Intrinsic connections in the anterior part of the bed nucleus of the stria terminalis
Hjalmar K. Turesson, Olga E. Rodríguez-Sierra, and Denis PareCenter for Molecular and Behavioral Neuroscience, Rutgers State University, Newark, New Jersey
Submitted 3 January 2013; accepted in final form 25 February 2013
Turesson HK, Rodríguez-Sierra OE, Pare D. Intrinsic connec-tions in the anterior part of the bed nucleus of the stria terminalis. JNeurophysiol 109: 2438–2450, 2013. First published February 27,2013; doi:10.1152/jn.00004.2013.—Intrinsic connections in the ante-rior portion of the bed nucleus of the stria terminalis (BNST-A) werestudied using patch recordings and ultraviolet (UV) glutamate uncag-ing (GU) in vitro. UV light was delivered at small BNST-A sites in agrid-like pattern while evoked responses were monitored in differentBNST-A regions. Three sectors were distinguished in the BNST-Ausing fiber bundles readily identifiable in transilluminated slices: theanterior commissure, dividing the BNST-A into dorsal and ventral(BNST-AV) regions, and the intra-BNST component of the striaterminalis, subdividing the dorsal portion into medial (BNST-AM)and lateral (BNST-AL) regions. Overall, GU elicited GABAergicinhibitory postsynaptic potentials (IPSPs) more frequently than excit-atory postsynaptic potentials. The incidence of intraregional connec-tions was higher than interregional links. With respect to the latter,asymmetric connections were seen between different parts of theBNST-A. Indeed, while reciprocal connections were found betweenthe BNST-AL and BNST-AM, BNST-AL to BNST-AM connectionswere more frequent than in the opposite direction. Similarly, whileGU in the BNST-AM or BNST-AL often elicited IPSPs in BNST-AVcells, the opposite was rarely seen. Within the BNST-AM, connec-tions were polarized, with dorsal GU sites eliciting IPSPs in moreventrally located cells more frequently than the opposite. This trendwas not seen in other regions of the BNST. Consistent with this, mostBNST-AM cells had dorsally directed dendrites and ventrally ramifiedaxons, whereas this morphological polarization was not seen in otherparts of the BNST-A. Overall, our results reveal a hitherto unsus-pected level of asymmetry in the connections within and betweendifferent BNST-A regions, implying a degree of interdependence intheir activity.
bed nucleus of the stria terminalis; glutamate uncaging; fear; anxiety
THE BED NUCLEUS OF THE STRIA TERMINALIS (BNST) is a poorlyunderstood brain structure that has been implicated in a variety offunctions, most relating to negative affects and stress. For in-stance, the anterior portion of the BNST (BNST-A) has beenshown to regulate the hypothalamus-pituitary-adrenal axis (Rad-ley and Sawchenko 2011; Ulrich-Lai and Herman 2009), to me-diate stress-induced relapse to drug seeking (Erb and Stewart1999), and to generate fear/anxiety responses to diffuse environ-mental cues (Davis et al. 2010) or predator odors (Fendt et al.2005).
Although its name suggests otherwise, the BNST is a collectionof nuclei. While there is disagreement regarding the number andboundaries of BNST nuclei (Andy and Stephan 1964; De Olmoset al. 1985; Ju and Swanson 1989; Moga et al. 1989), it is clearthat different BNST regions form contrasting connections. Withinthe BNST-A, for instance, hypothalamus-pituitary-adrenal-
regulating neurons are concentrated in its ventral (BNST-AV) andmedial (BNST-AM) portions (Dong et al. 2001b; Dong andSwanson 2006a; Prewitt and Herman 1998). In contrast, neuronsin the dorsolateral part of the BNST-A (BNST-AL) contributemost BNST outputs to brain stem structures regulating fear ex-pression (Holstege et al. 1985; Moga et al. 1989; Sofroniew et al.1983; Sun and Cassell 1993).
Similarly, many afferents to the BNST-A form heterogeneousconnections with these different regions. For instance, subicularand medial prefrontal inputs target the BNST-AM and BNST-AVbut not the BNST-AL, whereas insular axons show the oppositepattern or termination (for a review, see McDonald et al. 1999).Also, monoaminergic inputs are differentially distributed in theBNST-A (for a review, see Krawczyk et al. 2011). Furthermore,the main source of extrinsic input to the BNST, the amygdala, alsocontributes contrasting projections to different BNST-A regions(Dong et al. 2001a; Krettek and Price 1978b).
While this heterogeneous connectivity suggests a degree offunctional specialization within the BNST-A, a comprehensiveseries of tracing studies by Swanson and colleagues (Dong andSwanson 2003, 2004, 2006a, 2006b, 2006c) suggested thatdifferent BNST-A regions do not act as independent processingchannels but that they interact via inter-nuclear connections.For instance, they reported that components of the BNST-AL,particularly, the oval nucleus, strongly projects to parts of theBNST-AV, such as the fusiform nucleus (Dong and Swanson2004). However, interpretation of these findings is complicatedby the fact that the distance between different BNST regions issmall relative to the considerable extent of dendritic trees in theBNST (Larriva-Sahd 2006; McDonald 1983). Moreover, thisproblem is compounded by tracer diffusion from the injectionsite in the small volume of BNST, particularly along the tractof the pipettes used to inject the tracers.
Another unresolved question relates to the transmitter(s) usedby intrinsic BNST axons. Indeed, previous work has revealed thatthe BNST-A contains GABAergic and glutamatergic neurons(Cullinan et al. 1993; Polston et al. 2004; Poulin et al. 2009; Sunand Cassell 1993), with GABAergic cells accounting for themajority of BNST-A cells and glutamatergic neurons for a mi-nority. Thus, the present study was undertaken to shed light on theorganization of intrinsic BNST-A connections using a method thathas higher spatial resolution than tract tracing and allows identi-fication of the transmitters involved: glutamate uncaging (GU)coupled to patch recordings in vitro.
MATERIALS AND METHODS
The procedures used in the present study were approved by theInstitutional Animal Care and Use Committee of Rutgers Universityin compliance with the National Institutes of Health Guide for theCare and Use of Laboratory Animals. Our subjects were male Lewisrats (4–5 wk old, Charles River Laboratories, New Field, NJ) main-
Address for reprint requests and other correspondence: D. Paré, Center forMolecular and Behavioral Neuroscience, Rutgers State Univ., 197 Univ. Ave.,Newark, NJ 07102 (e-mail: [email protected]).
J Neurophysiol 109: 2438–2450, 2013.First published February 27, 2013; doi:10.1152/jn.00004.2013.
tained on a 12:12-h light-dark cycle. Rats were housed individuallywith ad libitum access to food and water. Before the experiments, ratswere habituated to the animal facility and handling for 1 wk.
Slice Preparation
Rats were anesthetized with ketamine (80 mg/kp ip), pentobarbital(60 mg/kg ip), and xylazine (12 mg/kg ip). After abolition of allreflexes, rats were perfused through the heart with cold (4°C) modifiedartificial cerebrospinal fluid (aCSF) that contained (in mM) 248sucrose, 2.5 KCl, 7 MgCl2, 23 NaHCO3, 1.2 NaH2PO4, and 7 glucose.Their brains were then extracted and cut into 400-�m-thick coronalslices with a vibrating microtome while submerged in the samesolution as detailed above. After the cutting process, slices weretransferred to an incubating chamber, where they were allowed torecover for at least 1 h at room temperature in control aCSF of thefollowing composition (in mM): 124 NaCl, 2.5 KCl, 1.25 NaH2PO4,26 NaHCO3, 1 MgCl2, 2 CaCl2, and 10 glucose (pH 7.3, 300 mOsm).Slices were then transferred one at a time to a recording chamberperfused with the latter solution (5 ml/min) plus caged glutamate[4-methoxy-7-nitroindolinyl-caged-L-glutamate (1.0 mM), TocrisBioscience, Bristol, UK]. Before the recordings began, the tempera-ture of the chamber was gradually increased to 32°C.
Electrophysiology
Under visual guidance with differential interference contrast and in-frared video microscopy, we obtained whole cell patch recordings ofBNST neurons using pipettes (7–10 M�) pulled from borosilicate glasscapillaries and filled with a solution containing (in mM) 130 K-gluconate,10 N-2-hydroxyethylpiperazine-N=-2-ethanesulfonic acid, 10 KCl, 2MgCl2, 2 ATP-Mg, and 0.2 GTP-tris(hydroxy-methyl)aminomethane
(pH 7.2, 280 mOsm). Biocytin (0.5%) was often added to the intracellularsolution to label the recorded cells. The liquid junction potential was 10mV with this solution, and the membrane potential was corrected accord-ingly. Current-clamp recordings were obtained with an Axoclamp 2Bamplifier and digitized at 10 kHz with a Digidata 1200 interface (AxonInstruments, Foster City, CA).
