Abstract
Dopaminergic signaling plays an important role in associative learning including fear and extinction learning. Dopaminergic midbrain neurons encode prediction error-like signals when threats differ from expectations. Within the amygdala, GABAergic intercalated cell (ITC) clusters receive the densest dopaminergic projections, but their physiological consequences are incompletely understood. ITCs are important for fear extinction, a function thought to be supported by activation of ventromedial cluster ITCs that inhibit central amygdala fear output. In mice, we reveal two distinct mechanisms how mesencephalic dopaminergic afferents control ITCs. Firstly, they co-release GABA to mediate rapid, direct inhibition. Secondly, dopamine suppresses inhibitory interactions between distinct ITC clusters via presynaptic D1-receptors. Early extinction training augments both, GABA co-release onto dorso-medial ITCs and dopamine-mediated suppression of dorso- to ventromedial inhibition between ITC clusters. These findings provide novel insights into dopaminergic mechanisms shaping the activity balance between distinct ITC clusters that could support their opposing roles in fear behavior.
Introduction
Mesencephalic dopaminergic neurons in the ventral tegmental area (VTA) and substantia nigra pars compacta (SNC) constitute a neuromodulatory system that has been linked to error prediction and salience coding (Bromberg-Martin et al., 2010; Horvitz, 2000; Schultz et al., 1997). Prediction error signals drive the formation and the updating of stimulus-outcome associations by detection of mismatches between actual and expected experiences (Pearce and Hall, 1980; Rescorla and Wagner, 1972), which also operates when conditioned stimulus (CS)-unconditioned stimulus (US) and CS-no US associations need to be formed during fear conditioning and extinction learning, respectively (Bouton, 2004; McNally et al., 2011; Salinas-Hernandez et al., 2018). More specifically, omission of an expected aversive stimulus during the early sessions of extinction training activates a subset of VTA dopaminergic neurons (Salinas-Hernandez et al., 2018). Consequently, optogenetic inhibition of the VTA dopaminergic neurons during extinction training impairs, and their excitation enhances the extinction learning (Luo et al., 2018; Salinas-Hernandez et al., 2018).
The amygdala is a key structure that plays a critical role in mediating fear and anxiety related behaviors, and is a primary site for acquisition and storage of fear memory (Davis, 2000; Duvarci and Pare, 2014; Luo et al., 2018; Salinas-Hernandez et al., 2018). Dopamine (DA) is released in the amygdala during affective states such as stress or fear (Inglis and Moghaddam, 1999; Yokoyama et al., 2005) and direct pharmacological intervention on dopaminergic receptors in the central amygdala (CeA) or basal amygdala (BA) affects the acquisition and storage of fear memories (Guarraci et al., 1999; Lee et al., 2017).
Interestingly, selective optogenetic stimulation of midbrain dopaminergic axons has been shown to release not only the neuromodulator DA, but also fast-acting classical neurotransmitters such as GABA and/or glutamate in the dorsal striatum and nucleus accumbens (Granger et al., 2017; Tritsch et al., 2016). In the amygdala, glutamate was shown to be co-released with DA in CeA (Groessl et al., 2018; Mingote et al., 2015) and in BA, though preferentially onto fast-spiking interneurons (Lutas et al., 2019), a co-release of GABA has not been reported so far.
Several lines of evidence suggest that within the amygdala, the intercalated cells (ITCs) could be a key target for dopaminergic regulation. ITCs constitute a specialized network of GABAergic cells, and in mice are organized in several clusters around the basolateral complex of the amygdala (BLA) (Busti et al., 2011; Geracitano et al., 2007; Marowsky et al., 2005). Dopaminergic axons densely innervate the somata and dendrites of ITCs (Asan, 1997; Fuxe et al., 2003), however, the functional consequence of this innervation has been largely unexplored. Furthermore, ITCs show a high-expression level of pre- and post-synaptic DA-receptors (Fuxe et al., 2003; Pinto and Sesack, 2008; Wei et al., 2018). DA application in brain slices directly hyperpolarizes neurons in the medial and lateral ITC clusters via D1 receptors (DRD1) and thereby suppresses their output to CeA and BLA, respectively (Gregoriou et al., 2019; Manko et al., 2011; Marowsky et al., 2005).
In that respect, ITC clusters are ideally positioned to integrate dopaminergic signals with the sensory information that is either conveyed directly (Asede et al., 2015; Barsy et al., 2020; Strobel et al., 2015) or that has been preprocessed in the BLA (Herry and Johansen, 2014; Kwon et al., 2015; Pare et al., 2004). Indeed, ITCs and their plasticity have been shown to play a significant role in extinction (Amano et al., 2010; Likhtik et al., 2008) and more recently also in fear learning and memory (Asede et al., 2015; Busti et al., 2011; Huang et al., 2014; Kwon et al., 2015). The classical view on ITC function within amygdala circuits posits an inhibitory action onto neighboring CeA and BLA nuclei to gate information flow (Ehrlich et al., 2009; Marowsky et al., 2005; Morozov et al., 2011; Pare et al., 2004). However, there is emerging evidence pointing toward a more complex picture that involves temporal separation in the activity of the individual clusters, i.e. dorsomedial (dm)- and ventromedial (vm)-ITC clusters, in fear recall and extinction (Busti et al., 2011). Considering the connections between the different clusters (Asede et al., 2015; Busti et al., 2011; Duvarci and Pare, 2014; Royer et al., 2000), it is plausible that ITCs’ inhibitory actions not only on CeA and BLA, but also onto other ITC clusters could be modulated by DA.
