Exp Brain Res (1992) 92:246-258
E.m
Brain Research 9 Springer-Verlag 1992
Effect of corticotectal tract lesions on relative motion selectivity in the monkey superior colliculus R.M. Davidson 1, T.J. Joly 2, and D.B. Bender 2 1Department of Oral Biology and 2Department of Physiology, State University of New York at Buffalo,4234 Ridge Lea Rd, Buffalo, NY 14226, USA Received February 12, 1992/Accepted June 11, 1992
Summary. Many cells in the superficial layers of the monkey superior colliculus are sensitive to the relative motion between a small target moving through the classic receptive field and a textured, moving background pattern that fills the visual field beyond the classic receptive field. The cells respond well when motion of the target differs from that of the background, but their responses are suppressed when the target moves in phase with the background. To determine whether this relative motion sensitivity depends on input to the colliculus from visual cortex, we studied colliculus cells in immobilized, anesthetized monkeys after unilateral thermocoagulation, or anesthetic blockade, of the corticotectal tract at the level of the pulvinar. In the colliculus ipsilateral to the corticotectal tract lesions, relative motion sensitivity was significantly reduced when compared either with the colliculus in intact animals or with the colliculus contralateral to the lesion. However, a moving-background stimulus still had a modest suppressive effect compared with a stationary background ("background motion sensitivity"), as is the case for intact animals. Anesthetic blockade of the corticotectal tract had similar effects; relative motion sensitivity, but not background motion sensitivity, was lost following injection of mepivacaine or bupivacaine. Pulvinar cell loss alone, induced by kainic acid injection, had no effect on relative motion sensitivity in the colliculus. The corticotectal tract lesions, but not the anesthetic injections, also had minor effects on flash-evoked responses and spontaneous discharge rates; these effects may reflect a retrograde response of some tectopulvinar cells to injury of their axons by the corticotectal tract lesions. In the colliculus opposite the corticotectal tract lesion, relative motion sensitivity was similar to that in normal animals. However, responses in the presence of a moving background were enhanced, suggesting that removal of cortical input to one colliculus may disinhibit the contralateral colliculus, a phenomenon reminiscent of the Sprague effect in the cat. We conclude that while cortical input to the colliculus may contribute little to the classic Correspondence to:
D.B. Bender
receptive field properties of superficial-layer cells, it clearly does contribute to relative motion sensitivity.
Key words: Vision - Pulvinar - Kainic acid - Center Surround interactions - Macaque
Introduction The classic receptive field properties of cells in the superficial layers of the monkey colliculus have been well established (Schiller and Koerner 1971; Cynader and Berman 1972; Goldberg and Wurtz 1972; Marrocco and Li 1977). When tested with small targets in an otherwise empty visual field, colliculus cells show little stimulus selectivity. With respect to motion selectivity, most are nondirectional and respond well to target movement in any direction over a wide range of speeds. Recently, we found that many colliculus cells are quite selective for relative motion between a target moving through the classic receptive field and a textured, movingbackground stimulus outside the receptive field (Bender and Davidson 1986; Davidson and Bender 1991). The cells discharge for any target movement that differs from background movement, and are suppressed when target direction and speed match that of the background. Furthermore, the magnitude of suppression depends only on the difference between target and background motion and is independent of absolute direction or speed. Cells showing this motion-specific selective suppression are found throughout the stratum griseum superficiale (SGS) and the stratum opticum (SO, together with the stratum zonale they make up the superficial layers) but are most common in the SO and lower half of the SGS. These findings led us to suggest that the lower half of the superficial layers may be specialized to detect relative motion. The superficial layers of the colliculus receive visual information from two major sources. There is direct input from the retina via the retinotectal pathway, and input from visual cortex via the corticotectal pathway. In the monkey, it is generally thought that cortical input contrib-
247 utes little to receptive field properties in the superficial layers; cooling of striate cortex, or ablation of areas 17, 18, and 19, has little effect on visual responses above the middle of the SO, although it does eliminate visual responses below this level (Schiller et al. 1974). Nevertheless, it was natural to think that cortical input might be a m a j o r contributor to relative m o t i o n selectivity. This property has been described in several cortical areas, including the m o t i o n areas M T and M S T in the superior temporal sulcus, V1, and V2 (Allman et al. 1985a, b, 1990; Saito etal. 1986; T a n a k a et al. 1986). Furthermore, cortical input, particularly that from MT, predominates in the lower half of the superficial layers where selective suppression is strongest (Lund 1972; Hubel et al. 1975; G r a h a m et al. 1979; G r a h a m 1982; Ungerleider et al. 1984). A retinal contribution might also be important, but this seemed less likely since relative direction selectivity has not been described in either retina or lateral geniculate body, although nonspecific s u r r o u n d effects have been noted in both structures (McIlwain 1964; K r u g e r 1977; Fischer and K r u g e r 1980; D a v i d s o n and Bender 1991). Furthermore, retinal input is m o s t dense in the upper half of the superficial layers (Hubel et al. 1975) where relative m o t i o n selectivity is less c o m m o n . O u r goal in the present study was to determine whether cortical input is essential for relative m o t i o n selectivity in the superior colliculus. Corticotectal cells are distributed over widespread areas of visual cortex, including areas V1, V2, V3, V4, and M T (Fries 1984; Colby and Olson 1985; L o c k et al. 1990). It would thus be impossible to eliminate cortical input at its source without destroying virtually all visual cortex. However, most corticotectal axons collect together in the corticotectal tract as they pass t h r o u g h the pulvinar nucleus of the thalamus to enter the colliculus t h r o u g h its b r a c h i u m (Bender and Baizer 1984). We thus attempted to sever the corticotectal tract by making radiofrequency lesions at the level of the pulvinar. This procedure of course interrupts a m a j o r ascending p a t h w a y from colliculus to pulvinar to cortex (Benevento and Fallon 1975; Benevento and Rezak 1976; Ogren and Hendrickson 1976; P a r t l o w etal. 1977; Harting etal. 1980), a projection system which might itself contribute, t h o u g h indirectly, to relative m o t i o n selectivity in the colliculus. To control for this possibility, in one animal we examined relative m o t i o n selectivity in the colliculus following a kainic acid (KA) lesion of the pulvinar. K A lesions effectively destroy pulvinar cells, thereby interrupting the ascending pathway, but do not d a m a g e the corticotectal tract (Bender and Baizer 1984). We found that d a m a g e to the corticotectal tract significantly reduces the degree of relative m o t i o n selectivity in the colliculus. D a m a g e to pulvinar cells alone, however, had no effect. Some of these findings have been reported previously (Davidson et al. 1986).
and recording methods are given in Bender (1982) and Davidson and Bender (1991). Briefly, each animal was ~tudied during a series of 8to 12-h semichronic recording sessions over a period of 6-12 weeks. The animal's skull had been implanted previously with hardware which provided painless fixation of the head and access to the brain. Surgery was performed under aseptic conditions. Animals were anesthetized with ketamine (35 mg/kg) or a mixture of ketamine (20 mg/kg) and xylazine (1.3 mg/kg); anesthetic adequacy was judged by the lack of reflex response to nociceptive stimuli, and anesthetic supplements (10-30% of initial dose) were given as required. At the start of recording the animal was premedicated with atropine, tranquilized with ketamine (6 mg/kg), anesthetized with 2.5% halothane in a 70% nitrous oxide oxygen mixture, and intubated with a xylocaine-coated endotracheal cannula. The eyes were immobilized with an intravenous infusion of pancuronium bromide. The pupils were dilated with cyclopentolate hydrochloride and the eyes were focussed with contact lenses on a tangent screen 0.57 m away. During recording the animals were maintained on 70% nitrous oxide-oxygen; the electrocardiogram and electroencephalogram were monitored to assess adequacy of the analgesia. The animals were recovered from immobilization at the end of the session under ketamine anesthesia. The activity of single isolated neurons was recorded using epoxycoated, tungsten microelectrodes with an impedance of 3-5 Mf~ at 500 Hz. The electrode shape was the same as that used in our previous study of relative motion in the intact colliculus. We elected not to adjust the shape, even though signal-to-noise ratios were often smaller and we sometimes had difficulty isolating cells on the side of the radiofrequency lesions, since such an adjustment could have introduced a more serious sampling problem. Three animals received a pulvinar lesion in the left hemisphere 5-13 weeks prior to recording. The lesions were intended to destroy the corticotectal tract and thus were made by radiofrequency heating at four sites within the pulvinar; the sites were selected after first making microelectrode recordings from the pulvinar and the caudal pole of the lateral geniculate nucleus (LGN). One of these animals (ST) also received a lesion of the right pulvinar. The lesion was made by microinjection of 5.2/~g KA, distributed over five sites, and was intended as a control for loss of pulvinar cells that accompanied the radiofrequency lesions. The procedures for making both types of lesions are described in detail in Bender and Baizer (1984). In the fourth animal we inactivated the corticotectal tract by injecting a local anesthetic into the pulvinar. To eliminate the possibly confounding effect of pulvinar cell inactivation, we first destroyed pulvinar cells by KA injection 10 months prior to the anesthetic injections. Anesthetic injections were made through a Hamilton syringe. The injection site was based on microelectrode recordings and was located about 1 mm posterior and 2.5 mm medial to the caudal pole of the LGN. Typically, 5 gl (range 1-20/~1) of 2% mepivacaine hydrochloride, or 0.25% bupivacaine hydrochloride with 0.001% epinephrine, was injected over a l-2min period. The injection site was marked by injecting rhodaminelabelled latex microspheres. The site was subsequently verified during histological processing and found to lie at the dorsal edge of the corticotectal fiber system (at the "L" of"PL" in Fig. 1A, AP level 4.0), relatively distant from the colliculus itself. During a single recording session, we usually studied only one colliculus cell before and after an injection. The experimental protocols for all procedures were approved by the Institutional Animal Care and Use Committee of the University at Buffalo, State University of New York.
