A micromodularity specific to the uppermost cortical layers, layers 1 and 2, is demonstrated in primates. This is most pronounced as patches of zinc-positive (Zn+) terminations, preferentially in the pre-Rolandic and limbic areas. The upper layer modularity can frequently be demonstrated by parvalbumin-immunoreactive (PV-ir) GABAergic terminations and by bundles of apical dendrites. Double-labeling or alternate section analysis shows that PV-ir and Zn+ terminations co-mingle at the layer 1,2 border and appear to coincide with dendritic bundles, proposed to originate from layer 2 pyramidal neurons. This model is basically similar to the prominent wall-and-hollow honeycomb organization in rat visual cortex. The organization of the PV-ir and dendritic components, however, is more difficult to define in primate than in rat. Moreover, the micromodularity is not uniform across areas. In some areas (motor and limbic), the modularity can be visualized by both zinc and PV. In other areas (i.e. primary sensory, sensory associational and prefrontal areas 46 and 8), although PV immunohistochemisty shows a periodic distribution, there is no detectable Zn+ modularity. These results add to the evidence for the complexity of layers 1 and 2 and raise the possibility that patches of Zn+ terminations correspond to zones of area-specific zinc-related plasticity. This might figure in the context of top-down or feedback influences, as often associated with layer 1.
The most superficial cortical layer, layer 1, has long been both under-investigated and puzzling (‘mysterious’; Mountcastle, 2003). This may be in part because of its thinness and location immediately beneath the pia surface, which has made it paradoxically inaccessible to conventional physiological techniques. Recently, we reported an unsuspected stratification and modularity in layers 1 and 2. The main components, investigated in detail in rat visual cortex (Ichinohe et al., 2003b), consist of a complex organization of different afferent and dendritic systems. In layer 1a, a uniform zone of thalamocortical (TC) terminations can be visualized by immunohistochemistry for vesicular glutamate transporter 2 (VGluT2); but this gives way in layer 1b to interdigitated patches of TC and zinc-enriched (Zn+) corticocortical terminations. In the upper part of layer 2, the Zn+ terminations co-mingle with parvalbumin-immunoreactive (PV-ir) GABAergic terminations and form the walls of a small-scale wall-and-hollow honeycomb mosaic. In layers 1b and 2, there is also a mix of apical dendritic components. These relate in an orderly way to the honeycomb mosaic. That is, bundles of apical dendrites deriving from deeper pyramidal neurons are concentrated in the honeycomb hollows, where they are likely to receive dense TC inputs. Those deriving from layer 2 pyramidal neurons, in contrast, are located preferentially in the walls and are hypothesized to form a specialized complex with Zn+ and PV-ir terminations.
Preliminary screening with PV revealed a honeycomb configuration in several other areas in rat, as well as in several areas in cat and monkey (Ichinohe et al., 2003b). As a next step, it seemed important to assess more widely the specific features of this organization. We have accordingly carried out more extensive analysis of cortical areas in the macaque. This further investigation has confirmed the finding of a modularity of the uppermost layers. In particular, histochemistry for zinc reliably demonstrates a small-scale upper layer modularity that is pronounced in the pre-Rolandic motor and limbic areas. Although our first impression with PV had suggested a honeycomb pattern, the modularity is not uniform. Rather, it shows a high degree of regional variability in both size and shape, ranging from small-scale honeycomb or reticulum, to larger-scale patches.
The upper layer modularity can also be demonstrated by immunohistochemistry for PV or for dendritic bundles labeled by microtubule-associated protein 2 (MAP2), but this tends to be more subtle (see ‘Single and Double Labeling’ in Materials and Methods). PV modularity is located more at the border between layers 1 and 2, while MAP2 labeling extends into layer 1. In areas of larger scale modules (such as the orbitofrontal and parahippocampal), the Zn+, PV-ir and MAP2-ir densities can be shown to co-mingle, as in the rat area V1. PV-ir modularity may also occur independently of Zn+ modularity.
In this report, we concentrate on four regions with pronounced modularity: parahippocampal, orbitofrontal, rostral cingulate and motor and dorsolateral prefrontal regions. The main result concerns zinc modularity. This is followed by sections on dendritic bundles, cell clusters in layer 2 and PV modularity.
Materials and Methods
Twelve adult macaque monkeys (Macaca mulatta and M. fuscata) were used in this study. All experimental protocols were approved by the Experimental Animal Committee of the RIKEN Institute and were carried out in accordance with the guidelines published in the NIH Guide for the Care and Use of Laboratory Animals (NIH publication No. 86-23, revised 1987).
Fixation and Tissue Preparation
Animals were anesthetized with ketamine (11 mg/kg, i.m.) and Nembutal (overdose, 75 mg/kg, i.p.). Six animals used for zinc histochemistry were perfused transcardially, in sequence, with saline containing 0.1% sodium sulfide for 5 min and then 0.1% sodium sulfide and 4% paraformaldehyde in 0.1 M phosphate buffer (PB, pH 7.4) for 30 min. The brains were removed from the skulls, trimmed and postfixed for 12–15 h in 0.1 M PB containing 4% paraformaldehyde. Then, brains were immersed into 30% sucrose in 0.1 M PB. Another six animals, used only for immunohistochemistry, were perfused transcardially, in sequence, with saline containing 0.5% sodium nitrite; 4% paraformaldehyde in 0.1 M PB for 30 min; and chilled 0.1 M PB with 10, 20 and 30% sucrose. The brains were removed from the skulls and were immersed into 30% sucrose in 0.1 M PB. After the brains sank (for both series), some blocks were frozen with dry ice for future processing. The remaining blocks were cut serially in either the coronal or tangential plane by frozen microtomy (at 40–50 µm thickness).
