Sink or swim? Bone density as a mechanism for buoyancy control in early cetaceans

Authors


Abstract

Previous analyses have shown that secondarily aquatic tetrapods, including whales, exhibit osteological adaptations to life in water as part of their complex buoyancy control systems. These structural specializations of bone span hyperostosis through osteoporosis. The past 15 years of paleontological effort has provided an unprecedented opportunity to examine the osteological transformation of whales as they make their transition to an obligate aquatic lifestyle over a 10-million-year period. It is hypothesized that whales manifest their osteological specialization in the same manner as extant semiaquatic and fully aquatic mammals. This study presents and analysis of the microstructural features of bone in early and late archaic cetaceans, and in a comparative sample of modern terrestrial, semiaquatic, and aquatic mammals. Bone histology was examined from the ribs of 10 fossilized individuals representing five early cetacean families, including Pakicetidae, Ambulocetidae, Protocetidae, Remintonocetidae, and Basilosauridae. Comparisons were then made with rib histology from nine genera of extant mammals including: Odocoileus (deer), Bos (cow), Equus (horse), Canis (dog), Lutra (river otter), Enhydra (sea otter), Choeropsis (pygmy hippo), Trichechus (sea cow), and Delphinus (dolphin). Results show that the transition from terrestrial, to semiaquatic, to obligate aquatic locomotion in archaeocetes involved a radical shift in bone function achieved by means of profound changes at the microstructural level. A surprising finding was that microstructural change predates gross anatomical shift in archaeocetes associated with swimming. Histological analysis shows that high bone density is an aquatic specialization that provides static buoyancy control (ballast) for animals living in shallow water, while low bone density is associated with dynamic buoyancy control for animals living in deep water. Thus, there was a shift from the typical terrestrial form, to osteopetrosis and pachyosteosclerosis, and then to osteoporosis in the first quarter of cetacean evolutionary history. Anat Rec, 290:638–653, 2007. © 2007 Wiley-Liss, Inc.

Mammals and reptiles that secondarily invade aquatic niches typically undergo remarkable structural transformations related to the mechanical constraints of locomotion in water versus land. Although the most obvious skeletal changes occur at the gross anatomic level, equally dramatic modifications occur in the structural properties of bone tissue, ranging across extremely high to extremely low density (for a review of literature, see Ricqles and Buffrenil, 2001). These adaptations are recognized as variable strategies for managing buoyancy (Taylor, 2000).

There are several well documented transitions of extinct reptiles to aquatic dependence; the past 15 years have provided us with unprecedented paleontological documentation of the early evolutionary transformation of the mammalian Order Cetacea, which includes modern whales, dolphins, and porpoises (see review in Uhen, 2007, this issue). Whereas recent fossil discoveries have led to many reports documenting cetacean phylogeny, functional morphology, sensory adaptations, and physiology, there has been no systematic examination of the structural properties of fossil cetacean bone.

This finding is of particular interest, since the earliest examination of fossil cetacean bone, that of 40 million-year-old, fully aquatic taxa Basilosaurus and Zygorhiza, indicated that both possess pachyosteosclerotic ribs, similar to modern members of Sirenia (Buffrenil et al., 1990). However, with few exceptions, modern cetaceans universally exhibit osteoporotic bone in all skeletal elements but the vertebrae and skull (Ricqles and Buffrenil, 2001).

The questions of interest here are ecomorphological. It is now possible to examine bone from the 10 million year period preceding Basilosaurus and Zygorhiza, which includes four additional archaic cetacean families: (from oldest to youngest) the Pakicetidae, Ambulocetidae, Remingtonocetidae, and the Protocetidae. The Pakicetidae are found in Early Eocene stream deposits of Northern Pakistan. While their skeletons were originally described as adapted for running (Thewissen et al., 2001), subsequent detailed analysis of the postcrania noted systemic hyperostosis, a condition incompatible with safe, energetically efficient cursoriality (Madar, 2007). Ambulocetus, a representative of the second family that occurred a million years later in the Middle Eocene of Northern Pakistan, lived in transitional marine environments, and showed more overt specialization toward aquatic locomotion, with lengthened distal limb segments, and axial modifications related to dorsoventrally propulsed aquatic locomotion (Thewissen et al., 1994, 1996). Osteosclerosis was noted in its long bones (Thewissen et al., 1996). Members of the Remingtonocetidae and the Protocetidae, from the middle Eocene of Indo-Pakistan and North America (protocetids only) bridge the chronological and morphological gap between the earliest freshwater and marine cetaceans. The Basilosauridae are middle to late Eocene archaeocetes, which contain the sister group to all modern cetaceans. Basilosauridae contain two subfamilies: Basilosaurus of the basilosaurinae (Buffrénil et al., 1990), and Zygorhiza, a member of the dorudontinae that exhibits essentially modern cetacean body form—more flipper like forelimbs, tremendously reduced hind limbs, and advanced feature of the vertebral column associated with oscillatory swimming and the presence of a tail fluke (Uhen, 1998).

The aim of this study is to document the microstructural features of bone over the first 10 million years of cetacean evolution in the fossil groups outlined above. Data from this study will help determine whether the mechanism for developing a static buoyancy control system apparent in early and late archaic cetaceans is the same, and whether other semiaquatic mammals that have documented high bone density achieve hyperostosis in a similar way. The ultimate goals are twofold: to gain clearer understanding of how transitional species made use of aquatic and terrestrial habitats, and to elucidate the physiological mechanisms involved in developing a static buoyancy control system.

Histological approaches can help address all of these questions. Previous research links certain bone tissue types to relative rates of growth. Plexiform or laminar bone formation as well as high and low bone densities have been associated with tetrapod transitions back to aquatic niches (Buffrenil et al., 1990).

MATERIALS AND METHODS

Paleontological Sample

The earliest published comparative work on archaeocete bone histology used rib sections, on account of the well-documented thoracic ballast of modern Sirenia. In addition to being used in the only previous analysis of archaeocete bone, rib specimens are also readily available for destructive sampling, due both to their numbers represented in the fossil record, and their less central role in studies of locomotion.

Ichthyolestes pinfoldi (Pakicetidae) is the smallest pakicetid archaeocete. Adult size approximately 1.4 m in length, known from the Ypresian (early Eocene) freshwater Lower Kuldana Formation of Punjab Province, Pakistan (Williams, 1998). H-GSP 96198 is a fragment of a mid-thoracic rib, distal half of a diaphysis, recovered in isolation from H-GSP Locality 62.

Pakicetus attocki (Pakicetidae) is the largest pakicetid archaeocete. Adult size approximately 2.5 m in length, known from the Ypresian (early Eocene) freshwater Lower Kuldana Formation of Punjab Province, Pakistan (Williams, 1998). H-GSP (Howard University-Geological Survey of Pakistan) 96355 is a mid-thoracic rib diaphysis, while H-GSP 96602 is the distal half of a more distal rib. Both specimens were isolated fragments recovered from the H-GSP Locality 62 and are not likely from the same individual.

