Abstract
Knowledge of thalamocortical (TC) processing comes mainly from studying core thalamic systems that project to middle layers of primary sensory cortices. However, most thalamic relay neurons comprise a matrix of cells that are densest in the “nonspecific” thalamic nuclei and usually target layer 1 (L1) of multiple cortical areas. A longstanding hypothesis is that matrix TC systems are crucial for regulating neocortical excitability during changing behavioral states, yet we know almost nothing about the mechanisms of such regulation. It is also unclear whether synaptic and circuit mechanisms that are well established for core sensory TC systems apply to matrix TC systems. Here we describe studies of thalamic matrix influences on mouse prefrontal cortex using optogenetic and in vitro electrophysiology techniques. Channelrhodopsin-2 was expressed in midline and paralaminar (matrix) thalamic neurons, and their L1-projecting TC axons were activated optically. Contrary to conventional views, we found that matrix TC projections to L1 could transmit relatively strong, fast, high-fidelity synaptic signals. L1 TC projections preferentially drove inhibitory interneurons of L1, especially those of the late-spiking subtype, and often triggered feedforward inhibition in both L1 interneurons and pyramidal cells of L2/L3. Responses during repetitive stimulation were far more sustained for matrix than for core sensory TC pathways. Thus, matrix TC circuits appear to be specialized for robust transmission over relatively extended periods, consistent with the sort of persistent activation observed during working memory and potentially applicable to state-dependent regulation of excitability.
Introduction
The prefrontal cortex (PFC) is a group of cortical areas considered critical for executive function and intimately linked to psychiatric diseases (Lewis et al., 2005; Ruzicka et al., 2007; Curtis and Lee, 2010; Lehman et al., 2011). It is thought to exert control throughout the brain to regulate goal-directed behaviors (Miller and Cohen, 2001; Uylings et al., 2003; Vertes, 2006; Sotres-Bayon and Quirk, 2010). There is a widely held belief that the PFC is itself regulated by an ascending activating system involving projections to layer 1 (L1) from the so-called “matrix” or nonspecific thalamocortical (TC) relay neurons located predominantly in the midline, intralaminar, and paralaminar thalamic nuclei (Krettek and Price, 1977; Herkenham, 1986; Vogt, 1991; Jones, 2001; Van der Werf et al., 2002; Rubio-Garrido et al., 2009). Large bilateral lesions encompassing the midline and intralaminar thalamic nuclei reportedly reduce behavioral responsiveness and may even induce coma, whereas electrical activation of these nuclei increases behavioral and cortical arousal (Moruzzi and Magoun, 1949; Jasper, 1960; Jones, 2007; Schiff, 2008; Shah and Schiff, 2010). The mechanisms by which matrix thalamic projections regulate the cortex are unknown, primarily because we know so little about the synaptic influences of these projections on the neocortex. For example, it is unclear whether they have weak and slow modulatory effects or, like core sensory TC projections, they can drive fast and robust cortical responses (Alonso and Swadlow, 2005; Boudreau and Ferster, 2005; Rose and Metherate, 2005; Bruno and Sakmann, 2006; Cruikshank et al., 2010). More fundamentally, the identity of their main postsynaptic target type within L1 is uncertain, i.e., pyramidal dendrites or local inhibitory interneurons. It is not even known whether matrix TC projections cause net excitation or net inhibition in cortex; the existence and magnitude of feedforward inhibition has not been characterized.
Investigation of matrix TC synapses has lagged, in part, because it has been impractical to study them using conventional electrophysiological methods. For example, TC pathways are often nonplanar so their axons tend to become severed in the types of slice preparations normally used for in vitro studies (Lee et al., 2007). Furthermore, selective electrical stimulation of matrix TC axons within the cortex is nearly impossible because of the proximity of these TC axons to other (nontargeted) cells and axons. Here we apply an optogenetic strategy to overcome these difficulties (Petreanu et al., 2009; Cruikshank et al., 2010; Yizhar et al., 2011a) and study the synaptic effects of thalamic matrix input on PFC for the first time. Channelrhodopsin-2 (ChR2) was expressed in matrix thalamic neurons projecting to L1 of medial areas of PFC (mPFC). ChR2 expression occurred throughout the membranes of the thalamic neurons, including their extended axons. Using optical stimuli, we selectively excited these TC axons within L1 of mPFC and characterized their synaptic influences on local cortical neurons and circuits.
