 |
||||
|
||||
|
||||
|
||||
The Journal of Neuroscience, August 6, 2003, 23(18):7021-7033
Previous Article | Next Article
Selective Elimination of Corticogeniculate Feedback Abolishes the Electroencephalogram Dependence of Primary Visual Cortical Receptive Fields and Reduces Their Spatial Specificity
Dirk Eyding,1 Jeffrey D. Macklis,2,3,4 * Ute Neubacher,1 Klaus Funke,1 * andFlorentin Wörgötter5 * 1Department of Neurophysiology, Ruhr University Bochum, D-44780 Bochum, Germany, 2Department of Neurology and Program in Neuroscience, Harvard Medical School, 3Division of Neuroscience, Children's Hospital, and 4Massachusetts General Hospital-Harvard Medical School Center for Nervous System Repair, Massachusetts General Hospital, Boston, Massachusetts 02115, and 5Department of Psychology, University of Stirling, Stirling FK9 4LA, United Kingdom
Abstract
Top
Abstract
Introduction
The role of corticogeniculate feedback in the organization, function, and state dependence of visual responses and receptive fields (RFs) is not well understood. We investigated the contribution of the corticogeniculate loop to state-dependent changes of characteristics of the primary visual cortex response by using a novel approach of eliminating corticogeniculate projection neurons with targeted neuronal apoptosis. Experiments were performed in anesthetized cats (N2O plus halothane) with parallel recordings of single units from experimental (right) and control (left) hemispheres2 weeks after induction of apoptosis. Within the experimental hemispheres, neurons of area 17 and of the dorsal lateral geniculate nucleus (dLGN) showed an unusually enhanced and prolonged tonic visual response during episodes of synchronized (syn) EEG activity, whereas response levels during less synchronized states were almost normal. In addition, dLGN cells showed a reduced tendency for burst firing and a less regular spike interval distribution compared with those of controls. These changes are likely attributable to a tonic depolarization of dLGN relay neurons or, more likely, to a decreased responsiveness of thalamic inhibitory processes to cortical feedback. Cortical neurons also displayed an activity-dependent increase in RF size, in contrast to an almost activity-invariant RF size of controls, a phenomenon likely related to the elimination of collateral, intracortical projections of layer 6 neurons. Together, these results demonstrate that selective chronic elimination of corticogeniculate feedback results in the loss of EEG-correlated differences of visual processing in the remaining thalamocortical network, accompanied by a significant increase in excitability during syn EEG, at the expense of noticeably reduced spatial receptive-field specificity in the remaining cortical neurons.
Key words: visual cortex; corticogeniculate loop; photoactivation of chlorin e6; neuronal apoptosis; state dependence; receptive-field characteristics
Introduction
The spectral composition of the electroencephalogram (EEG), a signal averaging the activity of large populations of cortical neurons, strongly covaries with behavioral states of an animal, primarily as a result of the activity of modulatory brainstem nuclei (Foote et al., 1991; Munk et al., 1996
-wave-dominated) EEG states, neurons in the visual thalamus [dorsal lateral geniculate nucleus (dLGN)] are preferentially in a burst mode. This mode is characterized by brief, transient, high-frequency (phasic) responses of only a few tens of milliseconds in duration, even to prolonged visual stimuli. The same neurons respond to the same stimuli with long-lasting tonic responses during activated (previously termed desynchronized,
-wave-dominated) EEG states. During this so-called tonic transmission mode, dLGN neurons can transfer additional information about the nature of the stimulus (e.g., its local contrast or duration) (Sawai et al., 1988
It has been suggested that corticogeniculate feedback plays an important role in the perpetuation of the tonic component of dLGN relay neurons during activated episodes (Kalil and Chase, 1970
Previously, it has not been possible to record visual cortical activity while inactivating the corticogeniculate feedback projection because of the global effect of all of the previously available inactivation procedures on the entire cortical network, also including neurons other than the layer 6 projection neurons. In this study, we applied a novel experimental approach of biophysical induction of neuron type-specific targeted apoptosis (Macklis, 1993
Materials and Methods
