|Home | About | Journals | Submit | Contact Us | Français|
Integrative brain functions depend on widely distributed, rhythmically coordinated computations. Through its long-ranging connections with cortex and most senses, the thalamus orchestrates the flow of cognitive and sensory information. Essential in this process, the nucleus reticularis thalami (nRT) gates different information streams through its extensive inhibition onto other thalamic nuclei; however, we lack an understanding of how different inhibitory neuron subpopulations in nRT function as gatekeepers. We dissociated the connectivity, physiology, and circuit functions of neurons within rodent nRT, based on parvalbumin (PV) and somatostatin (SOM) expression, and validated the existence of such populations in human nRT. We found that PV but not SOM cells are rhythmogenic, and that PV and SOM neurons are connected to and modulate distinct thalamocortical circuits. Notably, PV but not SOM neurons modulate somatosensory behavior and disrupt seizures. These results provide a conceptual framework for how nRT may gate incoming information to modulate brain-wide rhythms.
Clemente et al. deconstruct the reticular thalamus (nRT) at the cellular, circuit, and behavioral levels. They find that parvalbumin- and somatostatin-expressing neurons in nRT have distinct cellular and circuit properties, segregate along predominantly non-overlapping neuronal pathways, and differentially modulate thalamocortical rhythmogenesis, somatosensory behavior, and generalized seizures.
The nucleus reticularis thalami (nRT) is the “guardian of the gateway” in the thalamocortical circuit, given its role in modulating interactions between the cerebral cortex and thalamus (Crick, 1984; Halassa and Acsády, 2016). The nRT modulates thalamocortical oscillations that underlie functions like such as attention, sensation, sleep, and consciousness (Calabrò et al., 2015; Crick, 1984; Steriade, 2005). Disruptions in this circuit correlate with neurological (Paz et al., 2010; Slaght et al., 2002) and psychiatric disorders (Ferrarelli and Tononi, 2011; Zhang et al., 2010).
The nRT contains inhibitory neurons that release gamma-aminobutyric acid (GABA). It is the main source of inhibition for the excitatory relay thalamus in rodents (Gentet and Ulrich, 2003; Houser et al., 1980). The nRT has been thought to include mainly parvalbumin-expressing neurons (PV) (Csillik et al., 2005; Hou et al., 2016; Jones and Hendry, 1989) but somatostatin-positive cells (SOM) were recently found in rodent nRT (Ahrens et al., 2015; Wells et al., 2016). These cells have not been detailed in humans. PV and SOM neurons have distinct functions in the cortex and hippocampus (reviewed by Kim et al., 2016), while little is known about whether they mediate different functions of the nRT. Further, the existence of functionally distinct neuron types within the nRT has been suggested (Halassa et al., 2014; Lee et al., 2007) but not associated with molecular markers.
nRT neurons can orchestrate brain-wide network activity (Crick, 1984; Halassa and Acsády, 2016). Brief optogenetic stimulation of nRT neurons in vivo induces spindles in the cortex during non-rapid eye movement sleep (Halassa et al., 2011), while sustained stimulation increases delta band power similar to slow-wave sleep (Lewis et al., 2015). Variations in anatomical connections between nRT and thalamic relay cells differentially modulate cortical oscillatory activity and attention (Halassa et al., 2014; Macdonald et al., 1998), but it remains unknown whether these different functions are mediated by cellular substrates that can be distinguished by molecular markers.
Because the nRT does not project directly back onto the cortex, nRT modulation of specific cortical regions depends on its topographical organization and connections with thalamic relay nuclei (Crabtree, 1992; Lam and Sherman, 2011). However, it remains unknown whether the different cell types in the nRT (i.e. PV and SOM cells) project to distinct thalamocortical relay nuclei to produce different functions in vivo.
We hypothesize that the nRT’s multiple functions are mediated by PV and SOM neurons that are anatomically connected to different circuits. Here, we deconstruct nRT PV and SOM neuron populations by assessing their intrinsic electric and synaptic membrane properties, pre- and post-synaptic connectivity, and functions within the thalamocortical circuit during normal and pathological behaviors.
