The HDC-Cre line was generated by homologous recombination with an ires-Cre cassette inserted into exon 12 of the hdc gene, between the stop (TAG) codon and the polyadenylation (pA) signal (Zecharia et al., 2012). HDC-Cre mice were crossed with Rosa26-loxP-Stop-loxP-YFP mice (Srinivas et al., 2001) or a floxed γ2 mouse strain (γ2I77lox), separately. The γ2I77lox line was generated with the codon of phenylalanine (F) at position 77 mutated to isoleucine (I) in exon 4 of the γ2 subunit gene, and two loxP sites inserted in intron 3 and intron 4, respectively (Wulff et al., 2007). The F77I mutation resulted in the loss of zolpidem sensitivity from all cells tested (Cope et al., 2004, 2005). Importantly, the physiological properties of the GABA_ARs were unchanged in the F77I strain and there was no behavioral phenotype associated with this silent mutation (Cope et al., 2004, 2005). This line has been used to delete IPSCs from a number of cell types (Wulff et al., 2007, 2009a,b; Zecharia et al., 2012; Kerti-Szigeti et al., 2014; Rovó et al., 2014). To generate HDC-Δγ2 mice and littermate controls, homozygous γ2I77lox/lox mice were crossed with heterozygous γ2I77lox/+/HDC-Cre mice. The γ2I77lox mouse line was genotyped by PCR with the following primers: forward: 5′-GTCATGCTAAATATCCTACAGTGG-3′; reverse: 5′-GGATAGTGCATCA-G¬CAGACAATAG-3′ (213 bp wild-type; 250 bp floxed allele) and the HDC-Cre mouse line was genotyped using: forward: 5′-GTGTGGCTGCCCCTTCTG-CC-3′; reverse: 5′-AGCCTCACCATGGCCCCAGT-3′ (250bp).

For immunohistochemical localization, mice were deeply anesthetised with sodium pentobartital (in accordance with UK Home Office guidelines) and transcardially perfused with 4% paraformaldehyde (Thermo scientific) in phosphate buffered saline (PBS; Sigma). Coronal slices were cut at a thickness of 30 μm (Leica VT1000S vibratome) and incubated in rabbit anti-GFP (1:1000; molecular probes) and mouse anti-NeuN (1:300; Millipore) antibodies overnight. Slices were then incubated for 2 h at room temperature with Alexa Fluor 488 goat anti-rabbit (1:1000; Life technologies) and Alexa Fluor 594 goat anti-mouse IgGs (1:1000; Life technologies). Slices were then mounted in Vectashield mounting medium with DAPI (H1200, Vector labs) and the resulting red, green and blue signals were imaged on a Zeiss LSM 510 CLSM microscope (Facility for Imaging by Light Microscopy, FILM, Imperial College).

For electrophysiological recording, mice were routinely handled to reduce stress levels and brain slices were then prepared from adult (3–6 months postnatal) mice that were killed by cervical dislocation (in accordance with UK Home Office guidelines). The slicing solution contained (in mM) the following: NaCl 125, KCl 2.5, CaCl₂ 2, MgCl₂ 4, NaH₂PO₄ 1.25, NaHCO₃ 26, glucose 11, 1 kynurenic acid, pH 7.4, when bubbled with 95% O₂/5% CO₂. Slices were cut using a vibratome tissue slicer (Campden instruments) at a thickness of 250 μm and immediately transferred to a holding chamber containing slicing solution continuously bubbled with 95% O₂/5% CO₂. Once slicing was complete, slices were then transferred to a 37 °C heat block for 10 min, after which the slicing solution was exchanged for recording ACSF (in mM: NaCl 125; KCl 2.5; CaCl₂ 2; MgCl₂ 1; NaH₂PO₄ 1.25; NaHCO₃ 26; and glucose 11, pH 7.4, when bubbled with 95% O₂/5% CO₂). The slices were subsequently incubated in the recording ACSF at room temperature for at least another 30 min before electrophysiological recordings.

