| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Department of Cell and Developmental Biology, 2Institutes of Cognitive, Integrative and Molecular Neuroscience and 3Department of Psychology, Vanderbilt University, Nashville, Tennessee; and 4Department of Physiology, University of Szeged, Szeged, Hungary
Submitted 4 January 2006; accepted in final form 22 February 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
100-ms quiescent interval. The incidence of bursts was examined in several intensive-training Go-NoGo and target selection tasks as well as in training-free tasks where natural scenes with both familiar and novel contents were shown. Our chief findings were as follows. 1) Bursts occur in the majority of cells under every condition tested, 2) burst incidence is very low (<1 burst every 10 s), 3) bursts occur in association with a receptive field stimulus on average only once every 23 times in 65% of cells tested, 4) cells responding with bursts to the stimulus also tended to exhibit higher levels of spontaneous bursting, 5) the presence of bursts did not depend on the novelty of the stimulus or its behavioral relevance. When the monkeys explored static natural scenes, 6) bursts were not correlated with short-term changes in the image sampled by the cell's receptive field during saccades. Burst incidence 7) did not increase when images were novel or when they evoked an emotional reaction, and 8) bursts did not decrease when images were familiar. 9) Bursts were not correlated with saccades in the dark, but 10) more spikes participated in bursts in the dark. Although these results confirm the occurrence of LT bursts in LGN cells of awake monkeys, they do not support the hypothesis that these bursts are a privileged means of transferring sensory information, that they signal unexpected or significant visual events, or that they are involved uniquely in the coding of natural scenes. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Guillery and Sherman 2002).
Thalamic cell firing patterns vary with an animal's level of consciousness. These changes in firing have been proposed to result from both the interconnectivity of thalamus with other areas of the brain and the intrinsic properties of the thalamic cells themselves (Llinas 1988; McCormick and Bal 1997; Steriade and Llinas 1988; Steriade et al. 1993). One of the ionic conductances implicated in these different state-dependent firing modes is the low-threshold calcium (Ca2+), or IT, current (McCormick and Pape 1990). In vitro studies have shown that this current is activated by depolarizing the cell membrane after a long period of hyperpolarization (approximately 100 ms). Opening of IT Ca2+channels results in the production of clusters of high-frequency Na+/K+ action potentials (heretofore, "LT bursts"). A sustained depolarization would change the cell's response to tonic mode in which the cell responds with streams of single action potentials (McCormick and Pape 1990). Depending on the activation status of IT, LGN relay cells would then respond in burst or tonic modes to ongoing visual events (Denning and Reinagel 2005; Sherman 1996, 2001). Extrapolating from studies in anesthetized, paralyzed preparations, the two firing modes have been proposed to potentially code different aspects of visual information in awake animals. In tonic mode, the spike train would faithfully encode stimulus features, whereas in burst mode, spikes would be used to detect the presence of a new stimulus in the visual field. Thus bursts could be used as a "wake-up " call to visual cortex switching the thalamus out of the burst mode and into the tonic mode (Guido and Weyand 1995; McCormick and von Krosigk 1992; Scharfman et al. 1990).
The hypothesis that LT bursts signal significant visual or other sensory information in awake animals contrasts with other observations in slow-wave sleep and in anesthetized preparations. Under the latter conditions, sensory input is greatly diminished even though LT bursts are frequent, due presumably to the increased hyperpolarization of thalamic cells. Few studies have attempted to test directly the burst/tonic firing mode hypothesis in the LGN of awake preparations, and, to our knowledge, very few have involved awake monkeys (Martinez-Conde et al. 2002; Ramcharan et al. 2000, 2005). Studies in macaque monkeys are important because their visual system employs the "primate plan" common to humans. Primates, unlike many afoveate mammals, scan their environment bringing their fovea onto a region of interest. Therefore the goal of the present study was to examine the incidence of LT bursts in the LGN of awake behaving macaque monkeys using stimuli and tasks that should elicit LT bursts assuming such bursts are relevant to visual behavior. More specifically, we tested how the incidence of LT bursts varied with different epochs and conditions in several simple visuomotor tasks, free viewing of novel and familiar natural static scenes, image structure in natural scenes or blank fields, and the presence or absence of visual input. We also examined the relationship of LT bursts to saccades and how burst incidence evolves over time during the free viewing of static images.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Three male bonnet macaque monkeys (Macaca radiata, 6.0–8.0 kg) were used in this study. Because the methods for locating the LGN are described in detail elsewhere (Royal et al. 2006), only a short summary is presented here. We used magnetic resonance imaging (MRI) maps to position a recording chamber over the LGN. All the procedures, care and training of the monkeys conformed to the guidelines of the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the guidelines of the Vanderbilt University Animal Care and Use Committee under an approved protocol.
