topography of pyramidal neuron intrinsic connections in macaque monkey prefrontal cortex (areas 9...

THE JOURNAL OF COMPARATIVE NEUROLOGY 338~360-376 (1993)

Topography of Pyramidal Neuron Intrinsic Connections in Macaque Monkey Prefrontal Cortex (Areas 9 and 46)

JONATHAN B. LEVITT, DAVID A. LEWIS, TAKASHI YOSHIOKA, AND JENNIFER S. LUND

Departments of Neurobiology, Anatomy, and Cell Science (J.B.L., T.Y., J.S.L.), Psychiatry (D.A.L., T.Y., J.S.L.), and Behavioral Neuroscience (D.A.L.), University of Pittsburgh School

of Medicine, Pittsburgh, Pennsylvania 15261; Department of Visual Science, Institute of Ophthalmology, University of London, London EClV 9EL, England (J.B.L., J.S.L.)

ABSTRACT An understanding of the normal organization of prefrontal cortex is essential to the

recognition of pathology underlying human behavioral disorders believed to depend on this region. We have therefore studied the pattern of intrinsic intra- and interlaminar pyramidal neuron connectivity in prefrontal areas 9 and 46 (of Walker) in macaque monkey cerebral cortex (anterior to the arcuate sulcus between the principal sulcus and midline). We made focal (200-400 pm) injections of biocytin and mapped the pattern of orthogradely transported label. Injections made into the superficial layers label wide-ranging lateral projections within the same areas of prefrontal cortex. Projections local to such small injections form a narrow band of terminals in layers 1-3 (200-400 pm wide, 2-4 mm long) centered on the injection site. Collateral fibers spread orthogonal to this terminal band, making frequent bifurcations, to establish a series of parallel bands of terminals with uninnervated bands between, spaced regularly across the cortex (center to center 500-600 pm). The entire pattern of terminal label is stripelike, with occasional narrower interbands and crosslinks between the bands, and can extend over 7-8 mm across the cortex. These projections arise from pyramidal neurons in layers 2,3, and 5 and terminate in layers 1-3. The stripelike pattern contrasts with patchlike patterns in other cortical regions (Vl, V2, V4, motor, somatosensory) and is smaller in scale than stripelike zones of corticocortical afferent terminals to this region, reported to be 300-750 pm wide and spaced 1.0-1.5 mm center to center. B 1993 Wiley-Liss, Inc.

Key words: primates, biocytin, cerebral cortex, neuroanatomy, circuitry

Pyramidal cells, the major class of cortical excitatory cells, furnish the vast majority of extrinsic projections from the cortex (see review of Feldman, '84). The main axonal trunk of these neurons also gives rise to recurrent collaterals within a cortical region, many of which form synapses with the dendritic spines of other pyramidal neurons (LeVay, '88; White, '89; McGuire et al., '91). Thus, these collaterals may serve as the major propagator of intrinsic excitatory activity within the cortex. In visual, auditory, somatosensory, and motor cortices, the axon collaterals of pyramidal neurons in the superficial layers provide horizontally oriented, intracortical projections, which may spread for extensive distances within a region, or even into adjacent areas (Rockland and Lund, '83; Livingstone and Hubel, '84; DeFelipe et al., '86; Juliano et al., '90; Ojima et al., '91). These axon collaterals give rise to discrete columnar clusters of terminal boutons, which are most prominent in the superficial layers.

Although these patterns of intracortical excitatory connections appear to be a characteristic feature of sensory and motor regions of cortex, it is not known whether the same organizational scheme appears in other types of cortical areas, such as the higher order association regions of the prefrontal cortex. Regional differences in other aspects of cortical organization suggest that the pattern formed by local excitatory collaterals in prefrontal cortex might differ from that seen in primary sensory or motor cortex. In primary sensory and motor areas there is a clear functional topography to the regions, and different functional at-

Accepted June 25,1993. Takashi Yoshioka's present address is Krieger MindiBrain Institute,

Johns Hopkins University, Baltimore, MD 21218. Address reprint requests to Jonathan B. Levitt, PhD, Dept. of Visual

Science, Institute of Ophthalmology, University of London, 11-43 Bath Street, London EClV 9EL, England.

O 1993 WILEY-LISS. INC.

INTRINSIC CONNECTIVITY IN PREFRONTAL CORTEX 361

tributes are also interdigitated and geometrically arranged across the region. Recent work has suggested that the lateral punctate connectivity largely links regions of similar function. For example, in primary visual cortex, these clustered horizontal connections have been proposed to link columns of cells with similar orientation preferences (Ts'o et al., '86; Gilbert and Wiesel, '89). Similarly, in primary auditory cortex, they are thought to interconnect groups of neurons that represent similar characteristic tone frequen- cies (Matsubara and Phillips, '88; Wallace et al., '91). Physiological studies of prefrontal cortex have not identified a clear functional topography at the fine scale seen in the primary sensory or motor regions, so it is unclear for this region what scale or pattern might be expected from any system of intrinsic connections. Furthermore, in contrast to primary sensory cortex, which is dominated by excitatory inputs from the thalamus, the prefrontal cortex receives highly processed information from a large number of cortical regions (see Goldman-Rakic, '87; Barbas, '92; for reviews), which appears to be as prominent as input from the dorsomedial thalamic nucleus. However, components of prefrontal cortical circuitry do appear to have distinctive patterns of organization. For example, in macaque prefrontal cortex, associational and callosal afferent axons appear to terminate in an interdigitated stripelike fashion (Gold- man-Rakic and Schwartz, '821, and the thalamic projections to the prefrontal cortex are distributed in a discontinuous, columnar fashion (Giguere and Goldman-Rakic, '88). In addition, the axonal arbors of different classes of prefrontal local circuit neurons appear specialized to exert inhibitory control within distinct spatial domains (Lund and Lewis, '93).

In the present study, we used microinjections of the anterograde tracer biocytin to examine the three-dimensional geometry, and laminar source, of intrinsic axonal projections in Walker's ('40) areas 9 and 46 of macaque monkey prefrontal cortex. Our findings demonstrate that extensive horizontal projections are furnished by superficial layer populations of prefrontal cortical pyramidal neurons; the terminal fields of these collaterals are arranged in distinctive stripelike arrays that form a lattice structure different from that present in other cortical regions examined. The unique geometry of these connections may provide insight into how particular populations of prefrontal afferent axons and efferent neurons are linked to mediate the specialized functional properties of the region,

MATERIALS AND METHODS Surgical procedures

Seven adult cynomolgus (Macaca fascicularis) monkeys (3-4 kg) were used in this study. Some of these animals were also used in other tracing experiments unrelated to the present study. Animals were initially premedicated with atropine (0.05 mg/kg, i.m.), and tranquilized in their home cage with ketamine (10-15 mgikg, i.m.). In most cases, a catheter was inserted into either the saphenous or radial veins, and animals were anesthetized with an intravenous infusion of pentobarbital (Nembutal; 1-2 mgikglhr) in lactated Ringer's solution (5 ccihr); in one animal, an endotracheal tube was inserted, and anesthesia was maintained with halothane (0.5-1.5%). Animals were then placed in a stereotaxic apparatus. Local anesthetic (2% lidocaine) was applied topically at wound margins, and dexametha-

sone (0.5 mg/kg, i.m.1 was also administered to prevent cortical edema. The animals' temperature, respiration, and EKG were continuously monitored, and additional anesthetic was given as needed to abolish any signs of distress or arousal. On completion of injections, animals were given antibiotic (Ditrim; 0.11 cc/kg, i.m.1, and were then either maintained under anesthesia, or were allowed to recover and given analgesic (Stadol; 0.05 ccikg); animals survived 2-24 hours after biocytin injections and were closely monitored during this time. All surgical procedures were per- formed under aseptic conditions, and treatment of animals was in accordance with NIH guidelines.