We evaluated the location of PV and SOM neurons across the anterior-posterior (Figure 1A), medial-lateral, and dorsal-ventral axes of the nRT (Figure 1B) with genetic and immunohistochemical approaches. For validation of specificity and selectivity of Cre lines, see Supplemental Methods and Figure S1. PV and SOM cells were located across the entire anterior-posterior axis of the nRT. High-resolution confocal images distinguished singly labeled PV and SOM cells within the somatosensory nRT (Figures 1C–D), identifying a previously undescribed population of SOM neurons in the nRT that do not express PV. We divided the somatosensory nRT into tiers based on their projection patterns onto thalamic relay nuclei as described by (Lam and Sherman, 2011; Pinault and Deschênes, 1998) (Figure 1B & E). The distribution of SOM neurons was significantly different only in the medio-lateral axis of the somatosensory nRT. The number of singly labeled SOM cells was significantly lower than PV cells only in the central tier of the middle sector of the nRT (Figures 1E and 1B). Only a minority of neurons displayed both neurochemical markers (10–20%) in the somatosensory nRT (Figure 1E).
The electrophysiological properties of neurons vary depending on their location in the nRT (Lee et al., 2007). Not all nRT cells have the classic rebound bursting mediated by low-threshold, transient Ca2+ -dependent depolarizations, called low-threshold calcium spikes (LTS) (Contreras et al., 1992; Jahnsen and Llinás, 1984). We sampled neurons across anterior-posterior, medial-lateral, and dorsal-ventral axes of the somatosensory nRT (Figure S2).
To assess the properties of PV and SOM neurons between postnatal day 30 and 50 (P30–P50), we crossed PV-Cre and SOM-Cre mice with Ai32 mice expressing Cre-dependent channelrhodopsin-2 (ChR2). We identified cells via whole-cell patch-clamping based on their opsin expression (Figure S2). We found that the membrane input resistance (Rin), time constant (τm), resting potential (Vm) and capacitance (Cm) were similar in PV and SOM neurons (Table S1; Figure 2A,F), supporting that PV and SOM neurons have similar passive membrane properties. Also, PV and SOM neurons showed similar mean frequency-current (F-I) slope and rheobase [i.e., the minimal current intensity required to trigger an action potential (AP)] (Table S1; Figures 2E–2F)).
In all PV neurons (14 of 14 cells), a hyperpolarization to −95 mV or more was followed by a rebound of excitation characterized by a LTS crowned by a burst of APs. Of 14 PV neurons, 12 displayed repetitive rebound bursts and 2 produced single bursts (Figure 2A). Of 16 SOM neurons, 12 evoked a weak LTS after the same hyperpolarizing steps, and 4 did not evoke an LTS at all (Figure 2A).
We assessed the firing properties of these post-inhibitory rebound bursts. After similar responses to membrane potential hyperpolarizations, the maximal number of rebound bursts was threefold higher in PV than SOM neurons (Figure 2B). Also, the number of APs within the first burst and the intra-burst frequency of APs within the first burst were higher in PV than SOM neurons (Figure 2C–2D).
Next, we studied if differences in LTS and post-inhibitory rebound bursting were controlled by T-type calcium currents (Figure 2G–K). The peak T-current density was higher in PV neurons (Figure 2H) compared to SOM neurons with LTS burst. The voltage-dependence of the steady-state inactivation (SSI) of T-currents was similar in PV and SOM neurons (Figure 2J)—similar half-maximal SSI voltages (V50%) and decay time constants (Figures 2I and 2K)—suggesting that PV and SOM neurons have similar biophysical properties of T-currents. The differences in T-currents between PV and SOM neurons did not depend on the specific location of the neuron within the dorsal-ventral, anterior-posterior, or medio-lateral axes. Thus, small T-current produces the weaker post-inhibitory rebound LTS bursting in SOM neurons, suggesting that SOM neurons contain fewer and/or more dendritically located T-type calcium channels.
