SEARCH

SEARCH BY CITATION

Keywords:

  • allometry;
  • bird;
  • brain;
  • geometric morphometrics;
  • orbital shape;
  • partial least-squares

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

There is wide variation in brain shape among birds. Differences in brain dimensions reflect species-specific sensory capacities and behavioral repertoires that are shaped by environmental and biological factors during evolution. Most previous studies aimed at defining factors impacting brain shape have used volumetric or linear measurements. However, few have explored the quantitative indices of three-dimensional (3D) brain geometry that are absolutely imperative to understanding avian evolutionary history. This study aimed: (i) to explore the relationship between brain shape and overall brain size; and (ii) to assess the relationship between brain shape and orbital shape. Avian brain endocasts were reconstructed from computed tomography images and analyzed using 3D geometric morphometrics. Principal component analysis revealed dominant regional variations in avian brain shape and shape correlations between the telencephalon and cerebellum, between the cerebellum and myelencephalon, and between the diencephalon and optic tectum. Brain shape changes relative to total brain size were determined by multivariate regression analysis. Larger brain size was associated with a relatively slender telencephalon and differences in brain orientation. The correlation between brain shape and orbital shape was assessed by two-block partial least-squares analysis. Relatively round brains with a ventrally flexed brain base were associated with rounder orbits, while narrower brains with a flat brain base were associated with more elongated orbits. The shapes of functionally associated avian brain regions are correlated, and orbital size and shape are dominant factors influencing the overall shape of the avian brain.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Brain morphology reflects the behavioral repertoire, cognitive capacity and evolutionary history of a species. The relative importance of specific sensory modalities and cognitive processes is reflected by the relative volumes and sizes of brain regions that receive and process relevant stimuli (Jerison, 1973; Dubbeldam, 1998; Butler & Hodos, 2005; Striedter, 2005). Many studies have explored the relationship between brain morphology and behavior or sensory capacity using morphometric approaches. Volumetric methods, which typically compare total with regional brain volumes (Portmann, 1946; Jerison, 1973; Boire & Baron, 1994; Iwaniuk et al. 2005), and linear measurement methods have been used to describe the morphological characteristics of the avian brain (Cobb, 1960a; Ashwell & Scofield, 2008; Milner & Walsh, 2009; Ksepka et al. 2012). Portmann (1946, 1947), Boire & Baron (1994) and Rehkämper et al. (1991) concluded from volumetric and histological analyses that most of the variation in avian brain size is due to the size of the telencephalon. Thus, it would appear that the relative volume of the telencephalon and constituent regions (e.g. hyperpallium or eminentia saggitalis) is a major determinant of total brain volume and shape. More specifically, regional volume changes and the specific correlations between the volumes of different regions (Barton & Harvey, 2000; Barton et al. 2003; Whiting & Barton, 2003; Iwaniuk et al. 2004) result in regional differences in size that collectively determine the size and shape of the whole brain. Although volumetric and linear methods are valuable for revealing the absolute size or the size ratio of two brain regions, these measurement methods do not take into account overall brain shape and cannot describe brain shape quantitatively. Therefore, analyzing shape is critical for understanding how the brain evolves and the constraints imposed by other organs, such as the eyes. We analyzed brain shapes across Aves using three-dimensional (3D) geometric morphometrics. We focused on correlations between shape and size (shape allometry), and covariation between brain shape and orbital shape, as these are considered to be the most important factors determining variation in brain shape between species and thus the most useful for revealing the mechanisms that determine avian brain shape.

Correlation between shape and size (shape allometry)

This study aimed to determine the allometric changes that occur in the avian brain. When comparing biological organs, it is necessary to assess similarities or differences among taxa after size is taken into account (Jungers et al. 1995), because size influences and constrains shape (Gould, 1966). Indeed, lack of allometric information can lead to overestimation or underestimation of morphological changes. Many allometric relationships have been described in other animals; however, only a few allometric studies have investigated avian brain shape.

Marugán-Lobón & Buscalioni (2006, 2009) and Marugán-Lobón (2010) described a series of two-dimensional (2D) coordinates to define avian cranial shape, which approximately reflects brain shape, and subjected these coordinates to Procrustes, principal component (PC) and two-block partial least-squares (2B-PLS) analyses. They related changes in brain morphology to the dimensions of the cranial base and foramen magnum, and also found that morphological changes in the avian neurocranium were related to changes in cranial size. However, the precise quantitative relationship was not clearly defined. It is likely that these studies did not detect critical allometric shape changes because data were not collected directly from the brain cavity and because all measures were 2D. Some previous studies have suggested that the position of the foramen magnum varies according to cranial size in Aves (Kulemeyer et al. 2009; Marugán-Lobón & Buscalioni, 2009), likely because the position of the foramen magnum also correlates with brain shape and orientation, which are in turn related to cranial size (Duijm, 1951). Kulemeyer et al. (2009) suggested that the position of the foramen magnum may be associated with brain orientation and head posture. Clarifying the relationship between brain orientation and brain size will provide insight into the functional morphology of the head and neck in Aves.

