Distribution of corticotectal cells in macaque
- First Online:
- Cite this article as:
- Lock, T.M., Baizer, J.S. & Bender, D.B. Exp Brain Res (2003) 151: 455. doi:10.1007/s00221-003-1500-y
- 143 Downloads
We compared the cortical inputs to the superficial and deep compartments of the superior colliculus, asking if the corticotectal system, like the colliculus itself, consists of two functional divisions: visual and visuomotor. We made injections of retrograde tracer extending into both superficial and deep layers in three colliculi: the injection site involved mainly the upper quadrant representation in one case, the lower quadrant representation in a second case, and both quadrants in a third. In a fourth colliculus, the tracer injection was restricted to the lower quadrant representation of the superficial layers. After injections involving both superficial and deep layers, labeled cells were seen over V1, many prestriate visual areas, and in prefrontal and posterior parietal cortex. Both the density of labeled cells and the degree of visuotopic order as inferred from the distribution of labeled cells in cortex varied among areas. In visual areas comprising the lower levels of the cortical hierarchy, visuotopy was preserved, whereas in "higher" areas the distribution of labeled cells did not strongly reflect the visuotopic location of the injection. Despite the widespread distribution of labeled cells, there were several areas with few or no labeled cells: MSTd, 7a, VIP, MIP, and TE. In the case with an injection restricted to superficial layers, labeled cells were seen only in V1 and in striate-recipient areas V2, V3, and MT. The results are consistent with the idea that the corticotectal system consists of two largely nonoverlapping components: a visual component consisting of striate cortex and striate-recipient areas, which projects only to the superficial layers, and a visuomotor component consisting of many other prestriate visual areas as well as frontal and parietal visuomotor areas, which projects to the deep compartment of the colliculus.
KeywordsSuperior colliculus Prestriate cortex Parietal cortex Frontal eye fields Dorsal stream
The superior colliculus consists of two functional divisions: the superficial or "visual" compartment and the deep or "visuomotor" compartment; visual input is critical for both compartments (Hall and May 1984; Huerta and Harting 1984; Sparks 1986; Cusick 1988; Wurtz 1996). One source of visual information is the retina, which sends retinotopically organized projections to the superficial layers (Cowey and Perry 1980; Huerta and Harting 1984). A second source of visual input is from cortex; striate, prestriate, posterior parietal, and frontal cortex all receive visual input and all project to the colliculus. Electrophysiological studies suggest that these cortical areas may be roughly divided into two classes: visual areas and visuomotor areas. Visual areas have some degree of visuotopic order, the precision of which varies among areas, and their neurons have receptive fields and prominent visual responses but lack discharge strongly related to visuomotor activity (reviews in Kaas 1997; Rosa 1997). Visuomotor areas are also visually responsive, but in addition many neurons have activity that is related to some aspect of visually guided movement, typically eye movements or hand and arm movements (Bruce and Goldberg 1985; Andersen et al. 1990; Andersen et al. 1992; Murata et al. 2000; Sommer and Wurtz 2000).
Anatomical and electrophysiological studies of projections from cortex have led to the idea that the corticotectal system, like the colliculus itself, consists of two separate components: a visual subdivision consisting of visual areas projecting to superficial layers and a visuomotor subdivision consisting of visuomotor regions projecting to deep layers (Cusick 1988). In support of this are data from anterograde tracer studies showing that several cortical visual areas, including V1, V2, and MT, project to the superficial, or visual, compartment (Wilson and Toyne 1970; Campos-Ortega and Hayhow 1972; Maunsell and Van Essen 1983; Ungerleider et al. 1984; Cusick 1988). Anatomical and electrophysiological data show that two visuomotor zones, posterior parietal cortex and frontal cortex, project to the deep compartment (Lynch et al. 1985; Cusick 1988; Faugier-Grimaud and Ventre 1989; Pare and Wurtz 1997; Sommer and Wurtz 2000). However, there are suggestions from retrograde tracing studies that this scheme is too simple, and that some visual areas may project to the deep compartment (Fries 1984; Colby and Olson 1985). Subsequent to those reports, there has been a great expansion of our understanding of prestriate cortex, and many more areas have been identified and studied. We have used retrograde tracing methods to study the corticotectal system. We wished to chart the contribution of the various prestriate areas now recognized to the corticotectal system, to assess the degree of visuotopic order in those projections, and to reexamine the validity of the concept that the corticotectal system consists of a visual and a visuomotor component. Our data support the idea of the division of the corticotectal system into two components, but suggest that these are not simply visual and visuomotor. Projections to the superficial compartment arise from only a few cortical visual areas, namely striate and striate-recipient areas, and not from visuomotor areas. Projections to the deep compartment arise from a more extensive set of visual areas other than striate cortex and striate-recipient areas, as well as from visuomotor areas. Further, we found several areas, including the densely myelinated zone of MST, 7a, VIP, MIP, and TE, that do not contribute much to either component.