Slices were visualized using a fixed-stage upright microscope (BX51W1, Olympus) fitted with a high numerical aperture water-immersion objective and a digital camera. Patch pipettes were fabricated from thick-walled borosilicate glass capillaries (1.5 mm o.d., 0.86 mm i.d., Harvard Apparatus) using a two-step vertical puller (Narishige, PC-10). Pipette resistances were typically 3–4 MΩ when back filled with internal solution. The internal solution contained (in mM) CsCl 140, NaCl 4, CaCl₂ 0.5, HEPES 10, EGTA 5, Mg-ATP 2; the pH was adjusted to 7.3 with CsOH. Biocytin (1.5 mg/ml) was included in the pipette solution so that cell location in the slice could be confirmed. The amplifier head stage was connected to an Axopatch 700B amplifier (Molecular Devices; Foster City, CA, USA). Fine and course movement of the pipettes were controlled by micromanipulators (PatchStar, Scientifica) mounted upon a fixed platform. The amplifier current output was filtered at 10 kHz (–3 dB, 8-pole low-pass Bessel) and digitized at 20 kHz using a National Instruments digitization board (NI-DAQmx, PCI-6052E; National Instruments, Austin, TX, USA). Data acquisition was performed using WINWCP (Version 4.1.2) and WINEDR (Version 3.0.9) kindly provided by John Dempster (John Dempster; University of Strathclyde, UK).

For reconstruction of neuronal morphology from biocytin fills, the tissue was preserved in 4% paraformaldehyde for over 48 h. Paraformaldehyde was then washed off the tissue with ice cold PBS 3×, 10 min each time. Slices were then blocked and permeabilized with 0.2% Triton-X in PBS based solution at room temperature for 1–2 h. After further washing with PBS for 10 min, slices were submerged in 2 mg/ml Streptavidin, Alexa Fluor 555 Conjugate (Life Technologies) with 0.2% Triton-X for 3–4 h at room temperature. Slices were washed again in PBS (3×, 10 min each) and mounted on slides with mounting medium (H-1000, Vectashield).

Total membrane capacitance (Cm) was calculated from Cm = Q/ΔV, where Q was the charge transfer during a hyperpolarizing 10 mV step of the command voltage (ΔV). The total membrane conductance (Gm) was calculated from Gm = Iss/ΔV where Iss was the average steady-state current during the ΔV. Cells were excluded from further analysis if Gm < 1 nS as a low resting input conductance is a defining feature of thalamic interneurons. The electrode to cell series resistance (R_S) was calculated from the relationship R_S= ΔV/I_P where I_P was the peak of the capacitive current transient and recordings were excluded if Rs increased by >30%. Based upon biophysical and morphological criteria, a total of 47 recordings were made from thalamic relay neurons in control mice and 66 recordings were obtained from HDC-Δγ2 mice.

Spontaneous IPSCs (sIPSCs) were detected using scaled template matching and aligned on their initial rising phases. Waveform averages were constructed from sIPSCs that exhibited monotonic rises and uninterrupted decay phase. Average baseline current levels were calculated during a 10 ms epoch immediately before each detected event and the peak amplitude was determined relative to this value. The weighted decay of individual sIPSCs was calculated as the charge transfer during the baseline corrected sIPSC divided by the sIPSC peak amplitude. The increased holding current induced by DS-2 application was calculated from all-point histograms of the holding current using the fitted peak of a single Gaussian function to calculate the average amplitude of the holding current before and after DS-2 application.

All average values represent the mean ± the standard error of the mean (SEM). Data distributions were compared using Origin 8.5 and functions were fitted to data distributions using unconstrained least-squared fitting procedures. The type of statistical test used in each experiment is specified individually.

The HDC-Cre mouse line was crossed with the LoxSTOPLox Rosa-YFP mouse line (Figure 1A) and YFP expression was examined in the resulting HDC-CRE-YFP line (Figure 1B). As expected from previous studies (Zecharia et al., 2012), the YFP signal was associated with histamine-producing neurons of the Tuberomammillary Nucleus (TMN), ependymal cells lining ventricles and putative macrophages that were sparsely distributed throughout the neocortex. However, the attention of this study was focused on the dLGN where a high proportion of cells were shown to be YFP-positive (Figure 1C). On average we found that 41% of cells were NeuN positive in the adult dLGN (DAPI+: 926 cells, NeuN+: 379 cells) counted in representative slice sections; consistent with previous estimates of neuronal density in the mammalian dLGN (Wei et al., 2011). Co-fluorescence of YFP signal with the neuronal marker NeuN indicated that YFP proteins are exclusively confined in NueN-positive neurons and ~90% of NueN-positive neurons within the dLGN had undergone recombination and expressed YFP (DAPI+: 926 cells, NeuN+: 379 cells, YFP+: 340 cells), because the hdc-cre gene is transiently expressed during postnatal development of the dLGN (Zecharia et al., 2012). In the example volume of tissue analyzed in Figure 1D, 42 out of the 47 NeuN-positive neurons (~89%) expressed YFP (Figure 1E) from a total of 102 cells that were stained with DAPI (Figures 1D–F). These results demonstrate the usefulness of the HDC-Cre mouse as a method for altering gene expression within neurons of the dLGN.

