Thalamocortical Up States: Differential Effects of Intrinsic and Extrinsic Cortical Inputs on Persistent Activity

Spontaneous Up states in thalamocortical slices

Using thalamocortical slices from adult mice, we conducted simultaneous intracellular (whole-cell) and extracellular (field potential and multiunit) recordings from microelectrodes placed in layers IV–III of the SI cortex. All of the cells reported in this study (n = 45) had overshooting action potentials, a resting membrane potential more negative than −60 mV, and an average input resistance of 154 ± 65 MΩ (measured with a −0.1 nA current pulse). Figure 1 shows typical spontaneous activity recorded from a sample of three cells taken from different experiments. In these slices, spontaneous Up and Down states recur at ∼0.1 Hz. The Up state is readily detected in the field potential recording as a synchronous population (network) event that in the intracellular recording is characterized by barrages of postsynaptic potentials that depolarize the recorded neuron by 5–15 mV for the duration of the network event, which on average lasts 1.3 ± 0.5 s (n = 12 slices). During the Down state, cortical cells are relatively hyperpolarized, and there is little synaptic or network activity. These characteristics are indistinguishable from Up and Down states previously described in ferret slices (Sanchez-Vives and McCormick, 2000; Shu et al., 2003) and resemble those observed in vivo, although the Down state is usually longer and the Up state is shorter than in vivo (Steriade et al., 1993a). In some experiments, neurobiotin was included in the intracellular recording solution to label and subsequently trace the recorded cells (n = 5 cells labeled) (Fig. 1 B,C). The labeled cells were pyramidal-type cells located in layers IV–III.

To examine the characteristics of Up and Down states in individual cells, we moved the membrane potential to different levels with the intracellular injection of direct current (Fig. 2 A). At depolarized potentials, the cells fired spontaneous action potentials immediately before and during an Up state but were silent for several seconds after the Up state (i.e., during the Down state). The reduced excitability during the Down state after each Up state was demonstrated by applying short depolarizing current pulses (Fig. 2 B). Before and during the Up state, the current pulse reliably evoked action potentials, whereas during the Down state, the same pulse was not capable of eliciting action potentials. Such a reduction in cell excitability immediately after the Up state may be related to a slowly recovering after-hyperpolarizing potential that is particularly obvious when the cell is depolarized by current injection (Fig. 2 A, top traces). This potential after the Up state reversed at quite hyperpolarized potentials (more negative than −90 mV; n = 3 cells; data not shown), suggesting that it may be primarily mediated by a K⁺ conductance.

Up states in cortex are correlated with thalamic activity

Simultaneous recordings from both cortex and thalamus in thalamocortical slices should reveal whether the thalamus is in any way engaged with the cortical Up states. Here we distinguish between TC slices and non-TC slices. TC slices are those that have intact thalamocortical and/or corticothalamic connections, so that electrical stimulation within the thalamus produces some effect in the cortex. Non-TC slices are those cut adjacent to the TC slices in the same experiment but do not contain intact thalamocortical or corticothalamic fibers, so that electrical stimulation in the thalamus produces no effect in the cortex. We first conducted labeling experiments to determine the presence of intact corticothalamic fibers in TC slices.

In a group of TC slices (n = 8 slices from eight mice), we confirmed the presence of intact corticothalamic connections between the SI cortex and ventrobasal (VB) thalamus by applying neurobiotin with iontophoresis in the VB thalamus. The iontophoresis pipette was placed in the area of the VB thalamus, in which electrical stimulation triggered field potential responses in layers IV–III of SI cortex. If the corticothalamic cells in SI cortex have intact fibers that reach the VB thalamus, this procedure should result in retrogradely labeled cells in layer VI of SI cortex. Indeed, in every TC slice tested (n = 8), neurobiotin applied in the VB thalamus of TC slices resulted in labeling of a band of pyramidal cells in layer VI (i.e., corticothalamic cells). An example of this labeling is shown in Figure 3 A (left; the cortical area of retrograde labeling is marked by arrows).

