Recent studies have revealed striking differences in pyramidal cell structure among cortical regions involved in the processing of different functional modalities. For example, cells involved in visual processing show systematic variation, increasing in morphological complexity with rostral progression from V1 through extrastriate areas. Differences have also been identified between pyramidal cells in somatosensory, motor and prefrontal cortex, but the extent to which the pyramidal cell phenotype may vary between these functionally related cortical regions remains unknown. In the present study we investigated the structure of layer III pyramidal cells in somatosensory and motor areas 3b, 4, 5, 6 and 7b of the macaque monkey. Cells were intracellularly injected in fixed, flat-mounted cortical slices and analysed for morphometric parameters. The size of the basal dendritic arbours, the number of their branches and their spine density were found to vary systematically between areas. Namely, we found a trend for increasing complexity in dendritic arbour structure through areas 3b, 5 and 7b. A similar trend occurred through areas 4 and 6. The differences in arbour structure may determine the number of inputs received by neurons and may thus be an important factor in determining function at the cellular and systems level.
Two diametrically opposed views can be found in the literature regarding cortical organization. On the one hand, it is suggested that cortex is homogenous, all areas being composed of a basic repeated circuit (Creutzfeldt, 1977; Szentagothai, 1978; Rockel et al., 1980; Douglas et al., 1989; Hendry and Calkins, 1998). In the alternative view, cortex is considered to be heterogeneous, with distinct regional variations in the structure of its circuitry (Cajal, 1894; Lund et al., 1993; Hof and Morrison, 1995; DeFelipe et al., 1999) and in its development (Huttenlocher and Dabholkar, 1997; Donoghue and Rakic, 1999; Bishop et al., 2000; Fukuchi-Shimogori and Grove, 2001). According to the first view, functional specificity is determined by the source of inputs. The second view states that both intrinsic cortical circuitry, as well as the source of inputs, determines cortical function.
Recent studies have revealed distinct variations in the pyramidal cell phenotype in different cortical areas (Lund et al., 1993; Elston et al., 1996, 1999a, b, 2001; Elston and Rosa, 1997, 1998a, b, 2000; Elston, 2000; Jacobs et al., 2001). These variations are not random, but are systematic; that is, cells become more branched and more spinous when comparing primary sensory with sensory association and executive cortical areas. The extent of these differences is impressive: layer III pyramidal cells in prefrontal cortex of the macaque monkey have been reported to be up to 16 times more spinous than those in V1 (Elston, 2000). In addition, the extent of regional differences in cell structure is species dependent, being greater in human, for example, than in macaque and marmoset monkeys (Elston et al., 2001).
In the present study we determined the morphology of pyramidal cells in macaque somatosensory and motor cortex to extend the database and to provide further information relevant to the underlying trends and mechanisms that result in phenotypic variation. We found systematic differences in the pyramidal cell phenotype between cortical areas 3b, 5 and 7b, as well as between areas 4 and 6. These results extend previous findings in visual cortex and provide further support for the hypothesis that pyramidal cell structure is specialized for particular functional requirements. In addition, these data provide a basis for future comparisons between homologous areas in other species.
Materials and Methods
Methods of perfusion, slice preparation, cell injection, classification, morphological and statistical analysis have been detailed in previous studies (Sholl, 1953; Eayrs and Goodhead, 1959; Buhl and Schlote, 1987; Elston and Rosa, 1997; Elston et al., 1997). Briefly, two adult female macaque monkeys (Macaca mulatta) were used in the present study (4–4.5 kg). The animals were deeply anaesthetized with sodium pentobarbitol and perfused intracardially with physiological saline, followed by a solution of 4% paraformaldehyde in 0.1 mol/l phosphate buffer (pH 7.2). The protocol for these experiments was in accordance with those endorsed by the NIH (publication No. 86-23, revised 1985) and was approved by the RIKEN Animal Ethics Committee.
