Pyramidal neurons of granular prefrontal cortex of the galago: Complexity in evolution of the psychic cell in primates

Abstract

Typically, cognitive abilities of humans have been attributed to their greatly expanded cortical mantle, granular prefrontal cortex (gPFC) in particular. Recently we have demonstrated systematic differences in microstructure of gPFC in different species. Specifically, pyramidal cells in adult human gPFC are considerably more spinous than those in the gPFC of the macaque monkey, which are more spinous than those in the gPFC of marmoset and owl monkeys. As most cortical dendritic spines receive at least one excitatory input, pyramidal cells in these different species putatively receive different numbers of inputs. These differences in the gPFC pyramidal cell phenotype may be of fundamental importance in determining the functional characteristics of prefrontal circuitry and hence the cognitive styles of the different species. However, it remains unknown as to why the gPFC pyramidal cell phenotype differs between species. Differences could be attributed to, among other things, brain size, relative size of gPFC, or the lineage to which the species belong. Here we investigated pyramidal cells in the dorsolateral gPFC of the prosimian galago to extend the basis for comparison. We found these cells to be less spinous than those in human, macaque, and marmoset. © 2005 Wiley-Liss, Inc.

Pyramidal cells in the cerebral cortex vary in size, branching pattern, spine density, and the total number of dendritic spines (Lund et al., 1993; Elston and Rosa, 1997, 1998, 2000; Elston et al., 1999a, 1999b; Elston, 2000, 2003a; Jacobs et al., 2001). These structural differences are thought to influence sampling geometry, compartmentalization of processing, cooperativity between inputs, and the total number of excitatory inputs integrated within their arbors (for reviews, see Jacobs and Scheibel, 2002; Elston, 2003b, 2003c). In addition, marked differences have been reported in the patterns of axon projections of pyramidal cells in different cortical areas and species (Bugbee and Goldman-Rakic, 1983; Preuss and Goldman-Rakic, 1991; Lund et al., 1993).

In a series of studies, we have attempted to correlate the relationship between pyramidal cell structure, patterns of connectivity, and function. These studies were focused primarily in visual cortex, where a great deal is known about the response properties of neurons and visual function. These studies revealed systematic differences in pyramidal cell structure in visual cortical areas, which parallel systematic differences in their function (for reviews, see Elston, 2003b, 2003c). For example, there is a consistent trend for pyramidal cells of progressively more complex structure with anterior progression through occipitotemporal cortical areas, which parallel differences in their functional complexity and discharge properties. Systematic differences in pyramidal cell structure in somatosensory, motor, and limbic cortex also parallel functional specializations of neurons contained within (Elston and Rockland, 2002; Elston et al., 2005a).

In the present investigation, we extend our studies of pyramidal cell structure in prefrontal cortex. Presently available data reveal that pyramidal cells in adult human granular prefrontal cortex (gPFC) are considerably more branched and more spinous than those in gPFC of the macaque monkey, which are more branched and more spinous than those in gPFC of marmoset and owl monkeys. These data suggest that pyramidal cells in human receive more excitatory inputs than those in the macaque monkey and compartmentalize these inputs to a greater degree. Thus, the functional capacity of the individual cells is likely to be greater in human than in macaque,greater in macaque than owl monkey, and so on (cf Poirazi and Mel, 2001). These data led us to question how specialization in pyramidal cell structure observed in extant species may have evolved. Presently available data in humans, macaques, marmosets, and owl monkeys suggest at least two possibilities. Differences in the gPFC pyramidal cell phenotype could vary between different primate lineages; for example, highly complex pyramidal cell structure may be restricted to simians or anthropoids. Alternatively, the structural complexity of gPFC pyramidal cells may reflect the extent of gPFC expansion, irrespective of lineage.

In a bid to test these two possibilities, we studied pyramidal cell structure in the galago, a prosimian primate with a well-developed gPFC (Fig. 1). The lorisiform prosimians, which include the galago, have been separated from the simian lineage for over 50 million years (Martin, 1990). Importantly, the galago has a smaller granular prefrontal cortex than that in the marmoset (Fig. 2) (Preuss and Goldman-Rakic, 1991). Consequently, if pyramidal cells in the gPFC of the galago are less branched and less spinous than those in the marmoset, the complexity of the gPFC pyramidal cell phenotype is likely to reflect the relative expansion of gPFC. Alternatively, if pyramidal cells in the gPFC of the galago are more branched and more spinous than those in the marmoset, other features are likely to influence the cell phenotype.