Biocytin injections We made a small craniotomy in the vicinity of stereotaxic

coordinates A30 L10 (Szabo and Cowan, '84), then visualized the location of the arcuate and principal sulci. Our placement of injection sites was guided by the pattern of sulci; we aimed to place our injection sites into prefrontal areas 9 and 46, anterior to the arcuate sulcus, and between the principal sulcus and midline. Figure 1 is a schematic dorsal surface view of the brain, indicating the region of cortex in which we made injections; locations of all 39 injection sites are marked relative to local sulci. Most were on the convexity between the principal sulcus and midline (area 91, although some were intentionally placed more laterally, into the medial bank of the principal sulcus (area 46). The border between the areas is not a clear cytoarchitectural feature; instead, the transition is gradual. The border's approximate position is marked in Figure 1 by the dashed line.

We then made a small slit in the dura at each chosen location, and placed a glass micropipette (5-15 pm tip diameter) containing 4% biocytin (Sigma) dissolved in 0.9% saline at the chosen location and depth. At the appropriate site, we passed 6 pA of anodal pulsed current (7 seconds on, 7 seconds off) through the tip for 10-15 minutes to inject biocytin. This consistently yielded focal microinjections whose core diameters were 200-400 pm. We also made somewhat larger pressure injections by gluing a glass micropipette to the end of a microsyringe, and manually injecting 0.3-0.7 p1. We repeated this procedure several times in each hemisphere. Adjacent injections were separated by several millimeters to avoid overlapping label from different injections, except in cases where we were examin- ing interlaminar connectivity, in which case injections were spaced more closely, as we were less concerned with lateral overlap of label. In two other hemispheres (animal MK14, left; animal MK19, right), to label callosally projecting neurons, we made injections of the retrograde fluorescent tracer dye fast blue (FB) into prefrontal cortex. A 5 pl microsyringe containing 5% FB in distilled water was used to make a series of several (five to seven) closely spaced 300 nl injections, following which animals survived 10 days. Upon completion of surgery, the skull defect was covered with sterile Gelfoam, and the scalp was stitched closed.

Histology After the appropriate survival time, animals were deeply

anesthetized with Nembutal (50-75 mgikg, i.p.1 and per- fused transcardially with a saline rinse, followed by 3-4 liters of ice-cold (5OC) 4% paraformaldehyde in 0.1 M potassium phosphate buffer (KPB; pH 7.4) for 30 minutes, and finally 1-2 liters of 10% sucrose in 0.1 M KPB for 10 minutes. Brains were immediately removed from the skull,

362 J.B. LEVITT ET AL.

Rostra1

X MK13 W MK18 0 MK14 0 MK19

0 MK15

A MK16

A MK17

Caudal

Fig. 1. Summary of prefrontal injection sites. The cortex is viewed from above, rostra1 is to the top. Injections from each monkey (MK) are plotted (with different symbols) relative to the positions of the arcuate (AS) and principal (PSI sulci. On the insert, the dashed lines indicate the approximate location of the border between areas 9 and 46.

blocked, and placed in 30% sucrose in 0.1 M KPB in the cold (5°C) until they sank. They were then sectioned either coronally or tangential to the pial surface at 40-50 pm on a freezing microtome. Sections were processed for biocytin label and maintained in absolute serial order. Individual sections of interest were counterstained with thionin or cresyl violet to indicate laminar boundaries.

Our biocytin staining protocol is modified from those of Horikawa and Armstrong ('88) and King et al. ('88). Briefly, sections were incubated overnight in the cold in a solution of 0.7% Triton, 2% normal rabbit serum (NRS), and 0.02 M phosphate-buffered saline (KPBS). The tissue was then rinsed in 0.02 M KPBS, followed by a final rinse in 0.02 M KPBS containing 2% NRS; we then incubated sections in Elite-ABC solution (Vector Laboratories; 1: 130 dilution in 0.02 M KPBS) for 1-2 hours. Some sections were next soaked for 5-10 minutes in a 1% cobalt chloride solution in 0.02 M Tris buffer to intensify labeling (Adams, '77). Finally, label was visualized by peroxidase histochemistry with diaminobenzidine as the chromogen. Sections were then mounted on gelatin-coated slides, dehydrated, cleared, and coverslipped.

Reconstruction and analysis Sections were examined with the light microscope with

both bright- and darkfield illumination; details of axon terminal fields were generally better visualized in cobalt- intensified sections in brightfield. We examined sections at low power to draw the main section outline; it was generally necessary to examine tissue at high magnification because of the very fine diameter of fibers and terminals. Plots of labeled neurons and fibers were obtained by tracing the injections with the aid of a microscope drawing tube. Composite reconstructions in cortical depth were con- structed by superimposing drawings of adjacent sections, with blood vessels or other landmarks for alignment. Tangentially sectioned series were reconstructed from the pia to the white matter. Since many of the coronal sections

were not cut precisely along the true vertical axis between pia and white matter, serial reconstructions were also necessary in this case to determine the entire pattern of interlaminar (as well as lateral) connections local to the injection. The position of the midline was used as a means of aligning serial coronal sections, together with other features such as blood vessels at the surface or travelling obliquely through the tissue. Since any one section could pass through a terminal cluster at an oblique angle, or might not contain the entire cluster, thus obscuring its true dimensions, we measured terminal cluster width and spacing from complete serial reconstructions to prevent such distortions. Terminal stripe or cluster width was measured orthogonal to the long axis of the stripe, and interstripe or intercluster distances were measured center to center between nearest neighbor stripes or clusters.

RESULTS We made a total of 39 biocytin injections in seven

animals. Tissue from six injection sites was sectioned in the tangential plane to facilitate analysis of horizontal connectivity; the remaining 33 sites were sectioned coronally to examine interlaminar connectivity. The biocytin uptake gave very good orthograde label of axon processes, which appeared to derive almost exclusively from pyramidal neurons. However, very few pyramidal neurons were retrogradely labeled away from the injection axis; these lay in lower lamina 3 and could occur across any part of the region of orthograde terminal label.