Single and Double Labeling
In our preceeding rodent study (Ichinohe et al., 2003b), in order to visualize micromodularity at the border of layers 1 and 2, we used zinc, PV, MAP2 (for dendrites) and GABA receptor type A α1 subunit (GABAaα1; also for dendrites), VGluT2 (for TC terminations), glutamate receptor 2 and 3 (GluR2/3), glutamate receptor 5, 6 and 7 (GluR5/6/7), NMDAR1, calbindin (CB) and cytochrome oxidase. The present results are mainly based on the distribution pattern of zinc, PV, MAP2 and GABAaα1. In the monkey, VGluT2 has been less successful than in the rodent, where a good correspondence has been shown for the antibody and TC terminals by both electron microscopic and thalamic lesion studies (Fujiyama et al., 2001). In monkey, GluR2/3, GluR5/6/7 and NMDAR1 tended to stain only cell bodies and thick apical dendrites, with weak staining of neuropil. Calbindin is a marker for both pyramidal neurons and GABAergic interneurons in the monkey and rat; but in the monkey, CB-ir pyramidal neurons in the upper layers and their apical dendritic components are much fewer than in rat. More particularly, preliminary screening of CB and calretinin in the motor and parahippocampal areas did not show any discernible pattern at the border between layers 1 and 2 and the results are therefore not further described. Finally, cytochrome oxidase was not effective for examining the most superficial layers in the monkey, perhaps because of the frequency of edge artifact.
In rat V1, double immunofluorescence, with PV as one of the labels, was particularly useful in ascertaining spatial relationships of overlap or interdigitation. In monkey, however, labeling of PV-ir terminals was better achieved by the enhanced DAB method, rather than by immunofluorecence. In the present study, we therefore used a double labeling method combining zinc histochemistry and immunofluorescence for MAP2 or GABAaα1 (see below), but not PV. For analysis of PV-ir terminal distribution, we employed alternate tangential sections stained by the immunoperoxidase technique and compared PV and MAP2 (see below). This approach was not used for direct comparison of PV and zinc distributions because the perfusion method needed for zinc histochemistry (with sodium sulfide) weakened the PV-ir intensity
Sections perfused with a solution containing sodium sulfide were washed thoroughly with 0.1 M PB, followed by 0.01M PB. The IntenSE M silver Enhancement kit (Amersham International, Little Chalfont, Bucks, UK) was used to intensify zinc signals (Danscher et al., 1987; De Biasi and Bendotti, 1988; Akagi et al., 2001). A one-to-one cocktail of the IntenSE M kit solution and 33% gum arabic solution was used as a reagent. Development of reaction products was checked under a microscope and terminated by rinsing the sections in 0.01 M PB and, subsequently, several times in 0.1 M PB. Staining intensity was carefully controlled by periodic visual inspection under a low power microscope to avoid either over- or under-staining, which might obscure any modularity. Selected sections were further processed for Nissl substrate using NeuroTrace 500/525 green fluorescent Nissl stain (Molecular Probes, Eugene, OR) according to the company’s protocol. Other sections were further processed for double labeling with immunohistochemistry for MAP2 or GABAaα1 (see below).
Double Labeling Combining Zinc Histochemistry and Immunohistofluorescence for MAP2 and GABAaα1
After finishing the silver intensification, sections were immunoblocked with 0.1 M phosphate buffered saline (PBS, pH 7.3) containing 0.5% Triton X-100 and 5% normal goat serum (PBS-TG) for 1 h at room temperature and subsequently incubated for 40–48 h at 4°C in PBS-TG containing anti-MAP2 monoclonal mouse antibody (1:2000; Chemicon, Temecula, CA) or anti-GABAaα1 polyclonal rabbit antibody (1:5000; Chemicon, Temecular, CA). After rinsing, the sections were placed in PBS-TG containing Alexa Fluor 488 conjugated anti-mouse IgG polyclonal goat antibody (1:200; Molecular Probes, Eugene, OR) or Alexa Fluor 488 conjugated anti-rabbit IgG polyclonal goat antibody (1:200; Molecular Probes, Eugene, OR) for 1.5 h.
Immunoperoxidase Staining for MAP2 and PV
Sections were incubated for 1 h with PBS-TG at room temperature and then 40–48 h at 4°C with PBS-TG containing anti-MAP2 monoclonal mouse antibody (1:8000; Chemicon; Temecula, CA) or anti-PV monoclonal mouse antibody (1:50 000; Swant, Bellinzona, Switzerland). After rinsing, the sections were placed in PBS-TG containing biotinylated anti-mouse IgG polyclonal goat antibody (1:200; Vector, Burlingame, CA) for 1.5 h at room temperature. Immunoreactivity was visualized by ABC incubation (one drop of reagents per 7 ml 0.1 M PB; ABC Elite kits; Vector, Burlingame, CA) followed by diaminobenzidine histochemistry with 0.03% nickel ammonium sulfate. In three of six animals, MAP2 immunohistochemistry showed clear dendritic aggregation in layers 1 and 2 with very low interbundle background staining. In one monkey, dendritic clusters were detectable in layers 1 and 2, but the interbundle spaces also had background-like staining. The remaining two animals had weaker MAP2 staining in layers 1 and 2.
