| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Group in Vision Science, School of Optometry, Helen Wills Neuroscience Institute, University of California, Berkeley, California
Submitted 21 April 2006; accepted in final form 11 July 2006
 |
ABSTRACT |
|---|
|
|
|
INTRODUCTION |
|---|
|
, 1987). The mask does not have to be perpendicular as it suppresses response to the test grating over a wide range of orientations (Bonds 1989; DeAngelis et al. 1992; Morrone et al. 1982). This cross-orientation suppression (COS) has been studied extensively, and it has generally been thought to be caused by intracortical inhibition (Blakemore and Tobin 1972; Bonds 1989; Carandini and Heeger 1994; Carandini et al. 1997; DeAngelis et al. 1992; Heeger 1992; Morrone et al. 1982, 1987; Sengpiel et al. 1998). The mechanism of COS has functional implications because it may play a role in important properties of visual cortical neurons, such as the refinement of orientation selectivity (Carandini and Ringach 1997; Chapman and Stryker 1992; Lauritzen et al. 2001; Somers et al. 1995; Vidyasagar et al. 1996), and spatial frequency selectivity (Bauman and Bonds 1991), and as a component of contrast normalization (Carandini and Heeger 1994; Heeger 1992).
Because COS can be induced by a wide range of mask orientations and spatial frequencies, and cells in the LGN are also broadly tuned for these parameters, it is plausible that the mechanism for COS originates in the LGN (Bonds 1989; DeAngelis et al. 1992; Walker et al. 1998). However, this explanation is not consistent with the experimentally shown lack of COS in LGN cells (Bauman and Bonds 1991) and the fact that geniculate afferents to the visual cortex seem to be exclusively excitatory (Creutzfeldt and Ito 1968; Garey and Powell 1971; Tanaka 1985). Based on these arguments, several studies have concluded that the source of COS is intracortical inhibition from a pool of cortical neurons exhibiting a wide range of orientation and spatial frequency preferences (Bauman and Bonds 1991; Bonds 1989; DeAngelis et al. 1992; Heeger 1992; Morrone et al. 1982; Sengpiel et al. 1998; Walker et al. 1998).
In recent work, synaptic depression at thalamo-cortical synapses has been proposed as the mechanism underlying COS in the visual cortex (Carandini et al. 2002; Freeman et al. 2002; Mrsic-Flogel and Hubener 2002). Synaptic depression is a form of short-term synaptic plasticity in which the efficacy of a given synapse is temporarily reduced according to its recent activity level (Abbott and Nelson 2000; Chance et al. 1998). However, there are temporal aspects of synaptic depression in relation to COS that require examination. In the study reported here, we studied the recovery time course of COS and find it inconsistent with typical temporal properties of synaptic depression. This makes it unlikely that synaptic depression is a major mechanism underlying COS in the visual cortex.
As an alternative to the synaptic depression hypothesis, we propose a feedforward mechanism for COS, in which suppression originates in the LGN through a rectifying nonlinearity and contrast saturation, and that COS is strengthened in the visual cortex by a spike threshold nonlinearity. Our measurements of contrast saturation in LGN are consistent with this notion. After the conclusion of our experimental work, two relevant publications have appeared. In one, intracellular techniques were used to study COS (Priebe and Ferster 2006). The main finding was that cross-oriented stimuli suppressed both synaptic inhibition and synaptic excitation. A feedforward model provided a prediction of COS. In the second study, COS was found to be rapid. From other temporal measurements, COS was suggested to be caused by feedforward signals or rapid intracortical neuronal paths (Smith et al. 2006).
|
|
METHODS |
|---|
|
All procedures complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Extracellular recordings are made with epoxy coated tungsten microelectrodes (A-M Systems) in the striate cortex of anesthetized and paralyzed mature cats (5–8 mo, 2.5–3.5 kg). Anesthesia is induced with thiopental sodium intravenously and maintained at an appropriate rate determined individually for each cat. A tracheal cannula is positioned and the animal is artificially ventilated (25% O2-75% N2O) at a rate adjusted to maintain expired CO2 at 4–5%. Temperature is maintained at 38°C. After the anesthetic level is stabilized at a constant rate of thiopental sodium, the cat is paralyzed with an intravenous infusion of pancuronium bromide (0.2 mg/kg/h). EEG, ECG, heart rate, temperature, and end-tidal CO2 are monitored during the experiment. A craniotomy is performed over area 17 and LGN. The dura is resected and covered with agar and wax to prevent drying and to reduce pulsation. For area 17 recording, electrode penetrations are made along the medial bank of the postlateral gyrus, 4 mm posterior and 2 mm lateral from the Horsley-Clarke origin (Horsley and Clarke 1908) at an angle of 10° medial and 20° anterior. For LGN recording, vertical electrode penetrations are made 6 mm anterior and 9 mm lateral from H-C zero. Small position adjustments of our electrodes are made to enter the LGN near the representation of the area centralis.
Extracellular recordings
Single units are isolated by the shapes of their spike waveforms. Initial estimates of each neuron's tuning parameters are obtained qualitatively using computer-controlled manipulation of drifting sinusoidal gratings. Quantitative measurements of tuning functions for orientation, spatial frequency, temporal frequency, size, ocular dominance, and stimulus contrast are performed. Response amplitude is taken as the mean firing rate for complex cells or as the mean amplitude of the first harmonic of the response for simple cells and LGN cells. Because contrast gain and dynamic response range are apparently similar for X and Y cells (Hartveit and Heggelund 1992), we did not use this classification system for our LGN cell sample. The stimuli are presented to the dominant eye (responsive eye for LGN cells), whereas the nondominant eye (unresponsive eye for LGN cells) views a blank CRT screen of the same mean luminance as the gratings.
