inputs primarily target the medial CEA (CEAm) and ventrolateral BST (BSTvl; distribution of NA fibers shown in Figs. 1d, ,2d)2d) (Freedman and Cassell 1994). Thus, even within the “central extended amygdala”, afferent inputs to the CEAl and BSTdl differ to some extent from inputs to the CEAm and BSTvl.
Central afferents to the BSTvl in rats have recently been described (Shin et al. 2008). Based on its inputs, the BSTvl appears to integrate signals that impact emotional and motivational states such as pain, pleasure, hunger, thirst, and sickness (Gaykema et al. 2007; Harris and Aston-Jones 2007; Geerling and Loewy 2006; Gauriau and Bernard 2002; Ciccocioppo et al. 2003). Generally, in addition to inputs from the CEAm and other amygdala subregions, the BSTvl receives inputs from the hypothalamus, caudal medulla [including particularly robust input from NA neurons within the caudal nucleus of the solitary tract (NTS) and ventrolateral medulla (VLM)], thalamus, insular cortex, and infralimbic cortex. While most of these regions have also been described as innervating the CEA, the extent to which projections to the BSTvl and CEA arise from the same subregions and, potentially, from the same neurons has been largely unexamined. To date, only the VLM, medial prefrontal cortex, and insular cortex have been reported to contain individual neurons with collateralized axonal inputs to both the CEA and BST (Ciriello et al. 1994; Reynolds and Zahm 2005). However neither study specifically examined inputs to the CEAm and BSTvl, and there has been little experimental effort to discriminate between inputs that target the CEAm versus the CEAl.
The present study used iontophoretic delivery of two different retrograde neural tracers into the CEAm and BSTvl in order to examine the central distribution and potential overlap of neurons that provide axonal inputs to these specific subregions of the central extended amygdala. We hypothesized that brain regions with relatively large numbers of collateralized inputs to both the CEAm and BSTvl may provide an anatomical substrate for coordinating CEA and lateral BST outflow, while areas with few collateralized inputs may contribute to the unique functions of these two regions.
Adult male Sprague–Dawley rats (250–300 g BW; Harlan Laboratories, Indianapolis, IN, USA) were individually housed in stainless steel hanging cages in a controlled environment (20–22°C, 12:12 h light:dark cycle; lights off at 1900 h) with ad libitum access to water and pelleted chow (Purina 5001). Experimental protocols were approved by the University of Pittsburgh Institutional Animal Care and Use Committee, and were carried out in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, with efforts to minimize both the number of animals used and their potential discomfort.
Rats were anesthetized by halothane or isoflurane inhalation (Halocarbon Laboratories, River Edge, NJ; 1–3% in oxygen) and oriented into a Kopf stereotaxic device in the flat skull position. Pulled glass pipette tips (~20 µm outer tip diameter) were attached to the arm of the stereotax. A solution of 1% cholera toxin subunit B (CTB; List Biological Labs, Campbell, CA, USA) in 0.1 M phosphate buffer (pH 6.0) or a 1–2% solution of Fluorogold (FG; Fluorochrome, Denver, CO, USA) in 0.1 M cacodylic acid was backfilled through the pipette tip using negative pressure, then a wire connected to a current source (Stoelting) was inserted into the tracer solution. During the descent of the glass pipette into the brain, a −0.5 µA retaining current was used to minimize molecular diffusion of tracer from the pipette tip. CTB was unilaterally iontophoresed into the BSTvl (from bregma: −0.3 posterior, +2.8 lateral, −7.0 ventral; 10° angle) or CEAm (from bregma: −2.1 posterior, +3.9 lateral, −7.8 ventral) using a 7 s on/off pulsed current of +5 µA for 15 min. Afterward, FG was iontophoresed in the corresponding ipsilateral CEAm or BSTvl using the same pulsed current for 5 min. These iontophoretic parameters were based on the results of pilot studies to produce comparably sized tracer delivery sites. After the second tracer was delivered, the pipette was withdrawn and the skin closed with stainless steel clips. Rats were injected subcutaneously with 0.5 ml of a mild analgesic (Ketofen) and were returned to their cages after regaining consciousness and full mobility.
One to two weeks after tracer iontophoresis, rats were anesthetized with an overdose of sodium pentobarbital (Nembutal, 100 mg/kg BW, ip) and then transcardially perfused with 0.15 M NaCl followed by 500 ml of 4% paraformaldehyde. Brains were postfixed in situ overnight at 4°C then removed from the skull and cryoprotected in 20% sucrose solution before sectioning. Brains were sectioned coronally (35 µm) using a freezing microtome. Sections were collected sequentially into six adjacent sets and stored in cryopreservant (Watson et al. 1986) at −20°C for later immunocytochemical processing.