Materials and methods
The animal faced a 65 ~ x 65 ~ tangent screen on which two independent, computer-controlled, stimulus patterns could be presented. One pattern, which we call the "target", was a small spot or slit that could be moved through the cell's classically defined receptive field. The other pattern, which we call the "background", was an array of luminous, irregularly shaped spots that filled the tangent screen
Animal preparation and recordin9 Four male Macacafascicularis monkeys weighing between 3.2 and 5.4 kg were used. Detailed descriptions of the animal preparation
Visual stimulation and procedure
248 except for a central masked area. The mask was centered on the classic receptive field and was typically 3-4 times its diameter (median 3.8, interquartile range 2.3-5.2). The mask eliminated complex interactions between target and background contours which otherwise could have occurred within the classic receptive field. The stimulus arrangement is described in detail in Davidson and Bender (1991). In the animals with radiofrequency lesions, we studied all cells that could be reliably isolated. For each cell, we first used hand-held stimuli to test for visual responsiveness, and then plotted the cell's receptive field, made an informal estimate of direction preference, and selected an appropriate size and shape for the target. We then set up the computer-driven optical systems to move both target and background. The target was moved back and forth along a single axis, or along two orthogonal axes, through the receptive field. For each target direction, we compiled two histograms: an "inphase" condition in which the background stimulus moved at the same speed in the same direction as the target, and an "antiphase" condition in which the background moved in the opposite direction at the same speed. During informal testing, we adjusted the location of the target's trajectory within the receptive field, the target speed, and the intertrial interval (ITI) in an effort to maximize the difference between the inphase and antiphase responses. During the actual data run, we usually included additional conditions for which the background was stationary during target movement, and for which the target was presented as a 500- to 800-ms flash against a blank background. The histograms were based on five to ten trials, depending on the variance in the summed response, and the trials for all histograms in a run were randomly interleaved. The ITI was typically 10-15 s, but ranged from 3 to 25 s. Response strength for each histogram was taken as the total spikes evoked by the target when it passed through the receptive field, expressed as an average per trial, minus the spontaneous discharge during a period just preceding stimulus presentation. We also examined peak discharge rate as a measure of response strength. Both response measures yielded similar estimates of relative motion sensitivity (see Davidson and Bender 1991). Unless otherwise noted, we used total spikes as a measure, since this avoided the problem of histograms that had multiple peaks. This procedure differs slightly from that used in our study of normal animals (Davidson and Bender 1991), the results of which formed one basis for comparison with the present experiment. In that study, we had usually paired 4-8 directions of background movement with each target direction in order to assess tuning for relative direction, and we had also used different background speeds to determine selectivity for relative speed. Since in the present experiment it was important to study as many cells as possible after the lesions, we reduced the number of stimulus conditions per cell. In order to make direct comparisons between the operated and normal animals, we used only the inphase and antiphase conditions from the normal study in computing relative direction selectivity. Statistical comparisons were made using the Mann-Whitney U-test for differences in magnitude, and the chi-square or Fisher exact test for differences in proportions. In the animal that received local anesthetic injections, we studied only those colliculus cells that showed strong selective suppression. Each cell was tested with a single target direction paired with three conditions of background motion: inphase, antiphase, and stationary. A fourth condition, with the target flashed against a blank background, was also included. The conditions were randomly interleaved within a block of trials, five trials per condition. We first ran two to four blocks of trials to estimate the effect of background motion prior to injection, and then lowered the injection needle into the pulvinar, injected the anesthetic, and continued to run a block of trials every 6-7 rain until the effects of the anesthetic were no longer detectable or the cell was lost, typically 30 60 min after injection. The injection needle was left in place throughout this period. Microelectrode penetrations were made throughout the retinotopic map in the colliculus, with two exceptions. We tried to avoid eccentricities less than a few degrees since the pulvinar lesions
slightly damaged the representation of this part of the visual field in the LGN. We also tried to avoid eccentricities greater than about 15 ~ The mask size required at these eccentricities substantially reduced the amount of background stimulus on the screen, which in turn could have led us to underestimate the background's influence.
Histological methods and findings The animals were perfused intracardially under deep pentobarbital sodium anesthesia with 0.9% saline, followed by 10% formalin in saline. Frozen 33-#m sections through the thalamus and midbrain were cut in the vertical stereotaxic plane. Alternate sections were stained for Nissl substance with cresyl violet and for myelin with the Heidenhain-Woelcke method. The sections were examined microscopically, and regions of cell and fiber loss were plotted on a set of standard brain drawings. Electrode penetrations made during the last 2 weeks of recording were reconstructed, and recording sites were located within the SGS or SO on the basis of small electrolytic marking lesions made during recording. For the earlier penetrations, we estimated the depth of recording sites from microdrive readings. The radiofrequency lesions were placed with the intention of interrupting fibers of the corticotectal tract as they pass through the pulvinar before collecting together to enter the tectum through the brachium of the superior colliculus. Figure 1A indicates the approximate location of this fiber system within the pulvinar. The lesions were as intended and were well localized to the pulvinar; the common region of damage among the three animals amounted to a knife-cut through the lateral part of the lateral and inferior pulvinar (Fig. 1B). In no case did the lesions encroach on the colliculus itself, and there was no evidence of cell loss within the colliculus. Examination of myelin-stained sections showed a near total loss of staining in the region of the corticotectal tract in two animals (CH and ST); in the third (OB), there was a small island of tissue between levels 4.0 and 4.5 through which normal-appearing fascicles could be seen heading toward the brachium. In all animals, staining in the brachium itself was present but substantially fainter than on the opposite side; gross shrinkage of the brachium, as is seen with longer survival times (Bender and Baizer 1984), was not obvious. There was also minor and variable damage in some structures adjacent to the pulvinar. The most caudal tip of the LGN was slightly damaged, but only in the representation of the central 2-3 ~ of the lower quadrant, near the vertical meridian. The KA lesions were intended to produce comparable cell loss, without damaging fibers, in the same areas of the pul.vinar that had been injured by the radiofrequency lesions. Figure 1C shows a reconstruction of the KA lesion in animal ST. There was total cell loss throughout most of the inferior and lateral pulvinar, and in the lateral half of the medial pulvinar. The posterior tip of the LGN, representing the central 3 4 ~ of the lower quadrant, was slightly damaged, as it had been after the radiofrequency lesions. Cell loss was thus approximately in the same regions and just as extensive as for the radiofrequency lesions.
Results W e s t u d i e d 157 cells in the three a n i m a l s with radiofreq u e n c y p u l v i n a r lesions. R e c o r d i n g sessions b e g a n 5 - 2 0 weeks after the l e s i o n a n d c o n t i n u e d for 6 - 1 2 weeks; r e c o r d i n g s f r o m b o t h right a n d left colliculi were m a d e t h r o u g h o u t this period. All cells were l o c a t e d w i t h i n the S G S or SO. T w o t h i r d s of the cells (103/157) were in the colliculus ipsilateral to the r a d i o f r e q u e n c y lesion a n d t h u s c o u l d h a v e b e e n directly affected b y d a m a g e to the corticotectal tract; cells i n the c o n t r a l a t e r a l colliculus served as a c o n t r o l p o p u l a t i o n . Since receptive field p r o p e r t i e s were
249
A
(
.o !vi
,,Cd
-:::, ( PM -.
[M_~'LGN,
,~
B
.
II ?
T-t ( /Fr
"..
~-~_-~,~,
,-;F;J
W~.~-~
'..
RF COMPOSITE
)
.....
C
KA
I
5.5
5.0
4.5
4.0
3.5
3.0
Fig. 1. A Schematic rendition of the approximate paths taken by retinotectal fibers (solid black) and corticotectal fibers (dashed areas) on their way to the colliculus (adapted from Bender and Baizer 1984). B Reconstruction of the radiofrequency lesions of the left corticotectal tract as shown on standardized coronal sections through the posterior thalamus. Areas in solid black represent tissue destruction that was common to all three animals; stippled area represents the largest extent of the damage in any of the three animals. C Re-
construction of the kainic acid lesion in the right hemisphere of one animal. Solid black indicates areas of total cell loss. Numbers at the bottom indicate approximate stereotaxic levels of the sections. Cd, caudate nucleus; H, habenular complex; LGN, lateral geniculate nucleus; LP, lateral posterior nucleus; MD, medial dorsal nucleus; MG, medial geniculate nucleus; PI, inferior pulvinar nucleus; PL, lateral pulvinar nucleus; PM, medial pulvinar nucleus; R, reticular nucleus; SC, superior colliculus; VPL, ventral posterolateral nucleus
similar a m o n g the three animals, and at different survival times, we pooled data across animals on the side of the radiofrequency lesion. On the side opposite the radiofrequency lesion, one animal had received a KA lesion9 Since colliculus receptive field properties in this hemisphere were similar to those in the two intact hemispheres, we also pooled data for all three hemispheres opposite the radiofrequency lesion. Table 1 summarizes the number of cells and survival times for the three animals. For every cell that clearly responded to a light spot or slit, we attempted to assess both directional selectivity in the absence of a moving background ("classic directionality"), and selectivity for relative direction between the target and a moving background 9 When tested for classic directionality, most cells were nondirectional. In the following sections, we discuss these nondirectional cells first9 We compare selectivity for relative motion in
cells ipsilateral to the corticotectal tract lesion with two separate control populations: cells in the colliculus opposite the corticotectal tract lesion and comparably tested cells from our study in normal animals (Davidson and Bender 1991). We then describe some unanticipated effects of the lesions on both spontaneous and visually evoked discharge patterns, and the effect of anesthetic injections. In the last section, we describe the few cells that showed classic directionality.