Recovery time course of COS
After the initial characterization of a cell in area 17, we examined the responses to an optimal test stimulus before, during, and after the presentation of an orthogonal mask, as shown in Fig. 1. Test and mask stimuli are sinusoidal drifting gratings at optimal and orthogonal orientations, respectively. The temporal frequencies of the mask stimuli are predetermined through a separate protocol (Allison et al. 2001; Freeman et al. 2002; Li et al. 2005). In brief, the temporal frequency of the test grating is fixed to the optimal for the cell while that of the mask is varied from 0.5 to 25 Hz. The temporal frequency that elicits maximum suppression is used for the mask stimulus in this experiment. For our population of cells, 4–10 Hz produced maximum COS values. To examine the recovery time-course of COS, we measured the response to the test stimulus after presentation of a 0.5-s mask. After the mask is presented, five delay periods are used before the test stimulus (0, 0.5, 1, 2, and 4 s as depicted in Fig. 1C). We also measured responses to test-only (Fig. 1A) and to test plus mask (i.e., plaid) conditions (Fig. 1B). Presentations are randomized with 10-s interstimulus intervals. Blank trials (i.e., no stimulus) are also interleaved with stimulus presentations. The contrast of both the test and mask is 50%. The entire condition set is repeated 10–20 times, and responses are averaged across trials.
|
|
|
RESULTS |
|---|
|
Recovery time-course of COS
To consider synaptic depression as a principal mechanism of COS, we first note a predicted recovery time-course that this process would require. Specifically, responses to the test stimulus after a mask should recover exponentially toward the test-only response condition (Abbott et al. 1997; Chance et al. 1998; Varela et al. 1997). This prediction is quantified as
 | (1) |
; Chance et al. 1998; Varela et al. 1997).
Reported time constants for a fast form of synaptic depression are generally between 200 and 600 ms (Abbott et al. 1997; Chance et al. 1998; Varela et al. 1997). In our analysis, we use these values (i.e., T = 200 and 600 ms) as upper and lower bounds to predict the responses of cortical cells. To make these predictions suitable for our measurements, we averaged R(t) across the first 500 ms of each response.
Figure 2 A shows data from a representative cortical simple cell where the responses are plotted as a function of the delay between mask and test stimuli. The top and bottom horizontal dashed lines represent responses to the test and plaid, respectively. Dotted lines above and below the dashed lines represent ±SE of the mean value. All responses represent the average neural spike rate during the first 500 ms of the test stimulus. This example cell exhibits clear suppression to the plaid stimulus. The mean responses to the test (top dashed line) and plaid (bottom dashed line) are 42.8 and 23.9 spikes/s, respectively, i.e., there is a 44% response reduction when the mask is superimposed on the test stimulus. If the mechanism underlying this suppression was synaptic depression, we would expect the response to the test stimulus to be reduced for some period after removal of the mask because synaptic depression is known to have a prolonged recovery time (Abbott et al. 1997; Chance et al. 1998; Varela et al. 1997, 1999). However, responses to the test grating for all temporal delay conditions (0, 0.5, 1, 2, or 4 s, empty circles) are nearly identical to those of the test-only condition (top dashed line).
|
100%). This applies equally to both simple and complex cells (P = 0.5, t-test). A model of synaptic depression predicts percent recovery values between 60 and 30% (up and down triangles) for assumed time constants of 200 and 600 ms (Abbott et al. 1997; Chance et al. 1998; Varela et al. 1997), respectively. These results suggest that synaptic depression processes are too slow to account for the rapid recovery from a briefly presented mask stimulus. The data presented in Fig. 2, A–C, depict response magnitudes averaged over a 0.5-s time window. However, this analysis may be insensitive to rapid changes mediated by especially fast forms of short-term synaptic depression. To examine this possibility, we compare poststimulus time histograms (PSTHs) for the following three stimulus conditions: 1) the test-only stimulus (Fig. 2, D and E, empty circles), 2) the test stimulus after the mask (0-s delay condition; Fig. 2, D and E, filled circles), and 3) the plaid stimulus (Fig. 2, D and E, empty squares). Only responses from complex cells are included in this analysis because the onsets of simple cell responses are related to the phase of the drifting grating and are therefore not suitable for resolving fast response onset times. Results for a representative complex cell are given in Fig. 2D, and the population averaged responses are shown in Fig. 2E. Consistent with the above analysis, there is no significant difference between the test-only response time course (empty circles) and the test after mask response time-course (filled circles). This is the case even during the first 100 ms of the response. For comparison, the response to the plaid stimulus is clearly suppressed at these early time-points. If COS in visual cortex were mediated by synaptic depression, we would expect a mask stimulus to induce suppression during the initial part of a subsequent test stimulus. However, there is no evidence of this type of suppression.
Finally, we note that studies of cells in visual cortex in which stimuli have been presented in random sequence show that firing rate increases have relatively long latencies (
20 ms) compared with decreases (Bair et al. 2002; Smith et al. 2001). This delay has been cited as evidence in support of a model of synaptic depression (Carandini et al. 2002; Freeman et al. 2002). However, 20 ms is substantially shorter than the recovery times for synaptic depression typically reported in the literature (Abbott et al. 1997; Chance et al. 1998; Varela et al. 1997, 1999). Furthermore, the origin of this delay is not clear because a number of different mechanisms could cause a delay between the onset and offset of firing rates (Bair et al. 2002).
Linearity and the LGN
The analysis presented above shows that synaptic depression at thalamo-cortical synapses is unlikely to be the principal mechanism for COS in visual cortex. Other work on the temporal and adaptation properties of COS suggests that the primary mechanism underlying COS operates before the formation of cortical receptive fields (RFs) (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). Therefore COS in visual cortex is likely to originate from the LGN itself, rather than thalamo-cortical synapses. Previous studies of COS have generally overlooked the response nonlinearity of neurons in the LGN (Bonds 1989; Carandini and Heeger 1994; DeAngelis et al. 1992; Sengpiel et al. 1998; Walker et al. 1998).There are two major nonlinear inputs from LGN cells to cortical simple cells: spike rectification and contrast saturation in the LGN. The former refers to the fact that LGN responses cannot be decreased below zero and the latter obtains because LGN cells do not generally respond proportionally to stimulus contrast. Although contrast saturation in LGN neurons is less pronounced than that in visual cortex (Ohzawa et al. 1985; Sclar et al. 1990), an assumption of linearity is not warranted. To determine the degree of nonlinearity, we measured contrast response functions for a population of cells in the LGN (n = 74) for a wide range of contrast levels (1, 2, 4, 7, 14, 26, 50, and 100%).