To further validate PV-Cre, we used “INTRSECT” mice (Fenno et al., 2014) in which we selectively expressed ChR2 with the combinatorial virus (Cre-On, Flp-Off) in PVCre/SOMFlp (PV*) mice. This approach expressed ChR2 in PV neurons that do not express SOM. We examined the electrophysiological properties of neurons and found that PV* and PV neurons have similar intrinsic membrane properties, including post-inhibitory rebound bursting and T-currents (Figure S6A–D).
To determine the post-synaptic targets of PV and SOM neurons, we injected virus containing enhanced Yellow Fluorescent Protein (eYFP) (Figures 3A,C,E) or ChR2-eYFP (Figures 3B,D,F) in the nRT of PV-Cre and SOM-Cre mice. nRT axons robustly expressed ChR2-eYFP 2–4 weeks post-injection, so we could activate axonal terminals while patching from thalamocortical relay cells in different nuclei. By evoking IPSCs, we verified that the projections observed with eYFP were functional. The major targets of PV neurons were the ventroposteromedial relay thalamic nuclei (VPM), ventroposterolateral (VPL) (Figure 3A,B left), and PO (Figure 3C,D left; see also Figure S4B). The major targets of SOM neurons were the intralaminar (IL) thalamocortical nuclei, in particular the parafascicular nucleus (PF) (Figure 3E,F right; see also Figure S4A), and to a lesser extent, the ventroposterolateral (VPL) (Figure 3A,B right) and the ventromedial (VM) thalamocortical relay nuclei. At high magnification, we saw axonal boutons and varicosities in the targeted structures (Figure 3 insets). The eIPSCs were recorded with kynurenic acid (2 mM) to block glutamatergic transmission and were abolished by the GABAA receptor blocker picrotoxin (data not shown). Patched cells were not directly activated, supporting that the opsin was not expressed in thalamocortical cells.
Optical stimulation was specific to nRT cells. We validated that the virus specifically targeted the nRT by serially sectioning the brain and noting the eYFP somatic expression in all mice (Figure S4). We observed the same targets with different viral volumes (see Supplemental Experimental Procedures).
The development and organization of the primate thalamus, particularly the organization of the GABAergic system, are different from rodents (Jones, 2007). A whole brain study found SOM immunoreactivity in the human thalamus but provided no direct evidence of cell bodies or axons (Bouras et al., 1987), and the co-localization of SOM and PV neurons had not been examined. We immunostained postmortem human thalami (Figure 4) and found SOM neurons at both anterior (Figure 4A) and posterior levels (not shown), and at dorsal and ventral positions. We also found varicose SOM axons in human thalamic sections containing the higher-order thalamocortical relays, including the IL (Figure 4E), mediodorsal, anteroventral, ventral anterior, and midline nuclei (Figure 4D). Notably, we discovered a distinct SOM+PV- subpopulation in human nRT, similar to mice.
Because PV neurons exhibit stronger intrinsic oscillatory properties than SOM neurons, we hypothesized that PV neurons are more likely to initiate network oscillations within the nRT-ventrobasal (VB) thalamic microcircuit. We used horizontal thalamic slices that conserve the connectivity between the nRT and VB thalamus (VB = VPM + VPL), in which we could assess nRT-VB intra-thalamic network oscillations independent of other structures (e.g., cortex) (Huntsman et al., 1999; Paz et al., 2011a, 2013; Makinson et al., 2017).
We measured intra-thalamic oscillations evoked by a single optical stimulation of PV or SOM neurons in slices prepared from PV-Cre or SOM-Cre mice expressing ChR2 (Figure 5). With an array, we recorded multi-unit (MU) activity within the nRT and VB thalamus (Figure 5A). Optical activation of PV neurons evoked rhythmic oscillatory bursting activity in VB (Figure 5B), similar to previous work using electrical stimulations of the internal capsule (Huntsman et al., 1999; Makinson et al., 2017; Paz et al., 2011a, 2013; Barthó et al., 2014). Activation of SOM cells in the nRT evoked firing in VPL but not in VPM (Figure 5B). The number of active channels in VB (with a circuit oscillation evoked by an optical stimulation of nRT) and evoked bursts was higher in PV- than SOM-stimulated slices (Figure 5C,D). Evoked oscillatory firing in VB resulted from circuit oscillations in nRT-VB-nRT loops rather than direct activation of VB cells with light, because the light did not induce direct firing in VB channels (Figure 5B) and the VB neurons did not express ChR2 (see above). Optical activation of either PV or SOM cells also evoked a similar instantaneous firing in the nRT (Figures 5B,E), suggesting that weaker VB firing in SOM-stimulated slices is not due to weaker activation of SOM compared with PV neurons.