Selective suppression in nondirectional cells In normal animals, m a n y colliculus cells show a characteristic selectivity for relative motion. The response to a target moving through the classic receptive field is strongly suppressed when the background moves in the same
250 Table 1. Responsiveness, receptive fields, and test conditions for colliculus cells in animals with radiofrequency lesions of the corticotectal tract Left colliculus
Right colliculus
Survival time (weeks)
Number of cells
No response
5-18 9-18 13-20
33 28 42 103
6 3 5 14
Eccentricity (deg) Sizea (deg)
Median 11.0 3.0
IQR 6.0-13.0 2.0-5.0
Test conditions 2 target directions 3 target directions 4 target directions
Number of cells 36 7 46
Animal ST OB CH Total
Lesion RF RF RF
Number of cells
No response
28 8 18 54
0 0 0 0
Median 7.0 2.9
IQR 5.0-12.0 2.0-3.5
Lesion KA Intact Intact
Receptive fields
Number of cells 26 3 25
"Diameter of the receptive field's central activating area. RF, radiofrequency; KA, kainic acid; deg, degrees IQR, interquartile range direction as the target, but is only weakly suppressed, if at all, when the background moves in the opposite direction. This selective suppression for inphase background motion is about equally strong for all directions of target movement (Davidson and Bender 1991). The corticotectal tract lesions had two main effects. First, for most cells, they greatly reduced the magnitude of selective suppression for all target directions. Second, for the remaining cells, the lesions eliminated selective suppression for all but one target direction. In order to estimate the strength of selective suppression, we took the ratio of the response (total spikes evoked by the target - see Materials and methods) for the inphase condition to that for the antiphase condition. Almost all cells were tested using either two or four different target directions, thus yielding either two or four different ratios. F o r each cell, we used the smallest of these as an index of selective suppression (S). S is thus a conservative index, since it would reflect only the smallest decrease of suppression after the lesion. For comparison with normal animals, we computed the same measure for those cells in our previously published study (Davidson and Bender 1991) that had been similarly tested with either two or four target directions. Figure 2 shows the distribution of S for cells in normal and operated animals. On the side of the corticotectal tract lesion, suppression was substantially reduced (median S 0.63) when compared either with normals (median S 0.39; P < 0.0001) or with cells in the colliculus opposite the lesion (median S 0.36; P < 0.0001). The predominant effect of the lesion may have been to reduce suppression of inphase responses. After normalizing the inphase and antiphase responses to the response for a stationary background, inphase responses on the side of the lesion were substantially larger than in normal animals ( P
1
Selective Suppression Fig. 2. Distributions of selective suppression for cells from three intact animals (Norm) of a previous study (Davidson and Bender 1991), cells contralateral to the radiofrequency lesion (Contra), and cells ipsilateral to the radiofrequency lesion (RF). Selective suppression is the ratio of the inphase response to the antiphase response, both adjusted for spontaneous activity
0.0003), whereas the distribution of antiphase responses was not significantly affected. On the side opposite the lesion, the distribution of S did not differ from that in normal animals. It is apparent from Fig. 2 that some cells still showed a significant degree of selective suppression, as measured by S, after the corticotectal tract lesions. We found, however, that these cells had also been affected by the lesions. Although they showed significant suppression for the one direction of target movement selected by our index, most did not show significant suppression for other target
251
RF side anti-phase
Opposite RF lesion
in-phase
anti-phase
in-phase
lOO[L- i
.46~-=L _.
? .95
.80
. _
A__l_
28 L
24 j L
.30
directions, a marked change from cells in normal animals. Figure 3 shows a typical example of this. A cell on the side of the corticotectal tract lesion ("RF side") shows modest inphase suppression when tested with a target moving to the left, but there is no differential effect of background motion when the target moves in other directions. For a cell opposite the lesion, as shown to the right, the pattern of responses is characteristic of the normal colliculus: background motion has a similar effect for all target directions. To show this effect of the lesions for the population of cells, we first selected all cells with an S less than 0.6. (In normal animals, we used this value to indicate clear selective suppression since it represented a differential effect of background motion that was about twice the standard deviation associated with replicating a PST; see Davidson and Bender 1991.) For each of these cells, we then looked at the ratios of inphase to antiphase response for each of the target directions. On the side of the corticotectal tract lesion, few cells (17%, 6/35) showed clear selective suppression (i.e., had ratios < 0.6) for all target directions, whereas most cells in normal animals (71%, 39/55; P < 0.0005) did. In like fashion, many cells on the side of the lesion (57%, 20/35) showed clear selective suppression for only one target direction, whereas in normal animals only 20% (11/55; P < 0.0001) of the cells did. On the side opposite the lesion, the proportions of cells showing clear suppression in all, or just one, target direction were no different from those in normal animals. Figure 4 illustrates these results.
Antiphase responses. In normal animals, movement of the background in the antiphase direction typically has a weak suppressive effect when compared with the response evoked by the target in the presence of a stationary background. This was also the case for cells recorded on the side of the corticotectal tract lesion. Opposite the
Fig. 3. Inphase and antiphase responses for each of four different target directions. Histograms to the left are from a cell ipsilateral to the radiofrequency (RF) lesion, 890 ibm below the tectal surface. Significant selective suppression is seen only for leftward target movement. Arrows to the left of the histograms indicate target direction; numbers indicate the ratio of inphase to antiphase response. Receptive field eccentricity and mask diameter were 6.8 ~ and 8 ~ respectively; target speed 20~ Histograms to the right are from a cell contralateral to the radiofrequency lesion, 940/tin below the tectal surface. Note that selective suppression is strong for all target directions. Receptive field eccentricity and mask size were 14.4 ~ and 15 ~ respectively; target speed 30~ Calibration bar 8 ~ or 400 ms for the RF side, 8~ or 267 ms for the side opposite RF lesion
selective suppression
selective suppression in oll tgt directions
in only 1 tgt direction Norm(]Is RF Opposite RF I
100
r
i
50
0
50
i
100
go of Cells Fig. 4. Bars to the left of zero indicate percentage of cells showing significant selective suppression (suppression index < 0.6) in only one target direction. Bars to the right of zero indicate cells with significant suppression in all target directions tested
lesion, however, antiphase suppression was less common, and a number of cells even showed enhanced responses in the antiphase condition. Figure 5A shows a typical example of a cell whose response under the antiphase condition was significantly larger than the response with a stationary background. To illustrate this effect for the population of cells, we first normalized the antiphase responses for each target direction to the corresponding responses in the presence of a stationary background, and then used, for each cell, the largest of the normalized responses among those for the different target directions. Figure 5B, C shows the distributions of these normalized antiphase responses. For cells opposite the corticotectal tract lesion, the median response (1.1) was significantly larger than for normal animals (0.68; P < 0.0001). Furthermore, a larger proportion of cells (26 %, 8/31) showed clear enhancement (antiphase response > 1.4) than in normal
252
A
Stationary
B
Anti - phase
In - phase
Normal animals
with either two or four directions of target movement, each direction paired with inphase and antiphase background movement. We smoothed each poststimulus time (PST) histogram with a 5-point moving average, and then picked, for each cell, the largest of the four or eight responses as measured by peak discharge rate. There was no significant difference in these maximal responses between cells on the side of the corticotectal tract lesion and cells in normal animals. On the side of the lesion, the median peak response was 55 spikes/s, whereas in normals it was 49 spikes/s, and on the side opposite the lesion it was 67 spikes/s.
30
Other effects of the lesions 20 I0
I
0
C
Opposite RF lesion 30
]
N=31
,0j 0 0.0
0.5 Anti-phase
1.0
>1.4
Response
Fig. 5. A Enhancement of the antiphase response for a cell opposite the radiofrequency lesion, 540 #m below the tectal surface. Numbers to the left of the PSTs indicate response magnitude relative to that for the stationary background condition. Receptive field eccentricity and mask diameter were 3.2 ~ and 8 ~ respectively; target speed 20 ~ calibration bar 4 ~ or 200 ms. B Distribution of the magnitude of the antiphase response, normalized to the response with a stationary background, for colliculus cells in normal animals. C Corresponding distribution for colliculus ceils opposite the radiofrequency lesions
animals (4%, 3/74; P < 0.003). For cells on the same side as the lesion, the median antiphase response was 0.62 and did not differ significantly from normal.
