| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Department of Neurobiology and 2Kavli Institute for Neuroscience, Yale University, School of Medicine, New Haven, Connecticut; and 3The Salk Institute for Biological Studies, Crick-Jacobs Center for Theoretical and Computational Biology, La Jolla, California
Submitted 19 October 2006; accepted in final form 31 March 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Steriade et al. 1993, 2001), or may appear as persistent background activity in the awake cortex that maintains the membrane potential of cortical neurons in a depolarized yet highly variable state (Chen and Fetz 2005; Crochet and Petersen 2006; Steriade et al. 2001). Studies of rhythmic cortical activity during visual processing have mainly focused on fast oscillations (Singer and Gray 1995); however, slower fluctuations in background activity across widespread regions of the cortex have been implicated in changes of functional connectivity and neural response variability to sensory stimuli (Anderson et al. 2000a; Arieli et al. 1996; Azouz and Gray 1999; J. Massimini et al. 2005). Particularly interesting is the finding that increases in background activity and responsiveness occur in primate visual areas during selective attention (Luck et al. 1997; Reynolds et al. 2000; Williford and Maunsell 2006) a finding that can be explained by an increase in synaptic bombardment of these cells (Murphy and Miller 2003). Thus study of the cellular mechanisms associated with modulation of background activity levels may be relevant not only for sensory processing, but also for complex behaviors (Raichle 2006).
The cortical slow (<1 Hz) oscillation is a readily studied form of background activity, characterized by rhythmic cycles of synaptically mediated depolarization and firing (Up states), followed by diminution of synaptic barrages, hyperpolarization, and near cessation of firing (Down states). Examination of such states has yielded valuable information about basic mechanisms of cortical function during ongoing activity (Contreras and Steriade 1995; Cowan and Wilson 1994; Haider et al. 2006; Hasenstaub et al. 2005; Pare et al. 1998; Shu et al. 2003b, 2006; Waters and Helmchen 2004). Importantly, characteristics of the membrane potential (depolarization, variability, and increased conductance) observed during Up states are similar to those in awake animals (Crochet and Petersen 2006; Steriade et al. 2001). What is the effect of spontaneous network activity on cortical sensory responses? Up state network activity in vitro strongly facilitates responsiveness of cortical neurons to local inputs (Destexhe et al. 2003; Hasenstaub et al. 2005; Ho and Destexhe 2000; McCormick et al. 2003; Shu et al. 2003a) and also to thalamic activation (MacLean et al. 2005). Similarly, studies in cat sensori-motor systems in vivo showed increased responsiveness to electrical stimulation of prethalamic pathways (Timofeev et al. 1996) and to peripheral nerve stimulation (Rosanova and Timofeev 2005) during Up states, whereas studies in the rodent somatosensory system have shown strongly diminished responsiveness of cortical neurons during Up states (Petersen et al. 2003; Sachdev et al. 2004). Thus the effects that activated network states have on sensory responsiveness remains disputed. Here, using combined extra- and intracellular recordings, we first systematically examined whether the presence (Up states) or absence (Down states) of spontaneous local network activity enhanced or diminished the neuronal response to sparse visual stimulation in cat primary visual cortex. We found that, indeed, spiking responses to visual receptive field stimulation, and spiking responses to intracellular injection of excitatory postsynaptic potential (EPSP)-like conductances are significantly enhanced when stimuli are presented during Up versus Down states, because of the depolarization associated with the Up state. Closer examination of membrane potential dynamics during network activity revealed that this response enhancement varies smoothly with gradual increases in extracellular network activity and in parallel with progressive membrane potential depolarization. Importantly, we found that these spontaneous increases in network excitability also enhanced the responses to stimuli of varying contrasts, resulting in an upward scaling of the contrast response function, similar to the contrast response enhancement observed with attention in behaving primates. These results indicate that coherent increases in the level of background network activity are transformed into depolarizations that may enhance neuronal responsiveness to a wide variety of stimuli. Such a mechanism for gain modulation may be a basic feature of flexible cortical network operations.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Acute experiments were conducted on young adult female cats (Felis catus), weighing between 2.5 and 3.5 kg. Animals were initially anesthetized intramuscularly with a ketamine (15–20 mg/kg) and xylazine (1 mg/kg) mixture, and atropine (0.5 ml, sc) was administered to reduce secretions. A forelimb vein was cannulated for continuous intravenous infusion of ketamine-xylazine dissolved in lactated dextrose Ringer solution (1.8 mg ketamine and 0.04 mg xylazine/ml solution, respectively, infusion rate 16.0 ml/h). A cuffed endotracheal tube was inserted for active ventilation, and the EKG was monitored continuously. Additionally, we performed experiments in which the animal was anesthetized only with intramuscular supplements (instead of intravenous infusion) of the above described ketamine-xylazine mixture to replicate the anesthetic regimen that we used in previous studies (Haider et al. 2006; Hasenstaub et al. 2005). We found no major difference in expression of the slow oscillation or in visual response properties between the two differing anesthetic delivery methods. We therefore performed the majority of the experiments using the continuous intravenous regimen as described above. The animal was artificially respirated (
20 cycles/min) with oxygen, and end-tidal CO2 was maintained between 3.5 and 4.5%. A bilateral pneumothorax was performed to minimize brain pulsations arising from respiration. Depth of anesthesia was assured by continuously monitoring EKG waveforms along with heart rate (maintained at 120–180 bpm), rectal body temperature (maintained at 37–39°C), and by observing reactions to noxious stimuli (toe pinch) and reflexes. The animal was placed in a stereotaxic apparatus, the ribcage was elevated, a midline scalp incision was made, and the skin, underlying fascia, and muscles were retracted. A small craniotomy (1–2 mm diam) was performed 10–11 mm posterior to the earbar zero mark and 1–2 mm lateral to the midline, directly above the area centralis representation of area 17. A cisternal drainage was performed to relieve cerebrospinal fluid pulsations. The dura was dissected and retracted, and the craniotomy was filled with warm 4% agar. Ag/AgCl ground wires were also placed near the edges of the craniotomy and embedded in agar.
