THE JOURNAL OF COMPARATIVE NEUROLOGY 29Ek23-39 (1990)

Corticocortical Connections Predict Patches of StirnulUslEvokedMetabolic Activity in Monkey Somatosensory Cortex SHARON L. JULIANO, DAVID P. F‘RIEDMAN, AND DON E. ESLIN Department of Anatomy, USUHS, Bethesda, Maryland 20814 (S.L.J., D.E.E.), Laboratory of Neuropsychology, NIMH, Bethesda, Maryland 20892 (S.L.J., D.P.F.), and Division of Preclinical Research, NIDA, Rockville, Maryland 20857 (D.P.F.)

ABSTRACT Stimulus-evoked metabolic activity in the anterior parietal cortex (areas 3a, 3b, 1, and 2) occurs in the form of column-like patches. Similar patches characterize the connections to, within, and from these fields. The relation of the patches elicited metabolically to those formed by retrograde or anterograde transport, however, is not clear. If a type of projection connecting areas 3a, 3b, 1, and 2 transmits sensory information among these cortical fields, the resultant projection pattern may directly contribute to the definition of somatosensory metabolic “columns.” To test this possibility, electrophysiological recordings in areas 3b and 1 of Mucaca fuscicularis monkeys characterized stimuli that elicited the best neuronal response at a specific cortical site. Iontophoretic injections of wheat germ agglutinin-horseradish peroxidase (WGAHRP) were subsequently made into the identified cortical sites. Two days later, the animals were injected with 2-deoxyglucose (2DG) and received the somatic stimulus previously determined to best activate the neurons isolated at the injected cortical site. After injections of WGA-HRP into physiologically defined loci in area 3b, the patches of transported label within areas 3b and 1 were colocalized with evoked metabolic activity. Injections of WGA-HRP into area 1 produced anterogradely labeled terminals in areas 1 and 2 that also overlapped with patches of evoked metabolic activity, as did patches of retrogradely labeled cells located in area 3b. Patches of anterograde label found in area 3b after area 1 injections, however, were not coincident with the metabolically activated patches. These findings suggest that excitatory information is transmitted from area 3b to area 1in a way that connects clusters of cells with similar response properties. Key words: cortical column, serial processing, wheat germ agglutinin, horseradish peroxidase, 2-deoxyglucose

The idea that a column of neurons extending through all the cellular layers of the cortical mantle is the functional unit of information processing in neocortex was initially proposed by Mountcastle in 1957. Although the concept of a cortical column is generally well accepted, it has been difficult to clarify the precise features that characterize functional columns in somatosensory cortex, although a number of theories have been proposed (Sur et al., ’80, ’84; Dykes and Gabor, ’81; Favorov and Whitsel, ’88).Nevertheless, the connections to, from, and within the somatosensory cortex form individual patches that are strongly reminiscent of the functional columns proposed by Mountcastle (Friedman and Jones, ’80;Jones et al., ’82; for review see Jones, ’86). The focused aggregates of thalamocortical and corticocortical fiber terminations have been suggested as the morphologic basis of somatic functional columns

PUBLISHED 1990 BY WILEY-LISS,INC.

(Friedman and Jones, ’80; Jones et al., ’82; Jones, ’86; DeFelipe et al., ’86). Evoked metabolic activity in cat and monkey somatosensory cortex reveals that metabolic activity in response to somatic stimulation also forms column-like patches (e.g., Juliano et al., ’81, ’89; Juliano and Whitsel, ’87). Although both the connectional and metabolic forms of patchy label have been recognized for a number of years, a relationship between stimulus-evoked metabolic “columns” and patchlike columns of afferent terminations to or within the somatosensory cortex has not been established. In visual cortex, a number of studies have questioned whether columns of neurons characterized by specific response properties project to targets that share similar functional properties. In 1986, T’so et al. demonstrated by Accepted April 16,1990.

24

S.L. JULIANO ET AL.

TABLE 1. CorrespondenceBetween HRP and 2DG in Areas 3b and 1for All Analyzed Injection Sites.’ Totalarea3b+ 1 Monkey 1 Area 3h inJdon-foot HRP + no 2DG HRP + 2DG 2DG+noHRP Monkey 2 Area 3b Injectdon-hand H W + no 2DG HRP + 2DG 2DG+noHW Monkey 3 Area 3h injechon-foot H R P+ n o 2 D G H W t 2DG 2DG + no HRP Area 1mnjdon-hand HRP + no2DG HRP + 2DG 2DG + no HRP Monkey 4 Area 1mnjdon-hand HRP + nu 2DG HRP + 2DG 2DG t noHRP Monkey 5 Control injection-foot HRP + no2DG HRP + 2DG 2DG+noHW

Area 3b

Areal

reBon 1

1 7

0 17 8

24 19

11

7 33 15

3 17 9

4 16

25

4

21

41 8

28 1

19

6

4 3

regon 6

reson

7

regon 18 39

31

2 15 8

20 18 12

10 5 6

10 13 6

33 0 28

30 0 11

3 0 17

regon

regon

’Induded are counts of HRP and 2DG patches that overlap (HRP+ 2DG) and those that occur separately (HRP + no 2DG and 2DG + no HRP). The patches were counted on individual seninns and tallied .See text for details.

cross correlograms that in cat area 17, many columns of cells were interconnected and that facilitatory interactions were observed between clusters of cells sharing orientation preferences. In a follow-up study, Gilbert and Wiesel (’89) discovered that clusters of cells projecting to a physiologically identified cortical site aligned with the 2-deoxyglucose (2DG) patterns elicited by stimuli of the same orientation. Similarly, in cat auditory cortex, Matsubara and Phillips (’88) found that injections into cortical sites of neurons responding to known frequencies project to other cortical loci responding to the same, or similar, frequencies as the neurons in the injection site. By contrast, in a study of a “secondary” sensory area, Matsubara et al. (’87) indicated that HRP injections into locations of known response properties in cat area 18 yielded patches of label that coincided with regions of cortex responding predominantly to orientations orthogonal to those of the injection site. As a result of these discrepant findings, questions arise as to the functional relations between individual patches of cells and their targets . This study addresses these issues by asking the question: what are the functional relations between patchy corticocortical projections and their targets in somatosensory cortex? To examine this topic we compared the pattern of corticocortical projections arising from functionally defined cortical regions with 2DG uptake evoked by stimuli that best activated the injected regions. In individual experiments, wheat germ agglutinin-horseradish peroxidase (WGAHRP) was injected iontophoretically into physiologically identified sites of areas 3b and 1of macaque monkeys. The distribution of labeled terminals and retrogradely filled neurons was compared to 2DG uptake evoked by stimuli known to excite neurons in the injected location. The results of this investigation were reported in preliminary form (Juliano et al., ’87).