To characterize the electroresponsive properties of recorded cells,a graded series of depolarizing and hyperpolarizing current pulses(20-pA steps, 500 ms in duration) was applied from rest and otherprepulse potentials. The input resistance of the cells was estimated inthe linear portion of current-voltage plots.
GU
To study the intrinsic connectivity of the BNST-A, we usedultraviolet (UV) GU at various sites with respect to the recorded cells.UV light pulses (50 ms) were delivered at 0.1 Hz by a light-emittingdiode (LED) source (365 nm, 60 mW, CoolLED, Andover, UK) via a�60 immersion objective, yielding UV light spots of �150 �m indiameter. The microscope rested on a computer-controlled motorizedstage, allowing us to move the light spot in a grid-like pattern (50- or110-�m steps) with respect to the recorded cell (Fig. 1A). At leastthree UV light pulses were applied at each site while the cells werekept at �90 mV with direct current injection. If a synaptic responsewas observed, the prestimulus membrane potential of recorded cellswas sequentially set to two additional values (�80 and �65 mV),each for at least three light stimuli, and more when response latencieswere variable. With this approach, �60 min was required to scan theentire BNST-A region in search of sites where UV light applicationelicited responses in a given postsynaptic cell. Postsynaptic potentials(PSPs) of �0.2 mV from a membrane potential of �90 mV wereexcluded.
4 mV
5-94
-81
-64
100 ms
-73
-91
2 40 mV
-72
-91
3
-72
-91
4
1 -74-92
D
LM
VAL
AM
Patchpipette
Target cell
AC
0.5 mm
0.5 mm
D
LM
V
AL
AM
AVNylon net
A spot of UV-light was moved in agrid-like patternover BNST
Patchpipette
Target cell
AC
BBA
-95 -85 -65
2
0
-2
-75IPS
P a
mpl
itude
Vm (mV)
100 ms
123
4
65
Fig. 1. Approach used to study intrinsic connections of the bed nucleus of the stria terminalis (BNST). A: glutamate uncaging (GU) was used to study the intrinsicconnectivity of the anterior portion of the BNST (BNST-A). Patch recordings of BNST-A cells were obtained under visual guidance. Pulses (50 ms) of ultraviolet(UV) light were delivered (0.1 Hz) at sites of 150 �m in diameter (white circles), uncaging glutamate at the stimulation site. The site of UV light stimulationwas systematically moved over the entire BNST-A in a grid-like pattern. Multiple light pulses were applied at each site and from different membrane potentials.Three sectors were distinguished in the BNST-A using fiber bundles readily identifiable in transilluminated slices: the anterior commissure (AC), dividing theBNST-A into dorsal and ventral (BNST-AV) regions, and the intra-BNST component of the stria terminalis, subdividing the dorsal portion into medial(BNST-AM) and lateral (BNST-AL) regions. B: spatial selectivity of GU. A patch recording of a BNST-AL neuron was obtained, and the morphology of thecell was revealed with biocytin (black: soma and dendrites; red: axon). UV light was applied over, and in the vicinity of, the recorded cell. The center of theUV light spot was moved in steps of 50 �m. Sites evoking direct suprathreshold responses are shown by red circles; white circles indicate subthreshold responses.No suprathreshold responses could be elicited when the center of the UV light spot was �150 �m from the cell. Examples of direct responses from sites 1–4are shown in 1–4. GU at sites 5 and 6 elicited GABA-A inhibitory postsynaptic potentials (IPSPs). Responses elicited from site 5 from different membranepotentials are shown in 5.
GU-evoked PSPs could easily be distinguished from spontaneoussynaptic events because the latter occurred infrequently and showedno temporal relationship with respect to the light stimulus. Neverthe-less, since anterior BNST neurons display spontaneous synapticevents (Dumont et al. 2005, 2008; Gosnell et al. 2011; Guo andRainnie 2010; Kash et al. 2008) that could be erroneously interpretedas GU-elicited PSPs, the following approach was used to distinguishspontaneous versus GU-evoked PSPs. For each cell, we estimated theaverage interval between spontaneous PSPs [interspontaneous PSPinterval (IsPSPI)] during the prestimulus period. Across all recordedcells, the average was 331 � 30 ms. We then used the IsPSPI of eachcell to determine the duration of a temporal window within which werequired PSPs in at least eight of nine trials to consider them as GUelicited. This “detection window” was set to a quarter of the cell’sIsPSPI. Within this window, the probability of getting eight or morespontaneous PSPs in nine traces by chance (i.e., the false positive rate)was P � 0.000107 (binomial test). We searched for GU-elicited PSPs ina 300-ms period after the stimulus onset by moving our detection windowin nonoverlapping steps. The average number of nonoverlapping detec-tion windows was four, resulting in a false positive probability of0.000428 per stimulation site. The average number of stimulus sites percell was 77, resulting in an average false positive probability of P �0.0329 per cell. Since we recorded a total of 75 cells, the number ofstimulus positions with PSPs falsely labeled as GU elicited was equal to(0.0329 � 75 � 2.467). Given that we observed 277 connections, thisrepresents �1% of false positives.
Analyses were performed offline with IGOR (Wavemetrics),Clampfit (Axon Instruments), Stimfit (http://www.stimfit.org/), andcustom-written software using Numpy and Scipy (http://www.scipy.org). Values are expressed as means � SE. Three complementarystatistical approaches were used to assess the significance of theresults. The first approach consisted of comparing the proportion ofcells showing particular response types [e.g., response vs. no responseor excitatory PSP (EPSP) vs. inhibitory PSP (IPSP)] with a �2-test.The second approach compared the proportion of stimulation siteseliciting particular response types across all recorded cells combinedwith a �2-test. Third, the results obtained with the latter approach wereverified by first computing the proportion of effective stimulation sitesfor each cell individually, averaging these for different response types,and comparing them with a paired t-test.
Biocytin Revelation for Morphological Identificationof Recorded Cells
At the conclusion of the recordings, slices were removed from thechamber and fixed for 1–3 days in 0.1 M PBS (pH 7.4) containing 4%paraformaldehyde. Slices were then embedded in agar (3%) andsectioned on a vibrating microtome at a thickness of 100 �m. Sectionswere washed several times in phosphate buffer (PB; 0.1 M, pH 7.4)and then transferred to H2O2 solution (0.5%) in PB for 15 min. Afternumerous washes in PB, sections were incubated for 12 h in a solutioncontaining 0.5% Triton X-100 plus 1% of solutions A and B of anABC kit (Vector, Burlingame, CA) in PB. The next day, sections werewashed in PB (5 � 10 min). Biocytin was visualized by incubating thesections in a 0.1 M PB solution that contained diaminobenzidinetetrahydrochloride (0.05%, Sigma), 2.5 mM nickel ammonium sul-fate (Fisher), and H2O2 (0.003%) for 5–10 min. Sections were thenwashed in PB (5 � 10 min), mounted on gelatin-coated slides, and airdried. Sections were then counterstained with cresyl violet and cov-erslipped with Permount for later reconstruction.
Nomenclature Used to Designate Different BNST Subregions
Individual BNST subnuclei cannot be identified with precision inunstained, living slices. Therefore, we used a simpler subdivision ofthe BNST-A in three regions, based on the position of major fiberbundles that can be easily identified in transilluminated slices: the
anterior commissure, dividing the BNST-A into dorsal and BNST-AVsectors, and the intra-BNST component of the stria terminalis, subdi-viding the dorsal portion into BNST-AM and BNST-AL regions. Thecorrespondence between our subdivisions of the BNST-A and thesubnuclei identified by Swanson and colleagues (Ju and Swanson1989; Ju et al., 1989) is as follows: the BNST-AV corresponds toSawnson’s anteroventral, fusiform, parastrial, and dorsomedial sub-nuclei plus the subcommisural zone; the BNST-AL corresponds toSawnson’s oval, juxtacapsular, and anterolateral subnuclei; and theBNST-AM corresponds to Swanson’s anterodorsal subnucleus.