Figure 3 A shows an example LGN contrast response function that exhibits a modest degree of saturation (empty circles). The firing rates at 50 and 100% contrast are 86 and 91 spikes/s, respectively. A neuron without contrast saturation would exhibit a substantially larger response at 100% contrast (172 spikes/s) based on its response to the 50% contrast stimulus. To quantify the effects of contrast saturation, we calculated a contrast saturation index (CSI). This index is based on neural responses to two successive and adjacent contrast values that are used to determine contrast response functions. The lower of two successive levels is used to predict responses to the higher contrasts. This prediction, based on the assumption of a linear contrast response function, is compared with the measured response. CSI values of 0 and 0.5 represent, respectively, linear and full contrast saturation. Formally, CSI is defined as
 | (2) |
|
0.05, t-test). Consistent with the example cell for which data are shown in Fig. 3A, the strength of contrast saturation increases approximately monotonically with stimulus contrast. Predicted suppression through a feedforward model
The experimental data we present here show clearly that COS in visual cortex recovers rapidly after the presentation of a mask stimulus (Fig. 2) and that most neurons in the LGN exhibit some degree of contrast saturation (Fig. 3). The first result suggests that synaptic depression is unlikely to be the principal source of COS in the visual cortex. The second finding indicates that contrast saturation at the LGN level may play an important role in COS. To explore this possibility, we use a physiologically plausible model to propose that COS arises from response nonlinearities of LGN cells and is strengthened in visual cortex by spike threshold nonlinearities.
Our model follows the serial processing notion by which orientation selectivity is derived from excitatory LGN input (Hubel and Wiesel 1962; Reid and Alonso 1995). It follows closely the construction of a previous model used to explore synaptic depression in visual cortex (Carandini et al. 2002). The cortical simple cell RF is determined by thalamocortical excitaiton. ON and OFF cortical RF subregions arise from ON-center and OFF-center LGN cells. The excitation is accompanied by subtractive inhibition of a "push-pull" form. Both excitation and inhibition are included through ON-center and OFF-center neurons. Although intracortical inhibition could be included in this construction (Palmer and Davis 1981; Troyer et al. 1998), our approach is for inhibition to be derived exclusively from LGN. In our model, LGN cells operate linearly but have an added nonlinearity. The nonlinearity is established by assuming a resting spike rate of LGN cells of 10 spikes/s (Carandini et al. 2002) and a contrast saturation component as conveyed in Fig. 3. For each LGN cell in our model, the contrast saturation nonlinearity is estimated by using the mean normalized contrast response function from all LGN cells (n = 74). This contrast response function is well fit (the goodness of fit Rsquare = 0.99) with a hyperbolic ratio function, R = 1.0 x C/(C + C50), where C and C50 represent stimulus contrast and the contrast level that elicts half of maximum response, respectively. The estimated value of C50 from our population of LGN cells is 27.8%. Consistent with the model in the previous study (Carandini et al. 2002), we assume that the maximum modulation response of LGN cells is 100 spikes/s. However, in our model, thalamo-cortical synapses operate in a linear manner, i.e., synaptic depression is not included. We also assume that there is a spiking mechanism for cortical cells and that the firing rate is a rectified version of the membrane potential with a threshold of 5 spikes/s (Anderson et al. 2000; Carandini and Ferster 2000; Carandini et al. 2002).
In our model, depicted in Fig. 4 A, summation of LGN inputs occurs in a "push-pull" manner. In this configuration, excitation from an ON-center LGN cell and inhibition from an OFF-center LGN cell form an ON subregion. Likewise, excitation from an OFF-center and inhibition from an ON-center LGN cell form an OFF subregion (Carandini and Heeger 1994; Carandini et al. 1997; Glezer et al. 1982; Tolhurst and Dean 1990). Although inhibition is thought to operate through cortical inhibitory interneurons (Palmer and Davis 1981; Troyer et al. 1998), for our present purpose, it is assumed to come directly from LGN cells. This assumption is for simplicity. Intracortical inhibitory connections included in other models (Albrecht and Geisler 1991; Ben-Yishai et al. 1995; Carandini and Heeger 1994; Carandini and Ringach 1997; Heeger 1992; McLaughlin et al. 2000, 2003; Shelley and McLaughlin 2002; Shelley et al. 2002; Wielaard et al. 2001) are not considered in this analysis because they do not seem to be relevant to COS (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). We assume that test and mask stimuli are sinusoidal drifting gratings that are vertical and horizontal, respectively. The contrast, duration, and temporal frequency of both test and mask stimuli are 50%, 2 s, and 4 Hz, respectively. A plaid is formed by the superposition of these gratings. The model simple cell responds to the test stimulus but not the mask. If LGN cells respond linearly to a visual stimulus, a simultaneous presentation of the mask will not affect the synaptic input to the model simple cell. However, with spike rectification and contrast saturation in the LGN, synaptic input to the model simple cell is assumed to be reduced by the mask.