Tissue was taken from the caudal bank of the central sulcus (Vogt and Vogt’s area 3b), the rostral bank of the central sulcus [Brodmann’s area 4; corresponding to 4c of Preuss et al. (Preuss et al., 1997)], the exposed lateral portion of the precentral gyrus [Brodmann’s area 6; corresponding to premotor area PMv of Strick (Strick, 1985) or F4 of Matelli and colleagues (Matelli et al., 1985, 1991)], the rostral bank of the intra-parietal sulcus [Brodmann’s area 5; corresponding to I–II of Preuss and Goldman-Rakic (Preuss and Goldman-Rakic, 1989)] and the exposed rostral portion of the inferior parietal lobule (Vogt and Vogt’s area 7b) (Fig. 1). Tissue was taken from the right hemisphere of both animals. Since cortical areal maps are still in some flux, we have presented our data according to the anatomical location of the cortical areas as the most neutral and general option. In particular, data are presented with respect to the central sulcus. Tissue sampled in all cortical areas was taken along a transect perpendicular to the central sulcus, to increase the likelihood that only cells which represent similar body parts were studied (Fig. 1).
Blocks were prepared as flattened specimens by ‘unfolding’ the tissue, removing the white matter and postfixing between glass slides. Sections (250 mm, tangential to the cortical surface) were cut with the aid of a vibratome and prelabelled with the fluorescent dye 4,6-diamidino-2-phenylindole (DAPI; Sigma D9542). Under UV excitation (341–343 nm), individual DAPI-labelled somata could be visualized. Neurons were injected by continuous current (up to 100 nA) until the individual dendrites of each cell could be traced to abrupt distal tips and the dendritic spines were easily visible. Neurons were consistently injected at the base of layer III, near the transition with layer IV, and some of their basal dendrites projected into layer IV. 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) and 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 h) followed by a biotin–horseradish peroxidase complex (Amersham RPN1051; 1:200 in 0.1 mol/l phosphate buffer). Labelling was revealed using 3,3′-diaminobenzidine (DAB; Sigma D 8001; 1:200 in 0.1 mol/l phosphate buffer) as the chromogen (Elston et al. 1997). This method allowed reconstruction of cell morphology in fine detail, including the identification of individual dendritic spines (Fig. 2). In addition, we were able to determine which cells were completely filled and exclude those that were only partially filled.
Neurons were drawn with the aid of a camera lucida microscope attachment. The size of the basal dendritic arbours was determined by calculating the area contained within a polygon that joined the outermost distal tips of the dendritic arbour (using features of NIH image software; NIH Research Services, Bethesda, MD) (Elston and Rosa, 1997). Branching patterns were determined by Sholl analyses (Sholl, 1953), using a 25 mm incremental increase in the radii of successive concentric circles. Spines were drawn at high power (×100 oil immersion objective). All spine types, including sessile and pedunculate (Jones and Powell, 1969), were included in the spine counts, although no distinction was made between them. Correction factors used in other studies when quantifying spines (Feldman and Peters, 1979; Larkman, 1991) were not used in the present study, as the DAB reaction product allows the visualization of spines that issue from the underside of dendrites [e.g. Figs 1 and 3 of Elston et al. (Elston et al., 1999b)]. All spine data were obtained from a single case (R27) with the largest number of cells. Cell bodies were drawn with the aid of a Zeiss ×100 oil-immersion lens and their areas determined by tracing the outermost perimeter, whilst changing focal plane and using standard features of NIH Image. Statistical analysis was performed using SPSS (SPSS Inc., Chicago, IL).