RESULTS
Spatial Specificity of GU to Study Intra-BNST Connectivity
The usefulness of the GU method to study intrinsic BNSTconnections depends on whether it meets the following twocriteria: 1) that the rise in glutamate concentration produced byUV light be high and rapid enough to reliably fire neuronslocated where the light stimulus is applied and 2) that the decayof glutamate concentration with distance from the UV lightstimulus be sufficiently steep such that nearby neurons, notdirectly exposed to UV light, are not depolarized enough tofire. We first aimed to test whether the GU method met thesecriteria in the BNST.
To this end, the spot of UV light (150 �m in diameter and50 ms in duration) was centered over recorded cells (n � 10)and then gradually displaced away from this site in variousdirections (steps of 50 �m). In these recordings, the pipettesolution included biocytin to allow post hoc correlation ofmorphology and responsiveness to uncaged glutamate. A rep-resentative example of such an experiment is shown in Fig. 1B.The solid red and white circles in Fig. 1B mark the sites of UVlight application that elicited supra- or subthreshold responses,respectively. Application of UV light over the soma (Fig. 1B,4)and its immediate vicinity always elicited robust spiking. Thisis a direct response to uncaged glutamate. However, when thecenter of the UV spot was moved away from the soma, thesedirect depolarizing responses eventually became subthresholdor vanished. These variations (amplitude reduction vs. disap-pearance) depended on the exact position of the cells’ dendriteswith respect to the position of the light spot, with stimulilocated �150 �m from dendrites never eliciting spiking fromrest. For instance, in the case shown in Fig. 1B, GU elicitedspiking when the UV light was applied over the soma andproximal dendrites (red circles and Fig. 1B,2–4) but not whenthe stimulus was applied over more distal dendrites or at sites�150 �m from the soma and dendrites. At some of these sites(Fig. 1B,5 and 6), evidence of GABAergic synaptic connec-tions was obtained. Across all the cells tested in this manner,we did not observe a single case where spiking could beelicited from stimulation sites located �200 �m from theirsomata. Typically, direct responses vanished within 150 �m.Thus, these results suggest that the GU method has sufficientspatial selectivity to study intrinsic BNST connections.
Another important consideration when assessing the usefulnessof GU to study intrinsic BNST connections is whether depolar-izing PSPs can be distinguished from direct subthreshold re-sponses to uncaged glutamate. Indeed, when UV stimuli areapplied near recorded cells, it is possible that the evoked depo-larizations are not due to synaptically released transmitter but touncaged glutamate. Fortunately, these two types of responses
could be readily distinguished. Figure 2A,1 shows three superim-posed direct subthreshold responses to uncaged glutamate. As wastypically observed when the light stimulus was applied near therecorded cell, this direct response started shortly after the onset ofthe light pulse, rose gradually for its entire duration, and begandecaying shortly after its offset (Fig. 2A,1).
Compared with direct responses, synaptically evoked PSPshad a longer latency (PSPs: 79.0 � 4.4 ms vs. direct responses:8.3 � 0.1 ms), they peaked more rapidly (PSPs: 27.8 � 1.5 msvs. direct responses: 74.7 � 0.4 ms), the onset of their decayphase was not time locked to the offset of the light stimulus,and they sometimes showed conspicuous latency variations.The most direct illustration of the distinction between directand synaptic responses are cases where both phenomena areelicited by UV light application at the same site. In the exampleshown in Fig. 2A,2, the responses elicited by multiple consec-utive light stimuli are superimposed. All trials started with adirect response to uncaged glutamate that showed no latencyvariations. Superimposed on the decaying phase of these directresponses were depolarizing PSPs whose exact latencies andnumber varied from trial to trial. These latency variations,coupled to the differing time course of the two types ofresponses, leave no doubt as to their distinct origin. Of course,in the case of GABAergic PSPs, the distinction was furtherfacilitated by the fact that IPSPs reversed in polarity when thecells were depolarized (see below and Fig. 1B,5).
It should be noted that latency variations were typicallymuch smaller than those shown in Fig. 2A,2, as will becomeclear in the results shown in subsequent figures. However,since BNST-A neurons display spontaneous synaptic events(Dumont et al. 2005, 2008; Gosnell et al. 2011; Guo andRainnie 2010; Kash et al. 2008) that could be erroneouslyinterpreted as GU-elicited PSPs, the following approach wasused to distinguish spontaneous versus GU-evoked PSPs. Foreach target cell independently, we computed the frequency of
spontaneous PSPs and only considered PSPs that largely ex-ceeded the chance expected (see details in MATERIALS AND
METHODS). With the approach we used, the estimated falsepositive rate was 1%.
Distinguishing GABAergic and Glutamatergic PSPs Elicitedby Glutamate Uncaging
To identify the transmitters mediating GU-evoked PSPs, weprimarily relied on their reversal potentials. That is, PSPs withextrapolated reversal potentials near 0 mV were assumed to bemediated by ionotropic glutamatergic receptors, whereas PSPswith reversal potentials negative to �60 mV were classified asbeing mediated by GABA-A receptors (Fig. 1B,5). In severalcases, we verified these inferences by testing whether pre-sumed glutamatergic or GABAergic PSPs were sensitive todrugs that block non-N-methyl-D-aspartate (NMDA) glutamatereceptor [6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; 10�M); Fig. 2, B and C] or GABA-A [picrotoxin (75 �M); Fig.2D] responses. In all tested cases (EPSPs: n � 5 and IPSPs:n � 7), the pharmacological experiments confirmed our elec-trophysiological inference. Here, it should be noted that due tothe large rise in glutamate concentration produced by GU, thecompetitive receptor antagonist (CNQX) application did notblock (only delayed) direct suprathreshold responses to un-caged glutamate (Fig. 2B). In contrast, EPSPs elicited bysynaptically released glutamate were completely abolished(Fig. 2C). The differential sensitivity of direct versus synapti-cally mediated glutamatergic responses to CNQX has beenpreviously reported in a study relying on local pressure appli-cations of glutamate (Apergis-Schoute et al. 2007).
Mapping of Intrinsic BNST-A Connections With GU
We studied GU-evoked responses in 75 cells where long-term recording stability (�10% variations in input resistance
-60
-90
-60
-91
-96
100 ms
2 mV
-68
- 97
-66
Control
CNQX
Control
Picrotoxin
C
D
B
20 mV
20 ms
10 ms
Control
CNQX
2A1A Direct response Direct + PSPs5 mV
100 ms
Fig. 2. Distinguishing responses to uncaged glu-tamate versus synaptically released transmitters.A,1: direct subthreshold response to uncaged glu-tamate (three superimposed sweeps) from �90mV. Note the slow and invariant rising phase.Downward and upward arrows indicate the onsetand offset of UV light stimuli, respectively. A,2:case where a direct response and PSPs are trig-gered from the same stimulation site. Again, notethe slow and invariant rising phase of the directresponse that contrasts with the fast rising phase,variable latency, and number of evoked PSPs(four superimposed sweeps). B–D: example ofdirect suprathreshold response to uncaged gluta-mate and of PSPs in control artificial cerebrospi-nal fluid (aCSF; black), after the addition of6-cyano-7-nitroquinoxaline-2,3-dione (CNQX; Band C, gray) or picrotoxin (D, gray). The directsuprathreshold response to uncaged glutamate re-sisted CNQX (B), whereas indirect glutamatergicresponses were abolished (C). As shown in D,picrotoxin blocked IPSPs elicited by GU.
and �5 mV in resting potential) allowed extensive mapping oftheir intra-BNST connections with GU. These include 25BNST-AL, 28 BNST-AM, and 22 BNST-AV neurons. Con-sistent with earlier reports on the electroresponsive propertiesof BNST-A neurons (Guo et al. 2009; Hammack et al. 2007;Hazra et al. 2011, 2012), regular spiking and low-thresholdbursting neurons accounted for the vast majority of cells in thethree regions examined. However, because no differences inthe intrinsic connections were seen between physiological celltypes [�2(X, N � 75) � 7.64, P � 0.27], the results obtainedin the various cell types were pooled below. In all cellscombined, we tested the effects of UV light stimuli at 5,739sites, usually separated by 110 �m. Overall, 5.1% of thestimulation sites elicited a synaptic response. Typical examplesof intrinsic BNST-A connections evidenced with GU areshown for individual BNST-AM neurons in Fig. 3A, BNST-ALneurons in Fig. 3B, and BNST-AV neurons in Fig. 3C. We firstprovide a qualitative description of these response patterns;quantitative population analyses follow.