Tangential organization of orthogradely labeled intrinsic connections

Figure 2 illustrates the typical orthograde projection pattern resulting from restricted (200-400 pm) biocytin injections into the upper layers of prefrontal cortex. These


Fig. 2. A,B: Photomicrographs of two biocytin injection sites and resulting transported label. These sections through layer 3 were cut tangential to the cortical surface. Note the dense fiber emanations from each injection site and the periodic clustering of terminal label. Arrows point to elongated bands of terminal label. Scale bar = 200 km.

two photomicrographs are of sections cut parallel to the cortical surface through layer 3; each panel is a different injection site. Radiating away from the injection core, which appears as a uniformly dark central region, is an extensive spread of fiber projections; these periodically ramify and end in terminal clusters. These terminal clusters were separated by terminal-free zones and could be found over 4 mm away from injection sites. Generally, these terminal clusters were elongated into stripes and therefore were unlike the punctate lateral connections seen in other cortical areas (Lund et al., '93). The arrows in Figure 2 point to two such elongated terminal zones.

We emphasize the distinction in this discussion between fibers of passage emanating from an injection and terminal clusters that may be distinguished by synaptic varicosities and the higher density of axonal fibers within them. Figure 3 shows the pattern of lateral projections resulting from another bioctyin injection. At low power, several terminal clusters can be seen surrounding the injection site (Fig. 3A).

Fig. 3. Another biocytin injection site viewed tangentially in layer 3 at low (A) and high (B) magnifications. Asterisks indicate the same blood vessel in both photomicrographs. Extensive projections from the injection core are again evident, as are the distinct regions of higher terminal density separated by terminal-free zones. One elongated terminal field is shown at higher magnification in B. Note that labeled fibers (arrow) continue to spread horizontally beyond this terminal field. Scale bars = 200 bm.

In this field, these clusters appear more circular and less elongated; this is due to the particular plane of section in this field not showing each terminal cluster in its entirety. The distinctness of these terminal zones is apparent in Figure 3B, which shows one terminal cluster at higher magnification. Note how the cluster, consisting of a higher density of axons, is surrounded by a zone relatively devoid of fibers and terminals.

Since any single tangential section tended not to include all terminal clusters, nor even any one cluster in its entirety, we based our analysis of tangential projections on reconstructions from serial sections. Figure 4 shows drawings of three such reconstructions of the complete orthograde labeling pattern and their position relative to the


A. MK15 Rostral

0 I I Midline I I I I I I I I I I I I I I I

B. MK15 /

/ /

/ /

/ /

// Principal Sulcus /

/

Caudal

Rostral

\ \\ Principal SUICUS

\ \

\ \ \ \

1 mm Caudal

/ C. MK16

/ /

Rostral

/

,/ Principal SUICUS /

/

/ /

/

)I,- * 1 mm

Caudal

Fig. 4. A-C: Reconstructions from serial tangential sections of the complete pattern of terminal label resulting from different prefrontal injection sites. Plane of view is again tangential to the cortical surface. In each case, terminal label is indicated by the solid dark areas, and the injection core is indicated by the asterisked zone in the center. Injection

midline and principal sulcus. The injection core is indicated by the asterisked zone in the center, and terminal label is indicated by the dark regions. In each case, the projections consist of narrow bands of terminals, between 200 and 400 pm wide and between 2 and 4 mm long. These bands are spaced fairly regularly across the cortex (center to center spacing roughly 500 pm). The entire pattern of terminal label is stripe-like, with occasional narrower interbands, and crosslinks between the bands, and the overall extent of

I I I I I I

Midline! I I I I I I I I

I I I

Midline I I

I I I I I I I I I I I

core sizes: A, 450 pm; B, 380 pm; C, 235 pm. The orientation of these labeling patterns is shown relative to local sulci. Note the tendency for stripe-like terminal zones to run roughly orthogonal to the principal sulcus.

label across the cortical surface can exceed 8 mm. Also note the general tendency for these elongated zones of terminal label to point roughly orthogonal to the principal sulcus, particularly as they approach the sulcus. This was most clearly seen in the reconstructions from tangential series (cf. Figs. 4 and 5). This pattern contrasts with more patch-like patterns observed in other cortical areas (see Lund et al., '931, although isolated patches of terminal label are apparent in Figures 4 and 5.

INTRINSIC CONNECTIVITY IN PREFRONTAL CORTEX

A Rostral

365

/' I / I

I I I I I I Midline I I

I I

/ /

/ /

/ Principal Sulcus //

/ /

/ /

/ /

Caudal 1 mm

B Rostral / I

/ I / I

I / /

Principal Sulcus // i Midline / I

I I I

/ /

/ /

Caudal 1 mm

Fig. 5. Two different reconstructions from serial coronal sections of the tangential pattern of terminal label from another injection site (injection core size roughly 200 km). In A, all terminal label is indicated by the solid dark regions, regardless of density. In B, only those zones of highest terminal density are black. Stippling marks lightest terminal label; diagonal hatching marks moderate terminal density.

In the reconstructions of Figure 4, we plotted all terminal label as uniformly dark zones, suggesting that all clusters of terminal label were equally dense. However, this was not strictly true; the density of terminal label tended to vary across the cortical tissue. In Figure 5 , we show a reconstruction of terminal label from another injection. In Figure 5A, all label is indicated by uniform shading; in Figure 5B, for the same injection, only the heaviest terminal label is marked by solid shading. Lighter terminal areas are marked by the stippling; the diagonal hatching to each side of the injection core marks regions of substantial but not heaviest termination. There were clear boundaries between each of the shaded regions. Although there are certainly differences in detail between the two versions, it is clear that the elongated stripelike character of terminal zones in this region of cortex persists, even if differences in terminal density are ignored.

To convey a sense of how the spread of label from any locus on the cortical surface relates to the overall area occupied by prefrontal cortex, we show in Figure 6 the serial reconstructions of Figures 4 and 5 again but here

superimposed onto drawings of the cortical surface. These four injections were all into area 9 on the convexity between the midline and principal sulcus, and this Figure 6 shows that any point in area 9 projects over a substantial portion of the area. The orientation of elongated terminal zones perpendicular to the principal sulcus is also apparent.

We made measurements of the dimensions and spacing of these intrinsic terminal bands to determine if these dimensions had any regularity and how they might relate to other connection systems of the prefrontal cortex. Histograms of these measurements are presented in Figure 7A,B. As described above, we measured stripe widths perpendicular to the long axis of the stripe, and spacing was measured center to center between nearest neighbors. The mean stripe width was 267 * 114 pm (+S.D., n = 72), and the mean stripe center to center spacing was 536 2 236 pm (*S.D., n = 45). These dimensions are smaller in scale than has been reported for those of stripelike corticocortical afferent terminals in prefrontal cortex (300-750 ym wide, spaced 1.0-1.5 mm center to center; Goldman-Rakic and Schwartz, '82). We were curious, however, whether these

366

Rost ral

J.B. LEVITT ET AL.