Materials and Methods
Subjects and optogenetic virus injections.
Procedures were approved by the Brown University Institutional Animal Care and Use Committee. Viruses carrying fusion genes for ChR2 and fluorescent proteins were injected unilaterally into thalamus or cortex of mice of either sex in vivo between postnatal days 13 and 17. For TC synaptic experiments, paralaminar and midline thalamic nuclei that project to L1 of mPFC were targeted for the injections (Table 1) (Krettek and Price, 1977; Herkenham, 1986; Paxinos and Franklin, 2001; Rubio-Garrido et al., 2009). For corticocortical (CC) synaptic experiments, the anterior part of the mPFC (the prelimbic area, 1.98–2.68 mm anterior to bregma) was the target for virus injection.
Viral DNA was generously provided by Karl Deisseroth (Stanford University, Stanford, CA). For all of the TC injections and one-third of the PFC injections, we used a lentivirus [pLenti–synapsin–ChR2(H134R)–EYFP–WPRE] that had an enhanced ChR2–EYFP fusion gene driven by a synapsin1 promoter. For the remainder of the PFC injections, we instead used adeno-associated viruses (AAVs): pAAV–CaMKIIa–hChR2(H134R)–EYFP and pAAV–Ef1a–DIO–hChR2(H134R)–EYFP were each used for one-third of the CC experiments. Sequence information can be found at www.stanford.edu/group/dlab/optogenetics/. After amplification using Qiagen kits, vesicular stomatitis virus-G pseudotyped lentivirus or AAV2 or AAV2/5 viruses were produced at the University of Pennsylvania Vector Core. Injection volumes were 0.2–2 μl. All three viruses produced effective ChR2–EYFP expression in cells at the injection sites and in their downstream axons and terminals, including those in L1 of PFC.
Of the 45 mice studied here, 37 had ICR genetic backgrounds and eight had mixed backgrounds (either ICR/C57 or ICR/FVB). Three of the latter were Cre driver mice from GENSAT (the MR90–CRE line) in which we injected a Cre-dependent virus [pAAV–Ef1a–DIO–hChR2(H134R)–EYFP] into the PFCs to obtain ChR2 expression mainly in L3 pyramidal cells.
Slices and solutions.
After allowing 8–90 d for ChR2 expression (usually ∼14 d), acute brain slices (325–375 μm thick, coronal plane) containing mPFC areas ipsilateral to the virus injections were made for in vitro recording and optical stimulation using previously described methods (Cruikshank et al., 2010). The mPFC areas targeted for recording included the first and second cingulate cortices (Cg1 and Cg2), the prelimbic cortex (PrL), and occasionally the infralimbic cortex (IL), medial orbital cortex, or the border between Cg1 and the secondary motor cortex (M2) as defined by the atlas of Paxinos and Franklin (2001). Mean and median ages at the day of recording were P28 and P27, respectively. Experiments were conducted at 32°C in a submersion recording chamber. Slices were bathed in artificial CSF (ACSF) containing the following (in mm): 126 NaCl, 3 KCl, 1.25 NaH2PO4, 2 MgSO4, 26 NaHCO3, 10 dextrose, and 2 CaCl2 (saturated with 95% O2/5% CO2). Patch micropipettes were filled with the following (in mm): 130 K-gluconate, 4 KCl, 2 NaCl, 10 HEPES, 0.2 EGTA, 4 ATP-Mg, 0.3 GTP-Tris, and 14 mm phosphocreatine-K, pH 7.25 (∼290 mOsm). All drugs were applied through the bathing solution.
Whole-cell recordings and measurements of intrinsic properties.