MATERIALSANDMETHODS Experimentaldesign A monkey was prepared for electrophysiological recording in the anterior parietal cortex (areas 3a, 3b, 1, and 2). During the recording session, two distinct and widely separated cortical sites were identified and characterized according to: the location of the peripheral receptive field (RF) that activated the site, and the responsivity of multiunits to the frequencies, waveforms, and amplitudes of mechanical stimulators. Only loci with relatively small and discrete RFs on the glabrous skin of the hand and foot were considered. When a cortical site responding to a n appropriate RF was identified and characterized, a small iontophoretic injection of WGA-HRP was made into that site. Thirty-six to 48 hours later, a 2DG experiment was carried out using the stimulus previously identified to best activate the injected cortical site. The relations between the transported WGA-HRP and the evoked 2DG were then assessed.

SWwY Five Macaca fascicularis monkeys (male and female), weighing 3-4 kg, were used for these experiments. All surgical procedures were carried out under aseptic conditions using ketamine and pentobarbital anesthesia. Heart rate and respiration were continuously monitored; body temperature was kept within normal limits with servocontrolled heating pads. A bone flap was lifted over the central sulcus, with an exposure large enough to visualize both the hand and foot representations. The bone was saved for later closure of the wound. A well was formed from dental acrylic around the opening and the dura removed. From this point, sterile, artificial cerebrospinal fluid was kept in the well. After the recording session Gelfoam was placed on the surface of the brain, and the bone flap kept in place with sutures.

Recordingsandinjections After the surgery, the animal was intubated with a tracheal cannula, placed in a stereotaxic apparatus, and maintained in an adequate plane of anesthesia with halothane (0.5-1.0%). Heart rate and body temperature were continuously monitored and body temperature kept within normal limits. Single and multiunit activity was recorded through glass insulated tungsten electrodes with tip diameters of 2-5 pm and impedances of 2-5 MR. The signals were amplified and displayed in a conventional manner. A number of guidelines were used to select injection sites. Injections were made into cortical locations where the RFs: 1)were relatively small (because our mechanical stimulator could not effectively cover a large peripheral field); 2) were on glabrous skin; and 3) responded strongly to the stimuli produced by our mechanical stimulator. When an appropriate location was identified, an injection of 10% WGA-HRP, dissolved in tris buffer (pH 7.4 at room temperature), was made into the same location through a pipette with a tip diameter of 10-15 p,m. The injections were of 15 minute duration, electrode negative current, pulses 8 seconds on, 8 seconds off. In general, two sites were chosen for each animal, one in the foot representation and one in the hand representation.

2DG AND HRP IN SOMATOSENSORY CORTEX

25

Fig. 1. a: Typical injection site of WGA-HRP shown in a brightfield image. b An adjacent section demonstrating 2DG uptake. The 2DG autoradiograph 6)is printed directly from the film; thus labeled regions are white. The sections are cut in an oblique horizontal plane through the foot region of the anterior parietal cortex, and the injection

is in area 3b. A blood vessel common to both sections is indicated with double arrowheads. Cytoarchitectonic boundaries were determined on adjacent Nissl-stained sections and are indicated with arrows. The 2DG uptake is coincident with the injection site. Anterior is to the left. CS, central sulcus. Scale bar = 400 pm.

2DGexperiments

tained within normal limits. The anesthetic was changed to nitrous oxide (70%) and oxygen (30%).This condition was maintained for the duration of the experiment. In different experiments carried out in this lab, EEG recordings were made. These studies, which maintained the animals under similar conditions, led us to believe that the periodic cyclings observed between synchronized and desynchronized activity are typical of resting, unrestrained animals and indicate that these experimental conditions do not lead

Thirty-six to 48 hours later a 2DG experiment was carried out. The animal was reanesthetized with halothane, the trachea intubated and a catheter inserted in the lateral saphenous vein. Wounds were perfused with a long lasting topical anesthetic ointment (Americaine). At this point, gallamine triethiodide was administered and the animal maintained on a respirator; body temperature, heart rate, and expired CO, were continuously monitored and main-

26

S.L. JULIANO ET AL.

Figure 2

27

2DG AND HRP IN SOMATOSENSORY CORTEX

2

1 ,

\

\

\

2 0 6 STIMULATION SITE \

\

S

\ \

med

Fig. 3. On the right is a two-dimensional reconstruction of 2DG and HRP label. Each patch of 2DG activity corresponding to 70-90% above background is represented as a horizontal line; patches of activity 60-70% above background are indicated as horizontal dashed lines. The HRP label is shown with shading. Loci that were dominated by retrogradely labeled cells are illustrated with The region of the brain reconstructed is depicted in the inset in the upper left. The site

stimulated during the 2DG experiment and shown to activate best the injection site is shown in the lower left. The location of the fundus of the central sulcus (CS) is represented with dots. The injection site is bounded by heavy lines and the boundaries of cytoarchitectonic fields (3b, 1,and 2) indicated with dashed lines. The location of the sections shown in Figure 2 is indicated by the arrow. A good correlation occurs between the transported HRP label and stimulus-evoked 2DG uptake.

to undue stress or pain (Juliano, unpublished observations). Because halothane is known to decrease cortical 2DG uptake (Shapiro et al., '78), 1.5 hours was allowed to elapse between the termination of this anesthetic and the injection of 2DG. The animal was gently positioned on soft pads and the somatic stimulus initiated 5 minutes prior to the 2DG injection. The limbs to be stimulated were carefully posi-

tioned on a padded surface. The stimuli consisted of intermittent vertical displacements of 8-25 Hz, driven by a Ling stimulator with a rounded probe of a diameter ranging from 2 mm to 1 cm. The stimulus waveform was either square or sine wave, depending on which stimulus best excited the neurons in the recording site. In one animal, the site stimulated during the 2DG experiment was different from that identified during the recording session. This control allowed us to determine to what extent coincidence of HRP and 2DG label was random or due to conditions of the experiment. The ''CC-2-deoxy-D-glucosewas injected i.v. (100 FCdkg) and the stimuli continued for 45 minutes. At the end of this period, the animal received an overdose of pentobarbital Na (35 mgkg) and was perfused through the heart with 4% buffered paraformaldehyde. The brain was immediately removed, blocked, frozen in Freon 22, and stored in a freezer at - 70°C. Later, 30 pm thick sections were cut in a cryostat at - 16°C. The plane of section was oblique horizontal, roughly perpendicular to the central sulcus. Alternate

XIS.