Thus, while mounting evidence supports that midbrain dopaminergic inputs play important roles in controlling amygdala function in fear learning as well as extinction (Abraham et al., 2014; Lee et al., 2017), the cellular impact onto specific amygdala microcircuits is incompletely understood (Grace et al., 2007; Lee et al., 2017). To address this knowledge gap, we used specific optogenetic stimulation of dopaminergic axons from VTA/SNC to explore the functional impact onto ITCs. We observed co-release of GABA from dopaminergic axons onto ITCs mediating fast inhibition. Phasic stimulation of dopaminergic fibers suppressed spontaneous inhibitory inputs onto dm-ITCs. Consequently, when interrogating transmission between ITCs clusters, we observed a presynaptic depression mediated by DRD1. Extinction learning promoted the co-release of GABA in dm-ITCs and the DA-induced suppression of inhibition between dm- and vm-ITC clusters. Our results demonstrate a dual mechanism of action of dopaminergic inputs with a cooperative fast ionotropic and a slow metabotropic component, that jointly regulates the inhibitory network formed between ITCs. We suggest that this can tip the activity balance between individual clusters to support fear suppression.
Results
Dopaminergic fibers from VTA/SNC innervate ITC clusters in amygdala
We first examined the projections of midbrain dopaminergic inputs to amygdala ITC clusters and surrounding amygdala regions. To this end, we targeted dopaminergic cells selectively using a DAT-IRES-Cre mouse line and injected a Cre-dependent recombinant adenoassociated virus (AAV) encoding ChR2-YFP into the VTA/SNC (Figure 1–figure supplement 1A). Our injections resulted in expression of ChR2-YFP mostly in VTA and to a lesser extent in SNC neurons, the vast majority of which were immunopositive for tyrosine hydroxylase (TH) (Figure 1–figure supplement 1B-C). We next examined YFP+ fibers in the amygdala (Figure 1–figure supplement 2A). YFP+ fibers targeting FoxP2-positive neurons in the dm- and vm-ITC clusters were also immunoreactive for TH, indicating that they are dopaminergic in nature (Figure 1A-B). In line with previous reports (Lutas et al., 2019; Mingote et al., 2015), we also observed YFP+ fibers colocalized with TH-labeled axons in the CeA and more sparsely in the BA (Figure1–figure supplement 2B-C).
We next asked if YFP+ midbrain afferents made synaptic contacts with neurons in the dm- and vm-ITC clusters. To this aim, we filled ITCs with biocytin, which allowed us to visualize their somato-dendritic domain. We then labeled putative presynaptic active zones by immunostaining for Bassoon (Liu et al., 2018). Colocalization of Bassoon and ChR2-YFP was observed in close proximity to dendrites and somata of ITCs in dm- and vm-clusters (Figure 1C-D), suggesting that DA midbrain inputs make functional synaptic contacts.
Midbrain dopaminergic neurons from VTA and SNC co-release GABA and/or glutamate in the dorsal striatum and nucleus accumbens (Granger et al., 2017; Mingote et al., 2015; Stuber et al., 2010; Tecuapetla et al., 2010; Tritsch et al., 2012; Tritsch et al., 2016). In the amygdala, functional co-release of glutamate from dopaminergic VTA afferents has been demonstrated in CeA (Mingote et al., 2015) and to a lesser extent in the BA (Lutas et al., 2019). To adress whether this may also be the case for ITCs, we first checked for the presence of vesicular GABA or glutamate transporters in ChR2-YFP+ axons. We indeed observed immunoreactivity for the vesicular glutamate transporters vGluT1/2 and vesicular GABA transporter vGAT in a subset of ChR2-YFP+ presynaptic boutons within FoxP2-positive dm- and vm-ITC clusters (Figure 1E-H), corroborating the hypothesis of GABA and glutamate co-release from midbrain dopaminergic fibers. Because of incomplete penetration of the vGluT1/2 and vGAT antibodies into slices, a quantitative analysis of the colocalization with ChR2-YFP+ terminals was not carried out.
Dopaminergic fibers co-release mainly GABA in the medial ITC clusters
In light of our anatomical observations of a possible co-release of GABA or glutamate, we explored whether optical stimulation of ChR2-YFP+ fibers from confirmed dopaminergic midbrain injections (Figure2–figure supplement 1A-C) evoked fast postsynaptic currents (PSCs) in dm- and vm-ITCs. Indeed, we could detect PSCs in approximately 50-60% of the neurons recorded in whole cell patch clamp mode in ITC clusters as well as in the CeA (Figure 2A-B, D, G), but not in principal neurons of the BA (10 neurons recorded from 5 animals showing co-release in either CeA or ITCs, Figure 2–figure supplement 2D).
To classify fast PSCs according to neurotransmitter type, we first determined their reversal potential (Figure 2E and H). In a large fraction of neurons within dm- and vm-ITC clusters, we found PSCs that reversed close to the calculated reversal potential for chloride (−47.3 mV), suggesting that they were GABAergic (Figure 2C, E and H).
Another fraction of neurons displayed PSCs with more depolarized reversal potentials (>-40 mV and <-15mV), suggesting that they were mixed PSCs, and finally a small fraction of dm-ITCs displayed PSCs that reversed close to 0mV, suggesting that they were glutamatergic (Figure 2C, E and H). To ascertain our classification, we identified the neurotransmitter receptors involved using pharmacological blockers of GABAA-receptors, as well as AMPA and NMDA-type glutamate receptors in a subset of the recorded ITC neurons (Figure 2D, G). Applying the same approach to CeA neurons, in accordance with previous reports, we detected glutamatergic PSCs (Mingote et al., 2015), but also GABAergic and mixed PSCs (Figure 2C and Figure 2–figure supplement 2A-B). Overall, the relative contribution of neurotransmitters mediating the fast PSCs significantly differed between neurons in dm-ITC or vm-ITC clusters compared with CeA neurons (Figure 2C), with the latter having a higher proportion of glutamatergic PSCs.
