Abstract
The medial prefrontal cortex receives converging inputs from the mediodorsal thalamic nucleus (MD) and basolateral amygdala (BLA). Although many studies reported that the BLA also projects to MD, there is conflicting evidence regarding this projection, with some data suggesting that it originates from GABAergic or glutamatergic neurons. Therefore, the present study aimed to determine the neurotransmitter used by MD-projecting BLA cells in male and female rats. We first examined whether BLA cells retrogradely labeled by Fast Blue infusions in MD are immunopositive for multiple established markers of BLA interneurons. A minority of MD-projecting BLA cells expressed somatostatin (∼22%) or calretinin (∼11%) but not other interneuronal markers, suggesting that BLA neurons projecting to MD not only include glutamatergic cells, but also long-range GABAergic neurons. Second, we examined the responses of MD cells to optogenetic activation of BLA axons using whole-cell recordings in vitro. Consistent with our immunohistochemical findings, among responsive MD cells, light stimuli typically elicited isolated EPSPs (73%) or IPSPs (27%) as well as coincident EPSPs and IPSPs (11%). Indicating that these IPSPs were monosynaptic, light-evoked EPSPs and IPSPs had the same latency and the IPSPs persisted in the presence of ionotropic glutamate receptor antagonists. Overall, our results indicate that the BLA sends a mixed, glutamatergic-GABAergic projection to MD, which likely influences coordination of activity between BLA, MD, and medial prefrontal cortex. An important challenge for future studies will be to examine the connections formed by MD-projecting glutamatergic and GABAergic BLA cells with each other and other populations of BLA cells.
SIGNIFICANCE STATEMENT The mediodorsal thalamic nucleus (MD) and basolateral amygdala (BLA) send convergent projections to the medial prefrontal cortex. Although many studies reported that the BLA also projects to MD, there is conflicting evidence as to whether this projection is glutamatergic or GABAergic. By combining tract tracing, immunohistochemistry, optogenetics, and patch clamp recordings in vitro, we found that BLA neurons projecting to MD not only include glutamatergic cells, but also long-range GABAergic neurons. Differential recruitment of these two contingents of cells likely influences coordination of activity between the BLA, MD, and medial prefrontal cortex.
Introduction
Although the basolateral complex of the amygdala (BLA) is often likened to the cerebral cortex (Carlsen and Heimer, 1988), some of its features are inconsistent with this analogy. Reminiscent of cortex, the BLA contains a similar complement of cell types, including a majority of spiny glutamatergic projection cells and a heterogeneous group of aspiny inhibitory interneurons (McDonald, 1992; Smith and Paré, 1994; Mascagni and McDonald, 2007; Spampanato et al., 2011; Capogna, 2014; Vereczki et al., 2021). Also, as in cortex where inter-related areas tend to project to the same striatal territories (Alexander et al., 1986), the BLA innervates cortical areas that target the same regions of the striatum as it does (Groenewegen et al., 1990; McDonald, 1991).
However, the analogy breaks down when considering BLA's thalamic connections. Unlike the generally reciprocal connections between the dorsal thalamus and cortex (Jones, 2007), most dorsal thalamic nuclei projecting to the BLA receive little or no return afferents. Indeed, the BLA receives heavy inputs from parts of the ventroposterior and medial geniculate nuclei as well as multiple thalamic nuclei located near the midline, most of which do not receive significant projections from the BLA (Krettek and Price, 1977; Ottersen and Ben-Ari, 1979; LeDoux et al., 1990; Su and Bentivoglio, 1990; Turner and Herkenham, 1991; Vertes et al., 2015).
The best documented BLA projection to the thalamus ends in the medial portion of the mediodorsal nucleus (MDm), a projection that exists in rats and monkeys (Krettek and Price, 1977; Porrino et al., 1981; Aggleton and Mishkin, 1984; Russchen et al., 1987) but is absent or negligible in cats and mice (Krettek and Price, 1977; Mátyás et al., 2014). Importantly, the BLA projection to MDm originates from different neurons than those targeting the prefrontal or orbitofrontal cortex, with those innervating MDm having significantly larger somata (McDonald, 1987; Timbie and Barbas, 2015).
McDonald (1987) suggested that BLA cells projecting to MDm correspond to a specific subset of Golgi-defined class II neurons (McDonald, 1982), which generally correspond to GABAergic local-circuit cells. Consistent with this, studies using retrograde transport of tritiated aspartate found that BLA neurons innervating cortex (Carnes and Price, 1988) or striatum (Christie et al., 1987; Fuller et al., 1987) are labeled with this technique, whereas those targeting MDm are not (Ray et al., 1992).
Yet, other lines of evidence indicate that BLA projections to MDm do not originate from GABAergic neurons. First, it was reported that BLA neurons contribute large axon terminals that form asymmetric synapses in MDm, similar to those provided by subcortical inputs to sensory thalamic nuclei (Kuroda and Price, 1991; Timbie et al., 2020). Second, most BLA axons projecting to MDm are immunoreactive for vesicular glutamate transporter 2 (VGLUT2; Timbie and Barbas, 2015). However, since most MDm-projecting BLA neurons are located in its basolateral nucleus, and this nucleus is devoid of VGLUT2 expressing neurons (Hur and Zaborszky, 2005), these VGLUT2 axons might actually originate from a neighboring nucleus (Mátyás et al., 2014).
Overall, the nature (excitatory vs inhibitory) of the BLA projection to MDm remains unclear. Therefore, the present study aimed to shed light on this question using a combination of tract tracing, immunohistochemistry, optogenetics, and patch clamp recordings in rats.