Covariation between brain shape and orbital shape

In Aves, the brain and unusually large eyes (orbit) fill a large portion of the neurocranium. Therefore, these organs are likely dominant influences on cranial size and shape (Jerison, 1973; Dubbeldam, 1998; Bhullar et al. 2012). Furthermore, because there is a strong correlation between brain and eye size in birds (Garamszegi et al. 2002; Burton, 2008), it is possible that brain shape and orbital shape are interdependent. Indeed, Duijm (1951) found a strong correlation between brain shape and the size and position of the eyes; therefore, the interdependence of brain shape and orbital size is a reasonable assumption. Nonetheless, there is a need to quantitatively compare brain shape and orbital shape.

In general, the major factor determining an animal's shape is its overall size, and this is likely also true in the case of brain shape. After considering the size effect, the eye is an important factor creating variation in avian brain shape (Dubbeldam, 1989). Cobb (1960b) suggests that the large eyes of birds have ‘pushed’ the brain backward, producing a shift in orientation and thus allowing the eyes more space within the small skull. Hence, it is expected that avian brain shape is affected by eye shape or orbital shape. Although several studies have investigated the relationship between the brain and eye size in birds (Garamszegi et al. 2002; Thomas et al. 2006; Burton, 2008), few have quantitatively examined the relationship between brain shape and orbital shape. Revealing the relationship between brain shape and orbital shape is indispensable for an understanding of the shape variation and evolution of the avian brain; therefore, we explored the patterns of covariation between brain shape and orbital shape.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Specimens and computed tomography (CT) scanning

Our sample consisted of 61 adult skull specimens of extant birds from 22 orders (Table 1). The specimens were loaned from the National Museum of Nature and Science, Japan (Tokyo, Japan), Yamashina Institute for Ornithology (Chiba, Japan) and private collections. All skulls except one were CT scanned at the Natural Museum of Nature and Science and Ehime University (Matsuyama, Japan) using a LaTheta micro-CT scanner (Hitachi Aloka Medical). Acquisition parameters were 50 kV, 0.5 mA and a slice thickness of 48–480 μm. An ostrich skull was subjected to medical multi-detector CT (Pratico; Hitachi Medical) at Kochi University (Kochi, Japan) using the following acquisition settings: 130 kV, 100 mA and a slice thickness of 1 mm. We subsequently prepared virtual brain endocasts from acquired CT images. Those endocasts are more suitable for acquiring the 3D morphological data than soft raw or fixed brains. Although some little features of the brain (such as cerebellar folding) are obscured on the endocast, avian endocasts are widely known to reflect brain morphology with considerable accuracy (Jerison, 1973; Iwaniuk & Nelson, 2002; Witmer et al. 2003; Ashwell & Scofield, 2008). Details of the methods used in preparing and examining the brain models are given by Kawabe et al. (2009). Hence, virtual brain endocasts acquired from CT images are hereafter referred to as avian brain models or simply brains.

Table 1. Species analyzed and brain CSs
OrderSpeciesCSOrderSpeciesCS
  1. CS, centroid size.

Accipitriformes Aegypius monachus 90.990Passeriformes Passer montanus 30.779
Falco tinnunculus 49.875 Turdus pallidus 38.604
Gypohierax angolensis 72.059 Acridotheres tristis 43.215
Milvus migrans 69.203 Corvus corone 69.185
Anseriformes Aythya marila 59.840 Corvus macrorhynchos 62.305
Anas poecilorhyncha 60.367 Corvus macrorhynchos 63.451
Anas platyrhynchos 61.287 Serinus canaria 25.062
Tadorna ferruginea 58.292Pelecaniformes Pelecanus onocrotalus 99.636
Bucerotiformes Ceratogymna albotibialis 72.779 Phalacrocorax carbo 76.410
Casuariiformes Dromaius novaehollandiae 102.269 Phalacrocorax capillatus 86.014
Charadriiformes Tringa stagnatilis 34.159 Phalacrocorax pelagicus 63.444
Rostratula benghalensis 39.395Phoenicopteriformes Phoenicopterus minor 65.521
Fratercula cirrhata 61.470Piciformes Picus awokera 46.658
Larus crassirostris 58.956Procellariiformes Calonectris leucomelas 56.533
Burhinus superciliaris 53.535 Calonectris leucomelas 53.543
Scolopax mira 47.754 Diomedea nigripes 83.012
Scolopax rusticola 45.441 Pelecanoides urinatrix 37.060
Ciconiiformes Ardea cinerea 71.329 Puffinus tenuirostris 56.693
Butorides striatus 48.538Psittaciformes Amazona ochrocephala 65.677
Eudocimus ruber 56.313 Ara macao 85.996
Nycticorax nycticorax 59.417 Melopsittacus undulatus 30.615
Platalea alba 67.688 Nymphicus hollandicus 40.809
Threskiornis melanocephalus 67.406 Psittacula alexandri 59.901
Columbiformes Streptopelia orientalis 41.181 Psittacus erithacus 70.140
Columba livia 43.117Sphenisciformes Aptenodytes patagonicus 105.288
Coraciiformes Alcedo atthis 27.942 Pygoscelis adeliae 85.005
Halcyon chloris 36.677Strigiformes Strix leptogrammica 68.701
Galliformes Phasianus versicolor 56.148Struthioniformes Struthio camelus 117.550
Gallus gallus domesticus 62.512Tinamiformes Eudromia elegans 43.982
Gaviiformes Gavia pacifica 67.059Upupiformes Upupa epops 32.741
Gruiformes Fulica atra 49.779   

