Introduction

The cerebral cortex exhibits some features which are common to all mammals. Other attributes are peculiar to certain groups and may represent functional and/or evolutionary specializations; examples are the vibrissae barrel fields of certain rodents (Johnson, 1980), and the unique cortical lamination patterns of cetaceans (Glezer et al., 1988). However, our understanding of the diversity of mammalian cortical organization is deficient because certain orders, such as the Sirenia and Pinnipedia, have been studied far less than others. Investigation of sirenian brains is of particular interest because Sirenia are alone among mammals in being obligate aquatic herbivores.

Even the earliest reports indicated that sirenian brains exhibit a unique combination of traits. Elliott Smith (1902) noted that "Amongst the whole series of placental mammals there is no other animal in which the brain presents features so extraordinary and so bizarre as in the Sirenia". Included were: pronounced lissencephaly; large lateral ventricles; "aborted" rhinal fissure; small olfactory bulbs; small corpus callosum but with a complete genu; large pons and 5th, 7th and 8th cranial nerves; enormous flocculo-parafloccular lobes of the cerebellum. Dexler (1913) discovered in dugongs that some areas of cerebral cortex were well laminated, and that cell clusters ("Rindenkerne") were present in layer Vl of certain regions on the lateral surface. In more recent studies, our group has found that visual thalamic and brainstem nuclei are reduced in size in manatee brains, whereas auditory, trigeminal, somatosensory, and vagal nuclei are well developed; the facial motor nucleus is large and lobulated (Johnson et al., 1986; 1987; 1988; Welker et al., 1986). These findings concur with assessment of the relative importance of these systems in sirenian behavior. In particular, the trigeminal and facial nerve systems are associated with extensive use of the proboscis and mouth parts in the feeding behavior which dominates sircnian life.

All Sirenia, including extinct forms, have unusually small relative brain size (brain/body weight ratio), and this appears to be related to selection for large body size (O'Shea and Reep, 1990). This perspective is reinforced by three other findings: 1) the gyration index (a measure of cortical folding) is 1.06, representing a highly lissencephalic condition (Reep and O'Shea, 1990); 2) the telencephalon comprises 71% of total brain volume and is 90% cerebral cortex (Reep and O'Shea, 1990), values comparable to those in taxa having large relative brain size, including many primates; 3) manatee brains exhibit well-defined cortical lamination (Reep et al., 1989). Thus, small relative brain size and lissencephaly do not constrain the elaboration of internal structures.

Recently, we identified thirteen distinct areas of manatee frontal cortex (Reep et al., 1989). Two of these areas, representing the presumptive face region of somatic sensory cortex, contain neuron cell clusters (Rindenkerne) in the deeply situated layer VI. In area CL1 the clusters are about 1.0 mm in diameter and are spaced about 1 mm apart. In area CL2, which lies dorsally adjacent to area CL1, the clusters are 0.2-0.4 mm in diameter and are spaced 0.5-1.0 mm apart. Our preliminary observations indicate that Rindenkerne of varying shapes and sizes are also present in certain areas located more caudally. For example, in the postsylvian cortex immediately Caudal to area CL1, the clusters are sharply defined by a cell-sparse fibrous halo, are about 0.5 mm in diameter, are spaced about 1.5 mm apart, and are situated in the deepest portion of layer VI. Small clusters like those of area CL2 are seen in other caudal fields. It remains to be determined how many different kinds of Rindenkerne exist in the manatee brain, and that is one of the goals of the present studies.

Rindenkerne are unique to Sirenia since they are not present in the brains of approximately 100 other mammalian species examined (Johnson et al., l991a,b).). However, Rindenkerne are reminiscent of the 'barrels' in the vibrissae subfield of somatic sensory cortex in certain rodents and other mammals (Johnson, 1980), and they share histochemical attributes (Reep et al., 1989). But rodent barrels are hollow aggregates of neurons in layer IV, a major afferent zone, whereas manatee Rindenkerne are dense aggregates found in layer VI, an efferent zone. The restricted distribution of Rindenkeme is strong evidence of functional significance, especially considering that Sirenia are unique in being obligate aquatic herbivores, and that the presence of Rindenkerne is a unique sirenian trait.

