Synaptic properties of the feedback connections from the thalamic reticular nucleus to the dorsal lateral geniculate nucleus

Abstract

The thalamic reticular nucleus (TRN) is a shell-like structure comprised of GABAergic neurons that surrounds the dorsal thalamus. While playing a key role in modulating thalamocortical interactions, TRN inhibition of thalamic activity is often thought of as having an all-or-none impact. Although TRN neurons have a dynamic firing range, it remains unclear how variable rates of TRN activity gate thalamocortical transmission. To address this, we examined the ultrastructural features and functional synaptic properties of the feedback connections in the mouse thalamus between TRN and the dorsal lateral geniculate nucleus (dLGN), the principal relay of retinal signals to visual cortex. Using electron microscopy to identify TRN input to dLGN, we found that TRN terminals formed synapses with non-GABAergic postsynaptic profiles. Compared with other nonretinal terminals in dLGN, those from TRN were relatively large and tended to contact proximal regions of relay cell dendrites. To evoke TRN activity in dLGN, we adopted an optogenetic approach by expressing ChR2, or a variant (ChIEF) in TRN terminals. Both in vitro and in vivo recordings revealed that repetitive stimulation of TRN terminals led to a frequency-dependent inhibition of dLGN activity, with higher rates of stimulation resulting in increasing levels of membrane hyperpolarization and corresponding decreases in spike firing. This relationship suggests that alterations in TRN activity lead to graded changes in relay cell spike firing.

NEW & NOTEWORTHY The thalamic reticular nucleus (TRN) modulates thalamocortical transmission through inhibition. In mouse, TRN terminals in the dorsal lateral geniculate nucleus (dLGN) form synapses with relay neurons but not interneurons. Stimulation of TRN terminals in dLGN leads to a frequency-dependent form of inhibition, with higher rates of stimulation leading to a greater suppression of spike firing. Thus, TRN inhibition appears more dynamic than previously recognized, having a graded rather than an all-or-none impact on thalamocortical transmission.

INTRODUCTION

The thalamic reticular nucleus (TRN) is an important interface between thalamus and cortex, playing a key role in the attentional modulation of sensory processing, the generation and propagation of thalamocortical rhythms during sleep and wakefulness, and the prevention of hypersynchronous oscillations associated with certain disease states (Crick 1984; Fogerson and Huguenard 2016; Halassa and Acsády 2016; McCormick and Bal 1997; Pinault 2004; Steriade 2005). Comprised entirely of GABAergic neurons, the TRN surrounds much of the dorsal thalamus (Houser et al. 1980; Jones 2007). This nucleus receives excitatory drive from both thalamocortical and corticothalamic axon collaterals and, in return, TRN neurons project to thalamic nuclei to inhibit activity in a state-dependent manner (Fig. 1; Guillery et al. 1998; Guillery and Harting 2003; Halassa and Acsády 2016; Halassa and Kastner 2017; Pinault 2004).

Fig. 1.

Wiring diagram depicting the pattern of connectivity between the dorsal lateral geniculate nucleus (dLGN), retina, thalamic reticular nucleus (TRN), and visual cortex (VC). TRN receives excitatory input (red) from layer VI of visual cortex and axon collaterals of dLGN relay neurons (R). Inhibitory elements and TRN feedback inhibition to dLGN are shown in blue.

TRN-mediated inhibition and its impact on sensory processing are perhaps best understood by examining the connections between TRN and the dorsal lateral geniculate nucleus (dLGN), the thalamic relay of retinal signals to the visual cortex (Cox and Beatty 2017; Guillery and Sherman 2002; Hirsch et al. 2015; Usrey and Alitto 2015). Although the circuit linking TRN to dLGN has been the subject of intense investigation, what remains unclear is how variable rates of TRN activity affect retinogeniculate signal transmission. While many studies have focused on oscillatory activity and burst firing in TRN (Fogerson and Huguenard 2016; McCormick et al. 2015; McCormick and Bal 1997; Pinault 2004; Steriade 2005), the majority of neurons within the visual sector of TRN do not fire in bursts (Huh and Cho 2016; Lee et al. 2007). Moreover, sensory transmission occurs in the awake state, when burst firing is rare, and tonic firing prevails (Bezdudnaya et al. 2006; Ramcharan et al. 2000, 2005; Swadlow and Gusev 2001; Weyand et al. 2001). Indeed, tonic firing rates of TRN neurons vary widely during wakefulness (Chen et al. 2016; Contreras et al. 1993; Domich et al. 1986; Hartings et al. 2003; Lymer et al. 2019; Pinault et al. 2001). Nonetheless, TRN-mediated inhibition is generally thought to operate in a binary manner, selectively gating thalamocortical transmission in an all-or-none fashion (Crick 1984; Guillery and Sherman 2002; Halassa and Acsády 2016; Llinás and Steriade 2006; Pinault 2004). While it is clear that TRN activation can lead to a complete suppression of thalamocortical transmission, the bulk of the evidence supporting this view is restricted to manipulations that involve a single or a limited range of repetitive electrical stimulation (Crandall et al. 2015; Kayama 1985; Kim et al. 1997; Olsen et al. 2012; Shosaku et al. 1984, 1989; Shosaku and Sumitomo 1983; Thomson 1988; Yingling and Skinner 1976; but see Herd et al. 2013). Thus, while different behavioral states alter both TRN and dLGN activity, it is currently unclear how different rates of TRN firing impact retinogeniculate transmission.

The dLGN of the mouse offers a unique opportunity to explore this issue since the circuitry linking TRN to dLGN is highly conserved and it contains both thalamocortical neurons as well as intrinsic interneurons (Guido 2018). Here we addressed how variable rates of TRN activation influence the degree of dLGN activity by utilizing a variety of transgenic mouse models that allow for the visualization and manipulation of TRN input to dLGN. We made use of a transgenic line that expresses green fluorescent protein (GFP) in TRN neurons to delineate TRN synaptic profiles in dLGN (GAD65; Jurgens et al. 2012; López-Bendito et al. 2004). Additionally, we adopted an optogenetic approach by expressing channelrhodopsin 2 (ChR2), or one of its variants (ChIEF), into TRN terminals, to selectively activate postsynaptic inhibitory activity in dLGN neurons. Finally, the nature of TRN inhibitory activity was assessed in vitro by using an acute thalamic slice preparation, as well as in vivo by utilizing an awake, behaving head-fixed mouse preparation.

MATERIALS AND METHODS

Subjects.

