Evolution and Development of Cetacean Appendages Across the Cetartiodactylan Land-To-Sea Transition

Home , Ambulocetus, Anthracotheriidae, Aquatic locomotion, Archaeoceti, Bat wing development, Diacodexis

EVOLUTION AND DEVELOPMENT OF CETACEAN APPENDAGES

A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Lisa Noelle Cooper

December, 2009

Dissertation written by Lisa Noelle Cooper B.S., Montana State University, 1999 M.S., San Diego State University, 2004 Ph.D., Kent State University, 2009

Approved by

______, Chair, Doctoral Dissertation Committee J.G.M. Thewissen

______, Members, Doctoral Dissertation Committee Walter E. Horton, Jr.

______, Christopher Vinyard

______, Jeff Wenstrup

Accepted by

______, Director, School of Biomedical Sciences Robert V. Dorman

______, Dean, College of Arts and Sciences Timothy Moerland

TABLE OF CONTENTS

LIST OF FIGURES ...... v

LIST OF TABLELS ...... vii

ACKNOWLEDGEMENTS ...... viii

Chapters Page

I INTRODUCTION ...... 1

The Eocene Raoellid Indohyus ...... 2 Cortical Bone Evolution in Cetartiodactylans ...... 5 The Apical Ectodermal Ridge ...... 6 Hyperphalangy ...... 8 The Soft Tissue Flipper ...... 9

II MORPHOLOGY AND LOCOMOTION OF THE MIDDLE EOCENE RAOELLID INDOHYUS (ARTIODACTYLA: MAMMALIA) ...... 11

Introduction ...... 11 Materials and Methods ...... 15 Results ...... 16 Discussion ...... 47

III EVOLUTION OF BONE MICROSTRUCTURE DURING THE AQUATIC INVASION OF CETARTIODACTYLA (MAMMALIA) ...... 63

Introduction ...... 63 Materials and Methods ...... 69 Results ...... 77 Discussion ...... 88

IV EVOLUTION OF THE APICAL ECTODERM IN THE DEVELOPING VERTEBRATE LIMB ...... 95

Introduction ...... 95 Materials and Methods ...... 105 Results ...... 106 Discussion ...... 114

iii

TABLE OF CONTENTS (Continued)

CHAPTER Page

V DEVELOPMENT OF DOLPHIN FLIPPERS: MOLECULAR EVOLUTION OF HYPERPHALANGY AND INTERDIGITAL WEBBING ...... 119

Abstract ...... 119 Introduction ...... 120 Materials and Methods ...... 123 Results ...... 124 Discussion ...... 130

VI REVIEW OF THE EVOLUTION AND DEVELOPMENT OF CETACEAN APPENDAGES ...... 135

Developmental Processes ...... 135 Form and Function ...... 138 Performance ...... 144 Environment ...... 146 Biological Role ...... 147

REFERENCES ...... 148

APPENDIX: CATALOGUE OF INDOHYUS POSTCRANIAL ELEMENTS FROM THE A. RANGA RAO COLLECTION USED IN THIS ANALYSIS ...... 181

LIST OF FIGURES

Figure Page

1 Reconstruction of Indohyus ...... 3

2 A block of sediment showing several disarticulated bones of Indohyus ...... 4

3 Hyperphalangy in modern cetaceans ...... 8

4 Skeletal reconstruction of Indohyus ...... 13

5 Ribs and vertebrae of Indohyus ...... 17

6 Vertebrae of Indohyus ...... 22

7 Forelimb elements of Indohyus ...... 26

8 The manus and pes of Indohyus ...... 31

9 Pelvic limb elements of Indohyus ...... 36

10 Tarsal elements of Indohyus ...... 42

11 Phalangeal cross-sections in Indohyus, pakicetids, and Ambulocetus natans...... 56

12 Bone cross-sectional geometries in Indohyus compared to terrestrial artiodactyls with an equivalent body size (Tragulus and Hyemoschus) ...... 59

13 Mathematial model that calculates amount of bone as a function of the distance to the center of the bone ...... 75

14 Radiographs through long bone midshafts of cetartiodactylans ...... 78

15 Radiographs through long bone midshafts of fossil cetaceans from the archaeocete Family Remingtonocetidae ...... 79

16 Double logarithmic plots of P (a measure of bone thickness) ...... 81

17 Phylogenetic tree showing an ancestral character state reconstruction by squaredchange parsimony in fossil and extant cetartiodactylans ...... 84 v

LIST OF FIGURES (Continued)

Figure Page

18 A graphical summary of P-values calculated in this analysis ...... 86

19 Schematics of apical ectodermal (AE) morphologies ...... 98

20 Morphology of an approximately 110 day old (Carnegie Stage 23) pantropical spotted dolphin (Stenella attenuata, LACM 94285) ...... 103

