Optogenetic study of central medial and paraventricular thalamic projections to the basolateral amygdala

Abstract

The central medial (CMT) and paraventricular (PVT) thalamic nuclei project strongly to the basolateral amygdala (BL). Similarities between the responsiveness of CMT, PVT, and BL neurons suggest that these nuclei strongly influence BL activity. Supporting this possibility, an electron microscopic study reported that, in contrast with other extrinsic afferents, CMT and PVT axon terminals form very few synapses with BL interneurons. However, since limited sampling is a concern in electron microscopic studies, the present investigation was undertaken to compare the impact of CMT and PVT thalamic inputs on principal and local-circuit BL neurons with optogenetic methods and whole cell recordings in vitro. Optogenetic stimulation of CMT and PVT axons elicited glutamatergic excitatory postsynaptic potentials (EPSPs) or excitatory postsynaptic currents (EPSCs) in principal cells and interneurons, but they generally had a longer latency in interneurons. Moreover, after blockade of polysynaptic interactions with tetrodotoxin (TTX), a lower proportion of interneurons (50%) than principal cells (90%) remained responsive to CMT and PVT inputs. Although the presence of TTX-resistant responses in some interneurons indicates that CMT and PVT inputs directly contact some local-circuit cells, their lower incidence and amplitude after TTX suggest that CMT and PVT inputs form fewer synapses with them than with principal BL cells. Together, these results indicate that CMT and PVT inputs mainly contact principal BL neurons such that when CMT or PVT neurons fire, limited feedforward inhibition counters their excitatory influence over principal BL cells. However, CMT and PVT axons can also recruit interneurons indirectly, via the activation of principal cells, thereby generating feedback inhibition.

NEW & NOTEWORTHY Midline thalamic (MTh) nuclei contribute major projections to the basolateral amygdala (BL). Similarities between the responsiveness of MTh and BL neurons suggest that MTh neurons exert a significant influence over BL activity. Using optogenetic techniques, we show that MTh inputs mainly contact principal BL neurons such that when MTh neurons fire, little feedforward inhibition counters their excitatory influence over principal cells. Thus, MTh inputs may be major determinants of BL activity.

INTRODUCTION

Although a few thalamic nuclei located on or near the midline project very strongly to the amygdala (1), little is known about their impact on amygdala neurons. Several thalamic nuclei contribute to these projections. However, those originating from the central medial (CMT) and paraventricular (PVT) thalamic nuclei are especially dense, with CMT projecting almost exclusively to the basolateral nucleus of the amygdala (BL) and PVT projecting to BL and the central amygdala (CeA) (2–5).

Consistent with the strength of these projections, various lines of evidence suggest that CMT and PVT strongly influence BL and CeA activity. For instance, pharmacological or optogenetic inhibition of PVT neurons attenuates conditioned fear responses, an effect apparently mediated by PVT projections to CeA (6–9). Moreover, like amygdala neurons, PVT cells are activated by a variety of emotional stimuli, whether valenced positively or negatively (10–16). As to CMT, its connectivity suggests that it relays multimodal sensory information to BL. In particular, CMT receives strong inputs from the deep layers of the superior colliculus (17) and the parabrachial nucleus (18), potentially allowing CMT to provide BL with visual, auditory, gustatory, visceral, and nociceptive information (1).

So far, most studies on the thalamic innervation of the amygdala have dealt with inputs from the posterior thalamic complex (particularly the medial geniculate nucleus and posterior paralaminar nucleus) to the lateral amygdala (LA). In brief, neurons of the posterior thalamic complex contribute glutamatergic projections to LA (19–23), and their axon terminals only form asymmetric synapses with LA neurons, typically with the dendritic spines of principal cells and, in a third of cases, with dendritic profiles (24) that presumably belong to interneurons (25).

Although this termination pattern is consistent with that seen for cortical inputs (26–28), it contrasts with the ultrastructural properties of PVT and CMT axon terminals in BL. Indeed, Amir et al. (29) found that CMT and PVT inputs to BL end almost exclusively on principal cells, implying that when PVT or CMT cells fire, next to no feedforward inhibition opposes their excitatory impact on principal BL neurons. If this was the case, this property would be unique among glutamatergic inputs to BL, lending additional support to the notion that CMT and PVT are key determinants of BL activity. However, since limited sampling is always a concern in electron microscopic studies, the present experiments were undertaken to directly test this idea. To this end, we studied the impact of CMT and PVT inputs on BL neurons, using optogenetic methods and whole cell recordings in vitro.

MATERIALS AND METHODS

Animals and Virus Injections

Procedures were approved by the Institutional Animal Care and Use Committees of Rutgers University. We used Long-Evans rats of either sex (Charles River Laboratories, Fairfield, NJ; RRID: RGD_2308852; 300–350 g at the start of the experiments) with ad libitum access to water and food (regular rat chow). Rats were kept on a 12:12-h light-dark cycle (lights off at 7:00 PM), and experiments were conducted during the light phase of the cycle.

One week or more after delivery, rats were anesthetized with isoflurane and placed into a stereotaxic apparatus. Body temperature was kept at 37–38°C. Atropine sulfate (0.05 mg/kg im) was administered to aid breathing. The scalp was cleaned with Betadine and alcohol. The region to be incised was injected with the local anesthetic bupivacaine (0.125% solution sc). An incision was then made on the midline, and small burr holes were drilled above the thalamus.

Using glass pipettes pulled to an outer tip diameter of ∼70 µm by a PE-22 puller (Narishige Instruments, Amityville, NY), we made stereotaxically guided pressure injections (50 nL) of AAV5-CaMKIIa-hChR2(H134R)-mCherry (Addgene, Watertown, MA) in CMT (AP: −2.4, ML: 2.0, DV: −6.3; in mm relative to bregma) or PVT (AP: −3.2, ML: 1.6, DV: −5.15), both with a 18° lateromedial angle. The virus was infused at a rate of 9.6 nL/5 s with a Nanoject II (Drummond Scientific Company, Broomall, PA). The scalp was then sutured, a local antibiotic (Neosporin paste) was applied to the wound, and an analgesic was administered (ketoprofen, 2 mg/kg sc daily for 3 days).