Caudal

Rostra1

Caudal

Fig. 6. The four injection sites of Figures 4 and 5 are shown superimposed on the brain to give an impression of the scale of label spread relative to the overall area occupied by prefrontal cortex. Note again the tendency for elongated terminal zones to point perpendicular to the principal sulcus.

dimensions might relate to the size of clusters of particular efferent cell populations. In two animals, we therefore made a series of fast blue injections (to label maximally callosally projecting cells) in the hemisphere contralateral to the hemisphere in which we made the biocytin injections, to determine if intrinsic terminals and callosally projecting cells might interdigitate, as do ipsilateral corticorticortical and callosal terminals (Goldman-Rakic and Schwartz, '82). Following fast blue injections, clusters of cells were retrogradely labeled in the contralateral prefrontal cortex. Unfor- tunately, laterally spreading biocytin labeling in these animals was poor in regions homotopic to the fast blue injections, so we were unable to compare directly the distribution of the two labels (although biocytin injections nonhomotopic to the fast blue injections were of good quality and could still serve in our analysis of interlaminar connectivity). Nonetheless, we were able to make measure-

ments of the size and spacing of these fast blue-labeled callosal efferent cell clusters in individual coronal sections to relate them indirectly to the scale of intrinsic biocytin terminal zones (Fig. 7C,D). The mean diameter of these dense fast blue cell clusters was 535 % 245 pm (*S.D., n = 531, and the mean size of the poorly populated gaps between them (not center to center spacing) was 261 * 51 pm (tS.D., n = 33). The mean center to center distance between terminal stripes of the biocytin labeled intrinsic connectivity system matched the mean width of callosal efferent neuron patches (cf. Fig. 7B and D), and the width of the individual intrinsic connectivity terminal stripes or clusters closely matched the size of gaps between callosally efferent cell clusters (cf. Fig. 7A and C). These dimensions are summarized in the diagram of Figure 7E, which shows that alignment between the two systems should repeat approximately every 1,600 pm.

INTRINSIC CONNECTIVITY IN PREFRONTAL CORTEX

1 , T

367

n n n

0 200 400 600 800 1000 1200

Intrinsic system terminal stripe width (pm)

n=33 0.40

0.00 o-loLd 0 2

mean = 261 prn C

Width of gaps between callosal cell clusters (pm)

0.25

n=45 B 0.20-

0 200 400

Center-to-center intrinsic stripe distance (pm)

0.30, n=53

Y

0.15

z 0.10 0.05

0.00

r

D [mean = 535 pm

n

0 200 400 600 800 1000 1200 Width of callosal cell clusters (pm) - Intrinsic

System Terminals

Callosal Efferent -1 A A A A A A A A 1 Neurons

IA A A A A A A A1 5351.1mLL 261um

I 1,592pm 1

!3 I

Fig. 7. Distributions of the widths (A) and center to center (B) stripe spacing of biocytin-labeled intrinsic terminal zones. C indicates the width of gaps between clusters of fast blue labeled callosal efferent cells. D indicates the width of clusters of retrogradely labeled callosally projecting neurons. E: Diagram summarizing details of hioctyin labeled

Interlaminar specificity of intrinsic connections

Having established the topography and dimensions of this lateral connection system, we next examined the specificity of interlaminar connections to determine which layers contribute to this lattice and what circuits exist between input and output layers of areas 9 and 46. The discontinuous nature of intrinsic lateral connectivity in prefrontal cortex was apparent even in coronal sections.

terminal stripe and Fast Blue callosal cell cluster measurements. Arrows in A-D indicate means used in the summary diagram (El. Note that these systems do not interdigitate, since the intrinsic connections are continuously repeating from any point on the cortex. See text for further discussion.

Figure 8 is a photomicrograph of two terminal clusters (arrows in Fig. 8A) located several mm away from a biocytin injection site in layers 213. At higher magnification (Fig. 8B), the high density of fibers and synaptic varicosities can be seen. These terminal clusters extend from the pia to the base of layer 3; this is in contrast to corticocortical terminal zones, which extend in a columnar fashion through the depth of the cortex (Goldman-Rakic and Schwartz, '821, or thalamic afferents, which terminate principally in layers 4


Fig. 8. Photomicrographs of biocytin-labeled terminal fields in the medial bank of the principal sulcus viewed in coronally sectioned material. Principal sulcus is to the top, and dorsal surface of the prefrontal cortex is to the left in this figure. A Arrows indicate two

terminal clusters, which are seen to extend from the pial surface to the base of layer 3. The cluster to the right (solid arrow) is shown at higher magnification in B; note the extensive beaded endings throughout the cluster. Scale bars = 200 pm.

and deep 3 (Giguere and Goldman-Rakic, '88). We deter- mined the complete pattern of interlaminar projections around each injection site by serial reconstruction of coronal sections, and Figure 9 summarizes the results of all of our injections placed at different cortical depths. This laminar analysis is based on a total of 33 microinjections, with two or more injection cases centered on each lamina, plus additional injections laying on borders between laminae. Representative photomicrographs are shown in Fig- ures 11-16. Sections reacted for biocytin were then counterstained for Nissl substance as necessary to allow cytoarchitectural identification of laminar boundaries (see Fig. 11).

Figure 9A summarizes the pattern of biocytin label observed following injections involving layer 2 and the uppermost portion of layer 3. Figure 10 shows that the laminar architecture of prefrontal cortex as seen in Nissl counterstained preparations (of the sections reacted for biocytinf does not provide a distinct boundary between layers 2 and 3. However, layer 2 may be recognized as a cell-dense region roughly the same thick- ness as layer 1. A representative biocytin injection is shown in Figure 11. Fibers spread laterally away from the injection in layers 2 and upper 3 (fewer were found in the deeper portion of layer 31, making periodic terminations in the upper layers. One such terminal zone is apparent in Figure

Znjections into layers 2 and upper 3.

11 to the left of the injection core. Horizontal fibers suddenly turn toward the pial surface (solid arrow in Fig. 11B) and emit rising collaterals with terminals. Below the injection core, a descending column of fibers was seen that extended well into the white matter (this column is not visible in Fig. 11 because the plane of section was not precisely normal to the cortical surface). Fibers were also seen to descend at a shallow angle through layer 3 (open arrow in Fig. 11B). Retrogradely labeled pyramidal cells were found in layers 3 and 5 under the injection core, some of which were presumably filled via their apical dendrites. A second laterally spreading fiber plexus was seen in layer 5, although this was substantially less robust than the lateral spread in the superficial layers. Occasional rising and descending collaterals were seen to emerge from the layer 5 plexus to travel through layers 4 and 6, but there was rather little terminal contribution to layers 4 or 6, and very few horizontally running fibers in these layers.

Figure 9B summarizes the pattern of label around injections placed into layer 3, and Figure 12 is a photomicrograph of a typical injection site. A column of labeled cells was found above and below the injection in layers 3-5; large pyramids in layer 5 were particularly prominent, perhaps filled via their apical dendrites. Very prominent bundles of fibers spread laterally within layer 3, turning upwards at intervals to provide

Injections into layer 3.


A

C

D

E

3

Fig. 9. Summary diagram of anterograde and retrogrede label obsewed following small (200 pm) biocytin injections into Werent cortical laminae in prefrontal areas 9 and 46. Drawings are based on serial reconstruction from coronal Sections. A: Injections into layer 2

and upper layer 3 (see ale0 Fig. 11). B: Injection into lower layer 3 (see also Figs. 12 and 13). C Iqjection into layer 4 (see also Fig. 14). D Imtion into layer 5 (see also Fig. 15). E Injection into leyer 6 (see also Fig. 16).