Induction of apoptosis of corticogeniculate neurons in cortical layer 6. Apoptotic death of corticogeniculate feedback neurons in layer 6 of cortical areas 17-18 was induced using methods previously described by one of our laboratories (Macklis, 1993Stereotaxic injection of photoactive nanospheres. Under deep anesthesia with a combination of ketamine (20 mg/kg) and xylazine (2 mg/kg), cats [n = 9; five with cortical recordings; two with geniculate recordings after apoptosis; one control experiment with cortical and geniculate recordings but no final apoptosis (sham control); and one control experiment with apoptosis and multisite EEG recording instead of multiunit recording] were stabilized using standard stereotaxic methods, and craniotomies were performed unilaterally to allow access to the dLGN. All of the incisions were also locally anesthetized by xylocaine. Nanospheres carrying the targeting chromophore chlorin e6 [covalently bound to the 20 - 60 nm subfraction of rhodamine latex microspheres (Lumafluor, New York, NY)] were stereotaxically pressure microinjected under electrophysiological control into both dorsal layers (A and A1) of the dLGN of one hemisphere at three to four sites (20 nl in each site;
View larger version (23K):
[in this window]
[in a new window]
Figure 1. Schematic illustration of the experimental approach (for details, see Materials and Methods). a, Injection of fluorescent nanospheres carrying chlorin e6 (depicted in gray here) into both layers A and A1 of cat dLGN under electrophysiological control. b, The region of cortex with retrogradely labeled neurons in layer 6 was illuminated with deeply penetrating 674 nm light after electrophysiological verification of the correct topographic region. c, After completion of targeted corticogeniculate neuron apoptosis, electrophysiological recordings were conducted in both cortex and dLGN.
Induction of targeted apoptosis of corticogeniculate projection neurons. After 2-3 weeks, allowing sufficient time for retrograde transport of the photoactive targeting nanospheres, apoptosis of corticogeniculate neurons was induced in cortical layer 6 of primary visual cortex areas 17 and 18. The same basic surgical procedures were used as mentioned above. Neuronal apoptosis was induced via the noninvasive, transdural activation of chlorin e6 with 674 nm light (region diameter, 1.9 mm; 2.7 kJ/cm2; 2.2 W/cm2) delivered by a continuous-wave laser (Schaefter und Kirchhoff, Hamburg, Germany) coupled to custom beam-controlling optics via a quartz fiber. Light was applied after exposing the appropriate cortical surface representing the same visual field position as the injection sites, as reverified by RF mapping. After light exposure, neuronal apoptosis progressively developed over 10 -14 d (Fig. 1b).
Electrophysiology and receptive-field mapping (Fig. 1c). We applied conventional approaches for semiacute extracellular recording to allow RF mapping (Wörgötter et al., 1998b
Information about the cortical layer of the neurons is based on recording depth, because we avoided interference with the process of targeted neuronal apoptosis by not marking the electrode tracks.
For the control recordings in the dLGN, an optimally sized dot stimulus was flashed at a fixed position in the center of the RF (50% contrast; duty cycle, 0.8:1.2 sec), to obtain a standard peristimulus-time histogram (PSTH).
EEG analysis. The state of the EEG was determined by a sliding-window fast Fourier transformation (time base, 5 sec; step size, 1 sec) of the EEG trace, and the calculation of the
band were less effective, probably because
Analysis of cortical and geniculate single unit activity. Because of the RF-mapping procedure, cortical responses to visual stimuli can be represented as a three-dimensional (3D) plot, as shown in Figure 2a, with space and time as the x- and y-axes and the level of activity as the z-axis.
View larger version (39K):
[in this window]
[in a new window]
Figure 2. Schematic illustration of the response analysis. a, 3D representation of a cortical response with time and space on the x- and y-axes, respectively, and neuronal activity on the z-axis. The phasic maximum is the reference of the activity measure and—for graphical reasons—evenly divided into 10 vertical bins. The time axis has a resolution of 10 msec. Raw data maps were then smoothed with a 2D Gaussian profile with a of five bins along space and time. b, PSTH (top) obtained from the central cross-section of the 3D plot and a contour map (bottom; 2D map) viewed along the z-axis of the 3D plot. The calculation of the absolute width, the normalized WHH, and DHH of the tonic response is illustrated. Reference for the latter two is the tonic maximum (of 100 Hz in this schematic example). The lower bound for absolute width is twice the SD (2SD) of the spontaneous activity (gray shaded area). Activity levels above half-height of tonic maximum are shown in black, and those below are shown with white shading.