Response strength. Although lesions or blockade of visual cortex do not weaken visually evoked responses in the superficial layers (Schiller et al. 1974), it was conceivable that lesions of the cortieotectal tract might do so. We were thus concerned that our measure of selective suppression might be distorted by such a weakness, thereby making comparison with control data difficult. We found, however, that those cells for which we evaluated selective suppression were capable of responding just as strongly as cells in normal animals. To show this, we compared the strongest response evoked by moving stimuli for cells in the operated animals with the corresponding measure in normal animals. We used all cells that had been tested
As described below, we noticed that some of the cells recorded on the side of the radiofrequency lesion seemed to discharge in an abnormal fashion. Responses to flashed stimuli were somewhat more sustained and longer in latency, and high spontaneous rates were more frequent, than was typical for intact animals. The distributions of these measures tended to be bimodal after the lesions, hinting at the emergence of cell properties not present in the normal colliculus.
Response to flashed stimuli. In normal animals, the response of most colliculus neurons to the onset of an 800ms flash consisted of a transient burst of spikes, often followed by a sustained component. The response typically had a relatively short latency, a brief rise time, a small sustained component, and was well time locked. 1 On the side of the radiofrequency lesion, most cells gave PST histograms in response to flashed stimuli that were similar to those in normal animals. However, the distributions of latency, rise time, and sustained component all indicated an increase in these measures relative to normal animals. Thus, more cells (34%, 20/59) had onset latencies exceeding 80 ms than were found in normal animals (2%, 1/45; P < 0.0001; see Fig. 6A), and more cells (36%, 21/59) had rise times exceeding 72 ms than were found in normals (4%, 2/45; P < 0.0001). For some cells, the latter reflected a considerable trial-to-trial variation in latency that was rarely seen in normal animals (Fig. 6B). Also, more cells (20%, 12/61) gave only a sustained response, with no clear transient component, than in normals (3%, 2 of 65; P < 0.005), and for those cells that did have both transient and sustained components, the distribution of the sustained component was skewed toward larger values (P < 0.01; Fig. 6C). All of these changes were most apparent for cells in the lower half of the superficial layers. For cells opposite the radiofrequency
1Rise time was defined as the interval between response onset and the time to reach 90% of the peak discharge rate that was achieved within 200 ms of stimulus onset. The strength of the sustained component was expressed as a fraction of the peak response, (Rfi,al -Spon)/(Rpeak-Spon ) where Rnnal is the average discharge rate during the last 600 ms of the flash, Rp~ is the peak discharge rate during the first 200 ms of the flash, and Spon is the spontaneous discharge rate
253
40
[] Normal=,N=45 9 RFSide,N=52
~)30 ~20
o~ 10
4
36
20
52
68
>80
Latency (msec)
9 RFSide, N=59
. :~~-~I
I
I~176 E >80
Rise Time (msec)
[] Normals,N=43 9 RFSide, N=47
40 or)
_
..I,
dk,
103 ,
Jk,.
"~ 10
2,I
0 .1
.
o
.
.9
Sustained Component
D
"6
$0 l
i
[] Normals,N=82
40-
~ _
o
1
2
4
8
16
.29
JI >16
Spontaneous Rate (spk/s)
Fig. 6A-D. Comparison of flash response parameters for cells in normal animals (Normals) and cells on the side of the radiofrequency lesion (RF Side). A Distributions of response latency. B Distributions of response rise-time. Histogram to the right shows the response of a cell on the side of the radiofrequency lesion to an 800-ms flash (solid bar beneath PS:I); note variable onset latency in rasters at the top, and gradual rise-time in the histogram below. C Distributions of the sustained component, expressed as a fraction of the transient component, of the response. Histograms to the right are from two different cells on the side of the radiofrequency lesion. Numbers at the right of each PST give the index representing the sustained component; flash duration is 800 ms. D Distributions of spontaneous discharge rate
254
A
200
ii f
h
I
'
,
_
-
~-:_-
200f I
0.2s
I
Fig. 7 A, B. "Bursty" discharge patterns from a cell ipsilateral to a radiofrequency lesion, 870/~m below the tectal surface. A Response to an 800-ms flashed stimulus (solid bar beneath PST). B Response to the same stimulus swept through the receptive field at 30~ lesion, the distributions of latency, rise time, and sustained component were all similar to their counterparts in normal animals. Spontaneous rate and discharge pattern. In normal animals, most cells had spontaneous discharge rates less than 1-2 spikes/s. On the side of the radiofrequency lesion, spontaneous rates were somewhat higher (Fig. 6D). This was most obvious for cells in the lower half of the superficial layers where the median spontaneous rate was 10 spikes/s. We also encountered some cells, ipsilateral to the lesions, that had quite abnormal discharge patterns. The pattern consisted of a series of staccato-like bursts of impulses, separated by periods of little or no activity. The "bursty" pattern was present during both spontaneous and visually driven activity. Figure 7 shows an example of the pattern in the presence of both moving and stationary targets. For some cells, the burstiness was not so flagrant, and was difficult to classify as abnormal. For others, every trial in every histogram, for both moving and stationary targets, showed it. On the side with the radiofrequency lesions, 23% (18/78) of the cells were this bursty, whereas contralateral to the lesion, none were (0/43, P < 0.0002). The incidence of bursty cells did not vary with depth in the colliculus.
lesions. To control for this possibility, we selected colliculus cells that showed strong selective suppression in an animal with an intact corticotectal tract, and then injected a local anesthetic into the pulvinar in an attempt to reversibly inactivate the tract; pulvinar cells had been destroyed previously by a KA lesion. We studied ten cells before and after injection of mepivacaine or bupivacaine. The results paralleled the effects of corticotectal tract lesions. Nine of the ten cells were affected by the anesthetic injections. Selective suppression was eliminated in five and reduced by 30-50% in two. In the remaining two cells, selective suppression was unaffected by the injection, despite a 50-75% reduction in response strength. Figure 8A shows the loss of selective suppression for one cell. Immediately following the injection, there was a decreased response to the stationary background condition, a modest reduction of the antiphase response relative to the stationary background condition, and most notably the differential response to background direction had disappeared. Sixty-five minutes later, all responses had returned close to their preinjection levels. Figure 8B summarizes the effect of anesthetic injection on responses evoked for the stationary, antiphase, and inphase background conditions for the five cells that lost selective suppression. Since the onset and duration of the anesthetic effect varied among cells, we computed response strength from two successive blocks of trials prior to injection, and from the two successive blocks after the injection which showed the greatest loss of selective suppression. A block of trials took 6 7 rain, and the maximal anesthetic effect usually occurred within 15 min after injection. This procedure underestimated the maximum effect of the anesthetic since there was invariably some recovery of suppression within the postinjection blocks of trials that we used. All responses for a given cell were normalized to the preinjection response under the stationary background condition. Immediately following the injections, responses were usually smaller for all background conditions, and the responses evoked for inphase and antiphase background motion were no longer significantly different. For the five cells, S averaged 0.34 before injection and 0.82 after injection. In contrast, antiphase suppression, expressed as the ratio of responses for the antiphase and stationary background conditions, was not eliminated, averaging 0.87 and 0.76 before and after injection, respectively. The loss of selective suppression, without loss of antiphase suppression, was thus similar to the effect of corticotectal tract lesions. However, other effects of the radiofrequency lesions were not seen after anesthetic injections. Spontaneous rates were not elevated, nor was there any hint of bursty discharge patterns. Furthermore, there was no significant change in the latency or in the sustained component of the response to flashed stimuli, as there had been following radiofrequency lesions.
