Eighteen chimpanzee subjects were used in this study, including 8 females (mean age at death = 36 years, SD = 12.7, range = 13– 48) and 10 males (mean age at death = 26 years, SD = 10.7, range = 10 – 41). Six of the chimpanzee subjects were wild-caught before 1973 and lived in captivity since that time. The remaining 12 chimpanzees were born in captivity. All 18 subjects lived in social groups ranging from 2 to 13 individuals at Yerkes National Primate Research Center in Atlanta, Georgia. All subjects died from natural causes and were not part of any research protocol that may have contributed to their death.
Hand preference was considered for a task measuring coordinated bimanual actions, referred to as the tube task (Hopkins, 1995
). Handedness data from these subjects have been previously reported (Hopkins, 1995
; Hopkins et al., 2004
). For this task, peanut butter is smeared on the inside edges of polyvinylchloride tubes ≈15 cm in length and 2.5 cm in diameter. Each time the subjects reached into the tube with their finger, extracted peanut butter, and brought it to their mouth the hand used was recorded. We used measurements of hand preference on this task because it is stable across the lifespan and the strength of handedness elicited in chimpanzees by the tube task is significantly higher than for other actions, such as bimanual feeding or simple reaching (Hopkins, 2007
). Although the number of responses obtained from each subject differed for this task, a minimum of 30 responses were obtained for each individual.
Binomial z-scores were calculated for each subject on the basis of the frequency of left- and right-hand use. Subjects with z-scores greater than 1.95 or less than −1.95 were classified as right- and left-handed, respectively. Subjects with z-scores between −1.95 and 1.95 were classified as ambidextrous. In addition, a handedness index (HI) was derived for each subject by subtracting the number of right-handed responses from the number of left-handed responses and dividing by the total number of responses: HI = (R − L)/(R + L). Positive values reflect right-hand preference and negative values represent left-hand preference. The absolute value of the HI corresponds to the consistency of directional hand preference. Individuals that performed neuroanatomical measurements were blind to the HI score of the subjects.
MRI and measurement
Brains from each subject were obtained after death and were immersion-fixed in 10% formalin. In most cases the precise postmortem interval was not recorded at Yerkes National Primate Research Center; however, it was never greater than 14 hours. All of the brains were scanned postmortem using a T2-weighted protocol with a 1.5 T magnet. Images were collected in the transverse plane using a gradient echo protocol (pulse repetition = 22.0 s, echo time = 78.0 ms, number of signals averaged = 8 –12, and a 256 × 192 matrix reconstructed to 256 × 256). The “knob” area of the precentral gyrus that corresponds to the location of hand representation was identified in serial 1-mm slices in the axial plane following procedures previously used in human and ape brains (Yousry et al., 1997
; Hopkins and Pilcher, 2001
). Quantification of the knob region was obtained in the axial plane because this is the most common approach used in human studies (Hammond, 2002
) and it is difficult to reliably quantify this region from other planes in chimpanzees. Nonetheless, a previous study that quantified the hand knob in human brains from the sagittal plane found that these measures of asymmetry correlated with the subjects’ handedness (Foundas et al., 1998
). Because these results are largely consistent with the findings reported for measures from the axial plane, we reasoned that the axial measurement of the knob was sufficiently representative of the primary motor cortex hand area of chimpanzees.
Morphological measurements of the hand knob were performed using ANALYZE software (ANALYZE, Lenexa, KS). The horizontal epsilon or inverted omega that projected into the postcentral gyrus was traced on each image (). The dorsal and ventral edges of the knob served as the markers for defining the boundaries of the area. For each slice and hemisphere, an area measurement of the region was calculated by use of a mouse-driven pointer that traced the region of interest. The total of area measurements from all slices in which the knob was present were summed and used to derive a volumetric measure for each hemisphere (ranging from 5 to 13 slices in the sample).
Fig. 1 The region of the hand knob landmark as seen in four axial MRI sections of a chimpanzee brain from dorsal to ventral (A–D). Blue lines indicate the tracing of the central sulcus (CS). The horizontal epsilon or inverted omega that projects into (more ...)