21 Schematic of the transition from an apical ectodermal ridge (AE-1) to a finfold in the killifish (Aphyosemion scheeli, modified from Wood, 1982) ...... 107

22 Apical ectodermal morphologies associated protein signals in the forelimbs of dolphins ...... 111

23 Forelimb development in embryos of the dolphin Stenella attenuate ...... 121

24 Fibroblast growth factor (Fgf) protein signals in the developing forelimb of Stenella embryos and a fetus...... 125

25 Wingless type 9a (Wnt-9a) protein signals in the developing forelimb of Stenella embryos ...... 126

26 Gremlin and bone morphogenic protein (Bmp) signals in the developing forelimb of Stenella embryos ...... 128

27 Simplified phylogeny of cetaceans with key events in forelimb evolution indicated ...... 132

28 The Adaptive Pathway includes key theoretical steps used to determine if a biological structure is a biological structure is the result of an evolutionary adaptation ...... 136

29 Phylogenetic distribution of modes of phalangeal counts in the digits of extant cetaceans ...... 140

30 Averaged lift data for the minke whale flipper model ...... 145

LIST OF TABLES

Table Page

1 Indohyus Metapodial and Phalangeal Measurements ...... 34

2 Cortical Bone Thickness in the Long Bones of Artiodactyls ...... 60

3 Definitions of Bony Histological Specializations that Alter Skeletal Mass ...... 65

4 Taxonomic Identity and Compactness Profile Parameters of the Femur, Humerus, and Tibia. Based on Midshaft CT Scans and Histological Sections of Cetartiodactylans ...... 70

5 Habitats of Cetartiodactylan Taxa ...... 73

6 Comparisons between the Apical Ectoderm of Vertebrates and the Apical Epithelial Cap of Regenerating Salamander (urodeles) Limbs ...... 99

7 Patterns of Morphological Variation and Fibroblast Growth Factor (Fgf) Expression in the Limb Ectoderm of Developing Vertebrates during Normal Limb Development ...... 115

vii

ACKNOWLEDGEMENTS

First I must thank my advisor, Dr. Hans Thewissen for mentoring and encouragement throughout my dissertation and graduate career. Dr. Thewissen provided uncommon opportunities to work on rare cetacean embryos and find cetacean fossils in the remote deserts of western India. I am grateful to have found Hans early in my years before I started graduate school as he has a novel vision of the field of cetacean skeletal and sensory evolution, and has been an ardent supporter of an integrative approach to scientific questions.

Additionally, this project could not have been completed without the support of Walt Horton welcoming me into his molecular lab and serving with him on two effective and useful academic committees. Dr. Chris Vinyard is thanked for insightful advice and an introduction to the field of functional morphology. Dr. Wenstrup and Dr. Feldman are also thanked for serving on my committee, and offering support throughout the dissertation process. Dr. Van der Schyf is thanked for acting as moderator during my oral defense.

I am grateful to Dr. Sunil Bajpai, Dr. B.N. Tiwari, and the late Dr. F. Obergfell for allowing me to study the skeleton of the fossil Indohyus as part of my dissertation research.

This exciting fossil provided an outstanding opportunity for scientific study and discourse, and altered the scientific and public understanding of cetacean ancestry. Dr. Bajpai also provided loans of several cetacean fossil limb elements. Dr. S. T. Hussain and the Geological Survey of

Pakistan are also thanked for use of several fossil pakicetids skulls and limb elements, originally collected from Pakistan, for morphological comparison. Jennifer Sensor, Bobbi Joe Schneider,

Richard Conley, and Amy Maas are thanked for fossil preparation of Indohyus. viii

Study of cetacean embryos was made possible by the generous loan of rare embryonic dolphins from the Los Angeles County Museum. Jim Dines, Dave Janiger, and the late John

Heyning are thanked for loan of these specimens and for allowing the first molecular exploration into the developmental underpinnings of forelimb development in dolphins. Jim,

Dave and John gave our lab a tremendous and exciting opportunity.

I am thankful to Dr. Steve Ward and Dr. Eric Blum for CT scans of several limb bones at

Akron City Hospital. Without their support, my study on cortical bone evolution would not have been possible. Drs. Richard Ketcham and Matt Colbert of the High-Resolution X-ray

Computed Tomography Facility at the University of Texas at Austin are also thanked for high resolution scans of fossil limb bones.

I am extremely grateful for the loan of a car load of artiodactyl bones from the

Smithsonian Institution provided by Linda Gordon and Dr. Jim Mead for use in this dissertation.

I also thank Dr. Nancy Simmons, Darrin Lunde and Eileen Westwig of the American Museum of

Natural History for loan and prompt shipment of several artiodactyl limb elements used in this study. Dr. John Barry from the Yale Peabody Museum is also thanked for loan of rare fossil anthracothere limb elements.