Slice Preparation

To allow for adequate expression of the transgene, electrophysiological experiments were conducted ∼8 wk after the virus infusion surgery. Rats were deeply anesthetized with isoflurane. After abolition of reflexes, the chest cavity was opened and rats were perfused through the heart with an ice-cold solution containing (in mM) 103 N-methyl-d-glucamine (NMDG), 2.5 KCl, 10 MgSO₄, 30 NaHCO₃, 1.2 NaH₂PO₄, 0.5 CaCl₂, 25 glucose, 20 N-2-hydroxyethylpiperazine-N′-2-ethane sulfonic acid (HEPES), 2 thiourea, 3 Na-pyruvate, and 12 N-acetyl-l-cysteine. Brains were then sectioned in the coronal plane with a vibrating microtome (300-µm thickness) while submerged in the above solution. Slices were then kept submerged 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) in a holding chamber at 34°C for 5 min and returned to room temperature. One hour or more later, slices were transferred one at a time to the recording bath, which was perfused with the same oxygenated solution (6 mL/min; 32°C).

Electrophysiology

Whole cell recordings of BL neurons were obtained under visual guidance by infrared-differential interference contrast microscopy. Recording pipettes were pulled from borosilicate glass capillaries (resistance 5–8 MΩ). The intracellular solution contained (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). The liquid junction potential was 10 mV with this solution. Membrane potential values were not corrected for the junction potential. Cells were recorded in current- and/or voltage-clamp modes with a MultiClamp 700B Amplifier (Molecular Devices, San Jose, CA). The data were digitized at 20 kHz with a Digidata 1550 interface controlled by pCLAMP 10.3 (Molecular Devices, San Jose, CA).

To activate ChR2-expressing axons, blue light stimuli (5 ms) were applied at various frequencies (1, 10, 20, 50 Hz) with a blue LED coupled to the objective light path (Mightex, Toronto, ON, Canada). We also tested the effects of a longer light pulse (500 ms) to determine whether the decay of light-evoked responses was affected by the decay kinetics of the photo-current. Postsynaptic potentials or currents were evoked from different membrane potentials. The inhibitory postsynaptic potential (IPSP) or inhibitory postsynaptic current (IPSC) reversal potentials were calculated from a linear fit of fluctuations in IPSP or IPSC amplitudes as a function of membrane potential.

Measurement of Physiological Properties

To characterize the physiological properties of recorded cells, we applied graded series of current pulses (generally in ±40 pA increments; 500 ms; 0.2 Hz) from rest as well as from more negative and positive membrane potentials, as determined by DC current injection. The input resistance (R_in) of the cells was estimated in the linear portion of current-voltage plots. The duration of action potentials was measured at half-amplitude, only considering the first current-evoked spike elicited from rest by the lowest-amplitude suprathreshold current pulse. The rationale for this approach is that in many cells, particularly bursting cells, spike durations increased during repetitive firing. The same spike was used to measure the amplitude of each cell’s action potentials and spike afterhyperpolarization (AHP). Spike amplitude was defined as the difference between spike peak and spike threshold. AHP amplitude was defined as the difference between the membrane potential 1 ms before the spike peak and the most negative potential within an interval of 10 ms after the spike peak. Maximal firing rate was measured by averaging the interspike intervals of the spikes elicited by depolarizing current pulses of 200 pA and converting the value to frequency. Sag ratio corresponds to the ratio between the most negative membrane potential during the last 100 ms and the first 100 ms observed during negative current pulses (≥200 pA; 500 ms).

Drugs

Depending on the purpose of the experiments, one or more of the following drugs were added to the perfusate: 6-cyano-7-nitroquinoxaline-2,3-dione disodium salt (CNQX, 10 µM) to block α-amino-3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) receptors, (±)-3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP, 10 µM) to block N-methyl-d-aspartate (NMDA) receptors, picrotoxin (100 µM) to interfere with GABA-A 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. All drugs were obtained from Sigma (St. Louis, MO) and dissolved in the perfusate at the concentrations stated above.

Microscopic Observations

Prior to the recordings, we verified that the virus infusions had reached their intended target, using fluorescence microscopy (Axioscope; Zeiss, San Diego, CA). For more detailed examinations of the infusion sites, slices were immersed in 4% paraformaldehyde overnight and then examined with a fluorescence microscope (Nikon Eclipse E800, Melville, NY). The boundaries of CMT and PVT were drawn on bright-field images that were then superimposed on the fluorescence images to assess the location and extent of the virus infusion. Below, we only report data obtained in rats where the virus infusion site was centered on the intended target.

Analyses and Statistics

Analyses were performed off-line with the software IGOR (WaveMetrics, Lake Oswego, OR), MATLAB (MathWorks, Natick, MA), and Clampfit 10.3 (Molecular Devices, San Jose, CA). To minimize subjective bias in the data analysis, we adopted a standard criterion for the identification of response onsets (y-axis deviations > twice the noise, which then increased in amplitude without returning to baseline until the end of the response) and enforced this criterion with MATLAB scripts. Values are expressed as means ± SE. All cells with stable resting potentials that generated overshooting spikes were included in the analyses. No data were excluded. All statistical tests are two-sided. We used chi-square tests to compare the incidence of particular properties in unrelated samples. Unpaired t tests were used to assess significance of differences between different samples with a significance threshold of P = 0.05. Two-way repeated-measures ANOVAs were used to compare current-evoked firing rates between principal cells and interneurons.

