| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Department of Neuroscience, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
Submitted 25 September 2006; accepted in final form 24 June 2007
 |
ABSTRACT |
|---|
|
|
|
INTRODUCTION |
|---|
|
and Barlow (1953), the spatial weighting functions of both retinal and lateral geniculate nucleus (LGN) X-cell receptive fields have been viewed as the difference of two Gaussians (DOG) (Rodieck 1965). The DOG model consists of a strong and narrow central Gaussian and a weak but broad surround Gaussian. This linear model accounts for many properties of retinal ganglion cells (RGCs) and LGN cells, but numerous authors have described excitatory and/or suppressive effects in LGN cells that are not accounted for by the classic DOG model (Alitto and Usrey 2003; Cudeiro and Sillito 1996; Levick et al. 1972; Murphy and Sillito 1987; Solomon et al. 2002; Webb et al. 2002).
Many of these effects are attributed to a "nonclassical" receptive field and have proven to be controversial. Several authors have proposed that these effects are spatially localized to a region outside of the classical receptive field (e.g., Webb et al. 2002), although a recent study suggests that they are colocalized with the classical receptive field (Bonin et al. 2005). Sillito and colleagues described suppression in LGN cells that was dependent on the relative orientation of grating stimuli presented to the center and surround of the receptive field (Cudeiro and Sillito 1996; Sillito et al. 1993). They also showed that the suppression found with isooriented, but not orthogonally oriented, stimuli was partially released on ablation of visual cortex. Others have failed to see these orientation effects in the intact animal (Bonin et al. 2005; Solomon et al. 2002; Webb et al. 2002) and Webb et al. (2002) reported no differential release of suppression after cortical ablation. Some studies have attributed changes in the gain of the classical receptive field to stimuli in the nonclassical surround. Webb et al. (2002) observed no effect of cortical inactivation on the gain of the receptive field, whereas others showed that cortical feedback may enhance responses in the LGN (Przybyszewski et al. 2000). Solomon et al. (2002) described a divisive reduction in response in the presence of an annular surround stimulus, but did not inactivate visual cortex.
In this report, we focus on a particular shortcoming of the DOG model: suppression of responses of LGN cells at spatial frequencies above those to which the classical receptive field surround is responsive. We examine spatial and functional properties of this suppression, including the controversial effects just described. We find that both the feedforward input from RGCs and feedback input from the ipsilateral visual cortex contribute to suppression at high spatial frequencies observed in the LGN and that suppression in the LGN is stronger than that in its retinal input regardless of the spatial frequency. We compare many of our findings with those of Solomon et al. (2002), who demonstrated suppression at the optimal spatial frequency of primate LGN cells and those of Bonin et al. (2005) who showed that the high spatial frequency cutoff of the suppressive field was 1.27-fold the cutoff value of the classical receptive field in cat LGN. We also discuss how our findings influence current receptive field models of LGN cells and how our results affect our understanding of the circuitry involved in the corticothalamic projection.
These conclusions are reached using two approaches. The first is to record simultaneously from LGN cells and one of their excitatory RGC inputs. This is accomplished by recording S-potentials, the extracellularly recorded postsynaptic signature of RGCs (Bishop et al. 1958; Cleland et al. 1971; Freygang 1958; Hubel and Wiesel 1961; Kaplan and Shapley 1984). Our second approach is to record extracellularly from LGN cells before, during, and after cooling of a large, retinotopically corresponding portion of visual cortex (areas 17 and 18). Some of these results have been presented in abstract form (Nolt et al. 2005).
|
|
METHODS |
|---|
|
Animals were prepared for single-unit recording as described in detail elsewhere (Nolt et al. 2004). Briefly, adult cats were anesthetized with 3–4% halothane in a 70:30 mixture of N2O and O2. Venous catheters were placed in each hind limb, gas anesthesia was discontinued, and the animal maintained on intravenous sodium thiopental as needed through the remainder of the surgery. A tracheotomy was performed, the animal was placed in a stereotaxic frame, and it was paralyzed with an injection of gallamine triethiodide (Flaxedil, 60 mg). The animal was maintained on positive-pressure ventilation adjusted so as to hold the end-expired CO2 at 3.8%. Two venous catheters were used so that anesthetic (2–8 mg · kg–1 · h–1 thiopental) and paralytic (15 mg/h Flaxedil) could be infused throughout the experiment at independent rates. Cortical electroencephalogram (EEG) was monitored continuously throughout the experiment and the rate of anesthetic infusion regulated so as to maintain the animal in a state similar to light sleep characterized by frequent bursts of 7- to 10-Hz waves (spindles). Body temperature was also monitored and maintained at 38°C. A large craniotomy was made from Horsley–Clarke A11.0 to P10.0 and from the midline to L10.0 to allow access to the LGN and placement of the cooling plate over visual cortex. At the conclusion of the experiment, animals were given a lethal injection of sodium pentobarbital. This protocol was approved by the University of Pennsylvania's Institutional Animal Care and Use Committee and conforms to guidelines recommended in Preparation and Maintenance of Higher Mammals During Neuroscience Experiments (National Institutes of Health, publication 91–3207).
Neutral contact lenses were placed in each eye and spectacle lenses added to bring the reflection of retinal vessels into sharp focus on the screen of a CRT mounted in front of the cat. Biprisms were also placed in front of each eye to allow approximate superposition of the lines of sight, although cells were studied monocularly by covering the nondominant eye. Pupils were dilated with 1% ophthalmic atropine and the nictitating membranes retracted with 1% phenylephrine hydrochloride. Animals were given intramuscular (im) injections of glycopyrrolate (0.1 mg im once per day) to minimize secretions, dexamethasone (0.4 mg im once per day) to minimize cerebral edema, and ampicillin (10 mg/kg im) to prevent infection on a 12-h schedule. Most experiments lasted 1 day.