Funding for this research came from grants to me from the Ingalls-Brown Foundation,

NEOUCOM Skeletal Biology Fund, Kent State University Graduate Student Senate, American

Museum of Natural History Lerner-Gray Fund for Marine Research, and Sigma Xi Grant in Aid of

Research. Funding also was provided by grants to J.G.M. Thewissen from the National Science

Foundation- Earth Sciences.

Denise McBurney provided invaluable instruction in teaching molecular techniques and managing the lab of Dr. Horton. Sharon Usip and Richard Conley are thanked for providing scans of many artiodactyl limb bones. Sharon is also thanked for offering solid technical assistance regarding my immunohistochemical research. I also thank Dr. Phil Reno and

Ashleigh Nugent for assistance for molecular assistance in the lab of Dr. Horton.

I received wonderful administrative support and encouragement from Diana Dillon,

Margaret Sedensky, Debbie Heeter, Diana Dubinsky, and Judy Wearden. These administrators worked hard to make my time writing this dissertation relatively easy. Diana Dillon is thanked in particular for her support and friendship. The library staff at NEOUCOM, including Denise

Cardon, Laura Colwell, Lisa Barker and Heather McEwan, consistently went above and beyond the call of duty to help me gather information and are also thanked for their unending encouragement.

I am also thankful for the continued advice and support of Dr. Lindsey Leighton, who served on my Master’s thesis committee, and has continued to offer encouragement and sound strategic advice throughout this dissertation. Many people made significant contributions to this dissertation and a partial list includes: Drs. Frank Fish, Sentiel “Butch”

Rommel, Karen E. Sears, Maria Serrat, Phil Reno, and Tobin L. Hieronymus. Many graduate students have also contributed to this dissertation and a partial list includes: Amy Mork, Alison

Doherty, Ashleigh Nugent, Brooke Armfield, Mike Selby, Burt Rosenman, Erin Simmons, Angela

Horner, David Waugh, Mike Jorgensen, Verne Simmons, Gabrielle Radick, Dave Dufeau,

Meghan Moran, and Terry Lancaster. Amy and Angela are thanked in particular for many insightful for their continuing friendship.

Finally I am grateful for the support and love I have received from my parents, Barbara,

Tobin, Isabel, April, Erique, Cory, John, and Ray in making my time in graduate school possible and meaningful.

CHAPTER I

INTRODUCTION

Over the past decade, the origin of cetaceans (whales, dolphins, and porpoises) has become an excellent case study in mammalian evolution (Thewissen and Bajpai, 2001, Uhen

2007, Thewissen et al., 2009). A rich collection of fossil cetaceans, mostly from Asia, Africa and

North America, provides an ancient record of the cetacean transition from a terrestrial to a completely aquatic environment (e.g., Gingerich et al., 1994, 2001; Thewissen et al., 1994,

2001a,b; Hulbert, 1998; Hulbert et al., 1998; Bajpai and Thewissen 2000; Geisler, 2001; Uhen,

2004, 2007; Thewissen et al., 2009). Cetaceans initially took to the seas about 50 million years ago in an ancient shallow seaway, the Tethys Sea, that laid between the Indian subcontinent and

Asia (e.g. Thewissen and Bajpai, 2001). Within about 10 million years in the aquatic environment, cetaceans evolved specialized ears for hearing underwater, reduced the size of the hindlimbs, and developed tail flukes for propulsion. The cetacean body plan had so radically evolved from terrestrial artiodactyls (even-toed hoofed mammals) that modern cetaceans are no longer able to bear their weight on land and are obligatorily aquatic

(Thewissen and Fish, 1997; Buccholtz, 1998; Bajpai and Thewissen, 2000; Madar et al., 2002;

Gingerich et al., 2001; Gingerich, 2003).

Cetaceans are the only lineage of artiodactyls to become fully aquatic, and are now referred to as cetartiodactylans. However, recent fossil findings near the Jammu-Kashmir region

1 2

along the India-Pakistan border show that another artiodactyl family, Raoellidae, is closely related to cetaceans, lived in an aquatic environment, and never lost its ability to bear weight on land (Thewissen et al., 2007). Raoellid artiodactyls probably spent most of their lives wading in shallow aquatic habitats. These recent findings indicate that multiple groups of fossil artiodactyls occupied shallow water habitats, but only cetaceans evolved a body plan that facilitated their penetration of the marine environment.