These findings support that PV neurons are the main oscillation generators in the nRT-VB somatosensory circuit because PV nRT neurons can generate circuit oscillations in the nRT–VPL/VPM–nRT loop, while SOM nRT neurons can generate circuit oscillations in VB only via the nRT–VPL–nRT loop.
To learn whether nRT PV and SOM cells receive projections from distinct brain regions, we used a modified rabies, trans-synaptic tracing strategy (Wall et al., 2010, 2013) in PV or SOM Cre–expressing mouse lines. The first injection contained Cre-dependent adeno-associated viruses (AAV) expressing the avian tumor virus receptor A (TVA) tagged with mCherry and the rabies glycoprotein (RG). The second contained a modified rabies virus (EnvA) and was injected 21 days after the first injection (Figure 6A), which allowed rabies to infect cells in a Cre-dependent manner and be retrogradely trafficked to label inputs onto nRT neurons (Wall et al., 2010, 2013). We report the proportions of input cells in Figure 6B, because our helper virus did not allow us to simultaneously visualize starter and input cells. Inputs were normalized across each animal to prevent mice with many labeled inputs from biasing total input proportion. Only inputs that were detected in at least three mice were included in the analysis. For SOM-Cre (n=3) and PV-Cre (n=5) animals, we detected 150±46 and 307±40 total input cells, respectively (p>0.5, Mann-Whitney test).
We found distinct populations of afferent neurons targeting PV and SOM neurons in the nRT (Figures 6B–C). When targeting SOM nRT cells, presynaptic cells were labeled in VPL and anterior thalamic nuclei (AN) (Figure 6B). SOM cells also received input from extrathalamic sources, including the central amygdala (CeA) and basal ganglia [the external segment of globus pallidus (GPe)] (Figure 6C). When targeting PV nRT cells, we saw presynaptic labeling mainly in the VPM and S1 cortex (Figures 6B–C). We found a small number of input cells in GPe and none in CeA or AN when targeting PV cells (Figure 6C). We also found input cells in the nRT when targeting both PV and SOM cells, with SOM cells having significantly more input cells (Figure 6C), suggesting that dendro-denritic or axo-dendritic intra-nRT connections exist preferentially onto SOM cells.
These projections are not exhaustive (Jones, 2007), because the virus may not have infected all nRT cells across the medial-lateral, dorsal-ventral, and anterior-posterior axes. However, our results show that within the same nRT regions, SOM and PV neurons are targeted by different brain regions. Importantly, the starter cells infected with the TVA-mCherry virus were located specifically in the nRT (Figure S5). A subpopulation of VPM cells expresses PV, and larger volume injections in nRT led to viral expression in VPM (see Supplemental Methods), but this did not change the input cell location that was mainly restricted to S1BF.
As PV and SOM neurons segregate into different input-output circuits, we asked whether they could differentially control cortical states during free behavior. We designed devices containing multiple electro-corticogram (ECoG) screws in the cortex and a chronic multi-site optrode in the nRT (Figure 7A). A single, unilateral 20 ms-long optical activation of PV neurons evoked higher frequency and more robust oscillations than SOM neurons, mainly in the 7–15 Hz sigma range (Figures 7B–E). Such stimulations of nRT were previously shown to produce 7–15 Hz sleep spindles in the cortex (Halassa et al., 2011), but the cellular subtype of the nRT spindle generator remained unknown.
We also determined if we could reproduce the major effects of PV activation on the ECoG in PV-Cre mice by activating only PV* cells in the nRT. Using the INTRSECT approach (Fenno et al., 2014), we selectively activated PV* cells in the nRT, which induced sigma-range oscillations in the S1 cortex (Figure S6 F–I) similar to that seen in PV-Cre mice.