Fig. 2. Two sets of adjacent sections reacted for HRP (shown in dark field; a,b ) or exposed to film for 2DG (a',b').The 2DG sections were photographed directly through a microscope without image enhancement. The plane of section is oblique horizontal and taken entirely through area 1. Blood vessels (indicated with arrows) and other morphological features were used to align sections precisely. Cortical laminae (I-VI) and cytoarchitecture were determined by staining the sections used to produce the 2DG autoradiographs with thionin. The transported HRP reaction product overlaps with the stimulus-evoked 2DG label. A reconstruction of this experiment can be seen in Fig. 3. Anterior is to the left. Scale bar = 500 pm.

28

Fig. 4. Montage of HRP and 2DG label from adjacent sections cut in the coronal plane through the foot representation in area 1. The patch of HRP reaction product (b)is transported from an injection in area 3b and contains terminals and cells. The 2DG label shown in a is taken from a digitized section. Blood vessels common to both sections are

S.L. JULIANO ET AL.

identified with arrows. Panel (c)is a drawing of the overlapbetween the HRP and 2DG label. The 2DG label is indicated with shading and the anterograde and retrograde HRP indicated with dots. Common blood vessels are represented with circles. Lateral is to the right. Scale bar = 800 Fm.

2DG AND HRP IN SOMATOSENSORY CORTEX

29

labeled patches were also observed; these corresponded to 14C values 60-70% above background. Cortical regions surrounding the stimulus-evoked patches were 25-35% above white matter values. Occasional above background activity was observed in other forms, such as a band of Histologyanddataanalysis activity extending tangentially through a single cortical The sections saved for HRP histochemistry were treated layer, or very wide patches. It was common to see a band of according to the protocol described by Mesulam (’78) or label in layer IV of area 3b. A light band of activity can be using an adaptation of that described by Gibson et al. (’84). seen in Fig. 4; these tangential bands are not illustrated in The sections used for the autoradiographs were stained the reconstructions. The type of somatic stimulation used with thionin to identify cytoarchitectonic features. The in these experiments produced very little 2DG uptake in HRP sections were analyzed with brightfield and darkfield area 2; this pattern of metabolic activity is typical for microscopy and drawn for reconstruction purposes using a intermittent vertical displacement stimuli, which cause drawing tube. The 2DG sections were analyzed with a activation of cutaneous mechanoreceptors and thereby video-based image analysis system (Tommerdahl et al., ’85) maximally activated areas 3b and 1. WGA-HRP irqiection sites. The injection sites were or by direct prints from the autoradiographs. The autoradiographs containing above background regions of 2DG localized to areas 3b and 1 and ranged between 500 and uptake were drawn a t the same scale as the adjacent HRP 1,000 p m in diameter. A total of six injection sites were sections, using the video image analysis system, which recovered and analyzed for this study. In each case, the allowed us to convert the optical density values to I4C recording and injection site was labeled with 2DG. The 2DG concentrations. To align the above background 2DG uptake uptake was colocalized with the injection site, although the and HRP reaction product exactly, the histologic and HRP reaction product typically occupied a somewhat larger autoradiographic sections were compared by means of the volume of tissue. Fig. 1is an example of an injection site and drawings or photographs at the same magnification and by corresponding autoradiograph from the experiment reconmaking note of blood vessels and other morphologic fea- structed in Fig. 3. Area 3b WGA-HRP injections. The injections into tures. The two forms of label were reconstructed using the area 3b characteristically led to presumptive terminal drawings and by measuring the distance of the HRP or 2DG labeling in area 1. The HRP reaction product, which label from a morphologic landmark (e.g., fundus of central appeared as dust-like aggregations, was most dense in sulcus) and plotting the position of the label in relation to a laminae I11 and IV,often extended into lamina 11, and given morphologic feature. Both forms of label were in- usually occurred in patches (e.g., Fig. 2). From here on, we will refer to this label as “terminal.” Similar patch-like cluded on one reconstruction. aggregations were also found within area 3b. Most often, a Statistid analysis small number of scattered cells together with labeled The relationship between the two forms of label was terminals were found within area 3b or within area 1.Three further characterized by a tabulation of the labeled patches cases of area 3b injections were studied. Area 1 WGA-HRP irqjections. Like injections into (Table 1)and by chi-square analyses of the distributions of HRP and 2DG (Tables 2-4). To carry out this evaluation area 3b, injections into area 1 lead to anterograde label the number of patches of 2DG and/or anterogradely or within areas 1and 2, and this label was also found in layers retrogradely transported HRP were counted on adjacent 11-IV. Injections in this cortical field also led to HRP label in sections according to the following criteria: (1)the presence area 3b. The transport appeared as terminals in layers of an HRP patch of label with no corresponding 2DG label 11-IV or as patches of retrogradely labeled cells in laminae (HRP + no 2DG); (2) the presence of HRP and 2DG patches 11 and 111 (Figs. 7, 8). Two cases of area 1 injections were that were colocalized (HRP + 2DG); and (3) the presence of analyzed. a patch of 2DG uptake with no corresponding HRP label Relationshipbetweenstimulusev0ked2DG (2DG + no HRP). Because HRP label was not found in area andtransportedWGA-HR.P 2 after area 3b injections, we did not include information from area 2 in our quantitative analysis (i.e., Tables 1-4) so Area 3b injections. WGA-HRP injections into physiothat direct comparison of area 3b injections with area 1 logically identified sites of area 3b produced transported injections could be made. label that colocalized with evoked 2DG activity in areas 3b and 1.Fig. 2 illustrates two sets of adjacent sections, taken from area 1after an injection in area 3b, which were reacted RESULTS to visualize HRP and 2DG. In this experiment, the stimulus Generalcharacteristicsof2DGandHRPlabel found to best activate the neurons a t the site was a flutter 2 0 6 . As in previous experiments evaluating stimulus- (i.e., intermittent vertical displacement delivered at 7 Hz evoked 2DG activity, the dominant form of label was with an amplitude of 0.25 mm, using a square wave to drive patch-like. The patches ranged in width from 300 to 650 pm the stimulator) to the ventral surface of digit 1 of the foot. and usually extended from laminae I1 to IV.Occasionally, the label occupied laminae I or V. The distributions of activity were appropriately located within somatosensory Fig. 5. Two-dimensionalreconstruction of HRP (shading) and 2DG cortex for the body part and modality stimulated. As in (vertical lines) label. Conventions are as in Fig. 3, except the lines are previous studies (cf. Juliano et al., ’87, ’891, a patch of vertical in this case, because the plane of section was coronal. The activity corresponded to 14Cvalues that were 70-90% above fundus of the central sulcus (CS) is represented with dots. A good background, where background is defined as the white correlation between transported HRP label and evoked 2DG activity in matter underlying the postcentral gyrus. Less heavily areas 3b and 1 is demonstrated after an injection in area 3b. sections were saved either for 2DG autoradiography, or HRP histochemistry. The 2DG sections were exposed to x-ray film (SB-5, Kodak), with 14C standards for 7-10 days, and developed.