This dissertation utilizes a variety of methods to address the evolution and development of cetacean appendages. First, comparative anatomical methods are used to investigate locomotor abilities of a fossil cetartiodactylan. I then quantify bone cross-sectional area using a computer program and modern comparative phylogenetic methods to reconstruct character evolution in the limbs of fossil and extant cetartiodactylans. Experimental molecular techniques are also utilized to answer questions regarding the development of the dolphin forelimb. A final chapter summarizes finding of each chapter and integrates these data in the context of evolutionary adaptation.

The Eocene Raoellid Indohyus

One family of artiodactyls, Raoellidae, was for decades known solely on the basis of isolated teeth and fragmentary skull elements. However, within the past few years, paleontologists uncovered new skeletal elements (skulls, jaws, and numerous postcranial elements) of the raoellid Indohyus (Figure 1) in the Jammu-Kashmir region along the India-

Pakistan border (Thewissen et al., 2007); the full skeleton of the raoellid Indohyus was accessible for study for the first time. These skeletal elements were preserved in fossil mudstones in an archaic Eocene streambed, in which many individuals died and decomposed such that bones 3

were stacked atop one another (Figure 2). It is estimated that at least 25 individuals of varying ages were buried together in this death assemblage.

Chapter 2 of this dissertation furthers the work of Thewissen et al., (2007) by describing the postcranial skeleton of Indohyus recovered from this fossil assemblage. By comparing the skeletal anatomy of Indohyus with a modern artiodactyl of similar body size (Tragulus), with the earliest fossil cetaceans (pakicetids), and with the earliest artiodactyls (diacodexids), this study identifies unique characteristics of the postcranial skeleton of Indohyus and makes inferences about the locomotor abilities of Indohyus based on analyses of bone and joint shape and articulation.

Fig. 1. Reconstruction of Indohyus (Cooper and Thewissen, In press). Illustration by Carl Buell. © Thewissen lab, NEOUCOM.

Fig. 2. A block of sediment showing several disarticulated bones of Indohyus (Cooper and Thewissen, In press). At least 25 individuals of various ages were preserved in this death assemblage.

Cortical Bone Evolution in Cetartiodactylans

Mammals that wade and swim in shallow bodies of water, such as Hippopotamus, otters and manatees, have limb bones with thick cortices and reduced marrow cavities. These thickened bones act as ballast to weight the skeleton and counteract the the animal’s buoyancy when in water. Their marrow cavities are typically less than 55% of bone diameter, while terrestrial mammals have thinner cortices and a larger marrow cavity that is about 60-70% of the bone diameter (Thewissen et al., 2007). The limb bones of Indohyus display a small marrow cavity that was only 42% of the total bone diameter (Thewissen et al., 2007). This striking increase in cortical thickness at the expense of the medullary cavity is termed osteosclerosis.

Although dimensions of the long bones of Indohyus indicate it occupied an aquatic niche and was an amphibious wader, the earliest cetaceans displayed a comparatively extreme form of osteosclerosis in that the medullary cavity was only 10% of total bone diameter (Thewissen et al., 2007). It could be that the extremely thickened bones of cetaceans afforded them decreased buoyancy compared to their relative Indohyus, and the dramatic bone thickness of cetaceans may have ultimately contributed to their successful transition to a completely aquatic lifestyle.

Chapter 3 of this dissertation also furthers the work of Thewissen et al. (2007) by undertaking a taxonomically broad survey of long bone cortical thickness within Cetartiodactyla

(cetaceans and artiodactyls). The sample includes modern taxa that occupy terrestrial (e.g., deer), marsh y or swampy (e.g., moose and sitatunga), and river habitats (i.e., Hippopotamus), as well as fossils collected from river and stream habitats including Indohyus and the earliest cetaceans. Specifically, this study tested the hypothesis (Hyp. #1) that those taxa inhabiting 6

aquatic habitats (e.g., rivers and streams) will display thicker long bone cortices compared to taxa occupying a terrestrial habitat (e.g., Kriloff et al., 2008). In order to test this hypothesis, cortical dimensions were calculated based on CT scans (Girondot and Laurin, 2003) of long bones of fossil and extant cetartiodactylans, expressed relative to body size, and compared within the sample taxa. This study also tested the hypothesis (Hyp. #2) that increased cortical bone thickness evolved in the common ancestor to Indohyus, cetaceans, and hippopotamids

(Geisler and Theodor, 2009). In order to test this hypothesis, evolution of cortical bone thickness values were reconstructed on a composite of published phylogenies (i.e., Boisserie et al., 2005b, Marcot, 2007, Geisler and Theodor, 2009). Morphology of a common ancestor was determined by using parsimony analyses via ancestral character state reconstruction (Maddison and Maddison, 2009). This study therefore analyzes bone cortical thickness values in a taxonomically broad range of cetartiodactylans, and reconstructs patterns of cortical bone evolution based on analyses utilizing modern comparative methods.