370

1

2

3

4

5

6

WM

J.B. LEVITT ET AL.

Fig. 10. Example of biocytin reacted section after injection into layers 2iupper 3, counterstained with cresyl violet for cytoarchitectural identification of laminar boundaries. Scale bar = 200 bm.

terminal clusters in layers 1-3. A dense focus of fibers was also seen descending towards the white matter, and axon collaterals spread within layer 5 as well, although the spread within layer 5 was less extensive than in layers 1-3. We occasionally observed fine fibers turning upwards from layer 5 to make light contributions to the terminal clusters in layers 1-3. Laterally spreading fibers were again notably absent from layer 4. Labeling in layer 6 was mainly restricted to the columnar focus beneath the injection, although occasional obliquely descending fibers were also observed.

The photomicrograph in Figure 13 illustrates a biocytin injection into layer 4 (with some intrusion into deep layer 3 as well). The prominent feature of such injections is the dense columnar focus of fibers above and below the injection core (see also summary in Fig. 9C). We also noted labeled cells in layers 2-6, in vertical registration with the injection. Figure 13A also shows fibers spreading horizontally to the sides of the injection (due to layer 3 intrusion). At higher magnification, to the left of the injection (Fig. 13B), these fibers are seen to turn suddenly toward the cortical surface and end in a terminal cluster as described above. Figure 14 illustrates a case in which we succeeded in restricting the injection to layer 4. Following this injection, little or no label was found lateral to the injection core. Label was restricted to the columnar axis above and below the injection. The stout dendrites seen to the right of the injection in Figure 14A were labeled by the injection pipette coming into cortex from the right. The columnar focus of projections from layer 4 may be appreci- ated in Figure 14B, which shows a section located 150 km away from the central injection core. Here, too, label is


Fig. 11. A,B: Photomicrographs illustrating a biocytin injection into layer Ziupper layer 3. B is a higher magnification view, and the asterisks indicate the same blood vessel in A and B. Laterally spreading fibers are prominent in layers 1-3; these periodically turn toward the pial surface and ramify in terminal clusters; the solid black arrow in B points to one such cluster. Oblique fibers can also he seen (open arrow in B) descending to the base of layer 3 (see also summary diagram in Fig. 9). Scale bars = 200 pn.

restricted to the injection column, and little or none is found laterally.

Following biocytin injections into layer 5, we observed prominent horizontal fibers running within the layer (Fig. 15, see also Fig. 9D for summary). These fibers provided light rising contributions to layer 3, passing through layer 4. There was also a dense columnar focus of label above the injection all the way to the pial surface, with diffusely spreading fibers in layer 3. The lateral spread in layer 3 was more restricted than that seen in layer 5, or that seen in layer 3 after injections into the upper layers. The clustering of terminals was also much less prominent in the upper layers. Labeled cells formed a column above and below the injection; most were in layers 3 and 5. Layer 4 was again distinguished by its lack of horizontal projections. In Figure 15B, a higher power view



Fig. 12. Injection into layer 3. Laterally spreading fibers are again prominent in layers 1-3; a lighter lateral plexus is now seen in layer 5 as well. A dense columnar focus of projections extends through layers 1-6; descending fibers pass through layer 4 without emitting horizontal collaterals. Retrogradely labeled cells are found in layers 3-5. Scale bar = 200 pm.

of a different section, this is evident as a fiber-free gap, above and below which (in layers 3 and 5) horizontal fibers can be seen.

Figure 9E summarizes the result of placing an injection into layer 6, and Figure 16 is a representative photomicrograph. Labeled cells were seen above the injection as high as the base of layer 3. Projec- tions ascended in a focused column, then spread locally in layers 1-3. Laterally spreading fibers were seen within layer 6, but only light rising contributions were found in layer 5 , and there was no evidence for lateral clusters of terminals in any layer.

Znjections into layer 6.

DISCUSSION Technical issues

Apart from cells immediately local to the injection site, the biocytin labeling described in this study of areas 9 and 46 appeared to be due almost entirely to orthograde transport and for the most part to be restricted to fine axonal processes. From their morphology, in comparison to axon

processes of pyramidal and local circuit neurons observed in Golgi preparations of the same region (Lund and Lewis, '93), we believe these fine axon processes to arise almost entirely from pyramidal cells. Although in the immediate vicinity of injection sites we did observe rare examples of other cell types in addition to pyramidal neurons, such as inverted pyramidal or fusiform cells, away from the injection sites very few neurons were retrogradely labeled, and these were all pyramidal in morphology. We have no explanation of why local circuit neurons are so seldom labeled; their general failure to label in macaque monkey cortex after biocytin injections contrasts with the situation in the cat, in which (for visual cortex at least) basket neurons are not uncommonly labeled (Kisvarday et al., '93). The extremely sparse retrograde labeling of pyramidal neuron somata and dendrites contrasts with other cortical areas that we have studied in the same species using the same injection sizes and techniques (Vl, V2, V4, somatosensory, and motor areas), where many more pyramidal neurons were retrogradely labeled (Lund et al., '93). Compar- ing the diameter of fibers labeled in each area and the


Fig. 13. A: Injection into deep layer 3 (intrusion on layer 4). This injection also shows the prominent vertical fiber focus. B: Higher power view: note the horizontal fibers that then turn toward the pial surface in a terminal cluster. Scale bars = 200 wm.

numbers of retrogradely labeled neurons, it appears that retrograde labeling in pyramidal neurons may depend heavily on fiber diameter. For example, in motor cortex many pyramidal cells were retrogradely labeled at long distances from the injection sites, and the fiber diameters were the largest seen among the various regions studied, whereas in the prefrontal region, where the fibers were the finest observed, the fewest cells were found. This in turn suggests that the long distance projections seen surrounding each injection site in prefrontal cortex arose from neurons within the injection focus, and did not derive from labeling of axon collaterals of retrogradely labeled neurons whose axon trunks passed through the injection site. This observation is consistent with the majority of reports that biocytin label is not taken up by fibers of passage (Horikawa and Armstrong, '88; King et al., '89), although a recent

Fig. 14. Injection restricted to layer 4. A Orthogradely labeled fibers are seen above and below the injection core, but little or no label is found laterally. This is seen clearly in a section 150 pm away (B), where label is restricted to the injection axis. PS, principal sulcus. Scale bar = 200 wm.

study suggests that direct injections into fiber tracts can result in biocytin transport (Chevalier et al., '92). However, the fact that the laminar distribution pattern of the biocytin- labeled axon terminals does not match that of any known afferents to macaque prefrontal cortex, including the major corticocortical and thalamocortical projections (Schwartz and Goldman-Rakic, '84; Giguere and Goldman-Rakic, '881, further suggests that the observations of the present study did not result from labeling of axons of passage.


Fig. 15. Injection into layer 5. A: A lateral fiber plexus is seen in layer 5; vertically ascending fibers pass through layer 4, and terminate in layers 213. B: The lack of horizontally spreading fibers in layer 4 is apparent as a gap in the high-power view of a different section. Scale bars = 200 pm.