|
As shown above, contrast saturation is present in LGN neurons (Fig. 3). We applied the contrast response nonlinearity (spike rectification and contrast saturation) for each LGN cell in our model to predict the level of COS in visual cortex. Normalized responses of the model cortical cell are shown in Fig. 4C. The two bars labeled test and mask represent, respectively, the normalized responses of the model cortical neuron to test and mask presentations at 50% contrast levels. The bar labeled predicted represents the predicted responses to the plaid based on a population of LGN neurons exhibiting the mean contrast saturation observed in this study (CSI = 0.33). In this case, the predicted response is 0.60, i.e.,the plaid suppresses the response of the cell by 40%. This represents a strength of suppression that is 7% higher than the mean contrast saturation (0.33) we observe in LGN. This difference may be accounted for by a rectifying nonlinearity of LGN cells and a spike output nonlinearity of cortical cells. The fourth bar in Fig. 4C, labeled measured, represents the mean response to the plaid, which is normalized by the response to the test-only condition averaged across the entire 2-s stimulus duration (n = 36, see Fig. 2B). In this case, we observe a response reduction of 49% relative to the test-only condition. This is close to the predicted strength of COS (40%) from the model cortical cell and consistent with our previous report of COS (Walker et al. 1998). Therefore if we incorporate a rectifying nonlinearity for LGN cells and a spike threshold nonlinearity for cortical cells, contrast saturation in the LGN can account for most COS in cortical cells for high contrast conditions.
Because the CSI in LGN increases with stimulus contrast (Fig. 3C), our model predicts that the level of suppression increases with contrast. To test this prediction, we measured COS for 35 cortical cells at different test and mask contrasts. Figure 5 A shows the normalized responses to different combinations of test and mask contrasts. The degree of suppression clearly increases with mask contrast and at very low values (<10%), the level of suppression is minimal. This is qualitatively consistent with our model prediction. To quantify this result, we compare the measured and predicted COS for mask and test stimuli that have equal contrast levels of 12.5, 25, and 50% (Fig. 5B). The gray bars depict the measured levels of suppression, whereas the black and white bars denote the mean CSI of LGN cells and predicted COS values, respectively. Compared with measured COS levels, our model predicts 98.5, 90.3, and 78.9% suppression for 12.5, 25, and 50% contrasts, respectively. The data of Fig. 5 show that contrast saturation at the LGN level plays a reduced role at low contrast levels. For 12.5, 25, and 50% contrast levels, contrast saturation in the LGN can account for 24, 47, and 84%, respectively, of the predicted COS. This result suggests that contrast saturation in the LGN may be the principal source of COS for high contrast levels, but for low contrasts other nonlinear mechanisms are important.
|
COS can be obtained with high temporal frequency mask stimuli that do not typically activate cortical cells (Allison et al. 2001; Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). To explore this in this study and determine if this finding is consistent with our model, we measured contrast response functions at different temporal frequencies for a population of 15 LGN cells. Figure 6 A shows results for an example LGN cell. As the data show, there is contrast saturation for the tests at 7 and 10 Hz. Saturation also occurs at temporal frequencies of 2, 4, 15, and 25 Hz, but this is not obvious in Fig. 6A because we use a standard logarithmic scale for contrast. We use the data at 50 and 100% contrasts to calculate CSI as a function of temporal frequency. This function is shown in Fig. 6B for the example LGN cell. We quantify the tuning properties of this predicted suppression by fitting a Gaussian function to the temporal frequency tuning curve of CSI. We use this curve to calculate peak (TFpeak) and cut-off (TFcut-off) temporal frequencies. TFcut-off is defined as the temporal frequency at which the CSI value is reduced by one half of that at TFpeak. The example cell exhibits a TFpeak of 8.6 Hz and a TFcut-off value of 14.5 Hz. The distributions of TFpeak and TFcut-off values for 15 LGN cells are presented in Fig. 6, C and D. Although these temporal frequency tuning properties are broadly distributed, they are consistent with those of COS in visual cortex (Allison et al. 2001; Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). Considered together, these results are in accord with the hypothesis that contrast saturation in the LGN at high contrast levels may be the principal mechanism mediating COS in visual cortex.
|
|
|
DISCUSSION |
|---|
|
; Chance et al. 1998; Varela et al. 1997, 1999). Our current data show that the response to an optimal test stimulus is unaffected by a briefly presented orthogonal mask as long as the two stimuli do not overlap in time. This result therefore is inconsistent with a model in which synaptic depression is the principal source of COS. We find that contrast saturation in LGN cells, which has not been considered in previous studies (Carandini et al. 2002; Freeman et al. 2002), leads to strong COS in visual cortex for high contrast stimuli. This effect, in combination with the known rectifying nonlinearities in LGN and spike output nonlinearities of cortical neurons (Carandini 2004; Carandini and Ferster 2000; Priebe et al. 2004), can account for observed levels of COS in visual cortex. We studied the saturation mechanism in LGN. However, this effect is likely to have a retinal origin. Retinal contrast gain control mechanisms are well established and include the types of saturation effects we observe in LGN (Shapley and Victor 1978, 1979).
Recent work has shown that the temporal frequency and adaptation properties of cortical cells are not well matched to the characteristics of COS. COS can be induced by mask stimuli with higher temporal frequencies than those that can activate most cortical cells, and it is not affected by prolonged adaptation to mask stimuli (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). These temporal frequency and adaptation properties are consistent with those of LGN neurons but not cortical cells. Based on these results and the finding that COS is weak in the LGN (Bauman and Bonds 1991; Freeman et al. 2002), synaptic depression at thalamo-cortical synapses has been proposed as the mechanism underlying COS in the visual cortex (Carandini et al. 2002; Freeman et al. 2002; Mrsic-Flogel and Hubener 2002). Synaptic depression is a form of short-term synaptic plasticity in which the efficacy of a given synapse is temporarily reduced according to its recent activity level (Abbott and Nelson 2000; Chance et al. 1998). In vitro studies have revealed at least two separate components of short-term synaptic depression acting over different time scales. A fast form occurs within 5–10 presynaptic action potentials and requires 200–600 ms for recovery, whereas a second slower form requires many action potentials to reach full strength and approximately 10s to fully recover (Chance et al. 1998; Varela et al. 1997, 1999). If synaptic depression underlies COS in the visual cortex, its temporal properties should be evident in the responses of cortical neurons. For example, if the slow form of synaptic depression were mediating COS, we would expect to observe stronger suppression after prolonged adaptation to a mask stimulus. This is because synapses would become depressed during the adaptation period and remain depressed throughout the test period. In contrast, the response without prior adaptation would be high initially and exhibit suppression only at later time-points, after synaptic depression had set in. Previous studies of COS do not support this prediction. COS is just as strong with prior adaptation as it is without (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). This suggests that the slow form of synaptic depression with its prolonged onset time is not a viable mechanism for explaining COS.