In Fig. 3A,1–C,1, the colored circles mark UV stimulationsites that elicited IPSPs (blue), EPSPs (red), or no response(white). As was typically the case, these three cells respondedto a minority of stimulation sites. Also representative of theoverall response pattern, a majority of PSPs elicited with GUwere IPSPs. Examples of evoked IPSPs are shown in Fig. 3, A,2, B,2, and C,2. Examples of EPSPs are shown in Fig. 3C,3.The rise time and duration of evoked PSPs varied within andbetween cells. This variability probably reflects a number offactors, such as electrotonic distances between the activatedsynapse(s) and soma as well as differences in the number ofspikes (and instantaneous firing frequency) of presynapticneurons recruited by GU. Of course, it is also possible that thenumber of presynaptic neurons varied between stimulationsites.
The inset in Fig. 3A,1 shows a different representation of theresults obtained in the same three neurons shown in Fig. 3. Thesame representation was used to illustrate the responsiveness of20 additional neurons in Fig. 4. We will refer to the results
-81
-95
-68
-94
-84
-69
-88
-80
-68
-95
-82
-65
1B1A
A3
B2
Target cell
Target cell
AM
AV
AM
AV
A2
B3
(Position 1)
(Position 32)(Position 15)
(Position 14)
25 ms
2 mV
C1
C2
C3
(Position 69)
(Position 80)
AM
AL
Target cell
-97
-84
-61
-96
-82
-67
IPSPs EPSPs
AC
AM
AV
Fig. 3. Examples of response patterns observed in BNST-AM (A), BNST-AL (B), and BNST-AV (C) neurons with GU. 1: photomicrographs of transilluminatedslices with the positions of recorded cells (target cell, white circles) and UV stimulation sites (open circles). White circles indicate the sites of UV light applicationthat elicited no response. At sites marked by blue circles, UV light elicited IPSPs. At sites marked by red circles, excitatory PSPs (EPSPs) were evoked. Examplesof IPSPs (A,2 and 3, B,2 and 3, and C,2) and EPSPs (C,3) are shown. Numbers indicate the prestimulus membrane potentials (in mV) at which responses wereobserved. Upward arrows indicate the offset of 50-ms UV light stimuli that elicited the responses. The inset in A,1 shows a graphical summary of connectionsfound in the three cells shown in A–C. Blue: inhibitory connections; red: excitatory connections.
obtained in these cells later on, when we describe the generaltrends identified in this study.
Intrinsic BNST-A Connections: Population Analyses
Consistent with the fact that GABAergic cells represent themain cell type in the BNST-A, most of the intrinsic connec-tions disclosed with GU were inhibitory. Figure 5 shows this intwo ways: first by showing a plot of the proportion of cells inwhich only IPSPs (blue circles), EPSPs (red circles), or both
(intersection between the circles) were evoked by GU (Fig.5A), and second by showing the proportion of stimulation sitesthat were effective in eliciting EPSPs or IPSPs across all cells(Fig. 5, B and C). With the first approach, the prevalence ofinhibitory connections was apparent when we considered allBNST-A cells together [�2(1, N � 75) � 7.18, P � 0.0074;Fig. 5A,1] or neurons in different parts of the BNST-A sepa-rately (Fig. 5A,2–4). However, the differing incidence ofinhibitory and excitatory connections was especially marked in
A1
D1C1
B1A2
D2C2
B2 B3
A3 B4A4 B5 B6
C3 C4
AMd AL
AMv AVD3
D6D4 D5
Fig. 4. Plots of intra-BNST connections evidenced with GU in the dorsal (A) and ventral (B) parts of the BNST-AM (BNST-AMd and BNST-AMv, respectively),BNST-AL (C), and BNST-AV (D). Each of the images (A,1–4, B,1–6, C,1–4, and D,1–6) shows a different cell. Note that in dorsally located BNST-AM cells,intrinsic inputs prevalently ran dorsoventrally. This trend was generally not seen in cells recorded in the BNST-AL (B), BNST-AV (D), or ventral part of theBNST-AM (C). While excitatory connections to neurons in the BNST-AL and the dorsal part of the BNST-AM were rare, they were frequently encountered inthe BNST-AV and ventrally located BNST-AM cells.
the BNST-AL [�2(1, N � 25) � 4.53, P � 0.033; Fig. 5A,2]and progressively less so in the BNST-AM (P � 0.12; Fig.5A,3) and BNST-AV (P � 0.6; Fig. 5A,4). This result patternwas confirmed using a different statistical approach where theproportion of effective stimulation sites was first determinedfor each cell, averaged across cells separately for IPSPs andEPSPs, and then compared with a paired t-test (Fig. 6A). Thiswas used for all BNST-A neurons combined [t(74) � 3.87,P � 0.0002] as well as separately for BNST-AL [t(24) � 2.77,P � 0.011], BNST-AM [t(27) � 2.65, P � 0.013], andBNST-AV neurons [t(21) � 1.29, P � 0.212].
The same conclusions emerged from the overall analysis ofeffective stimulation sites (Fig. 5, B and C). For intraregionalconnections (Fig. 5B), that is, cases where the stimulation sitesand recorded neurons were in the same BNST-A region, a�2-test revealed a significant dependence between responsetype (IPSP, EPSP, no response) and BNST region (BNST-AV,BNST-AM, and BNST-AV) [�2(4, N � 2959) � 9.62, P �0.047]. Post hoc tests showed that the proportion of stimulationsites eliciting IPSPs was higher in the BNST-AL than BNST-
AV [�2(2, N � 1979) � 7.56, P � 0.023] and that the pro-portions in the BNST-AM are intermediate and not signifi-cantly different from either the BNST-AL (P � 0.07) or BNST-AV (P � 0.23).
For interregional connections (Fig. 5C), namely, instanceswhere the recording and stimulation sites were located indifferent BNST-A regions, a more complex picture emerged.First, irrespective of the type of response observed (IPSPs orEPSPs), the incidence of effective stimulation sites was muchlower than that seen in intraregional connections [intraregional:7.41% and interregional: 2.63%, �2(1, N � 5,775) � 84.34,P � 0.0001]. As shown in Fig. 6B, the same conclusion wasreached using a different statistical approach, namely, firstdetermining the proportion of effective stimulation sites percell for intra- versus interregional connections, averaging thesevalues, and then comparing them with a paired t-test [t(74) �2.93, P � 0.004].
Second, IPSPs did not prevail in all interregional connec-tions. They did in connections from and to BNST-AL neurons(IPSPs: n � 46 and EPSPs: n � 16, binomial test, P � 0.0001),whereas in connections from and to the BNST-AV, the inci-dence of inhibitory connections could be equal to (BNST-AVto BNST-AM) or even lower than (BNST-AM to BNST-AV)that of excitatory connections. However, because the propor-tion of effective stimulation sites was low in these interregionalconnections, the latter difference did not reach statistical sig-nificance (binomial test, P � 0.5).
Properties of GU-evoked PSPs (rise time, amplitude, andduration) did not vary depending on the BNST-A region wherethe target cells were recorded or where the light stimuli wereapplied. This was true of EPSPs and IPSPs, even with signif-icance levels uncorrected for multiple comparisons. Figure 7shows the frequency distributions of IPSP and EPSP propertiesusing results obtained in the three BNST-A regions combined.It should be noted that for these analyses, compound PSPswere not included, only well-isolated PSPs (presumed single-axon PSPs) that could be measured unambiguously. However,note that the rise times of compound events, particularly ofIPSPs, were markedly slower than those of isolated PSPs.
From: AM AV AL AV AL AMn = 860 423 506 230 466 295
Excitatory responses Inhibitory responses
n = 2959 1069 980 910
B C
ALL (n = 75)
57% 23% 20%
I EA1 AV (n = 22)
46% 32% 23%
I EA4AM (n = 28)
57% 21% 21%
I EA3AL (n = 25)
68% 16% 16%
I EA2
Fig. 5. Relative incidence of inhibitory andexcitatory connections within the BNST-A.A: proportion of cells that responded withglutamatergic (red; E) and/or GABAergic(blue; I) PSPs to GU. Graphs from left toright: all recorded cells irrespective of loca-tion, BNST-AL cells, BNST-AM cells, andBNST-AV cells. B and C: proportion oftested stimulation sites eliciting GABAergic(blue) or glutamatergic (red) PSPs when thestimulation and recording sites were in thesame (B, intraregional) or different (C, in-terregional) sectors of the BNST-A.