One hundred and sixty-one layer III cells in cortical areas 3b, 4, 5, 6 and 7b were included for analysis. Criteria for inclusion were that cells had an unambiguous apical dendrite characteristic of pyramidal neurons (DeFelipe and Fariñas, 1992), had their complete basal dendritic arbours contained within the section and were well filled. Although there was some interindividual variation in morphological features of cells for any given area, these differences were small by comparison to interareal differences seen in each case. The morphological parameters of cells in all cortical areas were both qualitatively and quantitatively similar in both cases (R25, n = 50 cells; R27, n = 114 cells). Moreover, trends in interareal variation were the same in both cases. As it is now well established that pyramidal cell structure varies systematically in different cortical areas and that these differences are reproducible in different individuals — marmoset monkeys (Elston et al., 1999b), macaque monkeys (Lund et al., 1993), humans (Jacobs et al., 2001) — we pooled data between cases and presented them according to cortical area.
Basal Dendritic Field Areas
Qualitative observation of pyramidal cells in the different cortical areas revealed differences in the appearance of their basal dendritic arbours related to arbour size, complexity of branching structure and spine density. To quantify differences in the size of the basal dendritic arbours, we determined their areas by drawing a polygon around the outermost distal tips of the dendrites and calculating the area contained within. These analyses confirmed that basal dendritic arbours of cells in area 3b were smaller (mean ± SEM, 49.15 × 103 ± 1.63 × 103 μm2) than those in area 5 (80.69 × 103 ± 1.71 × 103 μm2), which were smaller than those in area 7 (96.18 × 103 ± 2.65 × 103 μm2). In addition, cells in area 4 (74.39 × 103 ± 2.45 × 103 μm2) had smaller arbours than those in area 6 (90.33 × 103 ± 3.73 × 103 μm2) (Fig. 3A). A one-way analysis of variance (ANOVA) revealed significant differences in the areas of the basal dendritic arbours of pyramidal cells in the different cortical areas [F(4) = 47.2, P < 0.001]. Post hoc Scheffé tests revealed that 7 of 10 interareal comparisons were significantly different (P < 0.05; Fig. 4A).
Branching Patterns of the Basal Dendritic Arbours
Branching patterns of the basal dendritic arbours of pyramidal neurons are shown in Figure 3B, which plots the results of Sholl analysis (based on concentric circles of increasing radii: 25 μm intervals). As can be seen from the figure, the branching pattern of the basal dendritic arbours of pyramidal neurons was measurably different in all cortical areas studied. The peak branching complexity (the maximum number of dendritic intersections per circle) for cells in area 3b (mean ± SEM, 24.5 ± 0.8) was less than that of cells in area 5 (32.5 ± 0.9), which was, in turn, less than that in area 7b (38.1 ± 1.2). Cells in area 4 had a smaller peak branching complexity (29.8 ± 0.9) than those in area 6 (35.2 ± 1.3). Comparison of areas under the curves revealed that cells in area 7 had ∼20% more branches than those in area 5 and 120% more branches that those in 3b. In addition, cells in area 6 had ∼25% more branches than those in area 4. Despite differences in the number of branches in the basal dendritic arbours, the peak dendritic complexity of cells in all cortical areas was located approximately one-third of the distance from the cell body to the distal tips of the dendrites. A repeated-measures ANOVA revealed significant differences (P < 0.001) in the branching patterns of pyramidal cells between the different areas [intercept, F(1) = 3056; cortical area, F(4) = 33.24].
Spine Densities of the Basal Dendrites
Visual observation revealed differences in the density of spines on the basal dendrites of pyramidal neurons in the different cortical areas. In order to quantify the differences, spines were sampled, per 10 mm dendritic segment, from the cell body to the distal tips of the dendrites. More than 23 000 individual spines were tallied. The results of these analyses are plotted in Fig. 3C, which shows the mean and standard deviation of spine density for 20 randomly selected, horizontally projecting basal dendrites of different cells in each cortical area. In all cases, spine density was zero in the proximal 10 μm of the basal dendrites, rose to a peak density at approximately one-third the distance between the soma and the distal tips of the dendrites, and decreased with further progression toward the distal tips. A two-factor repeated-measures ANOVA (cortical area × distance from soma × spine density), over the entire dendritic length, revealed a significant difference (P < 0.001) in the distribution of spines between cells in all cortical areas studied [intercept, F(1) = 3959; cortical area, F(4) = 45.7]. Post hoc Scheffé tests revealed that 7 of 10 between-area comparisons of spine density were significantly different (P < 0.05; Fig. 4C).
