THE JOURNAL OF COMPARATIVE NEUROLOGY 294262-280 (1990)

Topographic Organization in the corticocortical Connections of Medial Agrmular Cortex in Rats R.L. REEP, G.S. GOODWIN, AND J.V. CORWTN Departments of Physiological Sciences and Neuroscience, University of Florida, Gainesville, Florida 32610 (R.L.R., G.S.G.); Department of Psychology, University of New Orleans, New Orleans, Louisiana 70148 (J.V.C.)

ABSTRACT Medial agranular cortex (AGm) is a narrow, longitudinally oriented region known to have extensive corticocortical connections. The rostral and caudal portions of AGm exhibit functional differences that may involve these connections. Therefore we have examined the rostrocaudal organization of the afferent cortical connections of AGm by using fluorescent tracers, to determine whether there are significant differences between rostral and caudal AGm. Mediolateral patterns have also been examined in order to compare the pattern of corticocortical connections of AGm to those of the laterally adjacent lateral agranular cortex (AGI) and medially adjacent anterior cingulate area (AC). In the rostrocaudal domain, there are notable patterns in the connections of AGm with somatic sensorimotor, visual, and retrosplenial cortex. Rostra1 AGm receives extensive afferents from the caudal part of somatic sensorimotor area Par I, whereas caudal AGm receives input largely from the hindlimb cortex (area HL). Middle portions of AGm show an intermediate condition, indicating a continuously changing pattern rathe? than the presence of sharp border zones. The whole of the second somatic sensorimotor area Par I1 projects to rostral AGm, whereas caudal AGm receives input only from the caudal portion of Par IT. Visual cortex projections to RGm originate in areas Ocl, Oc2L and Oc2M. Connections of rostral AGm with visual cortex are noticeably less dense than those of mid and caudal AGm, and are focused in area Oc2L. The granular visual area Ocl projects almost exclusively to mid and caudal AGm. Retrosplenial cortex has more extensive connections with caudal AGm than with rostral AGm, and the agranular and granular retrosplenial subregions are both involved. Other cortical connections of AGm show little or no apparent rostrocaudal topography. These include afferents from orbital, perirhinal, and entorhinal cortex, all of which are bilateral in origin. In the mediolateral dimension, AGm has more extensive corticocortical connections than either AGI or AC. Of these three neighboring areas, only AGm has connections with the somatic sensorimotor, visual, retrosplenial and orbital cortices. In keeping with its role as primary motor cortex, AG1 is predominantly connected with area Par I of somatic sensorimotor cortex, specifically rostral Par I. AG1 receives no input from visual or retrosplenial cortex. Anterior cingulate cortex has connections with visual area Oc2 and with retrosplenial cortex, but none with somatic sensorimotor cortex. Orbital cortex projections are sparse to AG1 and do not appear to involve AC. The rostrocaudal topography demonstrated within AGm is consistent with known functional differences between rostral and caudal AGm, suggesting that rostral AGm be considered a multimodal association area with properties of supplementary motor cortex. In addition, all of AGm appears to rep-

Accepted October 26,1989.

0 1990 WILEY-LISS, INC.

CORTICAL CONNECTIONS OF AGM

263

resent a frontal eye field. The mediolateral organization described indicates that AGm is distinct from the neighboring areas AGl and AC because of its more diverse corticocortical connections. Key words: cerebral cortex, axon tracers, topography

Medial agranular cortex (AGm) extends from the frontal pole to the rostral border of retrosplenial cortex, overlies the cingulum bundle, and is positioned between the anterior cingulate (AC) and lateral agranular (AG1) cortices. Our previous anatomical studies have demonstrated that AGm participates in corticocortical connections with somatic sensorimotor, visual, auditory, orbital, retrosplenial, and entorhinal areas (Reep et al., '84, '87). On the basis of its anatomical and functional properties, we have suggested that AGm is a multimodal association cortex with strong ties to the motor system (Reep et al., '87). The rostral portion of AGm has functional attributes not seen in more caudal parts of AGm. Results of microstimulation mapping studies indicate that AGm as a whole mediates movements of the vibrissae, eyes, and head (Donoghue and Wise, '82; Gioanni and Lamarche, '85; Hall and Lindholm, '74; Neafsey et al., '86; Sanderson eta]., '84; Sinnamon and Galer, '84). However, head orientation movements appear to be mediated primarily by rostral AGm (Sinnamon and Galer, '84). In addition, rostral AGm contains a second motor area for hindlimb movements that appears to be distinct from the main hindlimb representation in AG1 (Neafsey and Sievert, '82; Neafsey e t al., '86; Sanderson et al., '84). Furthermore, corticospinal projections of AGrn originate only from rostral AGm (Miller, '87). Area AGrn appears to be a component of the circuitry involved in directed attention. Unilateral destruction of AGm results in neglect of visual, somatic sensory, and auditory stimuli, whereas lesions located in AGl immediately adjacent to AGm do not (Corwin et al., '86; Cowey and Bozek, '74; Crowne and Pathria, '82; Crowne e t al., '83; Sinnamon and Charman, '88; Vargo et al., '88, '89). However, rostral and caudal AGrn appear to play different roles in directed attention, since neglect produced by lesions of caudal AGrn is more profound and longer lasting than neglect resulting from lesions of more rostral AGm (King and Corwin, '89). Although the exact neuroanatomical substrates involved in multimodal attention and neglect are not known, those cortical areas most often involved in neglect in primates are multimodal association fields with extensive corticocortical connections (Heilman et al., '85). Thus, it is possible that the relatively widespread corticocortical connections of AGrn play a role in these phenomena in rats. The above-described functional distinctions between rostral and caudal AGm suggested to us that the corticocortical connections of these portions of AGm might also be distinct. Thus the present study was undertaken in order to map in detail the afferent corticocortical connections of AGm and to delineate any significant rostrocaudal topography in these connections. Second, mediolateral topography has been investigated in order to determine whether the corticocortical connections of AGm differ markedly from those of the adjacent areas AG1 and AC. Mediolateral patterns are potentially significant because of the role of AGm as eyehead orientation cortex, in contrast to the somatic sensori-

motor role of AG1. Such patterns may also help define the circuitry involved in attention and neglect, phenomena which appear to depend on AGm but not AGl.

METHODS A total of 66 albino rats were used in this study. Each rat received an injection of Fluoro-Gold (FG) into the left cerebral cortex. The majority of injections were placed in AGm, but in 12 cases adjacent areas received injections so that rostrocaudal and mediolateral topographic patterns could be examined. In five brains separate injections were made into AGrn and AG1 by using various combinations of fluorescent tracers. These double injection experiments were useful in corifirming topographical findings derived from the results of several single-injected brains. A 2-4 0.0 solution of Fluoro-Gold (Fluorochrome, Inc.), Diamidino Yellow, or Fast Blue (Sigma) was made in double distilled water and delivered intracerebrally by pressure injection. In the first method, animals anesthetized with Ketamine/Rompun were given stereotaxic pressure injections of 0.05-0.10 pl through a 5 pl Hamilton syringe equipped with a 33 gauge beveled needle tip. In the second procedure, a glass monofilament micropipette (A-M Systems, Inc.) with tip diameter of 20-30 pm was back-filled with the aqueous FG or TB and attached to a filled 1 pl Hamilton syringe with 30 gauge needle tip. The junction was then sealed with soft melted wax and the pipette-syringe system checked for successful delivery of a drop from the tip of the pipette. This apparatus was then mounted on the stereotaxic unit and injections of 0.02-0.08 pl made into the brains of anesthetized animals. More recently we have used

Abbreviations ac AC AGI AGm AId AIP CL ERC FL HL MO OCl Oc2L Oc2M Par I Par I1 PR RSa RSg TE TT VLO

anterior commissure anterior cingulate cortex lateral agranular cortex medial agranular cortex dorsal agranular insular cortex posterior agranular insular cortex claustrum entorhinal cortex caudal forelimb region of somatic sensorimotor cortex caudal hindlimb region of somatic sensorimotor cortex medial orbital cortex visual cortex, granular portion visual association area Oc2, lateral portion visual association area Oc2, medial portion area Par I of somatic sensory cortex area Par I1 of somatic sensory cortex perirhinal cortex retrosplenial cortex, agraniilar portion retrosplenial cortex, granular p r t i o n auditory area TE taenia tecta ventrolateral orbital cortex

R.L.REEPETAL.

