The Brain Anatomy of the Brown Bear (Carnivora, Ursus arctos L., 1758) Compared to That of Other Carnivorans: A Cross-Sectional Study Using MRI

In this study, we aimed to provide a neuroanatomy atlas derived from cross-sectional and magnetic resonance imaging (MRI) of the encephalon of the brown bear (Ursus arctos). A postmortem brain analysis using magnetic resonance imaging (MRI – 1,5T; a high-resolution submillimeter three-dimensional T1-3D FFE) and cross-sectional macroscopic anatomy methods revealed major embryological and anatomical subdivisions of the encephalon, including the ventricular system. Most of the internal structures were comparably identifiable in both methods. The tractus olfactorius medialis, corpus subthalamicum, brachium colliculi rostralis, fasciculus longitudinalis medialis, nuclei vestibulares, velum medullare rostrale, nucleus fastigii, fasciculi cuneatus et gracilis were identified entirely by cross-sectional macroscopic analysis. However, the glandula pinealis, lemniscus lateralis and nuclei rhaphe were visualized only with MRI. Gross neuroanatomic analysis provided information about sulci and gyri of the cerebral hemispheres, components of the vermis and cerebellar hemispheres, and relative size and morphology of constituents of the rhinencephalon and cerebellum constituents. Similarities and discrepancies in identification of structures provided by both methods, as well as hallmarks of the structures facilitating identification using these methods are discussed. Finally, we compare the brown bear encephalon with other carnivores and discuss most of the identified structures compared to those of the domestic dog, the domestic cat, Ursidae and Mustelidae families and Pinnipedia clade.


INTRODUCTION
Relatively little is known about Ursidae brains' anatomy, in comparison with common domestic animals -the dog (Canis lupus familiaris), the cat (Felis catus) and the ferret (Mustela putorius furo). Among Carnivora, the domestic cat and domestic dog have been the most favored for an extensive neuroanatomical research employing the newest imaging techniques (e.g., Leigh et al., 2008;Mogicato et al., 2011b;Gray-Edwards et al., 2014), while for numerous species merely brain surface anatomy is known.
Most of the neuroanatomical analyses conducted so far have focused on sulci and gyri patterns in various Carnivora families (e.g., Mettler and Goss, 1946;Radinsky, 1973bRadinsky, ,a, 1975. Some studies can provide detailed information about internal brain structures due to employment of magnetic resonance imaging (MRI) combined with cross-sectional analyses. MRI method has been in use for about 20 years in studies of central nervous system features of given Carnivora species. An internal anatomical description of the canine brain using MRI (with 1,5 tesla magnet -1,5T) was performed by Leigh et al. (2008). There are also studies focused on comparison of the amount of data acquired with employment of different magnetic field MRI scanners: 1.5T to 7T (e.g., Kang et al., 2009) and 3T to 7T (e.g., Martín-Vaquero et al., 2010). Mogicato et al. (2011b) compared three transections using three methods: unstained and stained cross-sections, and virtual sections acquired by MRI (1T). Also the cranial nerves nearest to the brain parts with cranial foramina were investigated using MRI method (Couturier et al., 2005) Additionally, MRI techniques have been used to create neuroanatomical atlases for certain species. An innovative atlas of digitally smoothed canine brain surface showing area underneath the brain sulci using MRI has been performed by Datta et al. (2012), while online brain atlas published by Minnesota College of Veterinary Medicine (Fletcher, 2007) contains stained brain cross-sections and its MRI-acquired counterparts. Out of wild canids, an MRI and cross-sectional brain atlas of the red fox (Vulpes vulpes) has been created (Kassab and Bahgat, 2007). Sawada et al. (2013) studied the sexual dimorphism of neuroanatomical traits of the ferret telencephalon with use of MRI (7T) and immunohistochemistry, leading to a conclusion that total brain volume, cerebral cortex volume and subcortical white matter were bigger in males. The authors described the cerebral cortex's cytoarchitectonics by delineating motor and sensory areas of the pallium dorsale and the layers of bulbus olfactorius and visualized it in the colored map created with MRI-and immunohistochemistryacquired brain slides. Neal et al. (2007) used MRI to trace gyrification in the ferret with 4.7T and 8.5T magnetic strength. They visualized postnatal gradation of the sulci revealed toward the brain midline. Furthermore, postnatal neuronal proliferation, migration and level of myelination were studied on the basis of signal differences at T2 in the selected cortex and periventricular areas. It was shown that the growth rate of cortical volume exceeds the growth rate of the other analyzed subcortical structures. Barnette et al. (2009) studied alteration in MRI (4.7T) signal intensity during different stages of gyrification and reconstructed a 3D map of the gyri curvatures followed by comparison of gyrification in the ferret and human fetus. A featured brain MRI atlas (T2, 1,5T) of the California sea lion (Zalophus californianus) containing 20 transverse and 14 oblique cross-sectional MRI images was published by Montie et al. (2009). Many crucial brain components were depicted, and Pinnipedia signature traits were corroborated. MRI brain anatomy atlases were also developed for the cat (Yamada et al., 1994;Mogicato et al., 2011b;Gray-Edwards et al., 2014), and additionally combined with computer tomography (CT) to reveal the anatomical relation of cranial nerves with the brain and skull foramina (Gomes et al., 2009). A general brain anatomy of the young Bengal tiger (Panthera tigris tigris) could be also revealed by MRI (1.5T) in T1, T2 and diffusion MRI (DWI), and showed its utility for clinical assessment in veterinary neurology (Snow et al., 2004). As emphasized by this paragraph, MRI allowed revealing general and species-specific neuroanatomical traits among carnivores, supplementing knowledge acquired with traditional cross sections, while for some species the internal brain anatomy remains virtually unknown. Neuroanatomical features and volumetric analysis possible with MRI techniques available for all carnivores species, would allow inter-and intra-specific comparisons and its translation into functions in behavior and ecology of Carnivora.
Bears have evolved in Carnivora clade (Flynn et al., 2005;Supplementary Figure S1) during the late Oligocene and early Miocene, about 20-25 million years before present (McLellan and Reiner, 1994), attaining a wide geographical distribution range (Kumar et al., 2017). Family Ursidae comprises eight extant species, classified into three subfamilies: Ursinae (with polar bear Ursus maritimus, brown bear Ursus arctos, American black bear Ursus americanus, Asiatic black bear Ursus thibetanus, sloth bear Melursus ursinus, and Malayan sun bear Helarctos malayanus), Ailuropodinae (with giant panda Ailuropoda melanoleuca) and Tremarctinae (with Andean bear Tremarctos ornatus) (Wagner, 2010). Morphologically and taxonomically, bears possess all the traits of carnivores but, with the exception of the polar bear, have diets often comprised primarily of plant matter. A generalist omnivore strategy most of the bear species evolved with, allows them to successfully occupy a broad array of habitats (Robbins et al., 2004). Diverse habitats and its physical characteristics create challenges that influenced evolution of brain and sensory structures in bears as they have to detect through complex environment (e.g., Togunov et al., 2017). Additionally, bears are solitary and nonterritorial (Bellemain et al., 2006) and as such, they benefit from conveying multimodal cues to expedite breeding season, to recognize family and kin, and to avoid conspecifics. Available anatomical, histological and behavioral evidence suggests that bears are somatosensory specialists (Bacon and Burghardt, 1974;Nachtigall et al., 2007;Pastor et al., 2008;Owen and Bowles, 2011;Togunov et al., 2017;Tomiyasu et al., 2017), and possess neuroanatomical correlates for significant sensitivity of their extremities (Kamiya and Pirlot, 1988b).
We have performed analysis of a detailed brain atlas of the brown bear, from countable internal anatomy neurostructures acquired by two methods -macroscopic anatomy and MRI. We produced a series of virtual slices from MRI of postmortem brain, in which detailed structures could be recognized and then compared to them in manually performed cross-sections. Numerous structures have been recognized using both methods, and allow comparing to what extent the MRI scanning can reflect actual shape and expanse of given structure observed in cross sections. The atlas may be useful to veterinarians, as it is one of very few extensive sources of knowledge about the brain of a wild Carnivora species. In this paper we also present a comparative analysis of U. arctos brain with other Carnivora species, as well as with specimens available in Comparative Mammalian Brain Collections to further contribute to our understanding of brain organization. Those observations might be also used to infer ecological aspects that stem from particular brain traits.