6
4
2
0
Effe
ctiv
e S
timul
atio
n si
tes
All AL AM AV
(% o
f tes
ted
per
cell)
0Intra-
RegionalInter-
EPSPs IPSPsA B 12
8
4
Effe
ctiv
e S
timul
atio
n si
tes
(% o
f tes
ted
per
cell)
Fig. 6. Properties of intrinsic BNST connections. A: proportion of stimulationsites that elicited EPSPs (red) or IPSPs (blue) in BNST neurons. In contrast tothe results shown in Fig. 5, the proportion of effective sites was computed foreach cell separately and then averaged across cells (values are averages � SE).B: proportion of effective stimulation sites (EPSPs and IPSPs combined) inintranuclear (black bar) or internuclear connections (open bar). As in A, theproportion of effective sites was computed for each cell separately and thenaveraged across cells.
Heterogeneous Directionality and Polarity of IntrinsicBNST Connections
There were marked differences in the directionality of in-traregional connections in different sectors of the BNST-A.In the dorsal but not ventral portions of the BNST-AM, con-nections had a preferential directionality, with dorsal GU siteseliciting IPSPs in more ventrally located cells far more fre-quently than stimulation sites located ventrally to the recordedneurons (Fig. 4A). Although a few cells exhibited this phenom-enon in other BNST regions (Fig. 4, B,4, C,2, and D,3 and 5),no overall preferential directionality of connections emerged inthe BNST-AL (Fig. 4B,1–3 and 5 and 6), BNST-AV (Fig.4D,1,2,4, and 6), or the ventral part of BNST-AM (Fig. 4C,1,3,and 4).
Another obvious difference between the dorsal (Fig. 4A) andventral (Fig. 4C) parts of the BNST-AM was the incidence ofneurons receiving excitatory inputs. All but one of theBNST-AM neurons in which intrinsic glutamatergic connec-tions were disclosed (12 of 28 cells) were found in the ventralpart of the BNST-AM. As in the ventral part of BNST-AM, ahigh incidence of neurons receiving intrinsic glutamatergicinputs was found in the BNST-AV (12 of 22 cells, or 55%; Fig.4D), which was significantly higher than in the BNST-AL,where intrinsic glutamatergic inputs were infrequent (7 of 25cells, or 28%, Fisher’s exact test, P � 0.045).
Interregional connections were also asymmetric (Fig. 5C).Indeed, GU in the BNST-AM or BNST-AL elicited PSPs inBNST-AV cells (3.66%) much more frequently than in theopposite direction (1.38%, Fisher’s exact test, P � 0.007). Inaddition, while reciprocal connections were found between theBNST-AL and BNST-AM, BNST-AL to BNST-AM connectionswere more frequent than in the opposite direction (percent testedstimulated sites: BNST-AM to BNST-AL, 2.78%; BNST-AL toBNST-AM, 3.56%, Fisher’s exact test, P � 0.042).
Morphological Correlates
To test whether the contrasting directionality of intrinsicconnections observed in different BNST-A sectors was depen-dent on the morphology of BNST-A neurons, we filled 38neurons with biocytin (BNST-AM: n � 12, BNST-AL: n � 19,and BNST-AV, n � 7). Representative examples of biocytin-filled neurons are shown in Fig. 8A. After filling, slices wereplaced in fixative and resectioned at 100 �m, and biocytin wasrevealed. The morphology of recorded cells was reconstructedby performing drawings of all labeled elements found in thedifferent sections. Based on the matching position of bloodvessels and of the cut ends of dendritic and/or axonal segments,the labeling found in the different sections was then aligned.Figure 8B shows examples of such reconstructions.
Consistent with the prevalent dorsoventral connectivity seenin the dorsal portion of the BNST-AM, 11 of 12 cells recoveredfrom the BNST-AM (n � 12) had dendrites that extended morein the dorsal (370 � 157 �m) than the ventral (205 � 69 �m)direction (t-test, P � 0.047). Moreover, all cells located in thedorsal portion of the BNST-AM (n � 4) contributed ventrallydirected axons (BNST-AM; Fig. 8B,1 and 2). In contrast, cellsin other parts of the BNST-A displayed no consistent morpho-logical polarization (BNST-AL and BNST-AV; Fig. 8B,1 and2). Of note, whereas the dendritic arbors of BNST-AL (n � 19)and BNST-AM (n � 12) neurons were typically confined to theBNST region where their soma was located, BNST-AV neu-rons often (4 of 7 cells) had dendrites that extended dorsallybeyond the anterior commissure and into the BNST-AM orBNST-AL. This suggests that interregional connections target-ing BNST-AL or BNST-AM neurons typically depended onaxons that extended beyond the BNST sector where the parentsoma was located. In contrast, for BNST-AV neurons, this wasnot necessarily the case.
A IPSPsAmplitude noitaruDemitesiR
Amplitude Rise time DurationB EPSPs
0
10
20
30
0
10
20
30
40
10
20
30
0 2 4 6 0 10 20 30 40 0 20 40 60 80 100 120 1400
0
10
20
30
0
10
20
30
10
20
0 2 4 6 0 10 20 30 40 0 20 40 60 80 100 120 1400
)sm(noitaruD)sm(emitesiR)Vm(edutilpmA
)sm(noitaruD)sm(emitesiR)Vm(edutilpmA
% to
tal s
ampl
e%
tota
l sam
ple
% to
tal s
ampl
e%
tota
l sam
ple
% to
tal s
ampl
e%
tota
l sam
ple
n = 491.37 ± 0.13 mV
n = 4912.4 ± 0.8 ms
n = 4952.1 ± 3.8 ms
n = 531.89 ± 0.18 mV
n = 5346.9 ± 2.6 msn = 53
9.6 ± 0.42 ms
Fig. 7. Properties of IPSPs and EPSPs elic-ited by GU in intra- and inter-regional con-nections in all recorded neurons combined.A: frequency distributions of IPSP ampli-tudes (left), rise times (middle), and durations(right). B: frequency distribution of EPSPamplitudes (left), rise times (middle), anddurations (right). This analysis only includesconnections where individual PSPs could beresolved; compound events were excluded.All measures were performed from a mem-brane potential of �90 mV. PSP rise timescorrespond to times to half of peak ampli-tude.
It has been proposed that the BNST and central amygdala arepart of an anatomic entity termed the extended amygdala (Alheidand Heimer 1988; deOlmos and Heimer 1999). This concept isbased on similarities in neuronal morphology and transmittercontent (for a review, see McDonald 2003), common inputs fromthe basolateral amygdala (Krettek and Prince 1978a, 1978b; Pareet al. 1995; Savender et al. 1995; Dong et al. 2001a), as well as
overlapping projections to a network of motor and autonomicbrain stem nuclei thought to generate various aspects of fear/anxiety responses (Hopkins and Holstege 1978; Veening et al.1984; Holstege et al. 1985; Dong et al. 2000; Dong and Swanson2003, 2004, 2006a, 2006b, 2006c).
In contrast to the amygdala, however, the physiological orga-nization of the BNST is poorly understood. The BNST is com-posed of several subnuclei with much disagreement regarding
AC
AM AL
AV
AC
AM
AL
AV
0.5 mm
2B1B
2
1
100 µm
A1
50 µm
A2
Fig. 8. Morphological correlates of intrinsic connectivity. A: photomicrographs showing examples of BNST neurons labeled with biocytin (1, BNST-AV; 2,BNST-AL). B: drawings of eight BNST-A neurons (red, axons; black, somata and dendrites). The neurons labeled 1 in B,1 and 2 in B,2 are the same cells asshown in A,1 and A,2, respectively. Note that because all our recordings were performed in the BNST-A, the rostrocaudal position of recorded cells did not varymuch in our experiments (�250 �m).
their exact number and location (Andy and Stephan 1964; DeOlmos et al. 1985; Ju and Swanson 1989; Ju et al. 1989; Moga etal. 1989). However, it is commonly accepted that different BNSTregions form contrasting connections with the rest of the brain.This suggests a degree of functional specialization within theBNST, raising the question of whether different BNST regionsinteract with each other or whether they constitute independentprocessing modules.
The present study was undertaken to address this question,focusing on the intrinsic connections that exist in the BNST-A.Our study revealed the existence of asymmetric connectionswithin and between different parts of the BNST-A. Because theincidence of inhibitory and excitatory connections varied as afunction of the region contributing or receiving these intrinsicconnections, our findings raise the possibility that both coop-erative and competitive interactions take place within theBNST. Below, we summarize the pattern of intrinsic connec-tions evidenced in the present study and discuss these results inlight of earlier findings regarding the anatomy and physiologyof the BNST-A.