The HDC-Cre mouse was crossed with the γ2I77lox mouse to produce HDC-Δγ2 mice and littermate controls. Whole-cell voltage-clamp recordings were then made from identified thalamic relay neurons of the dLGN (Figure 2A). In littermate control mice, 46 out of 47 neurons exhibited sIPSCs whereas just under half of HDC-Δγ2 cells (31 out of 66) were devoid of sIPSCs (Figure 2B). Across all cells recorded, the average sIPSC frequency was 6.27 ± 0.66 Hz (n = 47) in control cells compared to 2.55 ± 0.4 Hz (n = 66) in HDC-Δγ2 cells resulting in a significant reduction (K-S test, P = 1 × 10⁻⁷) in synaptic drive following γ2 subunit removal. In those cells containing sIPSCs the average frequency was 6.41 ± 0.66 Hz (n = 46) in control cells compared to 4.80 ± 0.51 Hz (n = 31) in HDC-Δγ2 cells (K-S test, P = 0.21). Therefore, the main impact of γ2 removal is the loss of sIPSCs in a subset of thalamic relay neurons. As shown in Figure 2C, the remaining sIPSCs in HDC-Δγ2 cells were blocked by the GABA_AR antagonist picrotoxin (30 μM). This blocking action was associated with a reduction in the holding current and this tonic current was observed in the control and HDC-Δγ2 neurons irrespective of the presence or absence of sIPSCs (Figure 2C). To assay any change in the contribution of δ subunit-containing GABA_ARs to thalamic relay neuron excitability in the HDC-Δγ2 cells, we took advantage of the allosteric modulator DS-2 (Wafford et al., 2009; Ye et al., 2013). As expected the tonic current recorded from thalamic relay neurons was enhanced by DS-2 with little action on sIPSCs (Figure 2D). The DS-2 induced change in holding current was 83.7 ± 20.1 pA (n = 6 cells) in control cells compared to 100.7 ± 12.9 pA (n = 6 cells) in HDC-Δγ2 cells with no significant difference (two-tailed t-test, P = 0.49; Figure 2E). We also did not observe any change in IPSC kinetics or amplitude during DS-2 application (Figure 2F). Consistent with no change in the tonic conductance following γ2 subunit removal, the average input conductance from the cells used in different aspects of this study was 4.03 ± 1.01 nS in the control (n = 47 cells from 27 mice) compared to 3.35 ± 0.89 nS (n = 35 cells from 16 mice) in the HDC-Δγ2 that exhibited sIPSCs and 4.14 ± 0.88 nS (n = 31 cells from 14 mice) in the HDC-Δγ2 that did not exhibit sIPSCs (two-tailed t-test, P > 0.44 in all cases). There was also no significant difference of membrane capacitance between these cell groups (two-tailed t-test, P > 0.12 in all cases; control cells: 99 ± 7 pF, n = 47; HDC-Δγ2 cells with no sIPSCs: 93 ± 5 pF, n = 31; HDC-Δγ2 cells with sIPSCs: 108 ± 8 pF, n = 35), indicating the resting membrane excitability and the cell shape had not dramatically altered. Therefore, crossing the HDC-Cre mouse line with the γ2I77lox mouse line has resulted in the complete removal of sIPSCs from only 50% of thalamic relay neurons raising the possibility that alternative synaptic GABA_AR types contribute to phasic inhibition within the remaining cells of the dLGN.