Materials and methods
Subjects were one Macaca fascicularis (case RUL) and two Macaca mulatta, weighing 4.3–8.0 kg. Injection sites within the superior colliculus were selected by first recording multiunit activity with either a tungsten microelectrode or a glass micropipette (recording methods in Bender 1981). Fluorescent tracers Fast Blue (FB, 0.2–3.2 μl, 1–2%) or Diamidino Yellow (DY, 2.4 μl, 1%) were injected either with a Hamilton microsyringe or through the micropipette used for recording. Injections of FB were made in the left colliculus of all three monkeys, and DY was injected in the right colliculus of one monkey (case EDR). The Principles of Laboratory Animal Care, NIH publication number 86–23, revised 1985, were followed; the protocols for surgical and recording procedures were approved by the Institutional Animal Care and Use Committee, University at Buffalo.
After a survival time of 1 week (case RUL) or 3–4 weeks, the animals were deeply anesthetized with sodium pentobarbital and perfused through the heart with normal saline followed by 4–8% paraformaldehyde in phosphate-buffered saline. In case RUL, this was followed by 5% glycerol in paraformaldehyde. The brains were removed, blocked just behind the arcuate sulcus, photographed, cryoprotected in either 25% sucrose or in a mixture of 10% glycerol in 4% paraformaldehyde followed by 20% glycerol in 4% paraformaldehyde, and stored at 4°C. Parasagittal sections were cut at 50 μm on a freezing microtome. One series every 250 μm was immediately mounted onto gelled slides, and these were analyzed without coverslipping. Subsequently, two series of sections 1 mm apart were stained for cells with cresyl violet, or fibers with the method of Gallyas (1979).
The extent of the injections varied with depth in the colliculus. We estimated the extent of the injection separately for two different zones in the colliculus: the superficial layers (stratum zonale, stratum griseum superficiale, and stratum opticum) and the deep layers (stratum griseum intermedium, stratum album intermedium, stratum griseum profundum, and stratum album profundum). For each zone, we projected the region of intense fluorescence resulting from the injection onto a line tangent to the surface of the colliculus for a series of sections spaced 0.25 mm apart. The sections were then aligned to generate flat maps of the surface of the colliculus showing the extent of the tracer within each zone.
The retinotopic locus of tracer uptake within the superficial layers was estimated from the distribution of labeled cells in striate cortex. We superimposed the retinotopic map of striate cortex (Van Essen et al. 1984) onto our unfolding of striate cortex in such a way that the relative spacing between isoeccentricity lines remained the same in the superimposed version as in the original, and thus determined a rough measure of the retinotopic locus of labeled cells.
Distribution of labeled cells in cortex
The location of fluorescent cells in cortex was charted with a Leitz Dialux 20 microscope equipped with fluorescence optics, coupled to a computer plotting system. We plotted the location of every labeled cell on outlines of sections 0.5 mm apart. Subsequently, adjacent cell- or fiber-stained sections were projected onto these plots using a photographic enlarger, and we drew layer 4 and the borders of cortical areas onto the computer plots.
For cortex posterior to the central sulcus we unfolded layer 4 (Van Essen and Maunsell 1980), and projected the position of labeled cells and cortical areas onto this layer along lines normal to the cortical surface. For two hemispheres (cases RUL and EDR), sections 0.5 mm apart were used to generate the unfolded map. For the other two hemispheres, sections 2 mm apart (every 4th unfolded section) were used to generate the unfolding, and lines for the three intermediate sections were interpolated to complete the unfolded map at a section spacing of 0.5 mm. In case EDR, a relatively acute angle between the plane of section and the cortical surface within parts of the lunate and superior temporal sulci resulted in relatively wide spacing between the unfolded section lines; this in turn distorted the apparent density of labeled cells in those areas. We corrected the distortion by adding data from intermediate sections whenever the spacing between adjacent unfolding lines was more than about four times greater than the spacing in neighboring areas; in such regions, the map had a section spacing of 0.25 mm. To facilitate comparison, all maps are shown as right hemisphere unfoldings.