Surgery
Attachment of the recording chamber, the head post, and eye coil were performed under sterile conditions under general anesthesia (either ketamine and xylazine or 1.5–3% isofluorane in O2 as described in detail in Royal et al. (2006). Briefly, this involved securing head posts and wells with titanium screws and dental cement, and implanting an eye coil under the conjunctiva of the right eye (Judge et al. 1980).
Recording
Monkeys were seated in a primate chair in front of a computer monitor, within a frame generating a magnetic field received by the eye-coil. The head of the monkey was fixed by means of a head post attachment to the chair. Single-unit recordings were made with Parylene-coated tungsten microelectrodes (1–3 M
; FHC). The cellular activity was amplified and band-pass filtered, and spikes were discriminated with an amplitude-and-time window discriminator (BAK Instruments, Mount Airy, MD). A PC-based system running Tempo (Reflective Computing, St. Louis, MO) was used to control experiments, present stimuli, and store analog eye-position data and digital spike-timing data. Eye-position and neuronal activity were sampled at 250 Hz and 1 kHz, respectively.
Visual stimuli, LGN receptive field mapping, and cell classification
Stimuli for the first set of experiments (intensive-training visual tasks; see following text) were presented on an otherwise dark monitor in a light-tight room. The monitor was placed 57 cm from the eyes of the monkey and subtended 36 x 29° of visual angle. The monitor refresh rate and resolution were 70 Hz and 640 x 480 pixels, respectively. A grounded conductive-glass screen (3M, model AF250XXL) covered the monitor's face to reduce electrical noise.
The monkeys were trained to fixate a single-pixel (1/17°) fixation spot (FS) presented at the center of the monitor while a flashing bar (1–4 Hz) swept the right visual hemifield to locate and map LGN cell receptive fields (RFs). The stimulus' size, speed, flash rate, and location could be adjusted in real time allowing for precise RF mapping. Four equiluminant colored squares, white (CIE x = 304, y = 325), red (CIE x = 632, y = 336), green (CIE x = 290, y = 605), and blue (CIE x = 143, y = 60), (luminance = 1.8 cd/m2) were used to assess the response type (ON or OFF) and color preference of the isolated unit. The stimulus that produced the maximum excitation or suppression over baseline response was selected and used for the remainder of the session. During subsequent trained visuomotor tasks, the stimulus was enlarged beyond the edges of the RF center to compensate for microsaccades around the FS and provide for consistent stimulation across trials. Only ON-center cells were used in this study because their characterization and stimulation were more reliable in our setup (bright localized stimuli on a dark screen in a uniformly dark room). Only units that were well isolated and showed clear responses to the appropriate stimulus were chosen for further analysis.
The monkeys are still participating in experiments, thus exact histological confirmation of recording locations was not possible. Due to this limitation, strict criteria were applied to confirm that recorded single units were LGN cells and to classify the responses as belonging to parvocellular (P), magnocellular (M), and koniocellular (K) (see Royal et al. 2006 for details). Cells that could not be classified as M, P, or K but were clearly well-isolated ON-center LGN cells were still included in our sample.
Intensive-training tasks
The monkeys were trained to perform three tasks for reward: blocked Go-NoGo, interleaved Go-NoGo, and target selection. We first describe the Go-NoGo task (Fig. 1). Each trial started with a dark monitor and a white FS at the center of the screen (baseline). During this period, the monkeys were required to maintain fixation in a 1° x 1° window. After a random period of 500 ms ±30%, the color of the FS changed to cue the monkey about the task's behavioral requirements (cue epoch). After another 500 ms ±30%, a stimulus was displayed on the screen inside the single-unit's RF center, and the monkey either had to remain fixated or make a saccade to the stimulus, depending on the color of the cue. A red FS cued the monkey to remain fixated (NoGo condition), and a green FS instructed the monkey to shift gaze to the stimulus in the RF (Go condition). In both conditions, the stimulus remained on until the reward was given (350–650 ms). In the blocked Go-NoGo task, the Go or NoGo trials were presented in blocks of 20–30 trials each until completion of two blocks per condition in an ABAB or ABBA sequence. A new trial was presented every 2–4 s (mean = 3 s). Trials were aborted if the monkey failed to respond correctly. Only successful trials were included in the analyses reported here.
|
The time between the end of the RF characterization, and the beginning of the first task was 20 s to 1 min. This was also the interval between the end of a task and the beginning of a new task. The monkey could not predict task order. Both the monitor and the room were kept dark during pauses between tasks.