Injections of local anesthetic The apparent loss of selective suppression after lesions of the corticotectal tract could conceivably reflect a sampling bias induced by the lesions, rather than a loss of selective suppression in cells that had possessed it prior to the
Directional cells In normal animals, a small proportion of colliculus cells were directionally selective. When a target was moved
255
A
pre-inj
Stationary
r L
Anti-phase
In-phase
B
1.." 0.1
2 min post-inj
to
I
post-inj
0.01 0.01
i
i
i
i
i
i
i
I
I
i
i
0,1
i
a
i
i
i
i
]
1
Pre-iniecfion Response Fig. 8A,,B. Loss of selective suppression following injection of local anesthetic into the pulvinar. A Responses from a single cell before and after injection of 5/~1of mepivacaine. Each histogram is based on five trials. The cell was 250/~m below the tectal surface. Receptive field eccentricity and mask size were 2.4 ~ and 3.8 ~ respectively; target speed 10~ Calibration bar 4 ~ or 400 ms. B Responses from fivedifferent cells:for each cell, the in-phase and anti-phase responses
differed significantly before, but not after, injection. Dashed lines connect data for each cell. Each response is based on ten trials pooled from two successive blocks of trials; horizontal and vertical bars represent 2 SE about the mean preinjection and postinjection responses, respectively.Responses for each cell are normalized to the cell's preinjection response for a stationary background
back and forth along a single axis through the receptive field, in the presence of a stationary background, the response for one direction was more than twice that for the opposite direction. We had thought that this classic directionality, like relative motion selectivity, might depend on cortical input and thus might be less frequent following corticotectal tract lesions. This was not the case. On the side of the lesion, "the proportion of directional cells was about the same (23%, 11/57) as in normal animals (15%, 18/123). Opposite the lesion, the incidence of directional cells (24%, 10/42) was likewise no different from that in normal animals. Although the corticotectal tract lesions did not reduce the number of directional cells, they did substantially reduce selectivity for relative motion, just as they had for nondirectional cells. On the side of the lesion, the median S was 0.63 for directional cells, whereas in normal animals it was 0.32 for directional cells (P
E.m
Brain Research 9 Springer-Verlag 1992
Effect of corticotectal tract lesions on relative motion selectivity in the monkey superior colliculus R.M. Davidson 1, T.J. Joly 2, and D.B. Bender 2 1Department of Oral Biology and 2Department of Physiology, State University of New York at Buffalo,4234 Ridge Lea Rd, Buffalo, NY 14226, USA Received February 12, 1992/Accepted June 11, 1992
Summary. Many cells in the superficial layers of the monkey superior colliculus are sensitive to the relative motion between a small target moving through the classic receptive field and a textured, moving background pattern that fills the visual field beyond the classic receptive field. The cells respond well when motion of the target differs from that of the background, but their responses are suppressed when the target moves in phase with the background. To determine whether this relative motion sensitivity depends on input to the colliculus from visual cortex, we studied colliculus cells in immobilized, anesthetized monkeys after unilateral thermocoagulation, or anesthetic blockade, of the corticotectal tract at the level of the pulvinar. In the colliculus ipsilateral to the corticotectal tract lesions, relative motion sensitivity was significantly reduced when compared either with the colliculus in intact animals or with the colliculus contralateral to the lesion. However, a moving-background stimulus still had a modest suppressive effect compared with a stationary background ("background motion sensitivity"), as is the case for intact animals. Anesthetic blockade of the corticotectal tract had similar effects; relative motion sensitivity, but not background motion sensitivity, was lost following injection of mepivacaine or bupivacaine. Pulvinar cell loss alone, induced by kainic acid injection, had no effect on relative motion sensitivity in the colliculus. The corticotectal tract lesions, but not the anesthetic injections, also had minor effects on flash-evoked responses and spontaneous discharge rates; these effects may reflect a retrograde response of some tectopulvinar cells to injury of their axons by the corticotectal tract lesions. In the colliculus opposite the corticotectal tract lesion, relative motion sensitivity was similar to that in normal animals. However, responses in the presence of a moving background were enhanced, suggesting that removal of cortical input to one colliculus may disinhibit the contralateral colliculus, a phenomenon reminiscent of the Sprague effect in the cat. We conclude that while cortical input to the colliculus may contribute little to the classic Correspondence to:
D.B. Bender
receptive field properties of superficial-layer cells, it clearly does contribute to relative motion sensitivity.
Key words: Vision - Pulvinar - Kainic acid - Center Surround interactions - Macaque
Introduction The classic receptive field properties of cells in the superficial layers of the monkey colliculus have been well established (Schiller and Koerner 1971; Cynader and Berman 1972; Goldberg and Wurtz 1972; Marrocco and Li 1977). When tested with small targets in an otherwise empty visual field, colliculus cells show little stimulus selectivity. With respect to motion selectivity, most are nondirectional and respond well to target movement in any direction over a wide range of speeds. Recently, we found that many colliculus cells are quite selective for relative motion between a target moving through the classic receptive field and a textured, movingbackground stimulus outside the receptive field (Bender and Davidson 1986; Davidson and Bender 1991). The cells discharge for any target movement that differs from background movement, and are suppressed when target direction and speed match that of the background. Furthermore, the magnitude of suppression depends only on the difference between target and background motion and is independent of absolute direction or speed. Cells showing this motion-specific selective suppression are found throughout the stratum griseum superficiale (SGS) and the stratum opticum (SO, together with the stratum zonale they make up the superficial layers) but are most common in the SO and lower half of the SGS. These findings led us to suggest that the lower half of the superficial layers may be specialized to detect relative motion. The superficial layers of the colliculus receive visual information from two major sources. There is direct input from the retina via the retinotectal pathway, and input from visual cortex via the corticotectal pathway. In the monkey, it is generally thought that cortical input contrib-
247 utes little to receptive field properties in the superficial layers; cooling of striate cortex, or ablation of areas 17, 18, and 19, has little effect on visual responses above the middle of the SO, although it does eliminate visual responses below this level (Schiller et al. 1974). Nevertheless, it was natural to think that cortical input might be a m a j o r contributor to relative m o t i o n selectivity. This property has been described in several cortical areas, including the m o t i o n areas M T and M S T in the superior temporal sulcus, V1, and V2 (Allman et al. 1985a, b, 1990; Saito etal. 1986; T a n a k a et al. 1986). Furthermore, cortical input, particularly that from MT, predominates in the lower half of the superficial layers where selective suppression is strongest (Lund 1972; Hubel et al. 1975; G r a h a m et al. 1979; G r a h a m 1982; Ungerleider et al. 1984). A retinal contribution might also be important, but this seemed less likely since relative direction selectivity has not been described in either retina or lateral geniculate body, although nonspecific s u r r o u n d effects have been noted in both structures (McIlwain 1964; K r u g e r 1977; Fischer and K r u g e r 1980; D a v i d s o n and Bender 1991). Furthermore, retinal input is m o s t dense in the upper half of the superficial layers (Hubel et al. 1975) where relative m o t i o n selectivity is less c o m m o n . O u r goal in the present study was to determine whether cortical input is essential for relative m o t i o n selectivity in the superior colliculus. Corticotectal cells are distributed over widespread areas of visual cortex, including areas V1, V2, V3, V4, and M T (Fries 1984; Colby and Olson 1985; L o c k et al. 1990). It would thus be impossible to eliminate cortical input at its source without destroying virtually all visual cortex. However, most corticotectal axons collect together in the corticotectal tract as they pass t h r o u g h the pulvinar nucleus of the thalamus to enter the colliculus t h r o u g h its b r a c h i u m (Bender and Baizer 1984). We thus attempted to sever the corticotectal tract by making radiofrequency lesions at the level of the pulvinar. This procedure of course interrupts a m a j o r ascending p a t h w a y from colliculus to pulvinar to cortex (Benevento and Fallon 1975; Benevento and Rezak 1976; Ogren and Hendrickson 1976; P a r t l o w etal. 1977; Harting etal. 1980), a projection system which might itself contribute, t h o u g h indirectly, to relative m o t i o n selectivity in the colliculus. To control for this possibility, in one animal we examined relative m o t i o n selectivity in the colliculus following a kainic acid (KA) lesion of the pulvinar. K A lesions effectively destroy pulvinar cells, thereby interrupting the ascending pathway, but do not d a m a g e the corticotectal tract (Bender and Baizer 1984). We found that d a m a g e to the corticotectal tract significantly reduces the degree of relative m o t i o n selectivity in the colliculus. D a m a g e to pulvinar cells alone, however, had no effect. Some of these findings have been reported previously (Davidson et al. 1986).
and recording methods are given in Bender (1982) and Davidson and Bender (1991). Briefly, each animal was ~tudied during a series of 8to 12-h semichronic recording sessions over a period of 6-12 weeks. The animal's skull had been implanted previously with hardware which provided painless fixation of the head and access to the brain. Surgery was performed under aseptic conditions. Animals were anesthetized with ketamine (35 mg/kg) or a mixture of ketamine (20 mg/kg) and xylazine (1.3 mg/kg); anesthetic adequacy was judged by the lack of reflex response to nociceptive stimuli, and anesthetic supplements (10-30% of initial dose) were given as required. At the start of recording the animal was premedicated with atropine, tranquilized with ketamine (6 mg/kg), anesthetized with 2.5% halothane in a 70% nitrous oxide oxygen mixture, and intubated with a xylocaine-coated endotracheal cannula. The eyes were immobilized with an intravenous infusion of pancuronium bromide. The pupils were dilated with cyclopentolate hydrochloride and the eyes were focussed with contact lenses on a tangent screen 0.57 m away. During recording the animals were maintained on 70% nitrous oxide-oxygen; the electrocardiogram and electroencephalogram were monitored to assess adequacy of the analgesia. The animals were recovered from immobilization at the end of the session under ketamine anesthesia. The activity of single isolated neurons was recorded using epoxycoated, tungsten microelectrodes with an impedance of 3-5 Mf~ at 500 Hz. The electrode shape was the same as that used in our previous study of relative motion in the intact colliculus. We elected not to adjust the shape, even though signal-to-noise ratios were often smaller and we sometimes had difficulty isolating cells on the side of the radiofrequency lesions, since such an adjustment could have introduced a more serious sampling problem. Three animals received a pulvinar lesion in the left hemisphere 5-13 weeks prior to recording. The lesions were intended to destroy the corticotectal tract and thus were made by radiofrequency heating at four sites within the pulvinar; the sites were selected after first making microelectrode recordings from the pulvinar and the caudal pole of the lateral geniculate nucleus (LGN). One of these animals (ST) also received a lesion of the right pulvinar. The lesion was made by microinjection of 5.2/~g KA, distributed over five sites, and was intended as a control for loss of pulvinar cells that accompanied the radiofrequency lesions. The procedures for making both types of lesions are described in detail in Bender and Baizer (1984). In the fourth animal we inactivated the corticotectal tract by injecting a local anesthetic into the pulvinar. To eliminate the possibly confounding effect of pulvinar cell inactivation, we first destroyed pulvinar cells by KA injection 10 months prior to the anesthetic injections. Anesthetic injections were made through a Hamilton syringe. The injection site was based on microelectrode recordings and was located about 1 mm posterior and 2.5 mm medial to the caudal pole of the LGN. Typically, 5 gl (range 1-20/~1) of 2% mepivacaine hydrochloride, or 0.25% bupivacaine hydrochloride with 0.001% epinephrine, was injected over a l-2min period. The injection site was marked by injecting rhodaminelabelled latex microspheres. The site was subsequently verified during histological processing and found to lie at the dorsal edge of the corticotectal fiber system (at the "L" of"PL" in Fig. 1A, AP level 4.0), relatively distant from the colliculus itself. During a single recording session, we usually studied only one colliculus cell before and after an injection. The experimental protocols for all procedures were approved by the Institutional Animal Care and Use Committee of the University at Buffalo, State University of New York.
