Despite the high prevalence1,2
of anxiety disorders, the underlying neural circuitry is incompletely understood. Available treatments are inconsistently effective or, in the case of benzodiazepines, addictive and linked to significant side-effects including cognitive impairment and respiratory suppression17
, pointing to the need for deeper understanding of anxiety control mechanisms in the mammalian brain.
Although amygdala microcircuitry for conditioned fear has been optogenetically dissected18,19
, the causal underpinnings of unconditioned anxiety11
have not yet been investigated with cellular precision. Pointing to the need for precise optogenetic exploration, the amygdala is composed of functionally and morphologically heterogeneous subnuclei with complex interconnectivity. The BLA is primarily glutamatergic (~90%)20, 21
while the CeA, which encompasses the centrolateral (CeL) and centromedial (CeM) nuclei, consists of ~95% GABAergic medium spiny neurons22
. The primary output region of the amygdala is the CeM23,24
which (when chemically or electrically excited) mediates autonomic and behavioral responses associated with fear and anxiety via projections to the brainstem25
. Because patients with generalized anxiety disorder may have abnormal activity arising from the BLA and CeM11
, and since BLA neurons excite GABAergic CeL neurons26
that provide feed-forward inhibition onto CeM “output” neurons6,18,19
, we considered that the BLA-CeL-CeM circuit could be causally involved in anxiety. However, BLA pyramidal neurons as a whole could have varied and antagonistic roles in diverse projections throughout the brain, with targets including the bed nucleus of the stria terminalis (BNST), nucleus accumbens, hippocampus and cortex26
We therefore developed a method to selectively control BLA terminals in the CeA (Supplementary Methods
). BLA glutamatergic projection neurons were transduced with an adeno-associated virus serotype 5 (AAV5
) carrying codon-optimized channelrhodopsin (ChR2)-eYFP under control of the CaMKIIαeYFP promoter followed by unilateral implantation of a bevelled guide cannula to allow preferential illumination of the non-transduced CeL (Supplementary Figs. 1,2
). In vivo
electrophysiological recordings were used to determine illumination parameters for selective control of those BLA terminals in the CeA without nonspecific control of all BLA somata (Supplementary Fig. 3
To investigate the functional role of the BLA-CeA pathway in anxiety, we probed freely moving mice under projection-specific optogenetic control in two well-validated27
anxiety assays: the elevated-plus maze (EPM) and the open-field test (OFT; ). Mice display anxiety-related behaviors in open spaces; therefore increased time spent in EPM open arms or in the OFT center is interpreted as reduced anxiety27
. To test whether anxiety-related behaviors could be related to activation of the BLA-CeA projection, and not all BLA somata as a whole, we compared mice receiving projection-specific photostimulation (ChR2:BLA-CeA; ) to a group with identical illumination parameters transduced with a control virus (eYFP:BLA-CeA), and to another control group expressing ChR2 in the BLA receiving direct illumination of the BLA (ChR2:BLA (somata)). Photostimulation of BLA terminals in the CeA (ChR2:BLA-CeA) increased open-arm time (F1,42
=69.09, p<0.00001; ) and probability of open-arm entry from the maze center (F1,42
=24.69, p<0.00001: inset; Supplementary Movie
) on the EPM, as well as increased center time in the OFT (F1,105
=24.46, p<0.00001; ), reflecting anxiety reduction, without impaired locomotion (Supplementary Fig. 4
). In contrast, the ChR2:BLA(somata)group showed reduced open arm time (F1,42
=6.20, p=0.02; ) and probability of open arm entry (F1,42
=5.15, p=0.03) during photostimulation relative to eYFP:BLA-CeA controls, reflecting a distinct anxiogenic effect. Thus, selective illumination of BLA projections to the CeA, but not of BLA somata nonspecifically, produced an acute, rapidly-reversible anxiolytic effect.
Projection-specific excitation of BLA terminals in the CeA induces acute reversible anxiolysis
Next we investigated the physiological basis of this light-induced anxiolysis. We considered that preferential photostimulation of BLA terminals in the CeL could activate CeL neurons and exert feed-forward inhibition onto brainstem-projecting CeM output neurons18,19
to implement anxiolysis. To test this, we undertook in vivo
experiments, with light delivery protocols matched to those delivered in the behavioral experiments, using activity-dependent immediate early gene (c-fos
) expression to track neuronal activation. We quantified the proportion of neurons in the BLA, CeL and CeM (Supplementary Fig. 5
) within each group that expressed eYFP or showed c-fos immunoreactivity. Opsin expression was specific to BLA glutamatergic neurons, and was not observed in intercalated cells (Supplementary Fig. 6
). No group differences were detected in the proportion of eYFP-positive cells within each region (Supplementary Fig. 5
). We found a significantly higher proportion of c-fos-positive BLA cells in the ChR2:BLA(somata) group (F2,9
=10.12, p<0.01), relative to ChR2:BLA-CeA or eYFP:BLA-CeA groups (p<0.01 and p<0.05, respectively), but no detectable difference between the ChR2:BLA-CeA and eYFP:BLA-CeA groups, indicating that the bevelled cannula shielding effectively prevented BLA somata photostimulation. A higher proportion of CeL neurons expressed c-fos in the ChR2:BLA-CeA group relative to the eYFP:BLA-CeA group (F2,9
=4.54, p=0.04), but not the ChR2:BLA(somata)group (Supplementary Fig. 5
). Thus, the in vivo
illumination of BLA-CeA projections that triggered acute anxiolysis was found to excite CeL neurons without activating all BLA somata nonspecifically.