Recording and data acquisition
All recordings were made extracellularly with tungsten-in-glass electrodes. Signals were amplified and conditioned with a programmable amplifier (Alpha Omega, Nazareth, Israel). Amplified signals were routed to an oscilloscope, an audio monitor, and a spike sorter (Alpha Omega). The latter performed on-line template matching of action potentials and produced TTL pulses, which were detected at the computer interface during clock interrupt-service routines every 100 or 1,000 µs, depending on the application. The electrode was advanced with a Burleigh Inchworm (Burleigh Instruments, Victor, NY). Penetrations were limited to layers A and A1 based on electrophysiological criteria; no histology was performed. All receptive fields were limited to the central 5° and within the 12.5° below the horizontal meridian of the visual field because this was the area of visual field covered by the cooling plate in visual cortex.
Visual stimuli were presented on an Image Systems multisynch monochrome monitor at a frame rate of 125 Hz by means of a VSG-2/3 board (Cambridge Research Systems, Cambridge, UK). Mean luminance was 80 cd/m2 and lookup tables were linearized for contrasts in the range of ±100%. The monitor was situated 24.7 cm from the eye such that 1 cm represents 2.3° of visual space. The screen subtended 36° horizontally and 27° vertically. Stimuli were presented with a resolution of 22 pixels/degree. An identical monitor driven in series was situated at the experimenter's console.
The responses of each cell were thoroughly characterized before any experimental procedures. This was accomplished in two phases. First, computer-assisted hand-plotting procedures were used to estimate the position and general properties of the receptive field in both the two-dimensional (2D) spatial and spatial-frequency domains (Jones and Palmer 1987). In the second stage, the parameters of the receptive field were refined using a series of programs, each of which generated stacks of histograms as stimuli spanning some variable were presented in random order. Specifically, we generated at least one spatial-frequency tuning curve and a contrast-response function for each cell. The temporal frequency of the stimuli ranged between 2 and 4 Hz. In addition, several procedures were used to ensure that the exact center coordinates of the receptive field were available for subsequent data acquisition. Along with the preliminary hand-plotting procedures, we used reverse correlation of the spike train with a dense, 2D spatiotemporal white noise stimulus (Reid et al. 1997). The position of the stimulus was modified until the most vigorous response elicited from the receptive field was centered within the stimulus. The centering of the stimulus was periodically reexamined because eye drift can cause shifts in the location of the receptive field. Changes in the receptive field location are often detected by changes in response phase, but this was rarely observed. Additional details will be described in RESULTS.
Details of our recordings of S-potentials have been described previously (Nolt et al. 2004). Using on-line template matching, we were able to simultaneously and independently extract both LGN and S-potential events.
Cells were classified as X and Y according to a set of standard criteria (Wolfe and Palmer 1998). For most cells, this included latencies to optic chiasm stimulation. Cells were also classified as lagged or nonlagged according to their responses to a spot covering both center and surround, which cycled between mean, bright, mean, and dark contrasts at 0.5 Hz (Saul and Humphrey 1990; Wolfe and Palmer 1998).
Cortical cooling
A silver-platinum plate was placed over a large portion of the visual cortex (areas 17 and 18), extending 5 mm down the medial bank, laterally 10 mm across the cortical surface, and from A2.0 to P9.0. The plate was cooled by a large Peltier device mounted on its superior surface. The other side of the Peltier device was cooled by water circulating through a water jacket. A microthermocouple (Omega Engineering) was inserted through a small hole in the plate to continuously monitor the temperature of the cortex. Temperature was measured at the tip of the probe, which was placed 2 mm from the surface, in layer VI of the cortex. For all of our experiments, we continuously recorded the temperature and cortical activity in layer VI.
Previous work shows that neuronal activity in cortex is abolished when the temperature reaches about 20°C (Lomber et al. 1999). We verified this by using extracellular recordings of multiunit activity. The orientation and spatial frequency selectivity of the activity were obtained before cooling of the cortex. Once layer VI was cooled to about 20°C, the activity ceased, but returned after warming of the cortex and retained its selectivity. In later experiments local field potentials and EEG were recorded in layer VI to monitor the cortical activity before, during, and after cooling (Fig. 1 A). For all of our experiments, the temperature in layer VI was maintained between 15 and 20°C. After extended periods of cooling or after many cooling cycles, activity resembling seizures was often observed in the EEG. After cooling, if activity did not return to levels recorded before cooling or, if seizures were seen, the experiment was terminated.
|
|
|
RESULTS |
|---|
|
Suppression at high spatial frequencies
For each LGN cell, the analysis began by acquiring a spatial summation curve with contrast-modulated, spatially homogeneous disks (0 cycles/deg, DC) centered on the receptive field. The radius of the disk was varied in pseudorandom order. Each pass consisted of five cycles presented at each of 26 sizes and five to ten passes constituted a data set. There was no interstimulus interval but the first cycle for each size was discarded. An example for an X-cell in lamina A of the LGN is shown in Fig. 2 A, where response amplitude (F1) is plotted versus the radius of the stimulus. As the stimulus increases in size, the response of the cell increases to a maximum. With further increases in radius, inhibition from the surround overtakes excitation from the center and the response decays reaching an asymptote at three to five times the radius at which the peak occurs.
|
 |
 |
 |
 |
Spatial frequencies above those to which the surround is responsive, referred to here as "high spatial frequencies," were selected for further examination. Spatial summation curves were obtained at high spatial frequencies by presenting drifting gratings centered on the receptive field with a background of mean luminance. The gratings drifted within a circular aperture whose radii were identical to those used with the homogeneous disk and were varied in pseudorandom order. A set of spatial summation curves obtained at high spatial frequencies is shown in Fig. 3 A for the same cell as that shown in Fig. 2. Two key features of the set of curves in Fig. 3A are noteworthy: 1) the radius of the stimulus eliciting the peak response ("optimal size") in the spatial summation curve decreased as the spatial frequency increased and 2) a strong reduction of the peak response was evident for large stimulus sizes at these high spatial frequencies. The response reduction in a given spatial summation curve was determined as
 |
|
Contributions from the receptive field center
As the spatial frequency of the drifting grating increases, the optimal size of the spatial summation curve decreases slightly. This effect seems to follow from the linearity of the center mechanism alone. Having extracted the parameters of the best-fitting DOG, we computed expected spatial summation curves by calculating the inner products of the center Gaussian spatial weighting function and the sinusoidal gratings at those spatial frequencies tested in each cell. The grating was assumed to have even symmetry, i.e., it was a cosine centered on the classical receptive field and all spatial frequencies were at or above the optimal value for each cell.
