Original Articles
Three-dimensional endocranial shape variation in the modern Japanese population
2015 Volume 123 Issue 3 Pages 185-191
Details
2015 Volume 123 Issue 3 Pages 185-191
Quantifying variation in human endocranial shape is important for interpreting the morphogenetic mechanisms of endocranial morphologies, as endocranial morphology has emerged from modification of the ontogenetic processes in the course of human evolution. We therefore analyzed patterns of morphological variability in endocranial shape among the modern Japanese population using landmark-based geometric morphometrics. After generating virtual endocasts of cranial specimens based on computed tomography scans, we defined a total of 171 conventional anatomical and sliding semi-landmarks on the endocranial surface. The brachycephalic/dolichocephalic tendency was the most frequently identified endocranial shape variation. In addition, we found that a smaller endocranium tended to be associated with a relatively larger cerebellar region accompanied by a flat, depressed parietal region and a more superiorly located frontal pole. Asymmetric shape variability possibly resulting from petalia was also observed, indicating that global brain asymmetry is related to endocranial shape. The present description of endocranial shape variability may contribute to the comparative understanding of the evolution of the endocast morphology of fossil hominins.
As soft tissue such as brain is not fossilized, endocranial morphology is the only useful source of information known for estimating the brain morphology of fossil hominins. To clarify the evolution of the hominin brain, researchers have analyzed variations in endocast morphology between different species in the human lineage. Historically, research has focused on endocranial volume (as a proxy of brain size) (Falk, 2012); however, in addition to size, brain evolution or encephalization is also a matter of structure and organization. Therefore, to document evolutionary changes in brain structure, efforts have been made to identify sulcus patterns on the internal surface of fossil braincases (e.g. Holloway et al., 2004; Holloway, 2008). However, even though imprints of sulci and gyri extracted from the crania of human children (Zollikofer and Ponce de León, 2013) and non-human primates such as macaques (Kobayashi et al., 2014) are some-what pronounced, identification of cortical features on the endocranial surface is particularly difficult in those extracted from adult human crania.
Due to the advent of geometric morphometric techniques, detailed analyses of morphological variability in endocranial shape have recently been carried out. Geometric morphometrics comprises a set of methods that enable the statistical investigation of biological shape variations based on digitized homologous landmarks (Bookstein, 1991; Slice, 2005). After Procrustes superimposition, size and shape variations among specimens are quantified based on multivariate statistical techniques such as principal component analysis (PCA), allowing extraction and comparisons of covariation in endocranial shape. Using geometric morphometry, Bruner et al. (2003) attempted to analyze the endocranial morphology of fossil hominins. Morphological variations in sagittally projected endocasts (Bruner, 2004) and three-dimensional (3D) basicranial shape (anterior, middle, and posterior cranial fossae) (Bastir et al., 2008; Bruner and Ripani, 2008) based on anatomical landmarks have also been investigated. Bienvenu et al. (2011) studied 3D endocranial variations in humans and great apes based on 37 digitized anatomical landmarks on the entire endocranium. However, since imprints of sulci and gyri on the endocranial surface are difficult to identify, and hence only a few anatomical landmarks are definable on the endocranial vault, the landmarks used in that study were concentrated on the basicranium.
More recently, the semi-landmark method has been widely used for analyses of surface morphology that has limited definable landmarks such as cranial or endocranial vault morphology (Bookstein, 1997; Gunz et al., 2005, 2009). In this method, semi-landmarks on curves or surfaces are measured on a template specimen, projected onto all other specimens in the sample, and subsequently allowed to slide so as to minimize the thin-plate spline bending energy between each specimen and the mean shape in order to establish the geometric correspondence of the semi-landmarks (Gunz and Mitteroecker, 2013). This method has recently been used to provide detailed descriptions of ontogenetic endocranial shape changes in humans (Neubauer et al., 2009) and to compare the growth trajectory of humans with that of chimpanzees (Neubauer et al., 2010) and Neanderthals (Gunz et al., 2010, 2012). The semi-landmark method was also applied by Bastir et al. (2011) to clarify detailed differences in basicranial morphology between Neanderthals and fossil and extant humans. However, to the best of our knowledge, those are the only studies that have carried out detailed 3D shape analyses of the human endocranium. The patterns of shape variation of the adult human endocranium remain to be clarified to better interpret morphogenetic mechanisms of endocranial morphologies, as the variation of the endocranial morphology has emerged from modification of ontogenetic processes in the course of human evolution.