Free-viewing tasks
The monkeys were recorded in two tasks where they were free to move their eyes at will (i.e., they were not required to fixate). These tasks required no training, and so recordings began as soon as each task was introduced. The only other difference in the setup, from the intensive-training tasks, was that the monitor resolution was increased to 1,024 x 768 pixels, to display high-resolution images.
In the picture task, the monkeys were presented with natural scenes and blank fields (Fig. 2). Three categories of images were shown in an interleaved sequence. The first category contained scenes that the monkeys commonly see in the animal facility, including familiar people and other monkeys. We operationally defined this condition as the "familiar" condition. The second category was made up of unfamiliar images which we operationally defined as the "novel" condition. The novel condition consisted of a variety of scenes that these cage-born monkeys should never have seen including such items as street scenes, sky with clouds, and fish. The third category was a blank field (blank condition) that was presented at one of three luminance levels in different sessions, either matched to the average luminance of the pictures (4.6 cd/m2; gray blank), set to the maximum luminance of the monitor (27 cd/m2; white blank), or set to the intra-picture black background field (below our instrument limits, <0.1 cd/m2; black blank). Each image and the blank fields were presented for 7 s. The images were presented in the same order in every session, and the monkeys were rewarded between image presentations. An infrared camera was used to observe the monkeys' reaction to the images.
|
Burst-identification and epoch lengths
Bursts were defined as two or more spikes occurring within an interval of 
4 ms and preceded by a quiescent period of 100 ms. This criterion is conservative minimizing the chance of assigning spike clusters to LT bursts when they were not part of a burst. Firing patterns defined this way were found to be LT bursts 99% of the time in the anesthetized paralyzed cat (Lu et al. 1992; Ramcharan et al. 2000). A burst was considered complete when the interval between successive spikes became >4 ms (Alitto et al. 2005). The fraction of spikes in bursts was calculated by dividing the number of spikes belonging to bursts by the total number of spikes in the same epoch or condition. The analysis of the data also was performed using two more liberal criteria (50- or 25-ms quiescent periods). All the calculations were performed with custom programs written in MATLAB (Math Works).
In the intensive training tasks, we compared the number of bursts occurring during the baseline epoch with the number occurring during the cue and stimulus epochs. Because a trial is a continuous event, we operationally defined these epochs as periods of 500 ms, excluding any trials where epochs overlapped. The baseline epoch began with fixation, the cue epoch began 100 ms before the cue onset, and the stimulus epoch began 100 ms prior to stimulus onset (Fig. 1B). To maximize the possibility of counting stimulus-elicited bursts, we also analyzed bursts by narrowing the epoch window around the onset of the stimulus to 250 ms (100 ms before to 150 ms after stimulus onset). In this case, baseline and cue epochs were shortened similarly. In the picture and FreeSacDark tasks, each of the entire 7 s trials were analyzed without dividing trials into epochs.
Statistics
Statistical comparisons were made using a Mann-Whitney U test for two independent groups and the Kruskal-Wallis test for three or more independent groups. Differences between paired measurements were assessed with the Wilcoxon signed-ranks test, and differences among three related samples were evaluated with the Friedman's test for k dependent samples. Dependences between two variables were assessed evaluating Spearman correlations (Siegel and Castellan 1998). Alpha levels of P < 0.05 were considered significant. The statistical analyses were performed with custom programs written in SAS (SAS Institute).
Relation between bursts and saccades
We tested the association between bursts and saccades by measuring eye movements before, during, and after bursts in the picture and FreeSacDark tasks. To do so, we first identified the time (b) of the cardinal spike of each burst. Then for each burst, we defined three 100-ms windows centered 100 ms before the burst, on the burst itself, and 100 ms after the burst (c = –100, 0, +100 ms, respectively). For each burst, the eye-position change within each window was calculated as
 |
t is the window width (100 ms); and x and y are the horizontal and vertical components of the eye position. To examine the effect of shorter-latency image changes on burst incidence, the same calculations were repeated with narrower time windows (t = 30 ms), positioned closer to the bursts (c = –30 ms, 0 ms, +30 ms). |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Burst incidence in trained behavioral tasks
Figure 3 shows raw data from two typical cells while monkeys performed either the Go condition (A) or the NoGo condition (B) of the interleaved Go-NoGo task. In B, the arrow indicates the only burst detected in these 111 trials. Across all of our trained behavioral tasks, a total of 1,279 bursts were detected in 7,896 trials with 773 bursts occurring before the stimulus and 506 during or after the stimulus.