Two sets of tissue sections from each rat (each set containing sections spaced 210 µm apart) were used for single immunoperoxidase localization of FG and CTB to reveal tracer delivery sites and the distribution of afferent inputs to those sites. For this purpose, tissue sections were incubated overnight in buffer (0.1 M sodium phosphate, pH 7.4) containing 0.3% Triton X-100, 1% normal donkey serum, and either rabbit anti-FG (1:30,000; Millipore, Temecular, CA, USA) or goat anti-CTB (1:50,000, List Biological Labs, Campbell, CA, USA) antisera. Biotinylated secondary antisera (donkey anti-rabbit or donkey antigoat IgG; Jackson Immunochemicals, West Grove, PA, USA) were used at a dilution of 1:500. FG or CTB single immunoperoxidase labeling was revealed using Vectastatin ABC Elite reagents (Vector Laboratories, Burlingame, CA, USA) followed by a diaminobenzidine (DAB)-hydrogen peroxidase reaction to produce a brown immunoprecipitate in the cytoplasm of labeled neurons. In selected cases, additional series of tissue sections were processed for dual immunoperoxidase labeling of either FG or CTB together with the neuronal marker NeuN in order to better define tissue cytoarchitecture. In these cases, FG or CTB immunoperoxidase labeling was revealed by using nickel sulfate in the DAB solution to produce a black immunoprecipitate, followed by a second brown DAB immunoperoxidase reaction after incubating tissue sections in mouse anti-NeuN (1:5,000, Millipore, Temecula, CA, USA) and biotinylated donkey anti-mouse IgG (1:500; Jackson Immunochemicals, West Grove, PA, USA), as described above.
A third set of tissue sections from selected cases with well-placed tracer delivery sites into both the CEAm and the BSTvl was processed for dual immunofluorescent localization of both CTB and FG to identify double-labeled neurons. Tissue sections were incubated for 48–72 h at 4°C in a cocktail of primary antisera at 10 times the concentration used for immunoperoxidase (i.e., FG, 1:3,000; CTB, 1:5,000). Sections were then rinsed in buffer and incubated overnight (at 4°C) in a cocktail of fluorescently tagged secondary antisera [Cy3-conjugated donkey anti-rabbit IgG (1:500) and Cy2-conjugated donkey anti-goat IgG (1:300); both from Jackson Immunochemicals]. Fluorescent labeling of dopamine-beta hydroxylase (DBH) shown in Figs. 1c and and2c2c was performed following a similar protocol on non-experimental rat brain sections using mouse anti-DBH (1:6,000; Millipore, Billerica, MA, USA) and Cy3-conjugated donkey anti-mouse IgG (1:300; Jackson Immunochemicals).
Immunoperoxidase- or immunofluorescence-labeled tissue sections were rinsed in buffer and mounted onto Superfrost Plus microscope slides (Fisher Scientific), allowed to dry overnight, dehydrated and defatted in graded ethanols and xylene, and coverslipped using Cytoseal 60 (VWR).
The parcellation and nomenclature of the BST was first described in 1989 by Ju and colleagues based on cyto- and chemoarchitectonic studies of the BST and its surrounding region (Ju and Swanson 1989; Ju et al. 1989). Later anterograde tracing studies using discrete injections of PHA-L into each subnucleus led to an updated parcellation and nomenclature (Dong and Swanson 2006a). In our study, we attempted to iontophoretically target retrograde tracer to the fusiform subnucleus, which receives particularly dense NA input. However, retrograde tracer iontophoresis in the present study produced larger tracer delivery sites compared to those achieved using PHA-L, and so we describe these delivery sites as encompassing the BSTvl. We refer to the BSTvl as including the fusiform subnucleus and subcommissural part of the anterolateral subnucleus, as described in the Swanson atlas (Swanson 2004). In reference to Paxinos and Watson’s atlas, this generally corresponds to the fusiform subnucleus and subcommissural parts of the intermediate, posterior, and ventral subnuclei of the BST’s lateral division (Paxinos and Watson 2007). Shin and colleagues (Shin et al. 2008) used Ju and colleague’s 1989 nomenclature (Ju and Swanson 1989; Ju et al. 1989) in their report; thus, their retrograde tracer delivery sites were described as targeted to the fusiform nucleus but diffusing into the anterodorsal and subcommissural subnuclei (comparable to the subcommissural part of the anterolateral subnucleus as defined by Dong and Swanson), the dorsomedial subnucleus, and the parastrial nucleus (part of the preoptic hypothalamus adjacent to the BST). By comparison, the iontophoretic delivery sites in the present study were similarly targeted within the BSTvl, but avoided diffusion into the dorsomedial subnucleus and parastrial nucleus.
One selected case with the most accurate CEAm and BSTvl tracer delivery sites was used to fully document the distribution of retrogradely labeled neurons throughout the rostrocaudal extent of the brain. The distribution of CTB- and FG-positive retrogradely labeled neurons was plotted at 409 magnification using a Nikon light microscope connected to a computerized data acquisition system (StereoInvestigator; MBF Bioscience). Plots were made using two adjacent tissue series separately stained for FG and CTB. First, the distribution of FG-labeled neurons was plotted from one set of sections spaced by 420 µm. Plots of FG-labeled neurons were then matched and aligned to sections from the adjacent CTB-labeled tissue series. CTB-labeled neurons were plotted directly onto the initial FG plots with the digital markers for FG labeling hidden from view to prevent bias while plotting the location of CTB-positive neurons. The composite plots of FG- and CTB-labeled neurons were then overlaid onto corresponding Swanson atlas figures (Swanson 2004) using Adobe Illustrator software.