Materials and Methods
We performed two types of experiments. In the first, we aimed to identify what type of BLA cells project to MDm by combining retrograde tracing with immunohistochemistry. In the second, we studied the responses of MDm neurons to the optogenetic activation of BLA axons in brain slices in vitro.
Animals
We used Long–Evans rats of both sexes for all experiments (300–350 g on delivery; Charles River Laboratories; RRID: RGD_2308852). They had continuous access to water and food and were maintained on a 12/12 h light/dark cycle (lights-off at 7 P.M.). All procedures were conducted during the light phase of the cycle, with the approval of the Institutional Animal Care and Use Committee of Rutgers University, and in compliance with the Guide for the Care and Use of Laboratory Animals (Department of Health and Human Services).
Surgical methods for tracer and virus infusions
Seven days or more after delivery, rats were deeply anesthetized with isoflurane, mounted in a stereotaxic apparatus, and administered atropine sulfate (0.05 mg/kg, i.m.) to aid breathing. Their scalp was cleaned with Betadine and alcohol. A local anesthetic (bupivacaine, 0.125% solution, s.c.) was injected around the sites to be incised. Ten to 15 min later, an incision was made on the scalp and a burr hole was drilled above the regions of interest (MDm for tracer infusions; BLA for virus infusions). Body temperature was kept at 37°C throughout the surgeries.
For the intracerebral infusions, we used glass pipettes pulled to an outer tip diameter of ∼70 µm (PE-22, Narishige Instruments). Infusions proceeded at a rate of 9.6 nL/5 s using a Nanoject II (Drummond Scientific Company). For retrograde labeling of BLA neurons projecting to MDm, we infused 150–200 nl of Fast Blue (FB; Polysciences) bilaterally into the MDm of 24 rats. For the optogenetic experiments, we infused one of the two following viruses: pAAV-hSyn-hChR2(H134R)-mCherry (Addgene #26976-AAV5, 200 nl) or pAAV-mDlx-ChR2-mCherry-Fishell-3 (Addgene #83898-AAV9, 200 nl, 1:10) in the BLA of 39 and 30 rats, respectively.
Based on the stereotaxic atlas of Paxinos and Watson (2006), we used the following stereotaxic coordinates [in mm relative to bregma for the antero-posterior (AP), medio-lateral (ML), and dorso-ventral (DV) axes]: AP −2.4, ML 0.8, DV −5.0; AP −3.48, ML 0.6, DV −5.40 for MDm and AP −3.4, ML 5.1, DV −7.8 for the BLA. At the conclusion of the surgery, the wound was stitched with self-absorbing sutures, a local antibiotic was applied to the wound (Neosporin), and an analgesic was administered (ketoprophen, 2 mg/kg, s.c., daily for 3 d).
Immunohistochemical procedures
Tissue preparation
Two weeks after the tracer infusions, rats were deeply anesthetized with isoflurane and perfused through the heart with 200 ml of 0.1 m phosphate buffer solution (PB; pH 7.4; 4°C) followed by 200 ml of a fixative containing 4% paraformadehyde in 0.1 m PB at 4°C. The brain was then extracted from the skull, postfixed in the same fixative overnight, transferred to a 30% glucose solution until it sank, at which point, it was sectioned with a freezing microtome at a thickness of 60 µm. One section out of four was set aside for counterstaining with cresyl violet. Sections were placed in PBS (0.01 m; pH 7.4) until further processing.
Selection of GABAergic markers
The BLA contains a similar variety of GABAergic neurons as found in the cerebral cortex (McDonald and Betette, 2001; McDonald and Mascagni, 2001a,b, 2002; Mascagni and McDonald, 2003, 2007; Vereczki et al., 2021). In descending order of numerical importance, the main subtypes are: cells expressing (1) parvalbumin (PV+; ∼43%), (2) somatostatin (SOM+; ∼18%), (3) cholecystokinin (CCK+; ∼15%), or (4) 5-HT-3A receptors (∼7%). These peptides often colocalize with others. For instance, there are two main subtypes of CCK+ cells: small CCK+ cells that also express vasoactive intestinal peptide (VIP+), and large multipolar ones that do not. Similarly, a proportion of SOM+ cells also express nitric oxide synthase (NOS+; Vereczki et al., 2021). However, co-localization with calcium-binding proteins is particularly prevalent. For instance, calbindin (CB; ∼50%) is expressed by ∼80% of PV+ cells, ∼90% of SOM+ cells, and a low proportion of large CCK+ cells (McDonald and Betette, 2001; McDonald and Mascagni, 2002). By contrast, calretinin (CR; ∼20%) is expressed by many VIP+ and small CCK+ neurons (McDonald and Mascagni, 2001a, 2002; Mascagni and McDonald, 2003). Since we aimed to determine whether some MDm-projecting BLA neurons are GABAergic, we examined whether BLA neurons retrogradely labeled with FB also expressed various markers of GABAergic interneurons, which together are present in the vast majority of BLA interneurons: PV, CB, SOM, CR, NOS, CCK.