To classify neurons into principal cells and interneurons, we tried using an unsupervised clustering approach (kmeans function in MATLAB, 2 groups) based on spike half-width, AHP strength, and peak firing rate. This did not identify a clear bimodal distribution, so we instead adopted a threshold criterion based on prior literature (e.g., Ref. 30) of <0.6-ms spike half-width for classifying cells as interneurons. To determine whether this criterion distinguished variations in action potential waveforms more generally, we used a t-distributed stochastic neighbor embedding (t-SNE) algorithm to project action potential waveforms (1 ms before and after peak) onto a two-dimensional space. The tsne function in MATLAB was used for this, with perplexity set to 20 (similar results were obtained with perplexity values of 5, 10, and 30).

RESULTS

Virus Infusions and Resulting Labeling in the Amygdala

We performed infusions of AAV5-CaMKIIa-hChR2(H134R)-mCherry aimed at CMT or PVT in 36 and 29 rats, respectively. The data described below were obtained in a subset of these rats where the infusion sites were centered on their intended target (CMT, 24 rats, Fig. 1, A1 and A2; PVT, 19 rats, Fig. 1, B1 and B2) and in which robust light-evoked responses could be reliably elicited in BL neurons. Although some of the virus injections were not perfectly confined to CMT or PVT, this is not a major concern here, as neighboring thalamic nuclei contribute weaker projections to BL than CMT and PVT. Specifically, the only neighboring nuclei contributing projections to BL are the rhomboid nucleus, ventrally adjacent to CMT, and the paratenial nucleus, laterally adjacent to PVT. However, retrograde and anterograde tracing studies converge to show that CMT projects much more strongly to BL than the rhomboid nucleus (3, 4). Similarly, although anterograde tracer injections in the paratenial nucleus produce dense axonal labeling in the medial third of BL, PVT contributes dense projections to the entire mediolateral extent of BL (4). Nevertheless, we cannot exclude the possibility that some of the responses described below originate from the rhomboid and paratenial nuclei.

Figure 1.

AAV-ChR2 infusions in central medial (CMT) and paraventricular (PVT) thalamic nuclei lead to contrasting labeling in the amygdala. A1–A4: AAV-ChR2 infusion site in CMT (A1), adjacent section counterstained with cresyl violet (A2), and resulting labeling in basolateral nucleus of the amygdala (BL) at low (A3) and high (A4) magnification. B1–B3: AAV-ChR2 infusion site in PVT (B1), adjacent section counterstained with cresyl violet (B2), and resulting labeling in BL and central nucleus of the amygdala (CeA) (B3). Crosses in A3 and B3 indicate the orientation of these panels (L, lateral; M, medial; D, dorsal; V, ventral); cross in A3 also valid for A4. BM, basomedial nucleus of the amygdala; DG, dentate gyrus; EC, external capsule; H, habenula; LA, lateral nucleus of the amygdala; LD, laterodorsal thalamic nucleus; LP, lateroposterior thalamic nucleus; MD, mediodorsal thalamic nucleus; nRT, reticular thalamic nucleus; Pir, piriform cortex; Po, posterior thalamic nucleus; SM, stria medullaris; VM, ventromedial thalamic nucleus; VP, ventroposterior thalamic nucleus.

Consistent with earlier anterograde tracing studies (3–5, 29), the pattern of labeling observed in the amygdala after CMT and PVT virus infusions differed. That is, virus infusions in CMT (Fig. 1, A1 and A2) led to prominent anterograde labeling largely restricted to BL (Fig. 1, A3 and A4), whereas those in PVT (Fig. 1, B1 and B2) also led to labeling in CeA (Fig. 1B3).

Classification of Recorded Neurons

We tested the impact of activating CMT and PVT inputs in 196 BL neurons that generated overshooting action potentials and had stable resting potentials negative to −60 mV (Figs. 2 and 3). Since we aimed to compare the responsiveness of principal and local-circuit BL neurons, we examined the cells’ properties in search of variables that would betray the presence of two subsets of neurons in our sample and that we could use to distinguish them (Table 1). However, the electroresponsive properties of interneurons are extremely diverse (reviewed in Ref. 31). For instance, there are four physiological subtypes of parvalbumin-expressing (PV⁺) interneurons (30, 32) and three subtypes of local-circuit cells expressing cholecystokinin (CCK⁺) (33). Similarly, a single-cell RT-PCR study distinguished five subtypes of interneurons (34). Whereas the dynamics of current-evoked spiking in some of these interneuron subtypes clearly differ from those of principal cells (e.g., fast-spiking and stuttering PV⁺ cells; Ref. 30), others show similar activity patterns. Consistent with this, none of the parameters we considered was bimodally distributed (Fig. 2, A–C). Similarly, unsupervised k-means clustering with these parameters also failed to reveal clearly distinct clusters.

Figure 2.

Classification of recorded cells as putative projection cells or interneurons. A: frequency distribution of maximal firing rates in our sample of basolateral amygdala (BL) neurons (n = 196) during a depolarizing current pulse of 200 pA that lasted 500 ms. For each cell, interspike intervals were averaged and converted to frequency. B: frequency distribution of afterhyperpolarization (AHP) amplitudes computed as the difference in potential between 1 ms before the spike peak and the most negative potential within 10 ms after the spike peak. Positive values were obtained in cells generating spike bursts riding on a slow depolarizing plateau potential. C: frequency distribution of action potential (AP) duration at half-amplitude. For B and C, we only considered the first current-evoked spike elicited from rest using the lowest-amplitude suprathreshold current pulse. White and black bars (0.6 ms cutoff) indicate presumed interneurons (ITNs; n = 37) and principal cells (P-cells; n = 159), respectively. D: unsupervised clustering of AP waveforms (±1 ms around peak) from all neurons, performed using t-distributed stochastic neighbor embedding (t-SNE) projected onto 2 dimensions. E1–E3: firing rate ± SE (Hz, y-axis) of presumed ITNs (dashed lines and empty circles; n = 37) and P-cells (solid line and filled circles; n = 141) as a function of injected current (x-axis). E1: all available neurons with these current amplitudes [2-way repeated-measures ANOVAs, F_between(1,880) = 43.57, P < 0.001]. E2: subset of the cells shown in E1 with a bursting firing pattern [6 presumed ITNs and 12 P-cells; 2-way repeated-measures ANOVAs, F_between(1,80) = 16.08, P < 0.001]. E3: subset of the cells shown in E1 generating repetitive single spikes with frequency accommodation [15 presumed ITNs and 117 P-cells; 2-way repeated-measures ANOVAs, F_between(1,655) = 4.53, P < 0.05].