Examples of the predicted spatial summation curves for the same LGN X-cell at high spatial frequencies are shown in Fig. 3B. Response amplitude is plotted versus the radius of the grating patch in both cases. For any radius, the response falls with spatial frequency because the spatial frequencies used are all above the optimal. The optimal size of the spatial summation curve decreases and the inhibition evoked by the center mechanism increases with spatial frequency in the predictions. Population data are plotted in Fig. 3, D and E. Both effects are a consequence of the linear behavior of the center mechanism: as spatial frequency increases, the patch size where half-cycles of the "wrong" polarity (dark in an ON-center cell, bright in an OFF-center cell) are presented to the center mechanism decreases and the recruitment of these "wrong" half-cycles produces inhibition. Therefore the response reduction is at least partly the result of inhibition from the linear behavior of the center mechanism. Note that adding the classical surround to the computation (not shown) has no effect on either the optimal size or the response reduction because all gratings were at spatial frequencies above those to which the classical surround responds.
Measured and computed results are compared for the population of LGN cells in Fig. 3, G and H. Computation of optimal size from the center mechanism agrees well with measured values (Fig. 3G, P > 0.05, n = 20). In general, the observed response reduction was more than could be accounted for by inhibition from the classical center mechanism alone (Fig. 3H). Here the response reduction not accounted for by the center mechanism in LGN cells is plotted versus normalized spatial frequency. This is a measure of the suppression that results from stimulating the nonlinear component of the receptive field with a high spatial frequency grating. We find that this suppression tends to decrease with spatial frequency. To determine the suppression elicited from the nonlinear component in spatial summation curves, we used the following suppression calculation
 |
Sources of suppression at high spatial frequencies
RETINAL GANGLION CELLS (RGCS).
In principle, the suppression at high spatial frequencies could arise from at least three sources: the retinal input, processing intrinsic to the thalamus, or feedback from visual cortex. We examined the retinal input by simultaneously recording LGN cells and S-potentials, the extracellularly recorded excitatory postsynaptic potentials resulting from the firing of RGC input to the relay cell (Bishop et al. 1958; Cleland et al. 1971; Freygang 1958; Hubel and Wiesel 1961; Kaplan and Shapley 1984). Recording S-potentials allows us to simultaneously examine the receptive field properties of an LGN cell and one of the RGCs providing its input. Because the RGC and LGN cell receptive fields are accurately superimposed (Hubel and Wiesel 1961; Nolt et al. 2004) and have the same spatial frequency tuning (So and Shapley 1981), we acquired spatial summation curves simultaneously for both cells using a single stimulus.
Figure 4 A shows a set of spatial summation curves collected at high spatial frequencies for an RGC recorded simultaneously with the LGN cell in Figs. 2 and 3. In this example, the optimal size of the spatial summation curve decreased with spatial frequency and the response reduction increased with spatial frequency. These effects were also observed for our population data, shown in Fig. 4, B and C, where the optimal size of the spatial summation curve and response reduction are plotted versus normalized spatial frequency. We find that the population of RGCs exhibits the same behavior as the population of LGN cells: optimal size decreases and response reduction increases as the spatial frequency increases.
|
To estimate the amount of suppression at high spatial frequencies that RGCs contribute to LGN cells, we quantified the suppression in simultaneously recorded RGC–LGN cell pairs as shown in Fig. 5. Data from 16 pairs are plotted in the center panel; each pair was examined at multiple spatial frequencies, with spatial frequency normalized by the optimal frequency for each cell (x-axis). It is evident that, with a few exceptions, more suppression was found in LGN cells than in their RGC inputs, regardless of spatial frequency. From the box and whisker plot at the right it can be seen that, on average, the suppression in LGN cells was almost twice that for the RGCs [RGC = 29 ± 3% (SE); LGN = 56 ± 3%; P < 0.0001]. These data suggest that about half of the suppression seen in LGN cells at high spatial frequencies arises from their individual RGC inputs. Inhibition evoked by spatially homogeneous (DC) stimuli in LGN cells was also about twice that seen in their RGC inputs (left panel, RGC = 48 ± 3%; LGN = 89 ± 2%; P < 0.0001; Hubel and Wiesel 1961).
|
|
RGC data were also collected during the LGN cell recordings. Population data for RGCs and LGN cells are presented in Fig. 7 (6 RGC X-cells, 29 LGN X-cells). Inhibition before cooling is plotted versus inhibition during cooling for DC stimuli (Fig. 7, A and C) and suppression before cooling is plotted versus suppression during cooling for high spatial frequencies (Fig. 7, B and D). Not surprisingly, the RGCs do not show a significant difference in the amount of inhibition or suppression before versus during cooling for either the DC (Fig. 7A) or high spatial frequency (Fig. 7B) stimuli (P > 0.05 for both). For the population of LGN cells, there is significantly less suppression at high spatial frequencies during cooling of the visual cortex (Fig. 7D, P < 0.01), but inhibition at DC is not affected (Fig. 7C, P > 0.05). It is evident from the population data in Fig. 7 that cortical inactivation does not completely release the suppression observed at high spatial frequencies in LGN cells. We find that the cortical feedback is responsible, on average, for 27% of this suppression.
|
Spatial extent of suppression at high spatial frequencies
To determine the spatial locus of suppression at high spatial frequencies, we used a bipartite stimulus consisting of a small, central disk whose luminance was modulated sinusoidally in time and a surrounding annulus whose outer diameter was fixed between 5 and 10°. This annulus contained either a high spatial frequency drifting grating or a uniform field and its inner diameter was varied in pseudorandom order, at 10 equally spaced steps. The center stimulus was fixed in size at the optimal size determined from the DC spatial summation curve and elicited a constant response (F1).