Experiments were conducted in adult mice (>21 days postnatal) of either sex. We used C57BL/6J (Jax stock no. 000664, RRID:IMSR_JAX:000664) mice as well as two transgenic strains that express GFP in GABAergic neurons. The GAD67-GFP (line G42; Jax stock no. 007677, RRID:IMSR_JAX:007677) labels dLGN interneurons (Charalambakis et al. 2019; Golding et al. 2014; Jager et al. 2016; Leist et al. 2016; Seabrook et al. 2013a) and a GAD65-GFP strain that labels TRN neurons (Jurgens et al. 2012; López-Bendito et al. 2004). We also crossed somatostatin-Cre (SST-Cre) knock-in mouse (Jax stock no. 013044, RRID:IMSR_JAX:013044) with Ai32 mice (floxed-ChR2 and enhanced yellow fluorescent protein (EYFP); Jax stock no. 012569, RRID:IMSR_JAX:012569) and used the offspring to examine the postsynaptic responses of dLGN neurons evoked by blue light stimulation of TRN terminals (Ahrens et al. 2015; Clemente-Perez et al. 2017; Klug et al. 2018; Wells et al. 2016). Finally, to examine the proportion of SST-expressing neurons in TRN we crossed the SST-Cre mice with the Ai3 (EYFP) reporter line (Jax stock no. 007903, RRID:IMSR_JAX:007903). All breeding and experimental procedures were approved by the University of Louisville Institutional Animal Care and Use Committee.

Electron microscopy.

We used GAD65-GFP mice to study the ultrastructure of TRN terminals in dLGN. Adult mice were deeply anesthetized with Avertin (0.5 mg/g) and perfused transcardially with a fixative solution of 2% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer (PB). The brain was removed from the skull and 70-µm-thick coronal sections were cut using a vibratome (Leica Microsystems, Buffalo Grove, IL). Selected sections that contained GFP were incubated overnight in a 1:1,000 dilution (0.1 μg/mL) of a rabbit anti‐GFP antibody in PB with 1% normal goat serum (Millipore, Billerica, MA, catalog no. AB3080, RRID:AB 91337). The following day, the sections were incubated in a 1:100 dilution of a biotinylated goat-anti-rabbit antibody (Vector Laboratories, Burlingame, CA, 1 h, RRID:AB_2313606), followed by incubation in a solution of avidin and biotinylated horseradish peroxidase (ABC solution, Vector Laboratories, 1 h) and reacted with nickel-enhanced diaminobenzidine (DAB).

Sections that contained terminals labeled with the GFP antibody were postfixed in 2% osmium tetroxide, dehydrated in an ethyl alcohol series, and flat embedded in Durcupan resin between two sheets of Aclar plastic (Ladd Research, Williston, VT). Durcupan-embedded sections were first examined with a light microscope to select areas for electron microscopic analysis. Selected areas were mounted on blocks, ultrathin sections (70–80 nm, silver-gray interference color) were cut using a diamond knife, and sections were collected on Formvar-coated nickel slot grids. Selected sections were stained for the presence of gamma amino butyric acid (GABA) by utilizing a postembedding immunocytochemical protocol described previously (Bickford et al. 2010, 2015). Briefly, we used a 0.25 μg/mL concentration of a rabbit polyclonal antibody against GABA (Sigma‐Aldrich, St. Louis, MO, catalog no. A2052, RRID:AB_477652) and the GABA antibody was tagged with a goat‐anti‐rabbit antibody conjugated to 15‐nm gold particles (BBI Solutions USA, Madison, WI). The sections were air-dried and stained with a 10% solution of uranyl acetate in methanol for 30 min. Ultrathin sections were examined with a transmission electron microscope (Hitachi HT7700, Hitachi High Technologies, Clarksburg, MD) and the synaptic profiles containing labeled terminals were imaged. The diameters of pre- and postsynaptic profiles were measured (Image J, RRID: SCR_003070) and gold particles were counted to calculate the gold particle density overlying each profile.

Immunocytochemistry.

We used a mouse monoclonal antibody against NeuN (Chemicon, catalog no. MAB377, RRID:AB_2298772) on 70-μm-thick coronal sections through the thalamus of SST-Cre mouse crossed with an Ai3 reporter. Selected sections were incubated overnight in a 0.5 μg/mL concentration of the NeuN antibody. The following day, the sections were incubated in a 1:100 dilution of a goat-anti-mouse antibody directly conjugated to Alexa Fluor 546 (Molecular Probes catalog no. A11030, RRID:AB_144695). The sections were then mounted on slides and imaged using a confocal microscope (Olympus FV1200BX61).

AAV Injections.

For the voltage clamp in vitro recording experiments presented in Figs. 4 and 5, an adeno-associated virus (AAV) containing a ChIEF-tdTomato cassette (Jurgens et al. 2012) was injected into the TRN of adult C57BL/6J or GAD67 mice using methods described previously (Bickford et al. 2015; Jurgens et al. 2012; Sokhadze et al. 2019). Adult mice were deeply anesthetized with ketamine/xylazine (8 mg/0.6 mg/mL; 0.01 mL/g), placed into a stereotaxic frame, and 50 nL of AAV-ChIEF-tdTomato was injected into TRN (10 nL/min) using a Hamilton syringe attached to a nanopump. After surgery, animals were closely monitored for 3 days to assess wound healing and given carprofen (1.25 mg/mL; 0.1 mL/25g sc) if they displayed signs of pain or distress. Between 10 and 14 days after injection, mice were euthanized, and brain tissue was harvested for in vitro slice recordings. Mice receiving AAV-ChIEF-tdTomato injections were selected based on the accuracy of the stereotaxic injection and verified by examination of in vitro slices. Our targeted injections did not overlap with the more caudal location of dLGN (Fig. 4A). While in some instances the spread of the injection encroached on the boundary between TRN and VB, the latter does not project to dLGN.

Fig. 4.

Opsin expression in thalamic reticular nucleus (TRN) terminals and light-evoked synaptic responses in dorsal lateral geniculate nucleus (dLGN) neurons recorded in vitro. A: reconstruction of a biocytin filled relay neuron in dLGN containing ChIEF-tdTomato labeling of TRN terminals. B: whole cell voltage clamp recording of the same neuron showing that repetitive blue light stimulation evokes a train of inhibitory postsynaptic currents (IPSCs). C: similar recording showing that bath application of glutamate antagonists [DNQX (20 µM) and CPP (10 mM)] has no effect on light-evoked IPSCs. However, bath application of the GABA_A antagonist SR95531 (20 µM) blocks light-evoked IPSCs. D: reconstruction of a biocytin filled GFP+ interneuron in dLGN containing ChIEF-tdTomato-labeled TRN terminals from a GAD-67 mouse. E: recording from neuron in D showing that photostimulation of TRN terminals fails to evoke any postsynaptic activity. In B, C, and E blue traces below the synaptic responses depict the pattern TRN photostimulation. All recordings are done with cesium-based electrodes and at a holding potential of 0 mV.