In the selected plotting case and in two additional cases with the most accurate CEAm and BSTvl tracer delivery sites (see “Results”), brain regions with overlapping distributions of retrogradely labeled neurons were further analyzed in dual immunofluorescence-labeled sections to reveal double-labeled neurons. Labeling was visualized and digitally photographed using an Olympus photomicroscope with a 209 objective and filters to visualize Cy2 and Cy3 fluorescence. Counts of single- and double-labeled neurons within each region of interest were derived from images viewed on a computer screen using Adobe Photoshop software, while visualizing retrogradely labeled neurons within red and green color channels. The number of sections through each region that were photographed and used for cell counting, and the approximate bregma levels of quantified sections, are reported in Table 1. Within each afferent brain region, the percentage of tracer-labeled neurons with collateralized projections was calculated as [collateralized/collateralized + (CEAm-only and/or BSTvl-only)] × 100. Retrograde labeling was subjected to quantitative analysis only in regions which contained dense overlap of CEAm- and BSTvl-projecting neurons.
Neuroanatomical regions and nomenclature were defined using Swanson’s rat brain atlas (Swanson 2004). For each quantified brain region, statistical comparisons of the number of neurons projecting to the CEAm or BSTvl were performed using Student’s t test. Rostrocaudal distribution of retrograde labeling within the NTS and para-ventricular thalamus (PVT) was analyzed using two-way ANOVA with rostrocaudal level and tracer target site (CEAm or BSTvl) as independent variables. Differences were considered significant when P < 0.05.
Three rats (cases 09-133, 09-110, and 10-9) with dual iontophoretic delivery sites that were most accurately targeted and restricted to the CEAm and BSTvl were selected for quantitative analysis of retrograde labeling. In cases 09-110 and 09-133, FG was iontophoresed into the CEAm (see Fig. 1c) and CTB was iontophoresed into the BSTvl (see Fig. 2c). In case 10-9, the tracers delivered into each target were switched (see Figs. 1a, ,2a).2a). Retrograde labeling patterns were consistent across these three cases, although the actual numbers of retrogradely labeled neurons varied (Table 1). Different patterns of retrograde labeling resulted from tracer delivery sites that “missed” the CEAm and BSTvl, and were instead centered in closely adjacent regions. Those findings are presented at the end of the results section.
Of the three selected cases with the most accurate tracer delivery sites, case 09-133 consistently had the largest number of retrogradely labeled neurons across quantified brain regions. This representative case was used to digitally plot the distribution of tracer-labeled neurons projecting to the CEAm and BSTvl (see Figs. 3–12).
In general, the large majority of brain regions that contained CEAm-projecting neurons also contained BSTvl-projecting neurons, and vice versa. However, and as described in more detail in the following sections, some regions projected primarily to the CEAm with comparably less input to the BSTvl (e.g., insular cortex). Some regions displayed the opposite pattern, with relatively more input to the BSTvl compared to the CEAm (e.g., NTS), while some brain regions projected equivalently to both the CEAm and the BSTvl (e.g., PVT). Across all the regions in which retrograde labeling was quantified, double-labeled neurons with axonal projections to both the CEAm and the BSTvl accounted for as little as 2% to as much as 13% of the total population of tracer-labeled neurons (Table 1). The distribution of retrograde labeling is plotted in Figs. 3–10, in which red stars represent individual CEAm-projecting neurons, and green circles represent BSTvl-projecting neurons.
Retrogradely labeled neurons in the medulla were located primarily within the caudal, visceral portion of the NTS and the caudal VLM ipsilateral to the tracer delivery sites. Significantly more NTS neurons projected to the BSTvl compared to the number of NTS neurons projecting to the CEAm (136 ± 6 vs. 46 ± 15 neurons; P = 0.005). Rostrocaudal analysis of retrograde labeling patterns revealed a different distribution of BSTvl- versus CEAm-projecting NTS neurons (Figs. 3, ,4).4). Two-way ANOVA revealed a significant effect of both tracer target site [F(1, 29) = 90.39, P < 0.001] and rostrocaudal NTS level [F(4, 29) = 4.26, P < 0.05] on the number of retrogradely labeled NTS neurons, as well as a significant interaction between tracer target site and rostrocaudal level [F(4, 29) = 6.17, P < 0.01]. At levels caudal to the area postrema (AP), the NTS contained relatively few CEAm-projecting neurons (red stars in Fig. 3a, b). The number of CEAm-projecting NTS neurons increased at more rostral levels, reaching a peak just rostral to the AP (Figs. 3e, ,4).4). Conversely, the number of BSTvl-projecting NTS neurons reached a peak at the mid-AP level (green circles, Figs. 3c, ,4).4). NTS neurons with collateralized projections to both the CEAm and BSTvl were most prevalent in sections just rostral to the AP (Fig. 4). Approximately 9% of all tracer-labeled NTS neurons were double-labeled (Table 1). Overall, 33 ± 4% of all CEAm-projecting NTS neurons collateralized to also provide input to the BSTvl, whereas only 11 ± 4% of all BSTvl-projecting NTS neurons collateralized to innervate the CEAm (Table 1).