The total number of dendritic spines in the basal dendritic arbour of the ‘average’ pyramidal neuron in each area was calculated by combining data from the Sholl analyses with that of spine densities (Elston, 2001). The ‘average’ neuron in area 3b had 2987 spines in its basal dendritic arbour. Similar calculations revealed that the average pyramidal cell in area 5 had 4689 spines in its basal dendritic arbour, compared with 6841 spines for neurons in area 7b. The ‘average’ cell in area 4 contained 4568 spines, while that in area 6 had 8238 spines.
Somata were drawn, in the plane tangential to the cortical layers, and plotted in Figure 3D. A repeated-measures ANOVA revealed significant differences in cell body size between neurons in the different cortical areas [F(4) = 25.89, P < 0.001]. Somata of cells in layer III of area 3b (mean ± SEM, 187.7 ± 5.03 μm2) were smaller than those in area 5 (247.28 ± 6.70 mm2), which were smaller than those in area 7 (292.05 ± 8.23 mm2). Somata of cells in layer III of area 4 (286.46 ± 9.30 mm2) were larger than those in area 6 (268.24 ± 8.34 mm2). Post hoc Scheffé tests revealed that 6 of 10 between-area comparisons of soma size were significantly different (P < 0.05; Fig. 4D).
As can be seen from Figure 4, comparison of the degree of variation of the different anatomical variables tested here revealed that they are not necessarily correlated, but may vary independently of each other. In particular, cell body size does not necessarily reflect the size or branching patterns of the basal dendritic arbour, or the distribution of spines within the arbour (e.g. area 6).
Layer III pyramidal neurons were intracellularly injected in tangential slices taken from macaque sensorimotor cortex. We found two trends of increasing morphological complexity with progression from the central sulcus to adjacent cortical areas. First, the complexity of dendritic arbours increased with caudal progression from the posterior wall of the central sulcus (area 3b) to the rostral bank of the intraparietal sulcus (area 5) and the exposed rostral portion of the inferior parietal lobule (area 7b). Second, complexity in arbour structure increased with rostral progression from the anterior wall of the central sulcus (area 4) to the exposed lateral portion of the precentral gyrus (area 6). These data confirm previous reports of regional differences in pyramidal cell morphology in sensorimotor cortex (Lund et al., 1993; Jacobs et al., 2001) and extend the finding to show systematic variation between individual areas and functionally related regions.
Cortical Specialization and the Pyramidal Cell
The present data provide further evidence for principled trends underlying pyramidal cell structure and cortical function. While the effects of age, sex, hemisphere and rearing conditions (Scheibel et al., 1975; Nakamura et al., 1985; Jacobs et al., 1993, 1997; Anderson and Rutledge, 1996) have to be taken into account when comparing the present data with those from other studies, our results suggest that cells in polymodal sensory and motor association cortex of the macaque are considerably more spinous and more branched than their counterparts in primary areas. Furthermore, cells in polymodal sensory and motor association cortex are less spinous than their counterparts in macaque prefrontal cortex. Thus, data from macaque monkeys parallel trends reported in humans (Elston et al., 2001; Jacobs et al., 2001). The extent of the differences in macaque may not, however, be as great as those reported in man (Elston et al., 2001) and further studies are required in other species to determine whether or not a similar pattern is seen in brains that have undergone different types of specialization.