To test the hypothesis that preferential illumination of BLA-CeL terminals induced feed-forward inhibition of CeM output neurons, we combined whole-cell patch-clamp electrophysiology with two-photon imaging to visualize the microcircuit while simultaneously probing the functional relationships among these cells during projection-specific optogenetic control (). BLA neurons showed high-fidelity spiking to direct illumination (). Illumination of BLA terminals in CeL elicited excitatory responses in both weakly- and strongly-connected CeL cells (n=16; ).
Projection-specific excitation of BLA terminals in the CeA activates CeL neurons and elicits feed-forward inhibition of CeM neurons
To test whether illumination of BLA-CeL synapses could block CeM spiking via feed-forward inhibition from CeL neurons, we recorded from CeM neurons while selectively illuminating BLA-CeL synapses (). Indeed, we observed potent spiking inhibition in CeM cells upon illumination of BLA-CeL terminals (; F2,11
=15.35, p=0.004). shows CeM responses recorded during illumination of ChR2-expressing BLA neurons in the CeM; importantly, the very same CeM neurons (n=7) showed net excitation upon broad illumination of BLA inputs to the CeM (), but showed net inhibition upon selective illumination of BLA inputs to the CeL (). These data from a structurally- and functionally-identified microcircuit25
illustrate that the balance of direct and indirect inputs from the BLA to the CeM can modulate CeM activity. We then examined whether overlapping or distinct populations of BLA neurons projected to the CeL and CeM in the mouse (Supplementary Fig. 7
) by two-photon imaging in 350 μm-thick coronal slices. Of the BLA neurons sampled (n=18, Supplementary Fig. 7
); 44% projected to the CeL alone and 17% projected to the CeM alone, with only one cell observed to project to both the CeL and the CeM.
Since in vivo
c-fos assays had indicated that illumination of BLA terminals in the CeL sufficed to drive excitation of postsynaptic CeL neurons but not to recruit efficiently BLA neurons as a whole, we next investigated the properties of optogenetically-driven terminal stimulation in this microcircuit using whole-cell recordings. We first recorded from BLA pyramidal neurons expressing ChR2 and moved a restricted light spot (~125 μm-diameter) in 100 μm steps from the cell soma, both in a direction over the visually-identified axon and in a direction where there was no axon collateral, illustrating the spatial properties of light scattering in this circuit (); spiking fidelity in the BLA neuron and evoked inward currents are summarized (). Next, we found that typical photostimulation parameters drove reliable transmission when delivered to BLA-CeA synapses (assayed with recordings from postsynaptic CeL neurons; ; Supplementary Fig. 8
); in contrast, when recording from the BLA somata instead ~300 μm from the light spot, we did not observe reliable antidromically-driven action potential firing (only ~5% reliability) despite use of the very same BLA-CeA synapse illumination conditions that elicited 100% reliable transmitter release from the illuminated terminals and the same cells known to spike robustly in response to somatic illumination (). These results were consistent with the c-fos immunoreactivity and behavioral data () and held even upon bath application of GABA and glutamate receptor antagonists (n=7) to eliminate local circuitry effects. The marked difference between effective synaptic transmission and antidromic spiking fidelity (p=0.0039; , inset) reveals that optogenetically-driven vesicle release may occur in the absence of reliable antidromic drive, a potentially useful property that may relate to projection parameters such as axon caliber and myelination status (optogenetic stimuli will recruit thinner axons more efficiently than electrical stimuli), as well as experimental light intensity and spatial restriction properties.