Interlaminar pyramidal neuron projections The pattern of interlaminar projections of pyramidal

neurons at different cortical depths is summarized in Figure 17. The different patterns of axon label produced by injections into different laminae are not simply due to the presence or absence of pyramidal neurons; all layers are well populated with pyramidal neurons, and even the small cells of layer 4 are largely pyramidal in morphology. As is shown in Figure 17, pyramidal neurons in each layer of macaque prefrontal cortex tend to give rise to a particular arrangement of intracortical connections. The results ob-

Fig. 16. Injection into layer 6 . Laterally spreading fibers are seen within layer 6, with sparse ascending fibers through layer 5. Fibers also ascend vertically through layer 4 and ramify in the upper layers. Scale bar = 200 pm.

pyramidal neurons in superficial layers 2 and 3 of prefrontal cortex emit long horizontal collaterals that distribute their terminals in a discontinuous but highly ordered fashion across the neuropil of layers 1-3. Their descending axon trunks give off laterally spreading collaterals in layer 5 , some of which turn up again to enter the superficial layers and contribute to the bands of terminal label. Of cells in the deeper layers, only the axons of layer 5 pyramidal neurons seem to make laterally spreading and rising collateral contributions to the terminal arrays in the superficial layers. This laminar specificity distinguishes the intrinsic connection system from thalamocortical and corticocortical afferents, as these terminate, respectively, in layers 4 and deep 3, or in a column extending through all cortical layers (Goldman-Rakic and Schwartz, '82; Giguere and Goldman- Rakic, '88). The interlaminar circuitry is apparently very much like that found in other cortical regions, such as visual areas V2 and V4 (Levitt et al., '94; Yoshioka et al., '92); these regions share the feature of a narrow small- celled layer 4 and periodic connectivity in the superficial layers, to which layer 5 makes some lesser contribution.

Lateral connectivity of layer 2/3 pyramidal neurons

The patterns of terminals observed suggest that single pyramids of layers 213 project to a bandlike region, up to several millimeters in length, centered on the cell's own dendritic field. Collaterals of the axon spread out beyond

tained from the small injections show that the axons of this band passing through regions on- either side" and

3 74 J.B. LEVITT ET AL.

W M

Fig. 17. Summary diagram of intrinsic pyramidal neuron projections in prefrontal areas 9 and 46 suggested by serial reconstructions of orthograde biocytin transport. A Pyramidal neuron of layer 2 and superficial layer 3 makes long distance lateral connections to patches of tissue in layers 1-3. Laterally travelling fibers are primarily in layers 2 and superficial 3, with oblique descending collaterals through layer 3. Descending axons give off collaterals in layer 5; these travel horizontally, and some emit light rising terminal contributions to the upper layers. B: Pyramidal neuron of layer 3 also gives rise to prominent horizontal projections with periodic terminations in layers 1-3. The periodic terminal clusters from superficial layer pyramidal cells are approximately 270 Fm wide and are spaced 540 pm center to center.

roughly parallel to the central band, without terminating in them to then make a series of elongated terminal fields interleaved with uninnervated territories of equal dimensions, the whole stripe-like array covering 10-20 mm2 of cortical territory. This also suggests that in the reciprocal direction a single pyramidal neuron may be the target of axon inputs from large populations of pyramidal neurons, each distributed in a nonrandom stripe-like fashion across a large part of the same region. Since each of our injections in the superficial layers produced rather similar patterns of lateral projections, it suggests that these connections arise from every point across the cortex as a continuum, rather than there being a wide-spreading projection system at fixed positions interleaved with gaps between. In addition, we could see no evidence of a boundary or break in patterns of label crossing regions that would encompass the border between areas 9 and 46. This might suggest that the region is a functional continuum.

This type of connectivity is in some ways similar in pattern to that observed in the primary visual cortex of the tree shrew, in which beaded stripe-like arrays of labeled terminals are seen, even after very large injections of HRP (Rockland et al., '82; Mitchison and Crick, '82). The pattern does, however, differ from intrinsic connectivity in other areas of the primate cortex that we have examined (areas V1, V2, V4, primary somatosensory, and motor regions; Lund et al., '93), where the labeled terminal regions form circular or oval patches rather than stripes, and the collater- alization of single axon trunks labeling from similarly sized injection sites (particularly in areas V1 and V2) seems much

f

These terminal zones are elongated into bands 2 4 mm long and are generally oriented orthogonal to the principal sulcus. Collaterals descend diagonally though layers 3-5, and the descending columnar focus also emits collaterals in layers 5 and 6. C: Pyramidal neuron of layer 4 projects in a strictly columnar fashion with little or no lateral spread. D: Pyramidal neuron of layer 5 projects extensively within layer 5 and into layer 6. Ascending columnar projections give rise to more diffuse spreading terminations in layers 1-3. Periodic terminations are less prominent. E: Pyramidal neuron of layer 6 projects laterally within layer 6 and sends light rising fibers into layer 5. Lighter ascending columnar projections to the superficial layers are also observed.

less than that occuring in the prefrontal cortex. Similarities between the intrinsic connectivity systems in all regions, however, include the origin from pyramidal neurons, the primary localization to the superficial layers, the weak contribution from layer 5, the extensive lateral spread, the regular repeating nature of labeled and unlabeled regions with approximately equal width to each, and the apparently continuously distributed nature of origin of the arrays from every point across the cortex (Lund et al., '93).

Formation of discontinuous intrinsic connections

The formation of lattice arrays, in at least the visual cortical areas, has been suggested to depend on early postnatal afferent activity patterns, and the cortical domains linked together can apparently be changed by manipu- lation of the visual input (Callaway and Katz, '91; Lowel and Singer, '92). However, the precise constraints operating to build the regular but spatially discontinuous pyramidal neuron axon arrays in the superficial layers of each cortical area are not yet fully understood. We demonstrated in a previous study, in which we compared the scale of these arrays across different cortical areas in the macaque and between visual areas in tree shrew, cat, and macaque (Lund et al., '931, that terminal patch or stripe width was exactly matched to the width of the interspersed terminal-free gaps and, moreover, that both terminal patch and gap width matched the mean width of the dendritic spread of single pyramida1 neurons contributing to these connections in


each area, including the prefrontal region studied here. This suggests that some general principle of cortical organization is operating to build these arrays and also suggests the local scale on which it should operate.