Ambulocetus natans (Ambulocetidae) is approximately 5 m in length. A single, partially articulated skeleton H-GSP 18507 was recovered from the Lutetian (middle Eocene) marine Upper Kuldana Formation of Punjab Province, Pakistan (Williams, 1998). Four fragmentary rib samples were collected from H-GSP 18507 that preserved at least some portion of the cortical bone.

Kutchicetus minimus (Remingtonocetidae) is approximately 3 m in length. A single, partially articulated skeleton, IITR/SB (Indian Institute of Technology, Roorkee, curated by Sunil Bajpai) 2647 was recovered from the shallow marine Harudi Formation of District Kachchh, State of Gujarat, India (Bajpai and Thewissen, 1998). Four fragmentary rib specimens were collected from the type specimen, from Locality Godhatad.

Remingtonocetus sp. (Remingtonocetidae) IITR/SB 2653 is a fragmentary rib from a partial skeleton of a single individual. It was recovered from the middle Eocene Lutetian, Harudi Formation of Kachchh District, western India.

Georgiacetus vogtlensis (Protocetidae) is approximately 3–4 meters in length. Two rib fragments of unknown thoracic level from referred specimen GSM (Georgia Southern Museum) 350 (Hulbert, 1998) were sectioned. The specimens were recovered from the late middle Eocene Bartonian, “Blue Bluff Unit” of Burke County, Georgia (Hulbert, 1998).

Gaviacetus sahnii (Protocetidea) is greater than 5 m in length. Pieces of an isolated rib from a skeleton with associated skull, IITR/SB 2870, were collected from Dhedidi North, a locality within the middle Eocene Harudi Formation of District Kachchh, State of Gujarat, India (Bajpai and Thewissen, 1998).

Zygorhiza kochii (Basilosauridae: Dorudontinae) is approximately 6 m in length. A partial rib of unknown thoracic level, ALMNH PV (Alabama Museum of Natural History) 2000.1.1.1., was recovered from the Bartonian Pachuta Marl Member of the Yazoo Clay Formation near Melvin, Alabama.

Extant Sample

A broad array of extant, large-bodied, terrestrial and semiaquatic mammals were studied to determine regional variation in cortical and trabecular morphology in ribs obtained from different thoracic levels, to determine baseline variation along the lengths of individual ribs, and to compare immature vs. adult bone. The collection used for this study included adult and juvenile Odocoileus virginianus (deer), Bos taurus (cow), Equus caballus (domestic horse), and Canis domesticus (dog) obtained from the teaching collections of Hiram College. Lutra canadensis (river otter) was obtained from a trapper under ODNR (Ohio Department of Natural Resources) permit 211 to Hiram College; Enhydra lutris (sea otter) was obtained from California Department of Fish and Game; and Choeropsis liberiensis NMNH (National Museum of Natural History) 538815 (pygmy hippo), and Trichechus manatus NMNH 554180 (sea cow) were obtained from the Division of Mammals, Smithsonian Institution. Delphinus capensis, specimen JPD0444 (long-beaked common dolphin) was obtained from the Southwest Fisheries Science Center.

As noted above, ontogeny has clear implications for histological analysis of bone, which prompted examination of subadult bone in this extant sample. It should be noted that a juvenile condition of bone in the fossil sample does not suggest that the archaeocete was a juvenile, but rather that the morphology is commensurate with immature histomorphology in the modern specimens, where bone development of the entire skeleton can be evaluated. The developmental status of fossil taxa is often unclear, but will be stated when known.

Thin Sectioning Methods

Fossil and extant ribs were cut into 3-cm sections along the length of the rib and labeled accordingly from proximal to distal ends. Buehler low-viscosity epoxy was mixed according to recommended proportions of resin and hardener. All samples were placed into rubber mounting cups and embedded into 1–2 cm of the epoxy. Samples were then vacuumed for at least 5 minutes to remove any trapped air in the epoxy. Samples were then left to harden for 24 hours.

Embedded samples, or blanks, were first polished using a 240 grit disc, then with a 400 grit disc in perpendicular orientation with previous grit lines, followed by 600 and 800 grit, and allowed to dry. Blanks were attached to frosted slides using a thin layer of fresh epoxy. Using a belt grinder with a diamond saw, blanks were then cut from the slide. Slides were then ground down to ∼100 microns using a precision grinder. Samples were polished using the following grit progression: 400, 600, 800, to obtain sections of approximately 75 microns.

Histological Methods and Data Analysis

Using the Olympus BX-40 microscope with digital camera, full section photographs of all extinct and extant samples were taken in Olympus Microsuite version 3.2. Images were loaded onto a computer with ImageJ and Scion Image programs for data collection of cortical thickness and percentage medullary area. With ImageJ, a microscope calibration chart or in-frame scale was used to determine the number of pixels per millimeter. The straight-line tool of ImageJ was used to measure the cortical minimum and maximum thicknesses of the samples based on the tangents of the periosteal and endosteal surfaces. Scion Image was used to threshold the medullary cavity area and then total bone area of each section to obtain a percent medullary area. Patterns of osteogenesis were recorded for each section.

The image analysis software BIOQUANT (Bioquant Image Analysis Corporation © 2005) was used to analyze trabecular strut thicknesses and densities. Each histological section was viewed from the central most region (half of the length across the X and Y dimensions) of each trabecular cavity, or the central region where trabeculae were present in a portion of the medullary cavity, at a magnification of 20× (additional magnification from the microscope's digital camera was accounted for with the calibration of the BIOQUANT software). The measurements taken were total volume (TV), bone volume (BV), trabecular thickness (TTk), and trabecular thinness (TTn) using thresholding and software tools as follows.

TV is the total volume of the trabecular cavity when viewed in BIOQUANT. Measurements were taken with respect to the entire marrow cavity in the microscope field as a video area array. Any samples whose field was completely trabecular cavity was entirely thresholded; similarly, any cortical bone not at the endosteal edge of the medullary cavity in smaller slide sections was excluded from the thresholding region. The preview feature was used to create a linear outline of the selected regions before measurements were taken.

BV measurements indicate the volume of bone within the trabecular cavity. These were taken by thresholding the trabeculae to the medullary perimeter from the TV section as a video area array. Any bone that was broken in the field, but was obviously a continuous trabecular strut (postfossilization, for example, not after being broken due to fossil prepping) was thresholded along the straightest line segment possible or following the arcing pattern of the individual strut.

TTk and TTn measurements were used to determine the minimum and maximum strut diameters in the trabecular cavity. These were taken with the normal histogram tool as individual distance arrays. For each respective extreme measurement, the thickest or thinnest trabecula in the field was identified and thresholded. The exceptions to this method occurred when a trabecular nexus was in the field of view. A trabecular nexus is defined as a region of the bone where four or more trabecular struts converge together on a location to connect (like a node), but whose interconnecting segments are noticeably larger than the struts themselves. Only when the largest sized trabeculae were out of the microscope field were the interconnecting trabecular segments considered for measurement. The requirement for measurement is two well-defined origins (no third strut near either of the participating strut's ends) and substantial length for the normal histogram tool. Approximately 10, but no fewer than 6, normal lines were then measured for each extreme trabecular category and averaged.