Figure 1. A: Photomicrograph of a sagittal section of the galago brain illustrating, in particular, cytoarchitecture in the frontal and parietal lobes. Higher-power micrographs highlight differences in the cytoarchitecture between (B) granular prefrontal cortex, (C) premotor cortex, and (D) the primary motor area.

Figure 2. Photomicrograph of the lateral aspect of the brain of the human, macaque monkey, marmoset monkey, owl monkey, and galago. The extent of granular prefrontal cortex is represented by stipple. Scale bar = 2 cm for human and 1 cm for all other species.

MATERIALS AND METHODS

Galagos used in the present study (G1 and G2, 3.5 and 4 year old males, respectively) were the same as those used in a previous study (Elston et al., 2005b). Briefly, animals were overdosed with sodium pentobarbitone and perfused with 0.95% saline in PB (0.1 M phosphate buffer, pH 7.2), followed by 4% paraformaldehyde in PB. Cortices were flat-mounted and left in the 4% paraformaldehyde in PB for 12 hr at 4°C. Blocks including the entire frontal lobe were cut tangentially to the cortical surface and were sectioned (250 μm) with the aid of a Vibratome. The region corresponding to dorsolateral granular prefrontal cortex, anterior and medial to the frontal eye field (Wu et al., 2000; Wu and Kaas, 2003), was selected for study (Fig. 3). Individual neurons were visualized under fluorescence excitation by first prelabeling the sections with 10⁻⁵ mol/L 4,6-diamidino-2-phenylindole (D9542; Sigma) in PB. Cells were injected [8% Lucifer Yellow (L-0259; Sigma) in 0.1 M Tris buffer; pH 7.4] at the base of layer III. Because of the presence of a distinct granular layer (e.g., Fig. 1B) in the PFC of the galago [by definition prefrontal cortex is granular (Brodmann, 1913)], it was relatively easy to identify layer III and inject cells at the base of layer III. Three criteria are used to determine that cells are injected at the base of layer III. In the first instance, the unprocessed serial tangential sections cut from the flat-mounted blocks are placed on a black background and their color compared: supragranular sections are much whiter than the granular and infragranular sections. Second, the density of cell bodies clearly differs between supragranular, granular, and infragranular sections when visualized under UV excitation (e.g., see Fig. 3 of Elston and Rosa, 1997), allowing distinction between these layers. Third, by injecting cells in successive sections, it is easy to distinguish and identify the section that contains the granular layer by cell type. In addition, the base of layer III is relatively easily detected while injecting as the transition between layers III and IV can be detected visually while injecting.

Figure 3. Schematic of the galago brain and various cortical areas that have been identified by cyto- and myelarchitectonic, connectional, and mapping experiments. Note the motor areas that have been identified in the frontal lobe, including the primary (M1), supplementary (SMA), premotor dorsal (PMD), premotor ventral (PMV) areas. Anterior to the arcuate sulcus lies the granular prefrontal cortex, which includes the region of cortex extending from the frontal eye field (FEF) to the frontal pole. Neurons were injected in gPFC in both cases; however, the exact location in which neurons were injected varied slightly between galago 1 (G1) and galago 2 (G2). Modified from Kaas (2003).

Following cell injection, the tissue was processed with an antibody to Lucifer Yellow (LY) for 5 days at a concentration of 1:400,000 in stock solution [2% bovine serum albumin (Sigma A3425), 1% Triton X-100 (BDH 30632), 5% sucrose in 0.1 mol/l phosphate buffer]. Anti-LY was detected by a species-specific biotinylated secondary antibody (Amersham RPN 1004; 1:200 in stock solution for 2 hr), followed by a biotin-horseradish peroxidase complex (Amersham RPN1051; 1:200 in 0.1 mol/l phosphate buffer). Labeling was revealed using 3,3′-diaminobenzidine (DAB; Sigma D 8001; 1:200 in 0.1 mol/l phosphate buffer) as the chromogen (for details of cell, filling, immunohistochemical processing, confirmation of the laminar lacoation of injected neurons, and methods of quantification, see Elston and Rosa, 1997, 1998).