1

3b

IN

I I

I

2DG STIMULATION SITE

I I I I I

I

med

I

Lost

I

0 0 0 0

0

I I I I \

1mm

I

\

w

.c s

I

I

\

Figure 5

cs

0 0

0

0

a

0

0

0

0

a

a

0

0

0

-1

\-

I

I

I

I

I

-----

-

I

\

I

\

\

I

\

I 1 I-

\

I

3b

\ \

I-

\

/

I

I

I

I

I

- -

-

Fig. 6. Two-dimensional reconstruction of HRP (shading) and 2DG (horizontal lines). Conventions are as for Fig. 3. After an injection in area 1,areas of good correspondence between the two forms of label are

2DG STIMULATION SITE

-

3a

I

l m*m

med

2

O

S

t

seen in area 1,but the anterograde transport does not match well with evoked 2DG activity in area 3b. Retrogradely labeled cells in area 3b and 2DG activity overlap. CS, fundus of the central sulcus.

-

1

32

S.L. JULIANO ET AL.

33

2DG AND HRP IN SOMATOSENSORY CORTEX Fig. 2 shows that patches of transported HRP and evoked 2DG uptake overlap. As indicated previously, the injection site colocalized with evoked 2DG activity (Fig. 1). Fig. 3 is a two-dimensional reconstruction of the same experiment. This figure demonstrates that throughout the activated regions of cortex, patches of the two forms of label correspond closely. Table 1 reveals that the majority of HRP and 2DG patches are colocalized in areas 3b and 1 after an area 3b injection. Another reconstruction of the 2DG-HRP relationship after an area 3b injection can be seen in Fig. 5 . In this experiment, the stimulus that best activated the cortical site was a flutter stimulus (intermittent vertical displacement delivered with an amplitude of 0.25 mm at 7 Hz using a square wave) to the pad of digit 1 on the foot. We again found that the majority of patches in area 3b and 1colocalized (Table 1).An example of a patch of HRP reaction product containing both retrogradely labeled cell bodies and anterogradely labeled terminals that overlapped with evoked metabolic activity can be seen in Fig. 4. This label was produced after an area 3b injection and reconstructed in Fig. 5 . In experiments with injections into area 3b, the relationship between the two forms of label was generally good, whether the HRP reaction product occurred predominantly in the form of retrogradely labeled cells or terminals. Area 1 iqjections. The reconstruction of 2DG activity and HRP reaction product shown in Fig. 6 demonstrates the relation between the two forms of label after an area 1 injection. The data with injections in area 1 are summarized in Table 1.As indicated above, such injections generally lead to terminal label within areas 1 and 2. In area 1, the 2DG-HRP correspondenceis good. This is confirmed by Table 1, which indicates that the majority of labeled patches are colocalized in area 1, after area 1 injections. In area 3b, however, only the locations dominated by retrogradely labeled cells were colocalized with the 2DG activity. This can also be observed in Table 1,which shows that after an area 1 injection, the majority of labeled patches in area 3b do not overlap with HRP patches. When retrogradely labeled cells were found in area 3b, the patches of 2DG overlapped with the labeled cells. An example of this can be seen in Fig. 7, which demonstrates a cluster of retrogradely labeled cells and corresponding 2DG patch in area 3b of another experiment that is not reconstructed. A different example of a patch of retrogradely labeled cells coinciding with a 2DG patch can be seen in Fig. 8, with an accompanying Nissl stained section to demonstrate cortical layers. By contrast, the location of the anterogradely transported HRP found in layers 11-IV of area 3b showed little correspondence to the location of the evoked 2DG activity. In the experiments involving area 1 injections, rare instances of HRP transport to layers I and I1 occurred within area 1. In these cases, the layer I HRP label was generally found in close proximity, but not directly superimposed on, a patch of 2DG label localized to laminae 11-IV (Fig. 9). ~~

Fig. 7. Montage of 2DG and HRP label from adjacent sections taken through the central sulcus. a: Digitized 2DG section taken in an oblique horizontal plane through the hand region of area 3b. b: Darkfield image of the adjacent section containing a patch of retrogradely labeled cells. The boxed-in patch of cells is shown in higher power in d. Blood vessels . occurring in both sections are indicated with arrows. The blood vessels

Control andstatistical analysis Control. In one animal, the skin region stimulated during the 2DG experiment was different from the site identified and injected during the recording (Fig. 10). In this experiment, the neurons in the injection site responded best to a flutter stimulus applied to the ventral surface of digit 1of the foot. During the 2DG experiment, we chose a stimulation site that was clearly different from the region of skin activating neurons at the WGA-HRP injection. The ventral surface of digit 2 on the same foot was stimulated. Peripheral locations were selected that should have little overlap in their cortical representations; as a result little coincidence was expected between the two forms of label. Fig. 10 establishes that when the stimulus used during the 2DG experiment did not match the site activating the injected neurons, there was no coincidence between the two forms of label. These data are summarized in Table 1. Statistical analysis. The distributions of HRP and 2DG patches reported in Table 1 were analyzed using chi-square to address two null hypotheses.