Tissue preparation and immunohistochemistry
The region of hand representation in primary motor cortex was dissected from each hemisphere as a block ≈4 cm thick containing the pre- and postcentral gyri. To estimate the location of hand representation, the hand knob landmark was viewed on horizontal MRI scans of each brain (Hopkins and Pilcher, 2001
) and the corresponding dorsoventral level was noted on the lateral surface. In addition, prior to dissection the central sulcus was spread open to reveal the middle genu, which is anatomically equivalent to the hand knob seen in axial MRI sections in human brains (Yousry et al., 1997
). The position of the hand knob landmark generally accords with previous electrophysiological studies of motor maps in this species (Grünbaum and Sherrington, 1903
; Leyton and Sherrington, 1917
; Hines, 1940
; Dusser de Barenne et al., 1941
; Bailey et al., 1950
). Moreover, a recent functional imaging study on grasping using positron emission tomography (PET) in five chimpanzees found significant activation in the knob region in the hemisphere contralateral to the hand used (Hopkins et al., 2006
After dissection, tissue blocks were cryoprotected by immersion in buffered sucrose solutions up to 30%, frozen on dry ice, and sectioned at 40 μm with a sliding microtome perpendicular to the axis of the central sulcus. Every 10th section (400 μm apart) was stained for Nissl substance with a solution of 0.5% cresyl violet. For each individual, sections from both hemispheres were Nissl-stained together in order to ensure comparable staining conditions for subsequent analyses. Sections not used for immediate staining were cryoprotected in a storage solution consisting of glycerol, ethylene glycol, dH2O, and phosphate buffer (3:3:3:1 volume/volume) and archived at −20°C.
Immunohistochemistry was performed for each antigen on adjacent 1:20 series of sections. Free-floating sections were stained with mouse monoclonal antibodies to a non-phosphorylated epitope in neurofilament H (SMI-32 antibody; Covance International, Netherlands; dilution 1:2,000) and parvalbumin (Swant, Belinzona, Switzerland, Cat. no. 235; dilution 1:10,000). The SMI-32 mouse monoclonal IgG1
antibody was raised against the non-phosphorylated epitope of neurofilament H isolated from homogenized hypothalami of Fischer 344 rats. On conventional immunoblots, SMI-32 visualizes two bands (200 and 180 kDa), which merge into a single neurofilament H line on two-dimensional blots (Sternberger and Sternberger, 1983
). The antibody reacts with a nonphosphorylated epitope from 200-kD neurofilament heavy chain of most mammalian species. This protein is expressed in cell bodies, dendrites, and some thick axons within a subset of neurons that are mostly pyramidal cells (Campbell and Morrison, 1989
; Hof and Morrison, 1995
). Other cells and tissues are unreactive and the antibody does not recognize the phosphorylated 200-kD neurofilament heavy chain. The mouse monoclonal IgG1
PV antibody was raised against PV from carp muscle. It has been shown to bind with high affinity to the tertiary structure of PV from multiple species including macaque monkeys and humans, with binding eliminated by addition of exogenous PV (Celio et al., 1988
). No crossreactivity with other calcium-binding proteins was noted in radioimmunoassay and immunoblotting assays (Celio et al., 1988
). In primate brain tissue the pattern of staining with this antibody is consistent with that previously established for other PV antibodies (Conde et al., 1994
Prior to immunostaining, sections were rinsed thoroughly in phosphate-buffered saline (PBS) and pretreated for antigen retrieval by incubation in 10 mM sodium citrate buffer (pH 3.5) at 37°C in an oven for 30 minutes. For the SMI-32 antibody, antigen retrieval used the same buffer with pH 8.5 at 90°C in a water bath to achieve improved staining. Sections were then immersed in a solution of 0.75% hydrogen peroxide in 75% methanol to eliminate endogenous peroxidase activity. After rinsing again, sections were incubated in the primary antibody in a diluent containing PBS with 2% normal horse serum and 0.3% Triton X-100 for ≈48 hours on a rotating table at 4°C. After rinsing in PBS, sections were incubated in the secondary antibody (biotinylated antimouse IgG, Vector Laboratories, Burlingame, CA; dilution 1:200) and processed with the avidin-biotin-peroxidase method using a Vectastain ABC kit (Vector Laboratories). Immunoreactivity was revealed using 3,3′-diaminobenzidine (DAB). Sections were counterstained with cresyl violet to visualize nonimmunoreactive neurons and cytoarchitectural boundaries. Specificity of the reaction was confirmed by processing negative control sections as described, but excluding the primary antibody. No immunostaining was observed in control sections. Each set of sections from each hemisphere were stained together to control for interexperiment variation. In this way, because lateral asymmetry was of interest, differences in fixation protocols, duration of fixation, storage conditions, and postmortem delay within each individual case would affect both hemispheres equally. Examples of immunohistochemical staining results are shown in .