|
|
2 spikes within 4 ms, after 50- or 25-ms quiescent periods; see METHODS), we found that the measured burst incidence increased monotonically with the reduction of the quiescent period. No apparent differences across conditions or epochs emerged (Royal et al. 2003). We also examined whether different classes of LGN cells (K, M, P) showed differences in burst incidence. Given the large numbers of tasks, conditions and epochs, however, it was not possible to statistically tease apart cell classes for each condition. We were, however, able to compare P and M burst incidence in the target-selection task and found no differences in the fraction of spikes in bursts across cell classes during the stimulus epoch.
The 500-ms B, C, and S epoch lengths resulted in occasional overlaps between epochs in some trials (these were discarded from the analysis). Also eye movements would have interfered with the cell response to the stimulus in all the conditions other than the NoGo condition. Therefore, we reanalyzed the data shortening the three epoch lengths from 500 to 250 ms; e.g., for the stimulus epoch, the window was set from 100 ms before to 150 ms after the stimulus onset. The baseline and cue epochs were also shortened to 250 ms. The spike rate measurements in the baseline, cue, or stimulus epochs did not change. The burst rate and the fraction of spikes in bursts in the stimulus epoch did show an increase, however, when this shorter window was used, suggesting that more bursts occurred close to the stimulus onset. None of the spike rate, burst rate, or the fraction of spikes in bursts during the stimulus epoch, however, showed significant differences attributable to the behavioral relevance of the stimulus within any given task (P values between 0.17 and 0.94; Wilcoxon tests of the Go vs. NoGo conditions of the Go-NoGo tasks and RF vs. non-RF of the target-selection task). Importantly, stimulus-associated bursts were rare events occurring on average every 23 trials and not occurring at all in 35% of cells (see also Fig. 3).
Burst rate in the stimulus epoch was positively correlated with spontaneous bursts (baseline and cue epochs combined; r = 0.43, P < 0.0001 Spearman correlation; Fig. 5). Burst rate in the stimulus epoch was negatively correlated with the spontaneous spike rate (r = –0.21, P < 0.03) and uncorrelated with the spike rate during the same stimulus epoch (P = 0.66).
|
|
|
To further test the proposal that LGN bursts signal significant events in the visual world, we evaluated the incidence of LGN bursts as the monkeys freely inspected static natural scenes. Two categories of images were shown, "familiar" and "novel, " interleaved with blank fields (see Fig. 2). An example of the recordings obtained during this task is shown in Fig. 8. In each panel, the top trace represents the eye movements and, below, one can see the activity of the recorded cell and the detected bursts (
). Bursts occurred in every cell recorded in this task (n = 32) during all the conditions (familiar, blank, and novel), but their incidence was very low (median = 0.06 and mean = 0.09 bursts/s), equivalent to less than one burst every 10 s.
|
; Hinde and Rowell 1962; Kalin et al. 1991; Kenney et al. 1979; Maestripieri and Wallen 1997). In contrast, the novel images (clouds, unknown groups of people, unfamiliar natural scenes) as well as the blank fields never elicited such displays. Therefore, even if our classification of the images into novel and familiar was imprecise, the two sets of images clearly were not treated in the same way by the monkeys. Based on behavior, one could classify the images as engaging or nonengaging. The activity of LGN cells recorded in these tasks under different conditions is summarized in Fig. 9. We found no significant differences attributable to the novel versus familiar categories either in the spike rate (P = 0.34, Wilcoxon test, 32 cells; Fig. 9A), burst rate (P = 0.66; Fig. 9B), or fraction of spikes in bursts (P = 0.50; Fig. 9C).
|
2 test), neither the rate of bursts (Fig. 9B; P = 0.051) nor the fraction of spikes in bursts (Fig. 9C; P = 0.056) showed this relation. Moreover, despite parallel changes exhibited by the spike rate and burst rate with the luminance of the blanks (Fig. 9, A and B, respectively), the fraction of spikes in bursts did not change significantly (P = 0.55; Kruskal-Wallis test).
As before, the burst incidence depended on the criterion used to define a burst and increased monotonically with shorter quiescent periods. Thus the mean incidence of bursts in the picture task was 0.09 ± 0.09 bursts/s (100-ms quiescent period followed by 2 spikes in no more than 4 ms), 0.26 ± 0.20 bursts/s (50-ms quiescent period), and 0.70 ± 0.60 bursts/s (25 ms quiescent period).