We have suggested that local inhibition, in the form of basket neuron axons that we find to spread to three times the width of the pyramidal neuron dendritic arbors in at least visual and prefrontal cortices, may constrain the distribution of the pyramidal neuron axons to these discontinuous terminal zones (Lund et al., '93). The basic principle suggested is that afferents to the superficial layers simultaneously activate both excitatory pyramidal and inhibitory basket neurons that are spatially coincident. The basket neuron axon spread creates an inhibitory surround to the pyramidal neuron so that in trying to build contacts during development on other simultaneously active pyramidal cells it has to pass beyond this zone of inhibition. If the basket neuron axon is evenly distributed around the pyramidal neuron, and this configuration is the same for every pyramidal neuron across the cortex, then the end result is that each pyramidal neuron distributes terminals to evenly spaced patches around the cell of origin; indeed, perhaps as a consequence of the same inhibitory constraint, specific functional properties in area V1 repeat across the cortex at the same scale. We have suggested that the stripe-like domains of pyramidal neuron axon terminals found in the prefrontal cortex could be due to the local basket neuron axon fields being elongated orthogonal to the stripes of pyramidal neuron axon terminals (Lund and Lewis, '93); this would permit pyramidal neuron axons to find local uninhibited territory for terminals on one axis but would force axon terminals to step over inhibited territory along the opposite axis (Lund et al., '93). While we have yet to investigate the basket axons of the prefrontal region for presence or absence of such anisotropic arbor spread, basket neurons with slab-shaped axons have already been described in motor cortex (Marin-Padilla, '74).

Relation to extrinsic connections Our attempt to determine if there was any match in scale

between the layout of the intrinsic connection system and the distribution of efferent neurons projecting across the corpus callosum resulted in a surprising finding; there seemed to be a close match in size between sparsely populated gaps in the distribution of callosally projecting neurons and the size of the intrinsic system terminal stripes, or unlabeled gaps. Moreover, the mean repeat distance of the intrinsic connectivity system (center to center of stripes) is also the same dimension as the mean size of callosal efferent cell clusters (and a close match to callosal afferent terminal stripes as reported by Schwartz and Goldman-Rakic, '84). The technique of retrogradely labeling sets of efferent neurons by making multiple injections in the region of their axon arbors (as employed here as well as in prior studies of interhemispheric connections of prefrontal cortex; Schwartz and Goldman-Rakic, '84) is open to problems; it is impossible to guarantee a completely even deposit of label across a region, and it is impossible to guarantee an even uptake of the deposited label by the axon terminals of the projection pathway under investigation. Given such problems, it would not be unexpected to find an uneven distribution of retrogradely labeled callosally projecting neurons with regions of variable size lacking retrogradely labeled cells, as we and others (Schwartz and Goldman-Rakic, '84) have observed. However, the finding

of a close match between the mean size and distribution of the intrinsic terminal stripe width and the lacunae in the callosal efferent system, and between the mean repeat distance of the intrinsic system and callosal efferent cell clusters (Fig. 7), is of particular interest since these striking similarities seem unlikely to have arisen simply as a result of the uncertainties in the method of retrogradely labeling the callosal system. The overall repeat of alignments between the intrinsic system terminals and the callosal efferent system at 1,600 pm intervals may therefore represent the functional repeat distance of the region and one that physiological investigators might find useful to ex- plore.

We must stress, however, that we believe the intrinsic connection system is a continuum, such that every point across the cortex connects to a series of offset band-like zones; the intrinsic system therefore does not simply interdigitate with the interhemispheric relays, as is suggested for callosal and interareal corticocortical projections (Goldman-Rakic and Schwartz, '82). The difference in scale between the intrinsic connection system and the discontinuous terminal zones of other afferents to the prefrontal region suggests that neurons at single points in the superficial layers may sample many different afferents via the relays of the intrinsic system (rather than falling into step in distribution with any one afferent system). It also suggests that single points distributed across the cortex may selectively omit one or more of the various efferent projections known to arise from the region (see Barbas, '92, for review). I t is also possible that all the afferent and efferent systems are continuously distributed but with a repetitive, discontinuous distribution from single points, much as is seen in the intrinsic system. It is assumed that most, if not all, of the pyramidal neurons contributing to the intrinsic lateral connectivity also send axon trunks out of the cortex; however, this must be confirmed, since it is known that in some regions of cortex the superficial layers include pyramidal neurons whose axon projections are restricted solely to laterally spreading relays in the superficial layers (see Lund et al., '81, for description of such neurons in area V2 of the macaque cortex).

Functional correlates In considering the role of these intrinsic stripe-like

connections in the superficial layers, comparison with other cortical regions may be helpful. In the primary visual cortex, area V1, superfical layer connectional lattices have been suggested to connect regions of similar function (e.g., of the same orientation specificity; TS'O et al., '86; Gilbert and Wiesel, '89). It has been suggested that the prefrontal region may be crucial for maintaining an internal represen- tation of sensory cues during the delay period of delayed response behaviors (see Goldman-Rakic, '87; Fuster, '89; for reviews). Although the functions of area 9 are not as well studied as those of area 46, this region may utilize kinesthetic cues in mediating delayed response-type behaviors (Manning, '78; Passingham, '78). In area 46, individual neurons fire during the delay period of oculomotor delayed response tasks, but only in response to visual cues with a particular spatial location (Funahashi et al., '89). The intracortical axon collaterals furnished by pyramidal neurons in this area may serve to connect neurons with similar response properties. We find, however, that in area V2, a visual association region served by three sets of afferent fibers from area V1, each conveying different sensory


signals, the superficial layer connectional lattice links together points in different afferent terminal but of

" 9 the intrinsic connections may Serve to bring together different attributes of the same stimulus. Different subdivi- sions of area 46 have been reported to receive input from different sensory modalities (Barbas and Mesulam, '85). The intrinsic lateral connectivity system may therefore Serve to create associations between different sensory events presented by the afferents, creating unique patterns of lattice array activity' which are then active for a period of time while a "correct" motor response is also associated with the pattern. The distributed nature Of the intrinsic connections suggests that the significant activity of the region is encoded in distributed patterns of activity.

King, MA., P.M. Lewis, B.E. Hunter, and O.W. Walker (1989) Biocytin: A versatile anterograde neuroanatomical tract-tracing alternative. Brain Res. 497:361-367.

large basket cells in cat visual cortex (area 18): implications for lateral disinhibition. J. Comp. Neurol. 327r.398-415.

LeVay, S. (1988) Patchy intrinsic projections in visual cortex, area 18, of the eat: Morphological and immunocytochemical evidence for an excitatory function. J. ComP. Neural. 269:265-274.

Levitt, J.B., T. Yoshioka, and J.S. Lund (1994) Intrinsic cortical connections in macaque visual area V2: Evidence for interaction between different functional streams. J. Comp. Neurol. (in press).

Livingstone, M.S., and D.H. Hubel (1984) Specificity of intrinsic connections in primate primary visual cortex. J, Neurosci, 4r2830-2835.

Lowel, S., and W. Singer (1992) Selection of intrinsic horizontal connections in the visual cortex by correlated neuronal activity. Science 255:209-212.

Lund, J.S., and D.A. Lewis (1993) Local circuit neurons of developing and mature macaque prefrontal cortex: Golgi and immunocytochemical characteristics. J. Comp. Neurol. 328332-312.

Lund, J.S., A.E. Hendrickson, M.P. Ogren, and E.A. Tobin (1981) Anatomi- cal organization of primate visual cortex, area VII. J. Comp. Neurol. 202:19-45.

Lund, J.S., T. Yoshioka, and J.B. Levitt (1993) Comparison of intrinsic connectivity in different areas of macaque monkey cerebral cortex.