In line with a monosynaptic innervation, all components of the recorded inhibitory and excitatory PSCs had short latencies in both ITC clusters and CeA (Figure 2F and I, Figure 2–figure supplement 2C). In dm-ITCs, we also directly confirmed that PSCs were monosynaptic by demonstrating that they were abolished by the sodium channel blocker tetrodotoxin (TTX) and recovered in the presence of TTX and the potassium channel blocker 4-aminopyridine (4-AP) (Figure 2–figure supplement 3A-B) (Petreanu et al., 2009). Taken together, our results strongly suggest that projections originating from VTA and SNC dopaminergic neurons mainly co-release GABA in medial ITC clusters to induce GABAA-receptor mediated fast inhibitory PSCs (IPSCs).
Phasic stimulation of dopaminergic fibers alters sIPSC amplitude in dm-ITCs
Phasic activity of midbrain dopaminergic neurons has been shown to release dopamine in dorsal striatum and nucleus accumbens both ex vivo and in vivo (Schultz et al., 1997; Stuber et al., 2010). If this is also the case for ITCs, we should be able to detect an electrophysiological signature or a modulatory effect of dopaminergic inputs onto ITCs. We addressed this by stimulating dopaminergic midbrain afferents with 30 Hz light pulse trains to simulate phasic activity, while comparing the frequency and amplitude of spontaneous synaptic inputs (sPSCs) onto dm-ITCs before and after stimulation (Figure 3A). While we did not detect changes in the frequency for either sIPSCs or sEPSCs (Figure 3A-B), we observed a small but significant reduction in the amplitude of sIPSCs, but not sEPSCs (Figure 3C-D). Conversely, tonic 1 Hz stimulation had no effect on frequency or amplitude of sIPSCs (Figure 3–figure supplement 1A-B). Stimulation of dopaminergic midbrain inputs at 30 Hz also decreased the amplitude of sIPSC in vm-ITCs (Figure3–figure supplement 2A-B), corroborating our findings from dm-ITCs. Although we cannot completely rule out contributions from co-released transmitters, our results strongly suggest that phasic release of DA by midbrain neurons has a neuromodulatory effect on ITC inhibitory inputs.
DA modulates inhibitory synaptic transmission between ITC clusters
Following our observation that phasic stimulation of midbrain dopaminergic afferents affects sIPSCs, we wanted to more directly dissect the effect of DA on defined inhibitory inputs. Local inhibitory interactions have been shown between ITCs within one cluster, but also between medially located ITC clusters (Busti et al., 2011; Geracitano et al., 2007; Royer et al., 2000). To examine the effect of DA on between-cluster interactions, we used FoxP2-Cre mice to express AAV-ChR2-YFP specifically and selectively in one of the ITC clusters. In subsequent slice recordings, we stimulated axons and recorded in the innervated target cluster (Figure 4A, D, G). For the dm-ITCs, we hypothesized that the lateral ITC (l-ITC) cluster provides one of the inputs. Indeed, optical stimulation of ChR2-YFP+ fibers arising from l-ITCs reliably evoked PSCs in dm-ITCs (Figure 4B), that reversed around the equilibrium potential for chloride (Erev=−44.77 ± 1.77, n=6), and were largely abolished by the GABAA-channel blocker picrotoxin (PTX, % remaining current at 0 mV= 4.70 ± 1.09%, n=3). This suggests that l-ITCs provide fast GABAergic inputs to the dm-ITC cluster. Next, we assessed the effect of DA on inhibitory transmission in l-ITC→dm-ITC (Figure 4D) and dm-ITC→vm-ITC (Figure 4G) pathways. To select for effects of dopaminergic modulation of IPSCs without interference by GABA release from dopaminergic fibers, we opted to bath apply DA for 5 minutes. To gain insight into a possible presynaptic action, we used a paired pulse stimulation protocol (100 ms interstimulus interval). DA application significantly and reversibly suppressed IPSC amplitude to approximately 50% of baseline in l-ITC→dm-ITC and dm-ITC→vm-ITC pathways (Figure 4E and H), that was accompanied by a significant increase in the paired pulse ratio (PPR) (Figure 4F and I). Optically evoked IPSCs in the l-ITC→dm-ITC and dm-ITC→vm-ITC pathways were both of short latency, in keeping with monosynaptic connectivity (Figure 4C). Taken together, our data are consistent with a presynaptic site of action and indicate that DA dampens inhibitory interactions between l-ITC→dm-ITC and dm-ITC→vm-ITC clusters by decreasing GABA release. This is further supported by the fact that the decrease in IPSC amplitude was significantly correlated with the associated increase in PPR (Figure 5G).
Presynaptic DRD1 activation mimics the effect of DA in suppressing inhibitory interactions between ITC clusters
Dopamine receptors DRD1 and DRD2 have been localized in the amygdala, with ITC clusters being particularly enriched in DRD1 (Fuxe et al., 2003; Pinto and Sesack, 2008). To identify which receptor is involved in the DA-induced suppression of inhibitory synaptic transmission between ITC clusters, we used DH and quinpirole as selective agonists for DRD1 and DRD2, respectively.