1. The first null hypothesis tests differences between injections made into area 3b versus area 1. It states that there are no differences between the coincidence of HRP and 2DG after either an area 3b or area 1 injection made into a cortical site activated by the same stimulus used during the 2DG experiment. 2. The second hypothesis tests differences between area 3b (or area 1) injections versus control injections into cortical sites not activated by stimuli used during the 2DG experiment. The null hypothesis states that there is no difference between the coincidence of HRP and 2DG after either an area 3b or area 1 injection compared with the distribution found after a control injection (see Table 2). The first hypothesis was tested for the total label in areas 3b + 1(Table 2), as well as separately for the label found in each cytoarchitectonicfield (Tables 3,4).The second hypothesis was tested for the total label in areas 3b + 1 (Table 2) and for the label found in area 1(Table 4). 1. These calculations verify that there are differences in the distributions of HRP and 2DG after 3b injections versus area 1injections. When the injection site is activated by the stimulus used during the 2DG experiment, the coincidence of HRP and 2DG is significantly different after an injection into area 3b compared with an injection into area 1 (Table 2). When the distributions found in areas 3b and 1 are evaluated separately, we find that the distributions of HRP/2DG found in area 3b alone are significantly different when area 3b is injected compared with the area 3b distribution after an area 1injection (Table 3).The distributions found in area 1, however, are not significantly different when area 3b injections are compared with area 1 injections (Table 4). Table 1demonstrates that the distribu-

in the 2DG autoradiograph are somewhat obscured by the digitization process. c: Drawing of the relationship between the two forms of label. The sections were aligned using blood vessels and other morphologic features. The ZDG label in c is indicated with shading and the HRP is indicated with dots. The blood vessels are reuresented bv circles: anterior is to the right. Scale bar a-c = 800 Wm; d = 100 Wm.

34

S.L. JULIANO ET AL.

m

35

2DG AND HRP IN SOMATOSENSORY CORTEX TABLE 2. Coincidenceof HRP and 2DG in Areas 3b + 1 -

TABLE 3. Coincidenceof HRP and 2DG in area 3b

Area 3b injection (n = 2)

Area 1injection (n = 2)

Control injection (n = 1)

26 65 27

26 36

33 0 28

HRP + no2DG HRP+PDG 2DG + noHRP

51

Area 3b injection (n = 2)

Area 1injection (n = 2)

Control injection (n = 1)

5 35

14 8 37

30 0 11

HRP + no 2DG HRP 2DG 2DG + no HRP

+

12

~

Area 3b injeztioncompared with area 1 injection,total a3h + 1 Chi square analysm for the total distributionof HRP and 2M: in areas 3h + 1, comparing the totallabel resulting from the arra3b injectionswith that resulting from the area 1injedions

Area 36 injection compared with area I injection, distributionin area3b alone Chi square analysis for the total distributionof HRF' and 2DG in area3h alone, comparing the label resulting fmm the area3h injections with that resultingfmm the area 1injections

Expected frequencies

Expeded frequencies

HRP+no2DG HRP + 2DG 2DG + no HRP Chi square (95%)= 5.991 Chisquarecalc. = 15.61

Area 3b injection

Arm 1injection

Totals

26.56277 51.59308 39.84416

25.43723 49.40693 38.15585

52

101 78

2 D.F.

Significant,P 5 0.005

HRP+no2DG HRF' + 2DG 2 W + no HRP Chi square (95%)= 5.991 Chisquarecalc. = 33.664

Area 3b injection

Area 1injection

Totals

8.30251 20.14414 22.95496

10,0991 22.85586 26.04505

19 43 49

2 D.F. Si@cant,P

< 0.005

Area 1injeztion cornpard with control injection,total amas 3b + 1 Chi square analysis for the total distributionof HRF' and 2DG in areas 3h + 1, comparing the total label resulting from the area 1injectionswith that resulting from the control injection

Expeded frequencies

Control injection

under study had on the results; there are a number of issues to consider.

1.Injection site. Although the injection sites were similar in size to the size of cortical columns inferred from patches 31.8857 14.11043 46 HRP + no 2DG demonstrated by 2DG and anatomical methods (500-1,000 37.43559 16.56442 54 HRF' + 2DG pm), the probability of such an injection being confined to a 2DG+noHRP 43.677485 19.32515 63 single column is not great. As a result, we can expect some 2 D.F. Chisquare(9592)= 5.991 Chi square calc. = 7.731 Sigmfmnt,P 5 0.025 portion of label to be transported from a region surrounding the column in which the center of the injection and the 2DG activation was located. This spread of injection would be expected to result in patches of HRP label without tions of label in area 1 always contain a majority of accompanying 2DG activity. colocalized patches (i.e., HRP + 2DG), compared with ei2. Stimulation specificity. We feel confident that the ther form of label occurring alone (HRP + no 2DG, stimulation applied during the 2DG experiment excited 2DG no HRP). cells at the injection site. It is not possible to be sure, 2. These calculations indicate that significant differences however, that the same stimulus did not also excite neurons exist between the distributions of HRP/2DG after injections at other cortical locations. In fact, it seems likely that this into sites activated by the stimulus used during the 2DG occurred. Fig. 3 shows a case in which area 3b contains a experiment versus control injections, not activated by the number of 2DG patches that do not correspond with HRP 2DG stimulus. Comparison of the values found in areas 3b patches. In other cases the correspondence is nearly com+ 1 together (after injection sites either in area 3b or 1) plete (e.g., Fig. 5; also see Table 1).A number of studies of with the distribution found after an injection into a cortical primate somatosensory cortex indicate that a point on the site that was not activated by the stimulus applied during body is represented a number of times in the cortex (Nelson the 2DG study indicates that these distributions are signifi- et al., '80; McKenna et al.,'82; Favorov and Whitsel, '88). It cantly different (Table 2). The same is true for the label is therefore reasonable to expect locations of 2DG uptake found in area 1 alone (Table 4). This statistical analysis without corresponding HRP reaction product. As a result a reinforces the observation that the area 3b injections lead degree of imprecision is inherent in these experiments. to close correspondence between evoked 2DG and trans- Because some of the HRP and some of the 2DG label will ported HRP in both areas 3b and 1,whereas injections into occur independently, it is remarkable that in the experiarea 1lead to better correspondencebetween the two forms ments involving area 3b injections, we found such a high of label within area 1,than in area 3b. degree of correspondence between the 2DG uptake and HRP transport. Area 1injection