Immunohistochemical protocol
Sections were incubated overnight at room temperature with one of the primary antibodies listed in Table 1. These antibodies have been thoroughly characterized in previous studies on the amygdala, cortex, or related systems (PV, Trouche et al., 2019; Katona et al., 2020; Pesarico et al., 2022; CB, Larsson, 2018; Jablonski et al., 2021; SOM, Moga et al., 1989; Dabrowska and Rainnie, 2010; Suzuki and Bekkers, 2010; CR, Zhang et al., 2017; NOS, Clancy et al., 2001; Pla et al., 2006; Carrillo et al., 2007; Day-Brown et al., 2017; CCK, Whissell et al., 2015; Oláh et al., 2020; Poleksic et al., 2021). The pattern of labeling we obtained with each antibody conformed with prior descriptions (McDonald, 1989; McDonald and Mascagni, 2001a, 2002; Mascagni and McDonald, 2003). Moreover, omitting the primary antibody from the incubations resulted in a complete lack of differentiated labeling.
Primary and secondary antibodies used in the present study
The primary antibody solutions also contained 0.3% Triton X-100, 1% normal goat serum, and 0.02% sodium azide in PBS (0.01 m; pH 7.4). Sections were incubated in the primary antibody solution overnight. Then, after rinsing three times in PBS (0.01 m; pH 7.4) for 10 min, the sections were incubated at room temperature with one of the corresponding secondary antibodies listed in Table 1 for 2 h. The secondary antibody solution also contained 0.3% Triton X-100, 1% normal goat serum, and 0.02% sodium azide in PBS (0.01 m; pH 7.4). The sections were then washed repeatedly in PBS (0.01 m; pH 7.4) and mounted onto gelatin-coated slides. Finally, Prolong Gold anti-fade reagent (Invitrogen) was applied to the sections before covering them with a glass coverslip.
Criteria used for the identification of immunopositive neurons and search for double-labeled neurons
To minimize the incidence of false positives, a neuron was only considered immunoreactive for a given interneuronal marker if its entire soma, not only its periphery, displayed a level of immunostaining distinctly above background and similar to that seen in the most intensely labeled cells in the field of view. Also, since some interneuron subtypes are not homogeneously distributed in the BLA, we only assessed whether a cell labeled with FB also expressed an interneuronal marker if there were immunopositive neurons near the cell in question.
Electrophysiological experiments
Slice preparation
In vitro optogenetic experiments were conducted six to eight weeks after the virus infusions to ensure sufficient expression of the transgenes. Rats were deeply anesthetized with isoflurane until reflexes disappeared and then perfused through the heart with an ice-cold solution containing (in mm): 103 N-methyl-D-glucamine, 2.5 KCl, 10 MgSO4, 30 NaHCO3, 1.2 NaH2PO4, 0.5 CaCl2, 25 glucose, 20 HEPES, 2 thiourea, 3 Na-pyruvate, and 12 N-acetyl-L-cysteine. The brains were then extracted from the skull and sectioned (300 µm) in the coronal plane using a vibrating microtome while immersed in the above solution. Slices were then kept in an oxygenated solution containing (in mm): 126 NaCl, 2.5 KCl, 1 MgCl2, 26 NaHCO3, 1.25 NaH2PO4, 2 CaCl2, and 10 glucose (pH 7.3, 300 mOsm) at 34°C for 5 min and returned to room temperature. At least 1 h later, individual slices were transferred to the recording chamber, which was perfused (6 ml/min) with the same solution (32°C).
Electrophysiology
Using infrared differential interference contrast microscopy, we obtained whole-cell recordings of neurons in MDm. Recording pipettes were pulled from borosilicate glass capillaries (resistance 5–8 MΩ) and filled with a solution containing (in mm): 130 K-gluconate, 10 HEPES, 10 KCl, 2 MgCl2, 2 ATP-Mg, and 0.2 GTP-tris(hydroxymethyl)aminomethane, pH 7.2 (280 mOsm). Membrane potential (Vm) values were not corrected for the junction potential (10 mV with this solution). MDm neurons were recorded in current-clamp mode with a MultiClamp 700B Amplifier (Molecular Devices). The data were digitized at 20 kHz with a Digidata-1550 interface controlled by pClamp-10.3 (Molecular Devices).
To activate ChR2-expressing BLA axons in MDm, we applied blue light stimuli (2 ms) at 0.2, 4, 8, 20, and 50 Hz using a blue LED (Mightex) coupled to the light path. Postsynaptic potentials were evoked from different membrane potentials (Vms), as determined by current injection. To study the electroresponsive properties of recorded cells, current pulses increasing in steps of ±20 pA (500 ms; 0.2 Hz) were applied from rest and other Vms, as determined by current injection.
Drugs
As detailed in Results, one or more of the following drugs were added to the perfusate to identify the neurotransmitter and receptors involved in mediating light-evoked responses: DL-2-amino-5-phosphonopentanoic acid (AP5; 100 μm) to block NMDA receptors, 6-cyano-7-nitroquinoxaline-2,3-dione disodium salt (CNQX; 10 μm) to block AMPA receptors, picrotoxin (100 μm) to interfere with GABA-A responses, (3-Aminopropyl)(diethoxymethyl)phosphinic acid (CGP35348; 10 μm) to block GABA-B responses, tetrodotoxin (TTX; 1 μm) to block voltage-gated Na+ channels, and 4-aminopyridine (4-AP; 100 μm) to block K+ channels involved in repolarization following action potentials. These drugs were obtained from Sigma.
Statistical analyses
Analyses were performed with Clampfit 10.3 (Molecular Devices), off-line. All values are expressed as mean ± SEM. We only considered cells that had a stable resting potential and generated overshooting action potentials. No data were excluded. All statistical tests are two-sided. We adopted a significance threshold of p = 0.05 for all statistical tests. We used Fisher exact tests to compare the incidence of particular properties in different samples. Rank sum tests were used to assess the significance of differences between two independent samples. Signed-rank tests were used to assess the significance of differences between two related samples. We also ran nonparametric ANOVAs (Friedman test) when comparing three matched samples of MDm neurons. To minimize subjective bias in the data analysis, we adopted a standard criterion for the identification of response onsets (voltage deviations > twice the noise, which then increased in amplitude without returning to baseline until the end of the response).