Another prediction of the synaptic depression hypothesis is that test and mask stimuli need not to be presented simultaneously to observe COS. Because both fast and slow forms of synaptic depression exhibit significant recovery time constants, the suppressive effects of a mask stimulus should remain strong for at least a few hundred milliseconds after its removal. In this study, we evaluated this prediction for neurons in the striate cortex. In contrast to the prediction, we found that both test and mask stimuli must be presented simultaneously to observe COS. This result suggests that neither form of synaptic depression is likely to be a dominant mechanism underlying COS in visual cortex.
In previous studies, COS has been attributed to intracortical inhibition (Allison et al. 2001; Bauman and Bonds 1991; Bonds 1989; DeAngelis et al. 1992; Sengpiel et al. 1998). The primary evidence in support of this view is the finding that COS is blocked by application of the GABA agonist bicuculline over a large region of visual cortex (Morrone et al. 1987). However, this view has been questioned recently by data on the temporal frequency and adaptation properties of COS (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). COS in the visual cortex can be obtained with a mask drifting too fast to elicit responses from most cortical cells (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). In addition, COS is unaffected by adaptation to a mask grating, even though the responses of most cortical neurons are substantially suppressed by adaptation. These findings suggest that the primary mechanism underlying COS operates before the formation of cortical RFs (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005).
There are at least three plausible ways by which this could occur. The first possibility is that COS is present in LGN neurons and simply propagates to visual cortex. Although some cells in the LGN exhibit weak COS, it does not seem strong enough to explain the strength of suppression normally observed in visual cortex (Freeman et al. 2002). Furthermore, because LGN neurons are not selective for stimulus orientation, their responses to overlapping orthogonal stimuli are typically greater than those to either stimulus in isolation (Freeman et al. 2002). A second possibility is that COS may be mediated at thalamo-cortical synapses through synaptic depression (Carandini et al. 2002; Freeman et al. 2002). This is somewhat complicated to verify because synaptic depression seems to be stronger in vitro than in vivo (Boudreau and Ferster 2005; Chung et al. 2002; Sanchez-Vives et al. 2000). However, based on the data and analysis reported here, this possibility also seems unlikely. A third possibility is that COS in visual cortex may originate from other subcortical nonlinear properties, such as rectifying nonlinearities and contrast saturation in LGN neurons as we propose here and as considered briefly in a previous study (Freeman et al. 2002).
The feedforward model of COS postulated here is capable of explaining several of the properties of COS reported in previous studies (Bonds 1989; DeAngelis et al. 1992; Freeman et al. 2002; Li et al. 2005; Morrone et al. 1982). First, it accounts for the broad tuning of COS to mask orientation because LGN neurons do not exhibit orientation tuning. Second, it explains the fact that excitatory and suppressive RFs are largely co-localized (DeAngelis et al. 1992). It is clear that contrast saturation occurs only when the mask stimulus is presented simultaneously on an overlapped region of the excitatory test stimulus. Contrast saturation would not occur if the mask and test stimuli were presented to different regions of the RF. Third, it accounts for the immunity of COS to prolonged mask adaptation (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). LGN cells in the cat exhibit a small adaptation effect compared with cortical cells (Ohzawa et al. 1982, 1985; Sanchez-Vives et al. 2000; Shou et al. 1996). A prolonged mask grating will slightly shift the contrast response function of LGN cells to the right. This will reduce COS in visual cortex with a decrease of contrast saturation in the LGN (see Fig. 4B). On the other hand, the prolonged mask grating can decrease the firing rate of cortical cells (Carandini et al. 1998). These adaptation effects, which may reduce or increase COS, may offset each other. Finally, our model accounts for the temporal frequency properties of COS. It has been shown that COS can be obtained with mask stimuli drifting too rapidly to elicit substantial responses from cortical neurons (Allison et al. 2001; Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). Most LGN neurons exhibit strong contrast saturation to gratings at high temporal frequencies (Fig. 6). This result is in agreement with the temporal frequency properties of COS in visual cortex (Allison et al. 2001; Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005).
Results from this and previous studies of temporal frequency and adaptation properties of LGN and cortical cells point to subcortical mechanisms as the primary basis of COS (Freeman et al. 2002; Li et al. 2005; Sengpiel and Vorobyov 2005). With spike output nonlinearities attributable to cortical cells, these findings suggest that rectifying nonlinearities and contrast saturation in the LGN can account for most COS in visual cortex. Of course, we cannot rule out the possible participation of other mechanisms such as synaptic depression and intracortical inhibition.
A recent paper casts doubt on a purely excitatory feedforward mechanism for COS (Smith et al. 2006). Measurements were made of suppression, recovery, onset, and offset time-courses for cortical neurons. COS and its release occur quickly. However, it is delayed after the offset time response. This suggests a role for intracortical inhibition. However, to show this, inhibitory interneurons must be identified with the same temporal and adaptation properties as those for COS in visual cortex.
Another relevant issue concerns response phase advances with stimulus contrast increases. These have been reported in retina (Shapley and Victor 1978), LGN (Saul and Humphrey 1990), and visual cortex (Dean and Tolhurst 1986). When a mask is simultaneously presented with a test stimulus, the combination produces increased contrast. As a result, we should observe a phase advance of the response to the plaid in our experiments. However, phase advances are not evident in our data (Fig. 2, D and E). Two possible reasons may account for this. First, the resolution (Fig. 2, D and E) is not high enough to show phase advances in the response to the plaid. More spikes would be required from a number of repetitions to obtain higher resolution. Second, although synaptic depression is relevantly weak in vivo (Boudreau and Ferster 2005; Sanchez-Vives et al. 1998), phase delay from synaptic depression at thalamo-cortical synapses may cancel out the phase advance from the increased contrast in the plaid.