264

U

Q

4U

5E

F

2

-@ *al

--

CORTICAL CONNECTIONS OF AGM

265

Fig. 2. A Photomicrograph of a typical injection site. This is from case AGm 77, and corresponds to the caudal section represented in Fig. 1. Note the central zone of dense fluorescence surrounded by a halo of reduced extracellular labeling. Arrowheads denote mediolateral extent

of AGm. White matter indicated hy wm. B: Layer V labeling is seen immediately lateral to the injection site, in areas AGl and HL, after injections in mid or caudal AGm (case AGm 31).

a Picospritzer (General Valve Corp.) for controlled pressure injections (1-2 pulses, 20-40 psi, 3-10 msec) through micropipettes with tip diameters of 20-30 pm. We have used postoperative survival times ranging from 2 to 10 days with little observable difference in the FG results. After the survival period, animals were given 200 units of Heparin and a pentobarbital overdose, then perfused intracardially with phosphate-buffered saline at 37°C until the blood was fully cleared, followed by 4 7; phosphate-buffered paraformaldehyde with 5% sucrose. The descending aorta was clamped, and fluids gravity fed from a height of about 1 m through flexible tubing into an 18 gauge needle inserted into the left ventricle. The brain was extracted and postfixed in cold 30% sucrose fixative for 1-3 days. Coronal frozen sections were cut at 30-40 pm on a sliding microtome and collected at room temperature into dilute buffer. Sections were stored in the refrigerator and mounted within a few days. After mounting, slides were dried for 1-24 hours a t

room temperature in the dark. Two spaced series of sections were mounted. The first series was dehydrated through an ascending ethanol series (70, 95, 100, l00T8),defatted in xylene, and coverslipped with Fluoromount. A second series of slides was stained for cresyl violet and used for cytoarchitectural orientation. Slides were examined on a n Olympus BH-2 microscope. FG-labeled cells and True Blue-labeled cells were visualized under reflected light fluorescence using an excitation wavelength of 334 nm (UV filter), whereas cells labeled with Diamidino Yellow or Fast Blue were viewed hy using an excitation wavelength of 405 nm (V filter). Designation of most cortical areas has been done according t o the terminology of Zilles and Wree ('85). Areas AGm and AG1 are easily distinguished by their rostral location and agranularity. AGm exhibits a pale staining layer 111 compared to AGl (Donoghue and Wise, '82; Donoghue and Parham, '83). Most of the first somatic sensory cortex (as functionally defined) is contained in area Par I, which is characterized by a granular layer IV interrupted by dysgranular zones, and a layer V containing medium to large pyramidal cells and having a pale sublayer Va (Welker, '71). The representation field for the head and mystacial vibrissae composes the majority of area Par I. More rostrolaterally are representations for the nose, rostral vibrissae, and perioral region. All of these fields exhibit discontinuous granular zmeb in layer IV, organized as cell aggregates and harrels (Chapin and Lin, '84; Welker, '71,'76; Welker and Woolsey,

Fig. 1. Fluoro-Gold injection sites for twelve representative cases in five groups, with three coronal sections per brain. Within each group the cases are arranged in rostrocaudal order, and brain numbers are indicated in the upper left corner for each case. For a given injection site, the black zone represents the dense central deposit of Fluoro-Gold, whereas the stippled zone indicates a less dense surround. Significant retrograde transport occurs from both these zones, based on the distribution of contralateral homotopical labeling (see text). The solid line represents the overall extent of granular layer IV in somatic sensorimotor cortex.

266

R.L.REEPETAL.

AGm 29

Figure 3

CORTICAL CONNECTIONS OF AGM ’74). Smaller remaining portions of Par I are devoted to representations for the neck, trunk, and tail, and are located medial and caudal to the vibrissae area (Welker, ’71). The second somatic sensory area (Par 11)is far lateral in location, has a sharply reduced granular layer IV, and small pyramidal cells in layer V (Welker and Sinha, ’72). The primary (caudal) hindlimb and forelimb fields exhibit an amalgam of functional and anatomical attributes typical of sensory and motor cortex; thus they have been designated separately from Par I (Zilles and Wree, ’85). Caudal hindlimb cortex (HL) is identifiable hy its medial location, granular layer IV, and large pyramidal neurons densely packed in deep layer V. Caudal forelimb cortex (FL) lies rostrolateral to area HL and exhibits similar cytoarchitecture but with somewhat smaller, less densely packed pyramidal cells than in HL (Donoghue et al., ’79; Neafsey et al., ’86; Sanderson e t al., ’84; Welker, ’71, ’76; Wise and Donoghue, ’86; Zilles and Wree, ’85).

RESULTS Injection sites Figure 1illustrates the distribution of injection sites in 12 representative cases. Two zones, a dense core and less dense surround, are represented for each injection site. The dense core region fluoresces as a rather uniform bright zone, while in the surround individual labeled cells are prominent (Fig. 2A). The outer edge of the surround region is readily defined by a sudden decrease in the density of labeled cells. Retrograde transport appears to occur from both zones, judging by the distribution of contralateral homotopical labeling (see Figs. 3 5,6B). Injection sites centered in AGm range in location from pregenual levels to the beginning of retrosplenial cortex. For descriptive purposes, we consider three regions of AGm: “rostral AGM,” located rostral to the level of the genu of the corpus callosum; “mid AGm,” between the genu and the crossing of the anterior commissure (usually near bregma); “caudal AGm,” caudal to the anterior commissure and extending to retrosplenial cortex. Each of these regions is 2.0-2.5 mm in length (see Paxinos and Watson, ’86) and is represented by several cases with injections involving all layers of AGm. Therefore the resulting cortical labeling patterns may be considered accurate indicators of rostrocaudal topographic organization. In groups AGm/AC and AGm/AGl the injection site is centered in medial or lateral AGm and encroaches on anterior cingulate or lateral agranular cortex, respectively. These cases, in conjunction with those having injections confined to AGm, AC, or AG1, are useful in defining mediolateral topographic patterns. Brains eliminated from consideration were those in which there was extensive necrosis a t the center of the injection site, significant encroachment of the white matter by the injection site, and those having very light injections. In the Fig. 3 (A-H). T h e distribution of retrograde cortical labeling resulting from a large injection in medial agranular cortex. Spaced coronal sections illustrating the injection site and distribution of labeled cells in case AGm 29. Only cortical labeling has been plotted. Note the extensive labeling in caudal somatic sensorimotor cortex (E) and in visual cortex (F-H). There is also pronounced labeling bilaterally in orbital cortex (A-B). In this and subsequenb figures, a solid line represents the exten1 of it well-defined granular layer IV in somatic sensorimotor cortex. Breaks in the line represent dysgranular zones.