Terminology

It is important to include a list of the terminology used in this manuscript (Table 1), as there is little common usage in the recent literature on archaeocete bone, particularly when descriptions are made off of fossils without histological analysis. The terminology used in this study is primarily adapted from work of Fracillon-Vieillot et al. (1990) and is useful as it reflects similar terminology used in the clinical literature related to bone pathology.

Table 1. Basic terminology that will be used to describe the histological condition, vasculature, and tissue type of fossilized rib cross-sections adapted from Fracillon-Vieillot et al. (1990)
ConditionDescription
OsteopeniaThinning of cortical compact bone, reduction in number of trabeculae and general bone lose.
OsteoporosisSevere osteopenia, produced by intensive osteoclastic activity.
PachyostosisEspecially thick or massive bones. Histology: outward expansion of cortical bone and more massive cortical bone as compared to trabeculae and marrow cavity. Anatomic: general widening and general thickness of usually long bones.
HyperostosisAbnormally high amounts of primary bone deposition can be cortical or trabecular.
OsteopetrosisBrittle and hypermineralized bone. Late or no marrow cavity development and sometimes marked with the presence of remnants of calcified cartilage. May be caused by high osteoblast activity.
OsteosclerosisBone resulting in large amounts of calcified cartilage. Sometimes found with osteopetrosis. Can also be found with pachyostosis in the endoskeleton of nonrelated tetrapods as a secondary adaptation to an aquatic life style.
VasculatureDescription
Simple primary Vascular canalThin canal without surrounding bone lamellae. Usually enclosed in a mineral matrix.
Primary Haversian cavitiesLarge hole cavities usually between 50 and 250 microns may or may not be interconnected.
Primary HaversianVascular canal surrounded by concentric bone lamellae. Require no re-absorption of previous tissue to form.
Secondary HaversianSecondarily filled by concentric bone lamellae, with a surrounding cement line from re-absorption of primary tissue.
NonvascularBone lacking or having low amounts of vasculature bone tissue.
Vascular patternsNotably primary osteons with orientation in only one direction.
 a) LongitudinalVascular canals running in a longitudinal direction can be in circular bundles forming a circular pattern, radial bundles oriented radiating from the center, or in oblique bundles seen as a slanted structure in the cross-sectional view.
 b) CircularVascular canals in a three-dimensional circular pattern.
 c) RadialVascular canals in three-dimensional radial pattern.
 d) ObliqueVascular canals in three-dimensional oblique pattern.
Secondary vascular patternsThese all are primary but have more than one direction to the vascular canals.
 a) LaminarLongitudinal and circular vascular canals.
 b) PlexiformLaminar vascularization but with the addition of radial canals. Commonly found in fast growing massive land animals.
 c) ReticularRegularly but obliquely oriented vascular canals.
Haversian boneThese contain clear secondary osteons with clear cement lines.
Sliding Haversian systemsThese are usually secondary Haversian systems that contain more circumferential layers on one side of the vascular canal.
Tissue typeDescription
Woven boneFibrous or woven bone matrix forms the bulk of tissue. Generally found in fast growing individuals and usually contains primary tissue. When associated with primary osteons called fibrolamellar bone, which consists of remnants of the woven bone type.
Parallel-fiberedMutually parallel collagenous fibers. Associated with simple primary vasculature and with primary osteons.
Lamellar bonePlywood like pattern. Usually associated with inner and outer bone layers and usually laid in a circular pattern with growth lines present.
Laminar boneVascularization presents in circumferential layers of primary osteons. It is laid down in Haversian spaces primitively left open with initial woven bone. Remnants of both Haversian spaces and woven bone are present. Usually associated with a condition of pachyostosis in the phase of active growth of large vertebrates.
Endosteal lamellaeLamellar layers surround trabecular spaces. Bone matrix type usually associated with osteosclerosis and increasing bone density by increasing diameter of trabecular struts. Sometimes present with Howship's lacunae surrounding the endosteal layer.
Howship's lacunaeThe spaces within trabecular bone that would have contained osteoclasts responsible for remodeling the trabeculae by absorption of the bone matrix.

RESULTS

Extinct Taxa (Archaeocetes)

Pakicetidae: Ichthyolestes.

A section of H-GSP 96198, a mid-thoracic rib, was taken at the mid-section. It has a dense, thick cortical layer (Fig. 1A). The marrow cavity is also filled with dense trabecular struts. There is a distinct lamellar layer on the periosteal surface with 12 LAG lines (Fig. 1B). The cortical bone is divided in two nearly equal subsections: one consisting of laminar bone, and the other section consisting of nonvascular woven bone. The laminar bone is marked with primary Haversian systems and oblique vascular canals. The lacunae, all consistent in size and shape within the laminar bone, are not in any pattern, except when seen in primary Haversian systems. The woven bone has a few scattered primary Haversian systems as the only form of vasculature. However, the central canals in the Haversian systems are extremely small and some are oblong, in the process of formation (Fig. 1D). The rib bears an unusually small marrow cavity that lacks trabecular infill. The endosteal margin has several layers of lamellar bone.

Figure 1.

Histology of Pakicetidae ribs. A: Whole section of Ichthyolestes pinfoldi H-GSP 96198. B: View of periosteal cortex of Ichthyolestes pinfoldi H-GSP 96198. LAG, LAG lines in periosteal lamellar bone; PH primary Haversian system; SH secondary Haversian system; SSH sliding Haversian system. C:Pakicetus attocki H-GSP 96602 View of plexiform bone bordering the endosteal surface. RC, radial canal; RRC, remodeled radial canals. D:Ichthyolestes pinifoldi H-GSP 96198 view of woven bone bordering the endosteal surface.

Pakicetidae: Pakicetus.

Pakicetus histology is shown in Figure 1C. A section of a mid-thoracic rib approximately mid-shaft was taken from H-GSP 96355. The bone is marked with a grossly thick cortical layer showing little or no organization with densely packed trabecular struts. This bone is classified as osteopetrosis, with high amounts of primary bone deposits. Pakicetus ribs have cortical thickness and BV that fall outside of extant taxa. The periosteal surface has a few layers of lamellar bone, but with no distinct LAG lines separating them. The cortical bone is primarily woven bone with a few primary Haversian systems present. Approximately one fourth of the cortical bone is laminar bone interrupting the woven bone pattern; it is in the process of remodeling as shown by the sudden interruption of the laminar bone. This lamellar bone consists mostly of oblique vascular canals; some longitudinal canals are surrounded by evenly spaced and sized lacunae. This finding also suggests remodeling. When the lacunae are not evenly spaced, they are in sliding Haversian systems, with more layers present on one side of the central canal. Where the lamellar bone meets the woven bone, there is a transition state with randomly dispersed vascular canals, and is also marked by secondary Haversian systems with clear cement lines. Such vascular canals are mostly absent in the woven bone; the only vascular canals seen are in primary Haversian systems. The woven bone matrix is disorganized, made of primary bone, and juvenile in nature, as suggested by both the vasculature and the primary bone. There is no endosteal lamellar bone present.