Finally, we compared the incidence of bursts in the picture task with the burst incidence during long periods of absence of visual stimulation (FreeSacDark task, lasting from 1 to 4 min). The activity of 31 LGN ON-center cells recorded in the FreeSacDark task is represented by the rightmost box and whisker plots in the three panels of Fig. 9. We found that the spike rate was lower in the FreeSacDark task than in the picture task (P < 0.0001; Mann-Whitney test; familiar, novel, and blanks combined). No difference in the burst rate between tasks (P = 0.74) was found, but a higher fraction of spikes in bursts occurred in the FreeSacDark task than during the picture tasks (P = 0.01; picture task's median = 0.53% and mean ± SD = 1.1 ± 1.7%; FreeSacDark task's median = 1.7% and mean ± SD = 1.7 ± 1.9%). Therefore more spikes participated in bursts in the LGN in a task where there was no stimulus-driven input from the retina.
When the quiescent period was 50 ms, the fraction of spikes in bursts is larger in the FreeSacDark task than during the picture task (P = 0.037; Mann-Whitney test). When the quiescent period was 25 ms, no significant difference was found between the FreeSacDark and the picture tasks (P = 0.74).
Incidence of bursts associated with saccadic eye movements
LGN cell activity can exhibit changes associated with eye movements, namely a depression before or during a saccade followed by an enhancement once the eyes land at their new location (Bartlett et al. 1976; Jeannerod and Putkonen 1971; Lee and Malpeli 1998; Martinez-Conde et al. 2002; Ramcharan et al. 2001; Reppas et al. 2002; Royal et al. 2006). If the suppression is associated with the hyperpolarization of relay cells, one would predict an increase in burst probability after a saccade. Also because natural images would be expected to contain areas that both match and do not match the RF preferences of the recorded cell, an LGN cell the RF of which moves from a nonpreferred to a preferred location in the image would be expected to suddenly increase its activity after a period of inhibition and perhaps emit a burst.
We probed the association between bursts and saccades by measuring saccadic eye movements before, during, and after bursts identified in the picture and FreeSacDark tasks (eye movements were evaluated during 3 nonoverlapping time windows centered at t = –100, 0, and +100 ms, respectively, relative to the burst, with a window width of 100 ms). The results are shown in Fig. 10. The two panels show the absolute-value changes in eye position in the preburst, during-burst, and postburst time windows for the picture task (A, calculated from 552 bursts) and the FreeSacDark task (B, from 609 bursts). In general, there were very few eye movements near a burst as reflected by the close-to-zero medians and the close-to-zero 75% quartiles in all the box-and-whisker plots of both panels. No statistical differences were found among the preburst, during-burst, and postburst eye movement periods, either in the picture task (P = 0.83; Friedman's 2 test) or the FreeSacDark task (P = 0.47).
|
2 test) or the FreeSacDark task (P = 0.55). Does the incidence of bursts change over time?
It has been proposed that LGN bursts emphasize stimuli that are surprising in a given context (Sherman 2001). All three of our monkeys spent >1 yr performing the Go-NoGo and target-selection tasks already described, so the sudden presentation of natural images on the screen should have been novel and surprising. We thus expected that the burst rate would decline as the novelty of the pictures wore off. We examined the incidence of bursts on successive sessions of the picture task (Fig. 11). While some cells had more bursts than others, no trend in the fraction of spikes in bursts was observed for the blanks (P = 0.22; Spearman correlation) (A), familiar pictures (P = 0.92; B), or novel pictures (P = 0.064; C). Therefore, if the picture repetition over sessions modified the interest of the monkeys, this was not reflected in the incidence of bursts occurring in our sample of LGN cells.
|
in the same panels), which clearly decreased over the same 7-s trial.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Bursts: stimulus detection and novelty
According to the tonic/burst hypothesis (Sherman 2001), one would predict that a LGN relay cell, potentially silenced through hyperpolarization associated with stimulus absence/removal, or by the presence of an opposite-polarity stimulus, would respond with a burst to the onset of the preferred stimulus. The burst could thus act as a "wake-up call" to the rest of the visual system, "calling attention" to an unexpected stimulus, and/or to improve the detection of a stimulus (Sherman 2001). Not surprisingly, as reported by others in the somatosensory thalamus of rabbits (Swadlow and Gusev 2001), we found that more bursts occurred in cells that exhibited lower baseline activity because this, by definition, would allow more opportunity for the long periods of hyperpolarization necessary to produce an IT burst. Given the finding that bursts are more likely to fire cortical neurons (Swadlow and Gusev 2001), the high spontaneous burst rate implies a high stimulus-independent activation of cortical neurons in burst responding cells. If bursts were to act effectively as a wake up call to changing stimulus conditions, increased stimulus-independent bursting would appear disadvantageous.