In order to clarify normal patterns of human endocranial variability, the present study aims to provide a detailed description of endocranial shape variations among the modern Japanese population using the landmark-based geometric morphometric method. Specifically, we test hypotheses regarding the size–shape relationship, sexual dimorphism, and endocranial asymmetry. A basic description of endocranial shape variability may provide a fundamental framework for a comparative understanding of morphological similarities and differences of fossil hominin endocasts with those of humans.
We investigated endocranial shape variability in a total of 56 crania (23 female, 33 male) from the modern Japanese population housed at Kyoto University. Only crania that have not been cut transversely for the removal of the brain were used for this study. Each cranium was scanned using a helical computed tomography scanner in either the Laboratory of Physical Anthropology at Kyoto University or the Laboratory of Evolutionary Biomechanics at Keio University, Japan. Cross-sectional images were reconstructed at 0.5 mm intervals with a pixel size of 0.5 or 0.468 mm. The images were then transferred to medical imaging software (Analyze 9.0; Mayo Clinic, Biomedical Imaging Resource, Rochester, MN, USA), and a surface model of the cranium was generated as a triangular mesh model using the marching cube method. The outer surface of the cranium was then removed, and holes on the inner surface were filled using reverse engineering software (RapidForm 2006; INUS Technology, Seoul, Korea) in order to obtain a virtual endocast. The mesh surface was then regenerated based on its curvature flow using the same software. The 3D coordinates of midsagittal landmarks and midpoints of bilateral landmarks were digitized, also using the same software, and the midsagittal plane was calculated using the least-squares method.
To quantify the endocranial morphology of each specimen, we digitized the 16 anatomical landmarks on the endocranial surface of each cranium, as shown in Figure 1A. In addition, by approximating the curves using a seventh-order Bezier curve (Morita et al., 2013), we calculated the following: (1) 14 equally spaced points along the midsagittal curve between the foramen caecum (#1) and the internal occipital protuberance (#2); (2) 8 equally spaced points (4 on each side) along the anteroinferior border of the anterior cranial fossa between the foramen caecum (#1) and the most lateral point of the posterior border of the lesser wing of sphenoid (#6); and (3) 6 equally spaced points (3 on each side) along the lower border of the sulcus sinus transversi between the internal occipital protuberance (#2) and the intersection of the lower border of the sulcus sinus sigmoidei and transversi (#9). Therefore, in total 44 landmarks were extracted as non-sliding landmarks for each endocranium.
(A) Anatomical landmarks and (B) lines approximated using Bezier curves.
Based on the shortest paths between pairs of anatomical landmarks (Morita et al., 2013), we defined semi-sliding landmarks on one specimen chosen as a template (Figure 2). Specifically, we calculated the shortest paths connecting pairs of non-sliding landmarks. Along these paths, we obtained 127 equally spaced points, resulting in a total of 171 landmarks. The positions of all landmarks were symmetrized (Zollikofer and Ponce de León, 2002) to eliminate any possible asymmetric component of shape variation in the template specimen.
Anatomical landmarks and semi-landmarks defined for geometric morphometric analysis of endocranial shape. The endocranial shape is visualized using wireframe connecting landmarks.
We used this symmetrized template configuration to determine the semi-landmark locations on the other endocasts. First, a thin-plate spline function defining a mapping from the template configuration to the target endocast was created based on the 44 non-sliding landmarks, and the sliding landmarks on the template configuration were transferred to a target endocast using the mapping function. The sliding landmarks were then projected onto and slid along the endocranial surface of the target endocast in order to minimize the bending energy of the deformation from the template configuration to the target endocast. This process is iterated until the solution converges. Each of the sliding landmarks slides along a plane tangential to the endocranial surface; therefore, an offset exists from the surface and is thus projected back to the surface after each iteration (see Gunz et al., 2005) for a detailed calculation method). In this study, Templand in the EVAN Toolbox (www.evan-society.org) was used to calculate the location of the sliding landmarks.