Data and code accessibility
The full dataset has not been deposited in a public repository because of its size but is available from the corresponding author on request.
Results
Retrograde labeling found in the amygdala following FB injections in MDm
A total of 24 rats received bilateral infusions of the retrograde tracer FB in MDm. The data described below was obtained in a subset of 15 rats in which the FB injections were centered on MDm bilaterally (for a representative example, see Fig. 1A). In most experiments, FB diffused into one or more neighboring thalamic nuclei like the habenula, central lateral, paracentral, paraventricular, or central medial thalamic nuclei. However, since these sites are devoid of inputs from the BLA (Krettek and Price, 1977; Pitkanen, 2000), this tracer diffusion could not have resulted in the retrograde labeling we observed in the BLA.
Fast Blue injection in MDm and resulting retrograde labeling in the amygdala. A1, Bilateral FB infusion site in MDm. A2, Cresyl violet-stained section, close to the one shown in A1. B, Distribution of retrogradely labeled cells in the amygdala and immediately adjacent structures at three rostro-caudal levels (relative to bregma): (B1) −1.8 mm, (B2) −2.8 mm, (B3) −3.7 mm. In B1–B3, each dot marks the location of a retrogradely labeled cell observed in a single 60-µm section. C, Morphology of retrogradely labeled BLA neurons. Calibration bars in B3 and C4 apply to all panels of B and C, respectively. AA, anterior amygdala area; AHA, amygdalohippocampal area; APir, amygdalopiriform transition area; ASt, amygdalostriatal transition area; BL, basolateral nucleus of the amygdala; BLP, posterior part of BL; BLV, ventral part of BL; BM, basomedial nucleus of the amygdala; cc, corpus callosum; CeC, CeM, and CeL, capsular, medial, and lateral sectors of the central nucleus of the amygdala; Cl, claustrum; CMT, central medial thalamic nucleus; Co, cortical nucleus of the amygdala; cp, cerebral peduncle; CPu, striatum; EGP, external part of the globus pallidus; EN, endopiriform nucleus; EP, entopeduncular nucleus; Hb, habenula; I, intercalated cell cluster; ic, internal capsule; IMD, intermediodorsal thalamic nucleus; LA, lateral nucleus of the amygdala; LP, lateral posterior thalamic nucleus; MDL, lateral part of the mediodorsal thalamic nucleus; MDm, medial part of the mediodorsal thalamic nucleus; ME, medial nucleus of the amygdala; opt, optic tract; Pir, piriform cortex; PMCo, posteromedial cortical amygdala; Po, posterior thalamic nuclear group; PVT, paraventricular thalamic nucleus; STIA, intraamygdaloid division of the bed nucleus of the stria terminalis; LP, lateral posterior thalamic nucleus; V, ventricle; VM, ventromedial thalamic nucleus; VP, ventroposterior thalamic nucleus.
Figure 1B plots the distribution of retrogradely labeled amygdala cells in a representative case. Consistent with prior descriptions in rats (McDonald, 1987) and monkeys (Aggleton and Mishkin, 1984; Russchen et al., 1987), FB+ cells were not only observed in the BLA, but also in the medial, central, and cortical amygdala nuclei as well as several adjacent structures like the amygdalohippocampal area, endopiriform nucleus, and claustrum (Fig. 1B). In the BLA, retrogradely labeled cells were most concentrated in its caudal two thirds (Fig. 1B). They were present in all three components of the BLA (lateral, basolateral, and basomedial nuclei; Fig. 1B) and often occurred in conspicuous clusters of 4–12 closely spaced neurons (Fig. 1C). Even within such clusters, FB-labeled BLA neurons had heterogeneous morphologic properties, including pyramidal, stellate, and spindle-shaped neurons of up to 50 µm in their long axis (Fig. 1C).
Expression of interneuronal markers among MDm-projecting BLA neurons
We aimed to determine whether some MDm-projecting BLA neurons are GABAergic. However, prior studies reported that in many GABAergic projection neurons, somatic levels of GABA are too low for immunohistochemical detection (Tóth and Freund, 1992). Thus, to minimize the likelihood of false negatives, we examined whether BLA neurons retrogradely labeled with FB also expressed various markers of BLA interneurons, which together should label the vast majority of BLA interneurons (see Materials and Methods). For each GABAergic marker, we assessed co-localization with FB in 7–15 sections from two to seven different animals. See details in Table 2.
Proportion of MDm-projecting BLA cells (FB+) that were immunoreactive for interneuronal markers
We first examined whether FB+ BLA cells are immunoreactive for the Ca2+-binding proteins CB and CR, since these two types of interneurons account for a large proportion (∼70%) of GABAergic cells in the BLA (Mascagni and McDonald, 2007). As shown in the representative examples of Figure 2A, not a single instance of CB-expressing FB+ cell was observed. Consistent with the fact that a high proportion (∼90%) of cells expressing PV are also immunoreactive for CB (McDonald and Betette, 2001), no FB+ cells expressed PV (Fig. 2B). Surprisingly however, although many SOM+ cells are known to also express CB (McDonald and Mascagni, 2002), a substantial proportion of FB+ cells (21.7 ± 4.3%; 12 sections from four rats) were immunopositive for SOM (Fig. 2C). Since a proportion of BLA SOM+ cells express NOS (Vereczki et al., 2021) and some cortical NOS+ neurons contribute long-range projections (Tomioka et al., 2005; Christenson Wick et al., 2019), we next assessed whether FB+ cells express NOS. However, no instance of FB-NOS co-localization was observed (Fig. 2D). As to CR, a low proportion of FB+ cells (10.9 ± 2.3%; seven sections from three rats) were immunoreactive for this Ca2+-binding protein (Fig. 3A,B). Since CR is expressed by many small CCK+ neurons (Mascagni and McDonald, 2003), we also tested whether FB+ cells expressed CCK. However none of them did (Fig. 3C,D).