These raw 3D maps (bin width, 10 msec) were convoluted with an optimally suited two-dimensional (2D) Gaussian profile for smoothing in time and space. The Gauss had a
From the smoothed 3D maps, we determined the maximum of the phasic response component, and for the tonic component, the duration as well as two measures of the width and the mean amplitude. Two times the SD of the spontaneous activity level (shown in gray shading in Fig. 2b) was taken as the minimum response level, a conventional threshold for separation of background activity. The phasic response is the transient, high-frequency response first after stimulus onset, whereas the tonic response is the sustained response component afterward. The maximum of the phasic response is the firing rate of the highest bin. To determine the parameters of the tonic response, we first defined its temporal outline approximately by visual inspection. This is straightforward, because the shape of the responses was almost always very distinct (Fig. 2b, top). Within the tonic response, we determined the tonic response maximum from the smoothed profile. The tonic maximum was then used to more accurately define the duration of the tonic response as the difference between those two points at which the firing rate of the neuron crosses the half-height of the tonic maximum [duration at half-height (DHH)] (Fig. 2b). The normalized tonic RF width was determined as the maximal width of the tonic response portion (again, with the temporal limits given by the duration) at half-height of the tonic maximum [width at half-height (WHH)] (Fig. 2b, bottom).
It has been shown that acute silencing of the corticogeniculate feedback leads to a general depression of geniculate activity (Kalil and Chase, 1970
Geniculate visual responses were analyzed with the aid of peristimulus-time histograms and interspike interval histograms. Receptive fields of pairs of dLGN units were stimulated in an interleaved manner with spots of light of 33-50% contrast to background and 0.3-1.0° in diameter, flashed for 800 msec within the center of the RF (total sweep length, 2 sec).
Histology. After electrophysiological data collection, cats were perfused with heparinized Ringer's solution, followed by 4% paraformaldehyde in PBS at pH 7.4. After fixation, brains were removed and cryoprotected in 30% sucrose. Coronal frozen sections (30 µm) of area 17, area 18, and the dLGN were prepared. Sections were alternately prepared for fluorescence microscopy and cresyl violet staining (Nissl stain). Sections from the regions of targeted neuronal apoptosis and the homologous regions of the control hemispheres were processed for immunocytochemistry using primary antibodies directed against the mature neuronal marker NeuN (Chemicon, Temecula, CA). Labeling for NeuN was detected both histochemically and by fluorescence, in alternate sections: (1) to allow for quantitative neuronal counting, we used peroxidase anti-peroxidase (Dako, Glostrup, Denmark) and Ni-enhanced diaminobenzidine (DAB-Ni), and (2) to determine the ratio of retrogradely labeled neurons via fluorescence of the targeting chlorin e6 nanospheres (label ratio), we used secondary antibodies coupled to the green fluorochrome carbocyanine (Cy2) (Jackson ImmunoResearch, West Grove, PA).
Quantification of neuronal density in area 17-18. Quantitative analysis was performed using modified stereological methods (Guillery and Herrup, 1997
View larger version (112K):
[in this window]
[in a new window]
Figure 3. Highly selective elimination of corticogeniculate neurons in areas 17-18 of experimental cortex. a, Gyrus lateralis of an experimental cat 14 d after induction of apoptosis. The red arrow indicates the center and direction of photoactivation by incident light (1.9 mm diameter region of photoactivation); L, lateral; V, ventral; WM, white matter. b, c, Merged overlay of images for the rhodamine fluorochrome carried by the targeting nanospheres (red) and immunocytochemical labeling for the mature neuronal marker NeuN with Cy2 (green). b, High magnification image of the blue-boxed region in a, demonstrating the highly efficient elimination of corticogeniculate projection neurons in the experimentally targeted region. c, High magnification image of the red-boxed region in a, demonstrating the normal appearance of corticogeniculate neurons in control primary visual cortex. d-f, Quantification of corticogeniculate neuron loss by relative cell density profiles. Regions denoted by gray and white shading represent assignment to specific cortical layers (noted above the abscissa). d, Density profile in regions of corticogeniculate neuron apoptosis at the tip of the gyrus and in the homologous regions in the control hemisphere (see insets above). e, Control region at the tip of the gyrus 3-6 mm anterior to the affected region, within the same experimental hemisphere (see inset). f, Control regions deep in the medial or lateral banks of the experimental hemisphere. Note that layer 5A is relatively thicker than layers 5B-6 in these deep regions compared with that at the tip of the gyrus, as indicated by three sampling windows and thicker area of gray shading. Error bars in d-f indicate SDs from the means.