Animals and Brain Tissue Preparation
For the present study, brains were acquired from traffic-killed specimens collected within a frame of Brown Bear Management Plan for the Republic of Croatia (Huber et al., 2008). The use of tissues from animals found dead is exempt from approval by Ethical Committee for Animal Experimentation. The laboratory procedures began from brain immersion-fixing in 4% buffered formaldehyde immediately after its removal from the skull. Table 1 contains sex, age, and collection site data for each specimen. The age was determined using premolar teeth cementum age determination method (Matson et al., 1993). The brain of the bear RH 178-15 was used as a model brain for the analyses due to its best quality in terms of tissue preservation.

Magnetic Resonance Imaging
The isolated brain of RH0178-15 was imaged at Centre for Experimental Diagnostics and Biomedical Innovations in Wrocław. MRI scanning of the brain was carried out on a 1.5-T Philips MRI scanner. 3D Fast Field Echo (FFE) T1-weighted images in the transverse planes were acquired in Brain of the specimen RH0178-15 was used as a model brain for MRI imaging and cross-sections.
isotropic resolution with the following parameters: slice thickness 0,33 mm; matrix 660 × 660; FOV (field-of-view) 22 cm; TE (echo time) 9,7; TR (repetition time) 25,0; Flip Angle 30. After scanning, the brain was re-immersed in 4% buffered formaldehyde and sectioned within 24 h. Visualization of MRI images was performed with Medical Imaging Interaction Toolkit (MITK) software and images in different planes were selected to make the best correspondence with anatomical cross-sections. MRI. The recognition of examined structures was performed on the basis of previously examined anatomical cross-sections and also using MRI brain atlas of the dog (Leigh et al., 2008) and Zalophus californianus (Montie et al., 2009). Photo processing was performed with Canvas X 2017 GIS software with data acquired directly from MITK software.

Macroscopic Anatomy
Before examination, brains had been irrigated with running water for 24-48 h. The measurements of RH0178-15 specimen were performed with a caliper (Table 2). Subsequently, the RH0178-15 brain was cross-sectioned into 5-mm-thick sections with a stainless-steel knife. The examination was performed with use of Nikon SMZ800 stereomicroscope, overhead magnifying glass (3,5x) and with the naked eye. The images were taken with reflex camera Nikon D80. An internal preservation of the remaining specimens was of lower quality than the brain RH0178-15. Therefore, they were used as the point of reference for the examination of the brain surface variability and gyrification pattern and some other dubious traits mentioned in the text, including presence of the septum pellucidum, and were not analyzed in all details. The photo processing was performed with Canvas X 2017 GIS software. Structures of interest were identified by comparison with available atlases of the dog brain (Meyer, 1964;Whalen, 2003;Fletcher, 2007;Leigh et al., 2008;Uemura, 2015), textbooks (Meyer, 1964;Kobryń and Kobryńczuk, 2004;Uemura, 2015), and previously published observations of bear brains (Mettler and Goss, 1946;Kamiya and Pirlot, 1988b). Comparisons with other Carnivora were performed with illustrations in above mentioned texts and specimens available in Comparative Mammalian Brain Collections (CMBC).

Nomenclature
Most of nomenclature was acquired from Nomina Anatomica Veterinaria 5th edition (International Committee on Veterinary Gross Anatomical Nomenclature [ICVGAN], 2005). The remainder came from cited literature (Mettler and Goss, 1946;Meyer, 1964). The source of each term is shown in Table 3. Pallium nomenclature comes from Comparative Vertebrate Neuroanatomy -Evolution and Adaptation (Butler and Hodos, 2005). For pallium mediale homologs we used term "hippocampal formation" to remove a common ambiguity concerning the term "hippocampus"; using sensu lato (all cortical structures) or sensu stricto (cornu ammonis). "Hippocampal formation" stands for the subiculum, cornu ammonis and gyrus dentatus with associated white matter -fimbriae and alveus hippocampi. "Rhinencephalon" in our study stands for primary olfactory structures (bulbus olfactorius, pedunculus olfactorius, tractus olfactorius) and associated superficial pallium laterale structures (olfactory cortex, lobus piriformis, gyrus olfactorius lateralis, tuberculum olfactorium). The English and Latin names of species mentioned in the paper come from Cichocki et al. (2015).

Telencephalon -Gross Anatomy Including Sulci and Gyri Pattern
The brain surface of the brown bear is highly convoluted, suggesting an elaborate underlying structure. The most medially located structure is the gyrus marginalis (GM) ( Figure 1A). GM is divided by the sulcus parietalis (SP) into two parts -a smaller medial and larger lateral part. However, we found that the trait varies -in another individual (RH 155/14) only unilateral division was present and in another one (RH 194/14) the gyrus was not divided at all. GM continues rostrally as a short gyrus postcruciatus (GPc) and then two winding structures-gyrus sigmoideus caudalis (GSC) and gyrus sigmoideus rostralis (GSR). The border between the latter two is the sulcus cruciatus indentation. GM is bordered laterally by sulcus marginalis (SM) with its caudal part bending to the lateral surface of the hemisphere ( Figure 1B) and a shorter rostral part continued as the sulcus coronalis (SC) intercepting a short sulcus ansatus (SA).
Gyrus marginalis continues caudally with an apparent narrowing where the gyrus occipitalis (GOc) is bordered laterally by the sulcus occipitalis (SOc) (Figure 1E). The gyrus ectomarginalis (GEm) runs parallel to fissura pseudosylvia (FP) where its rostral section borders with it ( Figure 1A). Three sulci on the dorsal aspect take a transverse position. The most rostral is the sulcus cruciatus (SCr) dividing the gyrus frontalis dorsalis (GFD) from the gyrus frontalis medius (GFM). More caudally but close to SCr runs sulcus postcruciatus (SPc) separating GFD from GPc ( Figure 1A). Both abovementioned sulci are also visible from the lateral aspect ( Figure 1B) and are surrounded by GSR and GSC. The third sulcus is the sulcus cruciatus minor (SCrM) running rostro-laterally separating GM and GSC from a rostrally lying GPc (Figure 1A). On the lateral aspect a deep but very tight FP separates the gyrus ectosylvius rostralis (GEsR) from the gyrus ectosylvius caudalis (GEsC) -those gyri are separated from overlying gyrus suprasylvius rostralis (GSsR) and gyrus suprasylvius caudalis (GSsR) by sulcus suprasylvius rostralis (SSsR) and sulcus suprasylvius caudalis (SSsC) respectively. The abovementioned gyri and sulci converge at an acute angle that constitutes the border between rostral and lateral gyri and sulci on the lateral aspect ( Figure 1B). The rostral pole of the cerebral hemisphere is occupied mainly by the gyri frontales ( Figure 1C).
The most signature structure, a very distinctive lozenge, is formed by the GFD with the contralateral gyrus. The horizontal part of GFM runs medially until its vertical part running to the pole of the hemisphere, eventually becoming the ventral aspect of the brain (Figures 1C,D). The gyrus frontalis ventralis (GFV) lies more ventrally and runs horizontally converging with GFM and bordering with the gyrus proreus (GPr) dorsally. GPr is separated from a dorsally lying GSR by the sulcus proreus (SPr) with the sulcus praesylvius (SPs) converging with it ( Figure 1C).
The main external structures of the rhinencephalon are easily recognized in the ventral aspect of the brain ( Figure 1D). The bulbus olfactorius (BO) is present only as its caudal parts in the investigated specimen. The pedunculus olfactorius (PO) is apparent and continued caudally as the tractus olfactoriuslateralis (TOL), medialis (TOM) and intermedius (TOI). The gyrus olfactorius lateralis (GOL) is separated laterally from the dorsal pallium gyri by the sulcus rhinalis lateralis -pars rostralis (SRLPR). The sulcus is continued caudally as its pars caudalis (SRLPC) and separates lobus piriformis (LP) from caudolaterally lying parts of the temporal lobe with the dorsal pallium derivatives. A round and distinctive tuberculum olfactorium (TO) lies just in the front of LO and medially from GOL. The remaining area of the ventral surface of the telencephalon is occupied by the dorsal pallium. The caudal surface is occupied by the gyrus compositus caudalis (GCC) as a result of GEsC, GSsC and GEm ventral convergence. Rostrally, GFM is separated from PO by the sulcus rhinalis medialis (SRM). Also a small part of GFV is noticeable laterally to PO.
All of the abovementioned gyri, sulci and the rhinencephalon structures were confirmed in the cross-sectional anatomical images. The MRI images appeared devoid of sulci and gyrionly rhinencephalic structures showed good correspondence with macroscopic anatomy cross-sections. The lack of right BO in Figure 2D is caused by preparation defect, however, it is noticeable bilaterally in MRI (Figures 2A,C). Furthermore, in cross sections of the medial aspect of the hemispheres, the sulcus cinguli (SCi) was depicted separating the gyrus cinguli (GCi) from an above-lying GFD in the rostral part ( Figures 4D, 5B,D, 6B), from GM in the middle part (Figures 6D, 7B) and from the gyrus splenialis (GS) in the caudal    The list of figures of the structures without abbreviation but present in U. arctos is also included but the structures are not pinpointed in the figures, for example the telencephalon. a, abbreviation not used; b, term absent in NAV 5th or other references. c, structure absent in U. arctos but mentioned in the paper. d, structure present in U. arctos but not captured in the figures. e, structure probably present in U. arctos but not identified.