Functional Implications of Phenotypic Variation of the Pyramidal Cell
The present results suggest that arbours of pyramidal cells in different sensorimotor areas receive different numbers of excitatory inputs. This follows from the fact that each dendritic spine receives at least one asymmetrical (Colonnier, 1968; Jones, 1968) glutamatergic (DeFelipe et al., 1988; Kharazia et al., 1996) synapse and, therefore, the number of putative excitatory inputs can be estimated from spine number. Furthermore, differences in dendritic length and spine number are likely to influence the number of inhibitory inputs received by these cells [for a discussion, see Elston et al. (Elston et al., 1999c)], resulting in varying degrees of pre-integration inhibition in the arbours of cells in different cortical areas (Spratling and Johnson, 2001). In addition, differences in the branching patterns have been reported to influence the degree to which processing may be compartmentalized within the arbours (Koch et al., 1982, 1983) and the functional capacity of neurons (Poirazi and Mel, 2001).
Putative differences in the numbers of excitatory inputs received by supragranular pyramidal cells in the different cortical areas may determine convergence and divergence to individual cells, thus influencing their receptive field properties. For example, the receptive field size of cells in visual cortex is correlated with the size of their basal dendritic arbours (Colonnier, 1964; Gilbert and Wiesel, 1979; Elston and Rosa, 1998a,b; Elston et al., 1999b). In addition, the sampling profiles of neurons may be influenced by the size of their arbours (Lund et al., 1993; Malach, 1994). For example, the size of basal dendritic arbours of supragranular pyramidal cells in several visual areas is correlated with the size of intrinsic horizontally projecting axon patches (Lund et al., 1993; Elston and Rosa, 1998a,b). As previously suggested (Malach, 1994), such a correlation would result in a maximal sampling diversity. Further experiments are required to determine whether or not a similar correlation exists in sensori-motor cortex, which is also characterized by an intrinsic horizontal latticework of connections, or patches (Juliano et al., 1990; Huntley and Jones, 1991; Lund et al., 1993).
Two other examples come to mind regarding how cell structure might influence cortical function. In prefrontal cells, the long post-stimulus spiking activity, which is thought to underpin their role in memory, rule learning and reasoning (Fuster and Alexander, 1971; Kubota and Niki, 1971; Fuster, 1973; Wallis et al., 2001), may result from a high degree of interconnectedness and the integration of large numbers of inputs (Soloway et al., 2002). In addition, the work of Murayama and colleagues (Murayama et al., 1997) can be interpreted as evidence that interareal differences in intrinsic connectivity influence functional architecture. Namely, 40–100 Hz stimulation of layers 2/3 results in long-term potentiation (LTP) in temporal lobe cortex, but long-term depression (LTD) in V1. These different response characteristics might, in part, result from differences in the input configuration and interactions of the intrinsic excitatory connections to pyramidal cells (McGuire et al., 1991; Tanigawa et al., 1998).
Systematic differences in the phenotype of supragranular pyramidal neurons correlate, to an extent, with the proposed organization of cortical areas into hierarchies (Mishkin, 1979; Maunsell and van Essen, 1983; Pons et al., 1987, 1992; Felleman and van Essen, 1991; Gross et al., 1993; Graziano and Gross, 1997). The present results in somatosensory cortex show that the size of cells, their branching complexity and the total number of spines within their basal dendritic arbour increase through areas 3b, 5 and 7b, which reportedly form successive hierarchical levels in somatosensory processing (Friedman, 1983; Friedman et al., 1986; Felleman and van Essen, 1991), but see Neal et al. (Neal et al., 1987) and Andersen et al. (Andersen et al., 1990). However, our results in motor areas 4 and 6 conform less well to this interpretation. That is, motor effector cells in area 4, which is generally considered to be the end of a cortical motor decision–execution pathway, are smaller and less branched than those in area 6 (Jacobs et al., 2001), which is usually placed before area 4 in this pathway (Jones, 1986; Geyer et al., 2000). Furthermore, comparison of infragranular pyramidal cells reveals a systematic decrease in the size of, number of branches in and number of spines in their basal dendritic arbours through areas STP, TE and TEO (Elston and Rosa, 2000). Moreover, cells in prefrontal cortex, which modulate sensory processing (Vidyasagar, 1996; Buchel and Friston, 1997; Ito and Gilbert, 1999; Mehta et al., 2000), are larger, more branched and considerably more spinous than their target cells (Elston, 2000; Jacobs et al., 2001; Soloway et al., 2002).