The high spatial frequency annuli suppressed the center response in a size-dependent manner as shown in Fig. 8. These data are pooled from 14 LGN X-cells. Responses are normalized on the y-axis by the response to the center alone, shown as the horizontal line at 1.0. The inner radii of the suppressive annuli were normalized by the outer radius used for each cell. The high spatial frequency stimuli evoked only nonlinear suppression because they did not encroach on the center of the receptive field. Compared with high spatial frequency gratings, inhibition evoked by spatially homogeneous annuli caused a greater reduction of the response elicited by the center stimulus. Inhibition by the spatially homogeneous annuli arises predominantly from the classical surround, although there is evidence suggesting they may also evoke suppression from the nonlinear component of the receptive field (Bonin et al. 2005). Importantly, the suppression evoked from the nonlinear component of the receptive field by high spatial frequency gratings declines over the same spatial range as the inhibition from DC stimuli. Therefore we conclude that the region evoking suppression at high spatial frequencies is precisely colocalized with the classical surround.
|
To explore potential effects of suppression at high spatial frequencies on the gain of the classical receptive field, we obtained contrast-response functions with gratings limited to the receptive field center (radius = optimal size) in the presence and absence of a high spatial frequency grating that filled the surrounding region as defined earlier. The contrast of the center grating varied between 0 and 64%, whereas the surround grating was fixed at 30%. These measurements were made before, during, and after cortical cooling.
The results are summarized in Fig. 9 for a population of 15 LGN X-cells. Responses were normalized to their peaks and averaged (Fig. 9A). The contrast-response function of the center alone (
) was unaffected by cortical cooling (P > 0.05), and therefore the mean of the curves obtained before and during cooling is plotted for simplicity. On average, the high spatial frequency surround stimulus suppressed the response to the center stimulus by about half (
), except at the highest contrast. Cooling of the ipsilateral visual cortex (*) reduced the suppression at high spatial frequencies by about 30%. Because the center stimulus is the same in all cases, effects of the annuli arise from the action of the nonlinear component of the receptive field. On casual inspection, the curves seem to differ by a simple rotation about the origin, suggesting a divisive effect of the suppression at high spatial frequencies on the output of the neuron, also known as response gain control (Heeger 1992).
|
 |
is the semisaturation constant, or the contrast at which the response of the cell is half of its derived maximum (Rmax). If the nonlinear component of the receptive field was reducing only the Rmax, effectively dividing the output of the neuron, then the fitted curves should differ only in their Rmax values. In Fig. 9B, the three curves have been replotted after normalization by their respective Rmax values. The contrast-response function of the center alone () differs significantly from the other two curves, suggesting that the overall effect of the suppression at high spatial frequencies is not simply a change in the Rmax, and is therefore not response gain control alone. Suppression at high spatial frequencies is likely a combined effect of response gain control and contrast gain control. However, the contrast-response functions for the center plus surround stimulus before () and during (*) cooling are essentially identical after normalization by their respective Rmax values. Therefore these data suggest that the cortical component of the suppression elicits a change in Rmax alone, functioning only as a response gain control. Orientation tuning of suppression at high spatial frequencies
We examined the relative orientation dependence of the receptive field center and the surround at high spatial frequencies before and during inactivation of visual cortex. As shown in Fig. 10, we again used a bipartite stimulus to explore these issues. Both the center (at radius = optimal size) and surround stimuli were drifting, high spatial frequency gratings (Fig. 10A). The outer diameter of the surround stimulus was equal to the point where a plateau in the response was observed in the spatial summation curve and had either the same or orthogonal orientation relative to the center stimulus. The contrast of the surround was held constant at 30%, whereas the contrast of the center was varied between 0 and 64%, enabling us to obtain contrast-response functions for the center alone and for the center with both surround conditions.
|
), the center with the isooriented surround (), and the center with the orthogonally oriented surround (*) are plotted versus the center contrast. It is clear that the amount of suppression resulting from the two surround conditions is equivalent for all center contrasts. Suppression was calculated by comparing the response to the center stimulus alone with the response to the center stimulus with the appropriate surround condition. Inhibition evoked by the center mechanism is not subtracted from the response reduction in the suppression calculation because its effect is constant across all conditions. Therefore any observed response reduction was evoked by suppression from the nonlinear component of the receptive field. In Fig. 10C, orthogonally oriented suppression is plotted versus isooriented suppression for a population of 51 X-cells, with the center at 32% contrast. The points are grouped symmetrically about the line of unity (P > 0.05), indicating that, as a population, suppression was unmodified by changes in the relative orientation of the center and surround at high spatial frequencies. The same result was obtained for other contrasts of the center stimulus (data not shown). A similar result was obtained when the cortex was inactivated. This is shown in Fig. 10D, where the percentage suppression for isooriented and orthogonally oriented stimuli is compared during cortical cooling at multiple center contrasts for each cell. For the population, the two surround conditions show equal suppression (P > 0.05, n = 15). The iso- and orthogonally oriented stimuli show, on average across all center contrasts, 29 and 26% less suppression, respectively, compared with the suppression observed before cooling. The 3% difference between the two conditions is not statistically significant (P = 0.16). This release of suppression (27.5%) during cortical cooling is numerically indistinguishable from that observed in the spatial summation experiment described earlier.