Figure 3.

Contrasting electroresponsive properties of presumed principal cells (A1 and A2) and interneurons (B–E) recorded in the basolateral amygdala (BL) nucleus. Voltage responses of 6 different cells to a graded series of current pulses applied from rest. Insets show spikes generated by each cell (arrows). Current monitor in A1 applies to A2. Current monitor in C applies to B–E. Current pulses were applied from rest (in mV: A1, −71; A2, −73; B, −70; C, −72; D, −72; E, −64).

Table 1.

Comparison between the electroresponsive properties of presumed principal cells and interneurons

	Resting Potential, mV	Input Resistance, MΩ	AP Amplitude, mV	AP Duration, ms	Maximal Firing Rate, Hz	AHP Amplitude, mV	Sag Ratio
Principal cells (n = 159)	−69.7 ± 0.30	122.6 ± 4.86	82.9 ± 0.81	0.97 ± 0.026	40.2 ± 6.14	−6.7 ± 0.45	0.90 ± 0.13
Interneurons (n = 37)	−68.7 ± 0.78	121.3 ± 10.61	79.1 ± 1.85	0.46 ± 0.025	80.9 ± 11.5	−12.2 ± 1.32	0.92 ± 0.12
	t = 1.197 P = 0.238	t = 0.111 P = 0.9122	t = 1.882 P = 0.0658	t = 14.14 P < 0.00001	t = 3.122 P = 0.0028	t = 3.944 P = 0.0004	t = 0.113 P = 0.91

Values are averages ± SE. Action potential (AP) duration was measured at half-amplitude using the first spike generated in response to the lowest-amplitude suprathreshold current pulse applied from rest. Maximal firing rate is the average firing rate during a 500-ms current pulse of 200 pA applied from rest. The results of t tests comparing presumed principal cells and interneurons for the properties listed are shown at bottom. AHP, afterhyperpolarization.

However, in the above RT-PCR study (34), four of the five subtypes of interneurons generated briefer spikes than principal cells. This was found to be the case for all four PV⁺ cell subtypes (30), most somatostatin-positive (SOM⁺) cells (34), and many cells expressing vasoactive intestinal peptide (VIP⁺; Ref. 35) but not for CCK⁺ cells (33, 36). Since CCK⁺ cells account for a very small fraction of interneurons in BL, we adopted spike duration to classify the two cell types.

Because earlier studies relied on different ways to measure spike durations (e.g., full spike, half-width, duration at half-amplitude), we based our cutoff on the distribution of spike durations in our sample, measured at half-amplitude, only considering the first current-evoked spike elicited from rest by the lowest-amplitude suprathreshold current pulse. Specifically, cells with action potentials ≥ 0.6 ms were classified as projection cells (n = 159) and neurons with spikes < 0.6 ms as interneurons (n = 37). We selected a 0.6 ms cutoff because it coincided with a marked change in the incidence of cells with low versus high spike durations (Fig. 2C).

We determined whether the grouping of cells using this criterion was reflected in an unsupervised clustering of action potential waveforms. To do this, we used a nonlinear embedding technique, termed t-distributed stochastic neighbor embedding (t-SNE), which creates a low-dimensional representation of high-dimensional data while preserving the local structure (37). Action potential waveforms (±1 ms around peak) from all neurons were subjected to t-SNE projected onto two dimensions (Fig. 2D). Cells that were classified as interneurons overwhelmingly tended to occupy one region of the scatterplot. This suggests either that the spike half-width was a useful feature for parameterizing the shape of action potential waveforms or that additional features, correlated with our half-width cutoff, also distinguished interneurons. Both possibilities bolster our case that the demarcation of cell types based on spike width at half-amplitude generalizes to a data-driven classification scheme agnostic to our bespoke classification metric.

The two cell groups displayed contrasting firing patterns upon depolarization (Fig. 3), the main difference being that interneurons could sustain significantly higher current-evoked firing rates [Fig. 2E1; 2-way repeated-measures ANOVAs, F_between(1,880) = 43.57, P < 0.001]. Indeed, principal cells generated spike trains at low frequencies and with varying degrees of spike frequency accommodation. Some began firing with a high-frequency spike burst (25%, or 40 of 159; Fig. 3A1), but most generated only single spikes (75%, or 119 of 159; Fig. 3A2). In contrast, presumed interneurons were more diverse, some firing repeatedly at high rates with nearly no frequency accommodation [fast-spiking (FS) cells; 16%, or 6 of 37; Fig. 3B], others responding similarly to FS cells except for conspicuous firing pauses (stuttering cells; 14%, or 5 of 37; Fig. 3C), some generating high-frequency spike bursts at the onset of depolarizing current pulses (bursting cells; 14%, or 5 of 37; Fig. 3D), and a last group generating spike trains with modest frequency accommodation (57%, or 21 of 37; Fig. 3E). This heterogeneity is consistent with prior studies on BL interneurons (reviewed in Ref. 31).