Species differences may also be relevant. In the primate, magnocellular but not parvocellular cells are reported to show contrast saturation (Benardete et al. 1992; Shapley et al. 1981). In addition, parvocellular cells seem to constitute the major input to visual cortex (Malpeli et al. 1996). COS has been shown in primate visual cortex (Carandini et al. 1997). Therefore it may be that contrast saturation from LGN cells is less of a factor in COS in the primate visual pathway.
A recently published study, in which intracellular techniques were used, reaches similar conclusions to ours regarding the origins of COS in the visual cortex (Priebe and Ferster 2006). In this study, excitatory and inhibitory conductances are both reduced about equally by the overlapped mask. This suggests that cortico-cortical inhibition is not involved in COS. To explain COS, a toy model is also used in which there is feedforward excitation from thalamic relay cells to cortical simple cells. Within this model, contrast saturation and rectification of LGN cells are used to account for COS in simple cells for low and medium contrast levels (8 and 32%). However, this model produces elevated DC responses to the plaid stimulus (Table 1 of Priebe and Ferster 2006). Most likely, this is caused by the use of a "push" only model. In this study, we used a "push-pull" architecture in which feedforward excitation is balanced with antiphase inhibition separated in spatial phase by 180°. The inhibition is thought to be mediated by cortical interneurons (Carandini et al. 2002; Ferster 1988; Glezer et al. 1982; Palmer and Davis 1981; Tolhurst and Dean 1987; Troyer et al. 1998). Our model incorporates LGN rectification, contrast saturation, and cortical spike threshold in a "push-pull" format where antiphase inhibition balances feedforward excitation. For low contrasts, this arrangement is appropriate. However, it is relatively less effective at high contrasts where other factors including cortico-cortical effects may operate. Overall, however, a "push-pull" approach seems preferable especially because it seems to be required to account for fundamental response characteristics of cortical neurons such as orientation tuning invariance (Sclar and Freeman 1982; Troyer et al. 1998).
In summary, we showed that synaptic depression at thalamo-cortical synapses is unlikely to be the primary mechanism in COS because it requires considerable recovery time. The combination of spike rectification and contrast saturation in the LGN and a spike output nonlinearity of cortical neurons into a standard "push-pull" feed-forward model can explain COS in visual cortex.
|
|
GRANTS |
|---|
|
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: R. D. Freeman, 360 Minor Hall, Univ. of California, Berkeley, CA 94720-2020 (E-mail: freeman{at}neurovision.berkeley.edu)
|
|
REFERENCES |
|---|
|
Abbott LF, Varela JA, Sen K, and Nelson SB. Synaptic depression and cortical gain control. Science 275: 220–224, 1997.[CrossRef][Web of Science][Medline]
Albrecht DG and Geisler WS. Motion selectivity and the contrast-response function of simple cells in the visual cortex. Vis Neurosci 7: 531–546, 1991.[Web of Science][Medline]
Albrecht DG and Hamilton DB. Striate cortex of monkey and cat: contrast response function. J Neurophysiol 48: 217–237, 1982.
Allison JD, Smith KR, and Bonds AB. Temporal-frequency tuning of cross-orientation suppression in the cat striate cortex. Vis Neurosci 18: 941–948, 2001.[Web of Science][Medline]
Anderson JS, Lampl I, Gillespie DC, and Ferster D. The contribution of noise to contrast invariance of orientation tuning in cat visual cortex. Science 290: 1968–1972, 2000.
Bair W, Cavanaugh JR, Smith MA, and Movshon JA. The timing of response onset and offset in macaque visual neurons. J Neurosci 22: 3189–3205, 2002.
Bauman LA and Bonds AB. Inhibitory refinement of spatial frequency selectivity in single cells of the cat striate cortex. Vision Res 31: 933–944, 1991.[CrossRef][Web of Science][Medline]
Benardete EA, Kaplan E, and Knight BW. Contrast gain control in the primate retina: P cells are not X-like, some M cells are. Vis Neurosci 8: 483–486, 1992.[Web of Science][Medline]
Ben-Yishai R, Bar-Or RL, and Sompolinsky H. Theory of orientation tuning in visual cortex. Proc Natl Acad Sci USA 92: 3844–3848, 1995.
Blakemore C and Tobin EA. Lateral inhibition between orientation detectors in the cat's visual cortex. Exp Brain Res 15: 439–440, 1972.[Web of Science][Medline]
Bonds AB. Role of inhibition in the specification of orientation selectivity of cells in the cat striate cortex. Vis Neurosci 2: 41–55, 1989.[Web of Science][Medline]
Boudreau CE and Ferster D. Short-term depression in thalamocortical synapses of cat primary visual cortex. J Neurosci 25: 7179–7190, 2005.
Carandini M. Amplification of trial-to-trial response variability by neurons in visual cortex. PLoS Biol 2: 1483–1493, 2004.[Web of Science]
Carandini M and Ferster D. Membrane potential and firing rate in cat primary visual cortex. J Neurosci 20: 470–484, 2000.
Carandini M and Heeger DJ. Summation and division by neurons in primate visual cortex. Science 264: 1333–1336, 1994.
Carandini M, Heeger DJ, and Movshon JA. Linearity and normalization in simple cells of the macaque primary visual cortex. J Neurosci 17: 8621–8644, 1997.
Carandini M, Heeger DJ, and Senn W. A synaptic explanation of suppression in visual cortex. J Neurosci 22: 10053–10065, 2002.