Definitions of cortical areas
In general, we follow Ungerleider (Ungerleider and Desimone 1986; Baizer et al. 1991) for the locations and boundaries of prestriate visual areas. We used myeloarchitectonic criteria and fiber-stained sections (Gallyas 1979) to determine the borders of areas V1, V2, V3d, V3v/VP, V3A, V4, V4t, MT, FST, STP, and the densely myelinated zone (DMZ) of MST. We use the term V3v/VP in the text (and V3v in the illustrations) for the ventral area adjacent to V2 since its status is still controversial (discussion in Kaas 1997; Beck and Kaas 1999; Kaas and Lyon 2001; Lyon and Kaas 2001, 2002). We use the term DP for the cortex lying most medial and dorsal on the prelunate gyrus. We use the term V4/PM in the text (and PM in the illustrations) for cortex lateral to DP and just medial to the representation of the vertical meridian in V4 on the prelunate gyrus; this region is termed V4/PM (Maguire and Baizer 1984; Youakim et al. 2001) because its status as an area distinctly different from the more lateral V4 is still uncertain. For the inferior parietal lobule and intraparietal sulcus (areas 7a, 7b, LIP, and VIP) we have followed the cytoarchitectonic and myeloarchitectonic criteria of Andersen (Andersen et al. 1987; Blatt et al. 1990). For the areas in the territory where the intraparietal, parieto-occipital, and lunate sulci converge (areas MIP, PIP, PO, and V3A), we used the myeloarchitectonic criteria of Colby et al. (1988). However, there is still considerable uncertainty about the correct parcellation of this territory (e.g., Galletti et al. 1996, 2001; Kaas 1997; Lyon and Kaas 2001). For temporal cortex, we followed Ungerleider (Boussaoud et al. 1991) and divided the inferior temporal gyrus and the anterior bank of the STS into three regions, TEO, TE, and TF.
We first describe the general features of labeling in cortex, and then compare the distribution of labeled cells across striate and prestriate cortex in the different cases.
Laminar distribution of labeled cells
Areal distribution of labeled cells
After injections that affected all layers of the colliculus, labeled cells were seen in most areas of visual cortex. In contrast, when tracer was confined to the superficial layers, labeled cells were found in a much more restricted territory. We first describe the distribution of labeled cells in the three cases in which the injections included both superficial and deep layers, and then the pattern of labeling in a case with the injection confined to the superficial layers. For each individual case, we describe first the labeling in those prestriate areas for which the visuotopic organization is reasonably well known: areas V2, V3d and V3v/VP, V4, V4t, MT, and TEO (Desimone and Gross 1979; Gattass and Gross 1981; Gattass et al. 1981, 1988; Van Essen et al. 1981; Boussaoud et al. 1991). Next, we describe labeling in the remaining areas of the superior temporal sulcus (areas FST, MST, STP) and in that dorsal region of prestriate cortex where the lunate, intraparietal, and parieto-occipital sulci converge (areas V3A, PIP, PO, and MIP). Finally, we describe labeling in areas DP and V4/PM, in posterior parietal cortex (areas 7a, 7b, LIP, and VIP), and in the temporal lobe (areas TE and TF).
Injections involving both superficial and deep layers
Injection involving both upper and lower quadrant representations (case RUL)
In areas with reasonably well known visuotopic organization, we could assess the degree to which the distribution of labeled cells matched expectations based on the pattern in V1. The overall pattern of label in these areas was more or less consistent with that in V1, but the density of labeled cells differed among areas. In V2, the distribution of labeled cells roughly mirrored that in V1. The dorsal lower quadrant representation was labeled throughout its extent—except for the central few degrees and the far periphery—with a density slightly less than in V1. Within the upper quadrant representation, labeling was light and scattered, except for a strip along the representation of the horizontal meridian, which was densely labeled. Likewise, both V3d and V3v/VP were labeled, with a lighter density of labeled cells in V3v/VP. There was an increase in the density of labeled cells along the ventral representation of the horizontal meridian, contiguous with the area of increased density of labeled cells in ventral V2. Both dorsal and ventral V4 were labeled, and again the density was greatest in dorsal V4 outside of the representation of central vision, in which there was only a scattering of cells. Anterior to V4, there was a light scattering of labeled cells throughout area TEO. In area MT, however, both upper and lower quadrant representations were densely labeled, with slightly fewer cells in the foveal part of the upper quadrant representation. Labeled cells were found in the adjacent area V4t.
Within the superior temporal sulcus, the pattern of label was somewhat more complex. Labeled cells were present in areas FST and MST adjacent to MT, but there was a region in the posterior half of the DMZ of area MST which, unlike any of the surrounding areas, was completely free of labeled cells. Labeled cells were distributed throughout the anterior part of the upper bank of the superior temporal sulcus (which includes area STP) with the density of labeled cells decreasing markedly in the most anterior part of the sulcus.
Labeled cells were densely distributed through area V3A and with more variable density throughout areas PIP and PO, but only very sparsely in area MIP. There was a continuous scattering of labeled cells dorsal to V4 in areas V4/PM and DP. On the inferior parietal lobule, patches of labeled cells, variable in density, were seen throughout area 7a. Labeled cells were densely distributed throughout most of area LIP in the lower bank of the inferior parietal sulcus. However, both the rostral quarter of this area (in the rostral tip of the sulcus) and the neighboring area VIP (in the fundus of the sulcus) contained very few labeled cells. Further anterior on the inferior parietal gyrus, area 7b contained a light scattering of labeled cells.