On completion of the surgery, and assurance of adequate anesthetic depth by the field potential and absence of reaction to noxious stimuli, animals were paralyzed with vecuronium bromide (0.15 mg/kg induction dose, followed by continuous intravenous infusion of 0.1 mg/kg/h). The nictitating membranes were retracted, and the pupils were dilated with ophthalmic phenylephrine hydrochloride (2.5%) and atropine sulfate (1%). The eyes were focused onto a computer monitor 114 cm away using gas-permeable corrective contact lenses, and the area centralis and optic disks were located by back-projection. The location of the computer monitor was adjusted so that the center of the screen overlapped the representation of the area centralis. After the termination of experiments (12–36 h), animals were given a lethal dose of pentobarbital sodium. All procedures were approved by the Yale University Animal Care and Use Committee and conformed to the National Institutes of Health standards as recommended in National Institutes of Health Publication 94–3207.
Recording and analysis
Simultaneous intracellular and extracellular recordings were performed in area 17 within 10° of the area centralis representation. Tungsten microelectrodes (0.3–0.5 or 1–5 M
; Frederick Haer, Bowdoin, ME) were used to record extracellular multiple units (MUs) or single units (SUs), along with local field potentials (LFPs). The raw signal was differentially amplified (A-M Systems 3000, Sequim, WA) and recorded as a broadband (0.1–20 kHz) signal, as well as being sent to two filters (Krohn-Hite, Brockton, MA) set at 300 Hz to 20 kHz and 0.1–100 Hz, for MU and LFP signals, respectively. The electrode was lowered until either robust MU activity was detected or a well-isolated SU was encountered. The analog action potential (AP) waveforms were converted into a discrete SU spike channel with an on-line event detector. SU isolation was unambiguous, with signal amplitude routinely 5–10 times greater than the MU background. Extracellularly recorded single units were classified as fast spiking (FS) or regular spiking (RS) based on previously published criteria (Hasenstaub et al. 2005; McCormick et al. 1985). Intracellular electrodes were positioned to enter the cortex 200–500 µm away from the extracellular electrode. Intracellular electrodes were filled with 2 M potassium acetate and beveled to have a final impedance of 55–110 M. Current-clamp recordings were performed with an AxoClamp 2B amplifier (Molecular Devices, Sunnyvale, CA). Extracellular broadband, LFP, MU, membrane potential (Vm), and current (I) were sampled at 20, 1, 10, 20, and 20 kHz, respectively, and recorded using the Spike2 System (Cambridge Electronics Design, Cambridge, UK). Intracellularly recorded cells were classified as RS, FS, intrinsically bursting (IB), or chattering (CH), according to standard criteria (McCormick et al. 1985; Nowak et al. 2003). Recorded cells had to display robust Up states (
10 mV in amplitude) and had to exhibit stable (Up state negative to –50 mV) membrane potentials with minimal (0 to –0.1 nA) current injection. If a neuron failed these criteria, it was excluded from subsequent analysis. Every intracellularly recorded cell exhibited Up and Down states. When these oscillations were less robust or asynchronous in the intracellular recording, it was invariably also the case in the local MU and LFP network activity. This indicates that changes in the oscillations were likely related to the anesthetic state of the animal and not due solely to recording from a particular cell type, as has been previously suggested (Anderson et al. 2000a). Spiking responses to visual stimulation were recorded with zero current injection, whereas synaptic responses were recorded by hyperpolarizing the Up state membrane potential to near –75 mV (reversal for GABAergic Cl–-mediated inhibition) with negative DC injection. At this membrane potential, synaptic potentials should be dominated by glutamatergic EPSPs.
Dynamic clamp methods
Our dynamic clamp methods have been published previously (McCormick et al. 2003; Shu et al. 2003a). Briefly, using a real-time Linux system (Dorval et al. 2001) and a DAP-5216a board (Microstar Laboratory, Bellevue, WA), we injected artificial EPSPs at 10 Hz, in discontinuous current-clamp mode (to minimize any errors associated with series resistance compensation) with a minimum switching frequency of 2 kHz. During the protocols, the headstage output was continuously monitored with an oscilloscope to ensure adequate settling of the electrode voltage between samples. Injected conductances varied in size, from 2 to 80 nS, in steps of 2 nS. The amplitude time-courses of the artificial EPSPs were defined by a kinetic model of synaptic transmission, using only AMPA receptor kinetics (Destexhe et al. 2001).