267 text below results are sometimes discussed for brains that are not illustrated, but only when their injection sites are comparable to brains which are represented pictorially.

Mediolateral topography Three representative cases have been chosen to illustrate the overall pattern of cortical labeling seen with injections in AGm, laterally adjacent AGI, and medially adjacent AC. More detailed consideration of the labeling patterns in specific cortical areas is treated in separate sections below. Case AGrn 29. The injection site in this brain is large and centered in rostral AGm a t the level of the forceps minor (Fig. 3B), but also extends into mid AGm (Fig. 3 C ) . It does not encroach on the white matter. Rostrally there is extensive bilateral labeling in the ventrolateral (VLO) and medial (MO) areas of orbital cortex, heavier ipsilaterally (Figs. 3A, 6C). This orbital cortex labeling appears to be continuous with bilateral labeling found just caudally in the claustrum (Fig. 3B-D). In contralateral AGm there is heavy cell labeling whose rostrocaudal extent is somewhat greater than that of the injection site (Fig. 3A-D). The caudal taenia tecta just below the genu contains labeled cells ipsilaterally (Fig. 3C). Caudal to the injection site, at the level of the anterior commissure, a few labeled cells are present in the forelimb (FL) area (Fig. 3D). Caudal to the level of the fimbria, labeling is extensive in areas Par I and Par I1 of somatic sensorimotor cortex (Fig. 3E). This labeling is organized as clusters of labeled cells in layers IIDII, V, and a deep stratum we have termed layer VII (Reep and Goodwin, ’88). There is little labeling of layer VI and none in layer IV. Some of the labeled rlristers are in dysgranular regions of Par I. Labeling in Par I1 is continuous with that in the rostral part of auditory area TE (Fig. 3F). Labeled cells in caudal Par I1 and TE merge ventrally with labeling in the insular, perirhinal, and entorhinal areas (Fig. 3E-H), and the latter two regions are labeled bilaterally, heavier ipsilaterally. Finally, there is dense labeling seen in all areas of rostral visual cortex, organized as clusters of cells located mostly in layers IIiIII, V and VII (Fig. 3F-H). Light labeling is also seen in retrosplenial cortex ipsilaterally (Fig. 3E-H). As noted above, in isocortical regions exhibiting labeled cells we invariably find a thin stratum of labeled cells adjacent to the underlying white matter. This population of deep cells is often labeled in the absence of significant laheling in layer VI proper. This was one reason we designated this deep cellular zone as layer VII (Reep and Goodwin, ’88). Through applications of fluorescent tracers in a variety of regions, Divac et al. (’87) found that this was a general property of isocortex. Furthermore, they discovered that most of these deeply situated cells participate in local circuits with the overlying layer I. Altogether, these results imply that layer VII cells engage in local and long distance corticocortical connections. However, it is not known if single layer VII cells do both, or if instead there are subpopulations of local and projecting neurons within layer VII. Case AGZ 63. This brain also has an injection site centered at the level of the forceps minor, in area AG1 (Fig. 4B). Within orbital cortex a few cells are present in the lateral orhital (LO) area (Fig. 4A). Contralaterally, AGI is extensively labeled as is a portion of somatic sensory area Par I (Fig. 4A E). Ipsilaterally, just caudal to the injection site, labeling is quite heavy in area Par I and is organized into sheets and clusters (Figs. 4C-E, 7). Only some of these labeled cells are found in dysgranular zones of Par I, a pat-

R.L.REEPET AL.

Pa r

Po r

Pa I

Figure 4

CORTICAL CONNECTIONS OF AGM tern noted above for case AGm 29. The heaviest concentrations of labeled cells are located in layers II/III, V, and VII, with noticeably lighter labeling in layer VI and a virtual absence of labeled cells in layer IV. More caudally, a t the level of the fimhria, labeling appears in area Par I1 as it diminishes in Par I (Fig. 4G-H). There is no cortical labeling seen caudal to the level at which the dentate gyrus first appears. Case AC 48. Here the injection site is located in the anterior cingulate cortex medial to AGm, a t the level of the anterior commissure (Fig. 5C). Cortical connections are greatly reduced in comparison to cases with injections in AGrn or AG1. Labeling is seen in contralateral AC (Fig. 5BD), the claustrum bilaterally (Fig. 5B-D), the ipsilateral retrosplenial and visual cortices (Fig. 5E-H). Labeled cells are concentrated in layers 11,V, and VI in both the granular and agranular retrosplenial areas. Within visual cortex there is almost no labeling in the granular area Ocl, and areas Oc2M and Oc2L contain moderate numbers of labeled cells (Fig. 5G-H). The results of these three cases suggest the following overall pattern of organization regarding the connections of AGm, AGI, and AC with the somatic sensorimotor, visual, retrosplenial, and orbital cortices. First, while AGrn has connections with all four of these regions, AG1 and AC do not. AG1 has major connections with somatic sensorimotor cortex and sparse connections with orbital cortex. AC has no somatic sensorimotor or orbital connections but does have extensive ties with visual and retrosplenial cortex. Thus, the intermediately placed AGm shares attributes with the adjacent areas AGI and AC, and in addition has extensive bilateral connections with orbital cortex.

Rostrocaudal topography As described above, AGm has widespread connections with somatic sensorimotor and visual cortices, and less extensive connections with several other areas. Presented below are results on topographic organization of the corticocortical connections of AGrn and AG1 with somatic sensorimotor cortex, and those of AGrn with visual cortex. Somatic sensorimotor cortex-rostrocaudal topographg within AGm. Rostrally located injections, such as those in brains AGm 29, 41, and 77, result in extensive cell labeling in ipsilateral somatic sensorimotor cortex. Tn case AGm 29 (Fig. 3), this labeling is prominent in Par I and Par I1 caudal to the level of the fimbria. In the caudal part of Par I the labeling becomes clustered and continues caudally into visual cortex (Fig. 3F). In brain AGrn 77 (Fig. 8) a similar pattern is seen. Somatic sensorimotor cortex labeling begins a t the level of the anterior commissure, with a few labeled cells present in the forelimb area (Fig. HR). By the level of the fimbria labeling is present throughout Par I and Par 11, and continues ventrally into insular cortex (Fig. 8C). However, the heaviest labeling in Par I and Par I1 is found caudal to the level of the fimbria (Fig. 8D,E). Proceeding caudally, labeling in Par 1 becomes focused in a dorsolateral region, and is continuous Fig. 4 (A-H). Cortical afferents to lateral agranular cortex. Spaced coronal sections illustrating the injection site and distribution of labeled cells in case AG163. Somatic sensorimotor cortex label is focused in rostral Par I. Note bilateral labeling of Par I in sections D-E. In contrast to case AGm 29, there is no labeling in the orbital or visual cortices. No cortical labeling was seen in areas located caudal to the level of section H.