The capsula interna (CI) is visible both in MAC and MRI
TOL (Figures 3C,D, 4, 5B,D) and TOI (Figures 4, 5) were depicted in both MAC and MRI. However TOM proved to be very thin and was detected only in MAC ( Figure 4D). The olfactory cortex (OC) is visible in Figures 3D, 4B,D.
The thalamus (Th) with countable blood vessels was depicted as a broad centrally located gray matter or a hyperintense area (Figures 6C,D, 7). However, none of the hypothalamic nuclei were apparent, with the exception of the most posterior region -the pulvinar (Pv) (Figure 7). The corpus subthalamicum (CSu) was only depicted in MAC ( Figure 7B) and the nucleus endopeduncularis (NEp) in both methods (Figures 6, 7A,B). The zona incerta was not identifiable in any of the crosssections. The proper hypothalamus -nuclei hypothalamici (NH) is also depicted (Figure 6) but none of the particular nuclei were identifiable. The gray matter of superficial structures -TC (Figures 6C,D) and CMa was also visible in the cross sections (Figures 7A,B). Tractus opticus (TOp) (Figures 6C,D, 7) were also depicted. Both the corpora geniculata -corpus geniculatum mediale (CGM) et laterale (CGL) with their nuclei were identifiable in both methods (Figures 7C,D, 8A,B). The glandula pinealis (GlPn) is depicted only in MRI as highly hyperintense round area ( Figure 7C). The habenulae (Ha) were identified with both methods (Figures 7A,B) as well as the comissura caudalis (CCa) -above the ventriculus tertius (VT) -aquaeductus cerebri (AqC) border (Figures 7C,D). that in MRI the signal from the pedunculi is similar to that of gyri. (C,D) MRI and macroscopic anatomy cross-section through the anterior part of cerebral hemispheres just before the inception of ventriculus lateralis lumen. The insets in (A,C) shows the exact plane of cross-section in MRI and the best achieved correspondence in macroscopic anatomy (B,D). In (B,D) gyri and sulci of the telencephalon were written in green color. GFD, gyrus frontalis dorsalis; GFM, gyrus frontalis medius; GFV, gyrus frontalis ventralis; GOL, gyrus olfactorius lateralis; GSR, gyrus sigmoideus rostralis; GSsR, gyrus suprasylvius rostralis; OC, olfactory cortex; PO, pedunculus olfactorius; SC, sulcus coronalis; SCr, sulcus cruciatus; SPs, sulcus praesylvius; SRLPR, sulcus rhinalis lateralis, pars rostralis; TOL, tractus olfactorius lateralis.

Cerebellum -Gross and Internal Anatomy
The hemisphaeria cerebelli (HC) are quite laterally expanded and the fissura prima (FPr) takes a U-shaped course ( Figure 1A). The vermis (Vm) is noticeably divided into lobules ( Figure 1E). Lobules from the caudal and dorsal aspects are listed from ventral to dorsal: a poorly separated uvula vermis (UVm), a quite well separated pyramis vermis (PVm), an apparently protruding tuber vermis (TVm), two lobuli poorly separated one from anotherfolium vermis (FVm) and declive (Dc) are eventually distinctly separated from the culmen (Cu) by the abovementioned FPr.
The cerebellar nuclei were also detected. The nucleus fastigii (NF) was identifiable only in MAC ( Figure 11B). Another two structures could be recognized with both methods -the nucleus interpositus cerebelli (NIC) (Figures 10A,C,D, 11A,B) and the nucleus lateralis cerebelli (NLC) (Figures 10A,C,D). It is worth to mention that in Figures 11A,D parts of cerebellar cortex were depicted as indentations into the white matter from  the back -therefore they resemble the cerebellar nuclei. The pedunculi cerebellares were distinct: rostralis (PCR), medius (PCM) (Figures 9C,D) and caudalis (PCC) (Figures 10, 11B,D).

