Morphometric Study of Phylogenetic and Ecologic Signals in Procyonid (Mammalia: Carnivora) Endocasts
Heather E. Ahrens
Center for Functional Anatomy and Evolution, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
Correspondence to: Heather E. Ahrens, Center for Functional Anatomy and Evolution, Johns Hopkins University School of Medicine, 1830 East Monument Street, Room 305, Baltimore, Maryland, 21205, USA. Telephone: (410) 955–7172. E-mail: firstname.lastname@example.org
Neuroanatomy and the evolution of brain size have long been of interest to paleomammalogists. The foundations for interpreting differences in brain size and shape in Mammalia were established, in part, by work on the evolution of the brain throughout Mammalia (e.g., Marsh, 1884; Jerison, 1973), in horses and cetaceans (e.g., Edinger, 1948, 1955), primates (e.g., Biegert, 1963; Radinsky, 1967, 1975a; Jerison, 1979), and carnivores (e.g., Radinsky, 1968, 1971, 1975b). The neuroanatomy and relative brain size of Primates and Carnivora have been of particular interest because extant members of both clades exhibit encephalization. Encephalization is a larger brain size than predicted for a given body size (Jerison, 1973; Finarelli and Flynn, 2007, 2009; Silcox et al., 2009), which can be computed using a number of regression equations (e.g., Jerison, 1973; Eisenberg, 1981; Hurlburt, 1996). The evolutionary importance of encephalization and its relationship to diversity within a clade are not yet fully understood (Jerison, 1973; Harvey and Krebs, 1990; Finarelli and Flynn, 2007). However, relative brain size is thought to relate to numerous aspects of behavior and physiology (Eisenberg and Wilson, 1978; Mace et al., 1981; Hurlburt, 1996; Marino et al., 2004; Shultz and Dunbar, 2006), with many studies investigating the evolution of brain size and its relationship to ecology in Carnivora (Radinsky, 1978; Gittleman, 1986, Iwaniuk et al., 1999; Pérez-Barbería et al., 2007; Finarelli and Flynn, 2009). Given the range and occasionally contradictory results of these studies, analyses of brain shape may provide additional information regarding carnivore behavior and ecology, as well as a test of conclusions drawn from brain size alone.
Inferences about the evolution of behavior and ecology from brain morphology are based on fundamental principles of paleoneurology, including the localization doctrine and principal of proper mass. The localization doctrine states that neurological functions are controlled by localized structures in the brain (Jerison, 1973). The principle of proper mass states that the mass of neural tissue associated with a function is relative to the amount of information processing involved in carrying out that same function (Jerison, 1973). The principle of proper mass suggests that the absolute mass of specific neural tissue represents the importance of a given function, but Jerison (1973) indicated that the relative mass of given neural structures within a species represented the importance of those particular functions to individuals within that species. Though modern investigations into behavioral and ecological differences in brain size are based on these foundations, more recent analyses have recognized the complex functional and structural organization within the brain. There is significant evidence that specific interconnected neural systems, rather than isolated localized structures, share a given function and tend to evolve as an integrated unit (Barton et al., 1995; Barton and Harvey, 2000; De Winter and Oxnard, 2001). Understanding of the relative size and morphology of different functional divisions within the brain has revealed correlations with numerous behavioral and ecological traits, which are themselves subjected to selective pressures (Barton et al., 1995). In addition, species with convergent skeletal morphology and ecology possess convergent brain proportions (De Winter and Oxnard, 2001). Thus, it is important to understand which neural systems within the brain differ between taxa that possess different behaviors and ecologies in order to identify which systems experienced selection pressure (Barton and Harvey, 2000). These principles of neuroanatomy and paleoneurology have been used to interpret differences in the brain throughout the evolution Mammalia (e.g., Radinsky, 1968; Jerison, 1973; Rowe et al., 2011; Silcox et al., 2010).