Because PV cells more robustly modulate VPM and S1 cortical areas (Figures 5, ,7),7), and because VPM is preferentially involved in whisker-related sensation, we hypothesized that PV cells would be preferentially involved in whisker-related behavior. With the adhesive-tape task (Bouet et al., 2009; Fleming et al., 2013), we tested whether disrupting the output of PV or SOM cells onto thalamocortical nuclei would alter somatosensory behavior. In PV-Cre-Ai32 and SOM-Cre-Ai32 mice, we placed adhesive tape on the left whiskers and manipulated the activity of PV or SOM cells in the right nRT that is functionally related to left-whisker perception. We manipulated the activity of ChR2-expressing PV or SOM cells with an optogenetic approach that has been established to disrupt thalamocortical output onto the cerebral cortex in a behavioral context (Reinhold et al., 2015). Activating PV but not SOM cells induced abnormal grooming of the whisker pad area that outlasted the optical stimulation protocol. Indeed, PV-targeted mice continued scratching the whisker pad area even after the tape was removed (Figure 7L, M). No such behavior was induced with light alone (Figure 7M), suggesting that manipulating nRT PV but not SOM neurons alters sensory perception in the mice during the task.
Given that nRT PV activation generates more robust oscillations in the cerebral cortex, we asked whether targeting these cells during seizures could effectively disrupt generalized cortical oscillations. We used low doses of pentylenetetrazol (PTZ), an established model of pharmacologically-induced generalized seizures mainly involving thalamocortical circuits (Beenhakker and Huguenard, 2009; Malafosee, 1994). We then used an optogenetic approach to selectively modulate PV or SOM nRT neurons during free behavior. Such optogenetic manipulations have been used to silence relay thalamocortical output and assess their effect on cortical states (Lewis et al., 2015). We found that a unilateral optical stimulation of nRT PV cells readily disrupted bilateral generalized seizures (Figure 7G,I,J,K) in PV-Cre mice injected with ChR2. We validated this result in PV* mice, in which targeting the PV+SOM- population immediately aborted ongoing seizures (Figure S6J-I). In contrast, activation of nRT SOM cells in SOM-Cre mice injected with ChR2 had much more stochastic effects on ongoing seizures (Figure 7H,I,J,K). The differences between PV and SOM mice did not result from differences in behavioral states (see Methods).
The distinct effects of PV and SOM neurons on circuits have been extensively studied in other brain structures (Kim et al., 2016; Phillips and Hasenstaub, 2016; Wilson et al., 2012), but not in the nRT. Our study is the first to assign cellular and circuit properties to neuron populations of the nRT that can be differentially targeted with distinct molecular markers. We also provide the first evidence that singly labeled PV and SOM neurons co-exist in the human nRT. In mice, we found that PV and SOM neurons have distinct electrophysiological properties, segregate into different anatomical locations across the nRT, and participate in predominantly non-overlapping anatomical pathways that distinctly modulate thalamocortical circuits both in vitro and in vivo.
The nRT needs LTS-mediated burst firing to generate spindle oscillatory activity in the thalamocortical circuit (Bazhenov et al., 2000; Beenhakker and Huguenard, 2009; Deschenes et al., 1984; Steriade et al., 1987). The nRT-VB loop is a key node of thalamocortical rhythm generation, because thalamocortical VB cells also fire LTS-mediated rebound bursts after hyperpolarization induced by nRT GABAergic inputs (Jahnsen and Llinás, 1984). We found that PV, but not SOM neurons, have large T-currents, are rhythmogenic and recruit more robust oscillatory spindle-like activity in the cortex, suggesting that PV neurons are the main spindle rhythm generators in thalamocortical circuits. We also found that unilateral optogenetic targeting of PV nRT cells disrupts generalized seizures, supporting that PV cells are preferentially involved in rhythmogenic thalamocortical circuits and could be “choke points” for seizure disruption (Paz and Huguenard, 2015). In contrast, targeting SOM cells can either interrupt or enhance seizures, suggesting a more nuanced effect on seizure rhythms. Thus, the PV nRT–VPM loop is the main pacemaking circuit in the somatosensory thalamus, in part because PV cells can fire more powerful rebound bursts and have more robust reciprocal connectivity with VPM neurons that are wired to oscillate in the sigma range. We propose that SOM nRT projections to the relay thalamus generate different global forebrain activity, which requires further studies.