Of note, the difference in current-evoked firing rates we observed when comparing all available interneurons and principal cells irrespective of firing pattern (Fig. 2E1) persisted when we restricted this analysis to cells that had similar firing patterns, namely bursting interneurons versus bursting principal cells [Fig. 2E2; 2-way repeated-measures ANOVAs, F_between(1,80) =16.08, P < 0.001] or cells generating trains of single spikes with frequency accommodation [Fig. 2E3; 2-way repeated-measures ANOVAs, F_between(1,655) =4.53, P < 0.05]. The rationale for and limitations of using spike duration to distinguish principal cells and interneurons are considered in discussion.

Light-Evoked Responses in Principal BL Cells

Whether the virus was infused in CMT (Fig. 4, A1–A3) or PVT (Fig. 4, B1–B3), blue light stimuli (5 ms) elicited responses in a similar proportion of principal neurons (CMT 59%, or 46 of 78; PVT 58%, or 47 of 81; χ² = 0.02; P = 1). From rest, these responses had similar amplitudes [excitatory postsynaptic potentials (EPSPs): CMT 7.18 ± 0.87 mV, Fig. 4, A1 and A2; PVT 7.17 ± 1.15 mV, Fig. 4, B1 and B2; excitatory postsynaptic currents (EPSCs): CMT 144.02 ± 25.63 pA; PVT 171.04 ± 26.12 pA; t tests, P’s ≥ 0.489]. Moreover, light-induced activation of CMT or PVT axons fired a similarly low proportion of principal cells at rest (CMT 24%, or 19 of 78, Fig. 4A3; PVT 19%, or 15 of 81, Fig. 4B3; χ² = 0.81; P = 0.369).

Figure 4.

Optogenetic activation of central medial (CMT; A1–A3) or paraventricular (PVT; B1–B3) thalamic axons elicits similar responses in principal basolateral amygdala (BL) neurons. Four different principal BL neurons are shown in A1 and A2, A3, B1 and B2, and B3. Blue light stimuli were applied at rest (in mV: A1 and A2, −68; A3, −76; B1 and B2, −71; B3, −69). In A1 and B1, light stimuli (5 ms) were applied at 8 (top), 20 (middle), and 50 (bottom) Hz. In A2 and B2, the light stimulus lasted 500 ms. In A3 and B3, a 5-ms light stimulus elicited firing from rest. Calibration bars in B2 apply to A1 and A2 and B1 and B2. Calibration bars in B3 apply to A3 and B3. All traces in A1 and B1 are the average of 5 trials. Traces in A2 and B2 are the average of 3 trials.

Several factors may explain why only a proportion of principal cells were activated by CMT and PVT axons. However, the intensity/duration of the light stimuli we used is probably not one of them. Indeed, in prior experiments with the same opsin (38, 39), we found that light stimuli of the same intensity and duration activated all infected cells at the site of the virus infusion. Therefore, it is more likely that only a subset of BL neurons receive CMT or PVT inputs.

Repetitive activation of CMT or PVT axons at frequencies ranging between 8 and 50 Hz generally failed to transform subthreshold responses into suprathreshold ones (Fig. 4, A1 and B1). Similarly, lengthening the duration of light stimuli up to 500 ms did not alter the character of the response (Fig. 4, A2 and B2). The close parallel between the results obtained in the CMT and PVT experiments is consistent with the nearly identical ultrastructural features of CMT and PVT axon terminals found in BL (29). Accordingly, the results obtained in the two sets of animals are pooled below.

Consistent with the results obtained with posterior thalamic afferents to the lateral amygdala (20–22), the EPSCs or EPSPs elicited by activation of CMT or PVT inputs were nearly or completely abolished by addition of AMPA and NMDA receptor antagonists to the perfusate (CNQX, 10 µM; CPP, 10 µM), in all tested principal cells (n = 11; Fig. 5, A1 and A2). In a subset of responsive cells (48%, or 45 of 93), membrane depolarization beyond –70 mV revealed that the initial excitatory response was followed by an IPSP or IPSC that reversed at −67.4 ± 0.9 mV (Fig. 5, B1 and B2), suggesting a mediation by GABA-A receptors.

Figure 5.

Responses elicited by the activation of central medial (CMT) and paraventricular (PVT) thalamic axons are mediated by ionotropic glutamate receptors (A1 and A2) and are followed by synaptic inhibition (B1 and B2). Three different principal cells are shown in A1, A2, and B1 and B2. A1 and A2: recordings in voltage (A1)- and current (A2)-clamp modes. Responses evoked by activation of CMT axons before (Control) and 5 min after application of 6-cyano-7-nitroquinoxaline-2,3-dione disodium salt (CNQX) and (±)-3-(2-carboxypiperazin-4-yl)propyl-1-phosphonic acid (CPP). Light stimulus in A1 and A2 was applied while cells were at rest (−68 and −69 mV, respectively). B1: voltage-clamp recording of responses elicited by activation of PVT axons at 3 different holding potentials (−85, −70, and −50 mV). Rest in this cell was −71 mV. B2: using the data obtained in B1, relation between current amplitude at inhibitory postsynaptic current (IPSC) peak (vertical dashed line in B1) and holding potential. The oblique line is a linear fit to the data (circles). Dashed lines mark zero intercept (−70.7 mV in this cell). Traces in A1 and B1 are the average of 5 trials each. Control and experimental traces in A2 are the average of 3 and 5 trials, respectively.

Of note, the average latency of IPSPs and IPSCs (19.1 ± 1.20 ms; n = 45) was longer than that of blue light-evoked spikes (10.5 ± 1.90 ms; n = 34; t test, t = 4.05; P = 0.00011). In keeping with the lack of GABAergic neurons in the rat CMT and PVT (40, 41), we observed only one case of isolated IPSP or IPSC that is an inhibitory response with no preceding excitation. However, even in this case, the IPSC latency was 8.7 ms longer than the average latency of blue light-evoked spikes.