In order to study changes in brain morphology through time researchers have relied on endocasts, three-dimensional (3-D) representations of the endocranial space, which can be naturally preserved or artificially constructed with latex or digital methods. Because the brain fills the majority of the endocranial space during development in mammals, the brain leaves impressions on the inner surface of the skull (Macrini et al., 2007b). Although not all morphological features of the external brain are preserved in endocasts, major structures, including gross sulcal patterns, can be identified and are related to behavioral variation and specialization (Welker and Campos, 1963; Falk et al., 2009). Thus, endocasts serve as a proxy for the shape of the brain, and can provide clues to the evolution of sensory systems, ecology, and behavior (e.g., Edinger, 1948, 1955; Radinsky, 1968, 1975a, 1975b; Jerison, 1973; Rowe, 1996; Rowe et al., 2011). Endocasts also provide access to potential suites of phylogenetic characters (Macrini, 2006; Macrini et al., 2007a) and have been used to examine phylogenetic patterns in Carnivora (Radinsky, 1971, 1973, 1975b). Rarely, though, are endocasts used to explicitly test or build phylogenies beyond a comparative examination within a phylogenetic context due to their complex three-dimensional shape.
Methods for testing phylogenetic hypotheses using geometric morphometrics (Klingenberg and Gidaszewski, 2010) provide a means for explicit testing of phylogenies with brain shape as represented by the endocast. Because the brain is a complex structure, isolating meaningful, discrete phylogenetic characters is difficult. Three-dimensional geometric morphometric techniques can be used to quantitatively capture shape morphology and variation between taxa (Zelditch et al., 2004; Klingenberg, 2010), and are a potential avenue for testing phylogenetic hypotheses with complex morphological structures. The utility of geometric morphometric data in phylogenetic analyses has been discussed by many authors (Bookstein, 1994; Zelditch et al., 1995; Rohlf, 1998; Montiero, 2000; MacLeod, 2002), with Klingenberg and Gidaszewski (2010) introducing the use of a permutation test of shape mean to test for a significant phylogenetic signal in shape data. Using the methods of Klingenberg and Gidaszewski (2010), two phylogenetic hypotheses of the relationships within Procyonidae will be examined in light of new brain shape data from endocasts.
Crown Procyonidae is a family of mammals with only six genera: Bassaricyon, Bassariscus, Nasua, Nasuella, Potos, and Procyon and approximately 14 species (Wozencraft, 2005). The morphologic and molecular hypotheses of relationships within the family differ strikingly (Fig. 1), with the sister taxon relationships incongruent between the two topologies. In the morphological tree (Ahrens, 2012), Bassariscus is the most basal taxon, sister to all other procyonids, with the remaining members of Procyonidae forming two clades; one clade includes Ailurus and Bassaricyon plus Potos and the other includes Procyon and Nasuella plus Nasua. The molecular tree consists of Potos flavus as the sister taxon to all remaining members of Procyonidae, while Bassaricyon and Nasua compose a clade which is the sister group to a clade including Procyon plus Bassariscus (Fulton and Strobeck, 2007; Koepfli et al., 2007). Procyonidae possesses limited taxonomic diversity, but the members possess a wide range of feeding ecologies and behaviors. This ecological diversity has led some to suggest adaptive convergence as a possible explanation for the incongruent morphological and molecular phylogenetic hypotheses (Koepfli et al., 2007). Moreover, as members of the larger clade Carnivora they exhibit encephalization (Finarelli and Flynn, 2007, 2009), thus providing a test for patterns of encephalization within a single family.
Here, 3-D geometric morphometrics are used to examine phylogenetic and ecologic patterns in the endocasts of procyonids. Two phylogenetic hypotheses are compared using new morphological data, geometric morphometric data on brain shape derived from endocasts using methods proposed by Klingenberg and Gidaszewski (2010). For each test of a given phylogeny, the null hypothesis is the absence of phylogenetic signal in the sample morphometric data. In addition, it is expected that brain shape will provide better support for the phylogeny that was constructed using craniodental characters. The potential issues of testing phylogeny with brain shape are also discussed. Finally, brain shape variation is analyzed using principal components analyses and wireframes to extrapolate neuroanatomical differences to differences in behavior and ecology and examine aspects of encephalization within Procyonidae.
MATERIALS AND METHODS
FMNH, Field Museum of Natural History; MVZ, Museum of Vertebrate Zoology, University of California, Berkeley; ROM, Royal Ontario Museum; TMM, Vertebrate Paleontology Laboratory, Texas Natural Science Center, The University of Texas at Austin; USNM, National Museum of Natural History.