Visual stimulation and analysis
Detailed descriptions of the visual stimulation protocol have been reported previously (Nowak et al. 2005; Sanchez-Vives et al. 2000). Briefly, size, orientation, and direction preference of spiking responses for either intracellular or extracellular units were first determined by hand-mapping, after which a series of automated routines (VSG Series 3, Cambridge Research Systems) displayed on a 19-in color monitor (80 Hz noninterlaced refresh rate, 1,024 x 768 resolution; Sony, Tokyo, Japan) were used to quantitatively determine the neuron's optimal stimulus. We used a sparse noise stimulation technique similar to that developed by others (DeAngelis et al. 1993, 1994; Jones and Palmer 1987), and used previously in our laboratory (Nowak et al. 2005) to explore the role that ongoing network activity has on responses to stimulation within the receptive field (RF). A single bright (or dark) bar of 80% contrast and optimal direction, length, and spatial frequency was flashed once every 50 or 62.5 ms (stimulus duration, 50 ms) in 1 of 16 different randomly selected locations within the RF along a single axis that was perpendicular to the preferred orientation of the neuron. These single bright or dark bars, varying randomly in position, were presented in 2-s blocks that were followed by 2 s of uniform gray screen to record spontaneous activity in the absence of stimulation (Fig. 1).1
|
). We were assured of the proximity of our extracellular and intracellular recordings by the close relationship between the level of network activity and the intracellular membrane potential (see Fig. 6) and also by the observation that both recordings exhibited similar orientation and direction preference (data not shown). Bar responses, sorted by state, were used to construct separate three-dimensional space-time maps (Figs. 1, B–E, S1, and S2)1 in a manner similar to that described above for the on-line analysis. Each space-time plot shows 1) time on the x-axis, 2) spatial location on the y-axis, and 3) response amplitude (either spikes or membrane potentials) as color-coded values in the z-axis. For display purposes, space-time plots were smoothed with a normalized Gaussian filter (width, 3.6 ms; e.g., Fig. 1, B–E). RF responses elicited by bright or dark bars presented during Down states were compared with the responses generated by these same stimuli that were presented during Up states. Cells were classified as simple and complex based primarily on the space-time plots. Cells exhibiting either a single dark or single bright excitatory subfield (S1 cells) or multiple (usually 2–4) bright and dark excitatory subfields that are separated in space (DeAngelis et al. 1993; Field and Tolhurst 1986; Heggelund 1986; Hubel and Wiesel 1962; Jones and Palmer 1987; Movshon et al. 1978a,b) were classified as simple, whereas cells exhibiting bright and dark excitatory subfields overlapping in both space and time were classified as complex. In a subset of recordings, we also presented cells with optimally oriented sine-wave drifting gratings (data not shown) and found that categorization of both simple and complex cell types based on the space-time plots was consistent with classification based on the ratio of the modulated to maintained (F1/F0) portions of the spike rate response to drifting gratings (Nowak et al. 2005; Skottun et al. 1991). For a subset of experiments, we also presented small stationary gratings of the optimum phase, orientation, and spatial frequency, restricted in space to the center of the receptive field, as determined by the above one-dimensional mapping protocol. These grating patch stimuli were on screen for 85 ms with a 200-ms interstimulus interval, whereas the contrast for each stimulus was randomly assigned to be either 0, 5, 10, 20, 40, or 80%.
|
|
|
|
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
To study the dynamics of sensory responses in active cortical networks, we employed simultaneous extracellular and intracellular recordings in the primary visual cortex of ketamine-xylazine anesthetized cats while flashing optimal bright and dark bars. Ketamine-xylazine anesthesia produces robust, rhythmic slow oscillations, as shown in Fig. 1A. The recording configuration shows the LFP and the MU activity recorded from the same electrode. The Up state (solid horizontal lines) is characterized by local network MU firing and low-voltage, high-frequency fluctuations in the LFP. This period of activity is followed by a rapid transition into the Down state (dashed lines) where network firing ceases and is associated with a large, positive-going deflection of the LFP. The top vertical lines in Fig. 1A indicate the times of flashed bright and dark bars. Notice that the structure, frequency, and occurrence of Up and Down states are not obviously altered by presentation of visual stimuli.