In awake cats, Weyand et al. (2001) showed that the repeated presentation of a stimulus in the RF of LGN cells produced bursts in some cases and that the incidence of these bursts declined with repeated presentation of the stimulus. Our data do not support this observation for monkeys because we failed to observe any increase in the incidence of bursts associated with the first presentation of a stimulus in a series of stimuli versus later presentations of the same stimulus. It is possible that there are major species differences between monkeys and cats in the transfer of sensory information to cortex or that other differences in the behavioral demands of the tasks could account for these differences. The latter explanation seems less likely given the large variety of behavioral situations we tested.
It could be argued that the smaller incidence of bursts during the stimulus presentation in our experiments is still consistent with a wake-up call. Our monkeys were overtrained in the Go-NoGo and target selection tasks. They knew the sequence of events and the location where the stimulus would appear. Therefore the stimulus was anything but a surprise, and the LGN cells were more likely to be operating in tonic mode. This is why we presented novel images in a way that the monkeys could not have anticipated: natural scenes appearing on a monitor where none had ever been presented before. The low incidence of bursts, however, and their lack of dependence on the relative familiarity of the images does not suggest a special role of LGN bursts in novel or surprising events.
Bursts and temporal changes in the visual input
In anesthetized cats, different patterns of full-field luminance changes produce different proportions of bursts in the LGN (Denning and Reinagel 2005). Also in anesthetized cats, slow luminance changes trigger more bursts than fast luminance changes. Therefore bursts can be considered nonlinear amplifiers of slow changes in luminance in anesthetized cats (Lesica and Stanley 2004). These observations were interpreted as suggesting that burst incidence is larger for natural scenes and for luminance changes mimicking natural luminance changes and, extrapolating to unanesthetized preparations, that bursts are potentially suited to convey information about natural stimuli in awake animals. We did not characterize, in detail, the luminance and color changes at the exact locations sampled by the LGN cell RF while the monkeys scanned natural images. We could, nevertheless, make the following predictions: eye movements occurring on a blank screen should produce little relative change in RF stimulation and eye movements when the screen displayed rich natural scenes should have produced changes given the gradation of colors and luminance shifts. Thus, if bursts are optimized for coding natural luminance changes, bursts should have been more frequent while the monkeys viewed natural scenes (familiar and novel conditions) than during the gray blank conditions. Additionally, bursts should have correlated more with saccades during the presentation of natural scenes given the higher probability of RF stimulation. Contrary to these predictions, we found that the fraction of spikes in bursts did not change significantly when the monkeys viewed blanks versus natural scenes. More important, bursts were not associated with saccadic eye movements in the picture task.
Bursts and nonvisual inputs: saccades
According to the model of IT priming by hyperpolarization, bursts could be associated with saccades not only because of luminance changes occurring when the RF is stimulated but also because of nonretinal inputs suppressing LGN cell activity during saccades (Bartlett et al. 1976; Jeannerod and Putkonen 1971; Lee and Malpeli 1998; Martinez-Conde et al. 2002; Ramcharan et al. 2001; Reppas et al. 2002; Royal et al. 2006). For the vast majority of trials in both the FreeSacDark and picture tasks, burst production did not correlate with eye-position changes. Therefore we conclude that the saccadic suppression was insufficient to prime the low-threshold Ca2+ spikes, either under conditions when visual information was present or when it was absent.
Martinez-Conde et al. (2002), using a different criterion for bursts, report a higher probability of bursts after microsaccades. Using the same burst criterion we used here, however, a number of studies of LGN activity found no correlation between eye movements and burst incidence (Bartlett et al. 1976; Jeannerod and Putkonen 1971; Lee and Malpeli 1998; Ramcharan et al. 2001; Reppas et al. 2002; Royal et al. 2006). The rate of saccades per burst reported in the cat, a burst every 5.7 saccades (Lee and Malpeli 1998), is higher than what was observed in our monkeys (a burst every 45 saccades). This large difference could be attributable in part to methodological differences in saccade identification but most likely reflects the observation that cats make less frequent eye movements than primates.