Taxon Sampling and Phylogenetic Topologies
Eleven taxa representing both the ingroup, Procyonidae, and the outgroup were sampled. Ingroup taxa include Bassaricyon alleni (FMNH 41502), Bassariscus astutus (TMM 473), Nasua narica (FMNH 14013), Nasua nasua (FMNH 21400), Nasuella olivacea (FMNH 70746), Potos flavus (USNM 291066), and Procyon lotor (TMM 778), all of which were sampled for the morphological phylogeny (Ahrens, 2012). Outgroup taxa represent the remaining families of Musteloidea, including two mephitids, Spilogale gracilis (TMM 3746) and Mydaus marchei (FMNH 62878), one ailurid, Ailurus fulgens (ROM 180), and one mustelid, Martes pennanti (MVZ 29809). Only one endocast could be obtained for each species; though this small sample size is not ideal, phylogenetic information can still be gathered and assessed from small samples sizes of endocasts (Macrini et al., 2006; Racicot and Colbert, 2013). However, the amount of intraspecific variation could not be assessed because only one endocast was available for each species.
Two phylogenies of Procyonidae were tested, one derived from morphological data and the other from molecular data (Fig. 1). The morphological tree is based on 78 craniodental characters exclusive of endocast characters (Ahrens, 2012). The molecular tree tested is a composite of several congruent molecular topologies, with the ingroup topology based on Fulton and Strobeck (2007), Koepfli et al., (2007), and Wolsan and Sato (2010) and the outgroup topology based on Fulton and Strobeck (2006), Sato et al., (2009), Eizirik et al., (2010), and Yu et al., (2011); molecular phylogenies were based on nuclear introns and exons, as well as mitochondrial sequences.
There are several differences between the morphological and molecular phylogenies with respect to both ingroup and outgroup relationships. Ailurus fulgens, the red panda, is recovered with the ingroup by the morphological analysis, whereas it is recovered as the sister taxon to a clade consisting of Procyonidae plus Mustelidae in molecular analyses. Within the ingroup, the morphological analysis recovers Bassariscus astutus as the most basal taxon, whereas the molecular analysis recovers Potos flavus as the most basal taxon. Additionally, all subsequent sister taxon relationships are incongruent between the two phylogenetic hypotheses.
Twenty-one 3-D landmarks were collected from the surface of each endocast volume, serving as a proxy for the brain, in etdips 2.0 (Mullick et al., 1998). Landmarks collected are described in Table 1 and shown in Figure 2. Endocasts were oriented with the cranial base horizontal because habitual head orientation is unknown in Procyonidae and the cranial base represents a biologically homologous structure. Though endocasts preserve neural, vascular, and other structures, only those features representing neural structures were landmarked. Landmarks were collected from the left side and midline only in order to reduce the number of landmarks per specimen. The nature of endocast morphology, which does not generally preserve discrete structures or distinctive features, prevents the identification of Bookstein's Type I landmarks (1991). Rather, fuzzy landmarks located at minima or maxima of curvature relative to a local structure were used (Valeri et al., 1998). A test for outliers was performed using MorphoJ 1.03c (Klingenberg, 2011). A precision test was performed by repeating landmark selection three times for each specimen; standard deviations for each landmark are presented in Table 2. Landmark 9 (dorsal pole) represents the most imprecise landmark and the only landmark that deviated more than 0.5 mm. Additionally, reproducibility of the landmarks was verified using a principal components analysis (PCA) and tested using a one-way multivariate analysis of variance (MANOVA) on the Procrustes coordinates with specimen set as the factor in SPSS (IBM Corp., 2012). The PCA demonstrated that specimen repeats of the same species clustered closer to one another than to specimens of another species and the MANOVA verified that repeats of specimens are significantly different from specimens from another species (P value < 0.001). To further reduce error, all landmarks were averaged for the three trials for each specimen.
Olfactory bulbs, caudal constriction at the dorsoventral midpoint
Frontal pole, rostral-most point of the neocortex along the presylvian sulcus
Temporal pole, distal end of the temporal lobe
Piriform lobe, ventral terminus
Occipital pole, caudal-most point of the occipital lobe
Dorsal pole, dorsal to foramen lacerum located on the gyrus rather than the midline
Cruciate sulcus, midline
Lateral sulcus, dorsal terminus
Caudal midline terminus between the left and right occipital lobes
Vermis, caudal dorsal terminus
Vermis, caudal ventral terminus
Paramedian lobe, lateral terminus
Flocculus, lateral terminus
Paraflocculus, posterolateral terminus
Hypophoseal fossa, ventral terminus
Optic tract, medial margin, medial to the proximal-most differentiation of optic tract from the olfactory peduncle
Optic tract, lateral margin, at proximal-most differentiation of optic tract from the olfactory peduncle
Clinoid processes, midline of the dorsal surface of the opening of the clinoid processes
Table 2. Landmark Precision Reported as Mean Landmark Standard Deviation (mm)
All subsequent morphometric analyses and tests were conducted in MorphoJ 1.03c (Klingenberg, 2011). To compare endocast (brain) shape, generalized Procrustes superimposition was used to remove information relating to size, position, and orientation from the landmark coordinate data. Principal components analyses were used to examine shape variation among the taxa sampled. To test for an allometric relationship between brain shape and brain size, a regression analysis was performed for principal component (PC) scores versus log centroid size. Analyses were conducted twice, once with all taxa included and once with only ingroup (procyonid) taxa included. Wireframe comparisons were produced to compare brain shape between the root and Procyonidae node, as well as between Potos flavus and Nasua narica, two taxa with extreme relative brain shape.