Nature of the Synaptic Connections Evidenced in thePresent Study
Several factors suggest that the vast majority of the synapticconnections evidenced in the present study are monosynapticand had an intrinsic origin (the pre- and postsynaptic neuronswere located within the BNST). All the glutamatergic EPSPswe elicited with the GU method were �6 mV in amplitude(mode of 1 mV). Since all the BNST neurons we recorded hada resting potential negative to �65 mV, it seems extremelyunlikely that such low-amplitude EPSPs could cause enoughdepolarization to reach spiking threshold (�49.8 � 0.3 mV) ina neuron not directly exposed to uncaged glutamate. Indeed,our control experiments (Fig. 1) revealed that unless the UVlight stimulus was applied directly over the recorded soma orthe proximal portion of the dendritic tree, it never elicitedspiking. As a result, it seems extremely unlikely that theresponses we observed were polysynaptic. Regarding the in-trinsic versus extrinsic origin of the connections, the vastmajority of the UV light stimuli used to uncage glutamate wereapplied entirely within the BNST. While some of the effectivestimulation sites straddled BNST boundaries, they accountedfor a minority of the connections we observed. Furthermore,many of these peripheral stimulation sites were located in theinternal capsule, which is largely devoid of neurons, and thelateral ventricle.
One confound we cannot completely exclude, however, isthe possibility that uncaged glutamate affected axon terminalscontributed by neurons located in the BNST or elsewhere.Indeed, prior studies have revealed that the BNST contains asubpopulation of GABAergic axon terminals expressing NMDAreceptors (Gracy and Pickel 1995; Paquet and Smith 2000).Under this scenario, uncaged glutamate would bind to presyn-aptic NMDA receptors and cause sufficient depolarization totrigger GABA release. While this phenomenon cannot be re-sponsible for the glutamatergic EPSPs we observed, it couldaccount for some of the GABAergic IPSPs. However, for thiseffect to occur, the axon terminal expressing NMDA receptorsand its postsynaptic target would have to be located where thelight stimulus is applied. Thus, this effect could only be in-
volved in cases where both a direct response to uncagedglutamate and an IPSP were observed. However, such in-stances were rare in our database (�7.3% of the connections)and therefore cannot account for the pattern of results weobtained. Finally, while there is clear evidence that axon ter-minals in the BNST express metabotropic glutamate receptors(Gosnell et al. 2011; Grueter and Winder 2005; Grueter et al.2006; Muly et al. 2007), it is unlikely that activation of thesereceptors by uncaged glutamate generated the fast synapticevents we examined because metabotropic glutamate receptorsare G protein-coupled receptors that exert slow modulatoryeffects but do not mediate fast PSPs.
Overall Pattern of Intrinsic BNST-A Connections
Intraregional connections. With respect to intraregionalconnections, a marked difference was found between the dorsalpart of the BNST-AM and the rest of the BNST-A. In most ofthe BNST-A, intraregional connections displayed no preferen-tial directionality. However, in the dorsal part of the BNST-AM,intrinsic connections had a predominant dorsoventral orientation(Fig. 9). Importantly, we found a parallel for this in the morphol-ogy of individual BNST-A neurons. Indeed, our reconstructionsof biocytin-filled cells revealed that most BNST-AM neuronswere morphologically polarized in a way consistent with thedirectionality of intrinsic connections, that is, their dendrites ex-tended more in the dorsal than in the ventral direction and con-tributed axons that coursed ventrally. In contrast, neurons recov-ered from other sectors of the BNST-A showed no consistentorientation of their axons and dendrites.
Although a prior study has examined the connectivity of theBNST-AM with Phaseolus vulgaris leucoagglutinin (Dongand Swanson 2006a), it did not comment on the peculiarorganization we observed in the dorsal part of the BNST-AM.However, this is likely due to technical limitations inherent to
AM AL
AVGlu
GABA
Fig. 9. Overall pattern of intrinsic BNST-A connections revealed with GU. Redand blue arrows correspond to glutamatergic (Glu) and GABAergic (GABA)connections, respectively. For intraregional connections, the number of blueand red arrows approximates the relative frequency of inhibitory and excitatoryconnections, respectively. For interregional connections, the thickness of thearrows was adjusted to represent the relative incidence of connections.
tracing techniques. To disclose the type of organization weobserved with GU, one would need to perform extremely smalltracer injections, which is nearly impossible.
Another finding that emerged from our study is that therelative incidence of GABAergic and glutamatergic connec-tions varied markedly in the different regions examined. Al-though GABAergic connections were prevalent overall, insome BNST-A regions glutamatergic connections were nearlyas frequent. The incidence of glutamatergic connections waslowest in the BNST-AL and dorsal part of the BNST-AM. Incontrast, in the BNST-AV and ventral region of BNST-AM,they accounted for about half of the connections (Fig. 8).
The varying incidence of GABAergic and glutamatergic con-nections in different BNST-A regions is consistent with the resultsof previous reports that used immunohistochemistry (Esclapez etal. 1993; Hur and Zaborszky 2005; Sun and Cassell 1993) or insitu hybridization (Day et al. 1999; Poulin et al. 2009; Kudo et al.2012) to study the distribution of neurons that are GABAergic(expressing mRNA for glutamic acid decarboxylase 65 and/or 67)and/or glutamatergic [expressing mRNA for vesicular glutamatetransporter 2 (VGLUT2)] in the BNST. Considered together,these studies demonstrated that GABAergic neurons are abundantin all divisions of the BNST-A, whereas the incidence of gluta-matergic neurons show marked interregional variations. In partic-ular, consistent with our observations, no (or very few) VGLUT2-positive cells were seen in the BNST-AL, whereas a significantnumber was seen in the BNST-AV and BNST-AM.
Interregional connections. The connections between differ-ent BNST-A regions were asymmetric, that is, for all pairs ofregions examined, connections were significantly more fre-quent in one direction than the other (Fig. 9). This was the casefor all interregional projections involving the BNST-AL: therewas a higher incidence from the BNST-AL to BNST-AM andBNST-AV than from the latter two to the BNST-AL. Con-versely, all interregional connections ending in the BNST-AVwere stronger than the reciprocal connections: there was alower incidence from the BNST-AV to BNST-AM andBNST-AL than from the latter two to the BNST-AV.
This pattern of connections is consistent with the findings ofearlier anterograde (Dong and Swanson 2004, 2006a) andretrograde (Shin et al. 2008) tracing studies. Indeed, thesestudies revealed that components of the BNST-AV receiveconvergent inputs from the BNST-AL and BNST-AM and thatsubregions of the BNST-AL and BNST-AM are reciprocallyconnected. However, it is difficult to compare the relativestrength of the connections evidenced here with that seen intracing studies because the size of the various Phaseolusvulgaris leucoagglutinin injection sites was not constant. Nev-ertheless, the results of Swanson and colleagues appear gener-ally consistent with the notion that BNST-AL to BNST-AMconnections (Dong and Swanson 2004) are stronger than in theopposite direction (Dong and Swanson 2006a). Also consistentwith our findings, BNST-AL projections to the BNST-AV(Dong and Swanson 2004) appear denser than in the oppositedirection (Dong et al. 2001b).
In the present study, interregional connections also differedin the relative incidence of GABAergic and glutamatergicinputs (Fig. 9). Paralleling the intraregional connections, theprojections of the BNST-AL to BNST-AM or BNST-AV werealmost exclusively GABAergic. Similarly, return projectionsfrom the BNST-AM and BNST-AV to BNST-AL were also
characterized by a scarcity of glutamatergic connections. Incontrast, the connections between BNST-AM and BNST-AVincluded both glutamatergic and GABAergic projections.
The notion that a proportion of GABAergic and glutamater-gic BNST-A neurons contribute axons that extend beyond theconfines of the subregion where their somata is located issupported by results from earlier anatomic and physiologicalstudies. For instance, it has been reported that GABAergic cellsin the BNST-AM and BNST-AV project to the paraventricularhypothalamic nucleus (Herman et al. 2004; Radely et al. 2009;Radley and Sawchenko 2011). Similarly, it has been reportedthan GABAergic BNST-AL neurons project to the centralamygdala (Sun and Cassell 1993). Finally, with respect toexcitatory outputs, it has been shown that the BNST-AVcontains glutamatergic neurons that project to the ventraltegmental area (Georges and Aston-Jones 2001, 2002; Jalabertet al. 2009; Kudo et al. 2012).