Light-Evoked Responses in Presumed Interneurons

Overall, the results presented so far suggest that CMT and PVT axons excite principal BL neurons via ionotropic glutamate receptors, leading a subset of them to fire. This initial excitation is sometimes followed by GABAergic inhibition. This inhibition could be a consequence of the monosynaptic activation of local-circuit cells by CMT and PVT inputs (feedforward inhibition) or arise from the activation of principal neurons, which in turn excite interneurons (feedback inhibition). To examine these possibilities, we next consider the responses of BL interneurons to the activation of CMT and PVT inputs.

Activation of CMT or PVT axons by blue light stimuli (5 ms) elicited responses in 43% or 16 of 37 tested interneurons, a lower proportion than in principal cells (58%, or 93 of 159) but not significantly so (χ² = 2.83, P = 0.093; Fig. 6A, left, Fig. 6, B1–B4). Similarly, the proportion of interneurons fired from rest by activation of CMT or PVT axons (14%, or 5 of 37) was lower than in principal cells (21%, or 34 of 159), but again the difference did not reach significance (χ² = 1.17, P = 0.28; Fig. 6A, right). Whereas the amplitude of EPSPs (7.1 ± 1.4 mV) and EPSCs (94.4 ± 24.6 pA) elicited in interneurons did not differ significantly from those seen in principal cells (principal cells: 7.0 ± 0.8 mV, t test, t = 0.03, P = 0.974; 142.8 ± 16.7 pA; t test, t = 1.627, P = 0.11), their latency was significantly longer (interneurons: 3.90 ± 0.68 ms, n = 16; principal cells: 2.69 ± 0.18 ms, n = 45; t test, t = 2.35, P = 0.021; Fig. 6, C1–C3), yet the ranges of latencies observed in the two cell types overlapped (principal cells 1.1–7.8 ms; interneurons 1.1–10.6 ms).

Figure 6.

Comparison between the responses of principal cells (P-cells) and interneurons (INTs). A: proportion of P-cells (n = 159) and INTs (n = 37) responsive to light stimuli (left; χ² = 2.83, P = 0.093) or fired from rest (right; χ² = 1.17, P = 0.28). B1–B3: responses of an INT to trains of blue light stimuli (5 ms) delivered at 8 (B1), 20 (B2), and 50 (B3) Hz. B4: response of the same cell to a continuous 500-ms light pulse. Arrowheads in bottom 2 trials of B2 indicate spike truncation. Traces shown in B1 and B4 are the average of 3 trials. All traces shown in B2 and B3 are individual trials. C1–C3: light-evoked responses of a P-cell (C1) and 2 INTs (C2 and C3). Note longer response latency in INTs. Trace shown in B1 is the average of 5 trials, whereas the traces shown in C2 and C3 are individual trials. Part of the bottom trace of B3 is magnified in C3 (arrow). Stimulated axons originated in central medial (CMT; B1–B3 and C3) or paraventricular (PVT; C1 and C2) thalamic nucleus. All cells displayed in B1–B3 and C1–C3 were recorded in current-clamp mode. D: proportion of P-cells (left; n = 20) and INTs (right; n = 8) that remained (R) or did not remain responsive (UR) to light stimuli after application of tetrodotoxin (TTX) and 4-aminopyridine (4-AP) (χ² = 5.43, P = 0.019).

Impact of Abolishing Polysynaptic Interactions

Overall, the above comparisons are inconclusive as to the origin of the inhibition seen in principal cells when CMT and PVT axons are activated. In favor of the feedback inhibition model, activation of CMT and PVT axons excited a lower proportion of interneurons than principal cells, spike latencies in principal cells were much shorter than the latencies of IPSPs and IPSCs, and response latencies were on average lower in principal cells than interneurons. Yet, other observations are compatible with the feedforward inhibition model, namely the fact that response amplitudes were similar in the two cell types and that a few interneurons had response latencies as low as principal cells.

To shed further light on this question, we compared the impact of abolishing polysynaptic interactions by adding TTX to the perfusate. Since blocking voltage-gated Na⁺ channels is expected to reduce light-evoked transmitter release from thalamic axons, we also added the K⁺ channel blocker 4-AP to counteract this effect, as is standard in the field (for instance, see Refs. 42, 43). We compared light-evoked responses before versus after TTX in 20 principal cells and 8 interneurons.

As shown in Fig. 6D, TTX abolished light-evoked responses in a significantly higher proportion of interneurons (4 of 8) than in principal cells (2 of 20; χ² = 5.43, P = 0.019). In keeping with the idea that neurons whose responses were abolished by TTX had been indirectly recruited by principal cells, their pre-TTX response latencies (5.92 ± 1.39 ms, n = 6) were more than double those seen in neurons with TTX-resistant responses (2.43 ± 0.31 ms, n = 22; t test, t = 2.453, P = 0.0297). In terms of electroresponsive properties, interneurons with TTX-sensitive or -resistant responses were heterogeneous. Specifically, interneurons with TTX-resistant responses included a fast-spiking interneuron (FSI), a stuttering cell, and two accommodating interneurons. Interneurons with TTX-sensitive responses included two FSIs, a bursting cell, and an accommodating interneuron.

Figure 7 provides a few examples of pre (black)- versus post (red)-TTX responses in principal cells (Fig. 7, A1–A3) and interneurons (Fig. 7, B1–B3) recorded in current-clamp mode. The impact of TTX is apparent in the comparison between the black (control) and red (TTX) traces. As shown in Fig. 7A1, in the principal cells in which light stimuli elicited inhibition in control conditions (9 of 20), the inhibition disappeared after addition of TTX to the perfusate. Also, in both cell types, TTX reduced or abolished the gradual increase in response amplitude typically seen when light stimuli are applied at 8 or 20 Hz (Fig. 7, A2 and A3 and B2 and B3). Overall, light-evoked responses also had a significantly lower amplitude after TTX in interneurons (7.9 ± 1.3 pA, n = 8) than in principal cells (87.2 ± 22.4 pA, n = 20; t test, t = 3.53, P = 0.002).