While labeling was dense throughout visual areas in occipital and parietal cortex, temporal cortex, by contrast, was only lightly labeled. We saw no labeled cells in the most anterior part of TE and relatively few in the posterior part. Labeled cells were present, however, in the posterior part of area TF.
Injection site involving lower quadrant representation (case EDR)
In prestriate cortex, the pattern of labeling was similar to, but more restricted than, that for case RUL, as expected with the more restricted injection site. Thus, labeled cells were found throughout the lower quadrant representation of V2, V3d, and V4 in dorsal prestriate cortex, but were almost completely absent from the upper quadrant representation of V2, V3v/VP, and V4 within ventral prestriate cortex. Furthermore, there were now many labeled cells in the representation of the fovea in V2 and V4, whereas in case RUL this region had only sparse labeling, a pattern consistent with the difference in labeling within the foveal portion of V1 for the two cases. Anterior to V4, area TEO contained only sparse and patchy labeling, as in case RUL. However, the patches of labeled cells were now restricted to the anterodorsal part of the area, where lower quadrant receptive fields are found (Boussaoud et al. 1991), whereas in case RUL the patches had been present throughout TEO. Labeled cells were found in the representation of both quadrants in MT, although the density was slightly higher in the representation of the fovea and lower quadrant. Also as in case RUL, labeled cells were found in the adjacent area V4t.
Within the superior temporal sulcus, the labeling in areas FST and MST was similar to that in case RUL, and once again there was a distinctive absence of labeled cells within the DMZ of MST. The cortex just anterior to the DMZ in the upper bank of the superior temporal sulcus, including area STP, was heavily labeled, but the cortex at the rostral tip of the sulcus was only lightly labeled.
The distribution of labeled cells in the lunate and parieto-occipital sulci (areas V3A, PIP, PO, and MIP) was similar to, though less dense than, that for case RUL, but there was little reflection of the difference in visuotopic locus of the injection site between the two cases. Dorsal to V4, there was again labeling in area V4/PM and less densely in area DP. In posterior parietal cortex, labeling in area 7a was sparse and patchy: labeled cells were virtually absent from the crown of the inferior parietal gyrus, except in its most anterior part, but were scattered through those parts of area 7a which dip into the superior temporal and inferior parietal sulci. There were very few labeled cells in area 7b. As in case RUL, labeled cells were found in the representation of both quadrants in area LIP, but were comparatively rare in both the rostral tip of this area and in the adjoining area VIP.
As in case RUL, labeling in area TE was sparse and patchy. There were a few labeled cells in the posterior part of the superior temporal sulcus, but almost none on the surface of the inferior temporal gyrus. A few labeled cells were seen in area TF.
Injection site involving upper quadrant representation (case EDL)
In prestriate cortex, the pattern of labeling in V2, V3d and V3v/VP, and V4 was almost a mirror image of that seen for case EDR, reflecting the different visuotopic locus of the injection. There was dense labeling throughout the upper quadrant representation of these areas in ventral prestriate cortex, except for the most lateral parts of V2 and V4, which contain the representation of the center of gaze (Fig. 6b). Very few labeled cells were found in dorsal V2, V3d, or V4, consistent with the very sparse label within the lower quadrant representation of V1. Anterior to V4, labeling was again sparse and patchy in area TEO, and labeled cells were almost entirely confined to the ventral half of the area, which contains the periphery of the upper quadrant representation (Boussaoud et al. 1991). This contrasts with case EDR in which labeled cells were located in the dorsal half of TEO. In area MT, labeled cells were now most dense within the upper quadrant representation, whereas in case EDR they had been more dense in the lower quadrant representation, and there were now only very few labeled cells in area V4t.
The density of labeled cells in areas FST and MST was about the same as in cases RUL and EDR. Again there was a clear absence of labeled cells in the DMZ of MST. Anterior to the DMZ, cells were distributed quite densely throughout the upper bank of the superior temporal sulcus, including area STP, but quite lightly within the rostral end of this sulcus, just as they had been in cases RUL and EDR.
Once again, as in cases RUL and EDR, there were labeled cells scattered throughout areas PO and PIP, more densely in V3A, and very sparsely in area MIP. The distribution in those areas compared to the previous case did not reflect the different visuotopic locus of the injection. Only very few cells were now found dorsal to V4 in areas V4/PM and DP, but the pattern of labeling in posterior parietal cortex was almost identical to that in the previous two cases, with dense labeling in area LIP, sparing its rostral quarter. In area 7a there were few labeled cells on the crown of the inferior parietal lobule, but labeled cells were present in the posterior bank of the intraparietal sulcus adjacent to area LIP, and anteriorly adjacent to area 7b. Areas VIP and 7b were almost free of labeled cells.