We quantified the responsiveness of the network to visual simulation by constructing two-dimensional space-time maps from the MU responses to flashed bars that were presented during Down and Up states (n = 5 MU recordings). Both the structure (mean Up duration, 721.6 ± 372.0 ms; mean Down duration, 335.9 ± 157.4 ms) and the frequency (mean frequency, 0.39 ± 0.12 Hz) of network slow oscillations we recorded in cat area 17 are in agreement with previous reports in cat association cortex (Steriade et al. 1993), and broadly agree with a variety of other preparations (Anderson et al. 2000a; Cossart et al. 2003; Petersen et al. 2003; Sachdev et al. 2004; Timofeev et al. 2000; Waters and Helmchen 2004), including what we have observed both in vitro and in vivo in ferret prefrontal and visual cortical areas (Haider et al. 2006; Hasenstaub et al. 2005; Sanchez-Vives and McCormick 2000; Shu et al. 2003b). Throughout our study, visual cortical cells typically responded to the onset, and sometimes the offset, of the bright and dark bars. This was especially evident in the MU records (Fig. 1, C and E), which reflect the ensemble activity of cells with differing RF properties. We focused our analysis on the shortest-latency response to the onset of the bright or dark bar, particularly because the offset response to bars presented during the Down state often occurred during the spontaneous transition to the Up state and therefore should not be considered as a response that occurs during the Down state.
We observed that when bright and dark bars were presented during the Up state, the MU response to the onset of the bar was enhanced compared with when these same bars were presented during Down states. Space-time plots constructed from MU spikes show that both bright bar responses (Fig. 1, B and C; MU peak in Up state, 166.8 Hz, peak in Down state, 84.1 Hz) and dark bar responses (Fig. 1, D and E; MU peak in Up state, 183.6 Hz, peak in Down state, 84.0 Hz) exhibit enhancement during the Up state. Similar results were obtained in all MU recordings (n = 5).
Because of the relatively short duration of the Down state (e.g., average of 336 ms for MU recording in Fig. 1), space-time plots constructed from visual stimuli presented during this state exhibit a general increase in activity over time, especially during the second half of the period analyzed. This nonspatially localized increase in activity is caused by an increase in the probability of spontaneously entering into the Up state (dashed lines above space-time plots in Fig. 1, B and D). We found that in MU recordings, the network is capable of responding to the visual stimulus in both the Up and Down states, although the peak response rate to the optimal stimulus (best bar) is approximately doubled when stimuli are presented during Up states. Figure 1, F and G, shows the response amplitude as a function of time since stimulus presentation, where each of the curves represents the MU response to bright or dark bars in a particular spatial location (corresponding to the horizontal dashed lines on the space-time plots in Fig. 1, B and C). Because the response profile is constructed from multiple neurons with differing receptive field properties (i.e., increases or decreases in firing in response to the same bright or dark stimulus), the overall response profile exhibits phasic increases and decreases at the frequency of stimulus presentation. This in no way affects the main observation that the peak response to the best bar is significantly elevated during Up states. Furthermore, the response enhancement to bars presented in the Up state is not limited to the best bar in the optimal location but also applies to the suboptimal bars presented in adjacent spatial locations within the excitatory subfield (Fig. 1, F and G).
Up states enhance visual responses in extracellularly recorded RS and FS neurons
We next examined whether well-isolated SU recordings also exhibit response enhancement when visual stimuli are presented during Up states. Figure 2 shows representative examples of electrophysiologically identified RS (presumably pyramidal) and FS (presumably inhibitory interneuron) cell responses to bars presented during both Up and Down states. The RS cell responses to the optimal bar are clearly enhanced by network activity (peak in Up state, 15.6 Hz; Fig. 2B) compared with when stimulated during the Down state (peak in Down state, 1.2 Hz; Fig. 2A). As observed with MU activity, this response enhancement during the Up state is not limited to the best bar in the optimal location (Fig. 2B, arrow), but extends to multiple locations across the RF. Similar response enhancement was seen in the population of RS cells (n = 10). FS cells also show markedly enhanced responses to the optimal bar presented during the Up state (peak in Up state, 84.9 Hz; Fig. 2D) compared with responses to optimal bar presentation during Down states (peak in Down state, 23.6 Hz; Fig. 2C). As we have observed in frontal cortex (Hasenstaub et al. 2005), FS neurons typically exhibit a high spontaneous discharge rate during the Up state, which we also observe here in visual cortex (Fig. 2D). Similar response enhancement was seen in all identified FS cells (n = 3)
Up states more than double the population spiking response to visual stimulation
We next compared the average response elicited by presentation of the optimal bar (either bright or dark) in the best location, to the average response elicited to bars in "null" locations (locations that did not exhibit a significant change in activity; see METHODS) for the population of extracellular SU recordings (n = 14 neurons; 3 FS, 10 RS, 1 unclassified; 12 simple (6 S1), 2 complex). On the population level, the neurons were responsive in both the Up and Down states, and as expected, the peak responses to the optimal location (Fig. 3, A and B, solid red lines) were significantly greater than the responses in null locations (Fig. 3, A and B, solid blue lines; P < 0.01, paired Wilcoxon sign rank test). However, as with MU responses, the SU responses were on average significantly greater during the Up state (13.1 ± 10.3 Hz; Fig. 3A, solid red line), than during the Down state (5.0 ± 6.2 Hz; Fig. 3B, solid red line; P < 0.01, paired Wilcoxon sign rank test).