Histological and physiological control procedures. One sham control experiment was performed, specifically to test our procedures for inducing targeted neuronal apoptosis, in which all of the experimental steps were completed except for initiation of targeted neuronal death. We found no changes in cortical histology and no change in any electrophysiological parameter, compared with intact controls. In two more cats, we performed control measurements in the dLGN (n = 36 neurons) after induction of apoptosis, to compare the effects of chronic silencing of the corticogeniculate pathway with the previously published acute effects on phasic and tonic response components (Funke and Eysel, 1992
Results
Quantification of neuron loss in layer 6
After inducing the death of the targeted corticogeniculate neurons (Fig. 1a,b), histological analysis after the final recording session (c) revealed a substantial reduction of corticogeniculate projection neurons in the experimental regions of cortical layer 6. Because the targeting nanospheres also contain rhodamine as a fluorescent label, we could visualize the loss of previously labeled neurons. The substantial reduction in layer 6 neurons was visible by routine light-microscopic inspection of individual sections, which showed many fewer rhodamine-fluorescent corticogeniculate neurons in the experimental regions of layer 6, compared with the control regions (Fig. 3a-c).To determine quantitatively the loss of the labeled neurons, we first determined the percentage of total neurons in layer 6 that were initially labeled by the geniculate injections, without photoactivation and initiation of targeted neuronal death (label ratio). We then counted the total number of neurons in layer 6 of the experimental and the appropriate control cortices to quantify the total neuron loss (total loss ratio). From these two ratios, we calculated how many of the labeled neurons underwent apoptosis.
The label ratio of corticogeniculate neurons, compared with all of the NeuN-labeled neurons in control regions of experimental animals, was 45 ± 3% in the densest sublayers, in agreement with previous studies (McCourt et al., 1986
In both of the control regions immediately adjacent to the experimental regions (Fig. 3e), and in deep control regions (f), quantification of the density of layer 6 neurons was uniform with complete overlap of the density profiles, demonstrating no loss of neurons in the controls. Together, these results indicate that targeted induction of corticogeniculate neuron death was quite efficient and occurred only in the appropriate highly localized region of layer 6 in visual cortex. (See Discussion for additional details.)
We will now first discuss the effect of targeted apoptosis on the cortical network and later compare it with responses from dLGN neurons recorded after having applied the same experimental protocol.
Consequences of layer 6 neuron loss on cortical responses
We recorded the responses of 76 individual cortical neurons. It is known from previous studies that the spatial-response (Wörgötter et al., 1998bWe found that elimination of corticogeniculate neurons had striking effects on cortical response properties, abolishing the normal dependence on EEG state. In control cortex with intact corticogeniculate feedback circuitry, longer and more pronounced tonic responses occurred during less synchronized EEG states, compared with shorter, more phasic responses during synchronized EEG (
View larger version (45K):
[in this window]
[in a new window]
Figure 4. State dependence of tonic visual responses in control and experimental hemispheres. Four representative neurons are depicted from control (a-d) and experimental (e-h) hemispheres. The PSTHs show the activity of each cortical neuron along the central cross-section of its receptive fields as demonstrated in Figure 2. a-d, As expected, responses in control hemispheres are longer and more pronounced during less synchronized EEG. e-h, The situation is reversed in the experimental hemispheres, in which responses are stronger and have a longer duration during synchronized EEG. White boxes label the part of the tonic response that exceeds half-height of maximum. Twenty second-long EEG traces above the PSTHs show the typical EEG pattern during those episodes of spike activity collected in the corresponding PSTHs. To the left of e-h, coordinates are indicating the visuotopic location of the RF of the respective neuron (box), in relation to the RF location of the original nanosphere injection sites (gray dots) and the center of the region of targeted corticogeniculate neuron apoptosis (X), within the visual field (axes in degrees of visual angle; HM, horizontal meridian; VM, vertical meridian). Cell types: complex (a), complex (b), simple (c), simple (d), simple (e), simple (f), complex (g), and complex (h).