In area TE, there were only a few scattered cells in widely separated locations. Labeled cells were again found throughout the posterior part of area TF.
Injection restricted to the superficial layers
Injection involving lower quadrant representation (case KRL)
In prestriate cortex, the pattern of label in areas V2, V3d, and V3v/VP closely matched that predictable from the visuotopic locus of tracer uptake. Labeled cells were densely distributed in dorsal V2 and V3d, except in the representation of the periphery. There was also a patch of labeled cells in ventral V2 in the vicinity of the horizontal meridian representation, but there were no labeled cells in ventral V3v/VP. Labeled cells were distributed throughout area MT, although the labeling was a little more extensive in the representation of the lower quadrant.
Beyond V2, V3d and V3v/VP, and area MT, the pattern of labeling was strikingly different from that seen following injections that included all layers of the colliculus. In V4, there were a few labeled cells in the most lateral sections, but the rest of V4 was virtually free of labeling. Also in marked contrast to the previous cases, labeled cells were virtually absent from the rest of prestriate cortex. In the superior temporal sulcus, there were no labeled cells in V4t and FST, and the upper bank of the superior temporal sulcus, which had been heavily labeled in the three previous cases, contained fewer than a dozen labeled cells. Areas V3A, PIP, PO, MIP, V4/PM, and DP and the entire extent of posterior parietal cortex were also essentially free of label. Areas TEO, TE, and TF contained almost no labeled cells at all.
While bilateral labeling was easily detectable in the arcuate sulcus, in no other cortical area did we see cells labeled by the contralateral injection (and consequently, of course, there were no double-labeled cells).
Labeled cells were found in both banks of the lateral sulcus, but labeling was very sparse compared to the labeling seen in visual cortex. No labeled cells were found in the postcentral gyrus, and only an occasional cell was labeled within the central sulcus.
Our results suggest that the corticotectal system can be conceived of as two distinct subdivisions, reminiscent of a scheme proposed for the cat (Ogasawara et al. 1984). In this conception, one subdivision consists of striate cortex and striate-recipient cortical areas. It projects predominantly to the superficial, or visual, compartment of the colliculus, and projections are visuotopic. It thus is termed the "visual subdivision." The second subdivision of the corticotectal system arises primarily from areas relatively late in the cortical hierarchy, including many visual areas of the "dorsal stream" (Ungerleider and Mishkin 1982) as well as visuomotor and polysensory areas. It projects to the deep compartment of the colliculus, and maintains visuotopic order to a far lesser degree than the visual subdivision. We call it the visuomotor subdivision in view of its relation with the visuomotor compartment of the colliculus. We shall discuss our findings in light of earlier observations on the connections, visuotopic organization, and functions of the areas comprising the corticotectal systems in the monkey, beginning with the visuomotor subdivision.
Visuomotor subdivision of the corticotectal system
After injections affecting all layers of the colliculus, labeling in cortex was widespread. In general, the set of areas in which we saw labeled cells is in agreement with three earlier reports (Fries 1984; Colby and Olson 1985; Distler and Hoffman 2001). In all cases there were labeled cells in V1, V2, V3, and MT. Since those areas, and only those areas, were labeled after the injection restricted to the superficial layers, we consider them to comprise the visual component. We believe that the label in V1, V2, V3, and MT seen with the injections of tracer into all layers resulted from uptake of tracer only in the superficial layers; this is consistent with data from anterograde studies (discussed below). For the remaining areas in which there were labeled cells, there were differences in the degree to which topographic order was reflected in the pattern of projections. Areas in which projections were topographically arranged include V4 and its immediate neighbors V4t and TEO. There was a second set of areas, including FST, MST, V3A, PO, PIP, LIP, and DP, in which the distribution of labeled cells was far less predictable from the visuotopic locus of the injections. The absence of clearly different distributions of cells between cases could reflect either a lack of visuotopic organization within the cortical area itself or widely dispersed projections from an area which itself has some degree of visuotopy. Our results cannot resolve that question as they allow only a rudimentary assessment of topography based on injections involving quadrantic representations.
The distribution of labeled cells in V4 was consistent with descriptions of the visuotopic organization of V4 (Maguire and Baizer 1984; Gattass et al. 1988; Youakim et al. 2001). Areas V4t and V4/PM were labeled only following colliculus injections involving the lower quadrant, again consistent with those studies. Topographic order in the projection from TEO was less obvious than for V4, possibly for two reasons: first, label was much sparser in TEO than in V4, making it difficult to evaluate differences between the cases, and, second, the known variability in visuotopic organization in TEO in different animals could have obscured comparison (Boussaoud et al. 1991).