Materials and methods
The animal faced a 65 ~ x 65 ~ tangent screen on which two independent, computer-controlled, stimulus patterns could be presented. One pattern, which we call the "target", was a small spot or slit that could be moved through the cell's classically defined receptive field. The other pattern, which we call the "background", was an array of luminous, irregularly shaped spots that filled the tangent screen
Animal preparation and recordin9 Four male Macacafascicularis monkeys weighing between 3.2 and 5.4 kg were used. Detailed descriptions of the animal preparation
Visual stimulation and procedure
248 except for a central masked area. The mask was centered on the classic receptive field and was typically 3-4 times its diameter (median 3.8, interquartile range 2.3-5.2). The mask eliminated complex interactions between target and background contours which otherwise could have occurred within the classic receptive field. The stimulus arrangement is described in detail in Davidson and Bender (1991). In the animals with radiofrequency lesions, we studied all cells that could be reliably isolated. For each cell, we first used hand-held stimuli to test for visual responsiveness, and then plotted the cell's receptive field, made an informal estimate of direction preference, and selected an appropriate size and shape for the target. We then set up the computer-driven optical systems to move both target and background. The target was moved back and forth along a single axis, or along two orthogonal axes, through the receptive field. For each target direction, we compiled two histograms: an "inphase" condition in which the background stimulus moved at the same speed in the same direction as the target, and an "antiphase" condition in which the background moved in the opposite direction at the same speed. During informal testing, we adjusted the location of the target's trajectory within the receptive field, the target speed, and the intertrial interval (ITI) in an effort to maximize the difference between the inphase and antiphase responses. During the actual data run, we usually included additional conditions for which the background was stationary during target movement, and for which the target was presented as a 500- to 800-ms flash against a blank background. The histograms were based on five to ten trials, depending on the variance in the summed response, and the trials for all histograms in a run were randomly interleaved. The ITI was typically 10-15 s, but ranged from 3 to 25 s. Response strength for each histogram was taken as the total spikes evoked by the target when it passed through the receptive field, expressed as an average per trial, minus the spontaneous discharge during a period just preceding stimulus presentation. We also examined peak discharge rate as a measure of response strength. Both response measures yielded similar estimates of relative motion sensitivity (see Davidson and Bender 1991). Unless otherwise noted, we used total spikes as a measure, since this avoided the problem of histograms that had multiple peaks. This procedure differs slightly from that used in our study of normal animals (Davidson and Bender 1991), the results of which formed one basis for comparison with the present experiment. In that study, we had usually paired 4-8 directions of background movement with each target direction in order to assess tuning for relative direction, and we had also used different background speeds to determine selectivity for relative speed. Since in the present experiment it was important to study as many cells as possible after the lesions, we reduced the number of stimulus conditions per cell. In order to make direct comparisons between the operated and normal animals, we used only the inphase and antiphase conditions from the normal study in computing relative direction selectivity. Statistical comparisons were made using the Mann-Whitney U-test for differences in magnitude, and the chi-square or Fisher exact test for differences in proportions. In the animal that received local anesthetic injections, we studied only those colliculus cells that showed strong selective suppression. Each cell was tested with a single target direction paired with three conditions of background motion: inphase, antiphase, and stationary. A fourth condition, with the target flashed against a blank background, was also included. The conditions were randomly interleaved within a block of trials, five trials per condition. We first ran two to four blocks of trials to estimate the effect of background motion prior to injection, and then lowered the injection needle into the pulvinar, injected the anesthetic, and continued to run a block of trials every 6-7 rain until the effects of the anesthetic were no longer detectable or the cell was lost, typically 30 60 min after injection. The injection needle was left in place throughout this period. Microelectrode penetrations were made throughout the retinotopic map in the colliculus, with two exceptions. We tried to avoid eccentricities less than a few degrees since the pulvinar lesions
slightly damaged the representation of this part of the visual field in the LGN. We also tried to avoid eccentricities greater than about 15 ~ The mask size required at these eccentricities substantially reduced the amount of background stimulus on the screen, which in turn could have led us to underestimate the background's influence.
Histological methods and findings The animals were perfused intracardially under deep pentobarbital sodium anesthesia with 0.9% saline, followed by 10% formalin in saline. Frozen 33-#m sections through the thalamus and midbrain were cut in the vertical stereotaxic plane. Alternate sections were stained for Nissl substance with cresyl violet and for myelin with the Heidenhain-Woelcke method. The sections were examined microscopically, and regions of cell and fiber loss were plotted on a set of standard brain drawings. Electrode penetrations made during the last 2 weeks of recording were reconstructed, and recording sites were located within the SGS or SO on the basis of small electrolytic marking lesions made during recording. For the earlier penetrations, we estimated the depth of recording sites from microdrive readings. The radiofrequency lesions were placed with the intention of interrupting fibers of the corticotectal tract as they pass through the pulvinar before collecting together to enter the tectum through the brachium of the superior colliculus. Figure 1A indicates the approximate location of this fiber system within the pulvinar. The lesions were as intended and were well localized to the pulvinar; the common region of damage among the three animals amounted to a knife-cut through the lateral part of the lateral and inferior pulvinar (Fig. 1B). In no case did the lesions encroach on the colliculus itself, and there was no evidence of cell loss within the colliculus. Examination of myelin-stained sections showed a near total loss of staining in the region of the corticotectal tract in two animals (CH and ST); in the third (OB), there was a small island of tissue between levels 4.0 and 4.5 through which normal-appearing fascicles could be seen heading toward the brachium. In all animals, staining in the brachium itself was present but substantially fainter than on the opposite side; gross shrinkage of the brachium, as is seen with longer survival times (Bender and Baizer 1984), was not obvious. There was also minor and variable damage in some structures adjacent to the pulvinar. The most caudal tip of the LGN was slightly damaged, but only in the representation of the central 2-3 ~ of the lower quadrant, near the vertical meridian. The KA lesions were intended to produce comparable cell loss, without damaging fibers, in the same areas of the pul.vinar that had been injured by the radiofrequency lesions. Figure 1C shows a reconstruction of the KA lesion in animal ST. There was total cell loss throughout most of the inferior and lateral pulvinar, and in the lateral half of the medial pulvinar. The posterior tip of the LGN, representing the central 3 4 ~ of the lower quadrant, was slightly damaged, as it had been after the radiofrequency lesions. Cell loss was thus approximately in the same regions and just as extensive as for the radiofrequency lesions.
Results W e s t u d i e d 157 cells in the three a n i m a l s with radiofreq u e n c y p u l v i n a r lesions. R e c o r d i n g sessions b e g a n 5 - 2 0 weeks after the l e s i o n a n d c o n t i n u e d for 6 - 1 2 weeks; r e c o r d i n g s f r o m b o t h right a n d left colliculi were m a d e t h r o u g h o u t this period. All cells were l o c a t e d w i t h i n the S G S or SO. T w o t h i r d s of the cells (103/157) were in the colliculus ipsilateral to the r a d i o f r e q u e n c y lesion a n d t h u s c o u l d h a v e b e e n directly affected b y d a m a g e to the corticotectal tract; cells i n the c o n t r a l a t e r a l colliculus served as a c o n t r o l p o p u l a t i o n . Since receptive field p r o p e r t i e s were
249
A
(
.o !vi
,,Cd
-:::, ( PM -.
[M_~'LGN,
,~
B
.
II ?
T-t ( /Fr
"..
~-~_-~,~,
,-;F;J
W~.~-~
'..
RF COMPOSITE
)
.....
C
KA
I
5.5
5.0
4.5
4.0
3.5
3.0
Fig. 1. A Schematic rendition of the approximate paths taken by retinotectal fibers (solid black) and corticotectal fibers (dashed areas) on their way to the colliculus (adapted from Bender and Baizer 1984). B Reconstruction of the radiofrequency lesions of the left corticotectal tract as shown on standardized coronal sections through the posterior thalamus. Areas in solid black represent tissue destruction that was common to all three animals; stippled area represents the largest extent of the damage in any of the three animals. C Re-
construction of the kainic acid lesion in the right hemisphere of one animal. Solid black indicates areas of total cell loss. Numbers at the bottom indicate approximate stereotaxic levels of the sections. Cd, caudate nucleus; H, habenular complex; LGN, lateral geniculate nucleus; LP, lateral posterior nucleus; MD, medial dorsal nucleus; MG, medial geniculate nucleus; PI, inferior pulvinar nucleus; PL, lateral pulvinar nucleus; PM, medial pulvinar nucleus; R, reticular nucleus; SC, superior colliculus; VPL, ventral posterolateral nucleus
similar a m o n g the three animals, and at different survival times, we pooled data across animals on the side of the radiofrequency lesion. On the side opposite the radiofrequency lesion, one animal had received a KA lesion9 Since colliculus receptive field properties in this hemisphere were similar to those in the two intact hemispheres, we also pooled data for all three hemispheres opposite the radiofrequency lesion. Table 1 summarizes the number of cells and survival times for the three animals. For every cell that clearly responded to a light spot or slit, we attempted to assess both directional selectivity in the absence of a moving background ("classic directionality"), and selectivity for relative direction between the target and a moving background 9 When tested for classic directionality, most cells were nondirectional. In the following sections, we discuss these nondirectional cells first9 We compare selectivity for relative motion in
cells ipsilateral to the corticotectal tract lesion with two separate control populations: cells in the colliculus opposite the corticotectal tract lesion and comparably tested cells from our study in normal animals (Davidson and Bender 1991). We then describe some unanticipated effects of the lesions on both spontaneous and visually evoked discharge patterns, and the effect of anesthetic injections. In the last section, we describe the few cells that showed classic directionality.