3D geometric morphometrics

We digitized the 3D coordinates of 20 homologous landmarks and eight semi-landmarks (landmark 9, 10, 12, 13, 15, 21, 22 and 24) from the brain and lateral semicircular canal, and nine from the orbit (Fig. 1; Table 2). Those semi-landmarks were collected from a lateral view. Because the lateral semicircular canals provide a reliable guide for determining the normal head posture when they are horizontally arranged (Duijm, 1951; Witmer et al. 2003; Chatterjee & Templin, 2004; Taylor et al. 2009), the relative brain posture can be revealed using the semicircular canals as a standard reference for comparing the brain shape. Orbital coordinates were digitized from 58 specimens, and brain coordinates were digitized from all 61 specimens. Landmarks were digitized using amira (v 5.3.2; Mercury Computer Systems, San Diego, CA, USA). The resulting data set of 3D coordinates was subjected to generalized least-squares Procrustes analysis (GPA; Rohlf & Slice, 1990) using the morphoj software package (Klingenberg, 2011). In GPA, distances between homologous landmarks are minimized by translating, rotating and scaling all objects to a common reference. That is, the effects of size, position and orientation are eliminated so that remaining data reflect shape variation (Procrustes shape coordinates). Information about the absolute size of the specimen is preserved as centroid size (CS), which is calculated as the square root of the sum of squared distances of landmarks from their centroids (Bookstein, 1991). We set boundaries that passed through specific sites in brain regions, such as the telencephalon, diencephalon, optic tectum, cerebellum and myelencephalon (Fig. 1). Notably, even in a serial-sectioned brain, the boundaries of brain regions are difficult to delineate in some specimens (Boire & Baron, 1994; Iwaniuk et al. 2005), and our definition was based on the appearance of the brain endocast. Hence, our definition of regional boundaries does not completely coincide with that of histological boundaries. Yet, setting the boundaries of brain regions is essential when describing the changing pattern of external morphology of the brain or endocast.

Table 2. Landmarks used (Fig. 1) and anatomical descriptions (refer to Fig. 1) for profiled anatomical structures of (a) the brain and (b) the orbit
NumberAnatomical description
(a) Brain28 Landmarks
1Median anterior tip of olfactory bulb
2Median junction between telencephalon and cerebellum
3Median dorsal point of foramen magnum
4Median ventral point of foramen magnum
5Median junction between hypophysis and mesencephalon
6Median ventral tip of hypophysis
7Median junction between optic nerve and hypophysis
8Median junction between telencephalon and optic nerve
9Perpendicular at midpoint between landmarks 2 and 3 to dorsal margin of cerebellum in lateral view
10Perpendicular at midpoint between landmarks 4 and 5 to ventral margin of mesencephalon in lateral view
11, 20Most anterior tip of telencephalon, right and left
12, 21Perpendicular at midpoint between landmarks 11 (20) and 2 to dorsal margin of telencephalon in lateral view, right and left
13, 22Intersection of telencephalon, cerebellum and optic lobe, right and left
14, 23Most anterior point of optic lobe, right and left
15, 24Intersection of telencephalon, optic lobe and diencephalon, right and left
16, 25Most lateral point of the widest part of telencephalon, right and left
17, 26Most lateral point of the widest part of cerebellum, right and left
18, 27Most anterior end of lateral semicircular canal, right and left
19, 28Most posterior end of lateral semicircular canal, right and left
(b) Orbit9 Landmarks
8Median junction between telencephalon and optic nerve
29, 33Most distal point of the processus postorbitalis, right and left
30, 34Most dorso-lateral point of the os lacrimale, right and left
31, 35Most ventro-lateral point of the os lacrimale, right and left
32, 36Perpendicular at midpoint between landmarks B (F) and C (G) to dorsal margin of orbit, right and left
image

Figure 1. (A) The three-dimensional brain landmarks used for shape analysis as shown for the (A) dorsal (upper) and lateral (lower) views of the brain of Aegypius monachus, and (B) the lateral view of the skull of Corvus corone (description in Table 1).

Download figure to PowerPoint

Principal component analysis (PCA)

Procrustes shape coordinates were subjected to PCA to explore the patterns of major variation between avian brains. PCA summarizes the multidimensional information generated by the Procrustes alignment in linear combinations, which are the PCs. PCA was performed using morphoj, and Morphologika was used to illustrate 3D profiles. In all 3D multivariate analyses, the coordinate data of extremes were extracted and read into Morphologika. The coordinates were subjected to PCA using Morphologika without any alignments. The extracted 3D models correctly reflected the first extreme coordinates, which were extracted from the initial analysis by morphoj.

To explore the pattern of shape change in avian brain components, we divided the endocasts into what correspond to the major regions of the brain (telencephalon, diencephalon, optic tectum, cerebellum and myelencephalon; Fig. 2). We mainly focused on the anteroposterior or lateral elongation of each region, and described the magnitude and elongation pattern of the shape change.

image

Figure 2. The result of PCA. The first three principal components (PCs) of shape variation in the avian brain.