The endocranial form (i.e. size and shape) represented by the locations of the non-sliding and sliding landmarks was then analyzed using morphologika geometric morphometric software version 2.5 (O’Higgins and Jones, 1998). First, the landmark coordinates were superimposed by the method of least squares (Procrustes superimposition). Principal components (PCs) of endocranial form variations among the specimens were then computed in Procrustes form space (size–shape space) based on the variance–covariance matrix of the Procrustes residuals augmented with the logarithm of centroid size (O’Higgins and Jones, 1998; O’Higgins, 2000). To investigate allometric shape variation, correlations between the PC scores and the natural log of the centroid size (ln CS) were tested using Statistica 10 (Statsoft Inc., Tulsa, OK, USA). In order to test for significant differences in PC scores between female and male crania, multivariate analyses of variance (MANOVA) were conducted using the same software. To investigate possible endocranial shape asymmetry, we mirrored the original specimens using reflected relabeling (Gunz et al., 2009), and significant differences in PC scores between the original and reflected specimens were tested in the same fashion.
Equidistant landmarks are usually allowed to slide along curves or surfaces in geometric morphometrics since equidistance may not necessarily lead to geometrical correspondence of the landmarks across specimens (Bookstein, 1997; Gunz et al., 2005), resulting in possible artifacts in representing morphological variability. However, in the present study, the position of the equidistant landmarks defined along the curves was confirmed to move subtly before and after sliding, having very little influence on the extracted morphological variability. We therefore did not slide the equidistant landmarks along the Bezier curves here.
The results of PCA of the morphological variability in the endocasts of the modern Japanese population are presented in Figure 3 as plots of the first principal component (PC1) versus PC2, and as PC3 versus PC4. Based on a scree plot and cumulative variance, we considered the first four PCs to be dominant and thus retained them for further analyses. The remaining components were interpreted as secondary components showing low morphological variation as represented in the endocast. The first four PCs accounted for 65.9% of the variation (42.7%, 10.9%, 6.9%, and 5.4% for PC1, PC2, PC3, and PC4, respectively).
The results of principal component (PC) analysis. (A) PC1 (x-axis) versus PC2 (y-axis), and (B) PC3 (x-axis) versus PC4 (y-axis). White circle = female; black circle = male.
Only PC1 exhibits a significant positive linear relationship with ln CS (r = 0.997, P = 0.000). Plots of the PC scores in Figure 3 showed clear separation between female and male crania along the PC1 axis, but not along the higher PC axes, indicating that size-related variation due to sexual dimorphism is represented by PC1. Significant differences between the sexes were detected by MANOVA (Wilk’s lambda = 0.494, F = 13.05, P = 0.000).
Figure 4 shows the 3D shape variabilities along PC1, PC2, PC3, and PC4 by warping the endocranial shape represented by the wireframe that connected the landmarks. Here the landmark coordinates were normalized by CS for size-independent shape comparison. The shape difference along PC1 was much smaller than those represented by other PCs because the form variation represented by PC1 is mostly size variation.
Variations in endocranial shape represented by the first four PCs. Variations are visualized with 3D deformation of wireframe connecting landmarks. Solid line: PC1 = 0.1, PC2 = 0.05, PC3 = 0.04, PC4 = 0.04. Dotted line: PC1 = −0.1, PC2 = −0.05, PC3 = −0.04, PC4 = −0.04.
On the other hand, with increasing PC2, a relative contraction of the cranial length, a relative elongation of the cranial breadth, and a relative elongation of the cranial height were observed, as shown in Figure 4. Furthermore, the foramen magnum and the temporal pole were more inferiorly and posteriorly positioned with an increase in PC2. The relative position of the internal occipital protuberance, corresponding to the boundary between the occipital lobe and the cerebellum, was virtually unchanged. Therefore, in a dolichocephalic cranium, the occipital robe is specifically backward-protruding, whereas in a brachycephalic cranium, the cerebellum is specifically enlarged.
With increasing PC3, a relative elongation of the cranial length with a relative contraction of the cranial breadth was observed, indicating that the endocranium tends to be dolichocephalic. Furthermore, although the relative position of the temporal region remained unchanged with an increase in PC3, upward displacement of the frontal pole, downward displacement of the foramen magnum, relative depression of the parietal region, and relative contraction of the cranial breadth were observed.
With decreasing PC4, the left parietooccipital and the right fronttemporal regions were more lateroposteriorly and lateroanteriorly protruded, respectively, indicating the tendency for bilateral asymmetry along PC4. It was also noted that the endocranial contour of the forehead rises less vertically, and that the frontoparietal region is less upwardly projected with a decrease in PC4. Significant differences between the original and reflected specimens were detected by MANOVA (Wilk’s lambda = 0.406, F = 9.35, P = 0.000).