Totals

+

DISCUSSION Methodologicaland control considerations It is useful to consider what effect the amount of variability attributable to our methods and the neural systems

The control experiment and statistical analysis further support our impression that the coincidence between the two forms of label is not random. The control experiment (Fig. 10) demonstrates that the connections of a group of

Fig. 8. Illustration of a patch of 2DG uptake (b)in area 3b near the 3b-1 border. The autoradiograph is not digitized and is taken directly through the microscope. a is the section that produced the 2DG autoradiograph, stained for Nissl substance. The 3b-1 border is indicated with arrows in a x , area 1is on the left, and area 3b is on the right. In a, cortical layers are indicated with roman numerals on the left for area 1and on the right for area 3b. c is a darkfield photomicrograph of a

patch of retrogradely labeled cells on an adjacent section produced from a WGA-HRP injection in area 1. The same blood vessel is indicated with a dark circle in a and b; its location is indicated with double arrows in c. The blood vessel can be seen in a higher power view of the HRP labeled cells shown in d, also indicated with arrows. The patch of cells resides predominantly in layer I11 and can be seen to overlap with the patch of 2DG label. Scale bar a-c = 400 pm; d = 150 pm.

36

S.L. JULIANO ET AL. TABLE 4. Coincidenceof HRP and 2DG in Area 1

2DG

-

Area 3b injection

Area 1injection

Control injection

(n = 2)

(n = 2)

(n = 1)

21 30 15

12 28 14

3 0 17

HRP + no 2DG HRP + 2DG 2DG + no HRP

-

Area 36 injection compared with amo I injection, distribution in area 1alone Chi square analysis for the total distribution of HRP and 2DG in area 1 alone, comparing the label resulting from the area 3h injection with that resulting from the area 1injectLon

Expected frequencies HRP + no 2DG HRP + 2DG 2DG + no HEW Chi square (95%) = 5.991 Chi square d c . = 1.372

-

Area 3b injection

Area 1inwion

Totals

18.15000 31.90000 15.95000

14.85000 26.10000 13.05000

33 5n 29

-

2 D.F.

Not significant

-

Area 1 injection compared with contml injection, distrihution in area 1 alone Chi square analysis for the total distribution of the HRP and 2DG in area 1alone, mmparing the label resultingfrom the area 1injection with that resultingfrom the control injection

Expected frequencies

-

Control HRP + no 2DG HRP + 2DG 2DG + no HRP Chi square (95%) = 5.991 Chisouarecalc. = 22.904

Area 1injection

injection

Totids

10.94595 20.43243 22.62162

4.054054 7.567568 8.378378

15

-

28

31

2 D.F.

Sienificant.P< 0.005

the likelihood of no overlap. Alternatively one might select sites that are closer together, or one might select different stimulus parameters on an identical site. It is not clear, however, what a small degree of overlap would indicate as peripheral receptive fields are moved into closer proximity, and it is technically not feasible to activate two distinct receptor populations within a single receptive field. The chi-square analysis also supports our conclusion that the coincidence of HRP and 2DG is not a random event. The observation that the distributions of coincident patches of label in area 1did not differ significantly after either an area 1injection or an area 3b injection suggests that projections to the middle layers of area 1from either area 3b or area 1 carry the same kind of information. Despite the high degree of correspondence between the two forms of label, a number of instances of mismatch occur. We cannot rule out the possibility that an unknown principle of cortical organization governs the instances of poor correspondence between 2DG and HRP. It may be that a specific population of neurons responding to a given stimulus is interconnected, while a second neuronal population activated by the same stimulus is not connected to the first group of cells. The interconnected group may Fig. 9. Illustration of a digitized 2DG autoradiograph (top) and adjacent section reacted for HRP (bottom). The HRP label results represent one aspect of the cortical representation of from an injection in area 1, occurs predominantly as anterograde stimuli, such as place on the skin surface, whereas the transport in layers I and 11, and is located in a reciprocal relationship to second metabolically activated population may represent ia the 2DG activity. Arrowheads indicate the location of the HRP label. Cortical laminae are indicated with Roman numerals. These sections different attribute of cortical responsivity such as a feature of modality (e.g., directionality, velocity). The answer to are photographed almost entirely through area 1; the boundaries with areas 3b and 2 are at the borders of the photograph. Rostral is t o the these questions will be clarified by further experiments right. Scale bar = 400 km. involving anatomical tracers and functional correlates. There is also a question as to whether or not the increased metabolic uptake observed in the somatosensory neurons, which do not respond to the stimulus applied cortex in fact represents excitation. Although evidence during a 2DG experiment, do not overlap with metabolic exists that under certain circumstances 2DG label may activity evoked by that stimulus. Nevertheless, the determi- reflect inhibition (e.g., Ackerman et al., '84),we propose nation of an appropriate control in this case is problematic. that in sensory regions of the neocortex, data support the We chose stimulation sites on adjacent digits to maximize idea that metabolic patches reflect excitation (Juliano et al. I

37

2DG AND HRP IN SOMATOSENSORY CORTEX

2 D G STIMULATION SITE \ \

\

\ \

\

\ \

\ \

\ \

cs

med

3b

lmm Fig. 10. Two-dimensional reconstruction of 2DG (lines) and HRP (shading) label following a control injection into a cortical site not activated by the stimulus used during the 2DG experiment. The injection site is outlined with a heavy line. Conventions are as for Fig. 3. The skin region stimulated during the 2DG experiment is indicated

with an open arrow. The peripheral receptive field that activated neurons at the injection site is indicated with an asterisk. This figure demonstraies that no overlap occurs between the two forms of label during these experimental conditions.

'87; Tootell et al., '88). This is true for studies evaluating relations between 2DG uptake and neuronal activity in somatosensory and visual cortex (Juliano et al., '87; Schoppmann and Stryker, '81) and is also the interpretation obtained from studies of detailed topographic, chromatic, and binocular relations in visual cortex (Tootell et al., '88).