One difficulty with our analysis is that the Down states are relatively short, and for at least some visual stimuli presenting during the Down state, the visual cortical response occurred during the transition into the subsequent Up state, which may have artificially inflated our Down state responses. Likewise, Up state responses that occur during the transition to the Down state may be altered by this change in network activity (Leger et al. 2005). To examine these possibilities, we pooled those SU and MU recordings that exhibited any action potential responsiveness during the Down state and performed a similar analysis to that shown in Fig. 3, A and B, with the restriction that the action potential response for each neuron had to occur entirely during the Up or Down state (Fig. S3B). This analysis revealed that the magnitude of the Up state response was nearly identical, whether the criterion for analysis was the stimulus or the response occurring in the Up state. In contrast, for data collected during the Down state, responses that were entirely confined to the Down state were significantly smaller than those that were allowed to contain periods of transition to the Up state (Fig. S3). These results indicate that, in our experiments, the magnitude of Down state responses is likely an upper bound and add further support to our hypothesis that the presence of network activity significantly enhances visual responsiveness.
Flashed bar stimulation does not evoke transitions into or out of Up states
It is conceivable that visual stimulation with flashed bright and dark bars affected the patterns of ongoing activity by changing the probability of Up/Down state transitions compared with spontaneous transitions occurring in the absence of stimulation. For example, if presentation of the best bar increased the probability of starting an Up state, one would expect the best bar triggered histogram to display a sustained elevation of firing. That is, the response profile after best bar stimulation in the Down state (e.g., trace with arrow in Fig. 1F) would be followed by a sustained increase in firing that was greater than that occurring spontaneously at null bar locations. Note that in both the MU and SU space-time plots constructed from Down state stimulation, we do not observe a sustained increase in firing after best bar stimulation (i.e., Figs. 1F and 2B). However, it is possible that transitions were occurring without being visible with this analysis. Therefore to further examine the possibility that bar presentations caused transitions between Up and Down states, we first calculated the rate of state transitions before and after visual stimulation and compared the transition rate after best bar stimulation to the transition rate after bar presentation in null locations. Bar presentation in null locations not only provides an estimate of the average spontaneous spiking activity (Fig. 3, A and B, solid blue lines) but also provides an estimate of the spontaneous transitions between states (either Down to Up or Up to Down) occurring in the local network (Fig. 3, C and D, solid blue lines). We thus examined whether presentation of the optimal stimulus (Fig. 3, C and D, solid red lines) could affect the probability of starting or ending an Up state in the local network. We found no significant difference in the probability of generating an Up state in the local network after presentation of the optimal bar in the best location compared with the spontaneous transition probability after bar presentation at the null spatial locations (Fig. 3D; >4,000 best bar presentations in Down state, P > 0.05, paired Wilcoxon sign rank test; n = 12 SU and n = 2 MU). Similarly, Fig. 3C shows that the probability of ending the Up state after presentation of the best bar is not significantly different from the average probability of Up state ends after presentation of bars at the null locations (>4,000 bar presentations during Up state, P > 0.05, paired Wilcoxon sign rank test).
Given the rapid delivery of stimuli in our mapping protocol (see METHODS), it is conceivable that the probability of stimulus evoked transitions was affected by multiple sequential stimuli. To control for this possibility, we also presented the single best bright or dark bar in the best location at long interstimulus intervals (1.5 s), while keeping the stimulus duration (50 ms) the same as in the mapping protocols. Again, we found no significant effect of single flashed bars on the probability of evoking or terminating Up states in the local network (n = 3 neurons, 500 best bar presentations, P > 0.05, paired Wilcoxon sign rank test for both probability of ending the Up state after presentation of the best bar and for probability of starting the Up state after presentation of the best bar compared with spontaneous transition rates; data not shown). Importantly, because we recorded and segregated the local network activity into Up or Down states using the MU/LFP from the same electrode as that used to record visual responses of single cells (see METHODS), these findings strongly suggest that presentation of flashed bright and dark bars within the RF does not reliably evoke Up or Down state transitions in local cortical networks.
Up states enhance visually evoked intracellular spikes in simple and complex cells
Having established that extracellularly recorded responses to visual stimuli are enhanced by ongoing network activity, we next examined the effect that network activity has on intracellular spiking responses and synaptic potentials resulting from visual stimulation. We recorded local network activity with LFP and MU recordings (Fig. 4A, top and bottom traces) simultaneously with intracellular recordings from nearby cells (Vm; Fig. 4A). As previously described (Contreras and Steriade 1995; Cowan and Wilson 1994; Lampl et al. 1999; Metherate and Ashe 1993; Steriade et al. 1993), when the local network is silent (Down state, dashed lines below LFP trace), the intracellular membrane potential is markedly hyperpolarized, whereas during increased levels of network firing during Up states (solid line underneath LFP trace), the membrane potential of nearby neurons becomes depolarized to levels near firing threshold and is highly variable. The values that we measured here in cat primary visual cortex (mean depolarization of Up state, 14.8 ± 7.2 mV; SD of membrane potential during Up state, 2.4 ± 0.7 mV; SD of membrane potential during Down state, 0.6 ± 0.3 mV; average maximal range from Downmin to Upmax, 28.4 ± 7.2 mV; n = 7 neurons recorded in absence of visual stimulation and hyperpolarized to prevent APs) are similar to those measured in prefrontal cortical neurons (Haider et al. 2006; Shu et al. 2003b). Figure 4A is a representative example of an electrophysiologically identified RS simple cell (see Fig. S1 for space-time plot). The presentation of the best bright bar (Fig. 4A, arrow above tick mark at top) during the Down state (Fig. 4A, dashed line) results in subthreshold PSPs (Fig. 4A, arrow). These Down-stated evoked PSPs do not reach spike threshold, and the MU and LFP traces show very little activity. Presentation of this same optimal bright bar during the Up state (Fig. 4A, 2nd arrow at top) results in a PSP that reaches threshold and produces an AP (Fig. 4A, Vm trace, 2nd arrow).