Activation of DRD1 significantly suppressed IPSCs in the l-ITC→dm-ITC pathway, whereas DRD2 activation had no effect (Figure 5A-B). Furthermore, activation of DRD1 significantly increased the PPR in the l-ITC→dm-ITC pathway, whereas DRD2 activation did not (Figure 5A and C). The DH-induced IPSC suppression and PPR increase was comparable to the effect of DA application (Figure 5B-C). Likewise, DRD1 activation also suppressed IPSCs similar to DA (to about 50%) in the dm-ITC→vm-ITC pathway, whereas DRD2 activation had only a small, yet significant effect on amplitude, that was different from that of DA or DRD1 activation (Figure 5D-E). Activation of DRD1 also significantly increased the PPR in the dm-ITC→vm-ITC pathway, but DRD2 activation did not (Figure 5E-F). Moreover, we again observed a significant correlation between the decrease in IPSC amplitude and the increase in PPR upon application of the DRD1 agonist (Figure 5H), supporting a presynaptic site of action.
Another reported mechanism by which DA tunes ITC output in younger animals is a direct DRD1-mediated hyperpolarization (Manko et al., 2011; Marowsky et al., 2005). Therefore, we also tested for a direct postsynaptic action of DA by recording dm-ITC neurons in current-clamp mode from young and adult GAD67-GFP mice without stimulating any input (Figure 5–figure supplement 1A). We observed a significant DA-induced hyperpolarization in about 50% of recorded dm-ITC cells from both age groups (Figure 5–figure supplement 1B-D). Thus, irrespective of age, only a fraction of ITC neurons exhibits a small, direct postsynaptic response to application of DA. In conclusion, our data suggests that DA modulates ITC cluster interactions chiefly by presynaptic DRD1 receptors.
Early extinction training alters midbrain input properties and modulation of ITCs
Recent data show that activity in midbrain dopaminergic neurons correlates strongly with early extinction learning, suggesting that in the amygdala, DA provides a prediction error-like neuronal signal that is necessary to initiate fear extinction (Luo et al., 2018; Salinas-Hernandez et al., 2018). Thus, we tested if early extinction training alters co-release from midbrain dopaminergic inputs to ITC neurons and the dopaminergic modulation of ITC cluster interaction. To this end, we subjected one group of mice to fear conditioning on day 1 and a day later to fear retrieval and early extinction training (E-Ext) by presenting 16 CSs in the absence of the reinforcing US, while the control group was only exposed to the same number of CSs (CS-only, Figure 6A). In both DAT-IRES-Cre and FoxP2-Cre mice transduced with ChR2-YFP in midbrain and ITC neurons, respectively, this short protocol did not reduce fear expression significantly (Figure 6–figure supplement 1A-B and figure supplement 3A-B). Importantly, these mice can extinguish when subjected to a 2-day extinction training protocol (Figure 6–figure supplement 2). As our focus was on early extinction, we opted to obtained acute brain slices for recordings from ChR2-YFP transduced mice only after early extinction training (Figure 6A).
Early extinction training significantly decreased the reversal potential of fast PSCs evoked by stimulation of dopaminergic inputs from confirmed midbrain injection sites (Figure 6–figure supplement 1C) onto dm-ITCs compared to animals in the CS-only group (Figure 6B-C). When categorizing PSCs according to their reversal potential, as described for naïve animals (c.f. Figure 2C-F), we found that the pattern of responses shifted to more GABAergic PSCs with early extinction training (Figure 6D). This suggests that midbrain inputs are more likely to co-release GABA, that can contribute to inhibit dm-ITCs. Secondly, we found that early extinction training altered the efficacy of dopaminergic modulation of ITC cluster interaction in the dm-ITC→vm-ITC pathway (Figure 6E-G). The suppression of IPSC amplitude by DA was significantly larger in neurons recorded from animals in the E-Ext vs. CS-only group, with a concomitant significantly larger increase in the PPR of IPSCs (Figure 6F-G). This suggests that early extinction training decreases inhibition by enhancing a dopaminergic presynaptic mechanism, which can contribute to the disinhibition of vm-ITCs. In summary, our data indicate that changes in dopaminergic input action during early extinction learning can switch the activity balance between ITC clusters by enhancing inhibition of dm-ITCs and disinhibition of vm-ITCs.
Discussion
Here, we investigated the functional impact of midbrain dopaminergic inputs and dopaminergic modulation onto amygdala ITCs. Our key findings are that ITCs are controlled by two distinct mechanisms including fast inhibition resulting from a prominent co-release of GABA, and a slower DRD1-mediated presynaptic suppression of inhibitory interactions between distinct ITC clusters (Figure 7A). Upon early extinction learning, fast inhibition onto dm-ITCs is increased, and DA more potently suppresses dm cluster-mediated inhibition of vm-ITCs. This may support a shift in the activity balance between these two distinct ITC clusters by inhibiting dm-ITCs and disinhibiting vm-ITCs to enable fear suppression during extinction learning (Figure 7B). Although it is well established that amygdala ITC clusters receive dense TH+ projections (Asan, 1997; Fuxe et al., 2003), here, we show that these are part of the mesolimbic pathway from VTA/SNC, also providing afferents to amygdalostriatal transition zone (Astria), BA, and to a lesser extent to CeA. Our results are well in line with a recent report that 6-hydroxy-dopamine injections in BLA resulted in the loss of TH+ neurons in VTA/SNC that also denervated ITCs and ventral Astria, but not CeA (Ferrazzo et al., 2019). The CeA was shown to receive largely distinct dopaminergic inputs arising from dorsotegmental nuclei (Groessl et al., 2018). Consistent with previous ultrastructural investigations demonstrating that TH+ axons contact ITC somata, dendrites, and spines (Asan, 1997; Pinto and Sesack, 2008), we detected putative presynaptic terminals originating from VTA/SNC on the soma and along ITC dendrites suggesting functional connectivity.