Fig. 5.

Optically evoked thalamic reticular nucleus (TRN) inhibitory postsynaptic currents (IPSCs) in dorsal lateral geniculate nucleus relay neurons recorded in vitro. A, left: examples of IPSCs of a relay neuron evoked by stimulating TRN terminals at different temporal frequencies (0.2, 1, 5, 20, and 50 Hz). Right: corresponding responses depicting an expanded time scale of the initial responses shown in left panel. Blue traces below synaptic responses depict the pattern of TRN photostimulation. All recordings are done with cesium-based electrodes and at a holding potential of 0 mV. B: summary plot showing the degree of synaptic suppression of light-evoked TRN IPSCs as a function of stimulus number for each temporal frequency tested. To calculate the percent change in IPSC amplitude, the nth response was divided by the amplitude of the initial response and multiplied by 100. High rates (≥5 Hz) of stimulation lead to a synaptic depression of IPSCs (P < 0.0001). Each point represents the mean and SE from a total of 8 neurons.

In vitro slice preparation and optogenetic stimulation.

Brain slices were made from mice that were deeply anesthetized with isoflurane and rapidly decapitated using methods described previously (Charalambakis et al. 2019; Seabrook et al. 2013b). The brain was rapidly removed and placed into 4°C oxygenated cutting solution (in mM): 234 sucrose, 11 glucose, 26 NaHCO₃, 10 MgSO₄, 2.5 KCl, 0.5 CaCl₂, 1.25 NaH₂PO₄. Thalamic slices 270 µm thick were made using a vibratome (Leica). For 30 min, slices were incubated in warm (32°C) oxygenated artificial cerebrospinal fluid (aCSF) (in mM: 126 NaCl, 26 NaHCO₃, 10 glucose, 2.5 KCl, 2 MgCl₂, 2 CaCl₂, 1.25 NaH₂PO₄). Recordings were conducted in a chamber perfused continuously with 32°C aCSF at a rate of 2–3 mL/min.

Thalamic nuclei were visualized on an upright microscope (Olympus BX51WI) with DIC optics and fluorescent filters (GFP: Chroma 49002; tdTomato: Chroma 49005) using a ×10 or ×60 water immersion objective. A vertical puller (Narashige PC-10) was used to pull patch electrodes from borosilicate glass. For current clamp recordings, the electrode internal solution contained (in mM) 117 K-gluconate, 13 KCl, 1.0 MgCl₂ 1, 0.07 CaCl₂, 0.1 EGTA, 10 HEPES. For voltage clamp recordings, the electrode solution contained (in mM) 117 Cs-gluconate, 11.0 CsCl, 1.0 MgCl₂, 1.0 CaCl₂, 0.1 EGTA, 10.0 HEPES, 2 Na₂-ATP, 0.4 Na₂-GTP. The final tip resistance of all electrodes was 4–7 MΩ.

Whole cell recordings were made in current and voltage clamp using an amplifier (Multiclamp 700B, Molecular Devices), filtered at 3–10 kHz, and digitized (Digidata 1440A) at 20 kHz. Pipette capacitance, series resistance, input resistance, and whole-cell capacitance were monitored throughout the recording session.

The light-gated responses were activated using a light emitting diode (LED, Prizmatix) that delivered blue light through the ×60 objective. Blue light pulses of varying duration were 0.3 mm in diameter with a light power of 525 mW/mm². During some of the recordings, antagonists for α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) [6,7-dinitroquinoxaline-2,3-dione (DNQX), 20 µM, Tocris] and N-methyl-d-aspartate (NMDA) [(±)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid (CPP), 10 mM Tocris] or GABA_A receptors (SR95531, 20 µM Tocris) were bath applied to block postsynaptic activity.

For all recordings, biocytin (0.5%, Sigma) was included in the internal solution for intracellular filling and 3D neuron reconstruction using confocal microscopy (Charalambakis et al. 2019; El-Danaf et al. 2015; Krahe et al. 2011). Following the completion of the recording session, slices were fixed overnight in 4% paraformaldehyde in 0.1 M phosphate buffered saline (PBS), then washed with PBS and incubated overnight with AlexaFluor 647-conjugated streptavidin (Invitrogen, S21374) in a PBS solution containing Triton X-100 (0.1%). Slices were subsequently mounted with Prolong Gold with DAPI (Invitrogen, P36931) and coverslipped. Digital images of biocytin filled neurons were acquired using a multiphoton laser scanning confocal microscope (Olympus FV1200BX61). A HeNe laser (635 nm) was used to excite far-red, biocytin fluorescence. Neurons were imaged using a ×10 (0.3 NA) or ×20 (0.75 NA) objective lens at a scanning resolution of 1,600×1,600 pixels. A sequential series of optical slices with an optimal step size through the Z-axis of 1.26 μm (×20/0.75 NA lens) created a 3D, Z-stacked data set. Three-dimensional neuron reconstructions were compiled, and dendritic architecture was assessed to determine whether the labeled neuron was a relay cell or interneuron (Charalambakis et al. 2019; El-Danaf et al. 2015; Krahe et al. 2011).

In vitro assessment of postsynaptic activity in dLGN.

We measured the inhibitory postsynaptic activity of dLGN neurons evoked by blue light stimulation of TRN terminals. Viral injections of AAV-ChIEF-tdTomato were used for the voltage clamp in vitro experiments presented in Figs. 4 and 5, and SST-Cre × Ai32 (ChR2) mice were used for the current clamp in vitro experiments presented in Figs. 6 and 7. For inhibitory postsynaptic currents (IPSCs), neurons were held at 0 mV using a cesium-based electrode and presented with a train of 20 pulses of blue light (1 ms each) at different temporal frequencies (0.2, 1, 5, 20, and 50 Hz). The amplitude of IPSCs was measured from baseline levels obtained 1 s before photostimulation. To quantify any changes in IPSC activity evoked by repetitive stimulation, the amplitude of the nth response was divided by the amplitude of the initial response and then multiplied by 100 to calculate the percent change (Jurgens et al. 2012). To measure the voltage changes associated with repetitive stimulation of TRN terminals, dLGN neurons were recorded at −55 mV in current clamp using potassium-based electrodes. Stimulus trains consisted of 25 pulses of blue light (1-ms duration) presented at different temporal frequencies (2, 5, 10, 20, and 50 Hz). The mean voltage change was calculated during blue light stimulation and compared with a 1-s baseline period. To examine how TRN stimulation affects spiking activity, dLGN neurons were held at −70 mV and given a 1.5-s square wave depolarizing current pulse (mean ± SE = 175.8 ± 9.4 pA) of sufficient strength to evoke a steady train of spike firing > 5 Hz. Changes in spike firing were then calculated by comparing equivalent periods (0.5 s) of activity in the presence or absence of blue light stimulation (2, 5, 10, 20, and 50 Hz). To minimize the effects of frequency accommodation or burst firing, measures of spike firing were obtained 0.5 s after the onset of the 1.5-s current pulse (Fig. 7A). Typically, the measurements described above were based on the average of four stimulus presentations.