The methods outlined by Klingenberg and Gidaszewski (2010) and modules in MorphoJ 1.03c were used to test for phylogenetic signal in brain shape data. The key points of the methods are presented below; for a detailed description of the background and reasoning for the methods readers should refer to Klingenberg and Gidaszewski (2010). Tests were conducted for the morphological topology (Ahrens, 2012) and the composite molecular topology (Fulton and Strobeck, 2006, 2007; Koepfli et al., 2007; Sato et al., 2009; Eizirik et al., 2010; Wolsan and Sato, 2010; Yu et al., 2011) for all taxa included and for the ingroup (procyonid) taxa only. Given phylogenies were mapped onto all principal components and ancestral node values were computed based on unweighted squared-change parsimony and using the generalized least-squares method. Tree lengths were then calculated as the sum of squared changes for of all branches and ancestral nodes using the optimal configuration of the phylogeny in shape space. A permutation test with 100,000 permutations was used to simulate the null hypothesis of a complete absence of phylogenetic signal in shape data. The minimal P values are presented.
General Remarks on Morphology
General examination of the endocast surface models reveals noticeable variation in brain morphology (Fig. 3). Particular similarities between the species of Nasua include elongate brains with expanded olfactory peduncles and prorean gyri (dorsorostral expansion of the neocortex; Kreiner, 1961). Within Procyonidae, this elongate brain appears most different from that of Potos flavus, the kinkajou. The brain of Potos flavus is rostrocaudally compressed and has a dorsally expanded cerebrum. Procyon lotor, however, possesses a seemingly intermediate brain shape relative to Nasua and Potos flavus, with more dorsal expansion of the cerebrum than Nasua and a more elongate endocast than Potos flavus. Bassariscus astutus has few gyri and sulci, perhaps due to its small size (Macrini et al., 2007a). Outgroup taxa are visibly different from the procyonids, particularly the mephitids Mydaus marchei and Spilogale gracilis, which have flat cerebri and tall olfactory bulbs. The brain of Ailurus fulgens is similar to the procyonids in having a taller cerebrum and less pyramidal olfactory bulbs.
Morphometric Analysis of Brain Shape
A principal components analysis of brain shape variation was plotted for all taxa (Fig. 4A). The PCA recovers 10 PCs with the first six PCs each accounting for at least 5% of the total variance. The eigenvalues for PCs one through six as well as the five landmarks contributing the most to the variation on the first six PCs are presented in Table 3. The first two PCs represent 36.67% and 19.29% of the total shape variation. A regression analysis of PC scores versus log centroid size does not show a significant relationship between brain shape and brain size, with an r value of −0.02908 and a P value of 0.87923. One distinct feature of the PCA (Fig. 4A) was the separation in morphospace of the mephitids relative to all other taxa, with the two skunks as the highest loaders on PC1. For this reason additional analyses of shape variation and phylogenetic signal were run excluding all generally accepted outgroup taxa (Ailurus fulgens, Martes pennanti, Mydaus javanensis, Spilogale gracilis). Because the goal of this paper is to test phylogenetic hypotheses of Procyonidae, rather than relationships among Musteloidea, removing all outgroups provides a truer test of phylogenetic signal in procyonid brain shape.
Table 3. Eigenvalues for the Principal Components that Represent >5% of the Total Variance and the Five Landmarks With the Highest PC Loadingsa
Highest loading landmarks
Analysis of all taxa.