A distal end of a mid-thoracic rib from H-GSP 96602 has similar histology to the previous sample. However, there are a few noteworthy differences. The Haversian systems appear to be more advanced, some being secondary with distinct cement lines. Unlike the previous Pakicetus, the woven bone is present but there are a few primary Haversian systems in this woven bone. Another difference seen is the presence of plexiform bone type with remnants of longitudinal and radial canals. There are also a few layers of lamellar bone surrounding the endosteal surface, which was not seen in the mid-thoracic section. This lamellar bone can even be classified as plexiform bone seen in modern cattle, suggesting juvenile bone in fast development.

The trabecular bone in all the pakicetids is thick and packed tightly in the marrow cavity. Compared with semiaquatic and terrestrial extant mammals, pakicetids have greater overall BV relative to TV of the bone, but only slightly thicker trabecular struts. There are endosteal lamellae surrounding the trabecular spaces without signs of Howship's lacunae, suggesting active addition of bone matrix without absorption of previous primary matrix. This is also suggested by the presence of somewhat thinner struts that appear to represent primary deposition.

Ambulocetidae: Ambulocetus.

Four sections were analyzed from different portions of ribs from H-GSP 18507; two samples were taken from the distal end of an upper thoracic rib, a third from the distal end of a lower thoracic rib, and a fourth from a distal portion of a mid-thoracic rib. Consistent with the distal placement of the sections, these ribs have only a thin layer of cortical bone. Their thin cortices may be due to artificial loss of the outer layer of the fossil or the fact that the majority of specimens are derived from distal segments of ribs, but may also be due to the juvenile status of the type specimen. The cortical thickness and BV/TV of the Ambulocetus distal and proximal ribs are similar to mid-thoracic measures of extant taxa, while similar measures of the mid-thoracic rib falls out with the rest of the archaeocetes.

There is lamellar bone on the periosteal surfaces. The bone is marked with dense trabecular struts with little space between the struts. The bone exhibits osteopetrosis and pachyostosis. The entire thickness of the cortical bone is composed of woven bone. However, it has more organization then Pakicetus. The lacunae are seen making an “s” shaped pattern in the cortical bone (Fig. 2A). There are a few randomly placed Haversian systems within the wavy pattern of woven bone (Fig. 2B). The Haversian systems are mostly primary and have no consistent size or shape, including some sliding Haversian systems. Some are found in typical formation and the beginnings of some are seen with lacunae surrounding oblique and longitudinal vascular canals. Most of the Haversian systems being formed are oblong in shape and have only a few rings of even sized and shaped lacunae around them. Some of the oblong vascular canals are in a plexiform pattern. There is no lamellar bone present on the endosteal surface.

Figure 2.

Histology of Ambulocetus natans H-GSP 18507 ribs. A: View of periosteal surface. LAG, LAG lines present in periosteal lamellar bone; SW, “s” shaped woven bone. B: View of endosteal surface. W, woven bone. C: View of endosteal surface containing plexiform bone. The dotted line shows the original longitudinal canal. LC, longitudinal canal; RLC, remodeled longitudinal canal; RC, radial canal; SH, secondary Haversian system. D: View of the trabecular bone. EL, endosteal lamellae.

A fourth section from the same individual was taken from a mid-thoracic rib, mid-shaft. There was a thin layer of cortical bone around the periosteal surface. The periosteal surface has lamellar bone with a few faint LAG lines; the periosteal surface appears intact along most of the circumference, suggesting that bone was not lost in preservation and preparation. The lamellar pattern on the inner surface is occasionally interrupted by Haversian systems. The cortical bone consists of woven bone with many remodeled Haversian systems. There are only remnants of woven bone left. The Haversian systems are mostly primary but there are more secondary Haversian systems with distinct cement lines than seen in Pakicetus. The woven bone does not have any other vasculature beside the few Haversian systems. There are no oblong or longitudinal canals, so all the vasculature has been remodeled into primary Haversian systems in the plexiform bone pattern (Fig. 2C). Most of the Haversian systems are organized in radially oriented lines showing organization of the vasculature not seen in Pakicetus. There is no lamellar bone on the endosteal surface.

The Ambulocetidae ribs all have dense trabecular struts. Figure 3 shows that the mid-thoracic rib sample of Ambulocetus falls out of the extant taxa cluster. The lacunae, as in Pakicetidae, are surrounding the spaces between trabeculae, with endosteal lamellae that are not as distinct and thick as in other archaeocetes. The nature of the bone development in H-GSP 18507 is likely due to the immaturity of the individual, exemplified by the lack of complete fusion of long bone and vertebral epiphyses (Thewissen et al., 1996). The endosteal lamellae are being added without absorption of the previous primary bone structure (Fig. 2D).

Figure 3.

Rib cortical and trabecular thickness relative to overall bone volume. Bone volume is calculated according to trabecular volume/total volume (BV/TV) of the region of interest, taken from the center of the trabecular region of each rib. A: Bone volume versus maximum cortical thicknesses. Note measurements are not size corrected. Small archaeocetes cluster with large terrestrial mammals in cortical thickness, and all large archaeocetes have some rib segments with extremely thick cortices. All have bone volumes well outside terrestrial mammals, in line with the volume of the sea otter Enhydra. B: Bone volume versus maximum trabecular strut thickness. Trabecular thicknesses increases in archaeocetes in parallel with cortical dimensions.

Remingtonocetidae: Kutchicetus.

The ribs of Kutchicetus shows signs of osteopetrosis and pachyostosis, similar to the other archaeocetes. The bone is extremely dense, with little dense trabecular struts present in the marrow cavity (Fig. 4A). This finding can again be seen in Figure 3A where Kutchicetus is clustered with the other archaeocetes. The periosteal surface has dense lamellar bone with distinct LAG lines. Secondary Haversian systems are seen along one such LAG line, indicating extensive remodeling (Fig. 4B). The cortex is composed of highly organized laminar bone; however, there is a small remodeled area in which woven bone interrupts the laminar bone. This small section of woven bone has no vasculature, and the lacunae are in a wave-like pattern. The laminar bone on the periosteal surface of the rib contains lacunae that are relatively consistent in shape and size with one another. However, the laminar bone consists of longitudinal and oblique vascular canals that have been remodeled into primary Haversian systems, some of which are sliding Haversian systems (Fig. 4C). This is shown by lacunae that are encircling the vascular canal. However, no remnants of plexiform bone are found as in Pakicetidae and Ambulocetidae. There is no lamellar bone on the endosteal surface, but there is one layer of circling lacunae surrounding the marrow cavity.

Figure 4.

Histology of Kutchicetus minimus IITR/SB 2647. A:Kutchicetus minimus IITR/SB 2647 whole section view. B:Kutchicetus minimus IITR/SB 2647 view of periosteal lamellae. LAG, LAG line present in layers of periosteal lamellae; SH, secondary Haversian system. C:Kutchicetus minimus IITR/SB 2647 view of endosteal surface. The dotted line demarcates the differentiation between the endosteal surface and the medullary cavity. WR, woven remnants, EL, endosteal lamellae.