Fig. 6.

Optically evoked thalamic reticular nucleus (TRN) inhibitory postsynaptic potentials (IPSPs) in dorsal lateral geniculate nucleus relay neurons recorded in vitro. A: examples of IPSPs recorded from a relay neuron evoked by stimulation of TRN terminals at different temporal frequencies (2, 5, 10, 20, and 50 Hz). The bottom voltage recording shows that bath application of the GABA_A antagonist SR95531 (20 µM) blocks light-evoked IPSP activity. Blue traces below synaptic responses depict the pattern TRN photostimulation. All responses recorded at a holding potential of −55 mV. B: summary graph plotting the mean voltage response as a function of stimulus frequency for a total of 35 neurons (2 Hz: n = 6, 5 Hz: n = 35, 10 Hz: n = 14, 20 Hz: n = 24, 50 Hz: n = 24). Each point represents a separate response from a different neuron. The mean voltage response was calculated by measuring the average voltage response across the entire train of stimulus presentations (25 pulses). For each frequency tested, the individual data is plotted along with the mean (horizontal line ± SE). High rates of stimulation lead to a greater degree of hyperpolarization (P < 0.01 versus 2 Hz or 5 Hz; one-way ANOVA with Tukey’s multiple comparison test).

Fig. 7.

Thalamic reticular nucleus (TRN) stimulation suppresses spike firing of dorsal lateral geniculate nucleus (dLGN) relay neurons recorded in vitro. A: examples of voltages responses from a dLGN relay neuron showing a train of spikes evoked by current pulse injection (black). Responses in the left column show spikes in the absence of TRN stimulation, while those on the right depict the suppression of spike firing brought about by TRN stimulation (blue) at different temporal rates (2, 5, 10, 20, and 50 Hz). B: summary plot showing the changes in spike firing brought about by TRN stimulation. The firing rate is plotted as a percent control and based on matching periods (0.5s, black horizontal line in A) of dLGN activity. For each cell (n = 29; 2 Hz: 7, 5 Hz: 10, 10 Hz: 23, 20 Hz: 23, 50 Hz: 28), the percent of control firing frequency is plotted along with the group mean (horizontal line ± SE). C: similar plot for 7 neurons in which all five frequencies were tested. Each color represents a different neuron. The between- (B, compared with 2 Hz, *P < 0.05, ***P < 0.001, Kruskal–Wallis test) and within-cell analyses (C, compared with control, **P < 0.01, ***P < 0.001, Friedman test) indicate that high rates of stimulation lead to a greater suppression of spike firing.

In vivo recordings in awake head-fixed mice and optogenetic stimulation.

SST-Cre × Ai32 (ChR2) mice were prepared for in vivo recordings using methods similar to Tschetter et al. (2018). Mice were anesthetized with an intraperitoneal injection of ketamine (150 mg/kg) and xylazine (15 mg/kg), and if needed, supplemented with 25% of the initial dose. The mice were then fitted with a custom head plate that was used to stabilize the head in a stereotaxic device, and a craniotomy was made to permit subsequent electrode penetration through dLGN. The craniotomy was sealed with silicone (Kwik-Cast) and mice were allowed to recover. After surgery, mice were given an analgesic (meloxicam, 5 mg/kg) and monitored for 48 h for any signs of discomfort or distress.

Two to four days after the initial surgery, mice were secured to a head-fixed frame and acclimated to run on a Styrofoam ball. The silicone plug was removed, and the cortical surface was covered in a thin layer of agarose/aCSF to prevent drying. An optrode consisting of an insulated tungsten microelectrode (2–4 MΩ) glued to a fiber optic cable (100-µm diameter) was lowered into the brain using coordinates that target dLGN (bregma −2.25 AP, 2.25 ML). At the end of a recording session (45–90 min), the mice were deeply anesthetized and perfused transcardially with a fixative solution, and the was brain excised for electrode penetration reconstruction. The fiber optic was connected to a 473-nm blue-light laser (Laserglow). The light intensity was 20 mW/mm², and the duration and frequency of blue light pulses were under computer control.

During a recording session, multiunit activity was acquired using an extracellular amplifier (TDT BioAmp). The signal was amplified by ×2,000, bandpass filtered at 300–2,000 Hz, and sampled at 20 kHz. Single units were identified by spike waveform analysis using conventional protocols developed by Spike2 software. Units were determined to be in dLGN based on their dorsal-ventral coordinates and electrode tract reconstruction.

In vivo assessment of dLGN unit activity.

To examine how TRN stimulation affects dLGN spiking activity, units that were spontaneously active and stable were stimulated with blue light pulses (5 ms) presented at different temporal frequencies (5, 20, and 50 Hz). Changes in unit activity were then calculated by comparing the activity during equivalent periods just before (baseline) and during blue light stimulation (5 Hz: 2 s; 20 and 50 Hz: 0.5 s). Baseline unit activity was not significantly different between the three temporal stimulus frequencies used (mean ± SE = 30.6 ± 4.4 Hz, one-way ANOVA, F_2,52 = 2.588, P = 0.0849). Typically, the measurements described above were based on the average of five stimulus presentations.

Statistical analyses.

All statistical tests were performed using Prism 8.0 (GraphPad Software, La Jolla, CA). Parametric statistical tests were used to analyze gold particle density, IPSC amplitude, and mean voltage. We used both Tukey’s (one-way ANOVA) or Dunnett’s (two-way ANOVA) methods for correcting for multiple comparisons during post hoc tests. Nonparametric tests (Kruskal–Wallis and Friedman) were used to compare terminal size and percent changes in spike firing. Post hoc tests for these analyses employed a Dunn’s correction for multiple comparisons.

RESULTS

TRN terminals in the dLGN.