Shape change along the first two principal components reflects complex three-dimensional changes (Fig. 5). Principal component one depicts changes in a brain with a lower more rostrally placed olfactory bulb (landmark 3), a higher and more caudal dorsal pole (9), a well-defined temporal lobe due to the longer lateral sulcus (11), a more caudally located and narrower optic track (19 and 20), and a more dorsorostrally placed clinoid process (21) low on PC1 to a brain with dorsocaudal olfactory bulb (3), lower dorsal pole (9), rostrally short temporal lobe (11), more rostrally located and wider optic track (19 and 20), and ventrocaudal clinoid process (21) high on PC1. Principal component two reflects changes in a brain with a more caudally placed rostral terminus of the olfactory bulb (1), more rostrally shifted frontal pole (5), ventrorostral and more medial temporal pole (6), a more caudomedial flocculus (16), and enlarged paraflocculus (17) low on PC2 to a brain with a more rostral olfactory bulb (1), relatively caudal frontal pole (5), caudolateral temporal pole (6), a more rostrolateral flocculus (16), and reduced paraflocculus (17) high on PC2.
A principal components analysis of brain shape variation within Procyonidae was also plotted (Fig. 4B), with the first two PCs accounting for 46.30% and 21.13% of the total variation. The PCA recovers six PCs with the first five PCs each accounting for at least 5% of the total variance. The eigenvalues for PCs one through five as well as the five landmarks contributing the most to the variation on the first five PCs are presented in Table 4. PC1 contains the largest separation between taxa, with Nasua, Nasuella olivacea, and Procyon lotor distinct from Potos flavus, Bassaricyon alleni, and Bassariscus astutus. A regression analysis of PC scores versus log centroid size does not show a significant relationship between brain shape and brain size, with an r value of −0.18004 and a P value of 0.42324.
Table 4. Eigenvalues for the Principal Components That Represent >5% of the Total Variance and the Five Landmarks With the Highest PC Loadingsa
Highest loading landmarks
Analysis of Procyonidae only
Principal components one and two on the analysis of procyonid endocasts depict complex three-dimensional changes (Fig. 6) just as in the analysis of the endocasts of all taxa. The variation depicted along PC1 of the procyonid analysis is similar to the pattern of variation recovered by PC2 of the analysis of all taxa, with four of the top five highest contributing landmarks shared, including the rostral terminus of the olfactory bulbs, frontal pole, flocculus, and paraflocculus. Principal component one depicts changes in a brain with a ventrorostral located rostral terminus of the olfactory bulb (1), caudolateral frontal pole (5), a narrower temporal lobe and caudally displaced lateral sulcus (11), rostrolateral flocculus (16) and reduced paraflocculus (17) low on PC1 to a brain with a more dorsocaudal olfactory bulb (1), rostral frontal pole (5), wider temporal lobe with a rostrally located lateral sulcus (11), a more caudomedial flocculus (16), and enlarged paraflocculus (17) high on PC1. Principal component two reflects changes in a brain with a more dorsorostrally placed dorsal terminus and caudal constriction of the olfactory bulb (3 and 4), ventrorostrally placed midline of the cruciate sulcus (10), dorsocaudally placed paraflocculus (17), and dorsocaudally placed hypophyseal fossa (18) low on PC2 to a brain with a more ventrocaudally placed dorsal terminus and caudal constriction of the olfactory bulb (3 and 4), dorsocaudally placed midline of the cruciate sulcus (10), ventrorostrally placed paraflocculus (17), and ventrorostrally placed hypophyseal fossa (18) high on PC2. Though changes in the shape of the olfactory bulbs represent some of the greatest variation in shape among procyonids, most of the variation is restricted to the X and Z axes (seen in lateral view).
Phylogenetic Signal in Brain Shape
When the phylogenetic hypotheses are mapped into morphospace (Fig. 7) there are apparent differences between the two phylogenetic topologies. The morphological topology matches the morphological variation in brain shape fairly well, with the clade containing Nasua narica, Nasua nasua, Nasuella olivacea, and Procyon lotor present in the same area of shape space, and the basal taxa nearer to each other than to the coatimundis (Nasua and Nasuella) and the raccoon (Procyon). Ailurus fulgens, the red panda, is located in the shape space occupied by procyonids relative to PC1. Brain shape is not as congruent with the molecular topology. Long branches connect sister taxa across shape space, such as between Bassariscus astutus and Procyon lotor and between Bassaricyon alleni and the coatimundis. These qualitative observations are confirmed by the tree lengths, with a morphological tree length of 0.05630 compared to a molecular tree length of 0.06628. Despite this difference in tree length, there is significant phylogenetic signal in both topologies, with the P value of <0.0001 for the morphological topology and P value of 0.0039 for the molecular topology.