We found no clear evidence for topographic order in projections to the colliculus from FST or MST; the distribution of labeled cells did not differ systematically between the two cases with upper versus lower quadrant injections. Electrophysiological work has shown only crude visuotopic organization in FST and MST (Desimone and Ungerleider 1986) and an anterograde study even failed to reveal projections from either FST or MST to the colliculus (Boussaoud et al. 1992). The latter discrepancy from our findings may simply reflect differences in sensitivity of the two techniques.
We did not see clear differences among cases in the distribution of cells in the remaining areas: V3A, PO, PIP, LIP, and DP. Understanding of the parcellation of this cortex and the visuotopic organization of component areas is still incomplete and rather controversial (see discussion in Beck and Kaas 1999; Lyon and Kaas 2001, 2002). There is some evidence for topographic organization in PO (Colby et al. 1988), but the adjoining region termed "V6A" does not seem retinotopic and the borders between the areas are hard to determine (Galletti et al. 1996). Area PO may be partially co-extensive with a different topographically organized area, V6 (Galletti et al. 2001). Area V3A is thought to have disorderly representations of both quadrants, with the upper quadrant anterior and the lower posterior (Zeki 1978; Beck and Kaas 1999). Neither the borders nor the organization of PIP are known with any certainty. The visuotopic organization of DP likewise is not well defined, although there is some evidence for receptive fields in both quadrants (Maguire and Baizer 1984). For LIP, there is electrophysiological evidence for visuotopy (Blatt et al. 1990) and we found that the location of labeled cells differed between the two cases with single quadrant injections, but not in a way that was predictable from the report of Blatt et al. Our finding of labeled cells in LIP and DP is consistent with anterograde studies (Benevento and Davis 1977; Asanuma et al. 1985; Lynch et al. 1985). Evidence for at least some degree of order in the projections from V3A, PO, and LIP has been suggested by a brief report (Colby and Olson 1985) in which different fluorescent tracers injected into different locations of a single colliculus resulted in largely non-intermingled populations of cells within each of these cortical areas. It may be that the location and organization of these areas is sufficiently variable among animals that comparing locations of labeled cells across animals, as we did, would fail to reveal organized projections. It is also conceivable that projections from these areas are topological but not visuotopic, the topological order obeying some other dimension than retinal order.
The general distribution of corticotectal cells in the frontal lobe was similar to that described earlier (Distel and Fries 1982). We supplement their study, however, by showing that some single cells in the frontal eye fields send axonal branches to both colliculi. Anterograde studies have shown projections to the colliculus from prefrontal association cortex, the frontal eye fields, and the supplementary eye fields, terminating primarily within the intermediate gray (Kunzle et al. 1976; Leichnetz et al. 1981). Fibers can reach the contralateral colliculus through the tectal commissure or through the contralateral tegmentum (Leichnetz et al. 1981).
Visual subdivision of the corticotectal system
Of the many cortical areas labeled after injections affecting both superficial and deep layers, only four areas, V1, V2, V3, and MT, were labeled after an injection into the superficial layers of the colliculus, a result consistent with data from anterograde studies. The older data for V1 are summarized in Hall and May (1984) and Ungerleider et al. (1984). The projection from V2 to the superficial layers is consistent with data of Ungerleider (personal communication) and Campos-Ortega and Hayhow (1972, case MmVc 21), and two studies describe a projection from MT to superficial layers (Maunsell and Van Essen 1983; Ungerleider et al. 1984). Recent reports (Lia and Olavarria 1996; Abel et al. 1997) further analyzed the distribution of corticotectal cells in V1 and V2 in relation to compartments defined by cytochrome oxidase activity. These studies found that corticotectal cells are preferentially found in the interblob compartment of V1 and the thick stripes of V2. We have no direct measure of these compartments since we did not stain for cytochrome oxidase. Further, since we don't know the angle of our sections relative to the stripes we cannot compare our cell spacing with that reported previously. However, the regular spacing of labeled cells that we observed could well reflect the fact that the cells are indeed in compartments that are themselves more or less regularly spaced. The visuotopic organizations of V1, V2, and MT have been extensively studied and are well established (Gattass and Gross 1981; Gattass et al. 1981; Van Essen et al. 1981, 1984). There has been considerable controversy about the extent and organization of V3, but there is agreement on the basic plan that the upper quadrant is represented ventrally and the lower quadrant dorsally (Gattass et al. 1988; Felleman and Van Essen 1991; Kaas 1997; Lyon and Kaas 2001). Comparison of the distributions of labeled cells in V1, V2, V3, and MT among all of our cases supports the idea that the projections from these areas are visuotopic. In the case of MT, however, while labeled cells were most numerous in the representation of the injected quadrant, there was labeling in the other "uninjected" quadrant, and it was heavier than in the uninjected quadrant representations in V1, V2, and V3. This may reflect a greater degree of divergence in projections from MT, less precision in its visuotopic organization, or both. Extensive divergence in projections has not been noted in anterograde studies but weak projections to non-homotopic points might be hard to detect with the anterograde technique.