To examine the projections from TRN to dLGN, we utilized GAD65-GFP mice (Jurgens et al. 2012; López-Bendito et al. 2004). The GAD65-GFP mouse line exhibits robust labeling of TRN neurons (Fig. 2A) and their terminals in the dLGN (Fig. 2B). There is no evidence of somatic labeling in the dLGN (Fig. 2B) or pretectum (not shown), indicating that in this transgenic strain, the GFP terminal labeling in dLGN arises from the TRN.

Fig. 2.

Feedback projections from thalamic reticular nucleus (TRN) to dorsal lateral geniculate nucleus (dLGN). A and B: coronal sections through dorsal thalamus of the GAD65-GFP mouse illustrating green fluorescent protein (GFP)+ neurons (green) in TRN (A) and their projections to dLGN (B). Note the presence GFP+ neurons in ventral lateral geniculate nucleus (vLGN) but absence in dLGN. Scale bar, 100 µm. C: electron microscopic image of a TRN terminal (green overlay) in the dLGN labeled with an antibody against GFP (dark reaction product) and an antibody against GABA (high density of gold particles). The terminal synapses (arrows) on a non-GABAergic dendrite (low density of gold particles). Scale bar, 600 nm. D: a cumulative distribution showing the diameter of different presynaptic terminals in dLGN. E: a cumulative distribution showing the diameter of different postsynaptic profiles. TRN terminals and postsynaptic profiles in green (n = 72), all others measured from a previous study (Bickford et al. 2015; red: retinogeniculate (RLP, n = 179), black: tectogeniculate (TG, n = 108), blue: corticogeniculate (CG, n = 87).

To examine the nature of TRN contacts within the dLGN, we prepared GAD65-GFP tissue for electron microscopy. Figure 2C illustrates a TRN terminal (green), labeled with a GFP antibody (DAB reaction) as well as a GABA antibody (gold particles). This terminal forms a synapse (arrows) with a non-GABAergic (presumed relay neuron) dendrite (blue). We sampled 10 ultrathin sections through the dLGN of an adult GAD65-GFP mouse and detected a total of 72 TRN terminals that were labeled with DAB and numerous gold particles (≥40 particles per µm²; mean density ± SE: 105.9 ± 3.60). All sampled TRN terminals formed synapses with non-GABAergic postsynaptic profiles (mean density of overlying gold particles ± SE: 4.42 ± 0.36 particles/µm²). The profiles postsynaptic to TRN terminals were large dendrites or somata (range: 0.18–12.35 µm²; mean: 1.56 ± 0.31 µm²). Thus, GABAergic TRN terminals formed synapses with the somata and proximal dendrites of non-GABAergic relay neurons.

To compare the size and dendritic location of TRN terminals relative to other presynaptic terminals in the dLGN, we used data from a previous study (Bickford et al. 2015) in which we quantified the size of retinogeniculate (RG, n = 179), corticogeniculate (CG, n = 87), and tectogeniculate terminals (TG, n = 108) along with the size of their postsynaptic profiles. Figure 2D shows that TRN terminals (n = 72, green) were significantly larger than (blue) and TG (black) terminals (Kruskal–Wallis, CG: P < 0.0001; TG: P = 0.0009), but smaller than RG terminals (red; P < 0.0001). Figure 2E demonstrates that the diameter of profiles postsynaptic to TRN terminals (n = 72) was similar to the diameter of profiles postsynaptic to RG (n = 179) and TG (n = 108) terminals (Kruskal–Wallis, RG: P = 0.6902; TG P = 0.1818) and larger than profiles postsynaptic to CG terminals (P = 0.0003). Thus, TRN terminals contact large diameter postsynaptic elements, suggesting they innervate more proximal regions of dLGN relay neurons (Bickford et al. 2010, 2015).

Expression of opsins in TRN neurons.

To examine the nature of postsynaptic activity evoked by the TRN, we adopted an optogenetic approach. First, for the voltage clamp in vitro experiments presented in Figs. 4 and 5, we injected an AAV viral vector into the TRN (Fig. 3A) to express the channelrhodopsin variant CHIEF-tdTomato in TRN neurons (Bickford et al. 2015; Jurgens et al. 2012). In a second approach, we crossed somatostatin-Cre mice with the Ai32 line to target ChR2-EYFP in Cre-expressing neurons within TRN (Fig. 3D; Ahrens et al. 2015; Clemente-Perez et al. 2017; Klug et al. 2018; Wells et al. 2016). SST-Cre × Ai32 (ChR2) mice were used for the current clamp in vitro experiments presented in Figs. 6 and 7 and extracellular unit in vivo recordings in Figs. 8 and 9. To confirm that both approaches could activate TRN neurons, we conducted whole cell recordings using biocytin-filled pipettes in the TRN. A representative TRN neuron is reconstructed from a ChIEF-injected (Fig. 3B) and SST-Cre × ChR2 mouse (Fig. 3E). In ChIEF-injected animals, long single pulses (500 ms) of blue light evoked a sustained depolarization that led to spike firing (Fig. 3C). Furthermore, for each opsin, repetitive stimulation with brief pulses (1 ms) were sufficient to evoke a train of depolarizations with spikes riding the peaks (Fig. 3C: 10 Hz; Fig. 3F: 20 and 50 Hz). Figure 3C illustrates that this light-evoked activity was unaffected by the bath application of ionotropic glutamate blockers (20 µM DNQX and 10 mM CPP). Thus, when activated both opsins reliably evoked spike firing in TRN neurons. While these two opsins have slightly different kinetics (Lin 2011) their responses to brief pulses of blue light were similar.

Fig. 3.

Opsin expression and blue light activation of thalamic reticular nucleus (TRN) neurons. A: expression of tdTomato (red) in a coronal section through TRN, 8 days after injection of an adeno-associated virus (AAV) containing a ChIEF-tdTomato cassette. B: confocal reconstruction of biocytin filled TRN neuron (white) in a tdTomato-labeled section. C: whole cell in vitro recordings of the same neuron showing the voltage responses evoked by blue light stimulation. Left: a single 500-ms pulse of blue light produces a prolonged depolarization and train of action potentials. Right top: a 10-Hz (1-ms, 20-pulse) train of blue light generates a succession of unitary depolarizations and spike firing. Right bottom: light-evoked activity persists during bath application of glutamate antagonists (DNQX 20 µM and CPP 10 mM). D: expression of EYFP in TRN of a somatostatin-Cre × Ai32 (ChR2-EYFP) mouse. E: confocal reconstruction of biocytin filled TRN neuron in ChR2-EYFP mouse. F: corresponding recordings showing that high rates of stimulation (left 20 Hz; right 50 Hz) evokes reliable voltage responses and spike firing. In C and F, blue traces below the voltage responses depict the pattern of TRN photostimulation. G: expression of EYFP (green) in a coronal section through TRN of an SST-Cre × Ai3 mouse. H: high power view showing SST-expressing neurons in TRN. I: same section labeled with NeuN (magenta). J: merged image showing double-labeled neurons. Scale bars 200 µm and 10 µm.