The trabecular struts are extremely dense in Kutchicetus (Fig. 3B), more dense than all extant taxa. Trabecular strut thickness and BV are greater than the in semiaquatic and terrestrial individuals. The lacunae of the endosteal lamellae are highly organized with many concentric layers. There are no signs of remodeling of the endosteal surface, suggesting that bone was added to intact primary bone to create the extreme cortical thickness in the Kutchicetus rib.

Protocetidae: Georgiacetus.

The cortex of Georgiacetus is thick, but not as densely packed as the other archaeocetes, as the bone is punctuated by more vascular canals. As in all of the archaeocetes, there is no marrow cavity; it is filled with trabeculae and endosteal lamellae. Georgiacetus falls out among the other archaeocetes in bone density (Fig. 3A). Like the other archaeocetes the ribs are osteopetrotic, but with less gross increase in rib diameter, thus less pachyostotic. The periosteal surface contains a few layers of circumferential lamellae with a few distinct LAG lines. There are remnants of woven bone in the cortex with primary Haversian systems as the only form of vasculature in these regions (Fig. 5A). The only vasculature present in the remaining bone matrix is primary or secondary Haversian systems. The lacunae surround the Haversian systems in even, circular patterns. The primary Haversian systems interrupt the woven bone that was previously laid down. These primary Haversian systems were added secondarily to the bone tissue without resorption of the primary bone matrix. There are no longitudinal or radial canals present and no plexiform bone as seen in other archaeocetes. There are no circumferential lamellae on the endosteal surface.

Figure 5.

Histology of Georgiacetus vogtlensis GSM 350, Zygorhiza kochii ALMNH PV 2000.1.1.1 ribs, and Gaviacetus sahnii IITR/SB 2870. A:Georgiacetus vogtlensis GSM 350 view of cortical layer. B:Georgiacetus vogtlensis GSM 350 view of medullary cavity highlighting endosteal lamellae. C: Whole view of rib from Zygorhiza kochii ALMNH PV 2000.1.1.1. D:Zygorhiza kochii ALMNH PV 2000.1.1.1 view of periosteal surface. PH, primary Haversian system; SH, secondary Haversian system; SSH, sliding secondary Haversian system. E:Zygorhiza kochii ALMNH PV 2000.1.1.1 view of trabecular bone. F:Gaviacetus sahnii IITR/SB 2870 view of cortical bone with plexiform pattern. The dotted line demarcates the original path of the longitudinal canal. LC, longitudinal canal; RLC, remodeled longitudinal canal; RC, radial canal. G:Gaviacetus sahnii IITR/SB 2870 view of endosteal woven bone. W, woven bone; EL endosteal lamellae. H:Gaviacetus sahnii IITR/SB 2870 view of medullary cavity. PH, primary Haversian system, SH, secondary Haversian system; WR, woven remnants; EL, endosteal lamellae; EL, endosteal lamellae; HL, Howship's lacunae; CC, calcified cartilage.

The trabeculae filling the marrow cavity are thicker than terrestrial mammals (Table 2; Fig. 3); however, they are not as densely packed as those in earlier archaeocetes. There are some lamellae present in trabecular struts, but no more than three or four were found in any particular strut (Fig. 5B).

Table 2. Bone distribution in archaeocete and extant mammalian rib cross sectionsa
TaxonMax trabecular thickness (μm)Min cortical thickness (mm)Max cortical thickness (mm)BV/TV
  • a

    BV/TV, bone volume/total volume.

Pakicetus210.40.562.470.62
Ichthyolestes205.41.662.300.33
Pakicetus141.20.791.270.33
Ambulocetus 1222.70.483.570.17
Ambulocetus 2366.80.811.120.62
Remingtonocetus248.80.952.770.34
Remingtonocetus186.91.703.310.27
Kutchicetus693.81.405.630.63
Gaviacetus532.11.506.300.76
Georgiacetus435.21.866.670.54
Zygorhiza370.1∼0.002.590.53
Bos169.40.200.590.14
Equus219.01.203.100.26
Juv. Canis224.00.401.300.20
Juv. Odocoileus265.00.702.000.18
Juv. Odocoileus192.50.401.600.12
Ad. Odocoileus528.00.100.700.25
Lutra152.50.201.100.25
Lutra119.80.300.800.21
Enhydra522.41.973.680.68
Choeropsis176.60.381.810.17
Delphinus203.61.112.470.29

Protocedidea: Gaviacetus.

Both distal and mid-shaft sections were taken from a mid-thoracic rib from a single individual, IIR/SB 2870. Both samples were extraordinarily dense with little or no trabecular struts present in the filled marrow cavity. The bone is marked with osteopetrosis with no signs of pachyostosis. Both samples of Gaviacetus can be seen in Figure 3A as falling out with the other archaeocetes, with thicker cortical bone and increased BV. The mid-thoracic cross-section has a thicker cortex than the distal section, with a few layers of lamellar bone encircling the periosteal surface with LAG lines. The cortical bone consists of laminar bone, with more advanced Haversian systems than seen in Pakicetus, Ambulocetidae, and Remingtonocetidae (Fig. 5F). The majority of vasculature is formed by primary Haversian systems, some of them still oblong-shaped remnants of plexiform bone (Fig. 5G). There are only a few oblique and longitudinal vascular canals in the matrix. The lacunae are all different in size and shape.

Although fewer in number than other archaeocetes, the trabecular struts are thick and bear lacunae that are consistent in size and shape. The entire cavity is filled with endosteal lamellae that have been laid down with limited absorption (Fig. 5H). Figure 3B indicates that Gaviacetus clusters with Kutchicetus in the extreme thickness of cortical bone and trabecular density. As in other archaeocetes, trabeculae show no sign of osteoclastic activity.

Basilosauridae: Zygorhiza.

The overall bone density and cortical bone thickness in Zygorhiza is less than the protocetid and remingtonocetid archaeocetes that precede it (Fig. 5C). The marrow cavity is filled with trabecular struts, but the BV/TV is more similar to distal rib sections of these taxa. This bone can still be classified as pachyosteosclerotic, as in previous studies (Buffrenil et al., 1990), but the BV and cortical thickness shows that Zygorhiza falls out with the earlier, much smaller pakicetid archaeocetes (Fig. 3A). There is no lamellar bone on the periosteal surface; the cortex has mostly woven bone with sections of interrupting lamellar bone. There are also sliding secondary Haversian systems within the woven bone (Fig. 5D). There are a few layers of nondistinct lamellar bone on the endosteal surface. Within the woven bone, the lacunae and vasculature are mostly unorganized; a few oblique vascular canals have several rows of evenly spaced lacunae surrounding them. Longitudinal vascular canals also form a few primary Haversian systems, but most of the Haversian systems are secondary with clear cement lines. These secondary Haversian systems are not organized within the cortex; they are interrupting the woven bone pattern, suggesting substantial remodeling of bone to a greater degree in Zygorhiza than in any of the other archaeocetes.