Selective suppression in nondirectional cells In normal animals, m a n y colliculus cells show a characteristic selectivity for relative motion. The response to a target moving through the classic receptive field is strongly suppressed when the background moves in the same
250 Table 1. Responsiveness, receptive fields, and test conditions for colliculus cells in animals with radiofrequency lesions of the corticotectal tract Left colliculus
Right colliculus
Survival time (weeks)
Number of cells
No response
5-18 9-18 13-20
33 28 42 103
6 3 5 14
Eccentricity (deg) Sizea (deg)
Median 11.0 3.0
IQR 6.0-13.0 2.0-5.0
Test conditions 2 target directions 3 target directions 4 target directions
Number of cells 36 7 46
Animal ST OB CH Total
Lesion RF RF RF
Number of cells
No response
28 8 18 54
0 0 0 0
Median 7.0 2.9
IQR 5.0-12.0 2.0-3.5
Lesion KA Intact Intact
Receptive fields
Number of cells 26 3 25
"Diameter of the receptive field's central activating area. RF, radiofrequency; KA, kainic acid; deg, degrees IQR, interquartile range direction as the target, but is only weakly suppressed, if at all, when the background moves in the opposite direction. This selective suppression for inphase background motion is about equally strong for all directions of target movement (Davidson and Bender 1991). The corticotectal tract lesions had two main effects. First, for most cells, they greatly reduced the magnitude of selective suppression for all target directions. Second, for the remaining cells, the lesions eliminated selective suppression for all but one target direction. In order to estimate the strength of selective suppression, we took the ratio of the response (total spikes evoked by the target - see Materials and methods) for the inphase condition to that for the antiphase condition. Almost all cells were tested using either two or four different target directions, thus yielding either two or four different ratios. F o r each cell, we used the smallest of these as an index of selective suppression (S). S is thus a conservative index, since it would reflect only the smallest decrease of suppression after the lesion. For comparison with normal animals, we computed the same measure for those cells in our previously published study (Davidson and Bender 1991) that had been similarly tested with either two or four target directions. Figure 2 shows the distribution of S for cells in normal and operated animals. On the side of the corticotectal tract lesion, suppression was substantially reduced (median S 0.63) when compared either with normals (median S 0.39; P < 0.0001) or with cells in the colliculus opposite the lesion (median S 0.36; P < 0.0001). The predominant effect of the lesion may have been to reduce suppression of inphase responses. After normalizing the inphase and antiphase responses to the response for a stationary background, inphase responses on the side of the lesion were substantially larger than in normal animals ( P
1
Selective Suppression Fig. 2. Distributions of selective suppression for cells from three intact animals (Norm) of a previous study (Davidson and Bender 1991), cells contralateral to the radiofrequency lesion (Contra), and cells ipsilateral to the radiofrequency lesion (RF). Selective suppression is the ratio of the inphase response to the antiphase response, both adjusted for spontaneous activity
0.0003), whereas the distribution of antiphase responses was not significantly affected. On the side opposite the lesion, the distribution of S did not differ from that in normal animals. It is apparent from Fig. 2 that some cells still showed a significant degree of selective suppression, as measured by S, after the corticotectal tract lesions. We found, however, that these cells had also been affected by the lesions. Although they showed significant suppression for the one direction of target movement selected by our index, most did not show significant suppression for other target
251
RF side anti-phase
Opposite RF lesion
in-phase
anti-phase
in-phase
lOO[L- i
.46~-=L _.
? .95
.80
. _
A__l_
28 L
24 j L
.30
directions, a marked change from cells in normal animals. Figure 3 shows a typical example of this. A cell on the side of the corticotectal tract lesion ("RF side") shows modest inphase suppression when tested with a target moving to the left, but there is no differential effect of background motion when the target moves in other directions. For a cell opposite the lesion, as shown to the right, the pattern of responses is characteristic of the normal colliculus: background motion has a similar effect for all target directions. To show this effect of the lesions for the population of cells, we first selected all cells with an S less than 0.6. (In normal animals, we used this value to indicate clear selective suppression since it represented a differential effect of background motion that was about twice the standard deviation associated with replicating a PST; see Davidson and Bender 1991.) For each of these cells, we then looked at the ratios of inphase to antiphase response for each of the target directions. On the side of the corticotectal tract lesion, few cells (17%, 6/35) showed clear selective suppression (i.e., had ratios < 0.6) for all target directions, whereas most cells in normal animals (71%, 39/55; P < 0.0005) did. In like fashion, many cells on the side of the lesion (57%, 20/35) showed clear selective suppression for only one target direction, whereas in normal animals only 20% (11/55; P < 0.0001) of the cells did. On the side opposite the lesion, the proportions of cells showing clear suppression in all, or just one, target direction were no different from those in normal animals. Figure 4 illustrates these results.
Antiphase responses. In normal animals, movement of the background in the antiphase direction typically has a weak suppressive effect when compared with the response evoked by the target in the presence of a stationary background. This was also the case for cells recorded on the side of the corticotectal tract lesion. Opposite the
Fig. 3. Inphase and antiphase responses for each of four different target directions. Histograms to the left are from a cell ipsilateral to the radiofrequency (RF) lesion, 890 ibm below the tectal surface. Significant selective suppression is seen only for leftward target movement. Arrows to the left of the histograms indicate target direction; numbers indicate the ratio of inphase to antiphase response. Receptive field eccentricity and mask diameter were 6.8 ~ and 8 ~ respectively; target speed 20~ Histograms to the right are from a cell contralateral to the radiofrequency lesion, 940/tin below the tectal surface. Note that selective suppression is strong for all target directions. Receptive field eccentricity and mask size were 14.4 ~ and 15 ~ respectively; target speed 30~ Calibration bar 8 ~ or 400 ms for the RF side, 8~ or 267 ms for the side opposite RF lesion
selective suppression
selective suppression in oll tgt directions
in only 1 tgt direction Norm(]Is RF Opposite RF I
100
r
i
50
0
50
i
100
go of Cells Fig. 4. Bars to the left of zero indicate percentage of cells showing significant selective suppression (suppression index < 0.6) in only one target direction. Bars to the right of zero indicate cells with significant suppression in all target directions tested
lesion, however, antiphase suppression was less common, and a number of cells even showed enhanced responses in the antiphase condition. Figure 5A shows a typical example of a cell whose response under the antiphase condition was significantly larger than the response with a stationary background. To illustrate this effect for the population of cells, we first normalized the antiphase responses for each target direction to the corresponding responses in the presence of a stationary background, and then used, for each cell, the largest of the normalized responses among those for the different target directions. Figure 5B, C shows the distributions of these normalized antiphase responses. For cells opposite the corticotectal tract lesion, the median response (1.1) was significantly larger than for normal animals (0.68; P < 0.0001). Furthermore, a larger proportion of cells (26 %, 8/31) showed clear enhancement (antiphase response > 1.4) than in normal
252
A
Stationary
B
Anti - phase
In - phase
Normal animals
with either two or four directions of target movement, each direction paired with inphase and antiphase background movement. We smoothed each poststimulus time (PST) histogram with a 5-point moving average, and then picked, for each cell, the largest of the four or eight responses as measured by peak discharge rate. There was no significant difference in these maximal responses between cells on the side of the corticotectal tract lesion and cells in normal animals. On the side of the lesion, the median peak response was 55 spikes/s, whereas in normals it was 49 spikes/s, and on the side opposite the lesion it was 67 spikes/s.
30
Other effects of the lesions 20 I0
I
0
C
Opposite RF lesion 30
]
N=31
,0j 0 0.0
0.5 Anti-phase
1.0
>1.4
Response
Fig. 5. A Enhancement of the antiphase response for a cell opposite the radiofrequency lesion, 540 #m below the tectal surface. Numbers to the left of the PSTs indicate response magnitude relative to that for the stationary background condition. Receptive field eccentricity and mask diameter were 3.2 ~ and 8 ~ respectively; target speed 20 ~ calibration bar 4 ~ or 200 ms. B Distribution of the magnitude of the antiphase response, normalized to the response with a stationary background, for colliculus cells in normal animals. C Corresponding distribution for colliculus ceils opposite the radiofrequency lesions
animals (4%, 3/74; P < 0.003). For cells on the same side as the lesion, the median antiphase response was 0.62 and did not differ significantly from normal.