The Apical Ectodermal Ridge

Limb development in vertebrates typically follows a stereotypical pattern (Gilbert,

2006). First, mesenchymal cells proliferate from the somatic layer of the lateral plate mesoderm and somites. These mesenchymal cells continue to proliferate until they create an outward expansion along the body wall, called the limb bud. In terrestrial vertebrates, a total of four limb buds develop. Guiding outgrowth and patterning of this limb bud are two main signaling centers: (1) the apical ectoderm of the limb, which is a specialized region of cells at the limb tip that controls outgrowth and patterning along the proximodistal axis, and (2) the zone of polarizing activity, which regulates patterning along the anteroposterior axis (Gilber, 2006). 7

For decades, it has been assumed that a typical vertebrate apical ectoderm directs limb outgrowth and assumes the shape of a ridge, which is called the apical ectodermal ridge (AER).

In model taxa, such as chicks (e.g., Saunders, 1948; Jurand, 1965; Rubin and Saunders, 1972;

Pizette and Niswander, 1999; Talamillo et al., 2005) and mice (e.g., Jurand, 1965; Lee and Chan,

1991; Talamillo et al., 2005), the AER runs along the tip of the limb bud. Morphogens secreted from the AER function to: (1) direct the survival and proliferation of the mesenchymal cells deep to the AER, (2) maintain a positive feedback loop with the another limb signaling center, the zone of polarizing activity, to pattern the developing limb, and (3) to interact with those signaling molecules that control dorsal and palmar patterning (Gilbert, 2006).

Although an active apical ectoderm typically takes on the shape of an AER in model organisms (i.e., chicks and mice), recent studies in non-model animals (e.g., amphibians, marsupials) have shown a diversity of shapes in the active apical ectoderm. Grounded in the observation that the cetacean apical ectodermal morphology differs from the general mammalian pattern (Richardson and Oelschläger, 2002), Chapter 4 of this dissertation compares limb apical ectodermal morphologies and the associated signaling patterns across representative vertebrates. The morphology and function of the apical ectoderm is initially reviewed, followed by a description of apical ectoderm specific signaling. Finally, this study provides a taxonomically broad comparison of ectodermal morphologies and associated fibroblast growth factor (FGF) expression patterns across vertebrates (bony fish, lungfish, amphibians, squamates, birds and mammals) and presents new morphological and protein signalling findings of the pantropical spotted dolphin (Stenella attenuata).

Hyperphalangy

Although most fossil and living artiodactyls are supported by limbs that are elongated and hooved, modern cetaceans altered their forelimb from a weight-bearing limb to a lift- generating limb shaped like an aerofoil. The forelimbs, or flippers, of cetaceans function in lift, braking, turning, and to stabilize and maintain equilibrium while swimming (Fish and Rohr,

1999). Along with the development of a soft tissue flipper, cetaceans also developed unique bony morphologies of the fingers, namely supernumary phalanges (hyperphalangy) (Figure 3).

Fig. 3. Hyperphalangy in modern cetaceans (modified from Cooper et al., 2007a). Digits are made of multiple phalanges (shown in blue). Metacarpals shown in purple. Most mammals have three or fewer phalanges per digit. Cetaceans are the only mammals that have increased the number of phalanges beyond the plesiomorphic condition of 2/3/3/3/3. Roman numerals indicate digit identity. (A) Killer whale (Orcinus orca), (B) sperm whale (Physeter macrocephalus), (C) North Atlantic right whale (Eubalaena glacialis), and (D) humpback whale (Megaptera novaeangliae). Abbreviations: h, humerus; mc, metacarpals; r, radius; u, ulna.

Several mechanisms have been proposed regarding the developmental mechanisms underlying the acquisition of hyperphalangy in cetaceans (e.g., Caldwell, 2002; Richardson and

Oelschläger, 2002; Fedak and Hall, 2004) but these hypotheses remain untested as the protein signals generated during the formation of additional phalanges are unknown in cetaceans.

Chapter 5 of this dissertation therefore addresses the molecular underpinnings that generate cetacean hyperphalangy by testing the hypothesis (Hyp. #3) of Richardson and Oelschläger

(2002) that the genes, and their associated proteins, responsible for normal digit formation in terrestrial mammals are active over a longer period of developmental time when compared to mammals lacking hyperphalangy.

The Soft Tissue Flipper

Although several studies have described shape variation (e.g. Benke, 1993; Woodward et al., 2006) and embryonic development (e.g., Sedmera et al., 1997; Richardson and

Oelschläger, 2002) among cetacean flippers, no studies have identified the genes, and their associated proteins, that direct development of the cetacean flipper. This dissertation aims to identify the genetic mechanisms creating the cetacean flipper by studying protein expression in embryos and fetuses of the pantropical spotted dolphin (Stenella attenuata).