In order to visualize PV-ir neurons together with Nissl-stained neurons, we reacted some tissue by double fluorescent labeling for PV and Nissl substrate. Sections were immunoblocked with 0.1 M phosphate buffered saline (PBS, pH 7.3) containing 0.5% Triton X-100 and 5% normal goat serum (PBS-TG) for 1 h at room temperature and subsequently incubated 40–48 h at 4°C in PBS-TG containing anti-PV monoclonal mouse antibody (1:5000; Swant, Bellinzona, Switzerland). After rinsing, the sections were placed in PBS-TG containing Alexa Fluor 594 conjugated anti-mouse IgG polyclonal goat antibody (1:200; Molecular Probes, Eugene, OR) and NeuroTrace 500/525 green fluorescent Nissl stain (Molecular Probes, Eugene, OR) according to the company’s protocol for 1.5 h. Immunofluorescence for PV was ordinarily much weaker in our hands than the immunoperoxidase technique (however, see Fig. 5C).
Measurements and Analysis
In the first step of this investigation, we scanned a through-brain series of coronal sections (spaced at 250 µm) stained for zinc to ascertain the areas with zinc modularity. Four regions were chosen for more detailed analysis based on the occurrence of upper layer modularity (i.e. parahippocampal region, orbitofrontal region, rostral cingulate region and motor and dorsolateral prefrontal areas; Fig. 2).
Quantitative analysis of Zn+ patches was achieved with the aid of a Neurolucida System (MicroBrightField, Colchester, VT). First, multiple tangential blocks were prepared through architectonic areas of regions I–IV of Figure 2. For parahippocampal and orbitofrontal regions, two or three blocks were made and six or seven for the medial and dorsolateral regions (III and IV of Fig. 2). Secondly, the centers of Zn+ patches were identified and plotted in an area of 0.36–1.49 mm2, using a ×20 objective lens. The size of the area examined was constrained by the extent of flatness within the region. This was frequently too small to allow for random sampling with a grid. Therefore, rather than employ an unbiased sampling procedure, we instead measured all the available flat area from the four regions of interest from two animals. Then, we obtained the mean value of these stacked data. The nearest center-to-center distance (CCD) was obtained by using NeuroExplorer analysis software (MicroBrightField, Colchester, VT). In regions of large patch size, averages were based on 20–30 CCDs; in regions of smaller patch size, on 40–70. To assure that change in size is real and is not influenced by a variation in the plane of section, we were careful to scan through several serial sections (five sections on average).
Light and fluorescent images were photographed with an Olympus DP50 digital camera mounted on an Olympus BX60 microscope with an appropriate filter for green fluorescence (BA515IF; Olympus, Tokyo, Japan) and with a Zeiss LSM 5 Pascal confocal microscope (Jena, Germany).
Cortical areas were identified by reference to sulcal landmarks, by comparison with published maps and by architectonic analysis of selected histological sections stained for cell bodies. Where several nomenclatures are current, we usually adopted those using broader categories. We followed most closely: for prefrontal cortex, the nomenclature proposed by Barbas and Pandya (1989), with reference to that of Preuss and Goldman-Rakic (1991) and Carmichael and Price (1994); for motor cortices, that of Brodmann (1909), with reference to that of Matelli et al. (1985, 1991) and Dum and Strick (1991); for the parahippocampal and inferotemporal cortex, that of Yukie et al. (1990) and Saleem and Tanaka (1996), with reference to that of Suzuki and Amaral (2003); for visual cortices, that of Felleman and Van Essen (1991); and for parietal and cingulate cortex, that of Brodmann (1909).
In referring to zinc-enriched terminals and modular patches, we have for convenience used the shorter designation Zn+ (zinc-positive).
Modularity in the uppermost cortical stratum can be demonstrated in several areas by different marker substances, but the organization is complex and characterized by areal variability. We report the main features of this modularity as seen by zinc histochemistry, because this is so far the most reliable marker in primate. We concentrate on four regions with pronounced Zn+ modularity: parahippocampal, orbitofrontal, rostral cingulate and motor and dorsolateral prefrontal regions. We also report the modular distribution of MAP2 (for thicker dendrites), GABAaα1 (for thinner dendrites), Nissl (for cell bodies) and PV (showing terminals, as well as dendrites and cell bodies of some GABAergic neurons).