Our evidence suggests that projections to the superficial layers arise only from V1, V2, V3, and MT. However, data from other studies have suggested projections to the superficial layers from additional prestriate areas and from frontal cortex. Fries (1984, case 7) reported labeled cells in areas V3A, V4, 6, 7, and 8, as well as in V1, V2, V3, and MT in a case with a superficial layer injection. However, that injection may not actually have been restricted to the superficial layers as the dense core of the injection clearly extended into the intermediate layers. Webster et al. (1993) found projections from TEO to the colliculus using anterograde tracing, whereas we saw no labeled cells in TEO when the retrograde tracer injection was restricted to the superficial layers. The reason for the discrepancy is not clear. However, Webster et al. found anterograde label only deep to the middle of the stratum opticum, and projections so deep in the superficial compartment might not have been labeled in our superficial injection case; in our material, fluorescence clearly extended through the stratum griseum superficiale, and was not detectable below the stratum opticum. Finally, both anterograde and retrograde data have suggested a modest projection to the superficial compartment from frontal cortex; our injection may not have been large enough to engage it (Leichnetz et al. 1981; Huerta et al. 1986; Distler and Hoffman 2001).
The areas that project to the superficial layers of the colliculus also project to an adjacent nucleus implicated in eye movement control, the nucleus of the optic tract (NOT; Distler and Hoffman 2001). However, there appears to be a difference in the relative density of projections from these cortical areas to the two brainstem targets. For the NOT, the densest input is from MT (Distler and Hoffman 2001), whereas for the superficial layers of the colliculus we found the densest projection to be from V1. In addition, cortex in the STS outside of MT (including areas FST and MST) projects to the NOT, but was not labeled after our superficial layer injection.
The visual areas contributing to the visual subdivision of the corticotectal system thus appear to include V1 and only those prestriate areas that receive direct projections from the entire representation in V1: V2, V3, and MT (for review of V1 projections see Felleman and Van Essen 1991). Projections from striate cortex to areas V4, V3A, PO, and PIP have been described, but these projections are either inconsistent among cases or from only part of the representation, for example foveal for V4 and peripheral for V3A, PO, and PIP (Zeki 1969, 1978; Van Essen et al. 1986; Colby et al. 1988; Beck and Kaas 1999). Such recipients of partial V1 input were not labeled by our superficial layer injection. In that context we should note that Distler and Hoffman (2001) reported labeled cells in area V4 after a superficial layer injection of a retrograde tracer into the foveal representation in the colliculus. Further, since our superficial layer injection did not involve the representation of the most extreme periphery, we cannot fairly rule out the possibility of input to the superficial layers from the striate-recipient regions of V3A, PO, and PIP.
Corticotectal and corticopontine systems
Another major outflow of cortex is to the cerebellum via the pontine nuclei (Glickstein et al. 1980, 1985; Cohen et al. 1981). While the bulk of projections are from motor areas, there are sparse projections from the representation of the periphery in striate cortex and from visual areas in the superior temporal and intraparietal sulci, including MT (Glickstein et al. 1985). The visual areas that project to pontine nuclei in monkey, then, also project to the colliculus. In cat, electrophysiological data have shown that every visual corticotectal cell in two visual areas, area 18 and lateral suprasylvian cortex, also projects to pontine nuclei (Baker et al. 1983). It is therefore possible that visual corticotectal cells in monkey may likewise send an axon to the pontine nuclei. The converse is clearly not true: the majority of corticopontine cells do not project to the colliculus (Glickstein et al. 1980, 1985). Another difference between corticotectal and corticopontine projections is in preservation of magnification factor. Within labeled areas, our data showed that the density of labeled cells did not appear to differ across the representation, suggesting that cortical magnification factor is preserved in this projection. This is also true of the corticotectal projection in the cat (Cohen et al. 1981), whereas the corticopontine projection emphasized the visual field periphery.
Functions of the corticotectal system
Cortex does not provide the visual input critical for classical receptive field properties in the superficial layers (Schiller et al. 1974); this presumably comes from the retina. However, the response of these superficial layer cells to a stimulus within their classical receptive fields can be powerfully modulated by the visual context in which that stimulus occurs. The visual component of the corticotectal system may be the major player in such context-sensitive processing. For example, sensitivity to differential motion between a receptive-field stimulus and the surrounding visual field has been shown throughout the superficial layers (Bender and Davidson 1986; Davidson and Bender 1991) and that sensitivity is entirely dependent on input from MT (Davidson et al. 1992; Joly and Bender 1997). Other forms of local-global interaction are known in V1 and V2 (von der Heydt et al. 1984; Lamme 1995; Sillito et al. 1995; Zipser et al. 1996; Levitt and Lund 1997), and such context-sensitive information may also be sent to the colliculus, although evidence for this is still lacking. Additionally, the visual component of the corticotectal system may be even the major contributor to the well-known ascending tecto-pulvinar-cortical pathway (discussion in Hall and May 1984). If so, however, the nature of that contribution remains enigmatic. Finally, signals from the visual component may also find their way into the deep compartment via a multisynaptic intracollicular route for use in visuomotor control (Behan and Appell 1992; Helms et al. 2002).