269 from here into visual cortex (Figs. 8E, 11A). Throughout Par I, labeled cells are found in both granular and dysgranular zones (Fig. 8B-E). Cases AGm 41, AGm/AC 44 (Fig. 1) and AGm 73 exhibit fewer labeled cells in somatic sensorimotor cortex. Apparently this is due to more diffuse labeling a t the injection sites, which are located in AGm at the forceps-genu level as in cases AGm 29 and 77 above. As in these previously described cases, labeling starts as laterally spreading cells just caudal to the level of the anterior commissure, hut becomes densest caudal to the level of the fimbria. In these cases labeled cells are clustered or loosely scattered in Par I, rather than being organized into sheets. A prominent cluster of cells present in Par I1 appears continuous with a group of labeled cells which extends into insular cortex. As in cases AGm 29 and 77, there is bilateral labeling in the perirhinal/ entorhinal region. In brains with injections placed in the mid-portion of AGm, at the level of the anterior commissure, there is less labeling in somatic sensorimotor cortex than with rostral AGm injections. In cases AGm 30 (Fig. 8), 40, AGm/AC 42 (Fig. l),and AGm 74, labeling in this region begins as sheets of cells in layers V and VII, emanating laterally from the location of the injection site (Fig. 8B; see also Fig. 2B). Caudal to the level of the fimbria labeling is focused in two regions: one somewhat lateral to the cingulum bundle in the hindlimb area (HL) and another in Par 11. Cell labeling is present in, but not confined to, dysgranular zones (Fig. 8BD). With caudal AGm injections, at the level of the fimbria or fimbria-dentate, labeling in somatic sensorimotor cortex is even further reduced. There is no Par I label rostral to the level of the fimbria. This is shown for case AGm 43 (Fig. 8AB) and is also true for cases AGrn 31 and AGm/AGI 33 (Fig. 1).A t the level of the injection site, sheets of labeled cells spread for a short distance into the hindlimb region (Fig. 8C,D; see also Fig. 2B) and Par I, but there is much less extensive lateral labeling than that seen with injections in mid AGm or rostral AGm. Caudal to the level of the fimbria labeling becomes focused in the hindlimb region just lateral to the cingulum bundle (Fig. 8D-E). Labeling in this position continues into visual cortex. Cases AGm 31, AGm/AGI 33, and AGm 43 all have labeling in caudal but not rostral Par I1 (Fig. HD,E). The above observations demonstrate that there is a relationship between the pattern of labeling in somatic sensorimotor cortex and the rostrocaudal location of the injection site in AGm. In cases with injections in rostral AGm, labeling is quite extensive in Par I caudal to the level of the fimbria. In contrast, caudal AGm injections result in label that is largely confined to the hindlimb cortex. With respect to Par 11, rostral AGm injections result in labeled cells throughout Par TI, while caudal injections produce label confined to caudal Par 11. In general, injections in mid-AGm yield results which represent an intermediate condition. Somatic sensorimotor cortex-rostrocaudal topographg within AGI. Brains AGI 63 and 64 (Figs. 1, 4, 9) represent cases with injections in rostral AGl, at the level of the forceps minor. In both cases labeled cells spread medially and laterally from the injection site, and continue caudally in intermittent sheets and clusters throughout the lateral part of rostral Par I (Figs. 7C, 9A-C). Caudal to the level of the fimbria, labeling becomes sparse in Par I but is prominent in Par 11and continues ventrally into the insular region (Fig. 9D,E).

R.L.REEPET AL.

270

Figure 5

CORTICAL CONNECTIONS OF AGM

Cases AGl78 (Fig. 1)and 86 have injections located more caudally in AG1, at the level of the genu. Labeled cells spread laterally from the injection site toward Par I (Fig. 9A1. Within Par I labeling is densest medially in rostral Par 1(Figs. 7B, 9B-D). In Par TI labeling appears caudally, and a deeply situated sheet of labeled cells continues into the insular region (Fig. 9D). At this level labeled cells are also found in caudal AGm bilaterally (see also Fig. 6A). Case AG1 86 showed essentially the same patterns, having densest Par I labeling rostral to the level of the fimbria. In case AGm/ AGl 84 (Fig. 1) the injection site involves both AGm and AG1, and Par I labeling is dense rostral and caudal to the level of the fimbria. Cases AG1 49 (Figs. 1, 9) and 67 represent brains with injections in caudal AGI, just rostral to the level of the anterior commissure, and apparently encroaching on the forelimb sensorimotor overlap area (Sanderson e t al., '84; Neafsey et al., '86). In both cases labeled cells extend medially and laterally from the injection site (Fig. 9A). In Par I the heaviest labeling is located rostrally (Fig. 9B,C; see also Fig. 7A). Substantial labeling is also found throughout Par I1 and is continuous with that in the insular area (Fig. 9C-E). Finally, sparse to moderate bilateral label is present in caudal AGrn (Fig. 9R-D). The results described above show that for AG1, as with AGm, there is a relationship between the rostrocaudal location of the injection site and the pattern of labeling in somatic sensorimotor cortex. Rostra1 AG1 injections produce labeling focused in the lateral two-thirds of Par I rostral to the level of the fimbria. With more caudally placed AGI injections, labeling is densest in medial Par I and the forelimb area. In every case, labeling in Par I is relatively sparse caudal to the level of the fimbria. All brains with AG1 injections have labeling in Par I1 and there appears to be no topographical relationship to injection site location. Comparing the AGm and AGl groups as a whole, a consistent difference is apparent in the topography of Par I labeling. AGrn injections produce Par I labeling that is densest caudal to the level of the fimbria, while AGl injections result in Par 1 label which is densest rostral to this level. This is further illustrated by the double injection experiment of case AGm 98 (Fig. 10). A striking mediolateral relationship exists with respect to Par I labeling, when one considers that injections made into AC result in no labeling in somatic sensorimotor cortex (Fig. 5). In these cases (AC 47, 48) labeling is largely confined to the retrosplenial and visual cortices. In brain AC 45 the injection was centered in AC with slight encroachment on rostral AGm. This resulted in a few scattered labeled cells in Par I a t the level of the dentate gyrus. Thus the medially located area AC receives no afferents from Par I, centrally located AGrn receives input principally from caudal Par I, and the laterally located AGl is connected primarily with rostral Par I. Visual cortex-rostrocaudal topography within AGm. In cases with rostral injection sites, such as AGm 77, labeled cells in Par I become concentrated into a single broad cluster as the caudal boundary of Par I is approached,