Pons and Medulla Oblongata
The pons is visible on the ventral surface as distinctive bulge with transversely coursing fibers, including a medially coursing sulcus basiliaris (SBa) and the arteria basiliaris on its surface (Figure 1D). At the rostral margin of the pons, a portion of the nervus trigeminus (NV) was detected. A thin nerve visible on the one side is probably the nervus abducens (VI) (NVI). Two small bulges bordering with the caudal margin of the pons are the pyramides (Py) of the medulla oblongata (MO).
Only a few structures of the pons were identifiable. Within the pars ventralis pontis, we found the nuclei pontis (NPo) (Figures 8C,D, 9) and the fibrae pontis transverse FPoT) ( Figures 8C,D, 9). For the pars dorsalis pontis, we could find the rhaphe (Rh) (Figures 9C,D), but the nuclei rhaphe (NRh) were visible only in MRI ( Figure 9C). Some of gray matter within MO was also identifiable in cross-sections. The most prominent was the nucleus olivaris (NOv) (Figures 10C,D, 11) with its hilus (HNOv) apparent in MAC but also in one MRI image (Figure 11A). The nucleus cuneatus lateralis (NCL) forms a well identifiable gray matter over a long distance (Figures 10C,D, 11A,B). The nucleus cuneatus medialis (NCM) was also detected (Figures 11C,D, 12A,B). The nucleus gracilis (NG) was more apparent in MRI (Figures 11C,D, 12B) due to the bulge it forms on the fossa rhomboidea -tuberculum nuclei gracilis (TNG). A highly hyperintense asymmetrical area visible in Figure 9A is probably a lesion. The nuclei vestibulares (NVb) were collectively intentified only in MAC near the ventriculus quartus (VQ) (Figure 10B). TSNT along with the nucleus tractus  spinalis nervi trigemini (NTSNT) were apparent in MAC and MRI, across the medulla oblongata and end segment of the pons (Figures 9C,D, 10-12).
(Ursus maritimus) having a considerably more convoluted brain surface than other sampled Carnivora species, surpassed only by two Pinnipedia species (Lyras et al., 2016).
the American black bear Ursus americanus, the red panda Ailurus fulgens (Ailuridae) and the racon, Procyon lotor (Procyonidae)] have been investigated (Kamiya and Pirlot, 1988b). The shape of fissura pseudosylvia and adjacent gyri were compared, and brain regions with putative somatosensory and motor area (gyrus suprasylvius rostralis = gyrus coronalis) appeared more expanded than in Canidae, assigned to ursiformes' more skillful use of hands and fingers.
A signature trait of the Ursidae and Pinnipedia brain is a pronounced development of the gyri frontales (Figures 1A,C). Mettler and Goss (1946) marked the gyri with terms "superior, " "medius" and "inferior"; notably, the gyrus frontalis is not included in NAV. We used the terms "dorsalis" and "ventralis" in lieu of "superior" and "inferior" to unify the nomenclature. The left and right GFD together form the signature "ursine lozenge" as termed by Mettler and Goss (1946). The structure, GFM and GFV are all apparent in Ursidae (Mettler and Goss, 1946;Kamiya and Pirlot, 1988b;Welker et al., 2009) and Pinnipedia (Montie et al., 2009;Welker et al., 2009). On the other hand only one gyrus frontalis is present in the dog (Meyer, 1964;Uemura, 2015) and in the cat none of the gyri frontales is observed (Smith et al., 2001;Kobryń and Kobryńczuk, 2004). The course of SCr ( Figure 1A) is variable in our samples. SCr is defined as deep sulcus running from the medial surface of the hemisphere laterally (Meyer, 1964). In our specimen (RH0178-15) the marked SCr begins bilaterally in medial surface of the hemispheres, similar to U. maritimus in CMBC and H. malayanus (Kamiya and Pirlot, 1988b). The situation is complicated by somewhat wedged GFD into SCr (Figures 1B,C). In RH0178-15, RH0156/13, and in U. maritimus in CMBC the rostral sulcus to GFD begins in the medial surface of the hemispheres. In the same U. maritimus the caudal sulcus to GFD merge with the SPc and the same state we have observed on either side of the RH0178-15 brain ( Figure 1A). However,  in another specimen of U. arctos (RH194/14) the caudal sulcus to GFD is apparent at the medial surface. Dong (2008) also depicted the caudal sulcus as SCr in A. melanoleuca (2008), while Mettler and Goss (1946) depicted only the lateral section of SCr, suggesting the problematic medial section was omitted from their study. We suggest three possible models: (a) rostral sulcus to GFD is the homolog of SCr; (b) caudal sulcus to GFD is the homolog; (c) both caudal and rostral sulci to GFD are the homolog of SCr present in the other Carnivora that was separated by the developing GFD. The issue can be elucidated by further cytoarchitectonic or gyrification studies in U. arctos. Interestingly, the rostral area to SCr is supposedly involved in social behavior, as studied in Crocuta crocuta (Holekamp et al., 2007;Sakai et al., 2011) and in Panthera leo (Sakai et al., 2016). The area would be substantially larger if GFD is also included.
Another transverse sulcus -SCrM, is defined in the dog as the sulcus beginning from about the middle of SCi and reaching the dorsal margin of the brain at the level SA (Meyer, 1964). SCrM is well developed in the examined U. arctos specimen (Figure 1A) as well as in A. melanoleuca (Mettler and Goss, 1946), U. maritimus available in CMBC and H. malayanus (Kamiya and Pirlot, 1988b). It is plausible that SCrM is exceptionally well developed laterally in Ursidae as in all abovementioned species it reaches the dorsal surface of the brain and separates the caudally lying GM from FIGURE 12 | MRI imaging and cross sections through the most caudal part of cerebellum and the medulla oblongata -medulla spinalis transition. Note the decussatio pyramidum and gray matter typical for the medulla spinalis. The dorsal gray matter is typical for the nuclei of the medulla oblongata. The gray matter of cerebellum is no longer visible. In (B) the cerebellar structures were written in blue color. The inset in (A) shows the exact plane of cross-section in MRI and the best achieved correspondence in macroscopic anatomy (B the rostrally lying GPc (Figure 1A). SCrM is absent in the cat (Smith et al., 2001;Mogicato et al., 2011b;Gray-Edwards et al., 2014), Viverridae (Radinsky, 1975), and Mustelidae (Radinsky, 1973a). Pinnipedia species' SCrM is apparent, suggesting the presence of the sulcus is typical of Caniformia. We did not observe dimples in the anterior section of U. arctos telencephalon, although they are apparent in Mustelidae (Radinsky, 1973a) and Hyaenidae (Holekamp et al., 2007;Sakai et al., 2011). SPs in U. arctos is relatively short (Figures 1B,C) in comparison to that in the ferret (Mustela putorius furo) (Sawada et al., 2013), other Mustelidae including T. taxus, M. meles (Radinsky, 1973a) as well as numerous Viverridae including Hemigalus derbyanus, Diplogale hosei and Eupleridae, e.g., Fossa fossa (Radinsky, 1975). SPs is exceptionally long and arcuate reaching up to the dorsal surface of the hemisphere. GM in U. arctos ( Figure 1A) is more elaborate than that in the dog (Whalen, 2003;Uemura, 2015) because of its secondary division into two unequal parts. Some Ursidae show a longitudinal discontinuity, e.g., CMBC brains of H. maritimus and U. americanus and A. melanoleuca (Mettler and Goss, 1946) supporting the notion that the gyrus assumes a sigmoid course in many carnivores. The dog GM appears divided into nearly equal parts by the sulcus endomarginalis (Meyer, 1964). The rostral continuity of GM is GSC (Meyer, 1964), however the term "gyrus sigmoideus" is not included in NAV. According to Whalen (2003) in the online atlas a whole gyrus caudal to SCr is termed as the "GPc" and rostral to SCr as the "gyrus praecruciatus". Meyer (1964) defined GPc as an additional gyrus, located caudally to SPc if the sulcus is present. We retained the term "GSR et caudalis" for the rostral and caudal gyri to the SPc and "GPc" for the gyrus immediately caudal to SPc because we observed this anatomical trait consistently across Ursidae.
The caudal part of the cerebral hemisphere is the most problematic. We depicted GOc following the online atlas of Whalen (2003) considering it as continuity of GM onto the caudal surface of the hemisphere (Figures 1A,E). Smith et al. (2001) used the same term in their description of the cat brain surface. An alternative term might be the "gyrus postmarginalis" (in former papers "gyrus postlateralis") (Meyer, 1964). However, the sulcus separating GM from GEm in U. arctos does not constitute a natural continuity of SM, thus we prefer the terms GOc and SOc, denoting the lateral course of the SOc. The evident narrowing separating GM from GOc supports our thesis -in the studied specimen the narrowing is visible better on the left side (Figures 1A,E). Another doubtful gyrus is GEm-depicted by us after Mettler and Goss (1946) and Dong (2008) (Figure 1B). If the gyrus were to be a homolog of that in the dog (Uemura, 2015) only its caudal part would be present in U. arctos. Instead, both GM and GEm seem to constitute a single complex gyrus lacking a strict border between them. In other Ursidae the situation is also complicated (Mettler and Goss, 1946;Kamiya and Pirlot, 1988b). To fully resolve the issue of posterior gyri homologies, further cytoarchitecture studies are required.

Rhinencephalon
BO (Figures 1B,D, 2A,C,D) was shown only in single MAC due to technical difficulties. It is comparatively apparent in both types of cross-sections. Portions of BO were not preserved, precluding inference of the relative volume of the structure in other Carnivora. In general, the BO in Ursidae is relatively large (Mettler and Goss, 1946;Kamiya and Pirlot, 1988b) and its caudal region is cylindrical, as found in the dog (Whalen, 2003;Uemura, 2015), H. malayanus (Kamiya and Pirlot, 1988b) and U. americanus (Welker et al., 2009). A relatively large BO with relatively short PO is present in many lesser, macrosmatic Carnivora representatives -Mustelidae (Radinsky, 1973a), Viverridae (Radinsky, 1975) or Felidae available in CMBC. BO in A. melanoleuca is more spherical and rostrally centralized with a very prominent PO containing a distinct narrowing (Mettler and Goss, 1946). On the other hand, BO in Pinnipedia is relatively smaller, reflecting the decrement in olfaction importance in that group (Montie et al., 2009). PO is easily distinguishable in MAC (Figure 3B), but in MRI it is barely separated from the brain surface, similar to telencephalic gyri. TOL and TOI are both apparent in both MAC and MRI, due to their hypointensity (Figures 3C,D, 4, 5). Initial sections of TOI are more evident in MRI and terminal sections of TOL were distinct only in MAC (Figures 5B,D). TOI poses a hallmark for macrosmatic Carnivora (Kobryń and Kobryńczuk, 2004). TOM was found only in MAC ( Figure 4D); TO is visible in both methods (Figures 4D, 5) but the MRI signal interfusion with NC is possible (Figure 5A). TO in U. arctos is nearly perfectly round and that shape seems to be feature of Ursidae -the same is observed in A. melanoleuca (Mettler and Goss, 1946) and H. malayanus (Kamiya and Pirlot, 1988b). LP assumes the fist-like shape in U. arctos ( Figure 1D) as well as in other Ursidae (Mettler and Goss, 1946;Kamiya and Pirlot, 1988b;Welker et al., 2009). Its longitudinal measure is relatively shorter than that in C. lupus, Vulpes zerda, Canis latrans available in CMBC, Proteles cristata (Flower, 1869) and A. fulgens (Flower, 1870). LP seems to be smaller in relation to the whole brain in U. arctos than in abovementioned Canidae species as well as in Mustelidae (Radinsky, 1973a) and Viverridae (Radinsky, 1975). The LP is relatively larger than in Pinnipedia, although it shares a similar shape. Further volumetric analysis of U. arctos brain would provide information about relative volume of rhinencephalon and, consequently, about comparative importance of olfaction in Ursidae. Dong (2008) studied brains of panda ancestors, †A. baconi and †A. microta (both extinct), as well as giant panda and polar bear using virtual endocasts (endocranial casts acquired with CT). This volumetric analysis proved that the polar bear has the largest rhinencephalon among the abovementioned ursid species.