Using cross correlation techniques in cat visual cortex, T'so et al. ('861, found numerous connections between columns of cells sharing a similar orientation preference. The horizontal interactions between neurons appeared to be usually excitatory, although weak inhibitory interactions may be difficult to detect using cross correlation (Aertsen and Gerstein, '85). Studies using cytochrome oxidase as a measure of metabolic activity in visual cortex indicate that neurons located within blobs of high cytochrome oxidase staining project to other blob regions, and neurons located in the interblob regions project to interblob cortical regions (Livingstone and Hubel, '84). T'so and Gilbert ('88) also found, using cross correlation techniques, that blobs with similar color opponency are connected in monkey area 17. More recently, in a combined 2DG and retrograde tracing study, Gilbert and Wiesel('89) found that interconnected regions of visual cortex overlap with regions metabolically activated by stimuli of the same orientation that excited neurons at the injection site. Electron microscopic evidence in the visual cortex also suggests that the majority of long-range horizontal connections are excitatory (McGuire et al., '85). Results from a study of cat auditory cortex demonstrate that in AI, clusters of cells link other cells responding to specific frequencies (Matsubara and Phillips, '88). Thus evidence from the present study and from studies of other

Anatomic and physiologicmrrelah to

metabolic patches

In a study investigating the detail of corticocortical connections in somatosensory cortex, DeFelipe et al. ('86) made HRP injections that filled single pyramidal cells in the somatosensory cortex. They found long-range horizontal connections and suggested that the focused terminal arborizations identified in areas 3b, 1, and 2 are the morphologic basis for cortical columns. DeFelipe et al. ('86) also proposed that the horizontal connections are likely to be excitatory, and form a major route for information flow in somatosensory cortex, both between and within the cytoarchitectonic fields of the anterior parietal cortex. These long-range horizontal projections may provide the structure by which somatic information is relayed across fields resulting in the patchy pattern of metabolic label. A number of other studies in the visual and auditory cortex also postulate that horizontal connections within and between cytoarchitectonic fields are largely excitatory and transmit information of similar functional properties.

38

S.L. JULIANO ET AL.

sensory cortical regions indicate that long-range corticocortical projections are often excitatory and that the targets of projecting neurons respond to similar stimuli. These findings are in apparent discrepancy to the report of Matsubara et al. (’871, which indicated that in cat area 18 clusters of cells link cortical regions with orthogonal, rather than similar, orientation preferences. Although this study has not been replicated, it is possible that the connections identified by Matsubara et al. (’87) are akin to the type of projection found here from area 1 to area 3b, which were not usually correlated with increased 2DG label after the appropriate stimulation. A potential mechanism for this mismatch is discussed below. It is also possible that the functional relations between corticocortical connections in cat area 18 and those in monkey somatosensory cortex are not similar. In rare instances a reciprocal relationship occurred between HRP label in layers I and 11, and patches of 2DG activity in adjacent locations, after an injection in area 1 (Fig. 9). This finding holds open the possibility that a layer 1-11 projection may fill in the cortical regions in between the patches of metabolic label. Although such a distribution was rare in this study, its presence may help to explain earlier findings. A previous study demonstrated that the topical application of an antagonist of GABA during a 2DG experiment could “fill in” the cortical regions in between the densely labeled patches (Juliano et al., ’89).If the layer 1-11 termination occurs “in between” patches, it may participate in the suppression of activity in the patch-like region below it, via a direct or indirect GABAergic mechanism. The removal of such a mechanism, through the use of an antagonist of GABA, might result in the relatively homogeneous 2DG pattern that filled in the low activity regions seen in cat somatosensory cortex following such topical application experiments. If the corticocortical projections play a significant role in the transfer of information between cortical fields, we may ask, what does the thalamus do? The thalamus clearly projects on all fields of primary somatosensory cortex (see Jones, ’85, for review). Each field processes information that consists of different modality preferences, as well as increasing levels of complexity. It may be that discrete features of modality are conveyed through the ventrobasal nucleus of the thalamus, whereas the ability to integrate different aspects of sensation are relayed through the cortex. Experiments that injected anterograde tracers into the thalamus might also reveal projections coincident with 2DG activity, given a similar experimental design. This is a question that remains to be investigated and may help to identify the role of thalamus versus the role of corticocortical input in the processing of information in the fields of the anterior parietal cortex.

CONCLUSIONS The results from the present study lead us to the following observations about the functional organization of somatosensory cortex. 1. Projections from area 3b to area 1, and intrinsically within area 3b, appear to coincide with evoked metabolic activity and to connect clusters of cells with similar response properties; both of these projections terminate in layers 11-IV. 2. The projections originating from area 1 are less clear-cut. In general, projections from area 1to area 3b do

not appear to coincide with metabolic activity, even where the projection terminates in layer IV. Intrinsic projections within area 1, like those within area 3b, connect clusters of cells with apparently similar response properties. 3. Occasional projections found in layers I and I1 of areas 1 or 3b, after area 1 or 3b injections, do not appear to coincide with evoked metabolic activity. Although they are located in proximity to, they do not directly overlap with evoked metabolic activity. We do not have enough information to draw conclusions about area 2. Our data are most definitive for the projections from area 3b to 1. We cannot assign a function to these projections with certainty, although there appears to be a facilitatory flow of information from area 3b to area 1. The results of similar experiments within area 17 of the visual cortex suggest that excitatory corticocortical connections are not linking cells that normally respond to the same point in space, but rather link cells responding to a specific feature or features of stimulation, such as orientation, direction, and possibly ocular dominance (Gilbert and Wiesel, ’89). In the anterior parietal cortex, however, it seems likely that the putative excitatory connections between cytoarchitectonic fields that colocalize with stimulus-evoked metabolic label are connecting homotypic body parts represented in each field. These connections may therefore facilitate interactions between neurons in different fields of anterior parietal cortex that respond to stimulation of the same body part. Conversely, the connections between the fields that do not overlap evoked metabolic label (i.e., the area 1 to 3h projections) may represent a “feedback’ type of projection that is modulatory in nature and not connecting homotypic body parts. Perhaps the connections identified in visual cortex by Matsubara et al. (’87) that did not align with cortical locations responding to the same orientation are analogous to the “backward” projections of this study that did not match the evoked metabolic activity. It will be important to consider if these “hard-wired” relationships between structure and function can be altered by experimental conditions that disrupt specific input to the somatosensory cortex or during the development of the animal.

ACKNOWLEDGMENTS This work was supported by PHS grant NS24014 (SU) and the Laboratory of Neuropsychology, NIMH. The authors are grateful to Dr. Mort Mishkin for advice and support during the performance and analysis of these experiments. We also thank Drs. Leslie Ungerleider, Tim Pons, and Rosemary Borke for instructive comments on the manuscript.