Carandini M, Movshon JA, and Ferster D. Pattern adaptation and cross-orientation interactions in the primary visual cortex. Neuropharmacology 37: 501–511, 1998.[CrossRef][Web of Science][Medline]
Carandini M and Ringach DL. Predictions of a recurrent model of orientation selectivity. Vision Res 37: 3061–3071, 1997.[CrossRef][Web of Science][Medline]
Chance FS, Nelson SB, and Abbott LF. Synaptic depression and the temporal response characteristics of V1 cells. J Neurosci 18: 4785–4799, 1998.
Chapman B and Stryker MP. Origin of orientation tuning in the visual cortex. Curr Opin Neurobiol 2: 498–501, 1992.[CrossRef][Medline]
Chung S, Li X, and Nelson SB. Short-term depression at thalamocortical synapses contributes to rapid adaptation of cortical sensory responses in vivo. Neuron 34: 437–446, 2002.[CrossRef][Web of Science][Medline]
Creutzfeldt O and Ito M. Functional synaptic organization of primary visual cortex neurones in the cat. Exp Brain Res 6: 324–352, 1968.[Web of Science][Medline]
Dean AF and Tolhurst DJ. Factors influencing the temporal phase of response to bar and grating stimuli for simple cells in the cat striate cortex. Exp Brain Res 62: 143–151, 1986.[Web of Science][Medline]
DeAngelis GC, Robson JG, Ohzawa I, and Freeman RD. Organization of suppression in receptive fields of neurons in cat visual cortex. J Neurophysiol 68: 144–163, 1992.
Ferster D. Spatially opponent excitation and inhibition in simple cells of the cat visual cortex. J Neurosci 8: 1172–1180, 1988.[Abstract]
Freeman TC, Durand S, Kiper DC, and Carandini M. Suppression without inhibition in visual cortex. Neuron 35: 759–771, 2002.[CrossRef][Web of Science][Medline]
Garey LJ and Powell TP. An experimental study of the termination of the lateral geniculo-cortical pathway in the cat and monkey. Proc R Soc Lond B Biol Sci 179: 41–63, 1971.[Medline]
Glezer VD, Tsherbach TA, Gauselman VE, and Bondarko VM. Spatio-temporal organization of receptive fields of the cat striate cortex. The receptive fields as the grating filters. Biol Cybern 43: 35–49, 1982.[CrossRef][Web of Science][Medline]
Hartveit E and Heggelund P. The effect of contrast on the visual response of lagged and nonlagged cells in the cat lateral geniculate nucleus. Vis Neurosci 9: 515–525, 1992.[Web of Science][Medline]
Heeger DJ. Normalization of cell responses in cat striate cortex. Vis Neurosci 9: 181–197, 1992.[Web of Science][Medline]
Horsley V and Clarke RH. The struncture and functions of the cerebellum examined by a new method. Brain 31: 45–124, 1908.
Hubel DH and Wiesel TN. Receptive fields, binocular interaction and functional architecture in the cat's visual cortex. J Physiol 160: 106–154, 1962.
Lauritzen TZ, Krukowski AE, and Miller KD. Local correlation-based circuitry can account for responses to multi-grating stimuli in a model of cat V1. J Neurophysiol 86: 1803–1815, 2001.
Li B, Peterson MR, Thompson JK, Duong T, and Freeman RD. Cross-orientation suppression: monoptic and dichoptic mechanisms are different. J Neurophysiol 94: 1645–1650, 2005.
Malpeli JG, Lee D, and Baker FH. Laminar and retinotopic organization of the macaque lateral geniculate nucleus: magnocellular and parvocellular magnification functions. J Comp Neurol 375: 363–377, 1996.[CrossRef][Web of Science][Medline]
McLaughlin D, Shapley R, and Shelley M. Large-scale modeling of the primary visual cortex: influence of cortical architecture upon neuronal response. J Physiol Paris 97: 237–252, 2003.[CrossRef][Web of Science][Medline]
McLaughlin D, Shapley R, Shelley M, and Wielaard DJ. A neuronal network model of macaque primary visual cortex (V1): orientation selectivity and dynamics in the input layer 4Calpha. Proc Natl Acad Sci USA 97: 8087–8092, 2000.
Morrone MC, Burr DC, and Maffei L. Functional implications of cross-orientation inhibition of cortical visual cells. Proc R Soc Lond B Biol Sci 216: 335–354, 1982.[Medline]
Morrone MC, Burr DC, and Speed HD. Cross-orientation inhibition in cat is GABA mediated. Exp Brain Res 67: 635–644, 1987.[Web of Science][Medline]
Mrsic-Flogel T and Hubener M. Visual cortex: suppression by depression? Curr Biol 12: R547–R549, 2002.[CrossRef][Web of Science][Medline]
Ohzawa I, Sclar G, and Freeman RD. Contrast gain control in the cat visual cortex. Nature 298: 266–268, 1982.[CrossRef][Medline]
Ohzawa I, Sclar G, and Freeman RD. Contrast gain control in the cat's visual system. J Neurophysiol 54: 651–667, 1985.
Palmer LA and Davis TL. Receptive-field structure in cat striate cortex. J Neurophysiol 46: 260–276, 1981.
Priebe NJ and Ferster D. Mechanisms underlying cross-orientation suppression in cat visual cortex. Nat Neurosci 9: 552–561, 2006.[CrossRef][Web of Science][Medline]
Priebe NJ, Mechler F, Carandini M, and Ferster D. The contribution of spike threshold to the dichotomy of cortical simple and complex cells. Nat Neurosci 7: 1113–1122, 2004.[CrossRef][Web of Science][Medline]
Reid RC and Alonso JM. Specificity of monosynaptic connections from thalamus to visual cortex. Nature 378: 281–284, 1995.[CrossRef][Medline]
Sanchez-Vives MV, McCormick DA, and Nowak LG. Is synaptic depression prevalent in vivo and does it contribute to contrast adaptation? Soc Neurosci Abstr 24: 896, 1998.