Bursts in awake, sleeping, and anesthetized animals
The incidence of bursts in our preparation was less than one burst every 10 s under a variety of visuo-behavioral conditions. The incidence of bursts in the LGN of awake macaque monkeys in the absence of visual stimulation in our study is comparable to that reported by Ramcharan et al. (2000) (fraction of spikes in bursts = 1.7% in our FreeSacDark task and 1.3% in their monkeys in the waking state without any visual stimulation). These two values are five times smaller than the burst incidence in the monkeys studied by Ramcharan et al. (2000) during slow-wave sleep (fraction of spikes in bursts = 9.8%). Therefore burst firing is considerably lower in wakefulness than during sleep as concluded by these same authors (Ramcharan et al. 2000).
The higher incidence of bursts in monkeys during sleep parallels observations in cats. Weyand et al. (2001) found a 9:1 ratio of sleep to waking burst rates. Also in cats, McCarley et al. (1983) report relative burst rates of 
4:2:1 for slow-wave sleep, desynchronized sleep, and waking, respectively. Anesthetics increase the incidence of bursts in cats. Thus the rate of bursts in awake cats (0.09 bursts/s, in the cells that burst at all) (Guido and Weyand 1995) increases to 0.45/s in pentothal-anesthetized cats (average of all the cells reported by Denning and Reinagel 2005). It may be unjustified to compare bursts rates across species, but, if done, one finds that the incidence of bursts in pentothal-anesthetized cats is 5–10 times higher than the incidence of bursts in our awake behaving monkeys; the fraction of spikes in bursts reported by Lesica and Stanley (2004) is 13 and 26% under their two different conditions of stimulation. It is noteworthy that a qualitatively similar dependence of LT bursts on animal state was found in the medial geniculate nucleus of the guinea pig where the highest proportion of bursts occurred under ketamine/xylazine anesthesia and the least number of bursts were detected when the animals were awake (Massaux et al. 2004). Therefore there appears to be agreement that bursting is rare in the waking state in most sensory relay nuclei (Ramcharan et al. 2005) and that, regardless of the specific criterion used to define a burst, the incidence of LGN bursts increases with unconsciousness in a progression that seems to go from wakefulness to desynchronized sleep, to slow-wave sleep, and finally to deep pentobarbital anesthesia. One would then have to postulate that bursts have two completely opposite roles depending on the brain's state: to participate in "sensory disconnection" during sleep and anesthetic-induced unconsciousness, and to convey wake-up signals to the cortex while awake.
Bursts in other pathways
Our results do not necessarily apply to other nuclei of the thalamus or to other species. In rabbits, for example, more bursts (defined as we have done here) are produced in the somatosensory ventrobasal nucleus in awake drowsy rabbits than we see under any conditions in the LGN of our awake monkeys (Swadlow and Gusev 2001). Also in rabbits there seems to be a clear correlation between EEG-defined inattentiveness and an increase in burst rate (Swadlow and Gusev 2001). Additionally, in the somatosensory thalamus of the rat, Fanselow et al. (2001) show that bursts, defined with a criterion different from ours, occur more often during exploratory whisking than during quiet immobility. Bursts also appear to occur more frequently in "higher-order" nuclei (such as the pulvinar) of the thalamus of awake monkeys than is observed in the LGN, indicating that bursts may have some other, as yet-to-be identified, purpose (Ramcharan et al. 2005; see also following text). Perhaps our monkeys were never sufficiently bored to lapse into a state sufficient to show the large increase in burst rate in the LGN also seen in the LGN of inattentive rabbits (Bezdudnaya et al. 2005). The point, however, is that if bursts are meaningful under waking conditions in monkey LGN, we should have seen evidence of this given the wide variety of conditions our monkeys were exposed to and the hours and hours of recording time that we sampled.
Other functions for LGN bursts?
If the bursts studied here and in the thalamus of sleeping animals are the same, and if both are a consequence of LT current activation, what might their role be? The fact remains that bursts do occur in both the awake and sleeping thalamus. So why is this the case? Despite the absence of correlation with stimulus novelty, repetition, and behavioral relevance, a few bursts were found in association with stimulus onset. Therefore it could be the case that LT bursts are not utilized in any way that reflects the direct transfer of sensory information to cortex but for some other purpose. Given that LT bursts are associated with an influx of calcium, it is possible that bursts reflect some form of plasticity (Abarbanel et al. 2003; Karmarkar and Buonomano 2002; Shouval et al. 2002), perhaps utilized in calibrating the operating range of LGN cells or in the transfer of stored information as has been proposed for "sharp wave bursts" initiated in the hippocampus during slow wave sleep (Buzsaki 1998). Although it is not clear if the latter bursts are the same as LT bursts, the localized fast spikes involved are (like LT bursts) associated with calcium influx. Buzsaki (1998) has postulated that an important component of sharp wave bursts may be initiated subcortically and could be involved in release of modulatory transmitters (e.g., acetylcholine) necessary for synaptic plasticity. In this context, one might expect more bursting in higher-order thalamic nuclei such as pulvinar because it is likely that these nuclei are more closely involved with hippocampal circuits. Therefore it would not be surprising to find more bursting in higher-order thalamic nuclei as Ramcharan et al. (2005) report and more bursting in all areas during sleep when visual memories are being consolidated. This could also explain higher bursting during inattentive phases (Swadlow and Gusev 2001). Perhaps the brain is continuously engaged in such long-term changes but only during periods when demands for sensory processing and vigilance are low.