Fig. 2 Examples of histological staining in chimpanzee primary motor cortex (Brodmann’s area 4). Staining for Nissl substance with cresyl violet (A), nonphosphorylated neurofilament protein with cresyl violet counterstain (B), and parvalbumin with cresyl (more ...)
Histological identification of the region of interest
We identified the region of interest for all subsequent quantitative measurements as primary motor cortex (Brodmann’s area 4) based on previous descriptions of the cytoarchitecture of this area in chimpanzees (Bailey et al., 1950
; Sherwood et al., 2003
). In brief, the primary motor cortex is distinguished by giant Betz cells in the lower portion of layer V, low overall cell density, large average cellular sizes, a poorly defined layer IV, and a diffuse border between layer VI and the subjacent white matter. The border between area 4 and area 3a, which usually occurs close to the fundus of the central sulcus, is recognized by the development of a well-defined granular layer IV. Although the border between area 4 and premotor cortex (area 6) may occur along the convexity of the precentral gyrus, we restricted our analyses to the portion of area 4 on the anterior bank of the central sulcus ().
Fig. 3 The part of primary motor cortex (Brodmann’s area 4) that was sampled for quantitative measurements. Only the intrasulcal portion of the precentral gyrus was analyzed (A). Arrowheads illustrate the typical borders used in defining the region of (more ...)
Measurement of relative layer thickness
Using a portion of the precentral gyrus where cortical layers were most easily distinguishable and not obscured by tangential sectioning, three sections spaced 400 μm apart were selected to measure the relative thickness of cortical layers. Measurement sites were positioned to fall along a part of the anterior bank of the central sulcus that was not folded at the crown of a gyrus or within the depth of a fundus. At each measurement site a reference contour was drawn from the pial surface to the white matter interface following the radial orientation of cortical mini-columns at low magnification (4×) using a computerized stereology and morphometry system consisting of a Zeiss Axioplan 2 photomicroscope equipped with a Ludl XY motorized stage, Heidenhain z-axis encoder, an Optronics MicroFire color videocamera, a Dell PC workstation, and StereoInvestigator software v. 6 (MBF Bioscience, Williston, VT). The length of each cortical layer was then measured along the radial guideline. We segmented the primary motor cortex along the most obvious layer boundaries: I, II/III, and V/VI (). In general, these laminar subdivisions correspond to functional differences in connectivity patterns. Molecular layer I contains mainly apical dendrites and horizontally oriented axons. Neurons in supragranular cortical layers II/III are involved in corticocortical integrative processes and have axonal projections to other ipsi- and contralateral cortical areas. Infragranular layers V/VI are comprised of neurons participating in corticofugal systems projecting to the spinal cord, brainstem, striatum, and thalamus.
The border between layers III and V was identified by the presence of a poorly developed, yet detectable, inner granular layer IV. In each section measurements were performed at two locations spaced 2 mm apart. At each measurement site the fraction of the total cortical thickness occupied by the width of each laminar segment was calculated as the relative layer thickness. An average relative layer thickness for each hemisphere was obtained from these measurements.