Immunoreactivity for calbindin, parvalbumin, somatostatin, and nitric oxide synthase among MDm-projecting BLA neurons. Each row shows (1) BLA neurons retrogradely labeled following Fast Blue infusions in MDm (pseudo-colored green); (2) immunoreactivity for a particular interneuronal marker in the same BLA region (labels at bottom); (3) overlay of the previous two panels. A, Calbindin (CB). B, Parvalbumin (PV). C, Somatostatin (SOM). D, Nitric oxide synthase (NOS). Filled arrows point to double-labeled neurons. Empty arrows point to FB+ neurons that are not immunoreactive for the interneuronal marker considered. Calibration bars in panel 1 of A–D, apply to all panels of A–D.
Immunoreactivity for calretinin and cholecystokinin among MDm-projecting BLA neurons. Each row shows (1) BLA neurons retrogradely labeled following FB infusions in MD (pseudo-colored green); (2) Immunoreactivity for a particular interneuronal marker in the same BLA region; (3) overlay of the previous two panels. A, B, Calretinin (CR). C, D, cholecystokinin (CCK). Filled arrows point to double-labeled neurons. Empty arrows point to FB+ neurons that are not immunoreactive for the interneuronal marker considered. Calibration bars in panel 1 of A–D apply to all panels of A–D.
Responses of MDm neurons to optogenetic activation of BLA axons
The above findings suggest that the BLA sends a mixed GABAergic-glutamatergic projection to MDm. Seeking independent corroboration for these results, we examined the responses of MDm neurons to optogenetic activation of BLA axons using visually guided whole-cell recordings in vitro. Thus, six to eight weeks before the electrophysiological experiments, rats received infusions of pAAV-hSyn-hChR2(H134R)-mCherry or pAAV-mDlx-ChR2-mCherry-Fishell-3, viruses that drive expression of channelrhodopsin (ChR2) under the control of a synapsin promoter or a Dlx enhancer, respectively. For simplicity, we hereafter refer to rats infused with these viruses as “Synapsin rats” (n = 39) or “Dlx rats” (n = 30), respectively. The data described below was obtained in a subset of these rats (23 and 12, respectively) where robust mCherry expression was observed in the BLA and no infected somata were detected outside the BLA (Fig. 4A1,A2).
Virus infusion sites in BLA and resulting labeling in MDm. A, Examples of virus infusion sites with the synapsin promoter (A1) or Dlx enhancer (A2) and resulting labeling in MDm (B1, B2). Calibration bar in A2 also applies to A1. Calibration bar in B1 also applies to B2. ASt, amygdalo-striatal transition area; BL, basolateral nucleus of the amygdala; BM, basomedial nucleus of the amygdala; CeA, central nucleus of the amygdala; CMT, central medial thalamic nucleus; EC, external capsule; Hb, habenula; LA, lateral nucleus of the amygdala; MDL, lateral part of the mediodorsal thalamic nucleus; MDm, medial part of the mediodorsal thalamic nucleus; ME, medial nucleus of the amygdala; Pir, piriform cortex; PVT, paraventricular thalamic nucleus.
While a synapsin promoter drives transgene expression in glutamatergic and GABAergic neurons (Kügler et al., 2003), the Dlx enhancer is expected to do so preferentially (if not selectively) in GABAergic cells (Dimidschstein et al., 2016). Yet, the two viruses gave rise to anterograde labeling of similar intensity in MDm (Fig. 4B1,B2). Moreover, as detailed in Table 3, optogenetic activation of BLA axons with the two viruses elicited a similar response pattern in MDm neurons, including a similar proportion of evoked EPSPs. This unexpected observation suggests that while the Dlx enhancer may favor transgene expression in GABAergic cells (Dimidschstein et al., 2016), it does not prevent their expression in glutamatergic neurons. Since the results obtained in the two sets of animals were very similar, they are pooled together below.
Comparison between optogenetic responses of MDm neurons in synapsin and Dlx rats
We tested the impact of optogenetically activating BLA inputs in 177 MDm neurons that generated overshooting action potentials and had stable resting potentials negative to −60 mV. The electroresponsive properties of recorded cells matched earlier descriptions of thalamocortical neurons (Jahnsen and Llinás, 1984a, b). That is, they displayed two firing modes depending on their Vm: tonic firing when depolarized from positive to approximately –60 mV and a bursting firing pattern when depolarized from more negative Vms (Fig. 5A). Moreover, they generated rebound high-frequency spike bursts shortly after the end of hyperpolarizing current pulses (Fig. 5B).