Not only the amplitude and duration of the tonic visual response but also the spatial distribution of the activity differed between the experimental and control hemispheres during the different states of the system. RF widths in the experimental cortex appeared to covary with the amplitude of the response and were often wider with a high-amplitude response during syn episode than during a lower amplitude response in the less syn state. In contrast, in control cortex, no such covariation between amplitude and RF width could be observed. The RFs appeared to be equally wide or even narrower during the less syn episodes, despite the higher amplitudes. To more clearly demonstrate this effect, we replotted in Figure 5 the same data that are shown in Figure 4 (except for Fig. 4d because of missing tonic response) by showing 2D RF plots with normalized RF widths, which reflect WHH (areas filled in black) of the tonic response amplitude maximum (Fig. 2). This measure was chosen, because it is insensitive to mere scaling (iceberg) effects and was therefore also used for statistical quantification (Fig. 6). This representation shows that tonic RFs in control cortex were typically substantially narrower (early part) during less synchronized EEG than during synchronized EEG (a-c), in spite of the fact that the tonic amplitude was typically higher during the less syn state. In contrast, the normalized tonic RF widths in the experimental cortex showed a similar width during both EEG states (d and f), even wider RFs during less syn episodes (e), or, because of a very large amplitude difference, still wider RFs during syn episodes (g). In summary, these results suggest that, in normal cortex, there exists a normalizing mechanism that keeps the size of the receptive fields relatively constant. This mechanism appears to be impaired during chronic elimination of the dLGN projecting layer 6 neurons, most likely because of their collateral projection to layer 4.
View larger version (52K):
[in this window]
[in a new window]
Figure 5. Examples of cortical receptive-field maps with activity scaling normalized to the activity level at width at half-height of tonic response. Those examples of Figure 4 with sufficiently strong tonic responses are shown as 2D contour plots (only d is omitted). Shown in black is the area that had higher activity than one-half of the tonic maximum, which is depicted in absolute firing rate below the plots. WHH and DHH can be directly seen in this representation. max, Maximum of tonic response.
View larger version (34K):
[in this window]
[in a new window]
Figure 6. Quantitative analysis of response properties of experimental and control primary visual cortex. Responses of experimental (open squares) and control (filled squares) neuron samples are depicted during both synchronized and less synchronized EEG states, with regard to their duration (a), phasic-response (b) and tonic-response (c) amplitude (ampl.), tonic receptive-field width (e), and normalized (norm.) tonic receptive-field width (f). Logarithmic scatter plots show values obtained during less synchronized EEG, plotted against those observed during synchronized EEG. Enlarged dots depict the mean values for experimental (open) and control (filled) data. Histograms in the top right corners show the distributions of the data points (open bars, experimental; filled bars, control) with respect to the diagonal. This indicates the relative deviation ( ) from an EEG-independent behavior. The inset histograms at the bottom right of each panel show mean values and SEs of the distributions; large values are indicative of strongly EEG-dependent behavior; significance was evaluated by a one-sample Student's t test (mean against
Statistical analysis
Quantitative statistical analysis confirmed the qualitative observations (Fig. 6, Tables 1 and 2). The previous qualitative analysis suggested that responses in experimental cortex increased in both tonic amplitude and duration, particularly during synchronized EEG, and that tonic receptive fields scaled with amplitude. Figure 6 presents this quantitative analysis graphically, first with scatter plots of the raw data and their mean values (large dots). Histograms in the top right corner of each panel and their means (small insets) demonstrate the elimination of EEG-dependent behavior of cortical responses in the experimental hemisphere (Fig. 6a,c, open bars). Responses were only of 7 ± 7% longer (Fig. 6a) and 6 ± 5% higher (c) amplitude during less syn than during syn. This elimination of EEG-dependent behavior was primarily attributable to a significantly increased response duration (Fig. 6a) and amplitude (c) during synchronized EEG, compared with the control cortex (right shift of the large white dot only along the x-axis compared with the black dot). Because the change during less syn episodes appeared negligible, we calculated the shift only along the x-axis, which corresponds to an increase in duration of 64 ± 10% and an increase in amplitude of 77 ± 14% during syn episodes in the experimental hemispheres compared with the control cortices.