While co-release of glutamate and GABA from VTA/SNC dopaminergic neurons is well established in nucleus accumbens and dorsal striatum (Granger et al., 2017), glutamate co-release has only recently been demonstrated in CeA (Mingote et al., 2015). Here, we also find GABA co-release in CeA, although to a much lower extent than glutamate co-release. We did not observe glutamate co-release in the BLA, which is also in line with a previous study (Mingote et al., 2015), but see also (Lutas et al., 2019).
Our study is the first to directly examine dopaminergic afferents onto amygdala ITCs via optogenetic stimulation. Combined qualitative anatomical and quantitative electrophysiological evidence strongly supports the notion that glutamate and GABA can be released from dopaminergic afferents onto ITCs, with a major contribution of GABA. GABA co-release is a prominent feature of midbrain dopaminergic neurons targeting dorsal striatum and nucleus accumbens, where it can rapidly inhibit medium spiny neurons of the direct and indirect pathways (Tritsch et al., 2012). The presence of vGluT1/2 and vGAT in dopaminergic fibers innervating medial ITC clusters is in accordance with canonical release mechanisms for glutamate and GABA, respectively. In dorsal striatum however, GABA is transported into vesicles via the vesicular monoamine transporter VMAT2 suggesting a spatio-temporal synchronization of GABA and DA release from the same vesicles (Tritsch et al., 2012). Which of the two mechanisms is predominant in ITC afferents remains to be investigated. Furthermore, in dopaminergic striatal afferents, GABA is either synthesized in a non-canonical aldehyde-dehydrogenase1a1-dependent pathway, and/or relies on GABA uptake from the extracellular milieu via GABA transporters (Kim et al., 2015; Tritsch et al., 2014). Although we did not determine which mechanism provides GABA for inhibiting ITCs, regulation of GABA uptake could endow midbrain afferents with the flexibly to rapidly alter GABAergic co-transmission in a target-specific manner. It may, therefore, be a good candidate mechanism mediating changes in GABA co-release during early extinction learning.
ITCs receive a number of well characterized excitatory inputs from BLA, thalamus, prefrontal and sensory cortex that drive their activity (Amir et al., 2011; Asede et al., 2015; Cho et al., 2013; Pare et al., 2003; Strobel et al., 2015). Apart from local inhibitory interactions between ITCs, that were proposed to stabilize overall spike output (Geracitano et al., 2007; Manko et al., 2011; Royer et al., 2000), it remained unclear which extrinsic inputs inhibit ITCs. Co-release of GABA from dopaminergic midbrain could provide a rapid and temporally precise signal to inhibit ITC spike activity in response to salient stimuli and/or during learning. Furthermore, spatially targeted inhibition onto ITC dendrites or spines may shunt glutamatergic inputs to enable local control of synaptic plasticity (Tritsch et al., 2016).
Phasic activity of midbrain dopaminergic neurons may signal unexpected rewards or aversive events such as noxious stimuli and these activity patterns shape DA release in target structures (Brischoux et al., 2009; Bromberg-Martin et al., 2010). Phasic stimulation (> 20 Hz) of VTA neurons increases DA release in forebrain structures, including amygdala, and affects behavior (Holloway et al., 2018; Tsai et al., 2009). In our hands, phasic but not tonic stimulation of midbrain dopaminergic inputs to dm-ITCs depressed sIPSC, suggesting that released DA modulates inhibitory transmission. Our results are in agreement with a recent report showing that DA depressed IPSC amplitude in vm-ITCs (Gregoriou et al., 2019). Despite earlier studies reporting that DA directly hyperpolarizes medial and lateral ITCs via postsynaptic DRD1 (Manko et al., 2011; Marowsky et al., 2005), Kwon and colleagues did not replicate these findings (Kwon et al., 2015). Our data may resolve this discrepancy by demonstrating that only a fraction of dm-ITCs is directly hyperpolarized by DA, and indicate the presence of several distinct dopaminergic mechanisms gating ITC activity.