In some experiments (n = 8 slices in four mice), we also applied neurobiotin in the VB thalamus of adjacent non-TC slices (i.e., those slices in which electrical stimulation of the VB thalamus produced no response in the SI cortex). In non-TC slices, the iontophoresis pipette was placed in the middle of the VB thalamus or in areas equivalent to the TC slices. We did not find any retrogradely labeled cells in the neocortex of non-TC slices despite extensive labeling in the thalamus. Two examples are shown in Figure 3 A (middle, right). Thus, the labeling example in Figure 3 A shows three consecutive slices from the same animal (the middle 80 μm cryostat section of each of the three 400 μm slices is shown). Note that only the TC slice (left), which was the only slice that produced field potential responses in cortex when stimulated in the thalamus, showed retrogradely labeled cells in the cortex. It is worth mentioning that our slicing procedures are quite standardized, and we could predict with confidence, before histological and/or electrophysiological confirmations, which slice in the order of slicing was the most effective TC slice. It is logical that we found intact corticothalamic connections in these slices because electrical stimulation of the SI cortex in TC slices produces synaptic responses in VB thalamocortical relay cells that are mediated by corticothalamic synapses (Kao and Coulter, 1997; Castro-Alamancos and Calcagnotto, 1999, 2001). These results demonstrate that TC slices contain intact (descending) corticothalamic connections, in addition to the well known intact (ascending) thalamocortical connections. Through descending corticothalamic connections, cortical Up states may drive thalamocortical relay cells, and through ascending thalamocortical connections, relay cells may affect cortical Up states.

Using TC slices, we set out to determine whether spontaneous Up states in SI cortex had any relationship with the activity of thalamocortical relay cells in VB thalamus. Field potential and multiunit activity was recorded through one electrode in SI cortex, and single units were recorded through another electrode in the VB thalamus. In TC slices, spontaneous cortical Up states were accompanied by activity in the VB thalamus (Fig. 3 B), indicating that the cortical Up state was driving the activity of thalamocortical cells via corticothalamic connections. The VB thalamocortical cell shown in Figure 3 had an overall spontaneous firing rate of 2.3 Hz (measured over 105 min) and, as denoted by the peaks of <10 ms in the autocorrelogram (Fig. 3 C, right), tended to fire frequently in bursts. To measure the relation between the spontaneous cortical activity and the thalamic activity, we computed cross-correlations between the cortical Up states and the thalamic spikes (Fig. 3 C, left). Each cortical Up state was detected during its onset from the field potential recording. This time stamp was used as the marker to correlate with the thalamic spikes. The cross-correlation revealed large peaks that were well above a shifted correlogram (shift predictor) for the same activity (overlaid in green). In this experiment, the thalamic firing followed very faithfully the cortical Up state, indicating that the cortical Up state was driving synchronous thalamic activity. Moreover, the thalamic firing could begin ∼500 ms before the Up state was detected in the cortex. This suggests that thalamic activity may be triggering some of the cortical Up states, because it precedes them. An average of all of the detected field potential Up states for this experiment is also overlaid on the cross-correlogram (blue trace). In other experiments with TC slices, the cortical Up states were not at all preceded by any thalamic activity. Instead, the thalamic activity followed very reliably the cortical Up state (Fig. 4 A). In these cases, the cortex was driving the thalamocortical cells in the VB thalamus via corticothalamic connections, but there was no indication of thalamocortical cell firing preceding the cortical Up states.

In experiments with non-TC slices, there was no relation between the thalamic activity and the cortical Up states (Fig. 4 B). In these cases, the activity in the SI cortex and VB thalamus occurred independently of each other. Figure 5 shows population data from TC slices that had a significant cross-correlation between the cortical Up state and the thalamocortical cells in VB thalamus (25 ms bin; mean ± SD; n = 8). A gray dashed box marks the time before and after the Up state detection that showed a significant (p < 0.05) increase in thalamic activity above the shifted correlogram.

These results demonstrate that cortical Up states effectively drive related activity in thalamocortical cells via corticothalamic connections. Moreover, in some cases the activity of thalamocortical cells in VB thalamus precedes the cortical Up states, which suggests that thalamic activity may be triggering some of the cortical Up states. This last hypothesis was further tested next.