Response strength. Although lesions or blockade of visual cortex do not weaken visually evoked responses in the superficial layers (Schiller et al. 1974), it was conceivable that lesions of the cortieotectal tract might do so. We were thus concerned that our measure of selective suppression might be distorted by such a weakness, thereby making comparison with control data difficult. We found, however, that those cells for which we evaluated selective suppression were capable of responding just as strongly as cells in normal animals. To show this, we compared the strongest response evoked by moving stimuli for cells in the operated animals with the corresponding measure in normal animals. We used all cells that had been tested
As described below, we noticed that some of the cells recorded on the side of the radiofrequency lesion seemed to discharge in an abnormal fashion. Responses to flashed stimuli were somewhat more sustained and longer in latency, and high spontaneous rates were more frequent, than was typical for intact animals. The distributions of these measures tended to be bimodal after the lesions, hinting at the emergence of cell properties not present in the normal colliculus.
Response to flashed stimuli. In normal animals, the response of most colliculus neurons to the onset of an 800ms flash consisted of a transient burst of spikes, often followed by a sustained component. The response typically had a relatively short latency, a brief rise time, a small sustained component, and was well time locked. 1 On the side of the radiofrequency lesion, most cells gave PST histograms in response to flashed stimuli that were similar to those in normal animals. However, the distributions of latency, rise time, and sustained component all indicated an increase in these measures relative to normal animals. Thus, more cells (34%, 20/59) had onset latencies exceeding 80 ms than were found in normal animals (2%, 1/45; P < 0.0001; see Fig. 6A), and more cells (36%, 21/59) had rise times exceeding 72 ms than were found in normals (4%, 2/45; P < 0.0001). For some cells, the latter reflected a considerable trial-to-trial variation in latency that was rarely seen in normal animals (Fig. 6B). Also, more cells (20%, 12/61) gave only a sustained response, with no clear transient component, than in normals (3%, 2 of 65; P < 0.005), and for those cells that did have both transient and sustained components, the distribution of the sustained component was skewed toward larger values (P < 0.01; Fig. 6C). All of these changes were most apparent for cells in the lower half of the superficial layers. For cells opposite the radiofrequency
1Rise time was defined as the interval between response onset and the time to reach 90% of the peak discharge rate that was achieved within 200 ms of stimulus onset. The strength of the sustained component was expressed as a fraction of the peak response, (Rfi,al -Spon)/(Rpeak-Spon ) where Rnnal is the average discharge rate during the last 600 ms of the flash, Rp~ is the peak discharge rate during the first 200 ms of the flash, and Spon is the spontaneous discharge rate
253
40
[] Normal=,N=45 9 RFSide,N=52
~)30 ~20
o~ 10
4
36
20
52
68
>80
Latency (msec)
9 RFSide, N=59
. :~~-~I
I
I~176 E >80
Rise Time (msec)
[] Normals,N=43 9 RFSide, N=47
40 or)
_
..I,
dk,
103 ,
Jk,.
"~ 10
2,I
0 .1
.
o
.
.9
Sustained Component
D
"6
$0 l
i
[] Normals,N=82
40-
~ _
o
1
2
4
8
16
.29
JI >16
Spontaneous Rate (spk/s)
Fig. 6A-D. Comparison of flash response parameters for cells in normal animals (Normals) and cells on the side of the radiofrequency lesion (RF Side). A Distributions of response latency. B Distributions of response rise-time. Histogram to the right shows the response of a cell on the side of the radiofrequency lesion to an 800-ms flash (solid bar beneath PS:I); note variable onset latency in rasters at the top, and gradual rise-time in the histogram below. C Distributions of the sustained component, expressed as a fraction of the transient component, of the response. Histograms to the right are from two different cells on the side of the radiofrequency lesion. Numbers at the right of each PST give the index representing the sustained component; flash duration is 800 ms. D Distributions of spontaneous discharge rate
254
A
200
ii f
h
I
'
,
_
-
~-:_-
200f I
0.2s
I
Fig. 7 A, B. "Bursty" discharge patterns from a cell ipsilateral to a radiofrequency lesion, 870/~m below the tectal surface. A Response to an 800-ms flashed stimulus (solid bar beneath PST). B Response to the same stimulus swept through the receptive field at 30~ lesion, the distributions of latency, rise time, and sustained component were all similar to their counterparts in normal animals. Spontaneous rate and discharge pattern. In normal animals, most cells had spontaneous discharge rates less than 1-2 spikes/s. On the side of the radiofrequency lesion, spontaneous rates were somewhat higher (Fig. 6D). This was most obvious for cells in the lower half of the superficial layers where the median spontaneous rate was 10 spikes/s. We also encountered some cells, ipsilateral to the lesions, that had quite abnormal discharge patterns. The pattern consisted of a series of staccato-like bursts of impulses, separated by periods of little or no activity. The "bursty" pattern was present during both spontaneous and visually driven activity. Figure 7 shows an example of the pattern in the presence of both moving and stationary targets. For some cells, the burstiness was not so flagrant, and was difficult to classify as abnormal. For others, every trial in every histogram, for both moving and stationary targets, showed it. On the side with the radiofrequency lesions, 23% (18/78) of the cells were this bursty, whereas contralateral to the lesion, none were (0/43, P < 0.0002). The incidence of bursty cells did not vary with depth in the colliculus.
lesions. To control for this possibility, we selected colliculus cells that showed strong selective suppression in an animal with an intact corticotectal tract, and then injected a local anesthetic into the pulvinar in an attempt to reversibly inactivate the tract; pulvinar cells had been destroyed previously by a KA lesion. We studied ten cells before and after injection of mepivacaine or bupivacaine. The results paralleled the effects of corticotectal tract lesions. Nine of the ten cells were affected by the anesthetic injections. Selective suppression was eliminated in five and reduced by 30-50% in two. In the remaining two cells, selective suppression was unaffected by the injection, despite a 50-75% reduction in response strength. Figure 8A shows the loss of selective suppression for one cell. Immediately following the injection, there was a decreased response to the stationary background condition, a modest reduction of the antiphase response relative to the stationary background condition, and most notably the differential response to background direction had disappeared. Sixty-five minutes later, all responses had returned close to their preinjection levels. Figure 8B summarizes the effect of anesthetic injection on responses evoked for the stationary, antiphase, and inphase background conditions for the five cells that lost selective suppression. Since the onset and duration of the anesthetic effect varied among cells, we computed response strength from two successive blocks of trials prior to injection, and from the two successive blocks after the injection which showed the greatest loss of selective suppression. A block of trials took 6 7 rain, and the maximal anesthetic effect usually occurred within 15 min after injection. This procedure underestimated the maximum effect of the anesthetic since there was invariably some recovery of suppression within the postinjection blocks of trials that we used. All responses for a given cell were normalized to the preinjection response under the stationary background condition. Immediately following the injections, responses were usually smaller for all background conditions, and the responses evoked for inphase and antiphase background motion were no longer significantly different. For the five cells, S averaged 0.34 before injection and 0.82 after injection. In contrast, antiphase suppression, expressed as the ratio of responses for the antiphase and stationary background conditions, was not eliminated, averaging 0.87 and 0.76 before and after injection, respectively. The loss of selective suppression, without loss of antiphase suppression, was thus similar to the effect of corticotectal tract lesions. However, other effects of the radiofrequency lesions were not seen after anesthetic injections. Spontaneous rates were not elevated, nor was there any hint of bursty discharge patterns. Furthermore, there was no significant change in the latency or in the sustained component of the response to flashed stimuli, as there had been following radiofrequency lesions.
Injections of local anesthetic The apparent loss of selective suppression after lesions of the corticotectal tract could conceivably reflect a sampling bias induced by the lesions, rather than a loss of selective suppression in cells that had possessed it prior to the
Directional cells In normal animals, a small proportion of colliculus cells were directionally selective. When a target was moved
255
A
pre-inj
Stationary
r L
Anti-phase
In-phase
B
1.." 0.1
2 min post-inj
to
I
post-inj
0.01 0.01
i
i
i
i
i
i
i
I
I
i
i
0,1
i
a
i
i
i
i
]
1
Pre-iniecfion Response Fig. 8A,,B. Loss of selective suppression following injection of local anesthetic into the pulvinar. A Responses from a single cell before and after injection of 5/~1of mepivacaine. Each histogram is based on five trials. The cell was 250/~m below the tectal surface. Receptive field eccentricity and mask size were 2.4 ~ and 3.8 ~ respectively; target speed 10~ Calibration bar 4 ~ or 400 ms. B Responses from fivedifferent cells:for each cell, the in-phase and anti-phase responses
differed significantly before, but not after, injection. Dashed lines connect data for each cell. Each response is based on ten trials pooled from two successive blocks of trials; horizontal and vertical bars represent 2 SE about the mean preinjection and postinjection responses, respectively.Responses for each cell are normalized to the cell's preinjection response for a stationary background
back and forth along a single axis through the receptive field, in the presence of a stationary background, the response for one direction was more than twice that for the opposite direction. We had thought that this classic directionality, like relative motion selectivity, might depend on cortical input and thus might be less frequent following corticotectal tract lesions. This was not the case. On the side of the lesion, "the proportion of directional cells was about the same (23%, 11/57) as in normal animals (15%, 18/123). Opposite the lesion, the incidence of directional cells (24%, 10/42) was likewise no different from that in normal animals. Although the corticotectal tract lesions did not reduce the number of directional cells, they did substantially reduce selectivity for relative motion, just as they had for nondirectional cells. On the side of the lesion, the median S was 0.63 for directional cells, whereas in normal animals it was 0.32 for directional cells (P