Functional Implications
The pattern intrinsic connectivity disclosed in the present studyimplies that different BNST-A sectors do not act independently.In particular, because the projections of the BNST-AL toBNST-AM and BNST-AV are purely inhibitory and stronger thanthe reciprocating pathways, the BNST-AL is strategically posi-tioned to determine, or at least modulate, activity levels in the restof the BNST-A. This suggests an arrangement where the BNST-AL, via its inhibitory projections to the BNST-AM and BNST-AV, acts as a gating mechanism for many BNST-A outputs.When BNST-AL activity is high, this tends to reduce firing ratesin BNST-AM and BNST-AV neurons. Conversely, a reduction inBNST-AL activity could cause a positive (or self-reinforcing)feedback effect where disinhibition of the BNST-AM fromBNST-AL inputs would increase return inhibitory projectionsfrom the BNST-AM to BNST-AL, resulting in a further disinhi-bition of the BNST-AM, and so on. In addition to these compet-itive interactions, our findings raise the possibility that othersectors of the BNST-A entertain cooperative relations. Indeed, thehigh incidence of glutamatergic connections between theBNST-AV and ventral part of the BNST-AM suggest that neu-rons in these two regions may mutually enhance their excitability.
At present, it is difficult to assess how significant the impactof intrinsic BNST-A connections might be. While the inci-dence of interregional connections was relatively low, it waslikely underestimated because many connections, particularlythe longer ones, are lost when slices are prepared. In addition,it is likely that intrinsic inputs ending in the distal dendrites ofBNST neurons could not be detected due to electrotonicattenuation. Besides, the influence of intrinsic BNST connec-tions depends on a variety of factors, including moment-to-moment variations in the activity of extrinsic afferents as wellas modulatory inputs. In any event, the above considerationshighlight the difficulty of interpreting lesion and pharmacobe-havioral studies. Depending what exact BNST-A sector islesioned or inactivated, opposite behavioral consequences canemerge.
GRANTS
This work was supported by National Institute of Mental Health GrantR01-MH-098738.
No conflicts of interest, financial or otherwise, are declared by the author(s).
AUTHOR CONTRIBUTIONS
Author contributions: H.K.T. and O.E.R.-S. performed experiments; H.K.T.and D.P. analyzed data; H.K.T. interpreted results of experiments; H.K.T. andD.P. prepared figures; H.K.T. and O.E.R.-S. edited and revised manuscript;H.K.T., O.E.R.-S., and D.P. approved final version of manuscript; D.P.conception and design of research; D.P. drafted manuscript.
REFERENCES
Alheid GF, Heimer L. New perspectives in basal forebrain organization ofspecial relevance for neuropsychiatric disorders: the striatopallidal, amyg-daloid, and corticopetal components of substantia innominata. Neuroscience27: 1–39, 1988.
Andy OJ, Stephan H. The septum of the cat. Springfield: Thomas, 1964.Apergis-Schoute J, Pinto A, Pare D. Muscarinic control of long-range
GABAergic inhibition within the rhinal cortices. J Neurosci 27: 4061–4071,2007.
Cullinan WE, Herman JP, Watson SJ. Ventral subicular interaction with thehypothalamic paraventricular nucleus: evidence for a relay in the bednucleus of the stria terminalis. J Comp Neurol 13: 1258–1279, 1993.
Davis M, Walker DL, Miles L, Grillon C. Phasic vs sustained fear in rats andhumans: role of the extended amygdala in fear vs anxiety. Neuropsychop-harmacology 35: 105–135, 2010.
Day H, Curran EJ, Watson S Jr, Akil H. Distinct neurochemical populationsin the rat central nucleus of the amygdala and bed nucleus of the striaterminalis: evidence for their selective activation by interleukin-1�. J CompNeurol 413: 113–128, 1999.
De Olmos JS, Heimer L. The concepts of the ventral striatopallidal systemand extended amygdala. Ann NY Acad Sci 877:1–32, 1999.
De Olmos JS, Alheid GF, Beltramino CA. Amygdala. In: The Rat NervousSystem. Forebrain and Midbrain, edited by Paxinos G. Orlando, FL:Academic, 1985, vol. 1, p. 509–603.
Dong HW, Swanson LW. Projections from the rhomboid nucleus of the bednucleus of the stria terminalis: implications for cerebral hemisphere regu-lation of ingestive behaviors. J Comp Neurol 463: 795–799, 2003.
Dong HW, Swanson LW. Organization of axonal projections from theanterolateral area of the bed nuclei of the stria terminalis. J Comp Neurol468: 277–298, 2004.
Dong HW, Swanson LW. Projections from bed nuclei of the stria terminalis,anteromedial area: cerebral hemisphere integration of neuroendocrine, au-tonomic, and behavioral aspects of energy balance. J Comp Neurol 494:142–178, 2006a.
Dong HW, Swanson LW. Projections from bed nuclei of the stria terminalis,magnocellular nucleus: implications for cerebral hemisphere regulation ofmicturition, defecation, and penile erection. J Comp Neurol 494: 108–141,2006b.
Dong HW, Swanson LW. Projections from bed nuclei of the stria terminalis,dorsomedial nucleus: implications for cerebral hemisphere integration ofneuroendocrine, autonomic, and drinking responses. J Comp Neurol 494:75–107, 2006c.
Dong HW, Petrovich GD, Swanson LW. Organization of projections fromthe juxtacapsular nucleus of the BST: a PHAL study in the rat. Brain Res859: 1–14, 2000.
Dong HW, Petrovich GD, Swanson LW. Topography of projections fromamygdala to bed nuclei of the stria terminalis. Brain Res Brain Res Rev 38:192–246, 2001a.
Dong HW, Petrovich GD, Watts AG, Swanson LW. Basic organization ofprojections from the oval and fusiform nuclei of the bed nuclei of the striaterminalis in adult rat brain. J Comp Neurol 436: 430–455, 2001b.
Dumont EC, Rycroft BK, Maiz J, Williams JT. Morphine produces circuit-specific neuroplasticity in the bed nucleus of the stria terminalis. Neurosci-ence 153: 232–239, 2008.
Dumont EC, Mark GP, Mader S, Williams JT. Self-administration en-hances excitatory synaptic transmission in the bed nucleus of the striaterminalis. Nat Neurosci 8: 413–414, 2005.
Erb S, Stewart J. A role for the bed nucleus of the stria terminalis, but not theamygdala, in the effects of corticotropin-releasing factor on stress-inducedreinstatement of cocaine seeking. J Neurosci 19: C1–C6, 1999.
Esclapez M, Tillakaratne N, Tobin AJ, Houser CR. Comparative localiza-tion of two forms of glutamic acid decarboxylase with nonradioactive in situhybridization methods. J Comp Neurol 331: 339–362, 1993.
Fendt M, Siegl S, Steiniger-Brach B. Noradrenaline transmission within theventral bed nucleus of the stria terminalis is critical for fear behaviorinduced by trimethylthiazoline, a component of fox odor. J Neurosci 25:5998–6004, 2005.
Georges F, Aston-Jones G. Potent regulation of midbrain dopamine neuronsby the bed nucleus of the stria terminalis. J Neurosci 21: RC160, 2001.
Georges F, Aston-Jones G. Activation of ventral tegmental area cells by thebed nucleus. J Neurosci 22: 5173–5187, 2002.
Gosnell HB, Silberman Y, Grueter BA, Duvoisin RM, Raber J, WinderDG. mGluR8 modulates excitatory transmission in the bed nucleus of thestria terminalis in a stress-dependent manner. Neuropsychopharmacology36: 1599–1607, 2011.
Gray KN, Pickel VM. Comparative ultrastructural localization of theNMDAR1 glutamate receptor in the rat basolateral amygdala and bednucleus of the stria terminalis. J Comp Neurol 362: 71–85, 1995.
Grueter BA, Winder DG. Group II and III metabotropic glutamate receptorssuppress excitatory synaptic transmission in the dorsolateral bed nucleus ofthe stria terminalis. Neuropsychopharmacology 30: 1302–1311, 2005.
Grueter BA, Gosnell HB, Olsen CM, Schramm-Sapyta NL, Nekrasova T,Landreth GE, Winder DG. Extracellular-signal regulated kinase 1-depen-dent metabotropic glutamate receptor 5-induced long-term depression in thebed nucleus of the stria terminalis is disrupted by cocaine administration. JNeurosci 26: 3210–3219, 2006.