While ITCs form local inhibitory networks within, but also between medially located neighboring clusters (Amir et al., 2011; Busti et al., 2011; Geracitano et al., 2007; Manko et al., 2011; Royer et al., 2000), our novel finding that l-ITCs provide input to dm-ITCs suggests further complexity of inhibitory and disinhibitory interactions between ITC clusters that can provide a substrate for integration of information from distinct afferents (Asede et al., 2015; Morozov et al., 2011; Strobel et al., 2015). DA also modulated the interactions between ITC clusters. Its disinhibitory effect via presynaptic DRD1 in the target cluster resembles the situation in the nucleus accumbens, where presynaptic DRD1 attenuates IPSCs and consequently, lateral inhibition between medium spiny neurons, a mechanism proposed to diminish competitive interactions between single projection neurons or ensembles (Nicola and Malenka, 1997; Pennartz et al., 1992; Taverna et al., 2005). In analogy, the same mechanism may enable selection of specific ITC clusters or ensembles during distinct behavioral states. Our data extend current views on dopaminergic modulation of the ITC network with DRD1 playing a major role both pre- and postsynaptically. While direct inhibition of ITCs decreases overall inhibition onto diverse downstream regions, target-specific presynaptic disinhibition would promote inhibition from distinct ITC clusters onto defined downstream regions. These mechanisms may help to sharpen competitive interactions between clusters and to select a defined ITC network output. In the context of behavior, in vivo pharmacological studies in amygdala pointed to a role of DRD1 in the acquisition of cued fear memory (Guarraci et al., 1999; Lamont and Kokkinidis, 1998; Nader and LeDoux, 1999) as well acquisition of extinction (Hikind and Maroun, 2008). Cellular effects of DRD1 include altered excitability of BLA projection- and local interneurons, as well as DRD1-dependent LTP in the BLA-CeA pathway that have been discussed in the context of fear acquisition (Groessl et al., 2018; Lee et al., 2017). However, DRD1-dependent ITC network modulation could provide the required flexibility to support behavioral transitions during acquisition of extinction memory. Circuit and immediate early gene mapping studies converged on a model where dm-ITC become engaged and undergo plasticity during fear learning keeping vm-ITC activity in check. During extinction learning, this balance tips towards recruitment of vm-ITCs, which are required for extinction retrieval (Busti et al., 2011; Duvarci and Pare, 2014; Likhtik et al., 2008). Our data suggest that dopaminergic afferent control over ITCs could play a critical role in these transitions. DA neurons in the VTA are activated by unexpected omission of the US implying that they provide a prediction-like error signal to downstream targets including ITCs (Luo et al., 2018; Salinas-Hernandez et al., 2018). Consistent with the above model, enhanced GABA co-release onto dm-ITCs during early extinction could rapidly decrease their activity, and together with enhanced DRD1-mediated disinhibition of vm-ITCs allow for activation of the latter. Downstream, dm- and vm-ITCs most likely exert behavioral output by differentially inhibiting specific fear- and extinction related subpopulations of BLA and CeL or centro-medial (CeM) neurons (Figure 7B) (Asede et al., 2015; Duvarci and Pare, 2014; Gregoriou et al., 2019). Concurring or subsequent synaptic plasticity at BLA and other afferents onto dm- and vm-ITCs could help maintain this activation pattern during extinction retrieval (Amano et al., 2010; Asede et al., 2015; Huang et al., 2014; Kwon et al., 2015).
While our study delineates several novel cellular mechanisms how mesolimbic dopaminergic afferents control amygdala ITC networks, the observed ex vivo alterations are still correlative. Further studies are warranted to understand how these cellular mechanisms impact fear and extinction learning and memory. Beyond fear memories, DA is critical for processes related to incentive salience, motivation, and cue-reward learning (Abraham et al., 2014; Bromberg-Martin et al., 2010). For example, VTA dopaminergic inputs to BLA are activated by motivationally salient appetitive and aversive outcomes and acquire responses to predictive cues during learning (Lutas et al., 2019). DA is also released in the amygdala during affective states such as stress (Belujon and Grace, 2015; Yokoyama et al., 2005). Therefore, it is intriguing to speculate that dopaminergic modulation of distinct ITCs clusters may also play a role in other processes such as reward learning and addiction, as well as in the control of mood-related behavior such as anxiety- and depression-like behavior (Ferrazzo et al., 2019; Kuerbitz et al., 2018).
Materials and Methods
Animals
We used 6 to 14 week-old adult male DAT-IRES-Cre (JAX#006660, Jackson Laboratories, Bar Harbor, Maine, USA) (Backman et al., 2006) or FoxP2-Cre (JAX#030541, Jackson Laboratories) (Rousso et al., 2016) transgenic mice for Cre-dependent expression of viral vectors, and 20-28 day-old (young) and 8-10 week-old adult male GAD67–GFP (Tamamaki et al., 2003) mice for slice experiments without optical stimulation. All lines were heterozygous and backcrossed to C57BL6/J. Mice were kept in a 12-hour light/dark (6 am to 6 pm) cycle with access to food and water ad libitum. All behavioral experiments were conducted during the light cycle (between 7 and 10 am). All animal procedures were performed in accordance with institutional guidelines and with current European Union guidelines, and were approved by the local government authorities for Animal Care and Use (Regierungspraesidium Tuebingen, State of Baden-Wuerttemberg, Germany).
Surgical procedures
Stereotaxic injection of AAV-EF1a-DIO-HChR2(H134R)-eYFP (serotype 2/1 or 2/9, U. Penn Vector Core, Philadelphia, PA, USA or Addgene, Watertown, MA, USA) was performed in 6-8 weeks old DAT-IRES-Cre or FoxP2-Cre transgenic mice. Mice were anesthetized with isoflurane in oxygen-enriched air (Oxymat 3, Weinmann Medical Technologies, Hamburg, Germany) and the head was fixed in a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA or Stoelting, Wood Dale, IL, USA). Eyes were protected with ointment and the body temperature of the animal was maintained using a feedback-controlled heating pad with a rectal sensor (FHC, Bowdoin, ME, USA).
Lidocaine was used as a local anesthetic. An incision was made on the skin, the skull was exposed and coordinates of bregma were identified. The skull was drilled with a microdrill (Kopf Instruments, Tujunga, CA, USA) at the desired coordinates in reference to bregma. Borosilicate capillaries for VTA/SNC (1B150F-4, World Precision Instruments, Friedberg, Germany) and ITC injections (marked 1 to 5 μl, Drummond Scientific, Broomall, PA, USA) were pulled on a horizontal pipette puller (P-1000, Sutter Instruments, Novato, CA, USA) and used to pressure inject viruses at a volume of 300-500 nl for VTA/SNC and 25–50 nl for ITCs. The capillaries were slowly withdrawn, the skull disinfected and the skin sutured with silk. Postoperative pain medication included injection of meloxicam (Metacam, Boehringer Ingelheim, Ingelheim, Germany) at 5 mg/kg subcutaneously. The following coordinates were used in reference to bregma: for VTA/SNC in mm, AP: −3.00, ML: ±0.50, DV: 4.50; for dm-ITC in mm, AP: −1.40, ML: ±3.3, DV: 4.70; for l-ITC in mm, AP: −1.40, ML: ±3.45, DV: 4.70).