The demonstration of vesicular zinc requires a modified perfusion procedure, involving sodium sulfide; and relatively few studies have so far been carried out on the distribution of Zn+ terminations in monkey cortical regions. In terms of overall density, we found considerable regional variation. As a trend, zinc density is higher in agranular or dysgranular, and limbic cortices, including perirhinal and parahippocampal cortices; the temporal pole; agranular insular, rostral cingulate and caudal orbitofrontal cortices. There are generally two dense bands of Zn+ terminations, in the superficial (layers 1b, 2 and upper layer 3) and deep layers (layers 5 and 6). The middle layers (layer 4 and adjacent layer 3) and layer 1a have lower levels of zinc (see also Carmichael and Price, 1994; Franco-Pons et al., 2000). The detailed laminar distribution varies in relation to overall zinc density. In areas where zinc density is low [e.g. primary motor cortex (area 4), Fig. 1G], the infragranular band tends to be less dense in layer 5 than in layer 6, but the reverse occurs in zinc dense areas (e.g. parahippocampal cortex, Fig. 1E; rostral cingulate cortex, Fig. 1J). In zinc weak areas, the supragranular band in layer 3 is narrower and remains confined to the upper part of this layer.
Inspection of coronal and tangential sections reveals a region specific trend in both distinctness and size of Zn+ modularity at the border of layers 1 and 2. In some areas, this superficial Zn+ modularity is conspicuous and can easily be recognized in coronal as well as tangential sections [e.g. areas 4 and 6, caudal orbitofrontal cortex and parahippocampal area, just caudal to the entorhinal cortex (EC); see Fig. 1A–I, and yellow shading in Fig. 2]. Other areas have a more subtle modularity, which is obvious only in tangential sections (e.g. rostral cingulate cortex, Fig. 1J,K: gray shading in Fig. 2). In ‘conspicuous’ regions, one module can be identified through 2–3 serial sections (80–120 µm thickness), but in less conspicuous regions, a given module is usually limited to only a single section (40 µm thickness, with commensurate width). Finally, some areas, especially those with a cell-dense layer 4 (‘eulaminate’), do not exhibit modularity in the uppermost layers. These include most of the dorsolateral post-Rolandic cortex (primary and paraprimary sensory areas) and dorsolateral prefrontal areas, adjacent to the principal sulcus (areas 46 and 8; white regions in Fig. 2). Areas 46 and 8 thus form a small zone without Zn+ modularity, surrounded by regions with modularity (but see ‘PV’ section below).
In coronal sections, the patches of Zn+ terminations often appear dome-like, rising into layer 1b from a base in layer 2 (Fig. 1A,E). In areas of lesser density, they tend to be thinner and more delicate (Fig. 1B,G). In the tangential plane, the larger patches form a conspicuous grid in layer 1b (Fig. 1D,F), while the thinner pattern is rather chevron, reticular, or ‘honeycomb’ in shape (Fig. 1C,H,I,K). It is worth emphasizing that the details of shape are complicated. That is, even a clearly patchy pattern frequently looks reticular or honeycomb at the lower border of the patches, as these merge into a more continuous distribution.
Size differences in Zn+ modules are easily recognized in tangential sections (measured as CCD of Zn+ patches; see ‘Measurements and Analysis’ in Materials and Methods). Four regions are distinguishable on the basis of what appears to be a coherent pattern of size gradation.
. In the parahippocampal region, at the border of caudal EC and medial parahippocampal cortex (area TH), the average CCD is 217 µm. Proceeding caudally, at the mid-portion of area TH (1 mm caudal to the EC), the average CCD decreases to 140 µm and still more caudally (2 mm from EC), the average CCD is 110 µm (Figs 1F and 2C). The caudalmost part of area TH (3 mm or more from EC) has high levels of zinc, but no distinguishable periodicity. A second, mediolateral size gradient is also apparent in the parahippocampal region, where medial patches (mean CCD = 198 µm, in area TH, just lateral to caudal EC) are larger than lateral patches (mean CCD = 108 µm, in area TF in the lateral parahippocampal region, 2 mm from the EC/TH border; see also Figs 1E and 2C). With zinc histochemistry, the EC itself, well-known to have conspicuous cell islands in layer 2, shows a complex subregion dependent modular organization and will be treated elsewhere in more detail.
. In the orbitofrontal region, there is a similar size gradient as in the parahippocampal region, but reversed, in that patch size decreases rostrally. That is, Zn+ patches are large (mean CCD = 300 µm) in the proisocortical zone of orbitofrontal cortex (area OPro), at the rostral edge of the lateral fissure; smaller (mean CCD = 180 µm) in area 13, 5 mm rostral from the lateral fissure and even smaller more rostrally (mean CCD = 102 µm), in the caudal part of area 11, located 10 mm rostral from the lateral fissure (Figs 1D and 2C). Size gradation is obvious in coronal sections through the caudal orbitofrontal cortex (Figs 1A and 2C) and, as in the parahippocampal region, more lateral patches are smaller.
. In the rostral cingulate gyrus, patches are less conspicuous (Fig. 1J; gray shading in Fig. 2E), but still easily detectable in tangential section (Fig. 1K). Patches are larger (mean CCD = 171 µm) surrounding the genu of the corpus callosum (area 24) and then decrease in size both caudally and rostrally. Caudally, the mean CCD = 150 µm at the area 24/23 border and rostrally, this is 110 µm, in area 32, 4 mm rostral to the genu of corpus callosum (Fig. 2B).
. For most of the motor-related areas (areas 4 and 6) and dorsolateral prefrontal cortex, there is modularity in the uppermost layers, but of a more uniform small scale (mean CCD = 108 – 124 µm; Figs 1A–C,G–I and 2A,B).