The trabecular struts are composed mostly of lamellae, the borders of which are not distinct and, therefore, form Howship's lacunae (Fig. 5E). Zygorhiza appears to have less remodeling within trabecular lamellae, but as shown in Figure 3B, its strut thicknesses and BV are still well outside the range of extant taxa.

Extant Taxa

Odocoileus virginianus.

Both a juvenile and adult white-tailed deer, Odocoileus virginianus, were sampled (Fig. 6A). Four sections, approximately 1 cm apart, were taken along the length of a juvenile mid-thoracic rib. The periosteal bone along the anterior and posterior edges is not organized. Four or five layers of parallel lamellar bone are distinct along the medial and lateral periosteal surfaces of all four sections taken from the mid-thoracic rib. Following these layers inward, the cortex becomes highly packed with Haversian systems. Secondary Haversian systems are positioned toward the more endosteal region of the cortex. The trabecular spaces in the medullary cavity were enlarged by resorption. The development of this bone is characteristic of the juvenile osteogenesis as also observed in Canis and Equus (see below). Other than the ratio of cortical bone to trabecular bone present within these four sections, there is little morphological variation along the length of the juvenile deer rib. There was no semiwoven bone in this immature animal.

Figure 6.

Histology of extant mammal ribs. A:Odocoileus virginianus cortical view. WR, woven remnant; PH, primary Haversian system. B:Equus caballus cortical view. SH, secondary Haversian systems; IL, interstitial lamellae. C:Lutra canadensis cortical view. WR, woven bone remnants; PH primary Haversian systems; SH, secondary Haversian systems. D:Bos taurus cortical view of plexiform pattern. E:Enhydra lutris LC, longitudinal canals; RC, radial canals; RLC, remodeled longitudinal canals; PH, primary Haversian systems, WR, woven remnants, RRC remodeled radial canal.

Four sections, approximately 1 cm apart, were taken from a mid-thoracic rib of an adult Odocoileus virginianus, and seven sections were taken of a distal rib. In comparison to the immature deer, the cortex of the adult deer ribs is much thinner and more highly remodeled, bearing numerous overlapping secondary Haversian systems. The lateral, anterior, and posterior surfaces of the most proximal section of mid-thoracic rib have very compact Haversian bone relative the medial surface. Endosteal lamellae are found along the endosteal margin of all sections.

The distal rib is similar to the mid-thoracic rib. The primary differences are that its cortex is thicker throughout its length, trabecular density is similarly greater, and there appears to be greater organization of Haversian bone along the medial margin of the rib.

Equus caballus.

Two sections were taken along the length of a mid-thoracic rib of an adult horse, Equus caballus. The proximal section appears rectangular in shape, while the distal sample is elongated with more trabecular infill. Haversian systems within the proximal section are more densely packed than the more dispersed distribution distally (Fig. 6B). The periosteal cortex of the proximal sections is highly remodeled. Secondary Haversian bone is visible throughout the entire cortex with interstitial lamellae. The endosteal surface is composed of multiple secondary Haversian systems, some with very large circulatory canals.

Both the lateral and medial cortices of the distal rib section are thicker than its anterior and posterior margins, but are similar to other large terrestrial mammals examined (Table 2). Only primary Haversian systems are present, suggesting that remodeling has not occurred in this distal section. Haversian bone is present within the trabeculae as well; many bear large vascular canals suggesting that the bone is relatively immature. Howship's lacunae are found in the margins of the medullary cavity. As the specimen grew, the size of the medullary cavity was increased in response to circumferential growth.

Canis familiaris.

Three sections were taken from both a mid-thoracic and a distal rib of a single juvenile dog, Canis familiaris. Grossly, the most proximal rib sections taken from both mid-thoracic and distal ribs have rather hollow medullary cavities with only a few trabeculae spanning the entire width of the marrow cavity. The more distal sections of the mid-thoracic and distal ribs possess many more trabecular struts and less cortical organization than the more proximal sections. This finding suggests that, as one moves distally along the rib length, the trabecular bone increases in density, but is less organized.

A mixture of primary and secondary Haversian systems are visible throughout the cortical bone in each section. However, the medial periosteal edge is composed predominantly of primary Haversian bone, while distinct secondary Haversian bone occurs within the lateral periosteal surface of the ribs. Circumferential lamellae are not present in the periosteal region. The periosteum is jagged with concave pits present around the perimeter. Osteocytes are unevenly distributed within the cortex of both ribs, and distribution becomes even more random distally in each rib. The large number of Howship's lacunae found within the endosteal surface suggests that the cortical bone was being actively remodeled, and new layers of trabecular bone were added in an organized manner to the endosteal surface. These layers are continuous along the endosteal and trabecular edges, interrupting previously deposited bone.

Choeropsis liberiensis.

One mid-thoracic rib of a pygmy hippo, Choeropsis liberiensis, NMNH 538815 was sectioned near mid-shaft. This specimen was obtained from a zoo specimen, and signs of osteoarthritic growth were found in both axial and appendicular regions of the skeleton. The periosteal surface of the rib was covered with osteophytes. The rib cortex was thin and more porous than any mammal examined, suggesting that the specimen was pathological. It was, thus, not included it in the comparative analysis.

Trichechus manatus.

The histology of the manatee (Trichechus manatus) rib section matched the pachysoteosclerotic condition described in a previous study (Ricqles and Buffrenil, 2001). The manatee rib was entirely lacking a marrow cavity and any trabecular infill (see also Buchholtz, 2007, this issue, for anatomy of manatee vertebral column).

Lutra canadensis.

Two sections were taken from both a proximal and a distal rib of a river otter, Lutra canadensis (Fig. 6C). The cortex of all sections is thin as a result of increased osteoclastic activity. In the most proximal section of the mid-thoracic rib, a few patches of semiwoven bone are visible in the cortex. The majority of Haversian systems seen in this region are primary, showing no evidence of bone remodeling. However, a few secondary Haversian systems are seen toward the endosteal surface of the bone. The sizes and shapes of these Haversian systems also vary greatly. In the distal portion of the mid-thoracic rib, the cortex is composed of a mixture of parallel-lamellar and woven bone. A few primary Haversian systems are found in this region and along the endosteal margin, suggesting that some remodeling had occurred.

The most proximal section of the distal rib does not show signs of remodeling; all Haversian systems in this section are primary. As in the proximal rib sections, more layers of bone are being added to the endosteal surface of the bone. This finding suggests that the medullary cavity was actively being filled. The most distal section of the distal rib is also void of secondary Haversian systems, and trabecular bone filled the marrow cavity.

Bos taurus.

A single section was sampled from the mid-thoracic rib of a cow, Bos taurus. Cortical thickness is similar to other large terrestrial mammals sampled (Table 2). The periosteal surface of the sample is highly organized, with layers of lamellar zonal bone between longitudinal circulatory canals. Mostly primary Haversian systems were noted along the periphery of the bone. Secondary Haversian systems arise toward the endosteal surface. The endosteal surface of the rib neck is composed of unorganized woven bone in which osteocytes are laid down in a haphazard manner. Along the anterior edge of this section, there is plexiform bone with longitudinal and radial canals still present (Fig. 6D). The highly vascular plexiform bone is an indication of rapid bone development.