Cells in L1–L3 of mPFC were visualized with infrared differential interference contrast (DIC) microscopy and initially targeted by anatomical position and soma size/shape. After achieving whole-cell configurations, intrinsic membrane and synaptic properties of the neurons were characterized (Beierlein et al., 2003; Chu et al., 2003; Cruikshank et al., 2007; Lee et al., 2010). Recordings were conducted with Molecular Devices hardware and software (Multiclamp 700B, Digidata 1440A, pClamp 10). Series resistances (∼12–30 MΩ) were compensated online (100% for current clamp, 50–70% for voltage clamp). Voltages were corrected for a 14 mV liquid junction potential. Resting potentials were measured within 2 min of break-in, and then steady-state potentials were usually adjusted to −79 mV with intracellular current. Membrane time constants (τm), input resistances (Rin), and input capacitances (Cin) were calculated from voltage responses to small negative current injections (typically 10–20 pA, 600 ms). For τm, the voltage responses were fitted with a single exponential to the initial 100 ms of the response, omitting the first 1 ms. Rin was determined from Ohms law. Cin was calculated as τm/Rin. Sag potentials were determined using large negative currents (>100 pA, 600 ms) that reached peak negative potentials of −115 mV or greater. Peak sag amplitude was measured relative to the voltage at the end of the 600 ms current step. Latency-to-sag peak was measured from the onset of the step. Sag width was measured at half of the peak amplitude.
Intrinsic spiking properties were characterized by injecting suprathreshold positive current steps (usually ≥50 pA). Spike threshold was determined as the membrane potential occurring 50 μs before the peak of the third derivative of the voltage, verified by visual inspection. Peak spike amplitude was measured relative to threshold. Spike width was measured at half of the peak amplitude. Afterhyperpolarization was measured as the voltage change from spike threshold to the trough after the spike. Spike latency was measured relative to the onset of a 600 ms threshold intensity current injection. Spike frequency adaptation was quantified by calculating the adaptation ratio, defined as the number of spikes in the second half of a 600 ms current injection divided by the number of spikes in the first half, averaged across all sweeps in which the cell fired at mean rates of 10–40 Hz. Cells were tested for the presence of slow afterdepolarizations by injecting a 3–5 ms suprathreshold current and then determining whether or not the resulting spike was followed by an afterdepolarization (with a 30–100 ms latency to peak) (Hestrin and Armstrong, 1996; Chu et al., 2003).
Optical stimulation of ChR2-expressing pathways and postsynaptic recordings.
Synaptic responses to blue laser stimulation (447 nm) of the ChR2-expressing TC or CC axons were measured from postsynaptic cortical neurons (which were ChR2 negative) recorded in whole-cell current clamp and voltage clamp. The laser was focused as a 10-μm-diameter spot through a 40× water-immersion objective. Maximum total laser power of the optical stimulation system at the focal plane of the slice was ∼11.3 mW. In many experiments, the intensity was reduced (specified in Results). For synaptic stimulation, 0.1–2 ms flashes (usually 1 ms) were delivered either directly in or near the cortical column of the recorded cell to excite the ChR2-expressing axons and terminals in the TC or CC pathways that were presynaptic to the recorded cell. Error bars in the figures are SEM.
Results
The majority of virus injections into paralaminar and midline thalamic nuclei (16 of 24) resulted in ChR2/EYFP-expressing TC terminal arbors that were concentrated in the outer half of cortical L1. An example illustrating the fluorescence expression pattern is shown in Figure 1A. Using functional mapping (Petreanu et al., 2009), we found that excitable matrix TC axons and synapses were localized to outer L1, consistent with the fluorescence patterns (Fig. 1B–D). For example, optical stimulation of ChR2-expressing TC axons in outer L1, in which EYFP labeling was intense, evoked strong postsynaptic responses in L1 cortical neurons (Fig. 1B,C). In contrast, stimulation of deeper locations, such as the faintly labeled L1–L2 boundary region, was far less effective (Fig. 1C). A similar laminar gradient of effective stimulation sites was observed when TTX and 4-AP were applied to block axon conduction but allow direct optical stimulation of ChR2-expressing boutons (Petreanu et al., 2009; Cruikshank et al., 2010) (Fig. 1D). Under these pharmacological conditions, the locations of effective optical stimulation sites correspond with the locations of ChR2-expressing synapses; thus, it appears that the highest density of matrix TC synapses were in outer L1 (Fig. 1D). The locations of thalamic injections used in the synaptic experiments that follow are presented in Table 1. Consistent with previous anatomical studies, the thalamic origins of outer L1 projections included the ventromedial, anteromedial, reunions, or rhomboid nuclei (i.e., paralaminar and ventral midline nuclei; Table 1) (Krettek and Price, 1977; Herkenham, 1986; Vogt, 1991; Monconduit and Villanueva, 2005; Rubio-Garrido et al., 2009). We also observed that injections into several other thalamic nuclei, including the paratenial nucleus, resulted in projections to inner or middle tiers of L1 (n = 7; data not shown); these are not further considered here.