Spatial frequency tuning curves
Given that the DOG model does not account for all of the response reduction at spatial frequencies greater than or equal to the optimal, the spatial frequency tuning curve of a cell should exhibit a lower peak and cutoff frequency than those derived from the DOG model. Spatial frequency tuning curves derived from IDOG fits to spatial summation curves obtained with DC stimuli were compared with those collected experimentally with drifting gratings. For each cell, the diameter of the drifting grating was at least equal to the diameter at which the response of the spatial summation curve reached a plateau. The two spatial frequency tuning curves for an example LGN X-cell are shown in Fig. 11A, normalized by their peak responses. In this example cell, both the peak and the cutoff frequencies of the measured curve are lower than those predicted from the IDOG fits. The cutoff is defined as the spatial frequency at which the tuning curve fell to half of the maximum response. Peak and cutoff frequencies for our population of cells are plotted in Fig. 11, B and C, with the measured values plotted versus the predicted values. In both figures, most points lie below the line of unity. For our population of cells, the peak and cutoff frequencies of the measured curves are significantly lower than those of the predicted curves (n = 28; peak: P = 0.013; cutoff: P = 0.002).
|
|
|
DISCUSSION |
|---|
|
Suppression at high spatial frequencies
There is now a general consensus that the responses of LGN cells are suppressed by image components whose spectral composition extends well beyond that to which the classical surround is responsive. Using various methodologies, this has been demonstrated in the LGN for cat X-cells (Bonin et al. 2005); primate magnocellular, parvocellular, and koniocellular cells (Solomon et al. 2002); and most recently in primate magnocellular-projecting RGCs (Solomon et al. 2006). Using spatial summation curves collected at multiple spatial frequencies above those to which the classical surround is responsive, we show three effects are manifested as spatial frequency increases: 1) peak response diminishes, 2) the radius of the stimulus eliciting the peak response decreases, and 3) the fractional reduction of the peak response increases. The linearity of the center mechanism accounts for the shift in the optimal size, but the magnitude of the response reduction exceeds that which can be attributed to the inhibition from the center mechanism alone.
Comparison with RGC input
All of the effects of high spatial frequency gratings apparent in LGN X-cells are also evident in their retinal inputs as identified by recording S-potentials (Bishop et al. 1958; Cleland et al. 1971; Freygang 1958; Hubel and Wiesel 1961; Kaplan and Shapley 1984) in conjunction with many of our LGN cells. However, the suppression at high spatial frequencies in RGCs is only about half that observed in their simultaneously recorded LGN cells. A similar result was obtained by Cleland et al. (1983) using moving bars of varying length. Our finding extends Hubel and Wiesel's original observation, that surround suppression is greater in the LGN than in the retina, to the entire range of spatial frequencies to which the cells respond.
We never saw more than a single S-potential in our simultaneous recordings. Thus we were not generally able to see all of the retinal input to a given LGN cell. Nevertheless, greater suppression at high spatial frequencies was seen in LGN cells even though the efficacy of the single retinal input ranged widely (37–85%; data not shown). This suggests that each individual RGC that drives an LGN cell is equally suppressed and that suppressive effects do not add at the level of the postsynaptic LGN cell. Some of the remaining suppression at high spatial frequencies in LGN cells depends on input from visual cortex.
Cortical inactivation
The effects of cortical feedback on the responses of LGN cells have been found to be excitatory (Baker and Malpeli 1977; Funke et al. 1996; Hull 1968; Kalil and Chase 1970; Przybyszewski et al. 2000), inhibitory (Cudeiro and Sillito 1996; Murphy and Sillito 1987; Sillito et al. 1993; Waleszczyk et al. 2005; Webb et al. 2002), or both (Marrocco et al. 1996; Waleszczyk et al. 2005). Numerous methodologies have been used to silence the corticothalamic projection, including pharmacological inactivation, ablation, and cooling of the visual cortex. The results of many of these studies are difficult to compare due to the differences in visual stimuli. Furthermore, some studies used ablation (e.g., Murphy and Sillito 1987), with which it is nearly impossible to collect data from the same LGN cell before and after inactivation of visual cortex.
During cortical cooling, we found a systematic reduction of the suppression elicited by high spatial frequency stimuli in LGN cells. Cooling never affected the responses of simultaneously recorded RGCs and also had no effect on the inhibition elicited in LGN cells by DC stimuli. Thus at the high spatial frequencies used in this study, the cortical input appears to be suppressive. We are unable to account for the great diversity in the results described by others. We did not observe a single instance where the cortical input appeared to be excitatory. We also never observed any phase change in the response of the cell during a cooling cycle, implying that the receptive fields did not move during cooling and that the phase of the response was not affected by inactivation of visual cortex.
The combined effects of RGCs and cortical feedback account for 78% of the suppression at high spatial frequencies observed in our population of LGN X-cells. Approximately half of the suppression observed at high spatial frequencies and half of the inhibition observed for spatially homogeneous (DC) stimuli in LGN cells are already present in the retinal input. Sources of the remaining suppression at high spatial frequencies and the remaining inhibition at DC are unknown. Possible sources are the local networks limited to the LGN (e.g., interneurons) and recurrent connections with the thalamic reticular nucleus.