Enhydra lutris.

Two samples, a mid-shaft and distal section, of a mid-thoracic rib of the sea otter, Enhydra lutris, were prepared (Fig. 6E). Unlike Lutra, the sea otter exhibits highly remodeled, dense cortical bone (Table 2; Fig. 3). On the medial and lateral periosteal surfaces, thick layers of circumferential lamellae are present. Periodically, a circular row of Haversian systems interrupts the parallel layers of osteocytes. Several endosteal lamellae rim the endosteal cortical surface, filling multiple Howship's lacunae. However, it appears that bone was added faster than it was resorbed. Trabeculae are both thicker, and more densely packed than in the river otter, and the other terrestrial forms sampled (Fig. 3).

Delphinus capensis.

A single section was sampled from the mid-shaft region of a mid-thoracic rib of a long-beaked common dolphin, Delphinus capensis. The periosteal edge is similar to that of Canis; jagged in nature and void of circumferential lamellae or other organized bone. The cortex is thin, and highly remodeled with several secondary Haversian systems and endosteal lamellae. The Howship's lacunae present within the endosteal lamellae are clear evidence of bone resorption. Osteoclastic activity is also prevalent along margins of trabeculae struts, leading to the decreased trabecular volume and strut thickness (Table 2; Fig. 3B).

DISCUSSION

In their discussion of basilosaurid pachyostosis, Buffrenil et al. (1990) note the apparent “early” appearance of histological specializations of bone that are found in independent lineages of secondarily aquatic tetrapods. In this study, the relationship between bone histology and the anatomical correlates of locomotor specialization in each of these transitional archaeocete families is examined, and the biogeographical implications of the morphological shifts observed within this sample isdiscussed.

Comparative Osteogenesis

Pakicetidae.

Pakicetus rib sections, as well as all of the other archaeocetes sampled, are histologically distinct from the terrestrial mammals sampled. They developed osteopetrosis by replacement of the original cartilage matrix with bone, which continued to grow periosteally. Inhibited osteoclast activity limited remodeling of the original bone matrix. More bone was added to the existing space by filling in the medullary cavity with endosteal lamellae. Disorganization is the rule throughout the rib sections; layers of juvenile bone were laid over with adult bone but were not remodeled away as seen in extant terrestrial mammals. The primary bone deposits in juvenile terrestrial mammals are remodeled into more organized secondary bone. The plexiform bone near the endosteal surface suggests a fast growing individual, which resulted in disorganized, densely packed cortical bone. However, the circumferential lamellar bone seen on the periosteal surface suggests some organization, and the LAG lines suggest continued adding of lamellar layers to the periosteal surface. Pakicetids started with plexiform juvenile bone, like most other fast growing terrestrial mammals, but created ballast by continuing outward deposition, while not reabsorbing the original matrix on the endosteal surface. In the trabeculae, this persistent addition of new bone without re-absorption is seen in substantial development of endosteal lamellae. However, the endosteal lamellae are not as organized as in other archaeocetes, and the layers of bone deposited are not encircling the entire margin. There are original trabecular struts present, as well as trabecular struts with additional layers deposited upon them in a circular manner, filling in the trabecular spaces. There are no Howship's lacunae, further suggesting inhibition of osteoclast activity. The bone deposition was extended later in ontogeny to achieve increased density. This morphology is most consistent with the condition described in Enhydra among the modern aquatic and semiaquatic taxa sampled.

Ambulocetus.

The type specimen of Ambulocetus clearly represents a more immature state of osteological development than the other archaeocetes examined, consistent with the lack of fusion of many of the type specimen's epiphyses. Its rib cortical bone is thinner, but more organized than that of Pakicetus. Woven bone appears in the majority of the cortex, suggesting rapid bone deposition. Increased density is achieved by adding Haversian bone to trabeculae, as well as adding endosteal lamellae. The greater diversity of bone types deposited in Ambulocetus suggests more remodeling than in Pakicetus, or that the bone was still in a state of active formation, consistent with its juvenile state. This would explain the relatively thin cortical values of the four ribs sampled for Ambulocetus compared with both earlier and later archaeocetes. There was still active deposition in cortical and trabecular regions, consistent with the osteopetrotic state of many of the long bones in this specimen (Thewissen et al., 1996).

Remingtonicetidae.

Kutchicetus is more organized than both Ambulocetus and Pakicetus, indicated by the greater amount of secondary bone deposits. The bone has been highly remodeled; many secondary Haversian systems interrupt the circumferential lamellar bone. The periosteal circumferential lamellae also bear distinct LAG lines. The original matrix was likely laid down quickly, then added to by means of remodeling without absorbing the immature matrix. The trabeculae are abnormally thick and also bear signs of remodeling including the presence of endosteal lamellae without evidence of osteoclast activity. Increased bone density was accomplished by simply growing the bone outward without much re-absorption of the original cortex.

Protocetidae.

Georgiacetus is the intermediate condition between the earlier archaeocetes and Zygorhiza. The density is not nearly as great as in the earlier archaeocetes. The are many histological similarities seen in Zygorhiza and Georgiacetus, including more poorly developed endosteal lamellae, prominent Howship's lacunae, and endosteal lamellae present in the middle cortical layer. The replacement of woven bone with primary Haversian systems without prior re-absorption of the woven bone type is consistent with the description of Buffrenil et al. (1990), in which they suggested that Basilosaurus lacked osteoclastic activity. However, the bone density of Georgiacetus is not as high as the Basilosaurus as described by Buffrenil et al. (1990).

Basilosauridae.

Zygorhiza is both osteosclerotic and pachyostotic, but differs from the basilosaurine Gaviacetus and remingtonocetids in exhibiting fewer periosteal lamellae. It also has a thinner overall cortical layer relative to Gaviacetus and Kutchicetus. Endosteal lamellae were still added on to the trabecular struts, but there are Howship's lacunae along the interior margin of endosteal lamellae. There are also remnants of calcified cartilage and lacunae of chondrocytes in the medullary cavity, unlike other archaeocetes. This finding shows reduced osteoclastic activity in a manner unlike previous archaeocetes, perhaps at a much slower rate, and at an early stage in development.

Gaviacetus is similar in appearance and development to both Ambulocetus and Kutchicetus. Gaviacetus shows similar development of osteopetrosis, with clear evidence of osteoclastic inhibition. All sections have juvenile, plexiform bone and remnants of woven bone indicative of laying down secondary bone without re-absorption of the primary bone deposits. One remarkable difference seen in these three compared with Pakicetus is increased osteological organization. With the exception of the juvenile condition Ambulocetus, each displays layers from the periosteal lamellae to the endosteal surface. Ballast was increased by means of the consequent pachyostosis, as well as the addition of endosteal lamellae to the trabecular struts.