271 and these are continuous medially with labeled cells in visual area Oc2 (Fig. 11A). More caudally Par I disappears and the labeling which was contained in it is now found in area Oc2L (Fig. 11B). Also, the granular area Ocl appears between areas Oc2L and Oc2M, separating the labeling in these two areas by a label-sparse zone (Fig. 11B-D). In these more caudal levels of visual cortex, area Oc2L contains denser label than Oc2M, but the overall number of labeled cells is decreased from rostral visual cortex. This caudal drop in density is further illustrated by case AGm 98 (Fig. 10). Here moderate labeling in rostral visual cortex (Fig. 10F,G) becomes very sparse a t middle levels (Fig. 10H) and no label is present caudally. In case AGm 41 labeled cells are scattered throughout areas Oc2L, OcZM, and Ocl in the rostral half of visual cortex. More caudally they are focused in area Oc2L. In brain AGm 29, which represents a large injection affecting rostral and mid AGm, all areas of visual cortex are labeled, but labeling is densest in Oc2L (Fig. 3). In case AGm/AGI 84 (Fig. 1 ) the injection site affects the lateral half of rostral AGrn (Fig. 1) and labeling in visual cortex is confined to the rostral part of area Oc2. With mid AGrn injections such as that in case AGm 30, somatic sensorimotor labeling is found in the hindlimb area and in medial Par I (Fig. 8). Visual cortex labeling is first seen immediately caudal to this region, and is rather extensive (Fig. 8E). Proceeding caudally in case AGm 30, labeling becomes most concentrated in area Ocl as it appears between areas Oc2M and Oc2L (Fig. 11A). Labeling is rather evenly distributed among all three areas throughout the remainder of visual cortex and appears denser than in the rostral case AGrn 77, in spite of the fact that the injection site was substantially denser in case AGm 77 (Pig. 1).In case AGrn 40 (representing another injection site in mid AGm) labeling in caudal Par I is more diffuse than in case AGm 30. Visual cortex label first appears in the lateral part of area Oc2. As area Ocl becomes visible, labeling is fairly evenly distributed throughout areas Oc2L, Ocl and Oc2M. More caudally labeling is restricted to area Oc2L. The overall density of visual cortex labeling appears greater in this case than in rostral case AGm 41, though both injection sites are comparable in size and density. Caudal injections, represented by case AGm 43, produce densest somatic sensorimotor labeling in the hindlimb region (Fig. 8). This is continuous with robust labeling in area Oc2 of rostral visual cortex (Fig. 11A). As area Ocl appears, labeled cells are found in it and in areas Oc2L and Oc2M (Fig. 1lB). Labeling continues into the caudal portions of these areas (Fig. llC,U). In brain AGm/AGl 33 (Fig. 1) a very similar pattern is seen. The initially dense rostral labeling in area Oc2 is followed by uniform labeling throughout areas Oc1, Oc2M, and Oc2L, persisting caudally. T h e results above demonstrate that all portions of AGm receive afferents from visual cortex. However, injections of rostral AGrn produce labeling which tends to be focused in the rostral half of visual cortex, favors area OcZL, and minimally involves the granular area Ocl . In contrast, injections of mid or caudal AGm produce denser labeling which is relatively evenly distributed throughout the entire rostrocaudal domain of areas Ocl, Oc2L, and Oc2M.

Orbital cortex Fig. 5 (A-H). Cortical d e r e n t s to anterior cingulate cortex. Spaced coronal sections illustrating the injection site and distribution of labeled cells in case AC 48. Cortical labeling is sharply reduced compared to cases AGrn 29 and AGI 63, and is focused in visual and retrosplenial cortices (F-H).

The full rostrocaudal extent of AGrn receives bilateral afferents from medial and ventrolateral orbital cortex (MO and VLO), and the rostrocaudal location of the injection site

272

R.L. REEP ET AL.

Fig. 6. Photomicrographs of labeled cells in cerebral cortex after injections in AGm or AGI. A A cluster of labeled cells in caudal AGm after an injection of FG into AGl. Arrowheads denote mediolateral extent of AGm. Case AG178; corresponds to Figure 9D. B FG-labeled

cells in AGm contralateral to the injection site, case AGm 77. C: Orbital cortex labeling ipsilateral tn the injection site in case AGm 29. Correspnnds to Figure 3A.

in AGm has no obvious effect on the pattern of retrograde orbital labeling. As can be seen in Figures 3A and 6C, labeled cells are concentrated in two bands: one in layers 11/ 111and another in layer V. Superficial labeling is often more prominent rostrally, and the deeper band merges with claustral labeling caudally (Fig. 3B). Generally, but not always, ipsilateral labeling is heavier than contralateral. Although most brains exhibited the above pattern, some variations were seen. For instance, in a few cases contralateral labeling was very reduced, or limited to only the superficial or deep band. In the mid case AGm 73 there was no MO labeling; in the caudal case AGm 31 there was very little orbital labeling of any kind. When injections are placed far medially, on the border of AGm and AC (cases AGm/AC 45 and 46), there is a sharp reduction in orbital cortex labeling. In both cases the spatial

distribution described above is maintained, but there are fewer labeled cells and these are only lightly filled with Fluoro-Gold. In cases AC 47 and 48 the injections are centered in AC, and there is no orbital labeling (Fig. 5). Injections made in rostra1 AGI (cases AGI 63 and 64, and AGm 98) resulted in a few labeled cells in orbital cortex (Figs. 4A, lOA,B). Injections made in mid AG1 (cases AGI 78 and 86) produced very light labeling bilaterally in VLO. Caudally located AG1 injections such as cases AG149 and 67 were associated with scattered light label in MO and VLO bilaterally. Deep orbital labeling became continuous with claustral labeling caudally. It is clear that AGm receives a much more extensive afferent innervation from orbital cortex than does its lateral neighbor AGl; the medially adjacent AC has no orbital connections.

CORTICAL CONNECTIONS OF AGM

273

Fig. 7. Varieties of labeled cell clusters in area Par I of somatic sensorimotor cortex. A Laminar-specific labeling extends on either side of a central vertically oriented cluster. Asterisk marks the cell sparse zone between layers VI and VII, wm denotes white matter; case AC1 49. B:

Two adjacent clusters, principally involving layers In and V; case AGI 78. C: A cluster illustrating primary involvement of layers II/III, V, and VII, as well as lateral extension of the labeling in layer V;case AG164.

Retrosplenial and rhinal cortex

extensive throughout RSa and RSg. In case AGm 40 there was labeling in RSa and RSg but it was confined to rostral RS. Ethinal cortex labeling was seen in all AGm and AG1 cases, and the general pattern is illustrated in Figures 3,4,8,9,and 11. Typically, labeled cells in Par I1 are continuous ventrally with labeling in the posterior agranular insular region along the rhinal fissure (Fig. 3E). These ventrally located cells persist caudally into the perirhinal and entorhinal regions (Fig. 3F-H). In addition, perirhinal and entorhinal labeling is seen contralaterally, though it is usually less dense than that seen ipsilaterally. AG1 brains had less dense overall labeling in these regions than did AGm brains. In several AGm brains (cases 18, 29, 31, 41, 42, 77, 87) labeled cells in Par I1 are continuous caudally with labeling in auditory association cortex, area TE. In these cases labeled cells extend ventrally from rostral TE into perirhinal and entorhinal cortex (Figs. 3F, 11A).

Injections in AGm and AC result in retrosplenial (RS) cortex labeling, while injections in AGl do not (Figs. 3-5). The projection from RS to AGm is topographically organized such that injections in caudal AGm produce much more extensive labeling throughout RS than do injections in rostral AGm. As shown for cases AGm 29 (Fig. 3) and 77 (Figs. 8,11), rostral AGm receives aRerents from the rostral portion of RS, principally from the agranular subregion RSa. Case AGm 41 showed a similar labeling pattern. Caudal AGm injections produce labeling in RSa and the granular subregion RSg, throughout the rostrocaudal extent of RS, as shown in Figures 8 and 11 for case AGm 43. Similar results were found in cases AGm 33, AGm/AGI 33, and AGm/AC 42 (Fig. 1). With injections in mid AGm variable patterns were seen. In case AGm 30 (Figs. 8 , l l ) labeling was

274

R.L.RE:EPETAL.

m

CORTICAL CONNECTIONS OF AGM

DISCUSSION This study has investigated the topographic organization of the corticocortical connections of AGm, the basic patterns of which were defined previously by using horseradish peroxidase and autoradiographic techniques (Reep et al., '84, '87). Here we have demonstrated rostrocaudal topography within AGrn with respect to its afferent connections from somatic sensorimotor, visual and retrosplenial cortices. In the mediolateral domain, when the pattern of corticocortical connections of ACm is compared with those of the laterally adjacent area AG1 and the medially adjacent AC, it is clear that AGm is distinct from both these regions.