Download figure to PowerPoint

Shape allometry

We used independent contrasts (Felsenstein, 1985) to consider the phylogenetic structure of the data. The topology of the tree was adapted from Hackett et al. (2008), Suh et al. (2011) and Sibly et al. (2012). Branch lengths were set to one. The morphometrics program morphoj was used to produce a data set with independent contrasts on the shape and CS data (shape and CS contrast) for the subsequent regression analysis of shape change. To explore how shape covaries with size, multivariate regression of brain shape contrast onto CS contrast was performed (Fig. 4). A regression of the phylogenetically adjusted shape data and CS was performed to determine whether size alone was responsible for the differences in shape observed along the PC axes. Coefficients from the regression are vectors representing the correlations between shape change and size. Statistical significance was tested using permutation tests. P-values < 0.001 were considered statistically significant.

Exclusion of the size factor from brain shape variance

Because the result of PCA described above included size as a factor of allometric shape change, it is necessary to exclude allometric shape change from analysis to discuss the brain shape changes independent from size. To exclude the effect of size from the variance in brain shape change, the residual from multivariate regression analysis, which was already phylogenetically adjusted, was subjected to PCA. The residual from this regression was used as the ‘size-adjusted’ data set.

2B-PLS

The 2B-PLS method (Rohlf & Corti, 2000) is used to explore the pattern of covariation for linear combinations between two blocks of variables. The linear combinations are constructed so that the new variables account for as much of the covariation between the two original sets of variables as possible and to display the patterns of covariation between the two blocks (see Rohlf & Corti, 2000 for more information). Moreover, 2B-PLS is useful for exploring the covariation between two blocks of shape variables, or between one block with shape variables and another with any other type of continuous variable (see Marugán-Lobón, 2010 for details).

In the present study, 2B-PLS analysis was performed using morphoj to explore the covariation between shape data outlining the brain and the orbit. 2B-PLS searches for pairs of new explanatory factors (PLS dimensions), one for brain shape and one for orbital shape (singular warp), that maximize the covariance between the two blocks of data. The first pair of explanatory factors forms the first PLS dimension (PLS1) and explains the highest percentage of total covariance between the two blocks. Singular warp scores of brain shape and orbital shape can be plotted against each other. This visualizes changes in brain shape for a given change in orbital shape (Fig. 4).

To quantify the strength of covariation between blocks, we used the RV coefficient (Escoufier, 1973; Klingenberg, 2009), which can be interpreted as a multivariate generalization of the bivariate R2 value. The significance levels for the covariance between the blocks, and for the correlation between brain shape and orbital shape were evaluated using permutation tests (10 000 runs) on each PLS.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Patterns of shape diversification (unadjusted PCA)

The first three PCs from PCA accounted for 57.81% of the total shape variation and provided a reasonable approximation of the total shape variation (Figs 2and 3). Most explained variance was evenly distributed between the first two dimensions (PC1 and PC2 explained 23.10% and 21.73%, respectively, of the total variance). These two dimensions explained nearly half of the whole variation (44.83%). Because PC1, PC2 and PC3 were the only PC axes that accounted for more than 10% of variance (Fig. 2), these first three PCs will be the focus.

image

Figure 3. Variation in brain shape for each principal component (PC) score in unadjusted PCA. The lateral semicircular canals are shown by bold lines.

Download figure to PowerPoint

Shape variation along the PC1 axis involved variation in the telencephalon, characterized by both shrinking (positive PC) and expansion (negative PC), and it was characterized by ventrodorsal rotation in the posterior part of the brain (the cerebellum and myelencephalon; Fig. 3). Telencephalic shrinking moved the posterior portion of the telencephalon rostroventrally and the anterior portion rostrodorsally (Fig. 3). Shrinking the telencephalon produced dorsal rotation of the cerebellum and myelencephalon, elongated the brain base length posteriorly, making the brain base flatter. Expansion of the telencephalon resulted in ventral rotation of the cerebellum and myelencephalon and shortened the brain base and brain stem, making the brain base ventrally flexed. In addition, the cerebellum and myelencephalon elongated caudodorsally with shrinking of the telencephalon. On the other hand, the width of the diencephalon and optic tectum were not markedly changed by shape variation along the PC1 axis. Other regions of the brain responded to changes in telencephalon size. A brain with a large telencephalon had a relatively smaller cerebellum and myelencephalon, while a brain with a small telencephalon had a relatively larger cerebellum and myelencephalon. As a whole, PC1 primarily distinguished wide and rounded brains from narrow brains. A rounded brain with a ventrally flexed brain base had more space for the eyes under the telencephalon, while a narrower brain with a flat brain base had more space for the eyes rostral to the telencephalon.

The PC2 axis mainly corresponded to anteroposterior elongation or expansion of the telencephalon and flattening of the cerebellum and myelencephalon (Fig. 3). Accordingly, PC2 distinguished elongated and flat brains (positive PC) from round and ventrally arching brains (negative PC). With increased PC2 score, the rostral end of the telencephalon tended to widen and the posterior portion was pushed further posteriorly. In addition, the other parts of the brain became rounder with increasing PC2 score. Both the diencephalon and optic tectum narrow with increasing PC2 score. Similar to the effects of changes in the PC1 axis, the anteroposteriorly elongated brain created a space rostral to the telencephalon, while the round brain with a ventrally flexed brain base created a space under the telencephalon.