Functional Implications

Having histologically sampled archaeocetes throughout the first 10 million years of their evolutionary history, we know that microstructural changes in bone predate nearly all of the gross anatomical correlates of aquatic locomotion in early whales. Histologically, pakicetids (and all subsequent archaeocetes) are absolutely distinct from terrestrial mammals; all have either increased cortical thicknesses, trabecular strut thickness, and trabecular densities. Most show all three, and like the modern sea otter and sirenians, fall well outside the ranges of modern terrestrial taxa across all size ranges. Armed with this information, we can begin to examine the relationship between gross morphology and histomorphology. We can then use this understanding to examine the adaptive radiation of archaeocetes during their transition to modern form and distribution.

Hind limbs in whales were reduced during their ecomorophological shift from a swimming mode of quadrupedal paddling in the shallow, fresh water pakicetids to the caudal oscillation of modern cetaceans filling shallow and deep water niches. Figure 7 shows the phylogeny of the five archaeocete families and their skeletal ballast system, mapped against their evolving swimming acuity.

Figure 7.

Summary diagram of archaeocete bone histology and its functional implications over the first 10 million years of cetacean evolution.

While the postcranial morphology of pakicetids bears all of the hallmarks of their cursorial ancestry (Thewissen et al., 2001, Madar, 2007), their osteopetrosis, described here in the ribs, is systemic, and is thus incompatible with safe, energetically efficient, terrestrial locomotion, and cursoriality in particular (Madar, in 2007). Dense bone served as a buoyancy control, associated with bottom feeding adaptations in these fresh water, primitive cetaceans.

Earlier radiographic analyses also documented systemic ostelogical changes in Ambulocetus, (Thewissen et al., 1996), protocetids, remingtonocetids, as well as both basilosaurine and dorudontine members of the Basilosauridae (Madar, 1998). Long bones and ribs of each group lack marrow cavities, and have variably thick cortices. The development of this static buoyancy control system in these early archaeocetes is accomplished in three ways; osteosclerosis with inhibited osteoclasts, osteosclerosis combined with pachyostosis with inhibited osteoclasts, and osteosclerosis combined with pachyostosis, but with reduced osteoclast activity. Pakicetus develops by means of the first mechanism, Ambulocetus, Kutchicetus, and Gaviacetus develop by the second method, and Zygorhiza and Georgiacetus appear to develop by means of the third method. These differences are themselves linked to structural adaptations commensurate with increasingly aquatic specialization over the ten million years that were sampled, and fewer and fewer signs of any terrestrial dependence (Fish, 1996; Carter et al., 1996; Thewissen and Fish, 1997; Madar, 1998).

Buffrénil et al. (1990) suggested that the restriction of Basilosaurus' pachyostosclerosis to the ventral thorax was akin to a similar regionalization of pachyosteosclerosis seen in modern Sirenia. However, cortical bone was massively expanded in its highly reduced femora (Madar, 1998). Only in the dorudontinae do we see loss of long bone diaphyseal cortical bone, combined with hyperostotic ribs in a manner similar to modern Sirenia (Gingerich and Uhen, 1996; Uhen, 1998). The histological features of Zygorhiza indicate that bone remodeling played a much larger role in its development than in all other archaeocetes examined, and its trabecular volume and cortical densities do not reach the extremes of the other taxa. This finding indicates that regionalization is an advanced feature that precedes the systemic osteoporosis exhibited in modern cetaceans. Not surprisingly, it is dorudontines, not basilosaurines that have the body shape consistent with the dorsoventral oscillatory mode of locomotion found in all modern cetaceans (see also Fish et al., 2007, this issue, for hydrodynamic implications of cetacean fluke shape). As noted by Buffrénil et al. (1990), their morphology is consistent with hydrodynamic control of body trim, unlike Basilosaurus.

Thus, only in the members of the dorudontine basilosaurids do we see histological shifts in osteogenesis toward the osteoporotic condition seen in extant cetaceans. This finding would be commensurate with their clear adaptations toward a more fully derived form of dorsoventral oscillatory form of locomotion. Such increased swimming efficiency is associated with dynamic buoyancy control mechanisms in modern cetaceans and is commensurate with the more open water adaptations of this group, as compared with the shallow costal or freshwater niches of earlier taxa. In species inhabiting deep water, static ballast that helps an animal sink would be a detriment to animals that still need to rise to the surface to breathe (MacLeod, 2002).

In sum, the transition from terrestrial, to semiaquatic, to obligate aquatic locomotion in archaeocetes involved a radical shift in bone function at the microstructural level. The constraints of terrestrial weight bearing are very different from those related to locomotion in water, and the skeleton is clearly a highly plastic organ for solving some of the mechanical issues of aquatic existence. What remains staggering is that microstructural change predates gross anatomical shift in archaeocetes associated with swimming. The mechanisms of osteogenesis were flexible enough to accommodate the shift from typical terrestrial form, to osteopetrosis and pachyosteosclerosis, and then to osteoporosis in the first quarter of their evolutionary history.

Buffrénil et al. (1990) stated that the histological changes seen in bone of secondarily aquatic tetrapods should not be classified as pathology, but adaptations that were of immediate selective advantage, and that heterochronic mechanisms led to the hyperostotic bone that characterizes most of the archaeocetes and semiaquatic mammals that we describe. Further study may allow us to understand what environmental and mechanical triggers are associated with the down- and up-regulation of osteoclasts that appears to be at the heart of the structural adaptations found in secondarily aquatic tetrapods. Genetic mutations leading to inhibition of osteoclast differentiation can impact several points in the pathway, as shown in many mice and human models of osteopetrosis and osteosclerosis (for review, see Karsenty, 1999). Further studies should focus on the environmental triggers leading to selection-mediated hyperostosis in aquatic tetrapods.

Acknowledgements

Thanks to Dr. Joy Reidenberg for the invitation to submit this work. We are extremely grateful for access to archaeocete fossil material for histological study: Dr. J.G.M. Thewissen of NEOUCOM for Pakicetus, Ichthyolestes, and Ambulocetus; Thewissen and Dr. Sunil Bajpai of Indian Institute of Technology for the material of Kutchicetus, Remingtonocetus, and Gaviacetus; Dr. Mark Uhen of the Cranbrook Institute of Science for Zygorhiza, and Dr. Jonathan Geisler of Georgia Southern University for Georgiacetus. Extant material was graciously provided by: Dr. James Mead and Linda Gordon of the Smithsonian Institution, Division of Mammals for Trichechus, Choeropsis, and Hippopotamus; Sharon G. Toy-Choutka of the California Department of Fish and Game, Marine Wildlife Veterinary Care and Research Center for Enhydra; and Kerri Danil of the Southwest Fisheries Science Center for Delphinus. We are especially grateful to David Waugh and Dr. Rodney Feldman of Kent State University for training and open access to thin sectioning equipment, Dr. J.G.M. Thewissen for use of sectioning equipment and microscopy, and to Dr. Donna King of NEOUCOM for training and access to NEOUCOM's BIOQUANT facility for histological analysis. Thanks also to Robin Latkovich for artwork. Two anonymous reviewers greatly improved this manuscript. Thewissen and Madar were funded by the National Science Foundation, and Hiram College received a Howard Hughes Medical Institute grant.

Ancillary