Spatial extent
Surround suppression that is not attributable to the classical surround has been referred to as the "extra classical receptive field" (Alitto and Usrey 2003; Solomon et al. 2006; Webb et al. 2002), "extra classical inhibition" (Solomon et al. 2002), and, most recently, as arising from a "suppressive field" (Bonin et al. 2005). The term "extra" has taken on multiple meanings, from the region located beyond the classical receptive field to simply a reduction of the response elicited by a stimulus confined to the center that is not attributable to the inhibitory classical surround. We show here that suppression by gratings whose spatial frequencies exceed that to which the classical surround responds is colocalized with the classical surround. This is in agreement with the work of Bonin et al. (2005) who show that this is also true for low spatial frequencies. A possible exception to this colocalization principle is the periphery effect (McIlwain 1964), modulation from remote visual field stimulation, found in the retina and LGN of cats (Fischer and Kruger 1974; Kruger and Fischer 1973; McIlwain 1964), and, to a lesser degree, in primates (Kruger 1977; Kruger et al. 1975; Solomon et al. 2006). The periphery effect is predominantly found in cells that respond transiently and with short latencies: Y-cells in the cat retina and LGN (Cleland et al. 1971; Ikeda and Wright 1972; McIlwain 1964) and M-cells in the primate (Kruger 1977; Solomon et al. 2006).
Gain control
High spatial frequency gratings in the surround of LGN cells reduced the gain of stimuli limited to the receptive field center. The suppressive action of high spatial frequency gratings appears to reduce the gain in a complex manner, likely a combination of response gain control and contrast gain control. However, the cortical component of the suppression elicits a change in Rmax alone and therefore acts only as a response gain control (Heeger 1992). Solomon et al. (2002) also found that high spatial frequency stimuli in an annular grating placed beyond the classical receptive field divisively reduced responses elicited from the classical receptive field, but they did not inactivate cortex. Webb et al. (2002) found no effect of cortical ablation on the gain of the classical receptive field, but their stimuli were limited to the optimal spatial frequency and responses of the same cells were not recorded before and after ablation.
Our results are in essential agreement with Bonin et al. (2005), although they did not inactivate cortex and did not allow for feedback contributions such as those that we have identified. They observed suppression at spatial frequencies above those to which the classical surround is responsive and our data suggest that this suppression arises from both feedforward and feedback mechanisms.
Our observations are not in agreement with those of Przybyszewski et al. (2000). They observed diminished responses from LGN cells during cortical inactivation and suggested that the visual cortex multiplicatively enhances responses in the primate LGN.
We cannot account for this discrepancy. The stimulus used by Przybyszewski et al. (2000) may not have encompassed the entire classical receptive field, although we did not see a single instance of decreased LGN responses at any stimulus size during cooling. It seems unlikely that it can be accounted for by species differences, although we cannot rule out this possibility.
Orientation selectivity
We found that suppression of LGN X-cell responses by high spatial frequencies was independent of the relative orientation of the center and surround gratings. This is diametrically opposed to the findings of Cudeiro and Sillito (1996) who described stronger suppression when the center and annular gratings were isooriented than when they were orthogonal. Our methodologies appear similar, although they argued that this differential suppression was more evident at low spatial frequencies. However, recent studies in cat and primate (Bonin et al. 2005; Solomon et al. 2002; Webb et al. 2002) also failed to find any effect of relative orientation at any spatial frequency. For the moment, no obvious explanation for this discrepancy presents itself.
Extending the DOG model
We show that the peak and cutoff spatial frequencies obtained experimentally are lower than those derived from the DOG model. In many studies, spatial frequency tuning curves have been fit with a DOG function and the extracted parameters used to establish the characteristics of the center and surround of the classical receptive field (e.g., Enroth-Cugell et al. 1983). Our results show that spatial frequency tuning curves include suppression at high spatial frequencies that is not accounted for by the DOG model, thus calling into question the validity of fitting spatial frequency tuning curves with a DOG function.
Bonin et al. (2005) proposed a model for RGC and LGN cell receptive fields that combines the classical DOG mechanism with a suppressive field that includes a Gaussian-weighted bank of filters. This model has been shown to account for many aspects of RGC and LGN cell receptive fields. The authors suggested that the nonlinear component of the receptive field is suppressive at all spatial frequencies. As a result, the spatial summation curves collected using a spatially homogeneous stimulus would overestimate the strength and underestimate the size of the classical surround. This suggests that the classical surround is weaker and larger than we predicted and therefore the peak of the predicted curve in Fig. 11A would shift to lower frequencies, more closely matching the measured spatial frequency tuning curve. However, the cutoff of the predicted tuning curve would not be affected by a weaker and larger classical surround because the classical surround is not tuned to high spatial frequencies. The differences in measured and predicted cutoffs, although small, suggest that the role of suppression at high spatial frequencies is to restrict the response to large stimuli of high spatial frequencies.
Bonin et al. (2005) also showed that the action of the suppressive field is stronger at low spatial frequencies and high temporal frequencies, both incommensurate with cells in visual cortex, and therefore suggest that the suppressive field does not contain a cortical component. Our findings show that a cortical contribution to the suppressive field must be integrated into the model because cortical feedback modulates the gain of LGN relay cells, thereby contributing to the reduced response observed when receptive fields are stimulated with high spatial frequency gratings in the surround.
Circuitry
It is well established that outputs from layer VI of visual cortex make both direct, excitatory glutamatergic synapses on the distal dendrites of LGN relay cells and indirect, inhibitory connections by way of the thalamic reticular nucleus (TRN) (Ahlsen et al. 1982; Descheenes and Hu 1990; Montero 1989, 1990; Robson 1983). Weak excitatory influences of this cortical input have been inferred in several experiments (Baker and Malpeli 1977; Funke et al. 1996; Hull 1968; Kalil and Chase 1970; Przybyszewski et al. 2000; Waleszczyk et al. 2005; Worgotter et al. 1998), including decreased spontaneous activity during cortical inactivation (Marrocco et al. 1996; Waleszczyk et al. 2005). We found no evidence for an excitatory effect of cortical inputs to LGN X-cells and, as observed by Przybyszewski et al. (2000) and Webb et al. (2002), we found spontaneous activity to be unchanged during cortical cooling.