Sanchez-Vives MV, Nowak LG, and McCormick DA. Membrane mechanisms underlying contrast adaptation in cat area 17 in vivo. J Neurosci 20: 4267–4285, 2000.
Saul AB and Humphrey AL. Spatial and temporal response properties of lagged and nonlagged cells in cat lateral geniculate nucleus. J Neurophysiol 64: 206–224, 1990.
Sclar G and Freeman RD. Orientation selectivity in the cat's striate cortex is invariant with stimulus contrast. Exp Brain Res 46: 457–461, 1982.[Web of Science][Medline]
Sclar G, Maunsell JH, and Lennie P. Coding of image contrast in central visual pathways of the macaque monkey. Vision Res 30: 1–10, 1990.[CrossRef][Web of Science][Medline]
Sengpiel F, Baddeley RJ, Freeman TC, Harrad R, and Blakemore C. Different mechanisms underlie three inhibitory phenomena in cat area 17. Vision Res 38: 2067–2080, 1998.[CrossRef][Web of Science][Medline]
Sengpiel F and Vorobyov V. Intracortical origins of interocular suppression in the visual cortex. J Neurosci 25: 6394–6400, 2005.
Shapley R, Kaplan E, and Soodak R. Spatial summation and contrast sensitivity of X and Y cells in the lateral geniculate nucleus of the macaque. Nature 292: 543–545, 1981.[CrossRef][Medline]
Shapley RM and Victor JD. The effect of contrast on the transfer properties of cat retinal ganglion cells. J Physiol 285: 275–298, 1978.
Shapley RM and Victor JD. Nonlinear spatial summation and the contrast gain control of cat retinal ganglion cells. J Physiol 290: 141–161, 1979.
Shelley M and McLaughlin D. Coarse-grained reduction and analysis of a network model of cortical response. I. Drifting grating stimuli. J Comput Neurosci 12: 97–122, 2002.[CrossRef][Web of Science][Medline]
Shelley M, McLaughlin D, Shapley R, and Wielaard J. States of high conductance in a large-scale model of the visual cortex. J Comput Neurosci 13: 93–109, 2002.[CrossRef][Web of Science][Medline]
Shou T, Li X, Zhou Y, and Hu B. Adaptation of visually evoked responses of relay cells in the dorsal lateral geniculate nucleus of the cat following prolonged exposure to drifting gratings. Vis Neurosci 13: 605–613, 1996.[Web of Science][Medline]
Smith MA, Bair W, Cavanaugh JR, and Movshon JA. Latency of inhibition from inside and outside the classical receptive field in macaque V1 neurons. J Vis 1: 35a, 2001.
Smith MA, Bair W, and Movshon JA. Dynamics of suppression in macaque primary visual cortex. J Neurosci 26: 4826–4834, 2006.
Somers DC, Nelson SB, and Sur M. An emergent model of orientation selectivity in cat visual cortical simple cells. J Neurosci 15: 5448–5465, 1995.[Abstract]
Tanaka K. Organization of geniculate inputs to visual cortical cells in the cat. Vision Res 25: 357–364, 1985.[CrossRef][Web of Science][Medline]
Tolhurst DJ and Dean AF. Spatial summation by simple cells in the striate cortex of the cat. Exp Brain Res 66: 607–620, 1987.[CrossRef][Web of Science][Medline]
Tolhurst DJ and Dean AF. The effects of contrast on the linearity of spatial summation of simple cells in the cat's striate cortex. Exp Brain Res 79: 582–588, 1990.[Web of Science][Medline]
Troyer TW, Krukowski AE, Priebe NJ, and Miller KD. Contrast-invariant orientation tuning in cat visual cortex: thalamocortical input tuning and correlation-based intracortical connectivity. J Neurosci 18: 5908–5927, 1998.
Varela JA, Sen K, Gibson J, Fost J, Abbott LF, and Nelson SB. A quantitative description of short-term plasticity at excitatory synapses in layer 2/3 of rat primary visual cortex. J Neurosci 17: 7926–7940, 1997.
Varela JA, Song S, Turrigiano GG, and Nelson SB. Differential depression at excitatory and inhibitory synapses in visual cortex. J Neurosci 19: 4293–4304, 1999.
Vidyasagar TR, Pei X, and Volgushev M. Multiple mechanisms underlying the orientation selectivity of visual cortical neurones. Trends Neurosci 19: 272–277, 1996.[CrossRef][Web of Science][Medline]
Walker GA, Ohzawa I, and Freeman RD. Binocular cross-orientation suppression in the cat's striate cortex. J Neurophysiol 79: 227–239, 1998.
Wielaard DJ, Shelley M, McLaughlin D, and Shapley R. How simple cells are made in a nonlinear network model of the visual cortex. J Neurosci 21: 5203–5211, 2001.
This article has been cited by other articles:
|
|
 |
|
  R. Kimura and I. Ohzawa Time Course of Cross-Orientation Suppression in the Early Visual Cortex J Neurophysiol, March 1, 2009; 101(3): 1463 - 1479. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Ayaz and F. S. Chance Gain Modulation of Neuronal Responses by Subtractive and Divisive Mechanisms of Inhibition J Neurophysiol, February 1, 2009; 101(2): 958 - 968. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. J. Malone and D. L. Ringach Dynamics of Tuning in the Fourier Domain J Neurophysiol, July 1, 2008; 100(1): 239 - 248. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Duong and R. D. Freeman Contrast Sensitivity Is Enhanced by Expansive Nonlinear Processing in the Lateral Geniculate Nucleus J Neurophysiol, January 1, 2008; 99(1): 367 - 372. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. H. Cohen and Q. Zaidi Fundamental Failures of Shape Constancy Resulting from Cortical Anisotropy J. Neurosci., November 14, 2007; 27(46): 12540 - 12545. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Li and R. D. Freeman High-Resolution Neurometabolic Coupling in the Lateral Geniculate Nucleus J. Neurosci., September 19, 2007; 27(38): 10223 - 10229. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