Areal fraction of Nissl-stained tissue
We quantified the areal fraction (AF) of tissue comprised of Nissl-stained cell bodies of neurons, glia, and endothelial cells in layers II/III and V/VI from high-resolution images. The AF represents the proportion of stained cellular profiles that project onto a two-dimensional measuring plane. Using the same three Nissl-stained sections used for analysis of relative layer thickness, digital images were collected using fractionator sampling as implemented by the StereoInvestigator system. First, contours were drawn around layers II/III and V/VI at low magnification. Then a fractionator sampling design (grid spacing of 600 × 800 μm for layer II/III; 800 × 800 μm for layer V/VI) was used to obtain a series of 8-bit grayscale image frames in a systematic random fashion with a 20× (0.50 N.A.) Plan-Neofluar objective lens. Prior to collection of image frames in each section, the exposure of the digital camera was standardized to an average target intensity of 70%. Images covered 440 × 587 μm and were 1,600 × 1,200 pixels in size, yielding a resolution of 0.37 pixels per μm. Image frame acquisition was monitored during fractionator sampling and all images that fell outside of the laminar region of interest boundaries were omitted from further processing. On average, 24.3 ± 4.5 (mean ± SD) image frames representing each laminar region of interest were collected for AF analysis. To measure the AF, images were processed in ImageJ software v. 1.32j with background subtraction using a rolling ball algorithm (Sternberger, 1983
), converted to binary by an automated threshold routine based on Rider and Calvard (1978)
, and dilation and erosion were applied to fill small holes representing light staining of cellular nuclei (). After converting the image to binary, the percentage of the measuring frame area occupied by pixels representing stained elements was calculated. The AF value for each hemisphere is the section-weighted mean of AF measurements calculated from all frames.
Fig. 4 Image conversion method for measurement of areal fraction (AF). An image frame of Nissl-staining in layer II/III is shown (A). Images were processed with background subtraction and conversion to binary to measure the proportion of spaced occupied by stained (more ...)
Quantification of numerical densities of cells and their volumes within layer II/III was performed using the StereoInvestigator stereology system. We focused on layer II/III for these analyses because many prior studies concerning histological asymmetry in the human neocortex have identified significant hemispheric differences specifically in the superficial cortical layers (Hayes and Lewis, 1995
; Buxhoeveden et al., 2001
; Hutsler, 2003
; Garcia et al., 2004
), suggesting that the functional role of these layers may be particularly relevant for hemispheric specialization.
Strict stereological quantification of total cell numbers is not feasible within the restricted region of hand representation of primary motor cortex because distinct boundaries cannot be established to separate this area from the motor representation of adjacent body parts. Nonetheless, asymmetries of cell-type-specific numerical densities can provide a useful reflection of hemispheric specialization. In this way, our data characterize asymmetries in the cellular composition per unit of tissue. Densities of neurons and glial cells were estimated from sections stained for Nissl substance. The density of PV-immunoreactive (ir) interneurons was also estimated. Cell densities in layer II/III were obtained using the optical disector with fractionator sampling following methods described in a previous study (Sherwood et al., 2007
). The same three sections used for AF analysis, or those in adjacent series, were quantified. After outlining the boundaries of layer II/III at low magnification, a set of optical disector frames (30 × 30 μm for neurons and glia; 90 × 90 μm for PV-ir interneurons) were placed in a systematic random fashion to yield ≈30 frames per section. Disector analysis was performed under Koehler illumination using a 63× objective (Zeiss Plan-Apochromat, N.A. 1.4). The thickness of optical disectors was set to 6 μm to allow for a minimum 2-μm guard zone on either side of the section after z
-axis collapse from histological processing. Cellular densities (Nv
) were derived from these stereologic counts and corrected for shrinkage from histological processing by the number-weighted mean section thickness. On average, the coefficient of error (Schmitz and Hof, 2000
) of measurements was 0.07 ± 0.03 for neurons, 0.08 ± 0.03 for glial cells, and 0.11 ± 0.02 for PV-ir interneurons. These coefficients of error are somewhat larger than is common in stereologic studies because we restricted our counts to the part of the precentral gyrus where cortical layers were most easy to define. This amount of measurement error would be expected to reduce the probability of rejecting the null hypothesis in statistical tests; however, findings of statistical significance in spite of this error can be considered reliable.