We found that suppression by high spatial frequency gratings was diminished during cortical inactivation. This inhibitory effect of cortical input to the LGN is probably mediated by indirect action through the TRN. It is noteworthy that Golshani et al. (2001) found that the cortical input to the TRN is stronger than the cortical input to LGN relay neurons, suggesting that the indirect inhibitory input to the LGN may exert more influence on the responses of LGN cells than the direct excitatory input. It is also likely that the suppression not accounted for by either feedforward retinal input or feedback from the cortex is mediated by recurrent inhibition through the TRN. We also note that our experiments were conducted under anesthesia and it is possible that other feedback effects could be observed in experiments conducted in the awake and behaving state. For example, experiments in awake animals may be required to demonstrate a functional role for the excitatory component of the corticothalamic projection. O'Connor et al. (2002) also showed that attention can modulate neural activity in the human LGN, presenting another possibility for the role of the corticothalamic projection that could be tested only in an awake animal.
|
|
GRANTS |
|---|
|
|
|
ACKNOWLEDGMENTS |
|---|
|
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: M. J. Nolt, University of Pennsylvania School of Medicine, Department of Neuroscience, 421 Curie Blvd., 1127 BRB II/III, Philadelphia, PA 19104 (E-mail: nolt{at}mail.med.upenn.edu)
|
|
REFERENCES |
|---|
|
Alitto HJ, Usrey WM. Corticothalamic feedback and sensory processing. Curr Opin Neurobiol 13: 440–445, 2003.[CrossRef][Web of Science][Medline]
Baker FH, Malpeli JG. Effects of cryogenic blockade of visual cortex on the responses of lateral geniculate neurons in the monkey. Exp Brain Res 29: 433–444, 1977.[Web of Science][Medline]
Barlow HB. Summation and inhibition in the frog's retina. J Physiol 119: 69–88, 1953.
Bishop PO, Burke W, Davis R. Synapse discharge by single fibre in mammalian visual system. Nature 182: 728–730, 1958.[Medline]
Bonin V, Mante V, Carandini M. The suppressive field of neurons in lateral geniculate nucleus. J Neurosci 25: 10844–10856, 2005.
Cleland BG, Dubin MW, Levick WR. Sustained and transient neurones in the cat's retina and lateral geniculate nucleus. J Physiol 217: 473–496, 1971.
Cleland BG, Lee BB, Vidyasagar TR. Response of neurons in the cat's lateral geniculate nucleus to moving bars of different length. J Neurosci 3: 108–116, 1983.[Abstract]
Cudeiro J, Sillito AM. Spatial frequency tuning of orientation-discontinuity-sensitive corticofugal feedback to the cat lateral geniculate nucleus. J Physiol 490: 481–492, 1996.
Descheenes M, Hu B. Electrophysiology and pharmacology of the corticothalamic input to lateral thalamic nuclei: an intracellular study in the cat. Eur J Neurosci 2: 140–152, 1990.[CrossRef][Web of Science][Medline]
Enroth-Cugell C, Robson JG, Schweitzer-Tong DE, Watson AB. Spatio-temporal interactions in cat retinal ganglion cells showing linear spatial summation. J Physiol 341: 279–307, 1983.
Fischer B, Kruger J. The shift-effect in the cat's lateral geniculate neurons. Exp Brain Res 21: 225–227, 1974.[Web of Science][Medline]
Freygang WH Jr. An analysis of extracellular potentials from single neurons in the lateral geniculate nucleus of the cat. J Gen Physiol 41: 543–564, 1958.
Funke K, Nelle E, Li B, Worgotter F. Corticofugal feedback improves the timing of retino-geniculate signal transmission. Neuroreport 7: 2130–2134, 1996.[Web of Science][Medline]
Golshani P, Liu XB, Jones EG. Differences in quantal amplitude reflect GluR4-subunit number at corticothalamic synapses on two populations of thalamic neurons. Proc Natl Acad Sci USA 98: 4172–4177, 2001.
Heeger DJ. Normalization of cell responses in cat striate cortex. Vis Neurosci 9: 181–197, 1992.[Web of Science][Medline]
Hubel DH, Wiesel TN. Integrative action in the cat's lateral geniculate body. J Physiol 155: 385–398, 1961.
Hull EM. Corticofugal influence in the macaque lateral geniculate nucleus. Vision Res 8: 1285–1298, 1968.[CrossRef][Web of Science][Medline]
Ikeda H, Wright MJ. Functional organization of the periphery effect in retinal ganglion cells. Vision Res 12: 1857–1879, 1972.[CrossRef][Web of Science][Medline]
Jones JP, Palmer LA. The two-dimensional spatial structure of simple receptive fields in cat striate cortex. J Neurophysiol 58: 1187–1211, 1987.
Kalil RE, Chase R. Corticofugal influence on activity of lateral geniculate neurons in the cat. J Neurophysiol 33: 459–474, 1970.
Kaplan E, Shapley R. The origin of the S (slow) potential in the mammalian lateral geniculate nucleus. Exp Brain Res 55: 111–116, 1984.[Web of Science][Medline]
Kruger J. Stimulus dependent colour specificity of monkey lateral geniculate neurones. Exp Brain Res 30: 297–311, 1977.[Web of Science][Medline]
Kruger J, Fischer B. Strong periphery effect in cat retinal ganglion cells. Excitatory responses in ON- and OFF-center neurones to single grid displacements. Exp Brain Res 18: 316–318, 1973.[Web of Science][Medline]
Kruger J, Fischer B, Barth R. The shift-effect in retinal ganglion cells of the rhesus monkey. Exp Brain Res 23: 443–446, 1975.[Web of Science][Medline]
Kuffler SW. Discharge patterns and functional organization of mammalian retina. J Neurophysiol 16: 37–68, 1953.