View this table:
[in this window]
[in a new window]
Table 1. Statistical summary of state-dependent visual activity in control and experimental cortex
View this table:
[in this window]
[in a new window]
Table 2. Tonic responses of simple and complex cells
This was in sharp contrast to control cortices, in which the duration and amplitude of tonic responses were typically enhanced during the less synchronized EEG state [Fig. 6a,c, black bars, corresponding to 62 ± 12% longer duration (a) and 34 ± 8% higher amplitude (c)], consistent with previous reports (Ikeda and Wright, 1974
From the examples above, we expected a change in the relation between response amplitude and RF width in the experimental cortices. Statistical analysis demonstrated that covariation of receptive-field width and amplitude occurred only in experimental hemispheres, whereas RF width remained independent of the response amplitude in controls. This overall result is depicted more completely and with more detail in the next figure panels. For statistical analysis, we measured receptive-field widths (absolute and normalized; see Materials and Methods) during the tonic responses. This was possible only for a subset of neurons, in which the receptive fields were sufficiently smooth. The covariation of absolute RF width and response amplitude was analyzed in Figure 6d by plotting the less syn/syn ratio of the absolute RF widths against the ratio of the corresponding response amplitudes. In cases of a covariation, one would expect most data points to lie in the concordant quadrants (both ratios below or above 1 of the plot). We found exactly this result in the experimental cortex. Significantly more of the data points in Fig. 6d are in the concordant quadrants (31/38;
2 = 15.1; p < 0.001), demonstrating that RF widths scaled directly with amplitude. In control cortex, however, the observed amplitude increase during less synchronized EEG (Fig. 6c) was not typically accompanied by an increase in receptive field width; receptive fields in control cortex typically remained narrow. Therefore, the majority of data points in the controls lie in the discordant quadrant at the bottom right (15/23), although not significantly (
The insets in Figure 6, a-c and e-f, summarize the findings: the normally observed EEG-dependent changes in cortical-response characteristics were eliminated in experimental cortices because of an increase in the duration (a) and amplitude (c) of neuronal responses during synchronized EEG, whereas their spatial selectivity was diminished simultaneously (f).
Table 1 summarizes the means and SEM for the samples shown in Figure 6 (except for d). For tonic-response duration and amplitude, Table 2 shows that simple and complex cells were affected almost in the same manner, but the increase of tonic visual activity in the lesioned hemisphere is on average somewhat stronger in simple cells. The differences in tonic-response amplitude and duration, however, do not differ significantly between simple and complex cells, both in control and experimental hemispheres.