It is also clear that much remains to be determined about the anatomical and functional relationships between sensori-motor areas, particularly in terms of the functional weighting of connections for specific tasks (Jones, 1986; Paulesu et al., 1997). The results of recent imaging studies, for example, suggest that individual areas do not necessarily function in a successive temporal (or hierarchical) sequence, but, rather, many different cortical areas may be simultaneously involved in any given task (Kaas, 1990; Calvert, 2001). Thus, differences in pyramidal cell phenotype may be influenced by factors such as regional variation in gene expression during development (Rakic, 1988; Huttenlocher and Dabholkar, 1997; Donoghue and Rakic, 1999; Bishop et al., 2000; Fukuchi-Shimogori and Grove, 2001; Tochitani et al., 2001) and specialization during evolution (Cajal, 1894; Elston et al., 2001). The implications of interareal variation in pyramidal cell structure reported here will require objective analyses of cell structure (Elston and Jelinek, 2001; Jelinek and Elston, 2001) and quantification of patterns of connectivity between areas (Young, 1993; Jouve et al., 1998) across a number of species.
Cell Soma Size and Dendritic Arbour Structure: is there a Correlation?
The present results provide further evidence that cell soma size is not necessarily a reliable indicator of dendritic arbour structure. For example, there was only 50% concordance between the pair-wise statistical comparisons between cortical areas for branching pattern and somal size (Fig. 4). More specifically, we found no significant difference in the soma size of neurons between areas 6 and 5, 6 and 7, 4 and 6 or 4 and 7; but cells in area 4 had significantly smaller dendritic arbours than those in areas 6 and 7, and cells in area 7 had significantly more dendritic branches than those in areas 4 and 6 (Fig. 4). In addition, even when we found no significant difference in the size of the somata of cells in different cortical areas, we found marked differences in the number of spines contained within their basal arbours. For example, cells in area 6 contain at least 75% more spines than those in areas 4 and 5, but their somata were not significantly different in size (Elston et al., 1999b). Reverse comparisons also revealed a lack of consistency between cell body size and arbour structure. For example, soma size was significantly different between cells in areas 4 and 5, but there was no significant difference in their dendritic arbour size. Various groups have now reported a lack of correlation between cell body size of cortical neurons and the structure of their arbours — rat somatosensory cortex (Larkman, 1991), cat visual cortex (Matsubara et al., 1996), monkey visual cortex (Elston et al., 1999b) — suggesting widespread variance between these parameters.
By injecting neurons in different cortical areas we have demonstrated systematic variation in the pyramidal cell structure in sensorimotor cortex of the macaque monkey. Dendritic arbour size, number of branch points, spine density and soma size may vary independently of each other. As supragranular pyramidal cells form extensive intrinsic and interareal cortical projections, regional variation of the pyramidal cell phenotype is likely to have profound implications for areal and systems function. These findings are in direct contrast with the belief that all cortical areas are built of the same neural components linked in similar ways. Instead, the present results support the thesis that the pyramidal cell and the circuits it forms are modified in parallel with particular functional requirements. Further studies are required to determine the extent of regional variation of pyramidal cell apical dendrites, as well as interneuronal variation.
Thanks to Alberto Muñoz for many helpful discussions, Javier DeFelipe for providing the antibody to Lucifer yellow and Ruth Benavides-Piccione, Lidia Alonso, Inma Ballesteros and Azucena Oritz for technical help. G.N.E. was supported by a C.J. Martin Fellowship from the National Health and Medical Research Council of Australia; collaborative work was supported by research funds from the RIKEN Brain Science Institute.