PC3 explained 12.99% of total variance. Shape variation along the PC3 axis was mainly associated with the ratio of height to length (Fig. 3). The specimens with positive PC3 scores had ventrodorsally longer and anteroposteriorly shorter brains, whereas those with negative PC3 scores had ventrodorsally shortened and anteroposteriorly elongated brains. Ventrodorsally longer brains (positive PC) had a relatively rostrocaudally shortened optic tectum, cerebellum and myelencephalon, but a wider diencephalon in the anteroposterior direction. The telencephalic anteroposterior elongation was related to the dorsal rotation of the hindbrain in PC1 and PC2, whereas the posterior part of the brain rotated ventrally with the anteroposterior elongation of the telencephalon in PC3. Both the diencephalon and optic tectum concertedly widen with increasing PC3 score.

As described above, the first two PCs explained nearly identical shape variance, and most of the explained variance was evenly distributed between them. Thus, if more specimens were added or removed from the present sample, those two dimensions could replace each other. In addition, only the shape change pattern in the diencephalon and optic tectum differed between PC1 and PC2. Shape changes according to increase in both PC scores were related to the anteroposterior elongation of the telencephalon and dorsal rotation of the posterior part of the brain. Hence, the two dimensions are better interpreted together as a more comprehensive descriptor of the variation in avian brain shape.

Within the graph of PC1 and PC2 (Fig. 2), narrower and longer brains (such as Phalacrocorax and Gallus) were plotted on the upper right (+PC1 and +PC2), while rounded and upwardly tilted brains (such as Falco and Scolopax) are plotted on the lower left (−PC1 and −PC2). In addition, the brain position relative to the lateral semicircular canal changed dramatically in those PCs (Fig. 3). In the positive PC group, the myelencephalon was nearly parallel to the lateral semicircular canal; however, the myelencephalon and lateral semicircular canal were nearly perpendicular to the foramen magnum plane. The lateral semicircular canal was nearly parallel to the plane of the foramen magnum in the brain belonging to the negative PC group.

Shape allometry

Multivariate regression analysis using the contrast scores also revealed shape changes as the brain grew or diminished in size (Fig. 4). In other words, the avian brain showed significant (P < 0.001) shape allometry. Regression analysis of shape against size explained 8.81% of the shape variation. As the avian brain gets larger, it elongates rostrocaudally mainly because of caudodorsal elongation of the cerebellum and myelencephalon. These allometric shape changes corresponded to positive change along the PC axes, particularly the first two axes from unadjusted PCA. The dominant shape changes detected by unadjusted PCA clearly reflect the effect of size.

image

Figure 4. Results of multivariate regression analysis. (A) A scatter plot showing the regression of brain shape contrast vs. log centroid size (CS) contrast. Brain schematics illustrate negative and positive extremes. The lateral semicircular canals are shown by bold lines. (B) Sequence of schematic images of estimated brain shape as the brain is enlarged (to the right). The consensus shape is at the center, and the side images are negative and positive extremes of the horizontal axis.

Download figure to PowerPoint

Changes in shape related to the absolute brain size are summarized in Fig. 4. An increase in the absolute brain size resulted in: (i) anteroposterior elongation; (ii) caudodorsal rotation of the posterior part of the brain, including the foramen magnum plane; (iii) dorsoposterior extension of the medulla oblongata and brain base; and (iv) transformation of the brain base into a flatter structure, instead of being ventrally flexed. As a result, compared with populations having low scores, populations with high scores on the shape axes of Fig. 4 (i.e. species with large brains) showed ventroanterior rotation of the brain relative to horizontally positioned lateral semicircular canals, resulting in a relatively slender and anteriorly inclined brain. This pattern was analogous to that observed by increasing the unadjusted PC1 and PC2 scores in PCA.

Size-adjusted shape change (size-adjusted PCA)

The first three PCs from the size-adjusted data set accounted for 57.68% of the total shape variation and provided a reasonable approximation of the total shape variation (PC1, PC2, PC3 explained 24.93%, 20.68% and 12.07% of the total variance, respectively). These three PCs were the only PC axes that accounted for more than 10% of variance. Most explained variance is evenly distributed between the first two dimensions. As is the case in unadjusted PCA, PC1 and PC2 could switch places when adding (or removing) more specimens into (or from) the present data set. Hence, both shape changes according to the changing of the first two PC scores should be considered equally significant.

With an increase in PC1 score, the telencephalon/cerebellum pair shrank, the cerebellum/myelencephalon rotated ventrally, and the diencephalon/optic tectum moved ventrocaudally (Fig. 5). In the second dimension, the telencephalon/cerebellum also shrank with an increasing PC score; however, the cerebellum/myelencephalon rotated dorsally and the diencephalon/optic tectum showed slight change (Fig. 5). Shrinking of the telencephalon was also accompanied with the shrinking of the cerebellum in PC3; however, other parts of the brain showed major changes (Fig. 5). In all PCs, the telencephalic expansion or shrinking caused caudal expansion or shrinking of the cerebellum and shortened or lengthened the brain base, which includes the entire length of the brain, and brain stem. In addition, with an expanded telencephalon, the remainder of the brain was smaller than that with a smaller telencephalon in all PCs. Even without considering the effects of size, the shape change of the telencephalon and posterior part of the brain were large in those PCs. In contrast, the position of the brain relative to the lateral semicircular canal did not change dramatically without considering size. This clearly supports the notion that the brain position is dependent on size.

image

Figure 5. Variation in brain shape for each principal component (PC) score in size-adjusted PCA. The lateral semicircular canals are shown by bold lines.