Retrogradely labeled VLM neurons were present at the same rostrocaudal levels as retrogradely labeled NTS neurons (Fig. 3). Quantification of labeling within the VLM revealed a nonsignificant trend towards a greater number of BSTvl- versus CEAm-projecting neurons (25 ±11 vs. 8 ± 4 neurons; P = 0.63). VLM neurons projecting to the CEAm or BSTvl were distributed uniformly across rostrocaudal levels, and moderate numbers of double-labeled neurons with collateralized projections to both the CEAm and BSTvl were observed (Table 1). Approximately 11% of all tracer-labeled VLM neurons were double-labeled, similar to proportions of double-labeled neurons within the NTS (Table 1).
Pontine input to the CEAm and BSTvl was predominantly confined to the parabrachial nucleus (PB; Fig. 5). The locus coeruleus contained a small number of CEAm- and BSTvl-projecting neurons (i.e., 1–3 neurons per section, data not shown). Retrogradely labeled neurons were distributed bilaterally within several PB subnuclei, although a strong ipsilateral predominance of labeling was evident (Fig. 5). At the caudal end of the PB (Fig. 5a, b), retrograde labeling within the PB was heavily concentrated within the ventral lateral (PBlv), medial (PBm), and waist subnuclei (PBw). Within these subnuclei, CEAm-projecting neurons significantly outnumbered BSTvl-projecting neurons (PBlv = 79 ± 9 vs. 29 ± 6 neurons; P = 0.01; PBm = 139 ± 25 vs. 45 ± 7 neurons; P = 0.02; PBw = 25 ± 4 vs. 11 ± 1 neurons; P = 0.04). Relatively few (i.e., approximately 6 ± 1, 3 ± 1, and 8 ± 2%) of all tracer-labeled neurons within the PBlv, PBm, and PBw were double-labeled (Table 1). More rostral levels of the PB (Fig. 5c) contained larger numbers of CEAm- and BSTvl-projecting neurons that were primarily localized within the external lateral PB subnucleus (PBle; see Fig. 13a), with fewer retrogradely labeled neurons present within the PBlv, PBm, and PBw. Quantification revealed a nonsignificant trend towards larger numbers of PBle neurons projecting to the CEAm versus the BSTvl (298 ± 72 vs. 130 ± 9 neurons; P = 0.08). In contrast to other PB subnuclei, many of the retrogradely labeled PBle neurons collateralized to innervate both the CEAm and BSTvl: approximately 12 ± 3% of all tracer-labeled PBle neurons were double-labeled (Table 1). When double-labeled neurons were excluded, significantly more single-labeled PBle neurons projected to the CEAm compared to the number that projected to the BSTvl (Table 1; P < 0.05).
Within the caudal midbrain (Fig. 6a – c), CEAm- and BSTvl-projecting neurons were distributed ventral to the central aqueduct, within the pedunculopontine nucleus (PPN), dorsal raphé (DR), and ventral lateral periaqu-eductal gray (PAGvl). Within the DR, there was no significant difference between the numbers of CEAm- and BSTvl-projecting neurons (77 ± 20 vs. 65 ± 6 neurons; P = 0.61), and relatively few tracer-labeled DR neurons (7 ± 1 neurons, 6 ± 2% of total) had collateralized projections to both the CEAm and BSTvl (Table 1). In more rostral midbrain sections, retrograde labeling extended ventrally to include the central linear raphé (CLI; Fig. 6c, d). There was no significant difference between the number of CEAm- and BSTvl-projecting CLI neurons (29 ± 9 vs. 35 ± 6 neurons; P = 0.62), and there was very little collateralization of individual CLI neurons to both the CEAm and BSTvl (Table 1).
Rostral to the midbrain CLI, CEAm- and/or BSTvl-projecting neurons were located within the ventral tegmental area (VTA; Fig. 7a, b) and the compact portion of the sub-stantia nigra (SNc; Fig. 7c). Retrograde labeling within the VTA was very sparse and was not quantified. The SNc contained a fair number of CEAm-projecting neurons (not quantified, Fig. 7c), but no BSTvl-projecting neurons.