Cells were reconstructed from the 250 μm thick slices. Neurons that had an unambiguous apical dendrite were located at the base of layer III and had the complete basal dendritic arbor contained within the section were drawn with the aid of a camera lucida attachment coupled to a Zeiss Axioplan microscope. A polygon joining the outermost distal tips of the basal dendrites was drawn over each cell and the area contained within was determined with the aid of NIH Image (Elston and Rosa, 1997). Complexity of the dendritic branching structure was quantified by Sholl analysis (Sholl, 1953). Spine density was determined as a function of distance from the cell body to the distal tips of the dendrites for complete reconstructions of 10 randomly selected horizontally projecting basal dendrites of different pyramidal cells (e.g., Eayrs and Goodhead, 1959; Valverde, 1967; Elston, 2001). Correction factors used in Golgi studies when quantifying spines (e.g., Feldman and Peters, 1979) were not used in the present study as the DAB reaction product allows the visualization of spines that issue from the underside of dendrites. No distinction was made between the different types of spines: stubby, mushroom, and thin spines were all included in the counts. The tangential dimension of somata was determined using standard features of NIH Image. The total number of spines located in the basal dendritic arbor of the average pyramidal cell was calculated by multiplying the average number of dendritic branches per Sholl annul by the average spine density for the corresponding annul and summing over all successive annuls included in the dendritic arbor (see Elston, 2001).

RESULTS

Ninety-six neurons were injected at the base of layer III in dorsolateral prefrontal cortex (59 in B1 and 37 in G2), 47 of which were included for analyses according to selection criteria outlined in the Materials and Methods section (20 in G1 and 27 in G2). We made no attempt to determine which prefrontal areas these cells were sampled from, but it was likely that a large portion of them were sampled anterior to the frontal eye field (FEF; Fig. 4). Data for these cells was compared with those of the primary (V1), second (V2), dorsolateral (DL), and inferotemporal (IT) visual areas sampled in the same hemisphere of these animals (Elston et al., 2005b).

Figure 4. Photomicrographs of the basal dendritic arbors pyramidal cells injected with Lucifer Yellow at the base of layer III in the dorsolateral granular prefrontal cortex that were processed for a light-stable DAB reaction product. Scale bar = 50 μm.

Basal Dendritic Arbor Size

In galago 1, the average size of the basal dendritic arbors of layer III pyramidal cells was 91.3 ± 4.97 × 10³ μm² (mean ± standard error) and that in galago 2 was 118 ± 4.32 × 10³ μm² (Fig. 5). In both animals, the arbors of layer III pyramidal cells in gPFC were larger than those of cells in V1, V2, and DL. One-way analyses of variance revealed these differences to be significant in both G1 (F_(4,85) = 17.8; P < 0.001) and G2 (F_(3,71) = 78.8; P < 0.001).

Figure 5. A: Frequency histograms of the size of the basal dendritic arbors of layer III pyramidal neurons in the dorsolateral prefrontal cortex (PFC; black) dorsolateral visual area (DL; blue), second visual area (V2; green), and the primary visual area (V1; orange). B: Plots of the number of dendritic intersections of the basal dendrites of pyramidal neurons as determined by Sholl analysis. C: Plots of the number of dendritic spines per 10 μm segment of dendrite, as a function of distance from the cell body, in the basal dendritic arbors of layer III pyramidal neurons. D: Frequency histograms of somal areas of pyramidal neurons. Error bars = standard deviations.

Complexity of Basal Dendritic Arbors

Cells in PFC had a peak dendritic complexity of 27.1 ± 3.49 (mean ± SEM) in G1 and 27.9 ± 3.67 in G2 (Fig. 5). Repeated-measures ANOVAs revealed the branching patterns of cells to be significantly different between cortical areas in G1 (intercept, F_(1,81) = 700, P < 0.001; cortical area, F_(4,81) = 32.8, P < 0.001) and G2 (intercept, F_{(1, 68)} = 1351, P < 0.001; cortical area, F_(3,68) = 53.2, P < 0.001). The number of branches in the basal dendritic arbors of neurons in PFC was higher than that in V1, V2, and DL but lower than that in IT (Fig. 5).