Fig. 8.

Photostimulation of thalamic reticular nucleus (TRN) terminals in dorsal lateral geniculate nucleus (dLGN) during in vivo recordings. A: coronal section through dLGN of an SST-Cre × Ai32 mouse used during in vivo recording, showing the expression of EYFP (green) in TRN terminals. B: extracellular unit recordings in dLGN of an awake head-fixed mouse (black trace) from which single units can be isolated (3 cells, raster plot above). Unit activity is diminished during photostimulation of TRN terminals. Blue traces below unit activity depicts the pattern of blue light stimulation (left 5 Hz; right 50 Hz). Inset shows an expanded view of activity just before stimulation.

Fig. 9.

Thalamic reticular nucleus (TRN) stimulation suppresses lateral geniculate nucleus unit activity in vivo. A: raster plots and corresponding unit recordings for 3 dorsal lateral geniculate nucleus (dLGN) neurons showing the degree of suppression associated with stimulating TRN terminals at different rates (5, 20, and 50 Hz). Blue traces below unit activity depicts the pattern of blue light stimulation. B: summary plot for 37 neurons showing the changes in spike firing brought about by TRN stimulation. The firing rate is plotted as a percent control and based on matching periods of control, pre-stimulation activity. Each point represents a neuron and the percent of control firing frequency is plotted along with the group mean (horizontal line ± SE; 5 Hz: n = 11, 20 Hz n = 13, 50 Hz: n = 30). C: similar plot for 11 neurons in which 3 rates were tested (5, 20, and 50 Hz). Each color represents a different neuron. The between- (B, compared with 5 Hz, *P < 0.05, Kruskal–Wallis test), and within-cell analyses (C, compared with control, ***P < 0.001, Friedman test) indicate that high rates of stimulation lead to a greater suppression of spike firing.

To ensure that SST-Cre could be used effectively to examine light-evoked inhibition in dLGN, we obtained estimates of the number of SST expressing neurons in the visual sector of TRN by crossing SST-Cre with an Ai3 (EYFP) reporter line, and then using a NeuN antibody to determine the percentage of doubled-labeled neurons (Fig. 3, G–J). From three sections through visual TRN we found on average, 81.44 ± 5.04% of vTRN neurons were double labeled (n = 373 SS+NeuN, n = 85 NeuN). Thus, these results are consistent with other reports indicating that the vast majority of neurons in the visual sector of TRN express SST (Clemente-Perez et al. 2017).

Optogenetic activation of TRN terminals in dLGN.

Visualized, in vitro whole cell recordings were obtained from 60 relay neurons and 7 interneurons of dLGN. Relay neurons were identified by their large round somata, voltage responses to brief injections of current, and their dendritic architecture based on biocytin-filled reconstructions (n = 40) (Krahe et al. 2011). Interneurons were targeted using the GAD67-GFP transgenic mouse, which labels intrinsic interneurons (Charalambakis et al. 2019; Golding et al. 2014; Jager et al. 2016; Leist et al. 2016; Seabrook et al. 2013a). These had fusiform-shaped somata labeled with GFP and highly expansive dendritic trees that emanated from opposite poles of their cell bodies (Charalambakis et al. 2019; Seabrook et al. 2013a). To investigate TRN-mediated inhibition of dLGN neurons, we conducted voltage clamp recordings using a cesium-based internal solution while holding the neuron at 0 mV (Charalambakis et al. 2019; Govindaiah and Cox 2004, 2006). As shown in Fig. 4B, repetitive blue light stimulation of TRN terminals evoked large IPSCs in relay neurons (n = 8), and when temporal rates ≥5 Hz were used, the amplitude of IPSCs began to attenuate after the initial pulse (i.e., synaptic depression). These IPSCs were unaffected by bath application of glutamate blockers (20 µM DNQX and 10 mM CPP) but were inhibited by the GABA_A antagonist SR95531 (20 µM Fig. 4C). By contrast, similar patterns of stimulation failed to evoke any postsynaptic activity in interneurons (n = 7, Fig. 4E). Together, these data demonstrate that TRN terminal activation inhibits relay neurons but not intrinsic interneurons of dLGN.

To quantify the degree of synaptic depression noted among relay neurons, we presented trains of blue light pulses (1 ms) at different temporal frequencies (0.2, 1, 5, 20, and 50 Hz). An example of light-evoked responses to repetitive stimulation is shown in Fig. 5A. At low temporal frequencies (0.2 and 1 Hz) TRN terminal activation triggered large and relatively stable IPSCs for each pulse of the stimulus train. However, higher rates of stimulation (≥5 Hz) led to a form of synaptic depression, with the initial response evoking the maximal amplitude and subsequent responses showing a progressive decrease in amplitude that stabilized after the 5th pulse. A summary of these interactions is shown in Fig. 5B (n = 8). The initial response to all stimulus trains produced large IPSC activity (n = 40 IPSCs, mean ± SE: 736.6 ± 62.0 pA). At low rates of stimulation (0.2 and 1 Hz) little if any change in IPSC amplitude occurred throughout the entire stimulus train. However, frequencies of 5, 20, and 50 Hz showed a progressive decline in IPSC amplitude between the 2nd to 5th pulse (two-way repeated-measures ANOVA, F_19,285 = 98.67, P < 0.0001). These changes stabilized by the 5th pulse whereby the overall magnitude of inhibition was on average 47–56% of the maximal response evoked by the initial pulse. Thus, TRN terminal activation led to large amplitude IPSC activity, which began to attenuate in response to repetitive stimulation at temporal frequencies ≥5 Hz.