Projections to the CEAm and BSTvl were found throughout the rostrocaudal extent of the thalamus (Figs. 7–9) with large numbers of retrogradely labeled neurons within more caudally located thalamic nuclei (Fig. 7a – e). In general, thalamic CEAm-projecting neurons were more prevalent than BSTvl-projecting neurons. At caudal levels, many CEAm-projecting and few BSTvl-projecting neurons were present within the auditory thalamus (Aud; Fig. 7a – c). In more rostral sections (Fig. 7d), primarily CEA-projecting neurons were present within the medial parvicellular subparafascicular nucleus (SPFpm) and the parvicellular ventral posteromedial nucleus (VPMpc). Further rostrally, retrogradely labeled neurons extended more medially to join the midline thalamic nuclei group (MTN; Fig. 7e), including the central medial nucleus, intermediodorsal nucleus, and the PVT. CEAm- and BSTvl-projecting neurons were distributed throughout the rostrocaudal extent of the MTN (Figs. 7e–8b), but were most numerous within the PVT. When the PVT was considered as a whole, there was no significant difference between the number of CEAm- versus BSTvl-projecting neurons (618 ± 117 vs. 646 ± 35 neurons; P = 0.83). Two-way ANOVA revealed a significant effect of PVT rostrocaudal level [F(1, 83) = 2.75, P < 0.01] and an interaction between tracer target site and PVT rostrocaudal level [F(13, 83) = 6.45, P < 0.001]. When the PVT was divided into three rostrocaudal segments, there was a clear difference in the distribution of CEAm- versus BSTvl-projecting neurons across the caudal, middle, and rostral thirds of the PVT (cPVT = caudal to bregma level −3.25; mPVT = bregma levels −2.85 through −2.00; rPVT = rostral to bregma level −2.00; Fig. 9). Significantly larger numbers of BSTvl-projecting neurons were located within the rPVT compared to either the mPVT (310 ± 18 vs. 157 ± 7 neurons, P < 0.005) or the cPVT (310 ± 18 vs. 179 ± 12 neurons, P < 0.005). Conversely, the number of CEAm afferents within the cPVT was significantly greater than within the mPVT (320 ± 47 vs. 162 ±31 neurons, P < 0.05), but was not greater compared to the number of CEAm afferents in the rPVT (320 ± 47 vs. 196 ± 19 neurons, P = 0.72). Further analysis of retrograde labeling across the three rostrocaudal PVT levels revealed that the cPVT contained significantly greater numbers of CEAm afferents compared to BSTvl afferents (320 ± 47 vs. 179 ± 12 neurons, P < 0.05), while the rPVT contained a significantly greater number of BSTvl afferents compared to CEAm afferents (310 ± 18 vs. 135 ± 42 neurons, P = 0.02). The numbers of CEAm and BSTvl afferents within the mPVT were not significantly different (162 ± 31 vs. 157 ± 7 neurons, P = 0.87).
Hypothalamic regions generally contained many BSTvl-projecting neurons and smaller numbers of CEAm-projecting neurons. The hypothalamic distribution of these populations rarely overlapped, and therefore, were plotted (see Figs. 8, ,9)9) but not quantified. The density of BSTvl-projecting neurons in the hypothalamus appeared greatest within the lateral hypothalamic area (LHA; Fig. 8a) and the medial preoptic area (MPO; Fig. 9c). However, one exception to the largely separate hypothalamic distribution of BSTvl- and CEAm-projecting neurons was the parasubthalamic nucleus (PSTN; Fig. 7e, Fig. 13b), which contained similar numbers of CEAm- and BSTvl-projecting neurons (133 ± 54 vs. 136 ± 11 neurons, P = 0.96). Interestingly, despite the dense overlapping distribution of CEAm- and BSTvl-projecting PSTN neurons, relatively few were double-labeled (10 ± 4 neurons, 4 ± 2% of total; Table 1).
Within the amygdala, large numbers of CEAm- and BSTvl-projecting neurons were located within the posterior basolateral amygdala (BLAp; Figs. 8a, 12b, 13e) and basomedial amygdala (BMA; Figs. 10a, 13d), while relatively fewer retrogradely labeled neurons were present within the lateral or medial amygdala (Fig. 10a). There was a non-significant trend towards higher numbers of CEAm-versus BSTvl-projecting BLAp neurons (689 ± 155 vs. 388 ± 76 neurons; P = 0.16). Collateralized projections of BLAp neurons to both the CEAm and BSTvl were relatively common. Approximately one-third of BSTvl-projecting neurons and 20% of CEAm-projecting neurons within the BLAp were double labeled (13 ± 2% of total retrogradely labeled neurons, Table 1). In contrast, despite similarly large numbers of CEAm- and BSTvl-projecting neurons within the BMA (877 ± 334 vs. 413 ± 151 neurons; P = 0.27; Table 1), smaller proportions of these neurons were double-labeled as compared to the BLAp (6 ± 2% of total retrogradely labeled neurons, Table 1).
As expected, iontophoresis of retrograde tracer into the CEAm produced abundant retrograde labeling within the BSTvl, and vice versa, as well as retrograde labeling within other regions of the extended amygdala (Figs. 1b, ,2b,2b, ,8,8, ,9).9). CEAm-projecting neurons were located throughout the lateral BST, and BSTvl-projecting neurons were found throughout the CEAm and CEAl. Additional CEAm- and BSTvl-projecting neurons were scattered throughout the substantia innominata (SI; Fig. 9a), a region interposed rostrocaudally between the CEA and BST and described as part of the “central extended amygdala” (de Olmos and Heimer 1999). In addition, a dense cluster of CEAm- and BSTvl-projecting neurons was located within a discrete subregion of the SI called the interstitial nucleus of the posterior limb of the anterior commissure (IPAC; Figs. 9b, 13c) (Shammah-lagnado et al. 2001). Quantification of retrogradely labeled IPAC neurons revealed no significant difference between the number of CEAm and BSTvl afferents (227 ± 73 vs. 262 ± 48 neurons; P = 0.71). A moderate proportion of IPAC neurons projected axons to both the CEAm and BSTvl (Table 1). The caudal part of the nucleus accumbens shell (ACBsh), particularly its dorsal medial tip, contained large numbers of BSTvl-projecting neurons but relatively few CEAm-projecting neurons (Fig. 11a).