In those species having barrels, there is a one-to-one correspondence between vibrissae and barrels. By analogy, Rindenkerne in manatees may be related to the tactile bristles of the upper lips and lower buccal pad. These perioral facial structures are used extensively in labial prehension and palpation, often in conjunction with the forelimbs, for manipulation and ingestion of aquatic vegetation. Bristle counts in eight adult specimens averaged 121 per side, with 81 on the upper lip and perioral region, 40 on the lower buccal pad and perioral region. We have hypothesized that if Rindenkerne are related to bristle function in manatees, either the Rindenkerne arc involved in motor control of bristles, or layer VI is involved in the processing of thalamic afferent information to a degree greater than normally seen among mammals (Reep et al., 1989).

The presence of two size classes of Rindenkerne in the frontal regions suggested that the larger clusters (and hence area CL1) may relate to the long thick tactile bristles of the upper lips while the smaller clusters (and thus area CL2) may be associated with function of the more numerous and shorter bristles of the upper lips and lower buccal pad. This hypothesis is related to the finding in mice that the largest vibrissae are represented by the largest cortical barrels (Woolsey and Van der Loos, 1970). As mentioned above, Rindenkerne in manatees extend outside the limits of presumptive face cortex Therefore it is possible that Rindenkerne have no specific association with tactile bristle function, but rather represent some other functional attribute of particular cortical regions in Sirenia. Alternatively, our presumptions about the size of the face representation may be too restrictive.

In order to address these issues, we have begun making correlative counts of the various types of facial bristles and Rindenkerne. The purpose of the present study was to utilize computer based serial section reconstruction techniques in order to obtain accurate counts and morphometric measurements of Rindenkerne, and to produce three dimensional renderings illustrating the positions of Rindenkerne within cerebral cortex.

Methods

Brain 86-124 was utilized for this study. One hemisphere was embedded in celloidin, sectioned coronally at 30 m, and every section stained with thionin for Nissl substance. The most rostral 271 sections from a total of 1800 (15%) have now been analyzed using three dimensional reconstruction techniques. Each of the sections of interest is placed on a light box and digitized at low magnification so that the entire section is visible in one field of view. Each successive section must be placed in register with its predecessor, and this is done by the user entering fiducial marks referenced to anatomical landmarks such as the midline and ventricular spaces. The following features are detected automatically by gray level discrimination, or outlined using a mouse-controlled cursor: the cortical surface, gray matter-white matter boundary, ventricular spaces, and similar large anatomical landmarks; the upper boundary of layer VI, and the boundaries of relevant cytoarchitectonic fields such as areas CL1 and CL2. (Note: a feature is a structure which is visible on a section but is usually part of a larger object. Objects are eventually made whole again by recombining features from several sections into a 3D rendering or database) Next, each section is placed on the microscope and viewed at higher magnification so that smaller features of interest (i.e. Rindenkerne) are readily identifiable. Data on section number and thickness, and a magnification calibration factor are entered. At this higher magnification the entire section will not be visible in a single field of view, thus it must be scanned in order to identify features of interest so that their coordinate representations can be stored. A motorized scanning stage allows for continuous automatic alignment between the boundaries determined at low magnification and now displayed at the higher magnification, and the features of interest (e.g. Rindenkerne) being identified at the higher magnification. The coordinates of each feature are extracted and entered into the database for that section. Each of the classes of identified features (e.g. Rindenkerne, upper boundary of layer Vl) is user defined by separate codes used in the later 3D reconstruction.