Embryonic limb formation in mammals begins with a paddle-like handplate followed by initiation of digital development. Programmed cell death (apoptosis) separates digital rays by destroying any soft tissue webbing between the digits (Gañan et al., 1998), and if apoptosis is extensive, a hand with separate digits, much like that of humans, may be created. Limb development research has shown that bone morphogenic proteins (BMP) trigger apoptosis in the tissues between the digits (Gañan et al., 1998). However, the bat Carollia terminates 10

apoptosis by expressing both a BMP-inhibitor called Gremlin (GRE) and fibroblast growth factors

(FGF). Gremlin directly counteracts the apoptotic effects of BMP, while FGFs promote cell survival and proliferation of the interdigital tissues (Weatherbee et al., 2006).

Based on the presence of a soft tissue flipper in living cetaceans, it is evident that the cells between digits do not undergo apoptosis. Chapter 5 therefore also tests the hypothesis

(Hyp. # 4) that, in the dolphin, the signaling that would normally catalyze interdigital cell death is disrupted in a dolphin by the presence of a BMP inhibitor.

CHAPTER II

MORPHOLOGY AND LOCOMOTION OF THE MIDDLE EOCENE RAOELLID INDOHYUS (ARTIODACTYLA: MAMMALIA)

Introduction

Whale Ancestry

Over the past twenty years, extensive field work has yielded well-preserved fossil skeletons of cetaceans (whales, dolphins and porpoises) spanning a range of transitional morphologies (Gingerich et al., 1994, 2001; Thewissen et al., 1994, 2001a,b; Hulbert, 1998;

Hulbert et al., 1998; Bajpai and Thewissen, 2000; Fordyce and de Muizon, 2001; Geisler, 2001;

Uhen, 2004; Thewissen et al., 2007; Gingerich et al., 2009; Thewissen et al., 2009). These fossils document the extraordinary transition in locomotor organs from a deer-like terrestrial form to a streamlined body adapted for aquatic locomotion (Thewissen and Fish, 1997; Buchholtz, 1998;

Bajpai and Thewissen, 2000; Madar et al., 2002; Gingerich et al., 2001; Gingerich, 2003). Ankle bones of the earliest archaic cetaceans (archaeocetes) had a characteristic morphology indicating they belonged to the mammalian order of even-toed ungulates, Artiodactyla (i.e., pigs, camels, giraffes, hippos, deer, cows, sheep, and antelope) (Gingerich et al., 2001;

Thewissen et al., 2001a,b). Analyses also showed that within this order, the closest modern relative of cetaceans is the Hippopotamus (e.g. Shedlock et al., 2000; Geisler and Uhen, 2003;

Boisserie et al., 2005b; Geisler and Uhen, 2005; Price et al., 2005; Geisler et al., 2007; Marcot,

11 12

2007; Geisler and Theodor, 2009). The earliest fossil hippopotamid, Kenyapotamus, was recovered from middle Miocene deposits in Kenya dating from 14-16Ma (Pickford, 1983;

Behrensmeyer et al., 2002 Boisserie et al., 2005a,b), while the earliest archaeocetes were recovered from sediments from approximately 53Ma (Gingerich and Bajpai, 1998). Thus, a 30 million year gap separates fossil hippopotamids and cetaceans, leaving a critical gap in our understanding of the fossil relatives of archaeocetes and, by extention, all cetaceans. Central toward understanding whale evolution was testing for which fossil artiodactyls were the closest relatives to cetaceans. Previous phylogenetic hypotheses of the fossil sister taxon of cetaceans ranged from the entire artiodactyl order (Thewissen et al., 2001 b; Theodor and Foss, 2005), an anthracotheroid clade (Boisserie et al., 2005b), mesonychids (O’Leary, 1998; O’Leary and

Gatesy, 2008), or possibly to the poorly known artiodactyl family Raoellidae that was known only from fragmentary dental material (Geisler and Uhen, 2003; Geisler and Uhen, 2005; Geisler et al., 2007).