Matrix TC projections are widespread (Herkenham, 1986; Van der Werf et al., 2002; Monconduit and Villanueva, 2005; Jones, 2007; Rubio-Garrido et al., 2009), and there is a long-held assumption that their synaptic effects are slow and diffuse, potentially important for modulating cortical tone rather than for transmitting the kinds of temporally precise information carried by core sensory TC projections (Morison and Dempsey, 1941; Moruzzi and Magoun, 1949; Jasper, 1960; Shah and Schiff, 2010). In contrast to this assumption, we found that optical stimulation of matrix TC arbors within L1 drove robust and fast postsynaptic responses in mPFC neurons, including L1 interneurons and L2/L3 pyramidal cells (see Figs. 1⇓–3). Mean EPSP amplitudes were 9.8 ± 1.2 mV and onset latencies were 2.4 ± 0.1 ms (n = 53 and 50 cells, respectively; 1.0 ms laser stimuli applied to L1 at maximum intensity). As illustrated in Figure 1B, optical activation of the matrix TC synapses could, in some cases, drive action potentials in postsynaptic neurons (13 of 53 tested neurons).
The short latencies of the TC responses suggest that they are likely mediated by fast ionotropic rather than metabotropic (i.e., modulatory) glutamate receptors. To test this directly, we blocked ionotropic NMDA- and AMPA/kainate-type glutamate receptors while monitoring the effects on TC-evoked synaptic responses. Combined application of these glutamate receptor antagonists essentially eliminated the responses, with no obvious slow metabotropic receptor-mediated components remaining (Fig. 2). Thus, matrix thalamic pathways to L1 of mPFC excite postsynaptic cells mainly via fast ionotropic glutamate receptors, and these circuits could, in principle, carry high-fidelity information.
It has been suggested previously that TC synapses within L1 are mainly located on distal dendrites of pyramidal cells that extend apical processes into L1 (Vogt, 1991; Jones, 2001; Monconduit and Villanueva, 2005; Murayama et al., 2009; Rubio-Garrido et al., 2009). Thus, excitatory TC inputs to L1 would presumably depolarize the membrane potentials of pyramidal cells, enhancing their ability to respond to synaptic signals from other sources that are directed more proximally. However, other potential targets of TC terminals in L1 are the often-neglected inhibitory interneurons that inhabit the layer (Zhou and Hablitz, 1996a; Chu et al., 2003; Zhu and Zhu, 2004; Lee et al., 2010; Wozny and Williams, 2011; Arroyo et al., 2012). Unexpectedly, we found that matrix TC pathways were much more effective at exciting L1 interneurons than L2/L3 pyramidal cells. This is illustrated in Figure 3, A and B, with a simultaneously recorded cell pair. In that pair, the TC-evoked EPSCs were ∼3.5 times larger in the L1 interneuron than in the L2/L3 pyramidal cell, and there were even larger differences in EPSPs for the two cells (Fig. 3B; VC and IC, respectively). Across the tested population, L1 cells had approximately threefold larger TC-evoked EPSPs than the pyramidal cells (Fig. 3C). The L1 interneurons were also more likely than pyramidal cells to be excited above spike threshold by matrix TC input (29% of interneurons vs 8% of pyramidal cells). In addition to differences in response amplitudes, there were clear differences in short-term synaptic dynamics among postsynaptic cell types, with pyramidal responses typically undergoing initial facilitation during repetitive TC stimulation, whereas interneuron responses generally exhibited modest depression (Figs. 1C, 2A, 3B,C; discussed below).