Subcortical Gray Matter
Accurate borders of Pu are possible to delineate only in MAC (Figures 4B,D, 5B,D, 6B,D, 7B). It is also apparent in MRI, usually along with GP (Figures 4A,C, 5A,C, 6C) however, its borders are less clear, see Figure 5C. Both structures form a hyperintense zone in medio-ventral part of the telencephalon. It was possible to distinguish Pu from NC (converging in the initial section) in MAC ( Figure 4B) whereas in MRI both structures form one fused hyperintense zone -we labeled the zone as Pt + NC (Figure 4A). Pu in U. arctos always lies ventro-medially to Cm and assumes shapes from lentoid to triangular. Pu was also detected in the dog (Mogicato et al., 2011b) using unstained cross-sections and is discernible in MRI, in the cat (Gray-Edwards et al., 2014) and in Z. californianus (Montie et al., 2009). However, in the dog it was either undetected (Mogicato et al., 2011b) or collectively depicted as the nucleus lentiformis (Leigh et al., 2008). Borders of GP (Figures 4C,D, 5, 6) are also vague in MAC but it forms a consistently lighter area than Pu, in medial proximity. After staining, improved contrast between gray and white matter it is easily distinguishable (Mogicato et al., 2011b). GP in MRI is often discernible as the nucleus lentiformis but in the California sea lion, Z. californianus, it was possible to detect both the structures (Montie et al., 2009). We also managed to discern Pu and GP separately in one of MRI section (Figure 5C). Cm is lying under the sulcus rhinalis lateralis (that is the pallium dorsale -pallium laterale interface) and as a distinct structure, more evident in MAC (Figures 4B,D, 5B,D, 6B,D, 7B,D) with an apparent CEx laterally and capsula externa medially to it. Cm in MRI usually resemble the cerebral cortex thickening (Figures 5C, 6A,C, 7A). Cm signal intensity may be slightly lower than cortex intensity -then it is a little more difficult to spot but it does not form a deceptive cortex thickening (Figures 5C, 6A). It is extremely difficult to distinguish CEx ( Figure 5A) and CE ( Figure 6A) as hypointense narrow strips. Cm in its initial section is crescent-shaped in U. arctos (Figures 4B,D, 5B,D, 6B,D), whereas more lentoid in its caudal parts (Figures 7B,D). Cm was also visualized in the dog (Fletcher, 2007;Mogicato et al., 2011a) and in the cat (Smith et al., 2001;Gray-Edwards et al., 2014) and the vertical measure of Cm in both these species is greater than in U. arctos, more pickaxe-shaped than crescent-shaped. In other MRI studies, Cm is poorly discernible (Leigh et al., 2008;Kang et al., 2009;Mogicato et al., 2011b) or undetectable (Mogicato et al., 2011a). Anterior and middle parts of the NC (caput et cauda) are very well identifiable in both methods (Figures 4, 5, 6A,B,D) and in MRI are hyperintense. The signal appears faded for the cauda nuclei caudati and is more difficult to observe in MRI (Figures 6C, 7A). In MAC it is evident near the lateral ventricle wall but evinces lesser diameter (Figures 7B,D). The caput of NC occupies a broad periventricular area and assumes a somewhat triangular shape. The area decreases for the corpus of NC and the cauda of NC forms merely a thin periventricular strip of gray matter. NC was also visible in unstained cross-sections and MRI in the cat (Gray-Edwards et al., 2014) and the dog (Mogicato et al., 2011b) as well as in MRI of Z. californianus (Montie et al., 2009) and the ferret (Sawada et al., 2013). Only Mogicato et al. (2011b) have not visualized NC in MRI, but hardly any other subcortical telencephalic structure was apparent. The shape and relative volume of NC, in proportion to the whole telencephalon, is very similar between U. arctos, the dog (Mogicato et al., 2011b), the cat (Gray-Edwards et al., 2014) and the ferret (Sawada et al., 2013). The comparison of NC in Ursidae and the ferret is stymied - Sawada et al. (2013) measured NC and nucleus lentiformis volume to the whole encephalon, whereas Kamiya and Pirlot (1988b) measured the striatum proportion in A. melanoleuca, U. americanus H. malayanus without clarification of the term "striatum." The striatum may involve other structure in surrounding tissues, depending on interpretation (Butler and Hodos, 2005).
NAc is usually defined as separate structure within NC, in the form of a thin strip within medial and lateral wall of the VL (Kobryń and Kobryńczuk, 2004). A noteworthy identification of NAc in MRI, where its borders are sharp and signal is strong ( Figure 5C). In MAC its borders are unclear ( Figure 5D) and distinction between NAc and ventral part of NC is difficult. NAc in U. arctos extends in both walls of VL near TOI and appears congruent to the structure depicted by Fletcher (2007) in the dog. CAg was observed in both methods but is more evident in MAC, where its nuclei with interwoven strips of white matter are visible (Figures 6B,D). The mosaic is not observed in MRI but in general, the signal of CAg is quite strong within the LP rostro-laterally to the incipient hippocampal formation (Figures 5C, 6A). CAg was also observed in MRI in the dog (Leigh et al., 2008), the cat (Gray-Edwards et al., 2014), Z. californianus (Montie et al., 2009) and the ferret (Sawada et al., 2013) with similar localization, extent and general appearance. SeC is easily identified in both methods (Figures 4D, 5). SeP was absent -none of the traced MRI cross-sections contained the structure nor the other examined specimens with macroscopic anatomy method (RH0156/13, RH155/14, RH194/14). It is plausible that SeP is then highly reduced in U. arctos in a manner that CC is directly adjacent to SeC and Fx. However SeP is present in U. maritimus seen in CMBC and easily discernible in three unidentified cross-sections. Also it is present and prominent in Z. californianus (Montie et al., 2009) and appears relatively short in the dog (Meyer, 1964;Fletcher, 2007). Its presence was not mentioned in papers devoted to the cat (Mogicato et al., 2011a;Gray-Edwards et al., 2014) and we investigated CMBC (Welker et al., 2009) to corroborate that indeed it is also absent in the cat.
Hippocampal formation is easily identifiable in MRI due to alternate array of hypo-and hyperintense strips (Figures 6C, 7A,C, 8A). Cortical parts of hippocampus are hyperintense with the gyrus dentatus (GD) evincing the least intensity of them and therefore the GD is more difficult to spot (Figures 7A,C). The alveus hippocampi (AH) is also difficult to descry. GD on the other hand is very apparent in MAC, as are AH, cornu ammonis (CA) and subiculum (Sb) (Figures 6D, 7B,D, 8B). The dorsal part of the hippocampal formation, however, scant, is easy to identify in MAC. In MRI the dorsal part is hypointense, with a barely discernible hyperintense strips corresponding to cortical parts but the more accurate identification is impossible due to weak contrast ( Figure 7C). The tissue of the hippocampal formation evinces localization typical for Carnivora -both ventral and dorsal, which stems from the curvature of the structures along Fx. In fact, the cross-sections involve the same structures captured in two locations and referred to as the dorsal and ventral hippocampus (Butler and Hodos, 2005). The ventral hippocampus (VH) is more predominant in U. arctos than in the dog (Uemura, 2015), the cat (Gray-Edwards et al., 2014), T. taxus (Welker et al., 2009) and the ferret (Sawada et al., 2013) where the dorsal (DH) and VH are comparably well developed. On the other hand, in Z. californianus nearly all hippocampal tissue is localized ventrally with a scant amount dorsally (Montie et al., 2009). DH in U. arctos containing gray matter is localized largely caudally (Figures 7D, 8B). It distinguishes its organization from that in the dog, cat and Mustelidae and situate closer to Pinnipedia. Reduction of DH in U. arctos is not that advanced than in Pinnepedia.

White Matter
CC due to obvious topography is easily identified in both methods (Figures 4-7, 8B), being hypointense in MRI, like all the white matter of the hemispheres. However, in MAC, transverse course of fibers is apparent because of sharpening of the borders between the corpus callosum and the remaining white matter. In Figure 4 the genu, a knee-like structure of CC, was captured. The only difference is the vicinity of Fc in U. arctos with absent SeP unlike in most other Carnivora. All parts of Fx were depicted (Figures 6, 7). The signal intensity of Fx is comparable with CC, but a distinctive narrowing facilitates delineating the border between these two structures ( Figure 6A). Dorsal part of Fx is comparatively well discernible in both methods (Figures 6A,B), but ventral parts due to hypointensity are better discernible in MRI (Figure 6A). The crura fornicis (CrFx) are laterally expanded, however, as mentioned above, the hippocampal tissue is not present until the most caudal sections of Fx with its caudal part associated with hippocampal tissue is more evident in MAC ( Figure 7D). In MRI they are hypointense, similar to the gray matter of DH, resulting in less clear borders. The CEx is generally difficult to observe in MRI (this study). It was undetectable in the cat (Gray-Edwards et al., 2014) and in the dog (Leigh et al., 2008). The same goes for CE (Figures 4D, 5C,D, 6A,B,D, 7B). The most prominent is CI (Figures 4-6, 7A,B,D) and in Figures 7B,D its transitions into the left and right CrC were captured. Unlike the other capsules, CI is apparent in MRI (Leigh et al., 2008).