Figure 7.

Impact of tetrodotoxin (TTX) on light-evoked responses. Three different projection cells (A1–A3) and interneurons (B1–B3) recorded in current-clamp mode. Black and red lines indicate responses recorded in control conditions and after addition of TTX and 4-aminopyridine (4-AP) to the perfusate, respectively. Blue lines indicate when the light stimuli were applied. Calibration bars in A1 also apply to B1 and B2. All traces are the average of 3–5 trials except for the interneuron shown in B3, which illustrates the responses of the same cell to multiple trains of light stimuli in the absence (black) or presence (red) of TTX. Stimulated axons originated in central medial (CMT; A3, B2, B3) or paraventricular (PVT; A1, A2, B1) thalamic nucleus. All responses were elicited from rest (A1, −67 mV; A2, −70 mV; A3, −72 mV; B1, −74 mV; B2, −66 mV; B3, −69 mV).

DISCUSSION

The basolateral complex of the amygdala (BLA) receives inputs from multiple cortical areas (44) and thalamic nuclei (45). Understanding the impact of these afferents on the BLA requires that we determine what types of information they convey and how they are integrated in the BLA’s circuit. The present study addressed the latter aspect for CMT and PVT inputs to the BL nucleus.

Ultrastructural and Electrophysiological Studies on the Innervation of the BLA

With few exceptions, electron microscopic studies on the articulation of cortical and thalamic afferents with BLA neurons have yielded homogeneous results. In brief, cortical (26–28) and thalamic (24, 25) axon terminals almost always form asymmetric synapses, a majority of which end on dendritic spines (∼70–90%) and a lower proportion on dendritic shafts (∼10–30%). Since asymmetric synapses are generally associated with excitatory inputs (46) and interneurons are largely devoid of dendritic spines whereas principal neurons are densely spiny (47), these results suggest that cortical and thalamic afferents form excitatory synapses, most with principal cells but some with inhibitory interneurons.

However, the lower number of synapses onto interneurons is not evidence that cortical and thalamic afferents exert a weak influence on local-circuit cells. Because principal cells outnumber interneurons and their dendritic fields are extensive (47), it is expected that more cortical and thalamic synapses should contact principal cells than interneurons. In fact, physiological studies that combined optogenetic methods with whole cell recordings reported that thalamic and cortical inputs directly and powerfully excite interneurons, thereby generating feedforward inhibition in principal cells (43, 48–50).

CMT and PVT Contribute Few Synapses onto Inhibitory Interneurons

Contrasting with the homogeneous picture emerging from the above investigations, two ultrastructural studies on thalamic inputs to BL concluded that these afferents end almost exclusively on principal cells. The first examined inputs from another thalamic nucleus [interanteromedial thalamic nucleus (IAM)] and reported that >99% of IAM axon terminals form asymmetric synapses with dendritic spines (51), hence next to no innervation of local-circuit cells. The second took advantage of the fact that, as in cortex (52, 53), calcium/calmodulin-dependent protein kinase II (CaMKIIα) is selectively expressed by principal cells of the BLA (54) to study the articulation of CMT and PVT axons with principal and local-circuit cells in BL (29). None of the CMT axon terminals (of 162) and only one of the PVT terminals (of 126) was found to form a synapse with a CaMKIIα-immunonegative dendrite.

Since this termination pattern contrasts with that seen for other BL afferents, the present study was undertaken to examine the relative contribution of feedforward and feedback inhibition elicited by the activation of CMT and PVT inputs. Overall, our findings are consistent with the conclusions of Amir et al. (29). That is, although a few interneurons were activated at latencies suggestive of monosynaptic excitation, they were generally activated at longer latencies than principal cells. Moreover, a markedly lower proportion of interneurons than principal cells remained responsive to CMT and PVT inputs after blockade of polysynaptic interactions with TTX. This suggests that many interneurons are not activated directly by thalamic inputs but indirectly by principal cells. Yet, the presence of TTX-resistant responses in some interneurons indicates that CMT and PVT inputs do contact a proportion of interneurons directly.

To the best of our knowledge, our study is the first to probe the origin of the inhibition driven by CMT and PVT inputs in BL neurons. Previously, three studies reported on the influence of PVT (and none of CMT) inputs to principal BL neurons using optogenetic methods and whole cell recordings. First, Penzo et al. (9) reported that activation of PVT axons elicits EPSCs in most principal neurons. However, their recordings were performed near the GABA-A reversal potential, such that even if present, inhibition could not have been observed. In a second study, Chen and Bi (6) reported that PVT inputs “evoked very minor excitatory synaptic transmission and little net inhibition in BLA neurons” (p. 4858). Finally, Goedecke et al. (55) examined the modulation of PVT inputs to principal BL neurons by µ-opioid receptors. They observed PVT-evoked EPSCs of comparable amplitude as in the present study, which were markedly reduced by activation of µ-opioid receptors. However, they could not comment on PVT-evoked inhibition because their recordings were performed in the presence of GABA receptor antagonists. Overall, because of their specific focus and associated strategies, these three prior studies do not shed light on the origin of the inhibition driven by PVT and CMT inputs.

Limitations of the Present Study

A limitation of our study stems from the use of spike duration to distinguish interneurons from principal cells. Accordingly, we now consider its potential impact. Based on the expression of neuropeptides and calcium-binding proteins, it was found that BL contains a variety of GABAergic interneurons similar to that found in cortex (56). In rats, McDonald and colleagues distinguished five major interneuron subtypes in BL (57–61). In descending numerical importance, they are 1) PV⁺ cells (43% of GABAergic cells), 2) SOM⁺ cells (18% of GABAergic cells), 3) small neurons expressing CCK plus VIP, 4) large multipolar CCK⁺ cells (7% of GABAergic cells), and 5) neurons immunoreactive for 5-HT-3A receptors (7% of GABAergic cells), only a few of which express other interneuron markers. Calbindin is often coexpressed by three of these five subtypes of interneurons (PV⁺, SOM⁺, large CCK⁺ cells), particularly among PV⁺ (80%) and SOM⁺ (90%) cells. By contrast, a large proportion of VIP⁺ and small CCK⁺ neurons also express calretinin (54, 57, 60).