Levick WR, Cleland BG, Dubin MW. Lateral geniculate neurons of cat: retinal inputs and physiology. Invest Ophthalmol 11: 302–311, 1972.
Lomber SG, Payne BR, Horel JA. The cryoloop: an adaptable reversible cooling deactivation method for behavioral or electrophysiological assessment of neural function. J Neurosci Methods 86: 179–194, 1999.[CrossRef][Web of Science][Medline]
Marrocco RT, McClurkin JW, Alkire MT. The influence of the visual cortex on the spatiotemporal response properties of lateral geniculate nucleus cells. Brain Res 737: 110–118, 1996.[CrossRef][Web of Science][Medline]
McIlwain JT. Receptive fields of optic tract axons and lateral geniculate cells: peripheral extent and barbiturate sensitivity. J Neurophysiol 27: 1154–1173, 1964.
Montero VM. Ultrastructural identification of synaptic terminals from cortical axons and from collateral axons of geniculo-cortical relay cells in the perigeniculate nucleus of the cat. Exp Brain Res 75: 65–72, 1989.[Web of Science][Medline]
Montero VM. Quantitative immunogold analysis reveals high glutamate levels in synaptic terminals of retino-geniculate, cortico-geniculate, and geniculo-cortical axons in the cat. Vis Neurosci 4: 437–443, 1990.[Web of Science][Medline]
Murphy PC, Sillito AM. Corticofugal feedback influences the generation of length tuning in the visual pathway. Nature 329: 727–729, 1987.[CrossRef][Medline]
Nolt MJ, Kumbhani RD, Palmer LA. Contrast-dependent spatial summation in the lateral geniculate nucleus and retina of the cat. J Neurophysiol 92: 1708–1717, 2004.
Nolt MJ, Kumbhani RD, Palmer LA.Visual processing in the lateral geniculate nucleus of the cat. Program No. 506.13. 2005 Abstract Viewer and Itinerary Planner. Washington, DC: Society for Neuroscience, 2005. Online.
O'Connor DH, Fukui MM, Pinsk MA, Kastner S. Attention modulates responses in the human lateral geniculate nucleus. Nat Neurosci 5: 1203–1209, 2002.[CrossRef][Web of Science][Medline]
Przybyszewski AW, Gaska JP, Foote W, Pollen DA. Striate cortex increases contrast gain of macaque LGN neurons. Vis Neurosci 17: 485–494, 2000.[CrossRef][Web of Science][Medline]
Reid RC, Victor JD, Shapley RM. The use of m-sequences in the analysis of visual neurons: linear receptive field properties. Vis Neurosci 14: 1015–1027, 1997.[Web of Science][Medline]
Robson JA. The morphology of corticofugal axons to the dorsal lateral geniculate nucleus in the cat. J Comp Neurol 216: 89–103, 1983.[CrossRef][Web of Science][Medline]
Rodieck RW. Quantitative analysis of cat retinal ganglion cell response to visual stimuli. Vision Res 5: 583–601, 1965.[CrossRef][Medline]
Saul AB, Humphrey AL. Spatial and temporal response properties of lagged and nonlagged cells in cat lateral geniculate nucleus. J Neurophysiol 64: 206–224, 1990.
Sillito AM, Cudeiro J, Murphy PC. Orientation sensitive elements in the corticofugal influence on centre-surround interactions in the dorsal lateral geniculate nucleus. Exp Brain Res 93: 6–16, 1993.[Web of Science][Medline]
So YT, Shapley R. Spatial tuning of cells in and around lateral geniculate nucleus of the cat: X and Y relay cells and perigeniculate interneurons. J Neurophysiol 45: 107–120, 1981.
Solomon SG, Lee BB, Sun H. Suppressive surrounds and contrast gain in magnocellular-pathway retinal ganglion cells of macaque. J Neurosci 26: 8715–8726, 2006.
Solomon SG, White AJR, Martin PR. Extraclassical receptive field properties of parvocellular, magnocellular, and koniocellular cells in the primate lateral geniculate nucleus. J Neurosci 22: 338–349, 2002.
Waleszczyk WJ, Bekisz M, Wrobel A. Cortical modulation of neuronal activity in the cat's lateral geniculate and perigeniculate nuclei. Exp Neurol 196: 54–72, 2005.[CrossRef][Web of Science][Medline]
Webb BS, Tinsley CJ, Barraclough NE, Easton A, Parker A, Derrington AM. Feedback from V1 and inhibition from beyond the classical receptive field modulates the responses of neurons in the primate lateral geniculate nucleus. Vis Neurosci 19: 583–592, 2002.[CrossRef][Web of Science][Medline]
Wolfe J, Palmer LA. Temporal diversity in the lateral geniculate nucleus of cat. Vis Neurosci 15: 653–675, 1998.[CrossRef][Web of Science][Medline]
Worgotter F, Nelle E, Li B, Funke K. The influence of corticofugal feedback on the temporal structure of visual responses of cat thalamic relay cells. J Physiol 509: 797–815, 1998.
This article has been cited by other articles:
|
|
 |
|
  A. J. Camp, C. Tailby, and S. G. Solomon Adaptable Mechanisms That Regulate the Contrast Response of Neurons in the Primate Lateral Geniculate Nucleus J. Neurosci., April 15, 2009; 29(15): 5009 - 5021. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C.-I Yeh, C. R. Stoelzel, C. Weng, and J.-M. Alonso Functional Consequences of Neuronal Divergence Within the Retinogeniculate Pathway J Neurophysiol, April 1, 2009; 101(4): 2166 - 2185. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. L. Passaglia, D. K. Freeman, and J. B. Troy Effects of Remote Stimulation on the Modulated Activity of Cat Retinal Ganglion Cells J. Neurosci., February 25, 2009; 29(8): 2467 - 2476. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