Slice recordings and analysis
Three (for ITCs) to six (for VTA/SNC) weeks after viral injections, mice were deeply anaesthetized with 3% isoflurane (Isofluran CP, cp-pharma, Burgdorf, Germany) in oxygen and decapitated. The brain was rapidly extracted and cooled down in ice-cold slicing artificial cerebrospinal fluid (ACSF) containing (in mM): 124 NaCl, 2.7 KCl, 26 NaH2CO3, 1.25 NaH2PO4, 10 MgSO4, 2 CaCl2, 18 D-Glucose, 4 ascorbic acid, equilibrated with carbogen (95% O2/5% CO2). Coronal brain slices (320 μm) containing the amygdala were cut in ice-cold slicing ACSF with a sapphire blade (Delaware Diamond Knives, Wilmington, DE, USA) on a vibrating microtome (Microm HM650V, Thermo Scientific, Dreieich, Germany). Slices were collected in a custom-built interface chamber with recording ACSF containing (in mM): 124 NaCl, 2.7 KCl, 26 NaH2CO3, 1.25 NaH2PO4, 1.3 MgSO4, 2 CaCl2, 18 D-Glucose, 4 ascorbic acid, equilibrated with carbogen. Slices were recovered at 37°C for 40 minutes and stored at room temperature. Whole-cell patch-clamp recordings were performed in a submersion chamber under an upright microscope (Olympus BX51WI, Olympus Germany, Hamburg, Germany), where slices were superfused with recording ACSF at 30-31°C. Recordings were performed using an Axon Instruments Multiclamp 700B amplifier and a Digidata 1440A digitizer (both Molecular Devices, San Jose, CA, USA). Glass micropipettes (6-9 MΩ resistance when filled with internal solution) were pulled from borosilicate capillaries (ID 0.86 mm, OD 1.5 mm, Science Products, Hofheim, Germany).
Recordings in voltage-clamp configuration were performed with cesium-based internal solution containing (in mM): 115 Cs-methanesulphonate, 20 CsCl, 4 Mg-ATP, 0.4 Na-GTP, 10 Na2-phosphocreatine, 10 HEPES, 0.6 EGTA (290-295 mOsm, pH 7.2-7.3). Signals were low-pass filtered at 2 kHz and digitized at 5 kHz. Recordings in current-clamp configuration were performed with K-Gluconate based internal solution containing (in mM): 130 K-Gluconate, 5 KCl, 4 Mg-ATP, 0.4 Na-GTP, 10 Na2-phosphocreatine, 10 HEPES, 0.6 EGTA (290-295 mOsm, pH 7.2-7.3). Signals were low-pass filtered at 10 kHz and digitized at 20 kHz. Changes in series resistance <30% were accepted. Optical stimulation was achieved by triggering a light-emitting diode (LED; 470 nm, KSL70, Rapp Opto-Electronics, Hamburg, Germany or CoolLED pE, CoolLED, Andover, UK) coupled to the upright microscope objective (Olympus Germany, 60x 1.0 NA). PSC were recorded from LED stimulations (pulse length 0.2-5 msec) with 0.1 Hz frequency. Spontaneous and evoked IPSCs were recorded at 0 mV holding potential. For sEPSC recordings, GABAergic activity was blocked with PTX and cells were held at −70 mV. Phasic stimulation of dopaminergic fibers consisted of 10 pulses at 30 Hz repeated 10 times with 10 second inter-sweep interval. Tonic stimulation consisted of a single sweep of 100 pulses at 1 Hz. In some experiments biocytin (3-5%) was added to the internal solution for post-hoc visualization of the recorded cells.
Electrophysiological data were analyzed with Neuromatic (www.neuromatic.thinkrandom.com) (Rothman and Silver, 2018) and/or custom written functions in Igor Pro (Wavemetrics, Portland, OR, USA). PSC amplitudes were measured as the peak value in reference to a 5 msec baseline period. For co-release data, reversal potential (Erev) was calculated by regression analysis of the peak current at different holding potentials (−70 mV, −50 mV, 0 mV, +40 mV). Cells with Erev <-40 mV were considered to have GABAergic, −40 mV< Erev <-15 mV to have mixed, and Erev >-15mV to have glutamatergic PSCs.
Chemicals were purchased from Roth, Merck or Sigma. Drugs for pharmacology were obtained from Sigma, Tocris or Biotrend. All drugs were diluted in ACSF from concentrated frozen stocks on the day of recording. Drugs were bath-applied at a rate of 2 ml/minute using a peristaltic pump (Ismatec Products, Cole-Palmer, Wertheim, Germany).
Behavioral procedures
Fear conditioning was performed in a square chamber with a grid floor that was cleaned with 70% ethanol (Context A). Five pairings of conditioned (CS) and unconditioned (US) stimuli were presented. The CS was a 30 second tone (at 7.5 kHz, 80 dB) which coincided at its offset with the US (1 second foot shock, 0.4 mA). In the CS-only group, the US was omitted. Fear extinction was performed in a round chamber with a flat floor that was cleaned with 1% acetic acid (Context B). Early extinction training started 24 hours after fear conditioning and consisted of 16 CSs. Full extinction training was performed by presenting 25 CSs 24 hours and 48 hours after conditioning. All sessions started with a two-minute baseline assessment and the subsequent CSs were presented at random intervals of 20 to 180 seconds. Freezing was detected using an infrared beam detection system (Coulbourn Instruments, Holliston, MA, USA), with the threshold for freezing set to 2 seconds of immobility and quantified offline with custom-written macros in Microsoft Excel (Asede et al., 2015). Freezing data for fear conditioning are shown as average post-shock freezing for the last two CS-US pairings. CS-induced freezing during extinction is shown as averages four CSs. Animals were randomly assigned to experimental groups. Electrophysiology experiments were performed 80-90 minutes after the behavioral procedures by an experimenter blind to the training procedure of the animal.