Download figure to PowerPoint

In size-adjusted PCA, consistent shape change correlations were found between the telencephalon and cerebellum, between the diencephalon and optic tectum, and between the cerebellum and myelencephalon (Fig. 5). Within these pairs of brain regions, the shape of each brain region changed in concert with each other. However, the telencephalon/cerebellum, diencephalon/optic tectum and cerebellum/myelencephalon did not always show concerted changes but rather showed mosaic changes. The anteroposterior expansion of the telencephalon involved narrowing of the diencephalon/optic tectum pair in PC1, whereas no narrowing or widening of the diencephalon/optic tectum was observed in PC2 and PC3. In PC1 and PC3, telencephalic/cerebellar expansion was accompanied with dorsal rotation of the cerebellum/myelencephalon, whereas in PC2, telencephalic/cerebellar expansion was accompanied with ventral movement of the cerebellum/myelencephalon.

Covariation between brain shape and orbital shape (2B-PLS)

To explore the association between brain shape and orbital shape, we performed 2B-PLS analysis of the component of brain Procrustes shape coordinates against orbit Procrustes shape coordinates. The RV coefficient value was 0.5712. The first singular value explained 67.77% of the variance and showed a significant correlation between vectors (r = 0.8669; P < 0.001). The permutation test (10 000 permutations) yielded a highly significant (P < 0.001) covariance between the two blocks. The second singular value of 2B-PLS explained 12.882% of the variance (r = 0.8217; P < 0.001).

The results of 2B-PLS analysis of brain shape and orbital shape are shown in Fig. 6. Within the trend of covariation, changes in brain shape with orbital contraction were identical to brain shape changes with unadjusted PC1 decreases, including the expansion of the telencephalon, and ventral rotation of the cerebellum and myelencephalon (Fig. 3). Thus, orbital shape is a dominant factor impacting avian brain shape and vice versa. Changes in brain shape were accomplished by anterior movement of the lacrimal bone, which made the anteroposterior dimension of the orbit elongate in shape. The orbital shape tended to lengthen in the anteroposterior direction as the brain shape changed from round to longer and narrower according to 2B-PLS analysis (Fig. 4). That is, a flexed round brain had wide round orbits that are located laterally and caudally, while a brain with a relatively smaller telencephalon and lengthened cerebellum and myelencephalon had elongated orbits located rostrally.

image

Figure 6. Pattern of covariation between brain shape and orbital shape as revealed by two-block partial least-squares (2B-PLS). (A) A scatter plot of PLS1 for brain shape and orbital shape (r = 0.8669; P < 0.001). The covariance explained 67.77% of the variation in brain shape variables. The lateral semicircular canals are shown by bold lines. (B) Sequence of three-dimensional models of shape variables of the brain and orbit.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

Patterns of shape diversification

Because we set boundaries for each brain region on the basis of appearance of the endocast (Fig. 1), our definition of brain regions does not completely coincide with that of external histological boundaries of the brain. Therefore, notably, the observed changing patterns in these regions observed in this study are an ‘apparent’ pattern. Although the boundaries used in this study did not completely coincide with the histological boundaries, it is considered that the definition of brain regions adopted here is adequate to explore the external shape variation in the avian brain.

Our results indicate that a considerable amount of shape diversity found in the brains of modern Aves can be described by a small number of dimensions using PCA (Figs 2,3 and 5). Two main trends of concerted changes were observed both in unadjusted and size-adjusted PCA: (i) expansion or shrinking of the telencephalon; and (ii) elongation or shortening of the brain base and brain stem. Accordingly, expansion of the telencephalon and shortening of the brain base and brain stem led to a rounder and more upward brain orientation (posture), while shrinking the telencephalon and elongation of the brain base and brain stem led to a more elongated and anteriorly inclined brain. Duijm (1951) and Marugán-Lobón & Buscalioni (2006, 2009) reported that the angle of the brain base correlated with the orientation of the plane of the foramen magnum in the avian skull. These trends in shape change revealed by previous studies are generally consistent with our results. The anterior part of the telencephalon extends anteroventrally with decreasing PC1 and PC2 scores in unadjusted and size-adjusted PCA. Stingelin (1957) recognized two directions of telencephalic development: (i) basal–frontal development; and (ii) dorsal–frontal development. With dorsal–frontal development, the eminentia sagittalis (Wulst) extends to the rostral part of the telencephalon, while with basal–frontal development, the position of the eminentia sagittalis is largely unaltered and it remains in the rostral part of the telencephalon. Anteroventral extension of the telencephalon (i.e. Phalacrocorax, Gallus, Anseriformes, Ciconiiformes and many water birds), which was recognized by PCA, coincides with dorsal–frontal development. On the other hand, rounded telencephalons (−PC1 and −PC2; i.e. Corvus, Falco, Strix and Columbiformes) match basal–frontal development. Thus, the mode of telencephalon development is an important factor determining the variation in avian brain shape. The developmental mode of the eminentia sagittalis is consistent within avian orders (Stingelin, 1957), suggesting that the avian brain shape and the developing mode of telencephalon have been coevolved. It seems that the rounded upwardly tilted brain of taxa such as Passeriformes, Falconiformes and Strigiformes has been evolved with the basal–frontal development, while Psittaciformes closely related with Passeriformes and Falconiformes shows a different shape of brain – upward tilted flat brain. It would appear that the brain shape of Psittaciformes has had a unique experience of evolution as compared with passeriform and falconiform birds.