Another reason that LGN bursting may be more common in some other thalamic nuclei than in the LGN could be because the retina normally has a powerful driving input and shows relatively high spontaneous activity. This constant bombardment rarely allows the prerequisite hyperpolarization to occur (Rowe and Fischer 2001). In the ventrobasal complex of the somatosensory thalamus, for example, the situation appears to be different because the input pathways (the principal sensory and spinal thalamic nuclei) show much less spontaneous activity, hence allowing for more thalamic bursts.
An additional possibility is that LT bursts occurring in isolation are only minimally relevant and that our use of single-unit recordings is methodologically inadequate to tease apart the overall relevance of bursting. By this we mean to suggest that the behavioral relevance of bursting during conscious states may be tied strongly to the synchrony of bursts across small ensembles of thalamic neurons. It is conceivable that a small number of synchronous bursts could facilitate spatiotemporal binding of visual information. This possibility is interesting from the point of view of information processing but is seemingly inconsistent with the observation that synchronized bursts are actually more common during sleep (Destexhe and Sejnowski 2002). Furthermore, if bursts were involved in binding, one might predict a dramatic increase in burst incidence when visual discrimination demands were increased, yet this was not observed in our study.
Clearly further study is required to determine whether burst and tonic modes of firing patterns in the thalamus truly represent distinct and meaningful modes of thalamo-cortical communication.
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Address for reprint requests and other correspondence: V. A. Casagrande, Dept. of Cell and Developmental Biology, Vanderbilt Medical School, U3218 Learned Lab, Nashville, TN 37232-8240 (E-mail: vivien.casagrande{at}vanderbilt.edu)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Alitto HJ, Weyand TG, and Usrey WM. Distinct properties of stimulus-evoked bursts in the lateral geniculate nucleus. J Neurosci 25: 514–523, 2005.
Bartlett JR, Doty RW, Lee BB Sr, and Sakakura H. Influence of saccadic eye movements on geniculostriate excitability in normal monkeys. Exp Brain Res 25: 487–509, 1976.[ISI][Medline]
Bezdudnaya T, Cano M, Alonso JM, and Swadlow HA. Inattentive vision and "burst mode" in awake subjects: temporal tuning in the LGNd. Soc Neurosci Abstr 506.11, 2005.
Buzsaki G. Memory consolidation during sleep: a neurophysiological perspective. J Sleep Res 7: 17–23, 1998.[Medline]
Casagrande VA, Sary Gy, Royal D, and Ruiz O. On the impact of attention and motor planning on the lateral geniculate nucleus. In: Progress in Brain Research. Cortical Function: A View from the Thalamus. Edited by Casagrande VA, Sherman SM, and Guillery RW. Amsterdam: Elsevier, 2005, vol. 149, p. 11–29.[CrossRef]
Chevalier-Skolnikoff S. Facial expression of emotion in primates. In: Darwin and Facial Expression: A Century of Research in Review, Ekman P. New York: Academic, 11–89, 1973.
Denning KS and Reinagel P. Visual control of burst priming in the anesthetized lateral geniculate nucleus. J Neurosci 25: 3531–3538, 2005.
Destexhe A and Sejnowski TJ. The initiation of bursts in thalamic neurons and the cortical control of thalamic sensitivity. Philos Trans R Soc Lond B Biol Sci 357: 1649–1657, 2002.[CrossRef][ISI][Medline]
Fanselow EE, Sameshima K, Baccala LA, and Nicolelis MA. Thalamic bursting in rats during different awake behavioral states. Proc Natl Acad Sci USA 98: 15330–15335, 2001.
Guido W and Weyand T. Burst responses in thalamic relay cells of the awake behaving cat. J Neurophysiol 74: 1782–1786, 1995.<
in B) was detected in 111 trials.
) and spike rate (— and