Family Raoellidae: The Closest Relatives of Cetaceans

In 2007, skeletal material found in the Jammu and Kashmir regions, located along the border of India and Pakistan allowed for the first time a complete skeletal analysis of the Eocene raoellid Indohyus (Figure 4; Thewissen et al., 2007). Based on a cladistic analysis of morphological characteristics (i.e., presence of an involucrum in the bony ear, and presence of crushing basins in the molars) from multiple disarticulated skeletons, Thewissen et al. (2007) hypothesized that neither anthracotheres nor mesonchyids were the closest relatives to cetaceans, and instead Indohyus was the closest fossil relative to cetaceans. Thewissen et al.,

(2007) further hypothesized Indohyus occupied a freshwater niche based on analyses of limb 13

Fig. 4. Skeletal reconstruction of Indohyus (Thewissen et al., 2007, © Nature). Hashed elements are reconstructed. Scale bar is 10 cm in length.

bone thickness and dental isotopic data. These data suggested that at least one taxon from the family Raoellidae (Indohyus) was sister to cetaceans, and that the common ancestor to cetaceans and raoellids had an aquatic lifestyle as well (Thewissen et al., 2007, see reviews in

Thewissen et al., 2009, Cooper and Thewissen, In Press,). Additional support for this phylogenetic hypothesis came from recent combined molecular and phylogenetic analyses of artiodactyl relationships (Geisler and Theodor, 2009; Spaulding et al., 2009).

Fossils assigned to the family Raoellidae (Sahni et al., 1981) were recovered from middle

Eocene sediments in Pakistan and India (Russell and Zhai, 1987; Thewissen et al., 2001a). Most early fossils recovered were limited to dental material (Pilgrim, 1940; Dehm and Oettingen-

Spielberg, 1958; Ranga Rao, 1971; Sahni and Khare, 1971; Gingerich, 1977; West, 1980; Ranga

Rao and Misra, 1981; Sahni et al., 1981; Kumar and Sahni, 1985; Thewissen et al., 1987;

Thewissen et al., 2001a) which represented several genera, including Indohyus (Ranga Rao,

1971), Khirtharia (Pilgrim, 1940), Kunmunella (Sahni and Khare, 1971), and Metkatius (Kumar 14

and Sahni, 1985). These taxa can be divided into two broad morphological categories: a small, bunodont and slightly lophodont form (Khirtharia and Metkatius), and a larger, lophodont form

(Kunmunella and Indohyus) (Theodor et al., 2007). Indohyus (Ranga Rao, 1971) is currently represented by two taxa, Indohyus indirae (Ranga Rao, 1971) and Indohyus major that can be distinguishable based on size as I. major is twice as large as that of I. indirae (Thewissen et al.,

1987). Thewissen et al. (2007) reported the first skull and postcranial elements of Indohyus that were prepared from an assemblage of disarticulated skeletons. These numerous elements were collected by the Indian geologist A. Ranga Rao from a middle Eocene bone bed extending about

50m at the locality Sindkhatudi in the Kalakot region of Jammu-Kashmir on the Indian side of the

Line of Control. Only a handful of bones were described by Thewissen et al., (2007) however further preparation has unearthed at least 25 individuals (based on cranial elements), warranting a detailed description.

This study documents the skeletal morphology of Indohyus by comparing its postcranial morphologies to representatives of three artiodactyl lineages: (1) the earliest fossil families of cetaceans (pakicetids and ambulocetids), (2) the earliest fossil artiodactyl (Diacodexis), and (3) the modern small-bodied tragulid artiodactyl (Tragulus). Pakicetid cetaceans were recovered from middle Eocene sediments in Pakistan (e.g., West, 1980; Gingerich and Russell, 1981;

Gingerich and Russell, 1990; Thewissen and Hussain, 1998; Thewissen et al., 2001b). Multiple skeletal analyses indicated they are the earliest fossil cetaceans and that they occupied an aquatic niche (Thewissen et al., 1996; Roe et al., 1998; Clementz et al., 2006; Gray et al., 2007).

Diacodexis is the oldest-known diacodexeid artiodactyl with skeletons recovered from early

Eocene sediments (Theodor et al., 2007) of North America (Krishtalka and Stucky, 1985; Rose,

1985; Krishtalka and Stucky, 1986; Gingerich, 1989; Gunnell and Bartels, 2001), Europe (Sudre et 15

al., 1983), and Asia (specimens from Asia were renamed Gujaratia, Thewissen et al., 1983;

Thewissen and Hussain, 1990; Bajpai et al., 2005). This study will focus on those diacodexeid elements recovered from North America (i.e., Rose, 1985). The modern ruminant Tragulus javanicus (lesser mouse deer) is endemic to southern Asia and inhabits heavy lowland forests near shallow bodies of water (Nowak, 1991). This mouse deer bears a primitive hindlimb morphology relative to most ruminants, is one of the smallest modern artiodactyls (Endo et al.,

2006), and is approximately the same body size as Indohyus.

Institutional Abbreviations: AMNH, American Museum of Natural History; CMNH,

Cleveland Museum of Natural History; H-GSP, Howard University-Geological Survey of Pakistan;

GSP-UM, Geological Survey of Pakistan-University of Michigan; RR, A. Ranga Rao Collection of

India; UM-GSP, University of Michigan-Geological Survey of Pakistan; USGS, United State

Geological Survey.

Materials and Methods

Postcranial elements of the raoellid Indohyus described in this study are part of the A.