When the phylogenetic hypotheses are mapped into morphospace for the procyonids alone (Fig. 8), differences between the morphological and molecular topologies are apparent similar to the analysis of all sampled taxa. Brain shape agrees with the morphological topology quite well, with Nasua narica, Nasua nasua, Nasuella olivacea, and Procyon lotor separated on PC1 from the three other taxa Bassariscus astutus, Bassaricyon alleni, and Potos flavus. Potos flavus, the most derived taxa based on craniodental characters, is the highest loader in PC2. Brain shape is not as congruent with the molecular topology as with the morphological topology, with long branches uniting Bassaricyon alleni with the coatimundis and Procyon lotor with Bassariscus astutus. However, Procyon lotor and Bassariscus astutus both exhibit brain shape at the low end of PC2, and are sister taxa according to molecular data. The tests for phylogenetic signal corroborate these observations, with a morphological tree length of 0.02799 compared to a molecular tree length of 0.03300. There is significant phylogenetic signal in the morphological topology, with a P value of 0.0038, whereas the phylogenetic signal in the molecular topology is not statistically significant, with a P value of 0.0924.
Mapping the phylogeny of Procyonidae into morphospace also allows for the examination of the ancestral brain shape of Procyonidae relative to the more inclusive Musteloidea (root). There are several differences between the ancestral brain shapes (Fig. 9); changes are described with respect to the procyonid condition. The basal procyonid condition consists of smaller olfactory bulbs that are rostrally displaced dorsally relative to the rest of the brain. The piriform lobe is reduced, positioned more dorsomedially. The frontal lobe and parietal lobe (approximately represented by the dorsal pole) are expanded rostrally and dorsally, with the ventral margin shifted more ventrally. The paraflocculus is enlarged vetrolaterally relative to the rest of the brain. Variation within Procyonidae is greater than between the ancestral nodes discussed above, which is evident by the distance between the nodes versus the distance between procyonids on the first two principal components (Fig. 7A). Two taxa, Potos flavus and Nasua narica, represent different parts of the morphospace due to differences in brain shape. Morphological differences between the brains are listed below, including several present in regions of the brain associated with olfaction. The olfactory bulbs are elongate rostrocaudally and expanded ventrally in Nasua narica; however, the piriform lobe is enlarged ventrally in Potos flavus. The frontal and parietal lobes are expanded rostrally and dorsally, respectively, in Potos flavus. The caudally shifted frontal lobe of Nasua narica may rather indicate an enlarged prorean gyrus. The paraflocculus is greatly expanded laterally in Potos flavus.
Brain Shape and Phylogeny of Procyonidae
Patterns of brain shape variation preserved in the endocasts provide significant support for the phylogenies tested. When all four outgroup taxa are included, both the morphological and molecular topologies are supported by brain shape variation. However, when outgroup taxa are excluded, only the morphological topology is supported by the variation in brain shape. These results are promising and indicate that brain shape provides valuable information useful in testing hypotheses of musteloid relationships.
There are several possibilities for the support of the molecular topology when outgroup taxa are included. First, the molecular outgroup topology is correct, and the quantitative test of phylogenetic signal in brain shape provides morphological support for the molecular topology recovered by Fulton and Strobeck (2006), Sato et al., (2009), Eizirik et al., (2010), and Yu et al., (2011). However, the support may be biased by taxon sampling. The two mephitid taxa, Mydaus marchei and Spilogale gracilis, have drastically different endocranial morphologies (highest loaders on PC1) from the other taxa, and their sister-taxon relationship may be driving the support for both the morphological and molecular topologies. Additional studies incorporating endocasts representing a greater diversity of musteloid taxa could determine whether there is neuroanatomical support for the molecular outgroup or whether the results are an artifact of taxon sampling. If the lack of support for the ingroup relationships of Procyonidae is real, this may indicate homoplasy in the molecular data set of Procyonidae.