With the introduction of mice expressing green fluorescent protein (GFP) under the control of particular promoters, it became clear that the electroresponsive properties of interneurons were extremely diverse (reviewed in Ref. 31). Although the dynamics of current-evoked spiking in some of these interneuron subtypes differ strikingly from those of principal cells, others show similar activity patterns. However, one property characterizes a large proportion of interneurons: they generate briefer spikes than principal cells. In a single-cell RT-PCR study, for instance (34), four of five interneuron subtypes generated briefer spikes than principal cells.

Because earlier studies relied on different ways to measure spike durations, we opted to base our cutoff on the distribution of spike durations in our sample (Fig. 2C), where there is a marked change in the incidence of cells with low versus high spike durations. Nevertheless, we cannot exclude the possibility that a few principal cells were misclassified as interneurons and vice versa, particularly CCK cells (33, 36). However, given that CCK cells account for a minute proportion of BL neurons, such misclassifications had to be rare, ∼5% of cases, an error rate that compares favorably with the incidence of false positives in mice expressing GFP under a GAD67 reporter (12–20% of GFP⁺ cortical neurons are immunonegative for GAD67 or GABA; Ref. 62). Although we cannot exclude the possibility that some of the “principal cells” with TTX-resistant responses to CMT or PVT inputs are interneurons, given the low incidence of such misclassifications, they cannot account for the large difference we found between presumed interneurons and principal cells.

What Type of Information Do PVT and CMT Neurons Convey to the BL Nucleus?

It is well established that BL neurons respond to multiple kinds of sensory stimuli, especially if they predict rewarding or aversive outcomes (for instance, see Ref. 63). Moreover, it was recently shown that BL cells concomitantly encode many variables such as the sensory attributes of stimuli, their appetitive or aversive value, contextual information, behaviors like active avoidance, freezing, and reward seeking, movement speed, and even the hierarchical rank of conspecifics (64–68).

Currently, it is unclear what pathways transmit these signals to BL. The potential role of CMT and PVT is difficult to assess because most available studies focused on the expression of immediate-early genes (for instance, see Refs. 10–16). To complicate matters further, PVT, CMT, and BL have many common afferents, like the medial prefrontal cortex, a source of multimodal inputs (1, 69).

Nevertheless, parallels between the variables they encode are compatible with the possibility that PVT and CMT inputs are partly responsible for the complex responsiveness of BL neurons. For instance, CMT receives inputs from the deep layers of the superior colliculus, perhaps supplying BL with visual, somatosensory and auditory information (17, 70). Moreover, CMT is the recipient of parabrachial afferents, potentially explaining why BL neurons are responsive to noxious, visceral, and taste stimuli (18).

Congruently, PVT and BL neurons have similar response profiles. For instance, they both respond to positively or negatively valanced stimuli (reviewed in Ref. 1). Moreover, PVT inhibition interferes with conditioned fear responses (7, 8), perhaps because PVT conveys information about conditioned stimuli to BL. Finally, PVT is the recipient of suprachiasmatic inputs (71), which could explain why the expression of the clock protein Period2 shows a strong circadian rhythm in BL (72).

CMT and PVT: Key Determinants of BL Activity?

Although the cellular composition of the BLA is similar to cortex (47) and the BLA is endowed with an extremely divergent intrinsic connectivity (73, 74), single-unit recording studies in unanesthetized animals have emphasized that principal BLA cells fire at much lower rates (75, 76) than cortical neurons (77, 78). Congruently, spontaneously occurring and cortically evoked EPSPs in vivo are truncated by large-amplitude IPSPs that usually prevent principal BLA cells from spiking (79, 80). These findings, combined with the strong direct control of interneurons by most of BLA’s extrinsic afferents (43, 48–50), indicate that inhibition is a major regulator of BLA activity.

In this context, the contrast between the termination pattern of IAM, CMT, and PVT inputs (29, 51) versus other extrinsic afferents to BL (24–28) is particularly striking. As reviewed above, prior ultrastructural studies indicate that CMT and PVT inputs innervate BL interneurons more weakly than other extrinsic afferents. Therefore, when CMT and PVT cells fire, less feedforward inhibition opposes their excitatory impact on principal BL neurons than seen with other inputs. Based on this logic, if CMT or PVT inputs act on principal BL cells just before or at the same time as other inputs, they could depolarize recipient cells enough to make them fire before principal cells are hyperpolarized by interneurons. However, although these considerations suggest that midline thalamic nuclei may constitute key determinants of BL activity, it remains that CMT and PVT inputs do contact a proportion of interneurons directly, as evidenced in this study by the presence of TTX-resistant responses in some interneurons. Moreover, CMT and PVT axons can also recruit interneurons indirectly, via the activation of principal cells, thereby generating feedback inhibition.

GRANTS

This work was supported by R01 Grant MH-107239 to D.P. from NIMH. N.A, was supported by the Behavioral and Neural Sciences Graduate Program of Rutgers University.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

N.A., D.B.H., and D.P. conceived and designed research; N.A. and D.B.H. performed experiments; N.A., D.B.H., and D.P. analyzed data; N.A. and D.P. interpreted results of experiments; D.P. prepared figures; D.P. drafted manuscript; N.A., D.B.H., and D.P. edited and revised manuscript; N.A., D.B.H., and D.P. approved final version of manuscript.