To further explore the effects of repetitive TRN stimulation on dLGN relay neuron activity we recorded synaptic responses in current clamp mode using potassium-based electrodes and measured the mean change in the membrane potential. An example of TRN-mediated inhibitory postsynaptic potentials (IPSPs) is shown in Fig. 6A. Blue light stimulation evoked large IPSPs that diminished in amplitude when repeated stimuli were presented at rates ≥5 Hz. As expected, these responses were blocked during the bath application of the GABA_A antagonist, SR95531 (20 µM). Nonetheless, high rates of stimulation (10, 20, and 50 Hz) still resulted in larger and more sustained levels of hyperpolarization compared with lower rates of stimulation (2 and 5 Hz). Figure 6B summarizes these interactions for 35 dLGN relay neurons. For each neuron the mean hyperpolarizing voltage response evoked by a stimulus train is plotted as a function of the temporal frequency. The initial amplitude of light-evoked IPSPs were not significantly different across each of the temporal frequencies tested (one-way ANOVA, F_4,111 = 2.31, P = 0.062). More importantly, we found that a progressive increase in temporal frequency led to significantly higher levels of membrane hyperpolarization (one-way ANOVA F_{4, 71} = 19.50, P < 0.0001). Stimulation at 20 and 50 Hz led to larger changes than 2 or 5 Hz (mean ± SE = 2 Hz: 0.52 ± 0.15 mV; 5 Hz: 1.09 ± 0.09; 20 Hz: 2.39 ± 0.24 mV; 50 Hz: 4.03 ± 0.39 mV; P < 0.01).

Overall, these data suggest that, although the TRN-mediated inhibition exhibits a form of synaptic depression at high rates of stimulation (Figs. 5 and 6), it remains sufficient to strongly hyperpolarize dLGN relay neurons. Moreover, the overall impact of TRN inhibition is far greater at higher stimulation rates (Fig. 6).

Next, we explored how different rates of TRN stimulation affect the spike activity of dLGN neurons. To address this, we conducted in vitro dLGN recordings in current clamp mode and stimulated TRN terminals at different rates while injecting a depolarizing current pulse that triggered a steady train of spike firing (mean ± SE 17.5 Hz ± 0.72 Hz). An example is shown in Fig. 7A in which a square wave current pulse evoked tonic firing before (control, left) and during (right) blue light TRN terminal stimulation. While both low and high rates of TRN stimulation led to decreased firing, stimulation rates of 20 and 50 Hz often led to a complete suppression of spike firing (in 9/29 neurons). This pattern is summarized in Fig. 7B for 29 relay neurons which plots firing frequency as a percent change of control recordings. TRN stimulation at all temporal frequencies led to a significant decline in spike firing (two-way repeated-measures ANOVA F_1,86 = 311.8, P < 0.0001; versus control firing: 2 Hz: P = 0.0126, all other rates: P < 0.0001). Moreover, an increase in temporal frequency stimulation led to a progressive decline in spike firing (Kruskal–Wallis test, P < 0.0001; versus 2 Hz: 10 Hz: P = 0.0353, 20 Hz: P = 0.0003, 50 Hz: P < 0.0001; versus 5 Hz: 20 Hz: P = 0.0205, 50 Hz: P = 0.0012), so that stimulation rates of 20 and 50 Hz produced firing levels that were ≤12% of control values.

The relationship between dLGN spike firing and the frequency of TRN stimulation was also examined for 7 neurons that were stimulated at 2, 5, 10, 20, and 50 Hz (Fig. 7C). Again, TRN stimulation significantly decreased spike firing (Friedman test, P < 0.0001). Within each of the 7 neurons, compared with control, only rates of 10, 20, and 50 Hz significantly reduced cell activity (2 Hz: P > 0.999, 5 Hz: P = 0.4819, 10 Hz: P = 0.0091; 20 Hz: P = 0.0010; 50 Hz: P = 0.0001). Together, these data show that TRN terminal stimulation decreases the firing rate of relay neurons in a frequency-dependent manner, with rates of ≥10 Hz having the greatest impact on spike firing.

Finally, we tested whether the frequency-dependent effects observed in vitro also occur in vivo. To accomplish this, we adopted an optogenetic approach to stimulate TRN terminals and record unit activity in dLGN of an awake, head-fixed SST-Cre × ChR2 mouse (Fig. 8A). By targeting dLGN neurons with a custom-designed optrode consisting of a tungsten electrode attached to a fiber optic cable, we could stimulate TRN terminals in dLGN while simultaneously recording unit activity. Under these recording conditions, we found TRN neurons (n = 37) had a baseline spontaneous firing rate that ranged between 4 and 64 Hz. These values were similar to those reported by others, with tonic rates ranging between 5 and 50 Hz, and varying as a function of behavioral state or task relevance (e.g., 5–20 Hz Halassa et al. 2014; 20 and 50 Hz Wimmer et al. 2015; <10 Hz Lewis et al. 2015; <50 Hz Chen et al. 2016; <15 Hz Schmitt et al. 2017; <50 Hz Nakajima et al. 2019).

We recorded from a total of 37 dLGN neurons that responded to TRN stimulation. An example of a dLGN recording in which 3 neurons were isolated, before, during, and after blue light stimulation of nearby TRN terminals is shown in Fig. 8B. The raster plots for the 3 isolated neurons are shown above the multiunit recording. Repetitive blue light stimulation at 5 Hz diminished activity while stimulation at 50 Hz completely suppressed spike firing.

To further confirm this frequency-dependent suppression, we examined the extent to which 3 different rates of stimulation (5 Hz, 20 Hz, and 50 Hz) affect the spontaneous activity of dLGN neurons. Figure 9A provides examples of the spontaneous activity for 3 dLGN neurons recorded before, during and after repetitive stimulation of TRN terminals at 5 Hz, 20 Hz, and 50 Hz. Indeed, an increase in temporal frequency stimulation led to a progressive decrease in unit activity. These interactions are summarized in Fig. 9B for 37 dLGN neurons. For each neuron, the effects of TRN stimulation are expressed as a percent change in firing compared with baseline activity. Similar to in vitro experiments, in vivo TRN stimulation reduced dLGN activity in a frequency-dependent manner (Kruskal–Wallis test, P = 0.0400). Stimulation rates of 50 Hz reduced activity by twice as much as 5 Hz (median percent control, 5 Hz: 29.4%, 50 Hz: 12.6%; P = 0.0416). Furthermore, spike activity was completely abolished in 10/34 (29%) neurons stimulated at either 20 or 50 Hz. A within-cell comparison further illustrates these interactions (Fig. 9C). For 11 neurons that were stimulated with all three frequencies (5, 20, and 50 Hz) their spontaneous activity was significantly decreased by 20 and 50 Hz, but not by 5 Hz (Friedman test, P < 0.0001; versus control: 5 Hz: P = 0.1908, 20 Hz: P = 0.0004, 50 Hz: P = 0.0002). Together, the data show that dLGN cells are suppressed by TRN in a frequency-dependent manner and that activating TRN at higher temporal frequencies can reduce dLGN firing to levels that are <20% of baseline.