Interneurons of L1, which are believed to be exclusively GABAergic, are physiologically, morphologically, and neurochemically heterogeneous (Zhou and Hablitz, 1996a; Chu et al., 2003; Zhu and Zhu, 2004; Lee et al., 2010; Wozny and Williams, 2011). Although no consensus exists regarding nomenclature of L1 cells, we were able to use previously described criteria (Chu et al., 2003; Zhu and Zhu, 2004; Lee et al., 2010) to classify the majority of our recorded L1 interneurons into unambiguous late-spiking (LS) or non-late-spiking (NLS) types (Fig. 4, Table 2). The remaining L1 interneurons could not be definitively assigned to either group and were therefore left uncategorized. Strikingly, this sorting revealed that TC responses of LS cells were more than twice as large as those of NLS cells, on average (Fig. 4C,D). We also found that NLS cell bodies tended to cluster in the inner half of L1, whereas LS cells were more dispersed (Fig. 4E). This anatomical distinction could potentially explain the weaker TC responses in NLS cells. Recall that matrix TC terminals were located in outer L1 (Fig. 1, Table 1), relatively far from the NLS somata. In contrast, these TC terminals were spatially intermixed with a large fraction of LS somata, putting the LS cells in a potential position to receive effective TC synaptic input on their proximal dendrites. Consistent with the idea that terminal–somata proximity could affect response strength, there was a marked overall correlation between somatic position in L1 and TC-evoked EPSP amplitude, with cells in outer L1, near the TC terminals, responding most strongly (Fig. 4F).
Although NLS cells received relatively weak TC input, we observed that they could respond quite strongly to optogenetically activated CC excitatory input (Fig. 5). In contrast with the TC terminals, CC terminals were often observed in inner L1, near the NLS somata (Fig. 5) (Vogt et al., 1981; Vogt, 1991; Miró-Bernié et al., 2006; Smith et al., 2010). These data suggest a possible sublamination of inhibitory information processing in L1, with NLS cells concentrated in the inner tier, interacting preferentially with CC circuits, and LS cells handling the TC signals that arrive in the outer tier. It will be important to investigate the possible effects of such sublaminar processing in future studies, especially given previous observations that NLS cells are more likely than LS cells to send inhibitory axons to deep cortical layers (Chu et al., 2003; Zhu and Zhu, 2004).
It is clear from the present findings that matrix TC inputs can strongly excite L1 interneurons (especially LS cells; Figs. 1, 3, 4), so we next tested the consequences of this activation. As a first step, we used paired whole-cell recordings to characterize the synaptic connections between L1 interneurons and neighboring cortical cells in mPFC. L1 interneurons had high probabilities of forming inhibitory synapses onto L2 pyramidal cells and other L1 interneurons (∼50% incidence; Fig. 6A–C). IPSPs in both types of postsynaptic cells underwent similar short-term depression when stimulated repetitively at 10 Hz. The unitary IPSPs were smaller and had slower rates of rise in postsynaptic pyramidal cells than in interneurons, consistent with more distal synaptic locations for the pyramidal cells (Fig. 6A–C). Unlike inhibition, excitatory connections between L2 pyramidal cells and L1 interneurons were rare (1 of 19 pairs), similar to results from somatosensory cortex (Chu et al., 2003; Wozny and Williams, 2011).
The observations that L1 interneurons are excited by TC inputs and inhibit neighboring neurons leads to the prediction that matrix TC activation should initiate disynaptic feedforward inhibition. We confirmed this prediction: optical stimulation of TC axons in L1 resulted in feedforward inhibition in 7 of 15 L1 interneurons and 7 of 8 L2/L3 pyramidal cells tested (Fig. 6D). This inhibition was generally strongest early during a 10 Hz TC stimulus train and then depressed subsequently (Fig. 6D; responses at +6 mV). In pyramidal cells, the valence of short-term plasticity for disynaptic inhibition was opposite that for monosynaptic TC excitation, because excitatory responses facilitated slightly during the first few stimuli in a train (Figs. 2A, 3C, 6D).