AGm and somatic sensorimotor cortex Our present findings indicate two topographic trends with respect to connections involving somatic sensorimotor cortex. These patterns can be interpreted from a functional perspective by reference to the known organization of the body map in Par I, and it appears that the topography which we observe is correlated with functional parcellation within area Par I. The first trend we noted was that rostral AGm has more extensive connections with somatic sensory areas Par I and Par I1 than does caudal AGm. As discussed below, caudal AGm is more heavily connected with visual cortex than ie rostral AGm. Hence there appears to be topographic separation in AGm with respect to its connections with the cortical representations of the somatic sensory and visual modalities. Rostra1 AGm also contains a small secondary motor area for the hindlimb (Neafsey et al., '86) and is a source of corticospinal projections (Miller, '87; Neafsey and Sievert, '82). A small secondary forelimb area lies more laterally, in AG1 (Neafsey and Sievert, '82; Sanderson et al., '84). Together, these two secondary fields constitute a "rostral motor area," which some workers have suggested may be a supplementary motor area (Miller, '87; Neafsey et al., '86; Sanderson et al., '84). As Miller ('87) points out, primate corticospinal neurons are found in supplementary motor areas but not in the frontal eye fields. Similarly, he found that in rats corticospinal projections of AGm originate only from rostral AGm. While these considerations may argue in favor of rostral AGrn as part of a supplementary motor area, such a role for rostral AGm appears to be superimposed on the frontal eye field function characteristic of AGm as a whole (see below). The second topographic trend we noted was that AGm receives somatic sensorimotor afferents primarily from caudal Par I (caudal to the level of the fimbria) and the hindlimb cortex, while AG1 does so from rostral Par I and the forelimb cortex. Caudal Par I contains representations for the mystacial vibrissae and trunk while more rostrally in Par I are contained fields for the nose and small rostral

Fig. 8. Rostrocaudal topography in the connections between AGm and somatic sensorimotor cortex. For representative injections in rostral, mid and caudal AGm, spaced coronal sections (A-E) illustrate the distribution of labeled cells in ipsilateral somatic sensorimotor and retrosplenial cortex. Note that the rostral injection of case AGrn 77 resulted in widespread labeling of caudal somatic sensorimotor cortex, whereas there was reduced laheling with more caudally placed injections. Portions of the injection sites are shown in each case. Centers of the injection sites are illustrated in Figure 1.

275

vibrissae (Chapin and Lin, '84; Welker, '71; Welker and Woolsey, '74). Agl receives little input from the most rostral part of Par I, which contains representations for the perioral region. Connectivity between AGm and the mystacial vibrissae representation of Par I is consistent with the demonstrated role of AGm in vibrissae motor function (Gioanni and Lamarche, '85; Hall and Lindholm, '74; Neafsey et al., '86; Sanderson et al., '84). Afferents to AGm and AGl arise from granular and dysgranular zones of ipsilateral Par I. We have previously reported that the reciprocal efferents from AGm and AG1 to Par I also involve both zones (Reep e t al., '87). In contrast, homotopical commissural connections of Par I originate from and terminate in the dysgranular zones (Akers and Killackey, '78). Involvement of the granular zones may represent a fundamental difference in the organization of associational (ipsilateral) versus commissural connections involving Par I. The granular zones of Par I are sites of thalamocortical terminations. Thus associational projections arising from portions of Par I having a well defined granular layer 1V may be more directly involved with thalamic activity than the associational and commissural projections which originate in dysgranular parts of Par I.

AGm as a frontal eye field Microstimulation studies have shown that AGm is closely related to motor cortex, and may represent a frontal eye field. In AG1, which is the primary motor area, low current intensities (2-60 FA) produce movements and a largely complete body map is represented. In AGm much higher stimulation thresholds (60-300 FA) are needed to initiate movements, and these selectively involve the vibrissae, eyes, head, upper lips, and rhinarium (Hall and Lindholm, '74; Donoghue and Wise, '82; Sanderson et al., '84; Sinnamon and Galer, '84; Gioanni and Lamarche, '85; Neafsey et al., '86). These movements are often produced simultaneously as a coordinated contralateral orienting response, suggesting that AGm may function similarly to the primate frontal eye field. A frontal eye field role f o r AGm was first proposed by Leonard ('fig), and more recently by others (Leichnetz and Gonzolo-Kuiz, '87; Leichnetz et al., '87; Reep e t al., '84, '87), on anatomical grounds. These studies have demonstrated that AGm has connections with the superior colliculus and oculomotor brainstem, as well as with visual cortex. Our present findings also indicate that AGrn has substantial connections with visual cortex, but that those of rostral AGm are less extensive than those of mid and caudal AGm. Injection sites confined to rostral AGm produce visual cortex labeling that tends to be located rostrally, favors area Oc2L, and does not involve area Ocl. In contrast, the connections of mid and caudal AGrn involve the whole rostrocaudal extent of visual cortex and there is uniform participation of areas Ocl, Oc2L, and Oc2M. Reciprocal connections between AGm and all regions of visual cortex were described previously by Vogt and Miller ('83) and Miller and Vogt ('84). However, they found that whereas mid and caudal AGm participated in such connections, rostral AGrn did not. Similarly, a recent study (Sukekawa, '88) reports visual cortex connections involving that portion of AGrn caudal to a point +2.0 mm from bregma. These findings, like ours, demonstrate that rostral AGm is less involved in visual cortex connections than are mid and cau-

276

R.L.REEPETAL.

277

CORTICAL CONNECTIONS OF AGM

AGm

A.

B.

Figure 10.

(See page 278 for fig. legend.)

R.L. R E E P ET AL.

278 dal AGm. The reciprocal efferents from AGm to visual cortex reflect a similar organization. Thus, case 83058 of Reep et al. ('87) represents an injection of 3H-amino acids into rostral AGm, and terminal field labeling is focused over rostral visual cortex. In case ER5 of Deacon e t al. ('83), a similar injection into mid AGm produced labeling throughout all of visual cortex. These anatomical findings suggest that mid and caudal AGm play a greater role in cortical visual functions than does rostral AGm. However, brainstem regions involved in head and eye movements, such as the superior colliculus and oculomotor complex, have connections predominantly with rostral AGm (Leichnetz and Gonzolo-Ruiz, '87; Leichnetz et al., '87). Likewise, in microstimulation experiments, head orientation responses are elicited with lowest threshold from rostral AGm (Sinnamon and Galer, '84). Therefore, while a frontal eye field function may be distributed throughout AGm, there are significant topographic components to the connectivity of AGm, suggesting that the rostral and caudal portions of AGrn work in concert to produce normal eye-head orienting responses. Contralateral orienting responses are disrupted when AGm is lesioned unilaterally, resulting in hemispatial neglect (Cowey and Bozek, '74; Crowne and Pathria, '82; Crowne et al., '83; Corwin et al., '86; Sinnamon and Charman, '88; Vargo et, al., '88). Although lesions of caudal AGm produce more severe neglect than do lesions of rostral AGrn (King and Corwin, '89), lesions of either part of AGm result in neglect of visual, somatic sensory, and auditory stimuli. These findings suggest that while cortical representations may be topographic within AGm, sensory modality relationships as a whole are not so simply organized.