Thalamus and Subthalamus
Rostral parts of Th are more evident in MAC where the border between Th and a surrounding CI is easier to delineate ( Figure 6D). Caudal structures are belied by non-homogeneous texture stemming from the presence of multiple thalamic nuclei groups (Butler and Hodos, 2005). The borders between nuclei groups are unclear, so we did not make an attempt to detect them without staining. Even stained tissue lacked definite thalamic nuclei (Fletcher, 2007;Mogicato et al., 2011a). Penetrating blood vessels are apparent in MAC as black dots (Figures 6D, 7B,D). Th in MRI is discernible as a broad central hyperintense area. The signal intensity is the highest in more medial parts and it gradually decreases laterally finally reaching CI hypointensity (Figures 6C, 7A). The intensity of Pv is lesser than the remaining parts of Th (Figures 7A,C). The adhaesio interthalamica (ATh) is vast in U. arctos (Figures 6C,D, 7A,B) and it considerably reduces the lumen of the VT. ATh is also broad in the dog (Fletcher, 2007), the cat (Gray-Edwards et al., 2014), Z. californianus (Montie et al., 2009), and C. crocuta (Welker et al., 2009). CSu was identified only in MAC as a separated area below Th and over the CI -CrC transition ( Figure 7B). The label is consistent with that done by Fletcher (2007) and with the literature definition (Kobryń and Kobryńczuk, 2004). NEp is remarkably discernible in MRI as a hyperintense, oval zone medially to Pu -GP and dorsally to TOp (Figures 6A,C, 7A). NEp in MAC is also apparent, especially in its caudal segment (Figures 6B,D, 7B). NEp is the homolog of the lamina medullaris medialis of GP in Primates (Butler and Hodos, 2005). NEp was also depicted in the cat by Gray-Edwards et al. (2014) and in the dog (Fletcher, 2007). The zona incerta was not identified, suggesting why the amount of information about the other Carnivora subthalamus is scant.

Hypothalamus
Collectively, NHs are evident in both methods (Figure 6). No hypothalamic nuclei could be detected and depicted. Previously, it proved impossible to detect white and gray matter by contrast staining in the dog (Fletcher, 2007;Mogicato et al., 2011b). NHs are hyperintense in MRI and well contrasted with white matter in MAC. The hypophysis was not preserved during preparation and only If is depicted (Figure 1D). The latter was not captured in anatomical cross-sections (the cuts run exactly just before the rostral and caudal border). TC is prominent and features darker coloration in MAC than NH (Figure 6D). In MRI, the signal intensity is comparable (Figure 6C). CMas were cut in their caudal parts in a way that their gray matter is no longer visible in MAC ( Figure 7B). CMas in MRI are apparent as paired highly hyperintense areas thus nuclei mamillares were probably involved ( Figure 7A). Despite of the method CMa are easily identified due to paired bulges formed in ventral diencephalon. The feature typical for Carnivora (and a few other orders) is parity of CMa, as documented in the dog (Fletcher, 2007), the cat (Gray-Edwards et al., 2014), and Z. californianus (Montie et al., 2009). COp is discernible in MAC due to intensive white color highly contrasted with the surrounding tissues (Figures 5D, 6B) and in MRI it forms a round hypointense zone isolated ventrally from the rest of the brain (Figure 6A). TOp is more evident in MAC as it is highly contrasted and the border between it and ventral surface of the brain is sharp (Figures 6D, 7B,D). The contrast is poor in MRI and the border is more difficult to spot (Figures 6C, 7A,C).

Corpora Geniculata
The gray matter of CGL is featured by weak signal intensity in MRI (Figure 7C), whereas in MAC the contrast is very high, making identification of the pars dorsalis and pars ventralis nuclei geniculati lateralis (NGL) easier ( Figure 7D). Similarly, TOp flanking lateral CGL was depicted in MAC, but showed little contrast in MRI. The nucleus geniculatus medialis (NGM) is also poorly discernible in MRI ( Figure 7C) and in MAC the contrast is lower than that of NGL, however, it forms a lateral bulge. CGM that facilitates to indicate the structure accurately and it is apparent in the right side of Figures 7D, 8B and bilaterally in MRI (Figures 7C, 8A).
LGN of U. arctos is similar in its relative volume to that in the dog (Whalen, 2003). The cytoarchitecture was indiscernible without staining but distinct division into its dorsal and ventral parts is evident. The comparison with LGN of other Carnivora (Sanderson, 1974) requires staining. Topographic relation of both CGL and CGM is similar in the dog -CGL forms an ovate bulge in the caudal diencephalon and CGM is collocated with the medially lying CoR (Whalen, 2003). All the structures are also discernible in MRI in the dog (Leigh et al., 2008) and the cat (Gray-Edwards et al., 2014) but only NGM is apparent in Z. californianus (Montie et al., 2009).

Epithalamus
GlPn was depicted only in MRI ( Figure 7C). It forms a hyperintense round zone of considerable size, separated from the dorsal surface of the brainstem ( Figure 7C). However, in Z. californianus the gland was even larger, occupying a broad area above CCa (Montie et al., 2009). Large volume of GlPn is a typical trait for Pinnipedia (Jasiński, 1987). Ha was depicted in both methods and appear hyperintense in MRI ( Figure 7A). In MAC the contrast between Ha and Th is low but due to characteristic bulges, Ha is easily identified, regardless of the presence of its nucleus (NHa) (Figure 7B). NHa is discernible in MRI due to hyperintensity. CCa is apparent in both methods as a transverse hypointense in MRI band over VT -AqC transition (Figures 7C,D).

Tectum
Out of tectal structures two pairs of colliculi were depicted. CoR evinces striatal organization visible in MRI as horizontally alternating hypo-and hyperintense bands (Figures 8A,C). The strata were also clearly visible to the naked eye in macroscopic investigation, although difficult to resolve in Figures 8B,D. CoC with NCoC takes a more ventro-lateral position to the former and are highly hyperintense in MRI ( Figure 9A) and well discernible in MAC -captured in two cross sections (Figures 8D, 9B)whereas NCoC is discernible only in Figure 9B. It is noteworthy that despite the hyperintensity in MRI, NCoC is only slightly contrasted with surrounding tissues in MAC, although its border is very sharp. Both CoR and CoC were also visualized in MRI in the cat (Gray-Edwards et al., 2014), the dog (Leigh et al., 2008), and Z. californianus (Montie et al., 2009). The size of both pairs of colliculi is similar in U. arctos and that is also observed in Z. californianus (Montie et al., 2009). However, in the dog, CoR is considerably less than CoC (Kobryń and Kobryńczuk, 2004). CCoR and CCoC are hypointense in MRI (Figures 8C, 9A) and in MAC they form typical transverse and well-discerned strips of white matter (Figures 8D, 9B). BCoR were apparent only in MAC and BCoC in both methods. The latter is identifiable in MRI because of the bulge it forms on the brainstem surface ( Figure 8C). In MAC both pairs of brachia are apparent, including parts located within the mesencephalon (Figures 8B, 9B). Only CCoR and CCoC were discernible in MRI in the cat (Gray-Edwards et al., 2014). In other reviewed papers with MRI method white matter of the tectum was not depicted.