Figure 4B shows a representative example of another electrophysiologically identified RS cell, but with a complex RF (Fig. S2). As with simple cells, the membrane potential of complex cells (n = 4) depolarized and hyperpolarized with local Up and Down states (Fig. 4B, LFP and MU traces). Similar to simple cells, the presentation of the optimal stimulus (dark bar in this neuron, 1st arrow at top, Fig. 4B) leads to subthreshold PSPs during the Down state (1st arrow, Vm trace at bottom), whereas presentation of this same dark bar during the Up state leads to two APs (2nd arrow, Vm trace at bottom).
Intracellularly recorded population spiking responses to visual stimulation are significantly enhanced by Up states
On the population level, the average intracellular spiking response to the onset of a bar in the optimal location was significantly enhanced during the Up state (36.3 ± 9.1 Hz; Fig. S3A, solid red line) in comparison with the Down state [Fig. S3A, solid blue line; 14.6 ± 3.7 Hz, P < 0.01, paired Wilcoxon sign rank test; n = 16 neurons; 2 FS, 1 CH, 12 RS, 1 unclassified; 12 simple (4 S1), 4 complex]. We next compared the visually evoked change in firing rate, relative to the ongoing background activity, for the entire population (both extracellular and intracellular) of recorded neurons. As previously shown (Fig. 3, C and D), we were confident that the presentation of the optimal bar was not inducing transitions in the local network, and therefore used the activity levels present during the presentation of bars in null locations as estimates of the background activity present during the Up state (Fig. S3A, solid orange line) and also as an estimate of the spontaneous increase in firing caused by transition from the Down state into the Up state (Fig. S3A, solid light blue line). Thus for every neuron that we recorded, we subtracted the activity present during presentation of bars at null locations from the activity evoked by the best bar. The resulting plots of spiking responses take into account the background activity and state transitions occurring spontaneously over time, both during Up and Down states (Fig. 5A, baseline activity near 0 Hz for both Up and Down states, solid red and blue traces, respectively). For the entire population (combined extracellular and intracellular, n = 30), the relative magnitude of the onset (1st peak) spike response evoked by the best bar presented during the Up state (17.5 ± 4.3 Hz, Fig. 5A, solid red line) was more than double the relative spike response evoked by best bar presentations during the Down state (7.1 ± 2.1 Hz; Fig. 5A, solid blue line; P < 0.01, Wilcoxon sign rank test). Note that the peak population firing rate change evoked from the Down state occurs slightly later than the peak population response evoked by the same best bar presentations during the Up state. It is possible that elevated spiking responses occurred during Down state presentations, but with more temporal variability than the analogous responses evoked during Up states (resulting in a lower peak in the average Down state histogram response). To control for this, we examined the individual raw peak firing rates evoked by best bar stimulation in both Up and Down states for each neuron for the entire time period from the first significant deviation from baseline to return to baseline (i.e., from 30 ms to 125 ms in Fig. 5A). The average peak rate evoked by best bar stimulation during the Up state is significantly greater than the average peak rate evoked by stimulation during the Down state at any time-point during the initial visual response (Fig. 5B, mean of individual differences, solid dot, bars indicating 95% CIs; Kruskal-Wallis nonparametric ANOVA, P < 0.01). Examination of the individual times of this firing rate peak for the population revealed that the peak response to Up state best bar stimulation occurred significantly earlier (mean difference, 11.8 ms) than the corresponding peak response to Down state best bar stimulation (Fig. 5C, same convention as Fig. 5B; Kruskal-Wallis nonparametric ANOVA, P < 0.01).
Previous studies have shown that increases in the rate of change of the membrane potential immediately preceding an AP is associated with decreases in spike threshold (Azouz and Gray 1999, 2003). We made a similar finding here. Examination of the spike-triggered average of the membrane potential for the best stimulus evoked spikes showed that spike threshold exhibits significant variability and is well correlated with the rate of change of the membrane potential (dv/dt) before spike onset (Shu et al. 2006). For a window of 10 ms before spike onset, threshold was lowest (about –55 mV) for those spikes that were preceded by high values of dv/dt (1 mV/ms; data not shown). Taken together, these population results strongly support the hypothesis that synaptic bombardment during Up states is associated with more robust and more rapid spiking responses of cortical neurons to visual stimulation of the receptive field.