Spine Densities of Basal Dendrites

The peak average spine density in G1 was 15.7 ± 3.02 (mean ± SD) at a distance of 101–110 μm from the cell body. That in G2 was 16.0 ± 2.79 at the same distance from the cell body (Fig. 5). Repeated-measures ANOVAs revealed a significant difference in the distribution of spines along the dendrites of pyramidal cells in G1 (intercept, F_(1,32) = 707.8, P < 0.001; cortical area, F_(4,32) = 57.3, P < 0.001) and G2 (intercept, F_(1,36) = 1.62 × 10³, P < 0.001; cortical area, F_(3,36) = 112.3, P < 0.001). Posthoc analysis revealed that spine density was higher in PFC than all other areas sampled in both G1 and G2. By multiplying the spine density, averaged over 25 μm segments of dendrite, by the number of branches recorded over the corresponding segment, we were able to estimate the total number of dendritic spines in the basal dendritic arbor of the average layer III pyramidal neuron. In galago 1, the average cell had 3,219 spines in its basal dendritic arbor, which in galago 2 had 3,759 spines (Fig. 6).

Figure 6. Estimates of the total number of spines in the basal dendritic arbor of the average pyramidal cell in granular prefrontal cortex in the young mature adult galago, marmoset monkey, macaque monkey, and human. Note that there is a trend for a systematic increase in the total number of spines in the basal dendritic arbor of pyramidal cells with increasing size of gPFC. Individual values obtained from G1 and G2 are illustrated as individual columns headed galago (left and right column, respectively). Relative interindividual differences in the pyramidal cell phenotype in gPFC of the galago may be attributable to differences in the age of the animals (3.5 vs. 4 years old), the differing gender of the animals, or reflect natural variation or differences in the region from where cells were sampled. Data from Aotus (Elston, 2003a) are not compared here because of the marked difference in the age of the animal from which those data were sampled.

Somal Areas

The average size of the somata of layer III pyramidal cells in G1 was 202 ± 44 × 10³ μm² (mean ± standard deviation), and that in G2 was 247 ± 35 × 10³ μm², being on average larger than those of cells in other areas (Fig. 5). One-way analyses of variance revealed these differences to be significant in both G1 (F_(4,85) = 18.02; P < 0.001) and G2 (F_(3,71) = 188; P < 0.001). Posthoc analysis revealed that cells in PFC had significantly larger somata than those in V1, V2, DL, and IT.

DISCUSSION

The present study is one of several morphological investigations in which the structure of prefrontal cortical pyramidal cells has been quantified and compared with that of pyramidal cells in other cortical areas sampled from the same cortical hemisphere. This comparative approach has revealed new insights into specialization of the pyramidal cell phenotype in the granular prefrontal cortex of different primate species. The results of earlier studies suggested that the highly complex phenotype reported in gPFC of humans and macaque monkeys may be restricted to higher primates. The present data are consistent with this theory. By quantifying the differences in pyramidal cell structure in different species, we are gaining new insight into the thinking of Ramon y Cajal (1893) when, based on his comparative data, he concluded that the pyramidal cell is the “psychic cell.” Our findings suggest that specializaton of pyramidal cell phenotype may be an important factor during the evolution of increasingly complex intellect in primates.

How these differences in cortical circuitry may influence cognitive style, particularly that in primates, remains to be determined. However, there are multiple converging lines of evidence to suggest a parallel between pyramidal cell structure and cognitive ability. For example, decline in cognitive ability with aging is paralleled by a decrease in dendritic branching and spine loss (Scheibel et al., 1975, 1976; de Brabander et al., 1998; Page et al., 2002) and the relatively impoverished structure of pyramidal cells in the brains of individuals with Down syndrome and fragile X parallel cognitive abilities (Marin-Padilla, 1976; Huttenlocher, 1979; Suetsugu and Mehraein, 1980; Takashima et al., 1981; Leuba, 1983; Hinton et al., 1991; Dierssen et al., 2002; for review, see Kaufmann and Moser, 2000; Dierssen et al., 2003). In addition, pyramidal cells in experimental animals reared in an enriched environment have more complex structure, and the animals perform better on behavioral tasks, than animals that have been reared in a nonenriched environment (e.g., Greenough et al., 1973, 1985; Dierssen et al., 2002; for reviews, see Harris, 1999; Kintsova and Greenough, 1999; Wooley, 1999; Elston and DeFelipe, 2002).