dLGN data
As a control, we also performed recordings from 36 neurons in the dLGNs of two cats (18 pairs in experimental/control hemispheres), at or close to the injection sites as verified by stereotaxic coordinates and visual field positions. Similar to what was found for cortical responses, the geniculate visual responses also showed an increase in the tonic response component during syn episodes, leading to a loss of its normal EEG dependence (Sawai et al., 1988
View larger version (32K):
[in this window]
[in a new window]
Figure 7. Effect of chronic elimination of corticogeniculate feedback on thalamic visual activity. Peristimulus-time histograms (top rows; a, c, e, g) and spike interval histograms (bottom rows; b, d, f, h) showing the temporal properties of light responses evoked in two dLGN relay cells during different states of the system (top set, syn state; bottom set, less syn state), one recorded from the control hemisphere (a-d) and one simultaneously recorded from the experimental hemisphere (e-h). A spot of light slightly larger than the receptive-field center of the dLGN cells was flashed for the time indicated by the horizontal bar below histogram (a). The spike interval histograms composed of bars were calculated for the entire activity during the recording episode for spike intervals between 1 and 30 msec. The four bins to the left of the vertical line dissecting the diagram correspond to spike intervals usually composing a burst (1-4 msec). The smaller inset diagrams (solid lines) show the spike interval distribution during the tonic part of the visual response, for intervals between 1 and 50 msec. Arrows in histogram d indicate two interval modes, as described by Funke and Wörgötter (1995
Figure 8 summarizes the statistical analysis of the dLGN data. The mean ratio of tonic- to phasic-response amplitudes during the syn state clearly increases in the experimental hemispheres and approaches values usually obtained during less syn episodes in the control hemisphere (Fig. 8a). The ratio between less syn and syn states for this index is almost 1 in the experimental hemisphere and significantly smaller than the normal ratio obtained for the control hemispheres (Fig. 8b). For a subset of cells (n = 12 pairs in experimental and control hemisphere) that showed at least one clear mode in the spike interval distribution of the tonic response (Fig. 7d, inset diagram), we calculated the WHH of these modes. For the remaining cell pairs, either one or both cells of a pair did not show a clear mode in the interval distribution. In most cases, it was the cell of the experimental hemisphere that had a broad distribution without a distinct interval mode. Figure 8c demonstrates that mean WHH of modes calculated for the neurons of the experimental hemisphere is significantly broader than that of the neurons recorded from the control hemisphere during the less syn state. The WHH for the hemisphere with chronic elimination of corticogeniculate feedback is
5 msec. Figure 8d shows two major findings: (1) the typically higher incidence for burst firing (higher ratio) during the syn state compared with that of the less syn state as seen in the control hemisphere is abolished in the experimental hemisphere, and (2) the likelihood of burst firing is generally reduced in the experimental hemisphere, with a reduction of 80% in the syn state and 65% in the less syn state, indicating that the thalamic neurons may be tonically depolarized and thus are not able anymore to achieve a state suitable for generating low-threshold calcium spike-carried bursts.
View larger version (31K):
[in this window]
[in a new window]
Figure 8. Statistical analysis of the temporal structure of dLGN activity. a, Tonic/phasic ratio of light responses in control and experimental hemispheres in relation to the two system states of syn and less syn. b, The change of the tonic (ton.)/phasic (phas.) index with respect to EEG state. c, The mean WHH of spike interval modes during the tonic light response in control and experimental hemispheres. d, The ratio of spikes included in a burst (1-4 msec spike intervals) to single spikes (>5 msec spike intervals) in control and experimental hemispheres. For each cell pair (number of pairs analyzed is indicated at the bottom of histograms), one neuron was recorded from the unaffected hemisphere (control) while the other neuron was simultaneously recorded (with interleaved visual stimulation in the contralateral hemifield) from the experimental hemisphere. Asterisks indicate statistically significant differences between samples (*p < 0.05; **p < 0.01; Student's t test). Error bars indicate SEM.
Control experiments
Controls (sham controls) in which all of the experimental steps were performed except for initiation of targeted neuronal death showed the following: (1) no changes in cortical histology, compared with intact controls, (2) specifically, no change in the number of corticogeniculate neurons, and (3) no change in any electrophysiological parameter, compared with intact controls in dLGN and cortex.In one experiment with induced apoptosis of corticogeniculate projection neurons, we made no single-unit recordings but local EEG recordings directly above the lesioned cortex, 4 mm anterior to the lesion within the same hemisphere (still area 17) and at a position opposite to the lesion within the control hemisphere. The EEG was recorded via small (<0.5 mm) silver-ball electrodes attached to the cortical surface. We made long-lasting recordings with the cat either in the dark or a dimmed room, or during full-field flicker stimulation. Neither the spectral power of the EEG nor the coherence of EEG waves were found to differ for the three recording sites (data not shown). For additional details, see Discussion.