Three major regions of the cerebral cortex contained CEAm- and BSTvl-projecting neurons. The densest cortical input to the CEAm and BSTvl arose from a region situated near the rhinal sulcus, including the agranular and dysgranular insular cortex (AI and DI, respectively; Figs. 8, ,9,9, ,11).11). Within the AI, significantly greater numbers of retrogradely labeled neurons projected to the CEAm versus the BSTvl (868 ± 208 vs. 102 ± 45 neurons; P = 0.02). More than 30% of BSTvl-projecting AI neurons also projected to the CEA (i.e., were double-labeled), whereas double-labeled neurons comprised only 3% of all AI input to the CEAm (Table 1).
A dense overlap of CEAm- and BSTvl-projecting neurons was observed within deep layers of the caudal infralimbic cortex (ILA; Figs. 11d, 13f), with only a few scattered cells observed in the more dorsally situated prelimbic cortex (PL; Figs. 11d, 13f). Quantification of ret-rogradely labeled neurons within the ILA revealed similar numbers of CEAm- and BSTvl-projecting neurons (96 ± 37 vs. 129 ± 43 neurons; P = 0.6). Despite their overlapping distributions, BSTvl- and CEAm-projecting neurons within the ILA rarely collateralized to both the CEAm and BSTvl (3 ± 2% of total retrogradely labeled neurons; Table 1).
Within the ventral cerebral cortex, dense retrograde labeling was located within the postpiriform transition area (TR, sometimes referred to as the amygdala-piriform transition area; Figs. 12b, 13e), with fewer retrogradely labeled neurons scattered throughout the more medially situated posterior amygdala (PA), and also within area CA1 of the ventral hippocampus (Fig. 12a). Quantification of retrogradely labeled TR neurons revealed a significantly greater number of CEAm- versus BSTvl-projecting neurons (682 ± 94 vs. 192 ± 63 neurons; P = 0.01). Relatively few TR neurons collateralized to innervate both the CEAm and BSTvl (2 ± 1% of total retrogradely labeled neurons; Table 1).
The goal of this study was to target specific subregions of the central extended amygdala, i.e., the BSTvl and the CEAm, for retrograde tracing of their afferent inputs. This clearly is a challenging task, given the relatively small size of each structure. Thus, we sought to determine whether tracer delivery sites that were viewed as “accurately centered” and relatively confined to the BSTvl or CEAm produced retrograde labeling that was distinct from labeling produced by tracer delivery centered in adjacent “incorrect” regions. “Incorrect” tracer delivery sites located adjacent to the BSTvl included the ventral pallidum (VP, lateral to the BST) and the posterior BST (pBST). Iontophoretic delivery of tracer into the VP produced abundant retrograde labeling in the medial subthalamic nucleus (STN), whereas accurate BSTvl iontophoretic delivery produced retrogradely labeled neurons in the adjacent PSTN with no labeling in the STN. Iontophoretic delivery of tracer into the pBST produced dense retrograde labeling within the MEA, whereas MEA labeling was much more sparse after accurate tracer placement in the BSTvl. Tracer delivery into the VP or pBST also produced little or no retrograde labeling within other brain regions that contained labeled neurons after accurate BSTvl tracer delivery, including the NTS, VLM, BLAp, TR, and PSTN.
“Incorrect” tracer delivery sites adjacent to the CEAm included the dorsally situated amygdala-striatal transition area (AStr), the CEAl, and/or the BLA. In contrast to labeling produced by accurate CEAm-targeted sites, iontophoretic delivery of tracer into the BLA produced substantially more retrograde labeling within the pontine locus coeruleus (Asan 1998), and bilateral retrograde labeling of neurons within layer 3 of the nucleus of the lateral olfactory tract, consistent with previous findings (Santiago and Shammah-Lagnado 2004). Tracer delivery into the AStr retrogradely labeled neurons within secondary somatosensory cortex, which does not project to the CEA (Shammah-Lagnado et al. 1999). This cortical region was not labeled in rats with accurate CEAm tracer delivery sites in the present study. Cases in which the center of the iontophoretic delivery site was located within the CEAl rather than the CEAm produced little or no retrograde labeling within the NTS or VLM, in contrast to results following CEAm-centered tracer delivery, and consistent with the preferential distribution of NA fibers within the CEAm as compared to the CEAl (shown in Fig. 1d). NA inputs to the CEA arise primarily from the caudal medulla, and the large majority of NTS and VLM neurons that project to the CEA are NA neurons (Myers and Rinaman 2002).