Once a database is obtained consisting of features extracted from multiple sections, it is considered to be a 3D representation (in the case of data from serial sections, this is literal; in the case of data from spaced sections, interpolation algorithms are used to supply data from "missing sections"). Features of the same class which are continuous across sections are treated as parts of the same object. For example, all the outer cortical boundaries will have been given the same code and thus can be connected smoothly so as to appear realistically as a continuous surface. Reconstruction of Rindenkerne is interactive, since the operator specifies continuity of Rindenkerne features across sections. For purposes of visual display, observation, and discussion, the database of reconstructed objects (or subsets of specified objects) may be rendered in smoothed 3D form by combining the data from all sections or groups of sections and utilizing tiling algorithms. The operator specifies attributes such as the color of each object class, transparency and lighting angle. In this way one can produce a see-through 3D model which can then be rotated and viewed from any desired orientation on the Sun computer monitor.

Results

Several patterns of organization have been seen among the 208 Rindenkerne present in the rostral portion of cerebral cortex examined in the present study. As one proceeds caudally from the frontal pole, Rindenkerne first become visible in the cortex of the lateral surface (Figure 1). These clusters are relatively small in size (0.022-0.055 mm² in area) and are located superficially in layer VI. More caudally (Figure 2) the clusters become larger (0.0960.198 mm²) and extend throughout the depth of layer VI. There is a continuous increase in the areal size of Rindenkerne as one proceeds more caudally. Except for extremely rostral clusters, spacing between adjacent Rindenkerne in the coronal plane is remarkably consistent regardless of rostrocaudal location, averaging 931±40 m. The most rostral clusters are more sparse, spaced at intervals averaging 1268 m. Other attributes of the Rindenkerne examined are presented in Table 1, in which clusters have been grouped according to rostrocaudal length. Length appears to be a significant descriptor, since cross sectional diameter (and thus, area) increases in conjunction with it. This correlation could simply mean that the clusters are spherical regardless of size; however, the length/width ratio also increases with length, implying that as Rindenkerne become larger, length increases at a faster rate than width. In other words, larger Rindenkerne are ellipsoidal in shape, with the long axis orientated in the rostrocaudal direction. The smaller clusters have length/width ratios <1.0, indicating that they are also ellipsoidal in shape, but here the long axis is oriented in the mediolateral dimension. The larger two categories of Rindenkerne in Table 1 are relatively rare, and those of the ">500" class have almost twice the volume of those of the "400-500" class.

Click on the image to see a larger view

	Figure 1. Schematic representation of a coronal section of a manatee cerebral hemisphere. This section is located relatively rostrally where the clusters are still small and located superficially in layer VI of the cerebral cortex. Scale units are in mm.
	Figure 2. This section is located further caudally than that of Figure 1. Here the clusters have become larger and extend deeper into layer VI. Scale units are in mm.

Table 1. Attributes of Rindenkerne Clusters Organized According to Length

Discussion

The present quantitative and three dimensional results add to our previous qualitative two dimensional description of Rindenkerne. Even based on the partial data presented here, it is apparent that there are many more clusters than facial bristles. However, it is possible that only the largest Rindenkerne are involved in processing sensorimotor information related to bristle function, and Rindenkerne of this size class are relatively rare. Thus, it remains to be seen whether total counts of these larger Rindenkerne are consistent with counts of bristles. Even if this does turn out to be the case, it still leaves unanswered the question of the possible function(s) of the remaining Rindenkerne.