The relationship of AGm to prefrontal and premotor cortex Originally AGm was considered to be part of prefrontal cortex because of its afferent connections with the mediodorsal thalamic nucleus (Krettek and Price, '77). In addition to the fact that defining prefrontal cortex in this way is problematic (Divac et al., '78; Markowitsch and Pritzel, '79; Reep, '84; Goldman-Rakic and Porrino, '85; Passingham et al., '88),more recent anatomical studies have demonstrated that the dominant thalamic connections of AGm are with the ventral lateral, intralaminar, and ventromedial nuclei (Reep et al., '84, '87). However, what most distinguishes AGm anatomically from other regions of the rodent medio-

Fig. 9. Rostrocaudal topography in the connections between AG1 and somatic sensorimotor cortex. For representative injections in rostral, mid, and caudal AG1, spaced coronal sections (A-E) illustrate the distribution of labeled cells in ipsilateral somatic sensorimotor cortex. Note that labeling is densest in rostral Par I, and is seen throughout Par 11. Portions of the injection sites are shown in each case. Centers of the injection sites are illustrated in Figure 1. Fig. 10 (A-H). Double injection experiment illustrating differences in the cortical connections of AGrn and AG1. Fast Blue was injected into AGm, and Diamidino Yellow into AG1. Small dots represent labeling from the AGrn injection; large dots denote label related t o AG1. Note that somatic sensorimotor cortex label related t o AGm is prominent from the level of the fimbria caudally, while that resulting from the AGl injection is located rostral to this level. Within visual cortex, labeling related to AGm is moderately dense rostrally and is not present caudal t o the level of section H.

dorsal nucleus projection field are its extensive corticocortical connections (Reep et al., '84, '87). The pattern of these connections and its comparison to similar patterns seen in primates helped shaped our conclusion that AGm is a multimodal association area. In particular, the corticocortical connections of AGm combine attributes of primate supplementary cortex, the arcuate premotor area, and arcuate area 8 (Reep et al., '87). I t has been suggested that AGm, or portions of it, has attributes of primate premotor cortex. Passingham et al. ('88) found that bilateral lesions of AGm (apparently its rostral portion, judging from their Fig. 2) produced deficits on relearning a visual conditional motor task but not a spatial delayed alternation task. They relate this dichotomy to monkeys, wherein the behavioral properties of premotor area 6 resemble those of AGm and are dissociable from those of dorsolateral prefrontal cortex. Similarly, in rats AGm is functionally dissociable from medial wall cortex, lesions of which do produce deficits in spatial delayed alternation. Again, these findings suggest that rostral AGm is a multimodal association area with properties different from the rest of the cortex projected upon by the mediodorsal thalamic nucleus. In the context of our findings of rostrocaudal topography. it would be of interest to know how caudal AGm fits into this behavioral scheme.

Mediolateral topography Our finding of a mediolateral topography, wherein the corticocortical connections of AGm are significantly different from those of the laterally adjacent AG1 and medially adjacent AC, is consistent with previous work. With respect to AGl, Donoghue and Parham ('83) described cortical afferents from the rost,ral portion of somatic sensorimotor area Par I (principally dysgranular zones) and from area Par 11, but none from visual or retrosplenial cortex. Orbital cortex was sparsely labeled. Reep et al. ('87) reported a similar pattern of cortical efferents from AG1. With respect to anterior cingulate cortex, connections with visual and retrosplenial cortex were reported previously by Beckstead ('79), Finch et al., ('84), and Vogt snd Miller ('83). Reciprocal connections with the medial orbital area are suggested by Figure 4 of Reckstead ('79) and Figure 4 of Finch et al. ('84). However, we have seen nothing comparable in our material, perhaps because our injection sites d o not involve the most rostral part of AC.

AGm-one

area or many?

As we have discussed above, AGm engages in widespread corticocortical connections which are topographically organized along the rostrocaudal dimension. We found that injections into mid AGm generally produced labeling patterns representing a mixture of attributes typical of cases with injections in rostral or caudal AGm. This suggests that the topographic organization in AGm is characterized by gradual transitions between connectivity patterns characteristic of rostral AGm and those of caudal AGm, rather than by sharp boundaries. Functionally, area AGm as a whole exhibits attributes of a frontal eye field, and in addition the rostral part of AGm has properties of supplementary motor and premotor cortices. In primates, the cortical regions having these functional characteristics are well separated. However, in rats these functions are represented in a partially overlapping fashion concentrated within AGm. Such a condition is not unique; several species exhibit over-

CORTICAL CONNECTIONS OF AGM

279

ROSTRAL

Fig. 11. Rostrocaudal topography in the connections between AGm and visual cortex. For representative injections in rostral, mid, and caudal AGm, spaced coronal sections (A-D) illustrate the distribution of labeled cells in ipsilateral visual and retrosplenial cortices. Note denser

labeling in visual cortex after mid and caudal injections, and lack of label in area Ocl in rostral case AGm 77. Labeling of retrosplenial cortex is prominent after mid and caudal injections, and absent in the rostral case.

lap in various components of the somatic sensory and motor representations, for example as seen in the hindlimb and forelimb fields in rats (Wise and Donoghue, '86). In our view the anatomical and functional differences between rostral and caudal AGm. includine those we reDorted here. reoresent partial separation of AGm into regions with distinct properties. In order to understand the anatomical basis of functional overlap in rostral AGm, it would be of great interest to know whether separate or overlapping populations of

neurons participate in the various functions in which rostral AGm is engaged.

I

ACKNOWLEDGMENTS

Y

The authors thank Mike Baccala, Margaret Booth, and Laura Line Reep for their technical and artistic contributions to this project. This work was supported by the University of Florida College of Veterinary Medicine and the

R.L. REEP ET AL.