If the molecular hypothesis represents a better estimation of procyonid ingroup relationships, then there are implications for the evolution of the craniodental characters and brain shape within the clade. Because all sister taxon relationships are incongruent between the two topologies, rampant homoplasy among craniodental characters and brain shape must have occurred if the molecular topology represents the true relationships within Procyonidae. Studies describing morphological integration during development (Cheverud, 1995, 1996; Bastir and Rosas, 2005; Bruner, 2008) may reveal a mechanism for homoplasy in external characters of the skull and dentition and those preserved on the endocast. Molecular researchers have suggested adaptive convergence as justification for accepting the molecular topology (Koepfli et al., 2007); thus, a combination of adaptive convergence driving craniodental character morphology and integration between the external craniodental morphology and brain shape would need to occur to cause the homoplasy observed in both morphological datasets. This is in accordance with findings in several mammalian clades, where convergent brain proportions were correlated with several convergent ecological factors including diet, locomotor behavior, and activity timing (De Winter and Oxnard, 2001). In Procyonidae, sister taxa in the morphological topology share similar diets and locomotor behaviors (Koepfli et al., 2007) and therefore may possess convergent craniodental and brain morphologies.
However, some studies indicate different morphological modules exist in the carnivoran skull (Goswami, 2006), with increased disparity in certain modules of the face and dentition and lower disparity in the basicranium of carnivorans (Goswami and Polly, 2010). If the central nervous system and the surrounding osteological structures represent different evolutionary modules from those of the face and dentition and are subjected to different selective pressures or low disparity is present in regions that house the central nervous system, then testing phylogenetic hypotheses based on craniofacial data with neural characters may be informative. In addition, mosaic evolution has been documented within the mammalian brain (Barton and Harvey, 2000) and could explain why, despite correlations between ecology and the morphology of divisions within the brain, we would not necessarily expect the same signal in the skull, dentition, and brain. Using new morphological data to test a morphological hypothesis of procyonid relationships is not a fully independent test of phylogeny, but reduced integration between the skull, central nervous system, and postcranial characters may bolster the results regarding the morphological phylogeny. Importantly, this study tests and confirms the hypothesis that craniodental characters and brain shape will support the same phylogeny of Procyonidae.
Testing Phylogeny with Brain Shape
This is the first analysis testing phylogenetic hypotheses with brain shape data using these methods, and there are potential issues that should be acknowledged. One difficulty with testing phylogeny with complex morphologies is the often necessary use of Type II or III landmarks. These classes of morphometric landmarks may not represent homologous structures (Bookstein, 1991) and may not be ideal for testing hypotheses of similarity due to ancestry. The preserved endocast structure, which does not preserve unambiguously homologous features (Bruner, 2008), requires the use of fuzzy landmarks. Thus, researchers must operate under the assumption that phenetic similarity of brain shape in taxa is congruent with evolutionary relationships based on homology. Phenetic trees based on morphometric data are often not congruent with independent phylogenetic hypotheses, suggesting shape may be especially subject to homoplasy (Klingenberg and Gidaszewski, 2010). However, the presence of significant phylogenetic signal in procyonid brain shape consistent with the morphological phylogeny and outgroup phylogeny based on molecular data, suggests that brain shape can be used to test phylogenetic hypotheses and that brain shape within Procyonidae is no more likely to undergo homoplastic changes than conventional craniodental characters.
Additionally, there is significant quantitative support for the similarity in neuroanatomy between stink badgers (Mydaus) and other mephitids. These results support the relationships recovered by molecular data that recover Mydaus as a member of a mephitid clade (Dragoo and Honeycutt, 1997; Fulton and Strobeck, 2006; Sato et al., 2009; Eizirik et al., 2010). These results also corroborate those of Radinsky (1973), where the similarity between the brains of stink badgers to those of mephitids was recognized on the basis of qualitative comparisons. The similarities between the endocasts of Mydaus and Conepatus, Mephitis, and Spilogale were also suggested to represent retentions of a more primitive condition (Radinsky, 1973); however, the morphologies present in all other taxa included in this study are more similar to one another than they are to that seen in skunks and may indicate a shared ancestral brain shape among the nonmephitid musteloids. The quantitative support for qualitative observations of neuroanatomy provides further evidence that this method may be useful to phylogenetic studies.
Brain Shape, Ecology, and Encephalization of Procyonidae
Because of the complex relationship between neuroanatomical structures and their function, morphological changes along the principal components likely do not reflect one overall functional or ecological change, but may reflect several differences in behavior and ecology. Particularly, differences in brain shape between Procyonidae and basal Musteloidea may correlate with known aspects of their behavior and ecology. The relatively smaller olfactory bulbs and piriform lobe suggest that the basal procyonid had less acute olfaction than the basal musteloid. The expanded frontal and parietal lobes may be associated with better motor control and complex aspects of spatial orientation and perception, respectively (Welker and Campos, 1963; Buxton and Goodman, 1967; Radinsky, 1971). Better motor control and spatial perception may be expected because most procyonids spend more time in arboreal settings compared to Martes (Powell, 1981) and the mephitids (Kruska, 1990; Hwang and Larivière, 2004). The paraflocculus, which is thought to help in balance (Butler and Hodos, 1996), also correlates with better abilities to climb. The ancestral procyonid condition seems to emphasize areas related to arboreal and scansorial habits, and could be related to the diversification of procyonids in central and South America.