DISCUSSION

Our results indicate that the photostimulation of TRN terminals in dLGN led to a frequency-dependent inhibition of dLGN activity, with higher rates of stimulation resulting in increasing levels of inhibition and suppression of spike firing. Moreover, TRN-mediated inhibition of dLGN was restricted to relay neurons. We found no evidence, at the ultrastructural or functional level, of TRN input onto intrinsic interneurons. Such an arrangement allows for a powerful and efficient suppression of thalamic activity without engaging a disinhibition through intrinsic interneurons.

The absence of TRN input onto dLGN interneurons may be a unique feature of rodent TRN feedback connections, but additional studies in higher mammals (e.g., cat and galago) lend support to the view that overall, thalamocortical relay neurons of first order, sensory nuclei are the primary recipient of TRN input (Cucchiaro et al. 1991; Liu et al. 1995; Uhlrich et al. 2003; Wang et al. 2001). By contrast, connections between TRN and interneurons seem more prevalent among nonsensory (e.g., anterior and motor) thalamic nuclei (Ilinsky et al. 1999; Kultas-Ilinsky et al. 1995; Tai et al. 1995). Such differences suggest that the TRN is organized into subnetworks that are defined by their projection patterns onto specific cell types belonging to different thalamic nuclei (Halassa and Acsády 2016).

As expected, TRN stimulation evoked large inhibitory currents in dLGN relay neurons (Cox et al. 1997b; Crandall et al. 2015; Herd et al. 2013), a property that is consistent with their ultrastructural synaptic profile. For example, we found TRN terminals to be relatively large, making contacts primarily on proximal dendritic regions of relay neurons. Moreover, repetitive stimulation, especially at high rates, led to a form of synaptic depression with the initial response being the largest and subsequent ones showing a ~50% decline in amplitude by the 5th pulse in the train. It is important to note that light-evoked responses, especially when triggered by high rates of stimulation (>25 Hz), can lead to an artificial form of synaptic depression. This is especially true for ChR2, since this channel exhibits desensitization during prolonged activation (Lin et al. 2009), as well as an elevation in the initial probability of transmitter release when presynaptic boutons are directly activated (Schoenenberger et al. 2011). Some of these issues can be overcome by using ChR2 variants, optimizing expression strategies, or by varying pulse duration and intensity (Jackman et al. 2014; Lin 2011). Admittedly, we did not conduct the full arsenal of controls to determine the extent to which our choice of opsins distorted the degree of short-term synaptic depression noted in dLGN relay neurons. However, we showed that transgenic expression of ChR2 in TRN neurons gave rise to faithful and reliable spike firing even at 50 Hz. This suggests that our stimulation protocols did not produce substantial desensitization. Moreover, if opsin channel desensitization did have a significant impact on synaptic transmission, we would expect to see a frequency-dependent decrease in TRN inhibition, but instead we observed an overall increase in inhibition when high rates of stimulation were used. Alternatively, one can utilize electrical rather than optical forms of stimulation to further resolve this issue. Unfortunately, in thalamic slices that contain dLGN, it is not possible to preserve the projections from TRN to dLGN; hence our choice to use optogenetics to interrogate this synapse. However, in slices that preserve the connections between TRN and adjacent ventrobasal complex (VB), electrical stimulation of TRN input leads to a similar form of synaptic depression as noted here (Bessaïh et al. 2006; Cox et al. 1997a, 1997b; Crandall et al. 2015; Evrard and Ropert 2009; Herd et al. 2013; Ulrich and Huguenard 1996). Indeed, a similar form of synaptic depression was also observed when using optogenetics to stimulate TRN synapses in VB (Ahrens et al. 2015). Together, these observations suggest that TRN projections, at least those to first order thalamic nuclei, show a limited degree of convergence (Pinault 2004; Pinault et al. 1995; Pinault and Deschênes 1998) and a high probability of transmitter release (Zucker and Regehr 2002). Finally, the underlying pharmacology of these inhibitory responses suggest they are mediated by GABA_A receptor transmission. However, the involvement of GABA_B cannot be ruled out since others have shown such inhibition in other species (Kim et al. 1997) or between TRN and VB (Huguenard and Prince 1994; Warren et al. 1997).

It is important to note that even though high rates of repetitive stimulation led to synaptic depression, it also produced a greater degree of hyperpolarization. For example, 20- to 50-Hz rates of stimulation led to a four- to eightfold increase in membrane hyperpolarization when compared with 2–5 Hz. While these observations are consistent with the reported decay tau of TRN responses in thalamus (~55 ms), other factors, such as the activity-dependent recruitment of extrasynaptic GABA receptors, could further potentiate postsynaptic inhibition during high rates of stimulation (Herd et al. 2013).

These frequency-dependent effects on membrane hyperpolarization also had a corresponding impact on dLGN spike firing. For neurons recorded in vitro, increasing stimulus frequency between 2 and 50 Hz resulted in a progressive decrease in spike firing that ranged from 10 to 70% of control values.

Although more variable, a similar trend was observed during in vivo recordings from awake head-fixed mice, suggesting that even modest changes in TRN activity can have a dynamic impact on retinogeniculate signal transmission. It is important to note that not all neurons showed such graded effects, for many (9/29 in vitro; 10/34 in vivo) higher rates of stimulation (≥20 Hz) often led to a complete suppression of activity. Thus, while TRN stimulation at rates of ≥20 Hz can sometimes lead to a total suppression of dLGN activity, lower rates of TRN stimulation simply diminish relay neuron firing. This inflection point is intriguing and seems to correspond to the state-dependent changes in TRN activity that occur during sleep-wake transitions (Steriade et al. 1986) or from switching attentional states from one sensory modality to another (Chen et al. 2016; Wimmer et al. 2015). Overall, these data support the view that TRN inhibition allows for a more graded control of thalamic output, one that operates on a variable temporal scale to modulate thalamocortical transmission in a state-dependent manner (Halassa and Acsády 2016; Halassa and Kastner 2017).

GRANTS

This work was supported by EY12716 (WG), EY024173 (MB, WG), NS104807 (MB, WG), and EY026792 (PC).

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

P.W.C., G.G., S.P.M., M.E.B., and W.G. conceived and designed research; P.W.C., G.G., S.P.M., M.E.B., and W.G. performed experiments; P.W.C., G.G., S.P.M., M.E.B., and W.G. analyzed data; P.W.C., G.G., S.P.M., M.E.B., and W.G. interpreted results of experiments; P.W.C., G.G., S.P.M., M.E.B., and W.G. prepared figures; P.W.C., G.G., M.E.B., and W.G. drafted manuscript; P.W.C., G.G., M.E.B., and W.G. edited and revised manuscript; P.W.C., G.G., S.P.M., M.E.B., and W.G. approved final version of manuscript.