The thalamus affects the incidence of spontaneous Up states in the cortex

If spontaneous thalamic activity in VB thalamus is impacting the incidence of Up states in the SI cortex, then interrupting the connections between the thalamus and cortex should affect the occurrence of Up states in SI cortex. To test this possibility, we cut the connections between the thalamus and cortex in TC slices using a fine blade attached to a micromanipulator. Fifteen minutes after the cut, we began measuring the effect of this manipulation on the incidence of Up states in the SI cortex. We simultaneously measured the incidence of Up states in the medial portion of cortex (non-SI cortex) from the same slice, which does not receive intact thalamocortical input in TC slices (Fig. 6 A). Furthermore, we made the same cut in non-TC slices from the same animals, which served as a control of the potential effects of the cutting procedure. A cut of the thalamic radiation in TC slices significantly reduced the incidence of Up states in the SI cortex (n = 12 slices; p < 0.01) (Fig. 6 B) but not in the medial portion of the same slice, which was simultaneously recorded. Moreover, a cut of the thalamic radiation had no effect on spontaneous Up states in non-TC slices (n = 12 slices) (Fig. 6 B). IETHs of the spontaneous Up states in TC slices revealed that the cut significantly reduced the number of cortical Up states occurring at intervals of <12 s (n = 12 slices; p < 0.05) and significantly enhanced the number of Up states occurring at intervals of >30 s (p < 0.05) (Fig. 6 C). None of these changes were observed in the SI cortex of non-TC slices (n = 12 slices) (Fig. 6 C), indicating that the reduction of Up state incidence was caused by eliminating connections between thalamus and cortex and not by other nonspecific effects related to the cutting procedure. A power spectrum FFT analysis of the Up states in SI cortex of TC and non-TC slices showed no differences before and after the cut (n = 12 slices per group) (Fig. 6 D). This indicates that although the cut of the thalamic radiation suppressed the incidence of Up states, those Up states were similar before and after the cut (Fig. 6 D).

Effects of thalamic and cortical stimulation on triggering Up states

The previous results indicate that thalamic activity can precede or follow the onset of spontaneous Up states in SI cortex and that a cut that interrupts connections between thalamus and cortex reduces the incidence of spontaneous Up states in SI cortex. This suggests that thalamic activity may have a role in triggering cortical Up states. If this is the case, stimulating the thalamus should effectively trigger cortical Up states.

We compared the ability of thalamic and cortical electrical stimulation to trigger Up states in SI cortex. To avoid stimulating thalamocortical fibers within the neocortex, the cortical stimulating electrode was placed in the top layers, between layers II and III, and 0.5–1 mm away from the field potential and whole-cell recording electrodes. Thus, two different stimulating electrodes were placed in the slice, one in the VB thalamus and another in the top layers of SI cortex. Input/output (I/O) curves were derived by varying the intensity of the stimulation applied (10–150 μA). An example of such an experiment for a fast-spiking cell (FS cell; putative interneuron) is shown in Figure 7. Because Up states are population events, the effects of electrical stimulation on Up states were similar in regular-spiking and fast-spiking cells. Electrical stimulation of the cortex or thalamus evoked a typical short-latency EPSP that increased with intensity, and this could be followed by a longer-latency Up state. Interestingly, only low-intensity (10–20 μA) stimulation of the cortex triggered an Up state, whereas increasing the cortex stimulation intensity to >20 μA resulted in a stronger short-latency EPSP that triggered an action potential and was followed by a robust IPSP, but an Up state was not produced. In contrast, increasing the thalamus stimulation intensity simply increased the likelihood of triggering an Up state (Fig. 7 A).

Figure 7 B shows the average intracellular responses evoked by stimulating the cortex, the thalamus, or both together (cortex+thalamus) at different intensities for the FS cell in Figure 7 A (15 trials are averaged, and action potentials were removed using a median filter). As quantified later, results in regular-spiking cells were similar. The average depolarization evoked by each stimulation is compared with the average depolarization produced by the same number of spontaneous Up states. Thus, in this experiment, stimulation of the cortex triggered an Up state at 15 μA but not at higher intensities (Fig. 7 B, red trace). The triggered Up states were identical to those occurring spontaneously (i.e., an average of 15 spontaneous Up states is overlaid for comparison) (Fig. 7 B, light gray trace). In contrast, stimulation of the thalamus began triggering Up states at intensities as low as 10 μA, albeit with low probability. Hence, the average depolarization was lower than that of spontaneous Up states. As the intensity of the thalamic stimulation increased to >20 μA, the average depolarization evoked was similar to that of the spontaneous Up states, indicating that an Up state had been triggered in each stimulus trial. Simultaneous stimulation of the thalamus and cortex at low intensities resulted in a stronger depolarization than stimulating the thalamus or cortex separately, indicating a summation of their effects. However, simultaneous stimulation of the thalamus and cortex at higher intensities resulted in a reduced depolarization compared with the thalamus stimulation alone, indicating that the cortex stimulation was suppressing the ability of the thalamus to trigger Up states. Population data from multiple experiments are described below and shown in Figure 8.