Our major finding is that PV and SOM cells are embedded in distinct circuits: PV cells are mainly embedded in sensory circuits that involve inputs and outputs from specific thalamic nuclei (mainly VPM), whereas SOM cells are mainly embedded in circuits that involve non-specific thalamocortical relay nuclei (AN, IL, PF thalami) that have broader projections onto many cortical regions, including PFC (Saalmann, 2014). In previous work (Crabtree, 1992), somatosensory and intralaminar thalamic nuclei shared nRT input. Our results suggest that this input is mostly mediated by SOM neurons, because SOM but not PV neurons project onto the intralaminar nuclei in addition to VPL. Validating this hypothesis will require in vivo intracellular or juxtacellular biocytin labeling of SOM cells and a 3D reconstruction of their axonal projections. Although previous studies showed that S1 cortex and VPM thalamus are the main excitatory afferents to nRT (Crabtree, 1992; Lam and Sherman, 2011), our results identify that PV neurons are the cellular substrate in nRT that receive these inputs. Thus, PV nRT cells could be the main recipients of top-down cortical modulation of the thalamus and the main gatekeepers of somatosensory processing. Our finding that specifically manipulating PV cells induces excessive grooming after the tape was removed from the mouse’s whiskers suggests that PV neurons are the main cellular substrates in the nRT involved in whisker-related sensory perception and memory. We propose that SOM nRT neurons are less involved in whisker-sensory related behaviors, supporting a lack of projections onto VPM, which modulates the S1 barrel cortex. This result will motivate future studies on how SOM neurons affect behavior.
The cerebral cortex is thought to be the main excitatory afferent of the nRT (Williamson et al., 1993; Liu and Jones, 1999). Thus, our finding that, at least in the somatosensory nRT, S1 cortex exclusively targets PV and not SOM neurons, whereas subcortical structures preferentially target SOM cells, is novel. Our work suggests that the cortex is not the main modulator of a major subset of nRT neurons, and it urges the need to revise the framework of input analysis in the somatosensory corticothalamo-cortical circuits by including our findings that PV and SOM nRT cells differentially modulate information processing in these circuits.
Although studies have shown that the basal nucleus (BM/BL) and cortical nuclei (Co) of the amygdala project onto the nRT in non-human primates (Zikopoulos and Barbas, 2012), our study is the first to show a projection from the CeA to the nRT in mice. nRT SOM cells receive inputs from CeA and AN that are involved in selective attention (Holland et al., 2000) and emotional-attention interactions (Sun et al., 2015), respectively. Based on these inputs, we speculate that nRT SOM cells are the cellular substrates that impart emotional salience to attention. Furthermore, because SOM cells in the somatosensory nRT mainly receive projections from subcortical structures and target non-specific intralaminar thalamic nuclei that modulate the PFC (Saalmann, 2014), we propose that SOM cells are preferentially involved in the bottom-up limbic modulation of cortical circuits. Altogether, these inputs-outputs position the nRT SOM cells as the cellular targets to probe the limbic nature of nRT-mediated behavior.
Previous animal studies found that projections from the basal ganglia onto the nRT originate in the substantia nigra pars reticulata and GPe (Asanuma, 1989; Cornwall et al., 1990; Gandia et al., 1993; Hazrati and Parent, 1991; Jones, 2007). Our finding that GPe preferentially projects onto SOM cells in nRT suggests that SOM nRT cells could be central to communication between the basal ganglia and the thalamocortical relay nuclei. Thus, the GPe-SOM-TC pathway may affect behavioral switching and reward prediction (Gittis et al., 2014).
We studied PV and SOM neurons as they are the most abundantly expressed markers in rodent nRT (Lein et al., 2007). Future studies may include other neuron markers, such as neuropeptide Y (Brill et al., 2007), cholecystokinin (Cox et al., 1997), and vasoactive intestinal peptide (Burgunder et al., 1999).