Tegmentum
Out of the tegmentum the most prominent structures were SN and NR. SN is very distinctive in MRI forming a hyperintense band dorsal to a highly hypointense CrC (Figures 7C, 8A). In MAC it also displays considerable contrast with surrounding white matter (Figures 7D, 8B,D). The area occupied by SN is expanded in Figure 7D where one can descry its pars reticulata ventrally and pars compacta dorsally. NR is not apparent in its initial sections in MRI (Figure 7C), although it forms a hyperintense zone dorsally to SN in further sections (Figures 8A,C). It is discernible in MAC as it forms relatively well-contrasted round zone with sharp boundaries (Figures 7D, 8B,D). In U. arctos NR assumes a beanlike shape. Inference of the shape of NR in other Carnivora is stymied because of difficulties with imaging it accurate borders in MRI, e.g., in the dog (Leigh et al., 2008) or lack of visualizing the structure at all, despite beautiful preparations performed by Gray-Edwards et al. (2014) in the cat. Stained cross-sections of other Carnivora species available in CMBS do not improve the ability to infer the NR borders, but the structure is generally discernible (Welker et al., 2009). In Figure 7D it is plausible that only left NR is captured whereas the right structure may be CSu because of slight section asymmetry and their relative vicinity (Kobryń and Kobryńczuk, 2004). Around AqC a ring of gray matter, SGC, was depicted. Due to its characteristic topography and cordiform shape it is easy to notice in both methods. SGC proved to be hypointense in MRI and its outer borders are nearly impossible to delineate (Figures 8A,C, 9A) in contrast to well-defined borders and high general contrast in MAC (Figures 7D, 8B,D, 9B). NIp assumes dark coloration and is very distinct from CrC in MAC (Figures 8B,D). NIp is also discernible in MRI, even though the signal intensity is comparable with that of FRM; a signature hypointense rim makes it possible to unambiguously localize the structure ( Figure 8C). The IN is strongly stained in the dog where it takes a rectangular shape (Mogicato et al., 2011a), however in MRI in the Z. californianus (Montie et al., 2009), and the cat (Gray-Edwards et al., 2014) IN was not observed at all. The area typical for NMNO was detected (Figures 7D, 8B,D) and it was not possible to discern the structure in MRI. In Figure 7D NMNO forms an apparent wedged zone right below AqC and in further cross-sections it forms a less apparent zone ventrally to SGC. CrC due to its localization and uniform histology are easily identifiable in both types of cross-sections. They form hypointense semilunar or ovate zone in the ventral part of the mesencephalon (Figures 7A,C, 8A,C, 9A). Detection was more difficult in MAC, in more caudal sections with pars ventralis pontis inception -because they take up a more internal location (Figure 9B). A much smaller bundle -FLM, was discernible only in MAC as a subtle area of white matter just under SGC (Figures 8D, 9B). DPCR was detected in both methods. In MRI it appears to be a hypointense central zone with poorly defined boundaries ( Figure 8C). In contrast, DPCR seems to form a broad light-colored central zone in MAC in the ventral part of the mesencephalon and dorsally to IN (Figures 8D, 9B). LL was identified only in MRI ( Figure 9A) as a hypointense lateral band with relatively poor defined borders, as previously noted in MRI for Z. californianus (Montie et al., 2009) and the cat (Gray-Edwards et al., 2014) as one of very few structures better discernible in MRI than in unstained cross-sections.

Pars Ventralis Pontis
Relatively few pontine structures were depicted, but all of them are discernible in both methods. Out of the pars ventralis pontis NPo occupy relatively large area. In MRI they form horizontal, hyperintense bands in ventral part of the brainstem (Figures 8C, 9A,C) and are quite well contrasted with surrounding white matter in MAC (Figures 8D, 9B,D). In Figures 8C,D, 9A

Pars Dorsalis Pontis
None of nuclei were depicted in U. arctos nor in any other species reported in available literature. However, hypointense congregation of white matter is discernible in MRI as PCM, PCR and TSNT. PCM forms a broad band of white matter within the lateral aspect of the brainstem that can be identified in MRI (Figure 9C). The transition between FPoT into PCM was recognizable in MAC ( Figure 9D).

External Anatomy
The general shape of the cerebellum in U. arctos (Figures 1A,B,D,E) is conserved in other Ursidae -U. maritimus, U. americanus (Welker et al., 2009), A. melanoleuca (Mettler and Goss, 1946), and H. malayanus (Kamiya and Pirlot, 1988b). U. arctos differs from the dog in several anatomical features. The shorter relative length of the cerebellum relative to the decussatio pyramidum (DPy) lies below posterior part of the cerebellum (Figure 12), whereas in the dog it lies more caudally to its posterior ridge and the structures are never captured together in a cross-section perpendicular to the long axis of the medulla oblongata. HC are more expanded in U. arctos than in Mustelidae, including M. nivalis, Neovison vison, T. taxus (Welker et al., 2009) but considerably lesser than in Pinnipeds, possibly stemming from their aquatic lifestyle. The fraction of the cerebellum occupied by the cerebral hemispheres varies: in U. arctos about 1/3 volume, similar to that observed in other Ursidae and some Mustelidae (the ferret), and Pantherinae subflamily (P. leo, P. pardus) (Welker et al., 2009). In many Canidae including the dog, V. zerda, V. vulpes, C. latrans and in Pinnipedia, the cerebellum is nearly invisible from the dorsal aspect. Largely exposed cerebellum is observed in the cat (Smith et al., 2001). The cerebellar hemispheres division into lobuli was conducted only on the basis of macroscopic analysis (Figures 1A,B,D,E, 9D, 10B,D, 11B,D, 12B). LACR, LACC, LPm, PfD, PfV and Fc with PFc were clearly distinct. The lobules of the vermis were depicted also in MRI as part within the ventriculus and on the surface (Figures 1A,B,D,E, 9-12). Vm evinces relatively distinct division into its lobuli -all of them were depicted but the border between Dc and FVm is unclear. The noteworthy sections are in Figure 9D where PCM is continued into the pons and in Figure 10D wherein PCC is continued in MO. The folia cerebelli are apparent in both methods. In MRI they form alternating hypo-and hyperintense strips (Figures 9C, 10A,C, 11A,C, 12A). The hyperintensity are typical for cortex and nuclei of the cerebellum. In Figures 11A,D the cortex cerebelli due to indentation from the caudal aspect is also visible within the white matter so that it resembles nuclei.

Internal Anatomy
NF was depicted only in MAC ( Figure 11B) as a round zone below the vermis. The nucleus fastigii is located medially and more prominently separated from NIC than NIC from NLC. In both methods, the NIC and the NLC were discernible. The borders between them are very subtle in MAC (Figures 10D, 11B) and sometimes indistinguishable in MRI, so that NLC and NIC form a common hyperintense mass and are collectively labeled in Figure 10C. However, in more caudal sections the distinction between NLC and NIC was more prominent (Figure 11A). The isolated NIC is also discernible in its rostral section in Figure 10A, however, the signal intensity is much lower. The relative volume and topography of the nuclei is comparable in the dog, after staining (Fletcher, 2007). All three nuclei were identified in MRI in the cat (Gray-Edwards et al., 2014), but only NLC in MRI in Z. californianus (Montie et al., 2009) and none of them, in MR,I in the dog (Leigh et al., 2008). The staining makes the structures easily discernible (Welker et al., 2009) and all the nuclei are present in Carnivora species available in CMBC.

External Anatomy
In U. arctos, MO is dorso-ventrally flattened as in the dog (Whalen, 2003). Also Mettler and Goss (1946) found it in A. melanoleuca in similar shape. In comparison, MO of Z. californianus is nearly round in its cross-section in its caudal part (Montie et al., 2009). Cat MO seems to evince an intermediate state between those two extremes (Welker et al., 2009).

Gray Matter
The most prominent nuclei were NOv highly contrasted in MAC (Figures 10D, 11B,D) assuming a signature shape. Excellent contrast is apparent in MRI as a hyperintense, comma-shaped symmetrical zone with sharp borders (Figures 10C, 11A,C). In MAC and one MRI cross-section ( Figure 11A) HNOv was discernible as a subtle white or hypointense string. HNOv are not bulged into olivae on the surface in U. arctos. HNOv is typical for a few mammalian orders including Carnivora (Welento, 2002) and was also visualized in MRI in the cat (Gray-Edwards et al., 2014) but not found in the dog (Leigh et al., 2008) or in Z. californianus (Montie et al., 2009). The shape of NOv in the cross-section varies among Carnivora species available in CMBC. In U. arctos and U. maritimus its medial part is wider and shorter whereas the lateral part is thinner and longer. In the dog, both parts are nearly equal in length but the medial one is wider. In Z. californianus the medial part is scant and the lateral part is considerably broader. NCL is another quite well defined nucleus taking up more dorsal residence above an apparent TSNT. It is also visible in MAC over relative large distance where it forms an exceptionally conspicuous zone of gray matter (Figures 10D, 11B) in Figure 11B. It is relatively less apparent in MRI, however, an initial TSNT identification facilitates the pinpointing of NCL. Similarly, NCM is conspicuous in MAC (Figures 11D, 12B) whereas in MRI its borders are unclear. The topography of NCM is consistent with that in Fletcher (2007) but due to large concentration of nuclei in the dorsal region of the medulla oblongata, we cannot rule out confusion with other nuclei, including nucleus parasympathicus nervi vagi, nucleus motorius nervi hypoglossi or nucleus tractus solitarii. On the other hand, NG is more distinguishable, in MRI, due to hyperintensity ( Figure 11C) and as it forms TNG, identification in MAC is also effortless (Figure 11D). NVb were collectively depicted only in MAC as a distinct dark periventricular zone delineated by white matter of pedunculi cerebellares and NVII ( Figure 10B). They are at the level of CT and consistent with topography described in literature (Kobryń and Kobryńczuk, 2004) and pointed by Fletcher (2007) in the dog. There is a unilateral highly hyperintense zone measuring about 4 mm in one MRI section ( Figure 9A). It was unnoticeable in MAC, and due to its unilateral placement, we suppose it may be a lesion. NVb were depicted in the cat (Gray-Edwards et al., 2014) and in the dog (Leigh et al., 2008) and none of other nuclei of MO were depicted in abovementioned papers except for NOv.