Visually evoked PSPs do not differ in amplitude in Up versus Down states, but Up state PSPs bring Vm closer to spike threshold
Having established that the spiking output of neurons is enhanced by network activity during Up states, we next examined whether network activity causes a systematic difference in the synaptic potentials as recorded in or near the soma. For a subset of the neurons from which we recorded intracellular spiking responses, we also examined the synaptic potentials that are evoked by this same optimal visual stimulation. For these protocols, we hyperpolarized neurons with the intracellular injection of DC, so that the peak Up state membrane potential was between –85 and –75 mV. This effectively prevented generation of APs by the visual stimuli. Note that robust PSPs are evoked in this simple cell (same as shown in Fig. 4A) by bar presentation in either the Up (Fig. 6B) or Down state (Fig. 6A). The peak of the visually evoked response in the Up state (Fig. 6B, arrow, peak bright bar response during Up state, –77.5 mV), however, is significantly more depolarized than the PSP peak evoked in the Down state (Fig. 6A, arrow, peak bright bar response during Down state, –89.4 mV; P < 0.01, paired Wilcoxon sign rank test).
We also examined the synaptic potentials arriving because of optimal bar presentation in complex cells (Fig. 6, C and D; same cell as shown in Fig. 4B). Again, note that robust PSPs are evoked from visual stimulation occurring in either the Up or Down state. Similar to the previous example simple cell, the peak depolarization achieved by visual stimulation during the Up state in the complex cell (Fig. 6D, arrow, peak dark bar response in Up state, –64.4 mV) is significantly greater than the peak depolarization to the same visual stimulus presented during the Down state (Fig. 6C, arrow, peak dark bar response in Down state, –76.2 mV), even though for this particular cell, the amplitude of the evoked PSP is slightly larger in the Down versus the Up state.
We examined the effects of Up and Down states on evoked PSP responses more thoroughly on the population level (Fig. 7). We found that, in our sample of cells [n = 10; 1 FS, 9 RS; 6 simple (1 S1), 4 complex], optimal bar presentation evoked a response in the Down state that was 7.4 ± 9.5 mV with respect to the membrane potential at the onset of the PSP (Fig. 7B), which seems to be larger than that evoked by the visual stimulus during the Up state (3.8 ± 5.0 mV; Fig. 7A). However, a significant component of the Down state–evoked response is intermixed with the depolarization of the membrane potential associated with spontaneous transition to the Up state. To control for this Down-to-Up transition, we performed two manipulations. First, we subtracted null location membrane potential responses from best bar locations and found no significant difference in magnitude or latency of the evoked synaptic potentials between Up and Down states (Fig. 7C; P > 0.05, paired Wilcoxon sign rank test). However, on average, bar presentations in null locations (see METHODS) during the Up state resulted in a small (1.8 ± 0.5 mV) hyperpolarization, presumably because of surround inhibition that was difficult to detect at any particular bar location. Subtracting this hyperpolarization increases the apparent magnitude of the absolute evoked synaptic potential in Fig. 7C. Additionally, the population plots of membrane potential versus time since stimulus onset (Fig. 7, A ad B) combine the responses of neurons with varying latencies to peak membrane potential value.
To exclude the effect of the hyperpolarization in the null locations, we linearly interpolated the membrane potential of null location trials for the times spanning the initial best location visual response (25–150 ms). This procedure preserved the overall decrease in membrane potential over time during the Up state (Fig. 7A, dashed line with endpoints) and also preserved the increase in membrane potential depolarization for these same time-points during the spontaneous transition to the Up state (Fig. 7B, dashed line with endpoints). We compared the peak magnitude of the optimal bar evoked synaptic potentials at any time during the initial visual response (25–125 ms) for each neuron relative to the interpolated membrane potential value in the null location at the time of the stimulus evoked peak (Fig. 7D). Again, we found no significant difference in the population average for the peak amplitude of the evoked PSP in the Up state (5.7 ± 4.8 mV) compared with the Down state (6.1 ± 6.2 mV; Fig. 7D, bottom; P > 0.05, paired Wilcoxon sign rank test), and there was also no difference in the latency of onset or time to peak for the population PSPs (P > 0.05, paired Wilcoxon sign rank test).
Finally, by examining the level of depolarization evoked by visual stimulation in our population, we found that the peak membrane potential value of synaptic potentials evoked by the optimal bar during the Up state was significantly more depolarized than the same peak depolarization achieved from the Down state (Fig. 7E, bottom; Up state mean membrane potential value, –73.7 ± 9.1 mV; Down state mean membrane potential value, –81.6 ± 8.5 mV; P < 0.01, paired Wilcoxon sign rank test; hyperpolarizing DC was injected to prevent AP generation). These results indicate that, although the magnitude of the PSPs evoked by optimal visual stimulation during the Up and Down states was not significantly different, the peak depolarization achieved by these visually evoked synaptic potentials is greater during the Up state and significantly closer to the threshold for action potential generation.