Discussion
The combination of targeted neuronal apoptosis of corticogeniculate projection neurons with electrophysiological recording in visual cortex and thalamus made it possible for the first time to investigate directly whether elimination of a large proportion of the corticogeniculate projections would result in a loss of EEG-dependent response properties in the visual cortex itself. The approach of inducing targeted apoptotic neuronal death was established in mice (Macklis, 1993So far, the contribution of corticogeniculate feedback to state-dependent differences at the level of the dLGN has been documented only in acute experiments (Funke and Eysel, 1992
These results raise three major questions: (1) Why does elimination of corticogeniculate feedback lead to the abolition of state-dependent variations in visual responsiveness? (2) Why is there a response increase, despite the elimination of excitatory feedback? (3) What is the reason for the activity-dependent increase in receptive-field size in the experimental cortex?
First, it is likely that the observed loss of state dependence is the result of equally strong retinogeniculate signal transmission during syn and less syn episodes, attributable to either upregulated retinal input or other influences. The high retinogeniculate transfer ratio (Coenen and Vendrik, 1972
Second, previous studies investigating the effects of retinal deprivation demonstrated that the removal of excitatory input from the retina (deafferentation) leads to increased excitability of target neurons in dLGN and cortex (Eysel, 1979
Finally, modulatory inputs from the brainstem arousal centers might also come into play; however, this scenario is less likely, because these systems are not directly affected by our experimental procedures.
It is by now widely accepted that the interactions between thalamus, brainstem, and cortex are primarily responsible for the generation of the different EEG states, in part by influencing the response modes of thalamic cells. Intracortical mechanisms seem to be less important in this respect. This view is also supported here. The consistency of our results with the different mechanisms discussed above strongly suggests that effects occurring at the geniculate cells themselves are primarily responsible for the enhanced visual responsiveness during the syn state in cortex and dLGN. Intracortical plasticity reported by others (Kaas et al., 1990
The change in EEG-dependent activity in thalamus and cortex implies that also the spectral composition and coherence of the EEG within the lesioned cortex may have changed. This, however, was actually not the case. The diameter of the EEG electrode above the lesioned cortex (<0.5 mm) should have been small enough compared with the diameter of the lesioned area of layer 6(
Third, why does RF size scale with tonic activity in experimental cortex, but not in control hemispheres, which operate essentially in an RF size-invariant condition? Here, it is more likely that cortical rather than thalamic mechanisms are involved, because dLGN neurons show little if any state-dependent changes in RF size (Hartveit et al., 1993
Layer 4 neurons have been shown to receive direct excitatory input from layer 6, which is less efficient than the input from other layer 4 neurons, but with a tendency for facilitation with repetitive input (Stratford et al., 1996
Together, the experiments reported here demonstrate that selective chronic elimination of corticogeniculate feedback results in a loss of EEG-correlated differences in the remaining thalamocortical network, leading to a significant increase in excitability during syn EEG, at the expense of reduced spatial RF specificity in cortical neurons. Our results indicate that corticogeniculate feedback is critical for the establishment of differences in sensory processing, and that the observed changes are most likely related to a weakening of cortically driven inhibitory thalamic networks.
Footnotes
Received Nov. 1, 2002; revised May. 28, 2003; accepted May. 29, 2003.This work was supported by a collaborative grant from the Human Frontiers Science Program (F.W. and J.D.M.) and by a National Institute of Child Health and Human Development Mental Retardation Research Center grant (J.D.M.). Additional support was provided by grants from the National Institutes of Health (J.D.M.) and Deutsche Forschungsgemeinschaft (F.W. and K.F.). We thank Cindy C. Tai and Max Christian for excellent technical assistance, Nicolas Kerscher for experimental assistance, Ulf Eysel for helpful discussions, and Sanjay S. P. Magavi for helpful input from parallel mouse work.
Correspondence should be addressed to Dr. Klaus Funke, Department of Neurophysiology, Medical Faculty, Ruhr University Bochum, D-44780 Bochum, Germany. E-mail: funke{at}neurop.ruhr-uni-bochum.de.
D. Eyding's present address: PAION GmbH, Martinstrasse 10-12, D-52062 Aachen, Germany.
Copyright © 2003 Society for Neuroscience 0270-6474/03/237021-13$15.00/0
* J.D.M., K.F., and F.W. contributed equally to this work.
References