Cellular volumes of neurons immunostained for non-phosphorylated neurofilament protein (NPNFP) and PV were estimated using the nucleator with a vertical design (Gundersen, 1988
). Neurons were selected for volume measurement systematic randomly by applying optical fractionator sampling in two sections. In this way, the distribution of cell volumes obtained comprises an unbiased representation of the total population. The vertical axis of the probe was a line running superior-to-inferior to the pial surface. The centroids of neurons within the inclusion boundaries of optical disectors were marked and two transect lines from randomly selected directions were centered at the marker and superimposed over the neuron. The intersection of each line with the outer surface of the neuronal soma was marked and cellular volume was measured based on the nucleator principle. Because it was not feasible to perform isotropic-uniform-random sectioning in these rare behaviorally characterized chimpanzee materials, our mean cellular volume estimates contain a degree of bias due to a preferred sectioning orientation. However, due to normal variations in the orientation of the tissue in our preparations, not all cells were cut along the same axis, thereby generating a degree of randomness in the sample. Furthermore, coronal and sagittal sections have been shown to yield comparable results to isotropic-uniform-random sections using this probe (Schmitz et al., 1999
). In each hemisphere for each individual, 73.1 ± 13.6 cell soma volumes were sampled for NPNFP-ir neurons and 43.6 ± 16.6 for PV-ir interneurons. Mean cell volume was calculated for each hemisphere.
Lateral asymmetries in the various anatomical measurements were quantified by calculating an asymmetry quotient (AQ) using the formula: AQ = (R − L)/((R + L) × 0.5). Positive AQ values signify right hemisphere dominance, negative values signify left hemisphere dominance, and zero denotes symmetry. The absolute value of the AQ indicates the degree of asymmetry. provides AQ values for all neuroanatomical variables in each chimpanzee subject. Mann–Whitney U-tests did not reveal significant differences between sexes for HI score or any neuroanatomical AQ measurement; therefore, sexes were pooled in subsequent analyses. Furthermore, because none of the behavioral or neuroanatomical measurements showed a correlation with age, it was also not considered in the analysis. Nonparametric Spearman rank order correlations were used to examine associations with handedness because coordinated bimanual HI data was not normally distributed. We applied a sequential Bonferroni adjustment on a per hypothesis basis to adjust α for multiple comparisons in correlation and t-test analyses. However, because type II error is increased by this method, we report statistical significance at α′ = 0.1.
Forward stepwise multiple regression analysis was used to examine whether the combination of various anatomical AQ variables could predict variation in HI score. The forward stepwise approach sequentially selects the most highly correlated independent variable, removes the associated variance in the dependent, then enters further independents into the model which most correlate with the remaining variance in the dependent until selection of an additional independent does not increase r2 by a significant amount (P > 0.05). The final model includes a reduced number of predictor variables which collectively make the strongest, uncorrelated contributions to explaining variation in the dependent. Assumptions of multiple regression analysis were checked and each independent predictor variable was normally distributed as determined by Shapiro–Wilk’s W-tests. Furthermore, after multiple regressions, plots of studentized residuals versus unstandardized predicted values did not show nonlinearity or heteroscedasticity. Finally, binomial logistic regression was used to explore relationships between anatomical predictors and categorical handedness classification. For analyses that did not involve Bonferroni correction, statistical significance is reported at α = 0.05 (two-tailed).
Photomicrographs were obtained using an Optronics MicroFire digital camera mounted on a Zeiss Axioplan 2 microscope. Brightness and contrast of images were adjusted using Adobe Photoshop 6.0 software (San Jose, CA). Adobe Illustrator 8.0 was used for assembling and labeling figures.