MAP-2 staining was investigated most closely for those four regions showing strong Zn+ modularity. In our previous investigation of rat visual cortex, Zn+ patches were found to interdigitate with MAP2-ir thick dendritic bundles from layer 3 and 5 pyramidal neurons, as these pass through layer 2. The patches co-localized, instead, with more weakly labeled dendrites, interpreted as originating from layer 2 pyramidal neurons. In the monkey cortex, in contrast with rat, we found that MAP2 does not strongly stain the distal portions of apical dendrites of layer 3 and 5 pyramidal neurons (see also Peters and Sethares, 1991a). Conversely, at least in some areas, MAP2 staining of apical dendrites of the layer 2 pyramidal neurons is stronger in monkey (Peters and Sethares, 1991a). According to our model, we would expect these to co-localize with Zn+ patches.
In coronal sections of both the caudal orbitofrontal cortex and the parahippocampal area just caudal to the EC, MAP2 immunohistochemistry demonstrates distinct dendritic bundles rising into layer 1 from layer 2. These have a predominantly oblique orientation, and can often be followed back to pyramidal cells in layer 2 (Fig. 3A–C,F). Double labeling for MAP2 and zinc shows that the MAP2 bundles in fact co-localize with zinc patches, except that they tend to extend slightly higher into layer 1a (Fig. 3D–G). In motor-related and dorsolateral prefrontal cortices, MAP2 also demonstrates dendritic patches and, in double-labeled sections, these can again be seen to co-localize with Zn+ terminations (Fig. 3H,I). However, in these areas the MAP2 modularity is weak, even in tangential sections, and often not detectable in coronal sections, so that MAP2-ir dendrites are more difficult to trace back to their cell bodies. For areas with subtle or no zinc pattern (such as area 8, rostral cingulate and somatosensory areas), we could not find evidence for MAP2-ir dendritic bundles in layer 1 (data are not shown).
Dendritic bundles in layer 1 exhibit size differences in parallel with the gradations of the Zn+ terminations (CCD of 100–300 µm; see Figs 1A and 3A). It is worth noting that not all dendrites can be labeled with MAP2 immunohistochemisty (Peters and Sethares, 1996). Strongly MAP2-ir dendrites, on which our observations are based, are a subpopulation.
In order to investigate the organization of dendrites from deeper neurons, we used GABAaα1, which frequently yields an image of vertically oriented, presumably dendritic processes, especially in layers 1b–upper 3 (Fig. 3J). In monkey visual cortex, Hendry et al. (1994) have reported that densely GABAaα1-ir dendrites are mainly thin dendrites and that these frequently form vertical bundles. Double labeling for GABAaα1 and zinc shows that bundles of thin GABAaα1-ir dendrites at the layer 1, 2 border preferentially avoid Zn+ patches and instead occupy the interpatch zones (Fig. 3J–M). These bundles can be traced back to the middle of layer 3, before blending into densely immunoreactive neuropil (areas examined were areas 4 and 6, and orbitofrontal and parahippocampal areas; for area 4, Fig. 3J–M). These may be the tapered distal ends of apical dendrites from neurons deeper in layer 3 and possibly layer 5. Again, there is a good match between the spacing of GABAaα1 bundles and the scale of Zn+ modularities at the border between layers 1 and 2.
Cell clusters in layer 2 and/or a cell-dense ‘accentuated’ layer 2 have been reported in several cortical areas in Nissl stained material of several species (e.g. Sanides, 1970; DeFelipe et al., 2002; Suzuki and Amaral, 2003). These might be considered as a basis for micromodularity, but the relationship between cell clusters and micromodularity is not strict. In some areas with large zinc modules and dendritic bundles (i.e. caudal orbitofrontal cortex and parahippocampal cortex, just caudal to the EC; Fig. 4A,B), cell clusters at a comparable spatial scale can be discerned in layer 2. In contrast, in the perirhinal cortex, lateral to the rostral part of the EC (gray shading in Fig. 2), there is a very conspicuous clustering of cell bodies in layer 2 (especially in area 36 and, somewhat less, in area 35; see also Suzuki and Amaral, 2003; Fig. 4C), but only a weak manifestation of MAP2-ir dendritic or Zn+ modularity. In other areas, including motor-related and cingulate cortices, there are clear zinc modules, but no obvious cell aggregation. Motor but not cingulate cortex, as described above, has demonstrable dendritic bundles in layer 1.
In the upper layers in the monkey cerebral cortex, PV is present in subtypes of GABAergic, aspiny, non-pyramidal cells and their terminations (Hendry et al., 1989; Blumcke et al., 1990; Williams et al., 1992; Melchitzky et al., 1999). PV immunohistochemistry in monkey showed a honeycomb-like modularity (Fig. 5A,B,G–I) or occasionally a larger, patch-like pattern (Fig. 5D). This was usually situated in a thin stratum at the border between layers 1 and 2, but the larger scale modularity dipped further into layer 2. PV-ir terminations, as in rat V1, are dense (Fig. 5C) and can be taken as a main source of the modularity; but the distinction between walls and hollows (Fig. 5B) does not seem as sharp in the monkey as in rat V1. Another important difference between rat and monkey relates to PV-ir dendrites from layers 2 and 3. In monkeys, robust vertically oriented dendritic arbors reach up to the middle of layer 1 or higher (Blumcke et al., 1990; Williams et al., 1992; Gabbott and Bacon, 1996). They frequently penetrate through PV-ir weak hollows as well as PV-ir dense walls and obscure or confound any periodicity. As a consequence, PV-ir periodicity can more easily be evaluated in areas with weakly PV-ir neuropil and fewer PV-ir cell bodies.