While there is a growing body of evidence to suggest a link between pyramidal cell structure and behavioral abilities, less is known as to how aspects of pyramidal cell structure, such as the size or branching pattern of their dendritic arbors, or their spine density, may directly affect cellular and systems function. Our estimates of the number of spines in the basal dendritic trees of layer III pyramidal cells reveal that those in gPFC of the galago are 7 times more spinous than those in its V1, those in macaque gPFC are 17 times more spinous than those in galago V1, and those in human gPFC are on average 30 times more spinous than those in galago V1 (present results; Elston et al., 2001), making it highly likely that cells in these different species receive different numbers of asymmetrical (excitatory) inputs (for reviews, see Elston, 2002, 2003b, 2003c; Jacobs and Scheibel, 2002). In addition, differences in the number of branches may influence the degree to which processing of these inputs is compartmentalized within the dendritic arbors of pyramidal cells in gPFC of the different species. More spinous cells such as those in gPFC of human, for example, are more branched than the less spinous cells in galago gPFC, imparting different functional capabilities: cells with more dendritic branches have a greater capacity than cells with fewer dendritic branches (e.g., Poirazi and Mel, 2001). Differences in the branching structure also influence the propagation of inputs to the soma (Vetter et al., 2001). Furthermore, circuits composed of neurons with complex branching pattern such as those observed in human gPFC, which are also characterized by high spine density (peak of 32.5 ± 1.64 spines per 10 μm), have a greater potential for plastic change than circuits composed of less branched cells such as those in galago gPFC with lower spine density (16.0 ± 2.78) (Stepanyants et al., 2002). Differences in the dendritic structure of pyramidal cells in gPFC of different primate species, coupled with differences in patterns of intrinsic connectivity (Bugbee and Goldman-Rakic, 1983; Preuss and Goldman-Rakic, 1991), may influence the potential for resonant excitation via recurrent collaterals, which is believed to be important in mnemonic processing (for reviews, see Yuste and Tank, 1996; Koch, 1997; Häusser et al., 2000; Magee, 2000; Euler and Denk, 2001; Segev et al., 2001; Wang, 2001).

Given the variation in the structural complexity of the mature pyramidal cell phenotype in gPFC of different species (Fig. 7), and the behavioral implications of these differences, it was natural to ask why they occur. One interpretation is that pyramidal cells have become more spinous during expansion of gPFC (Fig. 8). The present data are consistent with this interpretation. For example, gPFC has a surface area of 148 mm² in marmoset, 1,733 mm₂ in macaque, and 39,287 mm² in man (Brodmann, 1913; for translation, see Elston and Garey, 2004), and complexity of pyramidal cell structure increases through these species. These data may be interpreted as evidence that pyramidal cell structure necessarily becomes more complex during cortical expansion. However, this is not the case. The primary visual area has also undergone considerable expansion in size in primates. That in the macaque (1,102 mm²) is considerably larger than that in galago (343 mm²), which is larger than that in the marmoset (198 mm²) (Brodmann, 1913), yet the number of spines found in the arbors of pyramidal cells in these species does not follow a similar progression [735, 556, and 699, respectively; the first number is the average from animals RM12, RM13, MF1, and MF2 (Elston et al., 2004a); the second number is the average from animals G1 and G2 (Elston et al., 2004b); the third number is the average from animal BS10 (Elston et al., 1999b)]. Thus, it appears as though circuitry or, more specifically, pyramidal cell structure has evolved differently in V1 and gPFC. In future studies, it will be worthwhile to study the apical dendrites of these neurones, as well as pyramidal cells in other cortical layers (e.g., Elston and Rosa, 2000).

Figure 7. Graphs illustrating differences in the number of dendritic branches (top) and spines density (bottom) in the basal dendritic trees of layer III pyramidal cells in granular prefrontal cortex of human (red), macaque monkey (orange), marmoset monkey (blue), and the galago (cream). Data taken from Elston et al. (2001) and the present results.

Figure 8. Graph illustrating the size of granular prefrontal cortex relative to the surface area of the entire cortex of the human, chimpanzee, gibbon, mandril, baboon, macaque monkey, long-tailed monkey, capuchin monkey, marmoset monkey, black lemur, dwarf lemur, fruit bat, dog, cat rabbit, hedgehog, armadillo, and opossum (taken from Brodmann, 1913). Regression analysis revealed that exclusion of the human (long dashed line) and the human and chimpanzee (short dashed line) result in slopes of lesser gradient than regression analysis of all species (solid black line), suggesting disproportionate expansion of gPFC in human and chimpanzee.

Acknowledgements

The authors thank Iwona Stepniewska, Laura Trice, Laura Ferris, and Brendan Zeitsch for technical help.