To compare the efficacy of thalamic and cortical electrical stimulation at triggering Up states as a function of intensity, we obtained several measures from intracellular and field potential recordings. The first measure was the probability of triggering an Up state based on 15–30 trials per stimulus intensity (n = 12 slices) (Fig. 8 A). This measure is simple to obtain, because Up states are easily identifiable events. To be counted as an Up state, the evoked response should meet two criteria. It should be phase locked to the stimulus (i.e., occur at a fairly constant latency on each trial) and have amplitude and duration similar to spontaneous Up states in the same experiment. As shown in Figure 8 A, the probability of triggering an Up state from the cortex became nil as the stimulation intensity increased. Thus, there was a significant difference between lower- (10–20 μA) and higher (50–150 μA)-intensity stimulation of the cortex in the probability of triggering an Up state (p < 0.01; n = 12 slices). In contrast, the probability of evoking an Up state from the thalamus increased as function of intensity (p < 0.01; lower vs higher intensity; n = 12 slices). For the cortex+thalamus stimulation, the probability of triggering an Up state at low intensities (10–30 μA) was approximately the sum of the cortex and thalamus stimulations delivered separately, but as the intensity increased, the probability was less than the thalamus stimulation alone. Thus, at intensities of >50 μA, the cortex stimulation significantly suppressed the ability of the thalamus to trigger Up states (p < 0.01; cortex+thalamus vs thalamus for 100–150 μA; n = 12 slices).

The second measure we obtained consisted of calculating the area of depolarization above the baseline membrane potential (prestimulus V _m) for the intracellular recordings (n = 7 cells) (Fig. 8 B). The area of depolarization evoked by each stimulus was compared with the area of depolarization of spontaneous Up states (Fig. 8 B, blue). This comparison produced results similar to the probability measure described above. Thus, for the cortex stimulation, at low intensities, the area of depolarization was equivalent to that produced by spontaneous Up states [not significant (n.s.); cortex vs spontaneous Up states for 10–20 μA], but as the intensity increased, the area of depolarization was strongly suppressed (p < 0.01; cortex vs spontaneous Up states for 30–150 μA). For the thalamus stimulation, the area of depolarization increased as a function of intensity and rapidly reached a value equivalent to spontaneous Up states. For the simultaneous stimulation of the cortex+thalamus, at low intensities, the area of depolarization was equivalent to that produced by spontaneous Up states (n.s.; cortex+thalamus vs spontaneous Up states for 10–30 μA), but as the intensity increased, the area of depolarization was significantly less than that of spontaneous Up states (p < 0.05; cortex+thalamus vs spontaneous Up states for 50–150 μA). This indicates that the cortex stimulation was suppressing the ability of the thalamus to trigger Up states.

The third measure we obtained from an additional group of experiments (n = 15 slices) (Fig. 8 C,D) consisted of measuring the peak amplitude of short- (2–10 ms) and long (15–500 ms)-latency field potential responses (Fig. 8 C,D) (n = 15 slices). The peak amplitude of the long-latency field potential response provides a simple way to measure Up states by comparing this amplitude with the amplitude of spontaneously occurring Up states (Fig. 8 C,D). Indeed, the results obtained with this measure were similar to those obtained by measuring the probability of triggering an Up state (Fig. 8 A) and the area of depolarization (Fig. 8 B). The short-latency response increased as a function of stimulation intensity and was larger for the cortex than for the thalamus stimulation. In contrast, the long-latency response, representing the Up state, showed different effects for the cortex and thalamus stimulation. For the cortex stimulation, at low intensities (10–20 μA), the long-latency response was variable but larger than at higher intensities (50–150 μA). Thus, the long-latency response evoked by cortex stimulation was suppressed as the intensity increased. In contrast, for the thalamus stimulation, the amplitude of the long-latency response simply increased as a function of intensity.

In conclusion, these results indicate that the likelihood of triggering an Up state from the thalamus increases rapidly with intensity, whereas the likelihood of triggering an Up state from the cortex is high at low intensities and rapidly decreases to nil as the intensity increases. Moreover, simultaneous stimulation of thalamus and cortex revealed a summation effect at low intensities on triggering Up states and an inhibitory effect of moderate-to-strong cortex stimulation on the ability of the thalamus to trigger Up states.