Thus, in the orbitofrontal and parahippocampal regions, where overall PV density is lower, PV-ir patchness is readily apparent (Fig. 5A,B,D for parahippocampal area). The PV modularity shows a distinct size gradation in these regions; that is, it progressively decreases in size laterally and either caudally (parahippocampal case) or rostrally (orbitofrontal case). This is in parallel with the change in scale in these same regions of the Zn+ and MAP2-ir patches (Fig. 5A, for parahippocampal area). Alternate section analysis showed that the PV-ir terminations co-localize, but at a slightly deeper level (i.e. more at the border between layers 1 and 2) with the MAP2-ir dendritic bundles (Fig. 5D,E). Since the MAP2-ir and Zn+ patches overlap in these areas, we suggest that the PV-ir pattern also overlaps with the Zn+ pattern (direct PV–zinc comparison was not undertaken; see ‘Single and Double Labeling’ in Materials and Methods).
In areas 4 and 6, the dorsolateral prefrontal area and the rostral cingulate, a PV-ir small scale pattern, more reminiscent of a honeycomb-like organization, is apparent (Fig. 5G). Center-to-center distance is ∼100 µm in areas 4 and 6 and in the dorsolateral prefrontal area. This is similar to the size of the Zn+ patches in these areas. In the rostral cingulate areas, size gradation is similar for PV-ir and zinc patches. The thinness and small size of both the PV-ir and MAP2-ir pattern in these areas again hindered direct comparison of spatial overlap.
Most of the post-Rolandic areas (e.g. primary auditory and somatosensory cortices), exhibited some degree of PV-ir honeycomb-like pattern (Fig. 5I; see also fig. 11D in Ichinohe et al., 2003b). Interestingly, prefrontal areas 8 and 46 also have this feature (Fig. 5H). The CCD of the honeycomb in these areas is again ∼100 µm. None of these areas, as reported above, exhibited any clear Zn+ modularity in layers 1 and 2.
In the parahippocampal region, just caudal to EC, we also identified a separate mosaic of PV-ir fibers, of unknown origin, in a thin stratum within layer 1 (border of layers 1a and b; Fig. 5F). The CCD was ∼200 µm. Comparison of alternate sections stained for PV or MAP2 showed that these PV-ir fibers surround clusters of MAP2 labeled dendrites at this level, in the middle of layer 1 (Fig. 5E,F). As this fiber mosaic is comparatively narrow and easily obscured by PV-ir dendrites and other dense PV-ir fiber systems in layer 1a, we do not know whether it is widespread in other areas.
These results provide further evidence for a general stratified and modular specialization specific to the uppermost cortical layers. As in the rat visual cortex, the anatomic components involve Zn+ terminations, PV-ir terminations and portions of distal apical dendrites. In areas where the modularity is conspicuous and large scale (i.e. caudal orbitofrontal and parahippocampal areas), the Zn+ patches can be verified to match well with MAP2-ir dendritic bundles in layer 1, thought to originate from pyramidal neurons in layer 2. In turn, by analysis of alternate sections, MAP2-ir bundles can be shown to co-localize with patches of PV-ir terminations at the border of layers 1 and 2. GABAaα1-ir dendritic bundles, presumed to be distal dendritic portions from deeper neurons, avoid Zn+ patches (see Fig. 6). This model is basically similar to the wall-and-hollow honeycomb organization demonstrated by several overlapping or complementary components in rat visual cortex (Ichinohe et al., 2003b).
The organization of the components is more difficult to define in primate than in rat, however; and the formulation of general principles is complicated by the fact that the modularity is not uniform across areas. For example, in some areas (motor and limbic areas), the modularity can be visualized by both zinc and PV. In other areas (i.e. primary and associational sensory post-Rolandic cortices and prefrontal areas 46 and 8), although PV immunohistochemisty shows a periodic distribution, there is no detectable Zn+ modularity. In monkey primary visual cortex, Zn+ patches have indeed been reported, but these are in layers 3 and 4A, complementary to cytochrome oxidase patches (Dyck and Cynader, 1993).
By way of comparison, areal specific differences occur in rat cortex as well. In the rat, areas with discernible upper layer Zn+ modules, in addition to V1, include V2, the caudal part of retrosplenial agranular cortex, and medial prefrontal cortex (fig. 2 in Ichinohe and Rockland, 2003a; fig. 1 in Mengual et al., 1995). Neither motor nor barrel cortices show modularity in zinc (unpublished observation), but do so with PV immunohistochemistry (Ichinohe et al., 2003b).