The present report is the first to fully map and compare the anatomical distribution of neurons projecting to the CEA and lateral BST, the two major components of the central extended amygdala. Our results are specifically focused on central neural inputs to CEAm and BSTvl subregions of the central extended amygdala. Although VLM and cortical neurons have been reported to provide collateralized axonal input to the CEA and BST (Roder and Ciriello 1994; Reynolds and Zahm 2005), previous studies did not investigate inputs that specifically target the CEAm and BSTvl. The present study reveals three patterns of retrograde labeling among brain regions that innervate the CEAm and BSTvl: high numbers of CEAm afferents with fewer BSTvl afferents, high numbers of BSTvl afferents with fewer CEAm afferents, or relatively even numbers of CEAm and BSTvl afferents (Fig. 14). Interestingly, neurons with collateralized inputs to both the CEAm and BSTvl exist within most of the CNS regions that project to either target, although the incidence of collateralization varies among regions (Fig. 14). These results generally support De Olmos and Heimer’s proposal that “all or most of the central extended amygdala would share similar inputs” in the sense that the CEAm and BSTvl receive inputs from the same brain regions (de Olmos and Heimer 1999). Indeed, the SNc appears to be the only brain region that provides input to the CEAm but not to the BSTvl. However, our new findings reveal that inputs from cortical and sensory-related regions appear to preferentially target the CEAm, while inputs from motor-related “behavioral control columns” [(Swanson 2000); see following section] appear to preferentially target the BSTvl.
In addition to considering our results as they pertain to the concept of the “central extended amygdala”, another way to interpret these findings arises from Larry Swanson’s descriptive model of how the brain regulates motivated behavior (Swanson 2000). Generally speaking, behaviorally relevant information from widespread regions of the cerebral cortex reaches motor output systems through a triple descending projection to hierarchically organized behavioral control columns, with each column dedicated to the production of a specific category of behavioral output (i.e., social, defensive, reproductive, or exploratory). Each column contains three levels of control that are common to the production of motivated behavioral output. The highestorder level of behavioral control is located within specific subregions of the hypothalamus and other rostral brainstem nuclei. This upper level of control defines an endogenous baseline activity level for specific subsets of brainstem motor pattern initiators and generators in order to regulate the series of movements that are necessary to produce organized behavior by controlling specific sets of brainstem and spinal motor neurons that initiate muscle contraction.
Overall, brain regions that contained larger numbers of neurons projecting to the CEAm versus the BSTvl (Fig. 14) are associated with cortical or sensory systems (e.g., AI, BLAp, BMA, PB, TR, cPVT, Aud, SPFpm/ VPMpc), while brain regions that contained larger numbers of neurons projecting to the BSTvl versus the CEAm include striatal-like regions and areas associated with Swanson’s behavioral control columns (e.g., most of the hypothalamus, NTS, VLM, ACBsh, LS, PPN). When incorporated with anatomical data from the literature detailing CEA and BST efferent projections (Dong et al. 2001b; Dong and Swanson Dong and Swanson 2003, 2004, 2006a, b), these new findings support an organizational hypothesis for two primary pathways through which behaviorally relevant information is processed by the CEA and BST. First, similar to Swanson’s model, we propose that cortical and sensory information necessary for initiating behaviors converges primarily within the CEA, including the CEAm, which then presumably recruits BST neurons that project to effectors in the motor system’s top-down behavioral control columns (see Fig. 15a) (Dong et al. 2001b; Dong and Swanson 2003, 2004; Dong et al. 2000; Swanson 2000). Secondly, we propose that bottom-up feedback about ongoing behavior is initially received and processed primarily by the BST, including the BSTvl, which then relays the information to the CEA in order to modulate ongoing motor outflow (see Fig. 15b). These top-down and bottom-up pathways may represent parallel but separate anatomical circuits within the CEA and BST, or may facilitate a bidirectional flow of information through the same circuit nodes.
In contrast to the “central extended amygdala” concept, Larry Swanson has noted that the network architecture of the CEA and BST shares many similarities to basal ganglia striatopallidal loop networks (Swanson 2000). He has suggested that the CEA is a caudal extension of the striatum and the BST is a rostral extension of the pallidum that together form a caudorostral striatopallidal circuit that is specially differentiated to regulate autonomic, neuroendocrine, and somatomotor output (Swanson 2000). Our results demonstrating differences in the organization of inputs to the CEAm and BSTvl lend supporting evidence for a striatopallidal-like organization of the CEAm and BSTvl. However, whereas motor feedback to the globus pallidus and ventral pallidum arises from brain regions involved in somatomotor control such as the STN, SNr, and the PPN (DeVito et al. 1980), the pallidal-like BSTvl receives more robust direct input from regions involved in somatomotor and visceromotor control, including the hypothalamus and pontine and medullary regions that receive and process visceral sensory inputs. Thus, the BSTvl (and the CEAm, to a lesser extent) receives moment-to-moment feedback about the physiological consequences of behavior, including autonomic and endocrine adjustments, consistent with evidence that changes in visceral and endocrine outflow can occur with little or no ongoing control by cortical structures. Given this abundant feedback, the CEA and BST are well-positioned to adjust somatomotor, autonomic, and neuroendocrine outflow as necessary to support ongoing and anticipated behavioral responses.
Individual neurons with collateralized axonal inputs to the CEAm and BSTvl were observed in nearly every brain region that contained retrogradely labeled neurons, although the incidence of collateralized projection neurons differed among regions. The largest proportions of retrogradely labeled neurons that were double-labeled were found within the lPBN, BLAp, and NTS (Fig. 14), suggesting that information transfer from these regions to the CEAm and BSTvl is more highly coordinated as compared to inputs from other brain regions. Previous work has demonstrated that BLA-driven neural responses within the CEA and BST are temporally synchronized in order to simultaneously activate target neurons within the brainstem, providing evidence for cooperative output of the extended amygdala (Nagy and Pare 2008). Interestingly, the BLAp, lPBN, and NTS are critical structures for acquisition in aversive learning paradigms (Sakai and Yamamoto 1998; Fendt and Fanselow 1999; Reilly 1999; Fanselow and LeDoux 1999), and the relatively high degree of collateralized input to the CEAm and BSTvl from these regions may contribute to the acquisition of newly learned behaviors through synchronized activity.
