Whole-Brain Map of Long-Range Monosynaptic Inputs to Different Cell Types in the Amygdala of the Mouse

As a hub responsible for the control of diverse behaviors and the modulation of different emotions such as fear, anxiety, and reward, the amygdala is reported to connect to a large number of brain regions [1‐7]. The amygdala consists of the basolateral amygdala (BLA) and the central nucleus of the amygdala (CeA). The BLA receives projections from the medial geniculate nucleus, auditory cortex (Au), and medial prefrontal cortex (mPFC), and projects to various nuclei such as the nucleus accumbens, ventral hippocampus, prelimbic cortex, and CeA [8‐13]. Previous studies have reported the long-range connectivity of the CeA, including inputs from the BLA, paraventricular thalamic nucleus (PVT), mPFC, bed nucleus of the stria terminalis (BST), and raphe nucleus, and output projections to the periaqueductal grey, nucleus tractus solitarius, locus coeruleus, and hypothalamus [14‐16]. Although various studies have explored the innervation of the amygdala and its outputs, research on whole-brain inputs to specific cell types in the BLA and CeA is lacking.

The BLA is composed of a majority (80%–85%) of spiny glutamatergic neurons and a minority (20%) of GABAergic neurons, including parvalbumin-expressing (PV⁺) interneurons [17‐21]. These PV⁺ neurons constitute 19%–43% of GABAergic neurons in the BLA, and > 90% of PV⁺ neurons are GABAergic [22]. The CeA is a striatum-like structure consisting primarily of GABAergic neurons, which can be specifically distinguished by their molecular markers – those expressing protein kinase C-δ (PKC-δ⁺, ~ 50% of GABAergic neurons) and those expressing somatostatin (SST⁺, ~ 50% of GABAergic neurons – and these markers are rarely co-expressed [23‐25]. For these reasons, we chose vesicular-glutamate transporter 2-positive (VGLUT2⁺) and PV⁺ neurons in the BLA and PKC-δ⁺ and SST⁺ neurons in the CeA as the main neuronal types in this study.

To provide anatomical evidence to help understand the precise connections and diverse functions of the amygdala, we developed a detailed map of the whole-brain long-range inputs to the amygdala and analyzed the distributions of inputs to the four main cell types in the amygdala.

We used a modified rabies virus retrograde tracing system to map whole-brain long-range inputs to the BLA and CeA. In total, 37 individual brain regions projected to the BLA VGLUT2⁺ glutamatergic neurons and 78 regions projected to the BLA PV⁺ GABAergic neurons. In the CeA, PKC-δ⁺ GABAergic neurons received projections from 104 regions and SST⁺ GABAergic neurons received innervation from 89 regions. These data together built a detailed map of the long-range projections from the whole brain to specific neuronal types in the amygdala.

Our trans-synaptic tracing data demonstrated that several nuclei exhibited input preferences to different cell types in the amygdala. For example, we found that the BLA VGLUT2⁺ neurons received significantly more projections from CA1 than the PV⁺ neurons. Considering the role that projections from CA1 to the BLA play in fear conditioning and the massive intrinsic connections in the amygdala, this projection preference possibly indicates that glutamatergic neurons are the main input sites for sensory information, and PV⁺ interneurons are the modulators of principal neurons in the BLA [17, 34‐36]. In addition, our data showed that the PV⁺ neurons received more projections from the Pir than the VGLUT2⁺ neurons. The Pir has a unique bidirectional connection with the BLA, which may be important in the association of different meanings with different odors [37]. Our findings suggest that PV⁺ neurons may be involved in odor information processing [36, 38]. In the CeA, we found that SST⁺ neurons received more projections from CA1 than PKC-δ⁺ neurons. Projections from CA1 to the CeA are necessary for the context-dependent retrieval of cued fear memories [35]. Considering the different roles of PKC-δ⁺ and SST⁺ neurons in conditioned fear [17, 39], these data suggest that CeA SST⁺ neurons are strongly associated with the retrieval of cued fear memories related to the hippocampus. Thus, these projection preferences indicate differences in connections and functions between different types of amygdala neurons.

The BLA is widely considered to be the sensory gateway to the amygdala [40]. As expected, our tracing results showed projections from the mPFC – consisting of the prelimbic and infralimbic cortex – to BLA PV⁺ interneurons (Table S1), in accordance with a prior study showing that opto-activation of the mPFC-to-BLA projection increases food intake behavior in mice [41]. The CeA is an important region of the amygdala, orchestrating a diverse set of behaviors, including fear, anxiety, and defensive responses [2, 42, 43]. Our rabies virus input-tracing identified several principal input regions reported in previous studies, such as the PVT [44], AI [45], BST [46], and substantia nigra pars compacta [47] (Figs 4 and 5, Table S1). Projections from the lateral parabrachial nucleus to the CeA may be essential for sodium intake [48‐50]. We observed that the lateral parabrachial neurons were labeled with RV-dsRed in the SST⁺ neuron retrograde tracing (Table S1), indicating that SST⁺ neurons may be involved in salt appetitive behavior. Our results help to understand the heterogeneity of the amygdala and provide potential directions for further studies.