The structure of the nRT does not support selective viral targeting, so we used multiple approaches to ensure targeting specificity. We showed that 1) in both AAV-eYFP and transynaptic rabies studies, our viral approach sufficiently targeted most cells in the nRT; 2) with large viral volumes, the virus spread to VB in PV-Cre mice but did not affect the results or data interpretation; and 3) the major electrophysiological features in vitro, and the major circuit oscillation and seizure disruption properties in vivo, were similar in transgenic PV-CrexAi32 and virally targeted PV-Cre and PV*ConFoff mice. Thus, our results are not specific to one Cre line and do not result from non-specific targeting of brain regions other than nRT.
Aberrant activity of the nRT affects many neurological and neuropsychiatric disorders (Paz et al., 2010; Slaght et al., 2002; Steriade, 2005; Ahrens et al., 2015; Ferrarelli and Tononi, 2011;Wells et al., 2016). We propose that PV cells generate rhythms that regulate sensory thalamocortical information processing, whereas SOM cells might gate limbic information and create informational salience. Within this framework, disruption of PV and SOM cells could lead to disorders with excessive rhythmogenesis (e.g., seizures) and psychiatric disorders (e.g., attention deficit disorder, schizophrenia), respectively. Our finding that pure, non-overlapping PV+SOM- and SOM+PV- cells co-exist in the human nRT supports these asymmetric roles.
We propose that PV and SOM cells in the nRT might be differentially involved in top-down versus bottom-up thalamocortical information processing. Future work will evaluate whether these cell types interact within the nRT to integrate the different information needed for attention, consciousness, and perception. Also, the nRT—with its densely packed and perhaps interconnected PV and SOM cells—could be an ideal target to study the high co-morbidity between neurological and psychiatric disorders.
Finally, our results suggest that PV and SOM cells may mutually inhibit each other, providing another mechanism to tune what information the nRT, and in turn the thalamus, controls (Crick, 1984; Sohal et al., 2000).
Integrative brain functions require the flexible distribution of neuraly encoded information across large-scale brain circuits. Through its cellular subtypes, the nRT could differentially tune incoming information to coordinate brain-wide rhythms.
We performed all experiments per protocols approved by the Institutional Animal Care and Use Committee at the University of California, San Francisco and Gladstone Institutes. Precautions were taken to minimize stress and the number of animals used in all experiments. Adult (P30-P180) male and female mice were used. Somatostatin (SOM)-Cre mice (SOM-IRES-Cre, IMSR_JAX:013044; mixed C57BL/6;129S4), Parvalbumin (PV)-Cre mice (PV-Cre, IMSR_JAX: 017320; C57BL/6 congenic), Rosa26-tdTomato reporter mice (Ai14, IMSR_JAX:007914; congenic C57BL/6 generously provided by Dr. Ken Nakamura, Gladstone Institutes); Rosa26-ChR2(H134R)-EYFP reporter mice (Ai32, IMSR_JAX:012569; mixed C57BL/6;129S4), C57BL/6J mice (wild-type, IMSR_JAX:000664), and PV-Cre/SOM-Flp mice (PV-p2A-Cre, SOM-IRES-Flp (Fenno et al., 2014) were used.
We performed stereotaxic injections of viruses into nRT as described (Paz et al., 2011, 2013). We targeted nRT with stereotaxic coordinates ranging from 1.3–1.6 mm posterior to Bregma and 2.0–2.1 mm lateral to the midline at two different injection depths (2.6 and 3.0 mm) ventral to the cortical surface. To determine the effects of PV or SOM neuron activation on cortical rhythms and behavior, we used PV-Ai32 and SOM-Ai32 mice, as well as PV-Cre, SOM-Cre, and PV-Cre/SOM-Flp mice previously injected with virus encoding ChR2, for simultaneous ECoG and thalamic LFP/MU recording combined with optical stimulation in nRT as described (Paz et al. 2011, Paz et al., 2013).