Kinetic analysis of sIPSCs revealed a small number of neurons in control mice (8 out of 47) that contained a single population of fast rising and fast decaying sIPSCs (Figure 3A). In two of these eight cells, we recovered fills with clear Y-like morphology similar to that reported previously for dLGN relay neurons with predominantly fast IPSCs (Bright et al., 2011). As expected from our previous studies, the majority of thalamic relay neurons (39 out of 47) exhibited a high proportion of slow rising and slow decaying sIPSCs (Figure 3B). As shown in Figure 3D, a clear reduction in sIPSC frequency was apparent across the entire population of cells with no overlap between the distributions of IPSCs in control cells and remaining IPSCs in HDC-Δγ2 cells. In order to estimate the proportion of fast sIPSCs (P_fast-sIPSC) present in any given cell, we defined a cut-off criterion (t_crit) for fast sIPSCs based upon data obtained from cells that exhibited a single population of fast rising and fast decaying sIPSCs. A single Gaussian fit was used to define a t_crit at which fast sIPSCs could be identified at a 95% confidence level. The average t_crit based upon Gaussian fits to the data obtained from all eight fast IPSCs-only cells (termed T_crit, to differentiate with t_crit from individual cells) was 1.7 ± 0.2 ms for the rise-time and 7.9 ± 1.0 ms for the decay. Figure 3C illustrates data from an HDC-Δγ2 relay neuron that contained both fast and slow IPSCs. Using the T_crit values obtained from the wild-type population, the P_fast-IPSC was 0.1 in this cell. To determine whether γ2 deletion has reduced the prevalence of fast IPSCs across all cells, the distribution of P_fast-IPSC was also compared (Figure 3E). There was a reduction in the prevalence of fast IPSCs with only one recording from the HDC-Δγ2 mice giving a P_fast-IPSC >0.5. Indeed, γ2 subunit removal was associated with a reduction in P_fast-IPSCs (K-S Test, P < 0.001) with an average P_fast-IPSC of 0.27 ± 0.03 in recordings from the control mice compared to 0.11 ± 0.02 in recordings from the HDC-Δγ2 dLGN. The loss of these fast sIPSCs in the knockout may reflect the loss of γ2-containing GABA_ARs in a mixed GABA_AR population in dLGN thalamic relay neurons. To test this hypothesis the pharmacological data associated with our whole-cell voltage-clamp recordings was analyzed.

To distinguish between GABA_AR heterogeneity or partial recombination in dLGN neurons, we assayed the diazepam sensitivity of the remaining sIPSCs recorded from HDC-Δγ2 neurons as the γ2F77I point mutation abolishes zolpidem sensitivity but diazepam sensitivity persists (Buhr et al., 1997; Cope et al., 2004). In the littermate control cells, 3 μM diazepam caused the average sIPSC weighed decay time to increase from 10.94 ± 0.37 ms to 15.39 ± 0.38 ms (n = 4; paired t-test, P < 0.001) with little change in the average 10%–90% rise-time (paired t-test, P = 0.26) or peak amplitude (paired t-test, P = 0.4) of sIPSCs. The average sIPSC waveform constructed from a control neuron before and during 3 μM illustrates this prolongation of the sIPSC decay (Figures 4A,B). 3 μM diazepam was then applied to those HDC-Δγ2 cells that contained sIPSCs. No change was observed in the average weighted decay time (n = 5; paired t-test, P = 0.11), 10%–90% rise-time (paired t-test, P = 0.89) or sIPSC peak amplitude (paired t-test, P = 0.27; Figures 4A,B). For example, the average weighted decay time was 10.62 ± 1.03 ms (n = 5) in normal ACSF vs. 11.27 ± 1.02 ms in the presence of 3 μM diazepam. The lack of diazepam sensitivity observed in the HDC-Δγ2 neurons (Figure 3B) clearly indicates that the remaining GABA_ARs, responsible for generating the sIPSCs, do not contain the γ2 subunit.

It is also possible that the remaining sIPSCs are meditated by αβ assemblies that lack γ subunits. However, this GABA_AR type should be potently blocked by Zn²⁺ ions (Draguhn et al., 1990), a feature not observed in either control or HDC-Δγ2 cells (see Figure 4C). For example, in HDC-Δγ2 cells, the sIPSC peak amplitude was 93.03 ± 41.82 pA (n = 9) in normal ACSF compared to 90.35 ± 41.82 pA in the presence of 10 μM Zn²⁺ (paired t-test, P = 0.63) and, on average, the IPSC weighted decay time did not significantly change (+8.4 ± 6.3%, n = 9; paired t-test, P = 0.93). As was the case for all pharmacological manipulations described so far in this study, the sIPSC frequency remained stable at 6.2 ± 1.2 Hz in control ACSF compared to 8.1 ± 1.8 Hz in the presence of Zn²⁺ (two-tailed t-test, P = 0.2).