The PSTN and ILA stood out as having relatively few collateralized inputs to the CEAm and BSTvl, despite abundant retrograde labeling from both target regions. The PSTN and ILA are strongly implicated in providing descending control over autonomic functions (Goto and Swanson 2004; Ciriello et al. 2008; Hurley et al. 1991; Fisk and Wyss 2000; Heidbreder and Groenewegen 2003). Transneuronal viral tracing of preautonomic circuits has revealed distinct parallel descending projections to specific visceral targets (Sved et al. 2001). The presence of relatively few collateralized inputs from the PSTN and ILA to the CEAm and BSTvl suggests that neurons in these regions may contribute to differential control over autonomic output to different visceral targets.
Although the NTS and PVT contained overlapping distributions of CEAm- and BSTvl-projecting neurons, rostrocaudal analysis revealed differing distributions of where these neurons were located within each nucleus. Within the NTS, BSTvl-projecting neurons were most prevalent at the level of the area postrema, while the largest number of CEAm-projecting neurons peaked just rostral to the area postrema (Fig. 3a), which also contained the largest number of neurons with collateralized axons to both the CEAm and BSTvl. Because vagal sensory inputs to the NTS terminate in a generally viscerotopic pattern (Kalia and Sullivan 1982; Altschuler et al. 1989), our findings of differing distributions of CEAm- versus BSTvl-projecting NTS neurons suggest differences in the type of viscerosensory feedback that may be relayed to the CEAm versus the BSTvl.
Within the PVT, BSTvl-projecting neurons were significantly more prevalent within the rPVT compared to the cPVT, consistent with a previous qualitative report of the distribution of BSTvl-projecting neurons within the PVT (Shin et al. 2008). Conversely, CEAm-projecting neurons were significantly more prevalent in the cPVT than the rPVT (Fig. 10a). Interestingly, the cPVT also provides dense axonal input to corticotrophin-releasing factor (CRF) neurons of the CEAl and BSTdl (Li and Kirouac 2008). A series of experiments have found that cPVT lesions affect behavioral and endocrine responses to chronic stress (Bhatnagar et al. 2002; Bhatnagar and Dallman 1998; Jaferi et al. 2003; Bhatnagar et al. 2003), while the rPVT appears to play a role in light-induced entrainment of circadian rhythms (Salazar-Juárez et al. 2002). Further investigation of the differential PVT input to the CEAm and BSTvl is needed to understand how these inputs may contribute to unique functions of the two limbic regions.
Lesion studies of the medial prefrontal cortex (mPFC) have revealed contrasting functional roles for the PL versus ILA in regulating the hypothalamic–pituitary–adrenal (HPA) neuroendocrine stress axis, suggesting that the ILA promotes HPA axis activity while the PL suppresses it (Radley et al. 2006). Two recent reports have directed focus on the BST as a relay for PL cortical inhibitory influence over the HPA axis (Radley et al. 2009; Radley and Sawchenko 2011). Within the BSTvl, the fusiform and dorsomedial BST subnuclei contain GABAergic neurons that innervate neuroendocrine neurons within the medial parvocellular subregion of the paraventricular nucleus of the hypothalamus (PVN) (Cullinan et al. 2008; Cullinan et al. 1993). In the present study, however, almost all of the mPFC neurons that project to the BSTvl were located within the ILA, in agreement with a previous retrograde tracing study of BSTvl inputs (Shin et al. 2008) and anterograde tracing studies of neural projections from the ILA and PL (Hurley et al. 1991; Vertes 2004; Chiba et al. 2001; Sesack et al. 1989)
The CEA and lateral BST have been described as constituent parts of an anatomical–functional macrosystem known as the ‘central extended amygdala’ (de Olmos and Heimer 1999). Our new findings challenge this view by revealing the anatomical organization of common and distinct sets of neural inputs to two discrete subregions of this proposed macrosystem, the CEAm and BSTvl. Cortical and sensory systems primarily target the CEAm, while input from behaviorally relevant motor systems and viscerosensory nuclei relaying interoceptive feedback from the body primarily target the BSTvl. Neurons with collateralized axonal inputs to both the CEAm and BSTvl are located within nearly all of the brainstem and forebrain regions that provide axonal input to either structure. The incidence of collateralization varies across brain regions, but is relatively minor compared to the number of neurons that provided distinct input to either the CEAm or BSTvl. Taken together, our new findings suggest an anatomical framework for information processing that may contribute to a better understanding of how CEA and BST circuits participate in organizing complex behavioral responses to cognitive and physiological challenges.
We thank Vicki Maldovan and Li Cai for their expert technical assistance in this study. This research was supported by grants from the National Institutes of Health (MH59911 and DK063922). Research supported by NIH grant MH59911 to L.R. and NIH grant DK063922 to M.B.