Immunohistochemistry on mouse brain sections was performed as described (Paz et al., 2010). For human tissue staining, we used control human thalamic tissues (n=3) from two male subjects (55- and 77-years-old) and one female subject (77-years-old) who had no history of brain disorders. All procedures were approved by the Regional and Institutional Committee of Science and Research Ethics of Scientific Council of Health (ETT TUKEB 31443/2011/EKU (518/PI/11)). See Supplemental Experimental Procedures for detailed image acquisition parameters.
Acute slice electrophysiology was done as described (Paz et al., 2010, 2011b). For thalamic oscillation recordings, we used a linear multi-electrode array (Neuronexus). Evoked oscillations were elicited by activating PV or SOM neurons of the nRT by delivering 450 nm light (0.8–5 mW, 100 ms pulses) to nRT via a 200μm optical fiber (Thorlabs).
Electrophysiological recordings in freely-behaving mice were performed as described (Paz et al., 2013). We recorded ECoG and thalamic LFP/MU signals using an RZ5 (TDT). To evoke cortical rhythms, we delivered 8–20 mW of 450 nm light.
We used a low dose of pentylenetetrazol (PTZ, 35–60 mg/kg) injected intraperitoneally to induce seizures in mice. Experimenters visually monitored ECoG and thalamic LFP/MUA signal for spike-and-wave seizures and brief generalized spiking seizure episodes (<10 s). During seizures that were >1 s in length, experimenters triggered optical stimulation (4 s, 450 nm, 25–35 mW).
To assess whisker-related somatosensory behaviors, the same mice were subjected to a modification of the adhesive removal task behavior involving the front paws and whisker pad (Bouet et al., 2009) with and without optical manipulation of nRT.
All numerical values are given as means and error bars are standard error of the mean (SEM) unless stated otherwise. Parametric and non-parametric tests were chosen as appropriate and were reported in figure legends. Data analysis was performed with MATLAB (SCR_001622), Origin 9.0 (Microcal Software, SCR_002815), GraphPad Prism 6 (SCR_002798), R-project, and SigmaPlot (SCR_003210). (*p<0.05, **p<0.01, ***p<0.001).
Detailed experimental procedures and statistical analyses for each experiment can be found in Supplemental Experimental Procedures.
J.T.P. is supported by NIH/NINDS R00NS078118, NIH/NINDS R01NS096369, Gladstone Institutes, Kavli Institute for Fundamental Neuroscience, Michael Prize, DoD EP150038, and NSF # 1608236. A.C.P and F.S.C. are supported by the National Science Foundation Graduate Research Fellowship Awards (NSF #1650113 and #1144247, respectively). S.S.H. is funded by the Achievement Rewards for College Scientists Scholarship. A.C.P, F.S.C., and S.S.H. are supported by NIH T32-GM007449 and the Weill Foundation. S.R.M. is supported by the American Epilepsy Society and the Dravet Syndrome Foundation Postdoctoral Research Fellowship. K.D and L.E.F. are supported by NIMH, NIDA, DARPA, and the Gatsby Foundation. C.D. and L.A. are supported by the National Research Development and Innovation Office (NKFI109754) and the Hungarian Brain Research Program (KTIA_13_NAP-A-I/1). We thank the Department of Pathology, Saint Borbála Hospital, Tatabánya at the Human Brain Research lab, and IEM HAS for providing human brain tissue. We thank John Huguenard and Jordan Sorokin for the custom Matlab software for thalamic oscillation analysis in vitro, and Meredith Calvert of the Gladstone Histology & Light Microscopy Core for help with confocal microscopy and cell counting. Thanks to Gary Howard and Crystal Herron for their insightful feedback on the manuscript.
Author Contributions:Conceptualization, A.C.P. and J.T.P.; Investigation, A.C.P., S.R.M., B.H., A.U., S.B., L.A., C.D.; Data analysis: A.C.P, S.R.M., B.H., F.S.C., M.W., A.U., J.T.P. Writing – Original Draft, A.C.P., and J.T.P.; Writing – Review and Editing, A.C.P., S.R.M., F.S.C., S.S.H., B.H., L.A. and J.T.P., Funding Acquisition, A.C.P and J.T.P.; Resources, L.E.F. and K.D (provided INTERSCT mice); Supervision: J.T.P.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.