280 Maxwell Fund, College of Veterinary Medicine, University of Florida Journal Series No. 229.

LITERATURE CITED Akers, R.M., and H.P. Killackey (1978) Organization of corticocortical connections in the parietal cortex of the rat. J. Comp. Neurol. 181:513-538. Beckstead, R.M. (1979) An autoradiographic examination of corticocortical and subcortical projections of the mediodorsal-projection (prefrontal) cortex in the rat. J. Comp. Neurol. 184:43-62. Chapin, J.K., and C-S. Lin (1984) Mapping the body representation in the SI cortex of anesthetized and awake rats. J. Comp. Neurol. 229r199-213. Corwin, J.V., S. Kanter, R.T. Watson, K.M. Heilman, E. Valenstein, and A. Hashimoto (1986) Apomorphine has a therapeutic effect on neglect produced by unilateral dorsomedial prefrontal cortex lesions in rats. Exp. Neurol. 94:683-698. Cowey, A,, and T. Bozek (1974) Contralateral 'neglect' after unilateral dorsomedial prefrontal lesions in rats. Brain Res. 725343. Crowne, D.P., and M.N. Pathria (1982) Some attentional effects of unilateral frontal lesions in the rat. Behav. Brain Res. 6:2&39. Crowne, D.P., C.M. Richardson, and G. Ward (1983) Brief deprivation of vision after unilateral lesions of the frontal eye field prevents contralatera1 inattention. Science 200:527- 530. Deacon, T.W., H. Eichenbaum, P. Rosenburg, and K.W. Eckmann (1983) Afferent connections of the perirhinal cortex in the rat. J. Comp. Neurol. 22u:i6aigo. Divac, I., A. Kosmal, A. Bjorklund. and 0. Lindvall(1978) Subcortical projections to the prefrontal cortex in the rat as revealed by the horseradish peroxidase technique. Neurosci. 3:785-796. Divac, I., S. Marinkovic, .J. Mogensen, W. Schwerdtfeger, and J. Regidor (1987) Vertical ascending connections in the isocortex. Anat. Embryol. 175r443455. Donoghue, J.P., and C. Parham (1983) Afferent connections of the lateral agranular field of the rat motor cortex. J. Comp. Neurol. 217:390404. Donoghue, J.P., and S.P. Wise (1982) The motor cortex of the rat: cytoarchitecture and microstimulation mapping. J. Comp. Neurol. 212:76-88. Donoghue, J.P., K.L. Kerman, and F.F. Ebner (1979) Evidence for two organizational plans within the somatic sensory-motor cortex of the rat. J. Comp. Neurol. 183:647464. Finch, D.M., E.L. Derian, and T.L. Babb (1984) A5erent fibers to rat cingulate cortex. Exp. Neurol. 83r46-85. Gioanni, Y., and M. Lamarche (1985) A reappraisal of rat motor cortex organization by intracortical microstimulation. Brain Res. 344:4941. Goldman-Rakic, P.S., and L.J. Porrino (1985) The primate mediodorsal (MD) nucleus and its projection to the frontal lobe. J. Comp. Neurol. 242:535-560. Hall, R.D., and E.P. Lindholm (1974) Organization of motor and somatosensory neocortex in the albino rat. Brain Res. 66:23-38. Heilman, K.M., R.T. Watson, and E. Valenstein (1985) Neglect and related disorders. In K.M. Heilman and E. Valenstein (eds): Clinical Neuropsychology. New York: Oxford University Press, pp. 243-294. King, V., and J.V. Convin (1989) Neglect following unilateral ablation of the caudal but not the rostral portion of medial agranular cortex of the rat and the therapeutic effect of apomorpbine. Behav. Brain Res. (In press). Krettek, J.E., and J.L. Price (1977) Cortical projections of the mediodorsal nucleus and adjacent thalamic nuclei in the rat. J. Comp. Neurol. I71:157-192. Leichnetz, G.R., and A. Gonzalo-Ruiz (1987) Collateralization of frontal eye field (medial precentraVanterior cingulate) neurons projecting to the paraoculomotor region, superior colliculus, and medial pontine reticular formation in the rat: a fluorescent double-labeling study. Exp. Brain Res. 68:355-364. Leichnetz, G.R., S.G.P. Hardy, and M K . Carruth (1987) Frontal projections to the region of the oculomotor complex in the r a t a retrograde and anterograde HRP study. J. Comp. Neurol. 263r387-399.

Leonard, C.M. (1969) The prefrontal cortex of the rat. I. Cortical projection of the mediodorsal nucleus. 11. Efferent connections. Brain Res. 12.321343. Markowitsch, H.J., and M. Pritzel (1979) The prefrontal cortex: projection area of the thalamic mediodorsal nucleus? Physiol. Psychol. 7;1-6. Miller, M.W. (1987) The origin of corticospinal projection neurons in rat. Exp. Brain Res. 67339-351. Miller, M.W., and B.A. Vogt (1984) Direct connections of rat visual cortex with sensory, motor, and association cortices. J. Comp. Neurol. 226:18& 202.

Neafsey, E.J., and C. Sievert (1982) A second forelimb motor area exists in rat frontal cortex. Brain Res. 232:151-156. Neafsey, E.J., E.L. Bold, G. Haas, K.M. Hurley-Guis, G. Quirk, C.F. Sievert, and R.R. Terreberry (1986) The organization of the rat motor cortex: a microstimulation mapping study. Brain Res. Rev. II:77-96. Passingham, R.E.. C. Myers, N. Rawlins, V. Lightfoot, and S. Fearn (1988) Premotor cortex in the rat. Behav. Neurosci. 102:lOl-109. Paxinos, G., and Watson, C. (1986) The Rat Brain in Stereotaxic Coordinates, 2nd edition, Sydney: Academic Press. Reep, R.L. (1984) Relationship between prefrontal and limbic cortex: a comparative anatomical review. Brain Behav. Evol. 2 5 - 8 0 . Reep, R.L., J.V. Corwin, A. Hashimoto, and R.T. Watson (1984) Afferent connections of medial precentral cortex in the rat. Neurosci. Lett. 44r247252. Reep, R.L., d.V. Corwin, A. Hashimoto, and R.T. Watson (1987) Efferent Cunnections of the rostral portion of medial agranular cortex in rats. Brain Res. Bull. 19:203 -221. Reep, R.L., and G.S. Goodwin (1988) Layer V11 of rodent cerebral cortex. Neurosci. Lett.. 9Or15-20. Sanderson, K.J., W. Welker, and G.M. Shambes (1984) Reevaluation of motor cortex and of sensorimotor overlap in cerebral cortex of albino rats. Brain Res. 292r251-260. Sinnamon, H.M., and C.S. Charman (1988) Unilateral and bilaterd lesions of the anteromcdial cortex increase perseverative head movements of the rat. Behav. Brain Res. 27:145-160. Sinnamon, H.M., and B.S. Galer (1984) Head movements elicit.ed by electrical stimulation of the anteromedial cortex of the rat. Physiol. Behav. 33:185-190. Sukekawa, K. (1988) Interconnections of the visual cortex with the frontal cortex in the rat. J. Hirnforsch. 29233-93. Vargo, J.M., J.V. Corwin, V. King, and R.L. Reep (1988) Hemispheric asymmetry in neglect produced by unilateral lesions of dorsomedial prefrontal cortex in rats. Exp. Neurol. 102:199-209. Vargo, J.M., M. Richard-Smith, and J.V. Corwin (1989) Spiroperidol reinstates asymmetries in neglect in rats recovered from left vs. right dorsomedial prefrontal cortex lesions. Behav. Neurosci. 103:178&1796. Vogt, R.A., and M.W. Miller (1983) Cortical connections between rat cingulate cortex and visual, motor and postsubicular cortices. J. Cornp. Neurol. 216:192-210. Welker, C. (1971) Microelectrode delineation of fine grain somatotopic organization of SMI cerebral neocortex in albino rat. Brain Res. 26:259-275. Welker, C. (1976) Receptive fields of barrels in the somatosensory neocortex of the rat. J. Comp. Neurol. 166:173-190. Welker, C., and M.M. Sinha (1972) Somatotopic organization of SmII cerebral neocortex in albino rat. Brain Res. 37:132-136. Welker, C., and T.A. Woolsey (1974) Structure of layer IV in the somatosensory neocortex of the r a t description and comparison with the mouse. J. Comp. Neurol. 158:437454. Wise, S.P.,and J.P. Donoghue (1986) Motor cortex of rodents. In E.G. Jones and A. Peters (eds): Cerebral Cortex Vol. 5, Sensory-motor Areas and Aspects of Cortical Connectivity. New York Plenum Press, pp. 243-270. Zilles, K., and A. Wree (1985) The Rat Nervous System. Sydney: Academic Press, pp. 375415.