Shape allometry

Multivariate regression analysis suggested that expansion of the telencephalon and caudodorsal rotation of the posterior part of the brain correlates strongly with increasing brain size (Fig. 4). Allometric changes in brain shape were related to the slenderness of the brain and the degree of rotation. The rotation of brain causes a dramatic change of the relative position between the brain and lateral semicircular canal. In general, allometric shape changes appeared to generally correspond with positive changes in unadjusted PC1 and PC2, and negative changes in PC3. The roundness or slenderness of the avian brain is related to skull shape (Duijm, 1951; Dullemeijer, 1960; Marugán-Lobón & Buscalioni, 2006); therefore, the allometric shape changes also relate to skull shape. In fact, skull shape is also subject to change with increased size (Kulemeyer et al. 2009; Marugán-Lobón & Buscalioni, 2009).

Size-independent change

Because some proportion of the variation in brain shape (8.81%) was explained by the brain size in unadjusted PCA, the effects of size on brain shape must be removed to understand the shape changes in the avian brain.

Using size-adjusted data sets, which were also adjusted for phylogeny, the consistent shape change correlations were found between three region pairs, the telencephalon and cerebellum, the diencephalon and optic tectum, and the cerebellum and myelencephalon (Fig. 2). Although the telencephalon and diencephalon might also show a weak shape correlation, this was not detected by our analysis. Iwaniuk et al. (2004) reported that the volumes of the following region pairs were consistently correlated in the avian brain: telencephalon and diencephalon; diencephalon and optic tectum; mesencephalon and optic tectum; and cerebellum and myelencephalon (‘mosaic brain evolution’: Iwaniuk et al. 2004; Striedter, 2005; Charvet & Striedter, 2009a,b). Their volumetric results are generally consistent with our shape analysis results. Specifically, both volumetric and shape analyses suggest mosaic brain evolution in birds. That is, it may be said that brain regions that share connections as part of the same functional pathway tend to covary in shape, while regions that differ in function change shape independently. However, there is a correlation in shape changes between the telencephalon, cerebellum and brain stem. Although they share few substantial neuronal connections (Iwaniuk et al. 2004), expansion of the telencephalon pushed the cerebellum posteriorly, causing the brain stem to rotate ventrorostrally in all PCs. This concerted shape change among the telencephalon, cerebellum and brain stem was caused mainly by the structural restriction than the strength of the neuronal connection. The strength of the neuronal connection is a dominant regulator of the brain shape, but location or structural constraints among each brain region also play a major role in determining the brain shape in Aves. However, it must be noted that geometric morphometrics pictures only the shape (not size) change of brains. Hence, it requires a much more detailed theoretical development of the method using geometric morphometrics to discuss mosaic brain evolution more clearly and precisely.

The brain with an expanded telencephalon (size-adjusted negative PC1, PC2, positive PC3, and non-size-adjusted negative PC1) had a relatively small optic tectum, cerebellum and myelencephalon in proportion to the whole brain. Hence, it is known that the brain with a larger telencephalon is highly encephalized (Iwaniuk et al. 2005), a large telencephalon correlates with an increase in relative brain volume. Iwaniuk et al. (2005) reported that psittaciforms and passerines whose telencephalon is significantly larger than that of other birds have a relatively smaller mesencephalon, optic tectum, cerebellum and myelencephalon than non-passerines. We clarified that the apparent size of the brain regions changes according to the degree of encephalization, and this finding corroborates the volumetric result of Iwaniuk et al. (2005). That is, encephalization is reflected in the brain shape of birds.

In addition, as described above, the brain with a large telencephalon (size-adjusted negative PC1, PC2, positive PC3 and non-size-adjusted negative PC1) tends to have a short cranial base (brain base), and this makes the brain ventrally flexed, which means that the anterior cranial base largely flexes ventrally. In other words, there is a correlation observed between cranial base angulation and encephalization in birds. This phenomenon has been previously reported in birds (Marugán-Lobón, 2010). A more ventrally flexed cranial base helps to accommodate a large brain volume in mammals (Biegert, 1963; Gould, 1977; Ross & Ravosa, 1993; Lieberman et al. 2008), and the flexion of the cranial base is a phenomenon common to both mammals and birds.

Covariation between brain shape and orbital shape

Only a small proportion of the variation in brain shape was explained by brain size, leaving approximately 90% of variance unexplained. Subsequent 2B-PLS analysis indicated that orbital shape can explain a significant proportion of the remaining variation in brain shape. Indeed, brain shape strongly correlated with orbital shape (Fig. 6). Thus, birds with a negative brain shape (narrower brains) have relatively elongated orbits, while birds with a positive brain shape (rounder, ventrally flexed) have relatively round orbits (Fig. 6). In addition, elongated orbits are more anterior in the skull while rounder orbits are posterolateral. Because the vertebrate retina is ontogenetically a part of the brain, the optic nerve is regarded as a central tract and the tectum as a coordinating or distributing center between other brain centers (Johnston, 1902). Thus, the anteroposterior elongation and contraction of the orbit is considered to be related to the change in length of the optic nerve or the relative position of the eyeball. This indicates that in rounder brains, the center of gravity lies above and nearer to the neck, while in narrower brains, the center of gravity is more anterior. Birds with narrow brains and elongated, anterior orbits (concentrated in the lower left in Fig. 6a) belong to ‘the straight skull type’ (Duijm, 1951), with relatively longer bills and a medulla that extends in the posterior direction. Duijm (1951) suggested that the skull and bill of birds of the straight skull type lie in line with the neck. The large shape change in the orbits stems largely from the major shift in the lacrimal bone, and this shift changes the direction of the viscerocranium and/or skull. The rounded orbits accompany the short and more ventrally tilted bill, whereas the bill with elongated orbits extends in a horizontal direction. Thus, brain shape may also be affected by head posture. In addition, Kulemeyer et al. (2009) noted a possible link between the position of foramen magnum and head posture, and suggested a correlation between head posture and the ability for sustained flight. Thus, brain shape may be affected by brain size, orbital shape and flight behavior.