Guo JD, Rainnie DG. Presynaptic 5-HT1B receptor-mediated serotonergicinhibition of glutamate transmission in the bed nucleus of the stria termi-nalis. Neuroscience 165: 1390–1401, 2010.
Guo JD, Hammack SE, Hazra R, Levita L, Rainnie DG. Bi-directionalmodulation of bed nucleus of stria terminalis neurons by 5-HT: molecularexpression and functional properties of excitatory 5-HT receptor subtypes.Neuroscience 164: 1776–1793, 2009.
Hammack SE, Mania I, Rainnie DG. Differential expression of intrinsicmembrane currents in defined cell types of the anterolateral bed nucleus ofthe stria terminalis. J Neurophysiol 98: 638–656, 2007.
Hazra R, Guo JD, Ryan SJ, Jasnow AM, Dabrowska J, Rainnie DG. Atranscriptomic analysis of type I-III neurons in the bed nucleus of the striaterminalis. Mol Cell Neurosci 46: 699–709, 2011.
Hazra R, Guo JD, Dabrowska J, Rainnie DG. Differential distribution ofserotonin receptor subtypes in BNST(ALG) neurons: modulation by unpre-dictable shock stress. Neuroscience 225: 9–21, 2012.
Herman JP, Mueller NK, Figueiredo H. Role of GABA and glutamatecircuitry in hypothalamo-pituitary-adrenocortical stress integration. Ann NYAcad Sci 1018: 35–45, 2004.
Holstege G, Meiners L, Tan K. Projections of the bed nucleus of the striaterminalis to the mesencephalon, pons, and medulla oblongata in the cat.Exp Brain Res 58: 379–391, 1985.
Hopkins DA, Holstege G. Amygdaloid projections to the mesencephalon,pons and medulla oblongata in the cat. Exp Brain Res 32: 529–547, 1978.
Hur EE, Zaborszky L. Vglut2 afferents to the medial prefrontal and primarysomatosensory cortices: a combined retrograde tracing in situ hybridizationstudy. J Comp Neurol 483: 351–373, 2005.
Jalabert M, Aston-Jones G, Herzog E, Manzoni O, Georges F. Role of thebed nucleus of the stria terminalis in the control of ventral tegmental areadopamine neurons. Prog Neuropsychopharmacol Biol Psychiatry 33: 1336–1346, 2009.
Ju G, Swanson LW. Studies on the cellular architecture of the bed nuclei ofthe stria terminalis in the rat: I. Cytoarchitecture. J Comp Neurol 280:587–602, 1989.
Ju G, Swanson LW, Simerly RB. Studies on the cellular architecture of thebed nuclei of the stria terminalis in the rat: II. Chemoarchitecture. J CompNeurol 280: 603–621, 1989.
Kash TL, Nobis WP, Matthews RT, Winder DG. Dopamine enhances fastexcitatory synaptic transmission in the extended amygdala by a CRF-R1-dependent process. J Neurosci 28: 13856–13865, 2008.
Krawczyk M, Georges F, Sharma R, Mason X, Berthet A, Bézard E,Dumont EC. Double-dissociation of the catecholaminergic modulation ofsynaptic transmission in the oval bed nucleus of the stria terminalis. JNeurophysiol 105: 145–153. 2011.
Krettek JE, Price JL. A description of the amygdaloid complex in the rat andcat with observations on intra-amygdaloid axonal connections. J CompNeurol 178: 255–280, 1978a.
Krettek JE, Price JL. Amygdaloid projections to subcortical structures withinthe basal forebrain and brainstem in the rat and cat. J Comp Neurol 178:225–254, 1978b.
Kudo T, Uchigashima M, Miyazaki T, Konno K, Yamasaki M, YanagawaY, Minami M, Watanabe M. Three types of neurochemical projection fromthe bed nucleus of the stria terminalis to the ventral tegmental area in adultmice. J Neurosci 32: 18035–18046, 2012.
Larriva-Sahd J. Histological and cytological study of the bed nuclei of thestria terminalis in adult rat. II. Oval nucleus: extrinsic inputs, cell types,neuropil, and neuronal modules. J Comp Neurol 497: 772–807, 2006.
McDonald AJ. Neurons of the bed nucleus of the stria terminalis: a golgi studyin the rat. Brain Res Bull 10: 111–120, 1983.
McDonald AJ, Shammah-Lagnado SJ, Shi C, Davis M. Cortical afferents tothe central extended amygdala. Ann NY Acad Sci 877: 309–338, 1999.
McDonald AJ. Is there an amygdala and how far does it extend? Ananatomical perspective. Ann NY Acad Sci 985: 1–21, 2003.
McElligott ZA, Klug JR, Nobis WP, Patel S, Grueter BA, Kash TL,Winder DG. Distinct forms of Gq-receptor-dependent plasticity of excit-atory transmission in the BNST are differentially affected by stress. ProcNatl Acad Sci USA 107: 2271–2276, 2010.
Moga MM, Saper CB, Gray TS. Bed nucleus of the stria terminalis:cytoarchitecture, immunohistochemistry, and projection to the parabrachialnucleus in the rat. J Comp Neurol 283: 315–332, 1989.
Muly EC, Mania I, Guo JD, Rainnie DG. Group II metabotropic glutamatereceptors in anxiety circuitry: correspondence of physiological response andsubcellular distribution. J Comp Neurol 505: 682–700, 2007.
Paquet M, Smith Y. Presynaptic NMDA receptor subunit immunoreactivityin GABAergic terminals in rat brain. J Comp Neurol 423: 330–337, 2000.
Paré D, Smith Y, Paré JF. Intra-amygdaloid projections of the basolateral andbasomedial nuclei in the cat: phaseolus vulgaris-leucoagglutinin anterogradetracing at the light and electron microscopic level. Neuroscience 69: 567–583, 1995.
Polston EK, Gu G, Simerly RB. Neurons in the principal nucleus of the bednuclei of the stria terminalis provide a sexually dimorphic GABAergic inputto the anteroventral periventricular nucleus of the hypothalamus. Neurosci-ence 123: 793–803, 2004.
Poulin JF, Arbour D, Laforest S, Drolet G. Neuroanatomical characteriza-tion of endogenous opioids in the bed nucleus of the stria terminalis. ProgNeuropsychopharmacol Biol Psychiatry 33: 1356–1365, 2009.
Prewitt CM, Herman JP. Anatomical interactions between the central amyg-daloid nucleus and the hypothalamic paraventricular nucleus of the rat: adual tract-tracing analysis. J Chem Neuroanat 15: 173–185, 1998.
Radley JJ, Gosselink KL, Sawchenko PE. A discrete GABAergic relaymediates medial prefrontal cortical inhibition of the neuroendocrine stressresponse. J Neurosci 29: 7330–7340, 2009.
Radley JJ, Sawchenko PE. A common substrate for prefrontal and hippocam-pal inhibition of the neuroendocrine stress response. J Neurosci 31: 9683–9695, 2011.
Savander V, Go CG, LeDoux JE, Pitkänen A. Intrinsic connections of therat amygdaloid complex: projections originating in the basal nucleus. JComp Neurol 361: 345–368, 1995.
Shin JW, Geerling JC, Loewy AD. Inputs to the ventrolateral bed nucleus ofthe stria terminalis. J Comp Neurol 511: 628–657, 2008.
Sofroniew MV. Direct reciprocal connections between the bed nucleus of thestria terminalis and dorsomedial medulla oblongata: evidence from immu-nohistochemical detection of tracer proteins. J Comp Neurol 213: 388–405,1983.
Sun N, Cassell MD. Intrinsic GABAergic neurons in the rat central extendedamygdala. J Comp Neurol 330: 381–404, 1993.
Ulrich-Lai YM, Herman JP. Neural regulation of endocrine and autonomicstress responses. Nat Rev Neurosci 10: 397–409, 2009.
Veening JG, Swanson LW, Sawchenko PE. The organization of projectionsfrom the central nucleus of the amygdala to brainstem sites involved incentral autonomic regulation: a combined retrograde transport-immunohis-tochemical study. Brain Res 303: 337–357, 1984.
![Anterior cingulate cortex hypofunction causes anti-social ... · 1 day ago · also referred to as the threat circuit [4], centres around the bed nucleus of the stria terminalis](https://static.documents.pub/doc/80x56/603a1e6e4d14772ffc200fc9/anterior-cingulate-cortex-hypofunction-causes-anti-social-1-day-ago-also.jpg)