Discussion
We have shown that matrix TC axons make robust synapses with both excitatory and inhibitory cell targets in outer L1 of mPFC. The matrix TC inputs most strongly excited inhibitory interneurons of L1, and it is highly probable that these interneurons mediated the rapid feedforward inhibition observed after TC stimulation.
L1 cells have only rarely been considered in discussions of TC circuitry (Mitani and Shimokouchi, 1985; Vogt, 1991; Zhu and Zhu, 2004). This is surprising given recent interest in these cells and studies of their developmental origins (Lee et al., 2010; Rudy et al., 2011), unique features of their physiology, pharmacology, and morphology (Christophe et al., 2002; Chu et al., 2003; Zhu and Zhu, 2004; Wozny and Williams, 2011; Arroyo et al., 2012), and their critical roles in learning and sensory processing (Letzkus et al., 2011; Palmer et al., 2012). L1 neurons are also interesting because of emerging evidence that disruption of inhibition and interneurons, including L1 interneurons, has a pivotal role in neuropsychiatric diseases (Lewis et al., 2005; Ruzicka et al., 2007; Yizhar et al., 2011b; Marín, 2012).
Recently, Letzkus et al. (2011) made the remarkable observation that L1 interneurons respond briskly to footshock stimulation even when they are located in cortical areas not generally associated with somatic sensation (i.e., auditory and visual cortices). This raises the possibility that L1 interneurons have common roles (perhaps signaling nociception or arousal) that are independent of the modalities of the cortical areas in which they are embedded (i.e., visual, auditory, etc.). It will be important to determine whether this is true for L1 cells of mPFC and other cortical areas. Much of the footshock-induced L1 cell activity reported by Letzkus et al. depended on nicotinic acetylcholine receptors, but the fastest component appeared to be glutamatergic. Based on previous studies of nociceptive processing in thalamus (Monconduit et al., 1999; Monconduit and Villanueva, 2005), it seems that this glutamateric component could be mediated by one or more of the nonspecific/matrix TC pathways studied here, suggesting potentially shared routes of activation across multiple cortical areas.
Surprisingly, Letzkus et al. (2011) reported that L1 cell activation produced strong inhibition of inhibitory interneurons in the underlying cortical layers but no observable inhibition of pyramidal cells. This was consistent with a previous study using cholinergic agonists to activate L1 cells (Christophe et al., 2002) but strikingly different from our findings using paired cell recording. We found that L1 interneurons monosynaptically inhibited neighboring L2/L3 pyramidal cells with a high (∼50%) probability. Likewise, other studies using paired cell methods also reported L1 cell inhibition of pyramidal cells (Chu et al., 2003; Wozny and Williams, 2011). These data alone might lead to a reasonable hypothesis that cholinergic activation somehow negatively gates L1 cell output to pyramidal cells. However, a recent study involving optogenetic activation of cholinergic axons (Arroyo et al., 2012) refutes this simple possibility by demonstrating that excitation of L1 cells via selective activation of cholinergic axons can produce inhibition of L2/L3 pyramidal cells.
To be sure, it will be important to reconcile the differences in L1 cell effects observed in different investigations. Nevertheless, it is clear from our experiments that matrix TC activation strongly excites L1 interneurons and can trigger feedforward inhibition of pyramidal cells. Some of this feedforward inhibition is almost certainly produced by L1 synapses on the apical dendrites of the pyramidal cells, in which many L1 interneurons (especially LS/neurogliaform cells) project dense terminal arbors (Hestrin and Armstrong, 1996; Zhou and Hablitz, 1996b; Chu et al., 2003; Zhu and Zhu, 2004). Such inhibition could have a variety of effects, including providing a potent means of regulating dendritic action potentials (Kim et al., 1995; Murayama et al., 2009). Along these lines, Palmer et al. (2012) recently presented evidence that sensory-evoked activation of L1 interneurons causes GABAB inhibition in apical compartments of L5 pyramidal cells, suppressing their dendritic action potentials.