Staining, immunohistochemistry, and imaging procedures
Slices containing midbrain and ITC injection sites were fixed in 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS), whereas recorded slices with biocytin filled cells were fixed in 4% PFA, 15% saturated picric acid solution, and 0.05% glutaraldehyde in PBS overnight at 4°C. All slices were resectioned on a vibratome (Microm HM650V, Thermo Scientific, Dreieich, Germany) at 60 to 65 μm and processed as previously described (Asede et al., 2015) with minor modifications.
Biocytin was revealed using fluorescently-conjugated Streptavidin (1:1000, Dianova, Germany). Immunostainings were performed using standard procedures. Upon blocking with PBS complemented with 0.3% Triton and 10% serum for 90 minutes, the following primary antibodies were used: goat anti-GFP (GeneTex, 1:500), rabbit anti-GFP (Invitrogen, 1:750), rabbit anti-FoxP2 (Abcam, 1:1000), mouse anti-FoxP2 (Merck Millipore, 1:1000), rabbit anti-TH (Millipore, 1:1000), mouse anti-TH (Merck, 1:1000), mouse anti-Bassoon (Abcam, 1:500). Secondary antibodies used were: Alexa488-conjugated donkey-anti-rabbit, Alexa555-conjugated donkey-anti-rabbit and goat-anti-mouse, Alexa647-conjugated donkey-anti-rabbit, Alexa633-conjugated goat-anti-mouse and goat-anti-rabbit. All secondary antibodies were obtained from Invitrogen and used in 1:1000 dilutions.
Overview images were taken with a fluorescence microscope (Axio Imager, Carl Zeiss, Oberkochen, Germany). Detailed images of injection and projection sites were taken with a LSM710 laser-scanning microscope (Carl Zeiss, Oberkochen, Germany) equipped with either a 40x 1.3 NA or a 63x 1.4 NA objective, with the pinhole set to 1 airy unit. Most of the confocal images are presented as maximum intensity projection of z-stacks. For deconvolution, Huygens suite 19.04 (Scientific Volume Imaging, Hilversum, Netherlands) was used. Cell counts were performed from confocal z-stack images with the cell counter plugin FIJI in ImageJ (https://imagej.net/Cell_Counter). Specificity of the infections in the dopaminergic midbrain was calculated as the ratio of double labeled cells (TH+GFP+ cells) to all infected cells (GFP+ cells).
For analysis of presynaptic markers, 60 μm coronal sections containing the ITCs, from DAT-IRES-Cre mice injected with AAV-EF1a-DIO-HChR2(H134R)-eYFP, were immunostained for vGluT1/2 or vGAT. Sections were incubated with the following primary antibodies: polyclonal rabbit anti-GFP (Invitrogen 1: 1000), sheep anti-FoxP2 (RnDSystems, 1:200) and polyclonal guinea pig anti-vGAT (Synaptic Systems, 1:500) or a combination of guinea pig-anti-vGluT1 (Millipore, 1:3000) and vGluT2 (Chemicon International, 1:5000). Secondary antibodies used were: AlexaFluor 488 donkey anti rabbit (life technologies, 1:1000), Cy5 donkey anti goat (Jackson Immunoresearch Lab., 1:400) and DyLight 405 donkey anti guinea pig (Jackson Immunoresearch Lab., 1:500). Images of vGluT and vGAT stainings were visualized using an Airy Scan LSM980 laser scanning microscope (Carl Zeiss, Oberkochen, Germany) with a 40x 1.3 NA objective. The pinhole was set to 1 airy unit. Z-stacks were obtained and then analyzed in 3D using Imaris Software (Bitplane, Zurich, Switzerland).
Data presentation and statistics
Most data are represented as individual points, box and whisker plots (reporting median, 25, 75, 10, and 90 percentiles), or bar graphs of average ± SEM. Where applicable, electrophysiology data were normalized to the first minute of baseline recording, similarly Z-scores were calculated from the average and standard deviation of the baseline.
No methods were used to predetermine sample size. Sample sizes are similar to other studies in the field. Statistical analysis was performed using the program SPSS (IBM, Germany). A p-value <0.05 was considered significant. Paired or unpaired Students t-tests were used for two dependent or independent comparisons, respectively. Comparison of multiple groups was done using one- or two-way ANOVA followed by a Bonferroni post-hoc test if appropriate. Repeated treatments were compared using repeated measure ANOVA followed by a Bonferroni post-hoc test if appropriate. Categorical data were compared with Fisher’s exact test.
Acknowledgements
We thank Johannes Ungermann and Marlly Achury for help with histology, members of the Ehrlich and Ferraguti labs for discussion and support, and Dr. Julien Genty, Martin Zeller, and Prof. Norbert Hajos for critical reading and commenting on an earlier version of the manuscript. This work was supported by the Charitable Hertie Foundation and the German Research Foundation DFG EH 197/3-1 (to I.E.) and by the Austrian Science Fund (FWF) grant # I-2215 (to F.F).