Electroresponsive properties of MDm neurons. A, B, Current-evoked responses in two representative MDm neurons recorded in Synapsin and Dlx rats, respectively. A, Voltage response to depolarizing current pulses applied from a depolarized Vm (A1; −52 mV) or rest (A2; −70 mV). Typical of MDm neurons, this cell fires tonically from the depolarized level but generates high-frequency spike bursts from the more negative potential. Inset, High-frequency burst of action potentials depicted with a faster time base. B, At the break of negative current pulses, MDm cell generates rebound high-frequency spike bursts. Time calibration in A1 also applies to A2. Voltage calibration in A2 also applies to A1.
Blue light stimuli (2 ms) were applied in a wide range of frequencies (0.2–50 Hz). When light stimuli evoked a synaptic response, the cells' Vm was set to different values (−55 to −100 mV) by current injection to distinguish EPSPs from IPSPs. Qualitatively identical results were obtained in male and female rats (Table 4). Optogenetic activation of BLA axons elicited synaptic responses in 21% or 37 of 177 MDm neurons (Figs. 6–8). In descending order of incidence, response types included an EPSP only (73% or 27 of 37 of responsive cells; latency of 3.84 ± 0.35 ms; Figs. 6, 7), an isolated IPSP (16% or 6 of 37 responsive cells; latency of 4.61 ± 1.37 ms; Fig. 8B,D), or an EPSP overlapping with an IPSP (11% or four of 37 responsive cells; Fig. 8A,C). The latency of EPSPs and isolated IPSPs did not differ significantly (rank sum test, p = 0.79).
Comparison between optogenetic responses of MDm neurons in male and female rats
Light-evoked EPSPs are monosynaptic and mediated by ionotropic glutamatergic receptors. A, B, Light-evoked EPSPs in MDm neurons recorded in Dlx and Synapsin rats, respectively. Blue trace at bottom: light stimuli. A, Superimposed light-evoked EPSPs in control conditions (top), after addition of TTX and 4-AP to the perfusate (middle), and further addition of CNQX (10 μm) and AP5 (100 μm; bottom, red). B, Superimposed light-evoked EPSPs in control conditions (top, black), and after addition of CNQX and AP5 to the perfusate (bottom, red). In both panels, light stimuli were applied while recorded cells were at their resting potential.
Dependence of light-evoked EPSPs on membrane potential and stimulation frequency. A, Response evoked by blue light stimuli when the MDm neuron was depolarized to −55 mV (A1; 6 trials), at rest (A2; 9 trials), or hyperpolarized to −85 mV (A3; 3 trials). In A1, suprathreshold (top) and subthreshold responses (bottom) are separated. Inset in A1, superimposition of light-evoked EPSPs elicited from different Vms, but scaled to match their amplitude. B, EPSPs evoked by blue light stimuli applied at different frequencies (0.2 Hz, top; 4 Hz, middle; 8 Hz, bottom). Stimuli were delivered at rest (−64 mV). C, EPSPs evoked by blue light stimuli delivered at 4 Hz. Red trace at the top is the average of the five individual trials (black) shown below. Stimuli were delivered at rest (−70 mV). Same cell as in A.
Properties of light-evoked IPSPs. A–D, Responses of four different MDm neurons to blue light stimuli delivered at 8 Hz (A1, B3), 0.2 Hz (B1, B2, D2), and 4 Hz (C, D1). A, MDm cell in which blue light stimuli elicited a mixed EPSP-IPSP response, before (black) and after (red) addition of picrotoxin (100 μm) and CGP35348 (10 μm) to the perfusate. Light stimuli (blue) were delivered at 8 Hz. The response to the first two stimuli in A1 is shown with a faster time base in A2. B, MDm cell in which blue light stimuli elicited pure IPSPs (black, Control). B1, The late phase of the IPSP was sensitive to the GABA-B antagonist CGP35348 (10 μm; red, CGP) whereas its early phase was blocked by picrotoxin (black, CGP + Picro). B2, Light evoked IPSPs at different Vms (numbers on the left). B3, Response of the same MDm neuron to trains of blue light stimuli at 8 Hz. Red trace at the top is the average of the five individual trials (black) shown below. C, MDm cell in which light stimuli elicited a mixed EPSP-IPSP response. In this cell, IPSPs showed a marked attenuation with increased stimulation frequency. D1, MDm cell in which light stimuli elicited pure IPSPs (black trace) that persisted after addition of CNQX and AP5 (red trace). D2, Light evoked IPSPs at different Vms in the same cell.
Properties of EPSPs elicited in MDm neurons by the optogenetic excitation of BLA axons
Light-evoked EPSPs persisted after addition of TTX (1 μm) and 4-AP (100 μm) to the perfusate (six tested cells; Fig. 6A), indicating that they were monosynaptic. Moreover, they were abolished or largely reduced by addition of the AMPA and NMDA receptor antagonists CNQX (10 μm) and AP5 (100 μm) to the perfusate (seven tested cells; Fig. 6A,B). When measured at the resting potential, light-evoked EPSPs had a 20–80% risetime of 5.27 ± 0.82 ms (n = 31) and a low amplitude (2.81 ± 0.44 mV; n = 31). They never elicited short-latency action potentials unless the cells were depolarized enough for the EPSP to reach spike threshold (usually requiring depolarization to at least −60 mV; Fig. 7A1).