In Procyonidae there is a similar pattern of variation that may be associated with increased arboreality in certain taxa. Potos flavus, which is highly arboreal (Ford and Hoffmann, 1988), has expanded frontal and parietal lobes, as well as a greatly expanded paraflocculus. The caudally shifted frontal lobe of Nasua narica indicates an enlarged prorean gyrus. The prorean gyrus has been associated with social behavior in canids (Radinsky, 1969; Finarelli and Flynn, 2009), and may be similarly important for social behavior in Nasua narica, which are known to exhibit highly social behavior (Gompper, 1995). There are also differences in olfactory parts of the brain with elongate olfactory bulbs in Nasua narica and an enlarged piriform lobe in Potos flavus. Both taxa have expanded areas of olfaction, with different regions of the olfactory system enlarged. The elongate olfactory bulbs may provide an example of integration between the face and the brain in Procyonidae, as Nasua has one of the most elongate skulls as well as rostrocaudally elongate olfactory bulbs. Additionally, Potos flavus has a relatively short skull that may correlate with the reduced size of the olfactory bulbs, yet still seems to possess an expanded piriform lobe. Understanding the size of the olfactory system relative to the rest of the brain, as well as relative sizes of different parts of the olfactory system, may provide new information regarding structural organization and integration within the brain and may relate to differences in ecology.
The allometric relationship between body size and brain size (Jerison, 1973; Eisenberg, 1981; Finarelli and Flynn, 2007) and between brain size and gyrencephaly (Edinger, 1948; Macrini et al., 2007a) must be considered when selecting geometric morphometric landmarks and when interpreting shape differences among endocasts of different species. Here, landmarks were only placed on sulci that were identifiable in all taxa, including Bassariscus astutus, which possessed the most lissencephalic endocast within the current phylogenetic sample. In addition, tests for an allometric relationship between brain size and brain shape using regression analyses of PC scores versus log centroid size in both the complete sample and sample of Procyonidae only did not recover significant relationships between brain size and shape. These results indicate that brain shape is not significantly correlated with size given these landmarks within this sample. However, a denser sampling of surface landmarks from the neocortex may correlate with brain size given the relationship between brain size and degree of gyrencephaly. Therefore, the relationship between size and shape should always be tested before interpreting results of shape variation.
The relationship between body size and brain shape is also notable because musteloids, except mephitids, possess significant encephalization (Finarelli and Flynn, 2009). Procrustes superimposition removes overall brain size differences among taxa and allows for the examination of relative, rather than absolute, size variation in different regions of the brain. Procrustes superimposition reveals a pattern of variation in brain shape within Procyonidae, which suggests that encephalization can result from the expansion of different regions of the brain even within one family. Also important, is the observation that several shape differences seem to correlate with known aspects of procyonid behavior and ecology. The examination of how encephalization is achieved within a given clade (i.e., which regions of the brain are emphasized) may reflect the relative importance of given structures and uncover differences in encephalization not revealed by measures of overall brain size. Thus, endocast shape is not only useful for testing phylogeny but also potentially useful for inferring behavior and ecology of extinct taxa, where preservation of soft neural tissue is unlikely.
This project was completed as part of a lab rotation during pursuit of my Ph.D. at Johns Hopkins University School of Medicine. I thank my rotation advisor Valerie DeLeon for all her help with the project. I also appreciate feedback on this manuscript from Mary Silcox and one anonymous reviewer. Thanks to Madeleine Chollet, Christian George, Francois Gould, Megan Holmes, and Katrina Jones for discussion throughout the project's course. I appreciate access to specimens provided by the Field Museum of Natural History, Museum of Vertebrate Zoology, Royal Ontario Museum, and Vertebrate Paleontology Laboratory branch of the Texas Memorial Museum. I am grateful to the staff of the High-Resolution X-ray Computed Tomography Facility at the University of Texas at Austin (UTCT) for scanning and initial image processing of the specimens. Finally, I wish to recognize Blaire Van Valkenburgh for additional loans of HRXCT scan data.