To test the possibility that other γ subunit-containing GABA_ARs contribute to the remaining IPSCs, we examined the actions of DMCM. As well as removing zolpidem sensitivity, the γ2F77I point mutation results in DMCM insensitivity (Buhr et al., 1997; Ogris et al., 2004; Leppä et al., 2005). Nevertheless, DMCM will act as an inverse agonist at γ3 subunit-containing GABA_ARs (Herb et al., 1992; Kerti-Szigeti et al., 2014), while DMCM will potentiate γ1-containing GABA_ARs (Puia et al., 1991; Khom et al., 2006; May et al., 2013). In control cells, 10 μM DMCM significantly enhanced the average sIPSC charge transfer by 39.8 ± 15.59% (n = 7) due to a combined action on the peak amplitude and decay of sIPSCs (paired t-test, P = 0.004; Figure 4D). A similar action of DMCM was observed in HDC-Δγ2 cells with a 21.57 ± 6.36% (n = 7) increase in charge transfer (paired t-test, P = 0.03; Figure 4D). These results suggest that the γ2 subunit is absent from all cells recorded from the HDC-Δγ2 dLGN, and γ1 subunit-containing GABA_ARs contribute to IPSCs in the dLGN.

The main GABA_AR receptor genes expressed in the thalamus are α1, α4, β2 and δ (Wisden et al., 1992; Pirker et al., 2000); little γ1–3 expression is detected by either in situ hybridization or immunohistochemistry. Indeed, extrasynaptic GABA_AR mediated tonic inhibition dominates in the thalamus (Jia et al., 2005; Bright et al., 2007). Nevertheless, the sensitivity of whole-cell recording is clearly able to demonstrate the presence of phasic inhibition mediated by αβγ-type synaptic receptors (Jia et al., 2005; Bright et al., 2007). We now present evidence that γ2 removal from the dLGN resulted in the complete removal of sIPSCs from half of all relay neurons and the γ2 subunit is in fact absent from the synaptic GABA_ARs that give rise to IPSCs in the remaining cells. This conclusion is based upon the observation that the remaining IPSCs in the γ2 knockout dLGN were diazepam insensitive (see Figures 3A,B). The diazepam induced potentiation of γ1- and γ3-containing GABA_ARs is much less pronounced than that known to occur at γ2-containing GABA_ARs (Puia et al., 1991; Herb et al., 1992; Wafford et al., 1993). The point mutation in the γ2F77Ilox line also results in DMCM insensitivity of γ2 subunit-containing GABA_ARs (Buhr et al., 1997). However, DMCM is an inverse agonist at γ3 subunit-containing GABA_ARs (Herb et al., 1992), and will potentiate currents generated by γ1 subunit-containing GABAARs (Puia et al., 1991). Therefore, the enhancement of sIPSCs we observe with DMCM (see Figure 3C) is consistent with the presence of γ1 subunit-containing GABA_ARs and offers a simple explanation for the sIPSCs that remain in the γ2 knockout (see Figures 4C,D). In contrast, an inhibitory action of DMCM in IPSCs of the neocortex was used to suggest that γ3 subunits are present in the synaptic GABA_ARs that remain following γ2 removal (Kerti-Szigeti et al., 2014). Given that only one γ subunit is present within the pentameric assembly (Olsen and Sieghart, 2009), we propose that the dLGN can express at least three distinct types of GABA_AR. An α1, α4, β2 and δ subunit combination contributes to extrasynaptic GABA_ARs that mediate the tonic conductance. The α1, α4, β2 and γ2 subunit combinations will contribute to fast synaptic inhibition and we now suggest that the α1, α4, β2 and γ1 subunit combination will generate a slower form of synaptic inhibition within the dLGN.