With the application of the rabies virus retrograde-tracing system, we identified several novel projections to the amygdala that have not been reported previously, including projections from the SI to both CeA PKC-δ⁺ and SST⁺ neurons (Figs 4 and 5). The lateral central amygdala (CeL) PKC-δ⁺-to-SI neural circuit modulates negative reinforcement learning [51]. Our findings suggest that the SI may have bidirectional connections with the CeA GABAergic neurons, possibly extending the roles of the latter in encoding motivation, fear conditioning, and conditioned reinforcement [51]. Interestingly, our trans-synaptic tracing results demonstrated a ZI-to-CeA PKC-δ⁺ neural circuit (Table S1). Stimulation of the ZI can be used to treat fear generalization [52]. Combined with recent findings on the functions of the ZI (eating, hunting, and sleeping) [53] and the role of CeA PKC-δ⁺ neurons in conditioned fear [17, 30], these novel projections may provide another explanation for how the ZI reacts in fear modulation and extend the potential roles of the amygdala in different functions. However, further optogenetic and electrophysiological studies are required to clarify these questions.

Previous study has revealed that direct glutamatergic projections from the cingulate cortex to the BLA participate in innate fear [54]. Our study showed that the BLA VGLUT2⁺ neurons did not receive inputs from the cingulate, whereas the PV⁺ neurons did. This discrepancy could be attributed to the different subtypes studied in the BLA. We did not cover other glutamatergic neuronal subtypes, such as VGLUT1⁺ glutamatergic neurons. The PV⁺ interneurons in the BLA can be divided into four subtypes based on their firing properties, and exhibit subtype-specific heterogeneity in their patterns of local synaptic connections, as well as in CeL early-spiking and late-spiking neurons [55]. It is difficult to distinguish the heterogeneities of these PV⁺ CeL early-spiking and late-spiking neurons in long-range connections with rabies virus tracing due to a lack of specific molecular markers for these subtypes. Thus, further studies combined with optogenetics and electrophysiological analysis are required to solve this issue. Using immunohistochemical and cholera toxin B tracing methods, previous studies have shown that the central and basal amygdala are targets of dopaminergic terminals [56‐60]. We observed inputs from the mesencephalic dopaminergic system, such as from the substantia nigra and ventral tegmental area (VTA) to the BLA PV⁺ neurons (< 1%, see Table S1) and from the substantia nigra pars compacta, VTA, periaqueductal gray, and dorsal raphe to the CeA PKC-δ⁺ and SST⁺ neurons (< 1%, see Table S1), supporting the idea that these dopaminergic areas may form synapses in the amygdala. However, these sparse inputs could be the result of the limitations of the monosynaptic rabies virus-tracing system used in neuromodulatory projections such as dopaminergic transmission. It is likely that the extracellular space between the dopaminergic terminals and neurons in the amygdala does not allow the rabies virus particles to traverse effectively [61]. Further studies are required to test this possibility.

There are some limitations in this study. First, it was difficult to obtain the complete olfactory bulb when removing the brain, so we did not count input cells in the olfactory bulb. Second, the amygdala has very strong intranuclear and internuclear connectivity, such as separate projections from the BLA to CeA PKC-δ⁺ and SST⁺ neurons [17]. However, due to the leakage of rabies virus at the injection site, the intra-amygdala connections were excluded from our data, and regions around the BLA and CeA were omitted from analysis (e.g., the medial and basomedial amygdalar nuclei) [62]. Third, due to the limitation of transgenic mouse strains, we could not cover all neuronal subtypes in the amygdala, e.g., calbindin and calretinin GABAergic neurons in the BLA [22] and corticotropin-releasing factor and neurotensin-expressing neurons in the CeA [25]. As such, additional research on these important neuronal subtypes is required to address this issue.

Taking the above data into consideration, most afferent inputs to the amygdala came from the cortex, striatum, hippocampus, and thalamus, with great variation among the different cell types. Our research supports the idea of the amygdala as a node that contributes to multiple behaviors, as it received a varied and wide range of input projections from the whole brain. Furthermore, our study revealed a wide range of inputs to different molecular marker-labeled neurons in the amygdala. These findings will help to fully understand the roles of glutamatergic neurons and different GABAergic subtypes in the amygdala in different behaviors.