Ackerman, R.F., D.M. Finch, T.L. Babb, and J. Engel, J r. (1984) Increased glucose metabolism during long-duration recurrent inhibition of hippocampal pyramidal cells. J. Neurosci. 4251-264. Aertsen, A.M.H.J.,and G.L. Gerstein (1985) Evaluation of neuronal connectivity: Sensitivity of cross-correlation. Brain Res. 340:341-354. DeFelipe, J.,M. Conley, and E.G. Jones (1986) Long-range focal collateralization of axons arising from corticocortical cells in monkey sensory-motor cortex. J. Neurosci. 6:3749-3766. Dykes, R.W., and A. Gabor (1981) Magnification functions and receptive field sequences for submodality-specific bands in SI cortex of cats. J. Comp. Neurol. 202597-620. Favorov, O., and B.L. Whitsel(1988) Spatial organization of the peripheral input to area 1 cell columns. I. The detection of ‘segregates.’ Brain Res Rev. 13t25-42.

2DG AND HRP IN SOMATOSENSORY CORTEX Friedman, D.P., and E.G. Jones (1980) Focal projections of electrophysiologically defined groupings of thalamic cells on the monkey somatic sensory cortex. Brain Res. 191t249-252. Gibson, A.R., D.I. Hansma, J.C. Houk, and F.R. Robinson (1984) A sensitive low artifact TMB procedure for the demonstration of WGA-HRP in the CNS. Brain Res. 2983235-241. Gilbert, C.D., and T.N. Wiesel (1989) Columnar specificity of intrinsic horizontal and cortico-cortical connections in cat visual cortex. J. Neurosci. 9t2432-2442. Jones, E.G. (1985) The Thalamus. New York: Plenum Press. Jones, E.G. (1986) Connectivity of the primate sensory-motor cortex. In E.G. Jones and A. Peters (eds.): Cerebral Cortex, vol. 5. New York: Plenum Press, pp. 113-183. Jones, E.G., D.P. Friedman, and S.H.C. Hendry (1982) Thalamic basis of place- and modality-specific columns in monkey somatosensory cortex: A correlative anatomical and physiological study. J. Neurophysiol. 48t545568. Juliano, S.L., and B.L. Whitsel (1987) A combined 2-deoxyglucose and neurophysiological study of primate somatosensory cortex. J. Comp. Neurol. 263:514-525. Juliano, S.L., P.J. Hand, and B.L. Whitsel (1981) Patterns of increased metabolic activity in somatosensory cortex of monkeys (Macacu fasczculurk-) subjected t.o controlled cutaneous stimulation: A 2-deoxyglucose study. J. Neurophysiol. 46t1260-1284. Juliano, S.L., D.P. Friedman, and D. Eslin (1987) Patterns of cortico-cortical connectivity can predict patches of stimulus-evoked metabolic activity in Neurosci. Abstr. 133470. monkey somatosensory cortex. SOC. Juliano, S.L., B.L. Whitsel, M. Tommerdahl, and S.S. Cheema (1989) Determinants of patchy metabolic labeling in the somatosensory cortex of cats: A possible role for intrinsic inhibitory circuitry. J. Neurosci. 9:l-12. Livingstone, M.S., and D.H. Hubel (1984) Specificity of intrinsic connections in primate primary visual cortex. J. Neurosci. 4:2830-2835. Matsubara, J.A., and D.P. Phillips (1988) Intracortical connections and their physiological correlates in the primary auditory cortex (AI)of the cat. J. Comp. Neurol. 268t38-48. Matsubara, J.A., M.S. Cynader, and N.V. Swindale (1987) Anatomical properties and physiological correlates of the intrinsic connections in cat area 18.J. Neurosci. 7tl428-1446.

39 McGuire, B.A., C.D. Gilbert, and T.N. Wiesel(1985) Ultrastructural characterization of long-range horizontal connections in monkey striate cortex. SOC. Neurosci. Abstr. llt17. McKenna, T.M., B.L. Whitsel, and D.A. Dreyer (1982) Anterior parietal cortical topographic organization in macaque monkey: A re-evaluation. J. Neurophysiol. 48t289-317. Mesulam, M.-M. (1978) Tetramethyl benzidine for horseradish peroxidase neurochemistry: A non-carcinogenic blue reaction product with superior sensitivity for visualizing neural afTerents and efferents. J. Histochem. Cytochem. 263160-177. Mountcastle, V.B. (1957) Modality and topographic properties of single neurons of cat’s somatic sensory cortex. J. Neurophysiol. 20t408-434. Nelson, R.J., M. Sur, D.J. Felleman, and J.H. Kaas (1980) Representations of the body surface in postcentral and parietal cortex of Mucucu fmczcularzs. J.Comp. Neurol. 192t611-643. Schoppmann A,, and M.P. Stryker (1981) Physiological evidence that the 2-deoxyglucose method reveals orientation columns in cat visual cortex. Nature 293t574-578. Shapiro, H.M., J.H. Greenberg, M. Reivich, G. Ashmead, and L. Sokoloff (1978) Local cerebral glucose uptake in awake and halothane-anesthetized primates. Anesthesiology 48397-103, Sur, M., M.M. Merzenich, and J.H. Kaas (1980) Magnification, receptive field area, and hypercolumn size in areas 3b and 1of somatosensory cortex in owl monkeys. J. Neurophysiol. 44:295-311. Sur, M., J.T. Wall, and J.H. Kaas (1984) Modular distribution of neurons with slowly adapting and rapidly adapting responses in area 3b of somatosensory cortex in monkeys. J. Neurophysiol. 513726744. Tommerdahl, M., R. Baker, B.L. Whitsel, and S.L.Juliano (1985)A method for reconstructing patterns of somatosensory cerebral cortical activity. Biomed. Sci. Instrum. 21:93-98. Tootell, R.B.H., S.L. Hamilton, M.S. Silverman, and E. Switkes (1988) Functional anatomy of macaque striate cortex. I. Ocular dominance, binocular interactions and baseline conditions. J. Neurosci. 8:15001530. T’so, D.Y., and C.D. Gilbert (1988)The organization of chromatic and spatial interactions in the primate striate cortex. J. Neurosci. 8;1712-1727. T’so, D.Y., C.D. Gilbert, and T.N. Wiesel (1986) Relationships between horizontal interactions and functional architecture in cat striate cortex as revealed by cross-correlation analysis. J. Neurosci. 6: 1160-1170.