White Matter
In the initial section of the medulla oblongata CT, we found a laterally extended band of white matter across MO width ( Figure 10B). It is hypointense and devoid of ventral border in MRI, so the signal intermingles with that of TPy lying below ( Figure 10A). The border of CT is also unclear in MAC. There are nerve fibers coursing from dorsal to ventral surface possible to spot in MAC but not in MRI: of NVII (Figure 10B), NXII (Figures 11B,D) and NVIII ( Figure 10B). However, the latter was depicted in MRI in its free part on the ventro-lateral surface of MO ( Figure 10A). The decussations of main neural tracts are discernable in both methods. LMs (Figures 10C,D, 11) are apparent centrally as a modest sphenoidal zone. LMs in MRI are hypointense with unclear borders; however, DLM occupies almost the entire central area of MO ( Figure 11C). In MAC, TPy and DLM are found with well-defined boundaries (Figures 11B,D). DLM is reverse-Y-shaped, but for DPy the Y-shape is not reversed (Figure 12). The closing section of MO contains typical funiculi: FuD, FuV and FuL (Figure 12). In MRI they form a hypointense rim ( Figure 12A). It is possible to notice a division of FuD into FaG and FaC in MAC ( Figure 12B) whereas in MRI the division is not apparent. TSNT with NTSNT ( Figures 9C,D, 10-12) are consistent across the entire course and they were also depicted in MRI in the cat (Gray-Edwards et al., 2014). The white matter of MO is generally consistent in all Carnivora groups according to specimens in CMBC.

Reticular Formation
The area occupied by the reticular formation is difficult to inequivocally identify in MAC (Figures 8B,D, 9B,D,  10B,D, 11B,D). The only intimation of its presence is slightly darker coloration of the area. In MRI, the signal intensity is high enough to notice the broad character of the area (Figures 9A,C, 10A,C, 11A,C), but we could not delineate the exact borders of the area. In two MRI sections (Figures 9C, 10A) the central area is featured by apparently higher signal intensitywe suppose it comes from NRh and the nucleus rhaphe magnus as depicted by Fletcher (2007) in stained crosssections in the dog.

Ventricular System
The ventricular system in general is well discernible in both methods. In MAC it forms obvious empty spaces, in some sections very tight -rostral parts of VL (Figures 4B,D, 5B,D, 6B) or VT (Figure 6B). A broader lumen was observed in the middle of VL (Figures 7B,D), initial section of VT (Figure 6B), AqC (Figures 7D, 8B,D, 9B) and VQ (Figures 9D, 10B,D, 11B,D). In MRI the same features are observed for abovementioned sections and the entire ventricular system is completely devoid of signal. The signal decrement is typical for areas occupied by larger vessels (Figures 4C, 5A,C, 6A). In Figures 10C, 11B, respectively, a hyperintense or black zone is present -PCVQ. The lumen of VL in U. arctos is tightened in the rostral part to an arcuate cleft (Figures 4; 5, 6A,B) and the similar state is observed in the dog (Leigh et al., 2008), the cat (Gray-Edwards et al., 2014) and Z. californianus (Montie et al., 2009). In the ferret, the arcuate cleft is more asymmetrical -the dorsal arm is shorter (Sawada et al., 2013). In Figures 6A,B the foramen interventriculare was captured. The lumen of VT in Carnivora is reduced due to a large ATh (Kobryń and Kobryńczuk, 2004) and in U. arctos that is also evident (Figures 6C,D, 7A,B). Figure 7D shows the aditus ad aquaeductus cerebri with CCa lying above it. The canalis centralis was not apparent in a thorough examination of the most caudal parts of MO.

CONCLUSION
1. Most of the recognized brain structures were apparent in both MRI and macroscopic anatomy. The tractus olfactorius medialis, corpus subthalamicum, brachium colliculi rostralis, fasciculus longitudinalis medialis, nuclei vestibulares, velum medullare rostrale, pedunculus flocculi, nucleus fastigii, fasciculus gracilis et cuneatus (nine structures) were depicted only in macroscopic anatomy analysis. The glandula pinealis, lemniscus lateralis and nuclei rhaphe (three structures) were visualized only in MRI. The lack of the glandula pinealis in anatomical analysis was omitted in cross-section. 2. The brain surface of U. arctos is more convoluted than in the cat, the dog and Mustelidae but lesser than in Pinnipedia. The homologies of the middle and rostral parts are less controversial than in caudal part. 3. A characteristic trait for Ursidae and Pinnipedia is presence of three gyri frontales and a lozenge in rostral part of the hemispheres. There is a difference in interpretation of the course of the sulcus cruciatus in U. arctos caudal or rostral to the gyrus frontalis superior. This is an important issue because higher development of rostral area to the sulcus cruciatus is assumed to take part in social behavior in C. crocuta and P. leo. 4. The rhinencephalon in U. arctos is organized in a manner typical for macrosmatic mammals. The pedunculi olfactori are stout, the tuberculum olfactorium is round and three tractus olfactorius are present including the tractus olfactorius intermedius. The lobus piriformis is relatively shorter and wider than in most Carnivora. 5. The claustrum in U. arctos brain cross-sections assumes a crescent-shape in contrast to the dog and cat where it assumes more pickaxe-shape. In MRI the claustrum usually resembles cerebral cortex thickening. The nucleus accumbens is more apparent in MRI. The boundaries between the globus pallidus and putamen are unclear in MRI and their signals usually mingle. The nucleus endopeduncularis is present and clearly evident in both methods. 6. The septum pellucidum in U. arctos is absent. Interestingly, it is present in most Carnivora including U. maritimus. 7. The hippocampal formation in U. arctos is divided into typical dorsal and ventral parts. The gray matter is largely reduced in dorsal part, however to a lesser extent than in Pinnipedia. 8. The epiphysis in U. arctos is relatively larger than in the dog but smaller than in Pinnipedia. It was visualized only in MRI. The hypophysis was not preserved. The corpora mammillaria are paired. 9. The layers of the colliculus rostralis are more evident in MRI in our study. The substantia grisea centralis is more contrasted in unstained cross-sections than in MRI where its signal intensity is weak. 10. Neither of methods used allowed to identify the nuclei of the pars dorsalis pontis in U. arctos. Out of the pars ventralis pontis the nuclei pontis were apparent. 11. The hemisphaeria cerebelli in U. arctos are prominent.
The lobuli of the vermis and most lobuli of the hemispheres were depicted. 12. The boundary between the nucleus interpositus cerebelli and the nucleus lateralis in U. arctos is unclear. The nucleus fastigii was identified only in macroscopic anatomy method. 13. The medulla oblongata in U. arcots is largely dorsoventrally flattened unlike the round in cross-section in Pinnipedia and slightly ovate in other Carnivora. Most of nuclei were impossible to discern in either method. The shape of nucleus olivaris in cross-section shows interspecies differences among Carnivora. The canalis centralis were undiscernible.

ACKNOWLEDGMENTS
The authors wish to thank the managers and Croatian Bear Emergency Team members who facilitated and assisted in sample collection. Professor Józef Nicpoń is thanked for the support and making it possible to conduct the MRI study at the Center of Experimental Diagnostics and Innovative Biomedical Technology of the Faculty of Veterinary Medicine at Wrocław University of Environmental and Life Sciences in Wrocław.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnana. 2019.00079/full#supplementary-material FIGURE S1 | Phylogenetic relationship among Carnivora families with position of Ursidae marked by * (after Flynn et al., 2005).