A simple relationship between γ subunit identity and IPSC kinetics is, however, unlikely given that fast rising and fast decaying IPSCs remain in the HDC-Δγ2 dLGN neurons, possibly as a result of different GABA_AR proximity to GABA release sites. However, our results clearly demonstrate that deletion of γ2 subunit reduced the proportion of fast rising and fast decaying IPSCs across all cells. The γ1 subunit influences GABA_AR clustering at central synapses (Dixon et al., 2014), giving rise to slow IPSCs in neurons of the central amygdala (Esmaeili et al., 2009). Macroscopic and single channel behavior of γ1 and γ2 subunit-containing GABA_ARs indicates little difference in activation and deactivation, but inclusion of the γ1 subunit was reported to slow both the rise and decay of sIPSCs and this was interpreted in terms of “loose clustering” of synaptic GABA_ARs (Dixon et al., 2014). We have previously concluded that spillover of GABA from local dLGN interneurons did not result in the activation of high-affinity δ subunit-containing extrasynaptic GABA_ARs within the dLGN in spontaneous activity recordings (Bright et al., 2011; Ye et al., 2013). Consistently, we have no evidence that the sIPSCs remaining in the HDC-Δγ2 dLGN neurons involve activation of these particular receptors as the δ subunit selective drug DS-2 (Wafford et al., 2009) has no action on IPSC properties even though the tonic conductance was clearly enhanced by this allosteric modulator (see Figures 2C,D). However, spillover of GABA onto these δ subunit-containing extrasynaptic GABA_ARs occurs onto VB relay neurons in response to stimulated burst firing of the nRT (Herd et al., 2013). Similarly, we recently reported that DS-2 application resulted in a slowing of ChR2-evoked IPSCs that are driven by optogenetic GABA release from dLGN interneurons (Jager et al., 2016). These results are not contradictory if the magnitude of the GABA transient associated with spontaneous release were much less than the GABA transient associated with evoked release.

Uniquely, GABA release within the rodent dLGN reflects afferent input from both the nRT and release from local dLGN interneurons. Other nearby first order thalamic nuclei such as the VB do not contain local interneurons, and GABA release in these nuclei is more restricted to the nRT input (Herd et al., 2013). Indeed, γ2 gene deletion from the VB nucleus resulted in a loss of IPSCs from all relay neurons examined (Rovó et al., 2014), which raises the possibility that the remaining IPSCs in the dLGN following γ2 deletion in our study are mediated by local interneurons not present in the VB. We do not exclude the possibility that the observed prevalence of the γ1 subunit can be a compensatory effect of γ2 deletion. Nonetheless, the presence of γ1-containing GABA_ARs following γ2 deletion in our study highlight the importance of γ subunit-containing GABA_ARs in the dLGN, compared to similar γ2 deletion studies mentioned above. Infrequent GABA_AR-mediated responses did remain in some cells in Rovó et al. (2014), but the extremely slow activation/deactivation of these events was interpreted in relation to extrasynaptic δ subunit-containing GABA_AR activation following GABA spillover. Indeed, simultaneous paired recording experiments have demonstrated that nRT burst firing can generate these slow GABA_AR-mediated responses within VB relay neurons (Herd et al., 2013). Importantly, this particular spillover response was absent when extrasynaptic GABA_ARs were genetically removed. Rhythmic activity in the neocortex was little altered following γ2 deletion in VB (Rovó et al., 2014), suggesting that these spillover currents can entrain thalamocortical oscillations in the absence of fast IPSCs (Rovó et al., 2014). Previously, we have also shown that global oscillatory activity across the neocortex was not affected in the HDC-Δγ2 mice during sleep/wake cycle (Zecharia et al., 2012). The IPSCs remaining in the dLGN may well be sufficient to maintain rhythmic activity, but the presence of δ subunit-containing GABA_ARs may also enable spillover-mediated inhibition to occur following nRT related burst firing in a similar manner to that suggested for VB (Rovó et al., 2014).

Comparing these results with similar studies highlights the complexity of synaptic GABA_AR targeting that is present in the mammalian brain. Purkinje cells (Wulff et al., 2009b), hippocampal parvalbumin interneurons (Wulff et al., 2009a) and VB neurons (Rovó et al., 2014) exclusively use γ2 subunit-containing GABA_ARs to generate fast IPSCs while some neocortical neurons make additional use of γ3 subunit-containing GABA_ARs to generate slow decaying IPSCs (Kerti-Szigeti et al., 2014). By combining quantitative analysis with pharmacological data in HDC-Δγ2 neurons, we have now demonstrated that deletion of γ2 subunit-containing GABA_ARs in the dLGN only results in complete deletion of IPSCs in half of dLGN neurons. The remaining slow rising and slow decaying IPSCs are not mediated by γ2 subunit-containing GABA_ARs, and they appear to involve activation of γ1 subunit-containing GABA_ARs. This highlights a possible requirement for distinct types of inhibitory control within the different pathways of the visual thalamus.