Areal Variability in Modular Size and Shape
One striking result in this study is the difference in size and shape of the upper layer modularity. The issue of shape is particularly complex, as this varies from an obviously patch-like to an obviously ‘honeycomb’ configuration. In areas with smaller-scale modularity (e.g. area 4), the module shape visualized by histochemistry for zinc, PV and MAP2 tends to resemble more a reticulum or wall-and-hollow honeycomb; but in areas with large-scale modularity (e.g. parahippocampal area), it tends to be more patchy. The ‘honeycomb’ itself can also be viewed as consisting of patches (‘hollows’) embedded in a thinner matrix. Thus, in the rat V1, the honeycomb could be described as ‘patches’ of TC connections surrounded by a thin matrix (‘wall’) of Zn+ and PV-ir terminations (Ichinohe et al., 2003b).
The issue of complementarity is better understood in the middle layers of primary sensory areas. In primate V1, the patch and matrix organization of layer 3 consists of CO ‘blobs’, corresponding to a subpopulation of TC connections, embedded in a matrix of Zn+ corticocortical terminations (Dyck and Cynader, 1993). A similar complementary pattern of Zn+ and TC terminals occurs in layer 4 of rodent somatosensory barrel cortex (Czupryn and Skangiel-Kramska, 1997). In layer 4A of monkey V1, a thin ‘wall’ of TC terminals surrounds zinc dense hollows (Dyck and Cynader, 1993). It would be helpful, in further interpreting our results, to know the distribution of a general marker for TC connections in the primate.
The difference in size is somewhat easier to see. For most areas, the mean CCD of Zn+ modularity is 110 µm. This is close to the size of honeycomb walls in rat cortex (mean CCD = 80 µm; Ichinohe et al., 2003b), as well as to that of the classic honeycomb in layer 4A of monkey V1 (mean CCD = 100 µm; Peters and Sethares, 1991b). In three regions, however, there is a marked size gradation; that is, in the orbitofrontal, parahippocampal and anterior cingulate regions. In the orbitofrontal zone, the change in size is particularly drastic, ranging from 100 to 300 µm over an expanse of 10 mm. For MAP2-ir dendritic bundles or PV patches, the same size differential is observed.
The factors underlying this variation in size and shape are not clear (but see next section). Other trendwise gradations within and across areas, however, have been frequently observed. Classic architectonic studies noted differences in the degree of development of cortical lamination and distinguished granular, dysgranular and agranular cortical types. These have been related to distance of particular areas from allocortex (Sanides, 1970; Barbas and Pandya, 1989; Dombrowski et al., 2001). Other differences have been observed in pyramidal dendritic architecture. Apical dendritic bundles are more prominent in some areas (Del Rio and DeFelipe, 1994) and the average size and complexity of basal dendritic arbors, as well as the degree of spininess, increase in prefrontal areas in comparison with early association or primary visual areas (for a review, see Elston, 2002). Gradient-like gene expression and protein distributions in developing and adult cortex have attracted increasing attention as influences on areal size and regionalization (Donoghue and Rakic, 1999; O’Leary and Nakagawa, 2002; Job and Tan, 2003). Of particular interest with regard to the uppermost layers, the monoclonal antibody 8B3, that recognizes a chondroitin sulfate proteoglycan, is expressed in a rostrocaudal gradient in monkey cortex. More specifically, 8B3 is concentrated in three bands, including a row of neurons at the border of layers 1 and 2 (Pimenta et al., 2001).
Composition of Dendritic Bundles
The issue of dendritic bundles is complex. Previous investigations have reported the occurrence of apical dendritic bundles in several neocortical areas and further suggested that these may be a general feature of cortical organization (e.g. Fleischhauer et al., 1972; Peters and Sethares, 1991a; Mountcastle, 1997). These investigations have also shown that apical dendrites of layer 2 pyramidal neurons may not join dendritic bundles arising from deeper pyramidal neurons (either from layer 3 or from layers 3 and 5; Schmolke and Viebahn, 1986; Peters and Sethares, 1991a; Peters et al., 1997). Our results from MAP2 and GABAaα1 labeling, in both rat and primate, are consistent with the interpretation that apical dendrites from layer 2 pyramidal neurons form a separate system of bundles in at least some cortical areas (see Fig. 6). A clear example of this type of dendritic formation is in rat granular retrosplenial cortex. Here the apical dendrites of layer 2 pyramidal neurons form distinct bundles in layer 1, which co-mingle with clusters of PV-ir dendrites (Wyss et al., 1990; Ichinohe and Rockland, 2002). Dendritic tufts from pyramidal neurons in layers 3 and 5, in contrast, co-localize with patches of calretinin-ir terminal-like puncta and with patches of corticocortical terminals (Wyss et al., 1990; Ichinohe and Rockland, 2003a; Ichinohe et al., 2003a).
Our results in monkey suggest that patches of dense Zn+ terminations selectively target dendrites belonging to layer 2 pyramidal neurons. In orbitofrontal and parahippocampal areas, distinct dendritic bundles can be seen to originate from pyramidal neurons in layer 2 and double-labeling further indicates that these co-localize with Zn+ terminals. In other areas, such as areas 4 and 6 and dorsolateral prefrontal areas, the source of the MAP2-ir dendritic patches is less obvious. Area differences in the organization of apical dendrites may be related to variations in the size and shape of upper layer modularity.