Based on our present data, it is interesting to consider the possible developmental history of Rindenkerne. Rindenkerne are found only in the cortex of the lateral surface; this implies that their presence does not represent some developmental process unique to sirenian cortex generally. The mammalian cerebral cortex develops from a proliferative population near the deep-lying ventricular margin (Rakic, 1981). Post-mitotic neuroblasts migrate along radial glial fibers to reach their appropriate laminar location, with neurons in layer VI being among the earliest to do so. Hence, aggregates such as Rindenkerne appear only after the laminar destination has been reached. A major question is whether neurons destined to belong to a given Rindenkerne are related as a clonal unit before aggregation. Most of the Rindenkerne of the present study are ellipsoidal rather than spherical in shape. This implies that Rindenkerne probably do not originate from "hot spots" of proliferative activity at specific loci on the two dimensional ventricular sheet, since in its simplest form this would produce spherical aggregates. On the other hand, the regularity of spacing among adjacent Rindenkerne regardless of rostrocaudal location does suggest some inherent spatial ordering. Another possibility is that the neurons of a given Rindenkerne are not clonally related, and that aggregation occurs as a result of the expression of cell adhesion molecules or extracellular matrix products.

Acknowledgements

The authors appreciate the continuing support of Mr. Dan Williams, Systems Programmer and manager of the College of Veterinary Medicine's Computerized Image Analysis Laboratory in which this work was done.

References

1.  Dexler, H. (1913) Das Him von Halicore dugong Erxl. Gegenbaurs Morphologisches Jahrbuch 45:97-90.

2.  Elliott Smith, G. (1902) Descriptive and illustrated catalogue of the physiological series of comparative anatomy contained in the museum of the royal college of surgeons of England, vol 2. London, Taylor and Francis.

3.  Glezer, I.I., M.S. Jacobs, P.J. Morgane (1998) Implications of the "initial brain" concept for brain evolution in Cetacea. Behav. Brain Sci. 11:75-116.

4.  Johnson, J.I. (1980) Morphological correlates of specialized elaborations in somatic sensory cerebral neocortex; in Ebbesson SOE (ed): Comparative Neurology of the Telencephalon, New York, Plenum, pp 423-447.

5.  Johnson, J.I., Welker, W.I., Reep, R.L. and Switzer, R.C. (1986) Dorsal column nuclei of the aquatic herbivorous manatee (Trichechus manatus). Soc. Neurosci. Abstr. 12:110.

6.  Johnson, J.I., Welker, W.I. and Reep, R.L. (1987) 'Fine motor nuclei of the cranial nerves in manatees, Trichechus manatus. Anat. Rec. 218:68A.

7.  Johnson, J.I., Kirsch, J.A.W., Reep, R.L., Switzer, R.C. and Welker, W.I. (1988) Well-developed brainstem auditory nuclei in manatees Trichechus manatus. Soc. Neurosci. Abstr. 14:491.

8.  Johnson, J.I., Kirsch, J.A.W., Reep, R.L., Switzer, R.C. and Welker, W.I. (1991a) Phylogeny through brain traits revisited: eight more characters for analysis of mammalian relationships. Submitted to Brain, Behav. Evol.

9.  Johnson, J.I., Kirsch, J.A.W., Reep, R.L., Switzcr, R.C. and Welder, W.I. (199lb) Phylogeny through brain traits revisited: distribution of states of new characters in extant mammals. Submitted to Brain, Behav. Evol.

10. O'Shea, T.J. and Reep, R.L. (1990) Encephalization quotients and life-history traits in the Sirenia. Journal of Mammalogy 71:534-543.

11. Rakic, P. (1981) Developmental events leading to laminar and areal organization of the neocortex. In: The Organization of the Cerebral Cortex (Ed. F.O. Schmitt), MIT Press, Cambridge, Mass.

12. Reep, R.L. and O'Shea, T.J. (1990) Regional brain morphometry and lissencephaly in the Sirenia. Brain, Behav. Evol. 35:185-194.

13. Reep, R.L., J.I. Johnson, R.C. Switzer, W.I. Welker (1989) Manatee cerebral cortex: cytoarchitecture of the frontal region in Trichechus manatus latirostris. Brain Behav. Evol. 34:365-386.

14. Welker, W.I., Johnson, J.I. and Reep, R.L. (1986) Morphology and cytoarchitecture of the brain of Florida manatees (Trichechus manatus). Soc. Neurosci. Abstr. 12:110.