Thalamic activity triggers cortical Up states

The excellent capability of thalamic electrical stimulation to evoke cortical Up states is in good agreement with the results showing that elimination of the thalamus reduces the incidence of Up states in the cortex and that Up states are correlated with thalamic activity. However, electrical stimulation of the thalamus will not only stimulate thalamocortical fibers ascending to the cortex but will also antidromically stimulate layer VI corticothalamic cells whose axons reach the thalamus. Thus, the intracortical synaptic collaterals of corticothalamic cells could be triggering the cortical Up states in response to thalamic electrical stimulation. To directly address this confound, we used puffs of glutamate delivered to the thalamus as a means of evoking cortical Up states. In TC slices, we first found a location at which electrical stimulation of the VB thalamus evoked an Up state in the cortex. This was followed by placing a glass pipette filled with glutamate in that location and then applying pressure pulses of increasing duration until a cortical Up state was reliably evoked or a maximum of 250 ms was reached. To assure that the glutamate was not spreading to the cortex, after finding an effective glutamate stimulation site in the VB thalamus, we moved the glutamate pipette to a location closer to the cortex (e.g., the thalamic radiation) and checked whether the glutamate pulse was still effective. In all cases, there was a localized region within the VB thalamus that would effectively evoke cortical Up states with glutamate puffs. Moving the glutamate pipette just a few hundred micrometers completely abolished its ability to evoke Up states.

Figure 9 shows two examples of cells that were subjected to these procedures. The left panels overlay thalamic stimulation trials produced by single electrical pulses, the middle panels overlay responses evoked by puffs of glutamate in the same thalamic site, and the right panels overlay spontaneously occurring Up states. Glutamate puffs were very effective at triggering Up states from the thalamus. The Up states triggered by thalamic glutamate had a relatively long but constant latency (within each experiment) from the onset of the glutamate pulse; the shortest latency that we obtained was 87 ms (221 ± 80 ms; n = 6 experiments). The long latency surely reflects the mechanical nature of the pressure-dispensed glutamate. Once an effective thalamic site was found, the Up states evoked by glutamate occurred very reliably (>90% of trials) trial to trial and at a constant latency (Fig. 9). In some cases (n = 4 of 6 experiments), the Up state evoked by glutamate was clearly longer lasting than either the spontaneous or electrically evoked Up states (Fig. 9 A). This likely reflects the sustained firing in thalamocortical cells produced by the long glutamate puff (data not shown), which contrasts with the synchronous and short jolt produced by the single electrical stimulus (200 μs). These results demonstrate that thalamic activity (uncontaminated by antidromic activation of corticothalamic cells) triggers cortical Up states. The results also suggest that sustained thalamic activity may have an added effect of prolonging the cortical Up states. In essence, thalamic activity has a net facilitating effect on cortical Up states.

Effects of thalamic and cortical stimulation during spontaneous Up states

The previous results indicate that thalamic activity has an overall facilitating effect on Up states, whereas cortical activity has a suppressive effect. To further test these hypotheses, electrical stimulation was delivered in the thalamus or in the cortex during the occurrence of spontaneous Up states. This should demonstrate whether thalamic stimulation enhances spontaneously occurring Up states and cortical stimulation suppresses them. To deliver the stimuli during the Up states, an analog threshold detector was used to detect the Up states on-line from the field potential recording. Also, once an Up state was detected, a lockout period assured that nothing was detected for 5 s. Thus, stimuli were tested on the Up state at a rate of <0.2 Hz. In these experiments (n = 5), moderate- to high-intensity cortex and thalamus stimulation (50–100 μA) was used. We compared the area of depolarization from intracellular recordings of spontaneous Up states unaffected by any stimulus and spontaneous Up states affected by either a cortical or a thalamic stimulus delivered during the Up state. A total of 30–50 trials were averaged per cell and condition. The results revealed that spontaneous Up states were enhanced by thalamic stimuli and were suppressed by cortical stimuli (Fig. 10). This effect was observed in every cell tested (n = 5 cells), so that the area of membrane depolarization was 32% larger for spontaneous Up states affected by thalamic stimulation than for unstimulated spontaneous Up states (p < 0.05). Moreover, the area of membrane depolarization was 38% smaller for spontaneous Up states affected by cortical stimulation than for unstimulated spontaneous Up states (p < 0.05). In conclusion, thalamic stimulation has a facilitating effect on spontaneous Up states, whereas cortical stimulation has a suppressing effect on spontaneous UP states.