The Origin of Zn+ Terminations
Zinc is known to distinguish a subpopulation of corticocortical excitatory terminals (Perez-Clausell and Danscher, 1985; Beaulieu et al., 1992). In rodents, this has been demonstrated by several workers using intra-cerebral or i.p. injections of sodium selenite (Slomianka et al., 1990; Christensen et al., 1992; Casanovas-Aguilar et al., 1998, 2002; Brown and Dyck, 2003). Injected Se2– forms reaction product with vesicular zinc which is transported retrogradely to cell somata. Notably, zinc-enriched neurons have not been reported in the thalamus. Sodium selenite injections in several areas in monkey confirm a cortical location, although some zinc-enriched neurons also can occur in the claustrum and amygdala (Ichinohe and Rockland, 2003b, 2004).
The amygdala projects widely to cortical areas. As amygdalo-cortical terminations target in layers 1 and 2 (Amaral and Price, 1984; Stefanacci et al., 1996), this is further evidence that the amygdala could contribute to Zn+ patches. However, in this matter also, area specializations are likely to be significant. For example, primary motor cortex exhibits Zn+ modularity, but does not receive projections from the amygdala (Avendano et al., 1983).
Dendritic Localization of Zn+ and PV-ir Terminals
In regions where there is spatial overlap of the PV-ir, MAP2-ir and Zn+ patches, PV-ir and Zn+ terminals might be supposed to share the same target dendrites (see Fig. 6). Several inferences follow.
A marked decrease of PV-ir terminals in layer 1 (Williams et al., 1992; Melchitzky et al., 1999; the present study) and of Zn+ terminals in layer 1a suggests that both probably avoid distalmost apical dendrites. As Zn+ terminals seem to continue higher into layer 1, they probably occur independently of PV-ir terminals in this portion of the apical dendrites. Another observation, from electron microscopic studies of monkey prefrontal cortex (Williams et al., 1992; Melchitzky et al., 1999), is that PV-ir terminals in layers 2 and 3a make symmetrical (presumably GABAergic) synaptic contact onto dendritic spines (44%) and shafts (39%). Thus, it is possible that Zn+ and PV-ir terminals can terminate even on the same spines in these layers (see Fig. 6C).
The putative close association of Zn+ and PV-ir terminations is important in the context of interactions between zinc and GABAergic synapses. Both pre- and post-synaptic effects of zinc on GABAergic transmission have been described. Bath application of zinc is known to enhance GABA release (Zhou and Hablitz, 1993; Smart et al., 1994). Smart et al. (1994) have proposed a possible mechanism of increasing excitability of GABAergic terminals through a direct effect of zinc on voltage-gated ion channels. Other studies show that zinc antagonizes the GABAa receptor complex (for a review, see Smart et al., 1994). Importantly, our results on the areal variability of Zn+ and PV-ir modularity suggest area-specific effects.
Importance of Layer 1
Our results support ongoing re-appraisals of the complexity of the uppermost cortical stratum (e.g. Rakic and Zecevic, 2003; Zhu and Zhu, 2004). This zone has attracted considerable interest by virtue of its high density of apical dendritic tufts. These are now known to have a potentially powerful impact on the generation of action potentials, owing to several mechanisms that counteract distance from the axon hillock/initial segment ‘trigger’ site (Magee, 2000; Zhu, 2000; Williams and Stuart, 2003). In addition, field potential recordings in mouse neocortex have demonstrated a significant contribution of layer 1 to the balance between excitation and inhibition in the underlying layers (Shlosberg et al., 2003).
Layer 1 in many areas receives cortical feedback and TC connections and, as noted above, amygdalo-cortical projections terminate widely at the border of layers 1 and 2 (Amaral and Price, 1984; Stefanacci et al., 1996). On the basis of single axon analysis, as well as retrograde tracer experiments, these projections to layer 1 have been considered as characteristically divergent, but the present results demonstrate that the subpopulation of Zn+ corticocortical terminations is frequently patchy.
An interesting possibility is that the patches of Zn+ terminations reported here may correspond to zones of elevated zinc-related plasticity in the upper layers. This is suggested by the reported action of zinc in primate visual and rat somatosensory cortices, where levels of synaptic zinc are rapidly and dynamically regulated in conditions of sensory deprivation (Brown and Dyck, 2002; Dyck et al., 2003). Activity dependent zinc regulation may be a general phenomenon involved in long-term potentiation (as in the hippocampus; Li et al., 2001a,b), or in other processes associated with plasticity (for a review, see Frederickson and Bush, 2001). In this case, the pronounced regional variation in modularity of Zn+ terminations may indicate areal differences in the potential for functional and/or structural remodeling. A similar suggestion has been made concerning the plasticity-related growth-associated phosphoprotein GAP43, which is preferentially distributed in layer 1 of association cortical areas (Benowitz et al., 1989). Some form of activity-dependent regulation may be significant for what are considered ‘top-down’ (or ‘feedback’) influences and further work on specialized dendritic and circuitry properties may help in further elucidating these contextual and attentional effects.
This work was supported by the Brain Science Institute, RIKEN. We would like to thank Ms Kyoko Shirasawa, Miyoko Bellinger, Yoshiko Abe, Hiromi Mashiko and Mr Adrian Knight for their excellent technical help.