In the present study, the morphological variability of endocranial form among modern Japanese populations was successfully extracted and visualized using a landmark-based geometric morphometric technique. The brachycephalic/dolichocephalic tendency (PC2) was found to be the most predominant endocranial shape variation in modern Japanese crania. Morita et al. (2014) studied the morphological variability of the ectocranial surface using the same 56 Japanese crania used in this study; they also found that the brachycephalic/dolichocephalic tendency was the most predominant shape variation. Therefore, variation in the shape of endocranial surfaces generally tends to correspond with that of ectocranial surfaces. However, this may not be true for apes or fossil hominins due to ectocranial superstructures such as crests and larger sinuses.
The shape variation represented by PC3 suggested that endocrania with a relatively larger cerebellum have a relatively flat, depressed parietal region and a superiorly located frontal pole, and vice versa. An ontogenetic study of human endocranial shape demonstrated that changes in early childhood are characterized by concomitant enlargement of both parietal and cerebellar regions (Neubauer et al., 2009). Therefore, we originally suspected that a pattern of endocranial shape variability would comprise enlargement of the both regions. However, that was not the case in the present study. Although this covariation pattern represents a possible direction for future study, we are currently unable to provide a reasonable explanation for this discrepancy.
| Number | Anatomical landmark | Type |
|---|---|---|
| 1 | Foramen caecum | m |
| 2 | Internal occipital protuberance | m |
| 3 | Opisthion | m |
| 4 | Basion | m |
| 5 | End of the processus clinoideus anterior | b |
| 6 | Most lateral point of the posterior border of the lesser wing of sphenoid | b |
| 7 | Petrosal apex | b |
| 8 | Intersection of the crista pyramidis and the upper edge of the sulcus sinus transversi | b |
| 9 | Intersection of the lower border of the sulcus sinus sigmoidei and transversi | b |
| 10 | Most lateral point of the margin of the foramen magnum | b |
Note: m = midsagittal landmark; b = bilateral landmark
The shape variability of bilateral symmetry along PC4 generally corresponds with petalia. Petalia is a significant asymmetry of the brain in humans (Balzeau et al., 2012), with protrusion of the right frontal and left occipital poles relative to their contralateral side (Toga and Thompson, 2003). Although the contribution ratio of PC4 is about 5%, the present study demonstrated that innate brain asymmetry is represented by the shape of the human endocranium. Therefore, global brain morphology generally corresponds well to the surface of the cranial endocast that encloses the brain; however, this association seems to be weak. Based on magnetic resonance imaging, it is generally accepted that the male brain shows a greater degree of asymmetry than the female brain (Hiscock et al., 1994; Shaywitz et al., 1995; Zilles et al., 1996; Good et al., 2001; Luders and Toga, 2010); the same tendency has also been reported for the Japanese brain (Zilles et al., 2001). If sex-based differences of the brain are represented in the endocranial shape, the male crania should be separated from female crania along the PC4 axis. However, no clear distinction between the sexes in endocranial shape was found in the present study. Therefore, the correlation between brain and endocranial shape is likely weak, and thus, sex-based differences in brain asymmetry are not well represented.
The present description of endocranial shape variability may serve as a fundamental framework for a comparative understanding of fossil hominin endocasts. This information is also necessary for the anatomically accurate statistical interpolation of missing parts in fossil endocasts (Gunz et al., 2009; Ogihara et al., 2015). The relatively small number of specimens in this study does represent a limitation. Furthermore, our sample was not geographically diverse. The fact that increases in stature are correlated to brachycephalization in Japan but to dolichocephalization in Europe (Kouchi, 2000) suggests that the pattern of morphological variability of the modern human cranium may be affected by environmental factors. Including cranial specimens from different populations for analysis is a crucial issue for future research. Lastly, although the present study successfully observed small directional asymmetric deviation in the human endocranial shape, the statistical treatment of asymmetry is actually known to be very difficult requiring special consideration (Mardia et al., 2000; Zollikofer and Ponce de León, 2002). Directional asymmetry of the human endocranium needs to be confirmed in further studies with a larger number of cases and more rigorous statistical treatment.
We wish to express our sincere gratitude to Prof. Takeru Akazawa (Kochi Institute of Technology) for his continuous guidance and support throughout the course of the present study. We also sincerely thank two anonymous reviewers for their constructive and thoughtful comments. This study was supported by a Grant-in-Aid for Scientific Research on Innovative Areas (#22101006) ‘Replacement of Neanderthals by Modern Humans: Testing Evolutionary Models of Learning’ from the Japanese Ministry of Education, Culture, Sports, Science and Technology.