The cortical areas projecting to the deep compartment include many visual areas. This diverse cortical input may be necessary to mediate the diversity of motor responses, including saccades, fixation, smooth pursuit, vergence, and coordinated head and eye movements, in which the superior colliculus participates (Wurtz 1996; Freedman and Sparks 1997a, b; Basso et al. 2000; Chaturvedi and Van Gisbergen 2000; Krauzlis et al. 2000). It is easy to imagine that each type of visuomotor response utilizes visual inputs in differing combinations from many different areas.
In addition to providing a rich visual input, some of these projections may provide nonvisual input (Jay and Sparks 1987; Wallace et al. 1996). Auditory responsiveness in the deep layers, for example, might depend on input from the upper bank of the superior temporal sulcus, a region where cells have polymodal responses (Bruce et al. 1981). Some inputs may facilitate multisensory integration as in cat (Jiang et al. 2001). Finally, the visuomotor input to the colliculus may participate more directly in the generation of motor commands.
Visual areas that are largely excluded from the corticotectal system
We have found several cortical territories which appear not to project to the colliculus. Two of these areas are visual (MSTd and TE) and the others are visuomotor. In each of our cases, there was an absence of labeled cells within a circumscribed region in dorsal MST. A similar "gap" in labeling in the upper bank of the superior temporal sulcus is also apparent in the data of Fries (1984, see Fig. 9B), but see Distler and Hoffman (2001), who show labeled cells in all of MST after a tracer injection into the deep layers of the colliculus. This region was called "MSTd" by Wurtz and collaborators (Komatsu and Wurtz 1988; Newsome et al. 1988). It occupies part of the DMZ of MST (Desimone and Ungerleider 1986). Several studies have found cells in MSTd that respond to complex movement of large random dot patterns, and the area is strongly implicated in the analysis of large-field motion (Saito et al. 1986; Tanaka et al. 1986). Thus, cells in MSTd probably participate in the analysis of optic flow (Duffy and Wurtz 1997a, b), and contribute to the perceptual analysis of movement of self relative to the environment.
Area TE appeared to be largely excluded from the corticotectal system. In the temporal lobe, labeled cells in our material were seen only sparsely in TEO and were virtually absent from area TE. While this conflicts with Fries (1984), who found labeled cells as densely distributed over the inferior temporal cortex as within V1, anterograde studies have shown only sparse projections from TE to the colliculus (Whitlock and Nauta 1956; Baizer et al. 1993). Since inferotemporal cortex seems concerned primarily with objects and patterns already fixated (Ungerleider and Mishkin 1982), perhaps it sets to work only after the colliculus has completed its task of shifting gaze to a relevant object, and thus has little to tell the colliculus. Neither does it need to engage the motor mechanisms of the cerebellum, apparently, as inferotemporal cortex lacks projections to the pontine nuclei (Glickstein et al. 1980, 1985). Nonetheless, inferotemporal cortex clearly accesses motor systems via projections to basal ganglia (Saint-Cyr et al. 1990; Baizer et al. 1993).
In posterior parietal cortex, five different visual areas have been proposed: LIP, VIP, MIP, 7a, and AIP; all participate in some aspect of visuomotor function (Andersen et al. 1987, 1990; Blatt et al. 1990; Colby and Duhamel 1991; Colby et al. 1993; Sakata et al. 1995; Murata et al. 2000). Of these areas, only LIP was densely labeled. LIP is distinguished by its strong connections with the frontal eye fields, which work in close association with the colliculus (Andersen et al. 1992; Lynch 1992; Lynch et al. 1994; Sommer and Wurtz 1998). LIP can thus influence the colliculus both directly and indirectly via the frontal eye fields. The other parietal areas, while largely excluded from direct tectal influence, may influence visuomotor function by way of connections to other subcortical motor structures such as the cerebellum or the basal ganglia (Glickstein et al. 1985; Saint-Cyr et al. 1990).
In summary, our results show that the corticotectal system, like the colliculus, has two distinct components. Cortical inputs to the superficial compartment arise from relatively few visual areas and are topographically organized. Inputs to the deep compartment arise from other visual areas, polysensory areas and visuomotor areas, suggesting that cortex and colliculus work together to mediate many different aspects of visuomotor responses.
We thank Karen Zernhelt-Wolf for skilled assistance with histology. Supported in part by grants EY02254 and NS32936 (D.B.B.), and MH42130 and the Whitehall Foundation (J.S.B.).