Manning, F.J. (1978) Dorsolateral prefrontal cortex lesions and discrimina- tion of movement-produced cues by rhesus monkeys. Brain Res. 149:77-

Marin-Padilla, M. (1974) Three dimensional reconstruction of the pericellu- lar nests (baskets) of the motor (area 4) and visual (area 17) areas of the human cerebral cortex: a Golgi study. 2. Anat. Entwicklungsgesch. 144:123-135.

Matsubara, J.A., and D.P. Phillips (1988) Intracortical connections and their physiological correlates in the primary auditory cortex (A11 of the cat. J. Comp. Neurol. 268r38-48.

McGuire, B.A., C.D. Gilbert, P.K. Rivlin, and T.N. Wiesel (1991) Targets of horizontal connections in macaque primary visual cortex. J. Comp. Neurol. 305370-392.

Mikchison, G., and F. Crick (1982) Long ayons within the striate cortex: Their distribution, orientation and patterns of connection. proc. Nail, Acad. Sci. USA 79r3661-3665.

Ojima, H., C.N. Honda, and E.G. Jones (1991) Patterns of axon collateraliza- tion of identified supragranular pyramidal neurons in the cat auditory cortex. Cereb. Cortex 1230-94.

passingham, R, (1978) Information about movements in monkeys ( M ~ ~ ~ ~ ~ mulatta) with lesions of dorsolateral prefrontal cortex. Brain Res. 152:313-328.

Rockland, K.S., and J.S. Lund (1983) Intrinsic laminar lattice connections in primate visual cortex. J. Comp. Neurol. 2I6r303-318.

Rockland, K,s,, J.S. ~ ~ ~ d , and A.L. Humphrey (1982) hatomical banding of intrinsic connections in striate Comp. Neurol. 209.4-58.

spheric connectivity of the prefrontal association cortex in rhesus monkeys: relation between intraparietal and principal sulcal cortex. J. camp, ~ ~ ~ ~ ~ l . 226;403-420.

Szabo, J., and W.M. Cowan (1984) A stereotaxic atlas of the brain of the cynomo~gus monkey (Mucacu fusciculuris). J, Camp. 222r265- 3oo,

Ts'o, D.Y., C.D. Gilbert, and T.N. Wiesel (1986) Relationships between

topography (Levitt et '94). Thus, in area Ksv&.day, Z.F., C, Beaulieu, and U,T, Eysel (1993) Network ofGmAergic

ACKNOWLEDGMENTS We thank Suzanne Holbach and Tom Harper for techni- assistance and Mary Brady for photographic assistance'

This work was supportedby NE1 Pants Ey05282, Ey10021 (J.S.L.), EY08098 (Pittsburgh Eye and Ear Institute); Cereb. Cortex 3:148-162. NRSA grant EY06275 (J.B.L.); NIMH grants MH00519 and MH45156 (D.A.L.); MRC grant 9203679N (J.S.L.); and an ARVOiALCON postdoctoral research fellowship (T.Y.).

LITERATURE CITED Adam% J.C. (1977) COnsiderati0n5 On the use Of horseradish

peroxidase as a neuronal marker. Neuroscience2:141-145. Barbas, H. (1992) Architecture and cortical connections of the prefrontal

cortex in the rhesus monkey. Adv. Neurol. 57r91-115. Barbas, H., and M.-M. Mesulam (1985) Cortical afferent input to the

principalis region of the rhesus monkey. Neuroscience 15619-637. Callaway, E.M., and L.C. Katz (1991) Effects of binocular deprivation on the

development of clustered horizontal connections in cat Striate cortex. Proc. Natl. Acad. Sci. USA 88t745-749.

Chevalier, G., J.M. Deniau, and A. Menetrey (1992) Evidence that biocytin is taken up by axons. Neurosci. Lett. 140r197-199.

DeFelipe, J., M. Conley, and E.G. Jones (1986) Long range local collateraliza- tion of axons arising from corticocortical cells in monkey sensory motor cortex. J. Neurosci. 6:3749-3766.

Feldman, M.L. (1984) Morphology of the neocortical pyramidal neuron. In A. and E.G. Jones (eds): Cerebral Cortex. Cellular Components Of

Funahashi, s., C.J. Bruce, and P.S. Goldman-Rakic (1989) Mnemonic coding the Cerebral Cortex. New York: Plenum Press, pp. 123-200.

of visual space in the monkey's dorsolateral re frontal cortex. J. Neurophysiol. 61r33 1-349.

Neuropsychology of the Frontal Lobe, 2nd ed. New York: Raven Press. Giguere, M., and P.S. Goldman-Rakic (1988) Mediodorsal nucleus: Areal,

laminar, and tangential distribution of afferents and efferents in the frontal lobe of rhesus monkeys. J. Comp. Neurol. 277t195-213.

Gilbert, C.D., and T.N. Wiesel (1989) Columnar specificity of intrinsic horizontal and corticocortical connections in cat visual cortex. J. Neuro- sci. 92432-2442.

regulation of behavior by representational memory. In: Handbook of Physiolo~-The Nervous System v, Washington, DC: American physi- as revealed by cross-correlation analysis. J. Neurosci. 6r1160-1170. ological Society, pp. 373-417. Walker, A.E. (1940) A cytoarehitectural study of the prefrontal area of the

Goldman-Rakic, P.S., and M.L. Schwartz (1982) Interdigitation of contralat- macaque monkey' J' 'Omp.

era1 and ipsilateral columnar projections to frontal association cortex in Wallace, M.N., L.M. Kitzes, and E.G. Jones (1991) Intrinsic inter-and primates. Science 216:755-757. intra-laminar connections and their relationship to the tonotopic map in

Horikawa, K., and W.E. Armstrong (1988) A versatile means of intracellular cat primary auditory cortex. EXP. Brain labeling: Injection of biocytin and its detection with avidin conjugates. J. White, E.L. (,1989) Cortical Circuits. Synaptic Organization of the Cerebral Neurosci. Methods 25t1-11. Cortex. Structure, Function and Theory. Boston, Basel, Berlin:

Juliano, S.L., D.P. Friedman, and D.E. Eslin (1990) Cortico-cortical connec- Birkhauser, pp. 5-223. tions predict patches of stimulus-evoked metabolic activity in monkey Yoshioka, T., J.B. Levitt, and J.S. Lund (1992) Intrinsic lattice connections somatosensory cortex. J. Comp. Neurol. 298.23-39. of macaque monkey visual cortex area V4. J. Neurosci. 122785-2802,

of tree shrews (Tuppaia glis). J.

Fuster, J.M. (1989) The 'refrontal Cortex: Anatomy, phYsiO1Oa, and Schwartz, M.L., and P.S. Goldman-R&ic (1984) Callosa] and interhemi-

Goldman-Rakic, P.S. (1987) Circuitry of Primate Prefrontal cortex and horizontal connections and functional architecture in cat striate cortex

73r59-86'

86r527-544.

topography of pyramidal neuron intrinsic connections in macaque monkey prefrontal cortex (areas 9...

Documents