Matrix-evoked feedforward inhibition suggests that matrix TC input may have a suppressive function within cortex, at least in part. In addition, it is possible that inhibition of other interneurons could result in a disinhibitory (i.e., net excitatory) effect on pyramidal cells, as discussed previously (Christophe et al., 2002; Letzkus et al., 2011; Arroyo et al., 2012). We observed clear inhibition of L1 interneurons (Fig. 6A,C), and it remains to be seen how the matrix TC input affects interneurons of lower layers.
Of course, matrix TC axons themselves make glutamatergic synapses on apical dendrites of pyramidal cells. As with core TC pathways, the effect of this direct excitation is sometimes suppressed after a short latency by feedforward inhibition (Fig. 6D). An interesting possibility is that the temporal precision of the suppression may be enforced by specialized spatial relationships between the TC and inhibitory synapses (Kubota et al., 2007). In any case, feedforward inhibition can impose a narrow “window of opportunity” for excitation (Gabernet et al., 2005; Cruikshank et al., 2007). During trains of matrix TC stimulation, the direct excitatory responses in pyramidal cells tended to facilitate, whereas the disynaptic inhibition depressed. This suggests a dynamic shift in the net effect of the matrix TC input during repeated activation, from net inhibition to net excitation (Gabernet et al., 2005). It is possible that this progressive increase in excitatory/inhibitory balance might play a role in the production of the “recruiting response,” an evoked potential generated near the cortical surface that can be triggered by repeated stimulation of certain nonspecific thalamic nuclei and that has eluded a satisfying mechanistic explanation for >70 years (Dempsey and Morison, 1941; Jasper, 1960; Castro-Alamancos and Connors, 1997).
A particularly interesting property of matrix TC input is its short-term synaptic plasticity. As just discussed, monosynaptic TC responses of pyramidal cells underwent short-term facilitation (Fig. 3C), and this contrasts with most TC synapses characterized previously. With rare exceptions (Tan et al., 2008; Viaene et al., 2011), core sensory TC synapses undergo robust short-term depression (Gibson et al., 1999; Beierlein et al., 2003; Boudreau and Ferster, 2005; Rose and Metherate, 2005; Bruno and Sakmann, 2006), including when they are tested with optogenetic methods identical to those applied here (Fig. 7) (Cruikshank et al., 2010). Furthermore, although matrix TC synapses on L1 inhibitory interneurons generally showed some short-term depression, it was markedly weaker than the depression observed in interneurons of the somatosensory system (Fig. 7) (Cruikshank et al., 2010). The striking differences in short-term dynamics of core sensory and matrix TC synapses are directly illustrated in Figure 7. Altogether, these observations about synaptic dynamics indicate that matrix TC synapses are far more capable of sustained responses during repeated activation than are core sensory TC synapses. The distinctive dynamics may be specializations tailored to the divergent functions of the circuits in which they are embedded. For example, sensory TC systems may be optimized for processing transient signals, such as onsets and offsets of sensory stimuli (Zhu and Zhu, 2004; Rose and Metherate, 2005; Bruno and Sakmann, 2006), whereas matrix TC systems could transmit information with more sustained profiles, including the arousal state of the organism (Harris and Thiele, 2011). However, the capacity of matrix circuits for maintaining responsiveness during repetitive activation should not be confused with kinetic sluggishness. On the contrary, the individual EPSPs within trains of matrix TC stimulation were relatively fast as well as strong. The ability of matrix TC circuits to carry rapid signals over sustained periods to L1 of PFC may contribute to the celebrated propensity of this cortical area to process high-frequency (e.g., gamma) oscillations and persistent activity associated with attention and memory (Miller and Cohen, 2001; Lewis et al., 2005; Curtis and Lee, 2010; Sotres-Bayon and Quirk, 2010).
Footnotes
This work was supported by National Institutes of Health Grants NS025983 and MH086400, National Science Foundation Grant EFRI-0937848, Department of Defense Grant DARPA-BAA-09-27, and the Brown Institute for Brain Science. We thank Karl Deisseroth and Charu Ramakrishnan for generous support with DNA and their time.
- Correspondence should be addressed to Scott J. Cruikshank, Department of Neuroscience, Box G-LN, Brown University, Providence, RI 02912. Scott_Cruikshank{at}brown.edu