Presumably because light-evoked EPSPs interacted with the neurons' intrinsic voltage-dependent conductances, their decay times (80–20% amplitude) varied markedly depending on the Vm from which they were elicited (−85 mV, 39.3 ± 6.7 ms; −70 mV, 80.60 ± 12.3 ms; −55 mv, 110.52 ± 11.86 ms; Friedman test, p = 0.0002). For instance, in a subset of MDm cells (23%), light-evoked EPSPs had a characteristic “shoulder” when elicited from around −65 to −75 mV (that is in the bursting regime; Fig. 7A2), which was absent when cells were near spike threshold (Fig. 7A1) or hyperpolarized to around −85 mV and beyond (Fig. 7A3). The contrast is particularly striking in the inset of Figure 7A, which shows a superimposition of light-evoked EPSPs elicited from different Vms, but scaled to match their amplitude.
Light-evoked EPSPs showed a dependence on stimulation frequency, their amplitude dropping markedly when the stimulation frequency was increased (Fig. 7B,C). With an interstimulus interval (ISI) of 250 and 125 ms, the second EPSP's amplitude was 66.9 ± 9.8% (signed-rank test, p = 0.005) and 63.5 ± 8.2% (signed-rank test, p = 0.002) of the first, respectively (Fig. 7B). If the stimuli were applied in a range of Vm values where the first stimulus elicited a “shouldered” EPSP, the second EPSP's shoulder was also largely reduced or abolished (Fig. 7C).
Properties of IPSPs elicited by optogenetic excitation of BLA axons
Activation of BLA axons with blue light stimuli also elicited IPSPs, albeit in a significantly lower proportion of MDm cells (5.6%) than EPSPs (17.5%; Fisher exact test, p = 0.0007). Although light-evoked IPSPs appeared to follow EPSPs in many cases (40% of cells with IPSPs; Fig. 8A), several factors suggest that these IPSPs were monosynaptic. First, the rat MDm does not contain local-circuit GABAergic cells (Jones, 2007). Second, when the IPSPs were blocked pharmacologically (n = 3), the rising phase of EPSPs preblockade versus postblockade diverged within ∼1 ms of EPSP onset (Fig. 8A), the amplitude of the isolated EPSP becoming much higher than in control conditions (Fig. 8A). Third, in six MDm cells, blue light stimuli elicited IPSPs with no coincident EPSP (Fig. 8B,D). Fourth, in such cases, IPSP latencies (4.61 ± 1.37 ms; n = 6) were nearly identical to EPSP latencies (3.84 ± 0.39 ms; n = 27), inconsistent with a dependence of these IPSPs on the excitation of reticular thalamic neurons that project to MDm. Fifth, in these cells, light-evoked IPSPs persisted after addition of CNQX (10 μm) and AP5 (100 μm) to the perfusate (two of two tested cells; Fig. 8D1).
Light-evoked IPSPs had two components, an early picrotoxin-sensitive phase (100 μm; Fig. 8B1) that reversed at, or slightly negative to, the resting potential (Fig. 8B2,D2) and is presumably mediated by GABA-A receptors, and a late phase blocked by CGP35348 (10 μm) and therefore mediated by GABA-B receptors (Fig. 8B1). In contrast with the EPSPs whose amplitude invariably dropped as the stimulation frequency increased (Fig. 7B,C), light-evoked IPSPs were heterogeneous. Some showed little attenuation during repetitive stimulation (change from first to second IPSP: −13.27 ± 5.82% with 250 ms ISI; 8.09 ± 4.47% with 125 ms ISI; n = 5; Fig. 8B3). Others showed a marked attenuation with increased stimulation frequency (Fig. 8C), revealing an underlying EPSP that was largely masked by the IPSP when it was not attenuated (amplitude drop from first to second IPSP: 58.30 ± 9.69% with 250 ms ISI; 82.97 ± 13.88% with 125 ms ISI; n = 4).
Discussion
The present study aimed to characterize BLA projections to MDm. The interest of this question stems from the fact that the medial prefrontal cortex, which is involved in higher order cognitive functions (Nett and LaLumiere, 2021; Friedman and Robbins, 2022), receives convergent inputs from MDm and BLA (Heidbreder and Groenewegen, 2003). Hence, projections from BLA to MDm could enhance the coordination of activity in this tripartite circuit.
Although many studies reported on BLA projections to MDm (Krettek and Price, 1977; Porrino et al., 1981; Aggleton and Mishkin, 1984; McDonald, 1987; Russchen et al., 1987), contradictory evidence has accumulated regarding the nature of this projection. It was initially suggested that it originates from long-range GABAergic cells (McDonald, 1987), a conclusion based on their distinctive morphology and the finding that they are not retrogradely labeled by deposits of tritiated aspartate in MDm (Ray et al., 1992), when the same technique labels BLA neurons that project to cortex (Carnes and Price, 1988) or striatum (Christie et al., 1987; Fuller et al., 1987).
However, subsequent studies cast doubt on this conclusion. First, it was found that the ultrastructural features of synapses formed by BLA axons with MDm neurons are similar to those of subcortical excitatory afferents ending in dorsal thalamic relay nuclei (Kuroda and Price, 1991; Timbie et al., 2020). Moreover, BLA axons projecting to MDm were reported to be immunoreactive for VGLUT2, a marker of glutamatergic neurons (Timbie and Barbas, 2015). However, BLA neurons do not express VGLUT2 (Hur and Zaborszky, 2005), which led Mátyás et al. (2014) to suggest that these VGLUT2 axons originate from the neighboring endopiriform nucleus.
Therefore, the present study was undertaken to determine whether BLA neurons projecting to MDm are GABAergic or glutamatergic. Overall, our results suggest that the BLA sends a mixed projection to MDm, with a majority of the BLA neurons at the origin of this projection using glutamate as a transmitter, and a minority, GABA.