Light-induced anxiolytic effects are attributable to activation of BLA-CeL synapses
To further confirm that the anxiolytic effect was due to the selective activation of BLA-CeL projections alone, and not BLA axons passing through the CeA, nor back-propagation of action potentials to BLA cell bodies that would then innervate all BLA projection target regions, we tested whether local glutamate receptor antagonism would attenuate light-induced anxiolytic effects. This distinction is critical, since previous reports that CeA lesions that alter anxiety are confounded by the ablation of BLA projections to the BNST passing through CeA28
. In a separate group of mice, we selectively illuminated BLA-CeA terminals as before (n=8; Supplementary Fig. 1
), but infused glutamate antagonists or saline via the fiberoptic guide cannula, before testing on the EPM and OFT. Confirming a local synaptic mechanism rather than control of fibers of passage, intra-CeA glutamate receptor antagonism abolished light-induced reductions in anxiety as measured by open-arm time (F1,35
=8.61; p=0.008) and probability of open-arm entry on the EPM (F1,35
=5.92, p=0.02), and center time during the OFT (F1,77
=13.99; p=0.0006; ). Importantly, drug treatment did not impair locomotor activity (Supplementary Fig. 9
), and in acute slices time-locked light-evoked excitatory responses were abolished upon bath application of 2,3-dihydroxy-6-nitro-7-sulfamoyl-benzo[f] quinoxaline-2,3-dione (NBQX) and (2R)-amino-5-phosphonopentanoate (AP5) (Supplementary Fig. 10
). These data demonstrate that the light-induced anxiolytic effects were caused by the activation of BLA-CeA synapses, and not attributable to BLA projections to distal targets passing through the CeA.
Finally, to test whether basal anxiety-reducing processes could be blocked by selectively inhibiting the BLA-CeA pathway, we bilaterally transduced either eNpHR3.0 -which hyperpolarizes neuronal membranes upon illumination with amber light15
- or eYFP alone, under the CaMKIIα promoter in the BLA, and implanted bilateral bevelled guide cannulae to allow selective illumination of BLA terminals in the CeA (Supplementary Fig. 11
). eNpHR3.0 expression was restricted to glutamatergic neurons in the BLA, and the eNpHR3.0:BLA-CeA group showed elevated levels of c-fos expression, relative to the eYFP:BLA-CeA and eNpHR3.0:BLA(somata) groups, in the CeM (p<0.05; Supplementary Fig. 12
), consistent with the hypothesis that inhibition of BLA-CeL synapses suppresses feed-forward inhibition from CeL neurons to CeM neurons, thereby increasing CeM excitability and the downstream processes leading to increased anxiety phenotypes. Selective illumination of eNpHR3.0-expressing axon terminals reduced the probability of both spontaneously occurring (frequency: F1,8
=32.99, p=0.00024; amplitude: F1,8
=21.96, p=0.001; Supplementary Fig. 13
) and electrically-evoked (F1,10
=10.79, p=0.006; ) vesicle release, without preventing spiking at the soma (Supplementary Fig. 14
). BLA somata inhibition did not induce an anxiogenic response, perhaps owing to the simultaneous decrease in direct BLA-CeM excitatory input. We also found that the eNpHR3.0:BLA-CeA group showed reduced open-arm time and probability of open-arm entry on EPM (F1,40
=21.08, p<0.00001; and F1,40
=19.93, p<0.00001, respectively; ) and center-time in the OFT (F1,100
=18.919, p<0.00001; ) during photostimulation when compared to the eYFP and BLA(somata) groups, without altering locomotor activity (Supplementary Fig. 15
). These data demonstrate that preferential inhibition of BLA-CeL synapses acutely increases anxiety-like behaviors.
Selective inhibition of BLA terminals in the CeA induces an acute and reversible increase in anxiety
Here, we have identified the BLA-CeL pathway as a neural substrate for real-time bidirectional modulation of the unconditioned expression of anxiety. The observation that selective illumination of specific BLA terminals produces distinct, and even opposite, behavioral responses from illumination of all glutamatergic BLA somata nonspecifically, points to the essential value of optogenetic control in causally dissecting intact neural circuitry, and suggests that multiple subpopulations or projections of BLA neurons can act in opposition (e.g. direct excitation of CeM along with feed-forward inhibition of CeM). Neural circuitry arranged in this way provides many opportunities for modulation of expression of anxiety phenotypes; for example, this microcircuit is well-positioned to be influenced by top-down cortical control from regions important for processing fear and anxiety, including the prelimbic, infralimbic and insular cortices that provide robust innervation to the BLA and CeL.
These data are consistent with reports implicating CeA involvement in anxiety9,11,12
, but it is important to note that our findings do not exclude downstream or parallel circuits including the BNST28
, the insular and prefrontal cortices29
, and the septal-hippocampal circuit30
; for example, stress induces CeL release of corticotropin releasing hormone (CRH) in the BNST28
. In the course of providing insight into native anxiogenic and anxiolytic processes, these findings demonstrate that anxiety is continuously regulated by balanced antagonistic pathways within the amygdala, and illustrate the importance of resolving specific projections in the study of neural circuit function relevant to psychiatric disease.