Responses to artificial conductance stimuli are significantly enhanced by Up states
Our results suggest that the enhancement of AP visual responses may occur through increases in neuronal excitability, as opposed to increases in the amplitude of synaptic responses. To test this hypothesis, we explored the input-output relationship of neurons by using the dynamic clamp technique to precisely control the conductance input into neurons, while the activity in the local network continued to oscillate between Up and Down states (n = 8 RS neurons). For these experiments, we recorded local network activity (Fig. 8A, LFP and MU, top traces), while simultaneously intracellularly injecting artificial conductance stimuli of various amplitudes (Fig. 8A, Gdyn) that mimicked single (AMPA receptor mediated) EPSPs into nearby neurons (Fig. 8A, Iinj), and recorded the ensuing APs (Fig. 8A, Vm, bottom trace). Note that when the local network is silent and the membrane potential is hyperpolarized, a small amplitude artificial conductance stimulus results in an injected current that produces an 10 mV depolarization that does not result in APs (Fig. 8A; 1st arrow), whereas the same conductance input presented during the Up state results in a slightly smaller injected current (caused by decreased driving force) but leads to a depolarization that does result in APs (2nd arrow). We quantified the probability of generating a spike in response to these varying amplitude conductance stimuli and found that the probability of an EPSP-like input initiating a spike was greatly enhanced in this neuron when the stimuli were injected during Up versus Down states (Fig. 8B, cf. red and blue traces). In addition, the Up state also caused a decrease in slope of the input-output curve (Fig. 8B). For this neuron, both the latency (Fig. 8C, top) and the jitter (SD of latency; Fig. 8C, bottom) of spikes were reduced in response to conductance stimuli injected during Up states (red trace) compared with Down states (blue trace; paired Wilcoxon sign rank test, P < 0.01).
|
; Shu et al. 2003a). Depolarization induced by network activity enhances responses to both artificial conductances and to visual stimulation in a smoothly graded manner
We next explored the enhancement of cellular responsiveness caused by network activity by examining the probability of evoking spikes to a medium-sized artificial PSP as function of membrane potential. Rather than segregating each occurrence of this single PSP into Up or Down state trials, instead, the number of spikes per trial was examined as a function of the membrane potential level immediately preceding the injection of an artificial EPSP (n = 4; Fig. 9A). As can be seen in Fig. 9A, the probability of a given sized PSP evoking an AP is a smooth function of the membrane potential just before injection of the artificial EPSP. These data were well fit by linear regression (dashed lines), and across the population of neurons (n = 7), there was a highly significant correlation (r = 0.94 ± 0.05, 
2, P < 0.05 for each neuron) between membrane potential and probability of the EPSP evoking an AP. A similar analysis examining the number of spikes per trial as a function of membrane potential was performed for flashed, optimal bars. Again, increased depolarization gradually increased the number of spikes evoked by the flashed, optimal bar (Fig. 9B). This was especially true for the range of membrane potential values that occurred in each neuron during the Up state (membrane potential histograms for each neuron in Fig. 9C). The gradual increase in spike output to the best stimulus as a function of membrane potential was evident for a representative sample of neurons (n = 5), and the average correlation coefficient to a linear regression best fit line was highly significant (r = 0.90 ± 0.07; 2, P < 0.05 for each neuron). Taken together, these results showed that increases in depolarization of the membrane potential are highly correlated with an increased probability of generating spikes to the same stimulus.
|
Finally, we determined whether the ongoing fluctuations seen in the network during the Up state correlate with increased neural responsiveness to optimal stimuli of different contrasts. For these experiments, a stationary grating patch of the best orientation (optimized for spatial frequency and phase) was randomly varied in contrast, and the responses to each contrast were plotted as a function of both the membrane potential of the recorded neuron and the level of network activity at the time of visual stimulation. Data obtained from a representative cell show the basic findings. Figure 10A shows a chattering (CH) neuron in which these varying contrast stimuli were presented during ongoing network activity of the activated (Up) state. The probability of generating spikes to a 40% contrast stimulus is markedly affected by membrane potential level at the time of visual stimulation, with a burst of spikes generated when the membrane potential was at –63 mV (1st arrow), whereas this same stimulus results in a subthreshold PSP and no spikes when the neuron is spontaneously less depolarized about 200 ms later (–68 mV, 2nd arrow). Shortly afterward, the neuron again becomes spontaneously depolarized to –63 mV, and a high-contrast stimulus (80%, 3rd arrow) results in more spikes than the 40% contrast stimulus delivered previously at the same membrane potential. Note that we included only responses that occurred during the body of the Up state, and therefore our visual responses do not represent the initiation or termination of this state of recurrent network activity. During the activated (Up) state, fluctuations in membrane potential are significantly correlated with fluctuations in the activity of the local network, as measured simultaneously with nearby LFP recording (Fig. 10A, red trace at top). During the body of the Up states, the LFP and membrane potential fluctuations were significantly correlated, and the relationship between these two variables was well described by a linear regression (Fig. 10B; r = –0.61; 2, P < 0.01; 30 sequential Up states).
|
2, P < 0.01 for both fits using a modified Hill equation). These results showed that a greater level of depolarization during network-activated states significantly scales upward the entire contrast response function at all contrast levels (from Rmax = 0.60 ± 0.2 to Rmax = 0.91 ± 0.15; P < 0.01, t-test) without a statistically significant change in the contrast needed to produce a half-maximal response (hyperpolarized: C50: 19.28 ± 2.24; depolarized C50: 16.86 ± 10.96; P > 0.05, t-test). |
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