The BLA contains a small contingent of long-range GABAergic neurons that project to MDm
In the present study, converging results from two different experimental approaches support the conclusion that a proportion of MDm-projecting BLA neurons are long-range GABAergic cells. We first tested whether BLA neurons retrogradely labeled by FB infusions in MDm were immunopositive for various markers of GABAergic neurons. These analyses revealed that MDm-projecting BLA neurons do not express CB, PV, NOS, or CCK. However, ∼22% of them were immunopositive for SOM and ∼11% for CR. These findings are consistent with many prior studies that demonstrated the existence of long-range GABAergic cells in a variety of systems (Caputi et al., 2013; Melzer and Monyer, 2020), including the BLA (McDonald et al., 2012; McDonald and Zaric, 2015). Of note, while long-range GABAergic cells are neurochemically heterogeneous (Caputi et al., 2013), several studies (Bakst et al., 1986; Jinno and Kosaka, 2000; Tomioka et al., 2005, 2015; Melzer et al., 2012), including one on long-range GABAergic cells of the BLA (McDonald and Zaric, 2015), reported that long-range GABAergic cells express SOM, as found in the present study. By contrast, long-range CR neurons have been reported less frequently and they accounted for a minority of long-range GABAergic cells (Tóth and Freund, 1992; Zappone and Sloviter, 2001; Tomioka et al., 2005).
In a second approach, we examined the responses of MDm cells to optogenetic activation of BLA axons using visually guided whole-cell recordings in vitro. Consistent with the results of our immunohistochemical experiments, optogenetic activation of BLA axons in MDm elicited mixed glutamatergic-GABAergic responses or isolated GABAergic responses in 11% and 16% of responsive cells, respectively. Since there are no local-circuit GABAergic neurons in the rat MDm (Jones, 2007), the IPSPs could have resulted from the optogenetic activation of GABAergic axons arising in the BLA or from the glutamatergic recruitment of reticular thalamic neurons that project to MDm. Inconsistent with the latter possibility however, the latency of light-evoked EPSPs and IPSPs was statistically indistinguishable. Moreover, addition of antagonists of ionotropic glutamate receptors did not abolish the IPSPs.
Impact of BLA inputs on MDm neurons
Together, our results indicate that the BLA contains glutamatergic and GABAergic neurons that project to MDm. In vivo, the impact of these neurons on MDm cells will depend on a variety of factors, many of which are at present uncharacterized. These factors include the number, strength, and location of the synapses along the somatodendritic axis of MDm neurons, the state of the recipient neurons (tonic vs bursting regime), the dynamics of synaptic transmission (facilitating vs depressing synapses), and the connections (if any) formed by glutamatergic and GABAergic BLA cells with each other or with different populations of BLA cells. We consider these various aspects in turn below.
Consistent with the fact that a minority of MDm-projecting BLA neurons expressed markers of GABAergic interneurons, light stimuli typically elicited EPSPs, alone or coinciding with IPSPs; light stimuli also occasionally elicited isolated IPSPs. Together, these data suggest that the BLA projection to MDm mainly originates from glutamatergic neurons, but that glutamatergic and GABAergic BLA cells often converge on the same MDm neurons. In our optogenetic experiments, although light stimuli presumably caused a synchronous activation of most glutamatergic BLA axons to MDm, the EPSPs they elicited had a low amplitude and did not evoke spiking unless MDm neurons were depolarized beyond −60 mV by intracellular current injection. While the temporal overlap between light evoked EPSPs and IPSPs contributed to reduce EPSP amplitudes in some MDm cells, EPSPs continued to be of low amplitude after pharmacological abolition of GABA responses. Hence, when MDm cells are in the tonic regime, glutamatergic BLA inputs cause a slight depolarization, which requires summation with other inputs or regenerative dendritic conductances to fire MDm cells. By contrast, when MDm cells are in the bursting regime, light-evoked EPSPs can elicit high-frequency spike bursts, likely because the depolarization they cause activates the low-threshold Ca2+ conductance.
Overall, the low amplitude of light evoked EPSPs suggests that glutamatergic BLA axons form few synapses with each target MDm neurons or that the efficacy of these synapses is low because of presynaptic or postsynaptic factors such as low release probability or receptor expression, respectively. Further reducing the impact of glutamatergic BLA axons onto MDm neurons is the dynamics of transmission at the synapses they form. That is, the amplitude of light-evoked EPSPs drops markedly when stimulation frequency is increased. Of note, it is unlikely that the low amplitude of light-evoked EPSPs resulted from a remote dendritic location of these synapses given prior electron microscopic studies indicating that BLA axon terminals are found on large dendritic shafts (Kuroda and Price, 1991; Timbie et al., 2020).
While light stimuli simultaneously activated glutamatergic and GABAergic BLA axons in our in vitro experiments, they may be recruited independently in vivo. For instance, the two types of MDm-projecting BLA cells may be driven by different inputs. Moreover, long-range GABA cells may form inhibitory synapses with MDm-projecting glutamatergic BLA cells. Finally, given precedents indicating that long-range GABAergic cells project to multiple sites (Jinno et al., 2007; Unal et al., 2018), we should not discount the possibility that the recruitment of MDm-projecting BLA cells is part of a large-scale coordination of activity at distributed sites.
Footnotes
This work was supported by the National Institutes of Health National Institute of Mental Health Grant R01 MH-107239 (to D.P.). N.A. was supported by the Behavioral and Neural Sciences Graduate Program of Rutgers University.
The authors declare no competing financial interests.
- Correspondence should be addressed to Denis Paré at pare{at}rutgers.edu