Because the eye is a part of the brain, the eye is also considered to be a module of the brain. Thus, understanding the relationship between the optic organ and brain components sheds light on the evolution of the brain shape in birds. Changes in brain shape expressed by covariance are almost identical to those expressed by unadjusted PC1. Because unadjusted PC1 explained the greatest proportion of the variation in avian brain shape, it can also be assumed that orbital shape considerably affects brain shape. There is also a possibility that the pattern of telencephalon development (Stingelin, 1957) correlates with the orbital shape, i.e. a brain with a relatively rostroventrally situated eminentia saggitalis (involved in integration and processing of the visual signal; Iwaniuk & Wylie, 2006) is associated with the more posterolaterally situated round orbit. In fact, the roofs of both hemispheres of the telencephalon showed the biggest change according to the PLS axis change (Fig. 6), and these parts corresponded to the position of the eminentia sagittalis. In other words, the shape or size of the eminentia sagittalis changes with the orbital shape. These orbital and telencephalic changes correlate with an integrated change in the brain shape, such as ventrodorsal rotation of the cerebellum and myelencephalon.

Compared with body and cranial sizes, birds have relatively large eyes because vision is the predominant sensory modality. Birds also have relatively large brains to process visual input. Several studies have shown a significant correlation between brain size and eye size (Garamszegi et al. 2002; Thomas et al. 2006; Burton, 2008), and one study has speculated a complex interdependence of eye size or shape, skull shape, and brain shape (Dubbeldam, 1989). Thus, the structure of the avian brain is strongly influenced by eye size and shape, and it is significant that this study shows a clear quantitative relationship between brain shape and orbital shape. In light of recent studies showing a correlation between eye size and behavior (Brooke et al. 1999; Garamszegi et al. 2002; Thomas et al. 2006; Hall, 2008; Hall & Heesy, 2011), the present study results may lead to a better understanding of brain shape evolution and the relationship between eye shape and behavior.

The cranial base angle covaried with the shape change of the orbits. Brains with a more ventrally flexed cranial base (positive PLS1; Fig. 5) had more anteroposteriorly shorter orbits than those with a more dorsally flexed cranial base. In the case of mammals, cranial base angulation is closely related to brain size and facial length (Moss & Young, 1960; Biegert, 1963; Ross & Ravosa, 1993; Ross & Henneberg, 1995; Lieberman, 1998; Lieberman et al. 2000a,b; McCarthy & Lieberman, 2001; Ross et al. 2004; Bastir & Rosas, 2006; Bastir et al. 2006). Architecturally, the cranial base provides growth space for the brain and face. Hence, cranial base angulation also results in variation in facial prognathism and orientation (Ashton, 1957; Biegert, 1963). The cranial base may help accommodate the spatial packing of the face in mammals. As is the case of the mammalian face, the avian orbit is also lengthened according to the angulation of the cranial base. The face is the prime component of the mammalian skull and influences the entire skull shape, whereas the eyes are the prime component of the avian skull. Considering that the mammalian face and avian orbit are the important components of the viscerocranium, the cranial base angulation may also play an important role in the spatial packing in the avian viscerocranium.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

We highlight the two main findings of this study. First, larger brain size was associated with the relative shrinking of the telencephalon and ventral rotation of the posterior part of the brain. Second, upwardly tilted ventrally flexed brains tended to have rounder orbits, while narrower and more anteriorly inclined brains tended to have anteroposteriorly elongated orbits. Combining the result of PCA and other analyses, we conclude that the size and shape of the orbit is the important factor impacting brain shape and the diversity of brain shapes across avian species.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References

The authors thank Tetsuo Higashi (Hitachi Aloka Medical, Ltd.), Minoru Ikehara (Kochi University), Toshiaki Kuramochi (National Museum of Nature and Science, Japan) and Shizu Yanagimoto (Kochi University) for access to CT scanning. The authors also thank Norihisa Inuzuka (The University of Tokyo), Makoto Manabe (National Museum of Nature and Science, Japan), Takanobu Tsuihiji (The University of Tokyo) and Takeshi Yamasaki (Yamashina Institute for Ornithology) for providing specimens. This study was supported by the JSPS Research Fellowship to S.K. Part of this study was performed under the cooperative research program of Center for Advanced Marine Core Research, Kochi University <09B037>. This manuscript was greatly improved by comments from Julia Clarke (University of Texas at Austin), Jesús Marugán-Lobón (Universidad Autónoma de Madrid) and 2 anonymous reviewers.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References