Ranga Rao (RR) fossil collection (see Appendix), and are currently on temporary loan to the

Anatomy and Neurobiology Department at the Northeastern Ohio Universities College of

Medicine.

Anatomical comparisons were conducted between Indohyus and elements of pakicetids

(Madar, 2007) and ambulocetids (Ambulocetus natans H-GSP 18507 (Thewissen et al., 1996,

Madar et al., 2002) recovered by the Howard Geological Survey of Pakistan. Further comparisons with primitive artiodactyls were conducted with representative fossil diacodexid artiodactyls including published material and casts of Gujaratia pakistanensis (originally 16

described as Diacodexis, Thewissen et al., 1983) and Diacodexis metsiacus (Rose, 1985).

Comparisons with extant artiodactyls includes those artiodactyl taxa that are equivalent in body size to Indohyus, including Tragulus javanicus (CMNH 17917, 21826). Locomotor comparisons were also made with representative cursorial artiodactyls including Odocoileus virginianus

(CMNH 2026, 21888) and O. hemionus (CMNH 4980).

This study employs detailed description of postcranial elements and a discussion of bone microstructural properties based on high resolution CT scans.

Results

Ribs – Ribs of Indohyus (Figure 5 A,B; RR 168, 217, 221, 222, 235, 243, 244, 245) and

Tragulus are gracile compared to the robust ribs of pakicetids (e.g., H-GSP 92060, 96436,

30131). The cranial ribs of Indohyus have tall tubercles that extend far above the neck and head

(RR 243; Fig. 5A), while those of pakicetids (H-GSP 96436) and Tragulus are shorter. Ribs of

Indohyus (RR 168, 245), pakicetids (H-GSP 30131, 92060), and Tragulus all have an equivalent curvature.

Cervical Vertebrae –

Atlas – The atlas (RR 165; Fig. 5C,D) of Indohyus is short and wide with the left transverse process extending caudally compared to the right, unlike other examined taxa. Both transverse processes are thin in Indohyus and Tragulus, while in pakicetids (H-GSP 96021,

96566) these processes are comparatively robust. Ventral surfaces of the transverse processes are deeply excavated in Indohyus, pakicetids, and Tragulus, but only shallowly concave in 17

Fig. 5. Ribs and vertebrae of Indohyus. A, anterior rib (RR 243), B, posterior rib (RR 244). Atlas (RR 165) in C, caudal and D, ventral views. Cervical vertebra 5 (RR 38) in E, dorsal view and F, ventral view. Cervical vertebra 5 (RR 136) in G, cranial, H, caudal, and I, lateral views. Cranial thoracic vertebra (RR 247) in J, cranial and K, caudal views. Caudal thoracic vertebra (RR 25) in L, lateral view. Two articulated terminal thoracic vertebrae in articulation with the cranial-most lumbar vertebra (RR 239), showing shift in angle of inclination of spinous processes in M, left lateral view, N, dorsal view, and O, right lateral view. Scale bar 1 cm in length.

Diacodexis (Rose, 1985). The dorsal arch is longer in its craniocaudal aspect than the ventral arch in Indohyus (RR 86) and pakicetids (the ventral arch of H-GSP 96021, 96566 were originally interpreted as the dorsal arch (Madar, 2007)). The apex of the dorsal arch in Indohyus and pakicetids bears a slightly raised, conical projection along its cranial surface, while this structure is a transverse ridge in Tragulus. Flanking the sides of this dorsal projection are two large lateral vertebral foramina (RR 86) that are present in all examined taxa. A ventral tubercle projects from the caudal aspect of this arch in all examined taxa. The cranial facets for articulation with the occipital condyles of Indohyus form a wide and deep socket surrounding the vertebral canal with facets extending along the lateral surfaces of the canal. Around this vertebral canal,

Tragulus displays two sets of articular facets (dorsolateral and ventrolateral), while Indohyus and pakicetids combine these facets. The caudal facets for articulation with the axis are shallow and slightly concave in Diacodexis (Rose, 1985), Indohyus and pakicetids, but gently convex in

Tragulus. Dimensions of the atlas (RR 165) are: mediolateral length is 42.0 mm; dorsoventral height is 10.7 mm; cranial view neural canal width is 8.7 mm; and cranial view neural canal height is 2.8 mm.

Cervical Vertebra 3 – Cervical vertebra 3 (RR 33) of Indohyus is identified based on the relative size of the body and articular facets. This vertebra displays a dorsal arch that is elongate in the craniocaudal aspect, and bears a longitudinal crest that gives rise to a low spinous process, as in Tragulus (CMNH 21826). Cranial facets extending from the dorsal arch are ventral-facing, slightly convex, and elliptical in shape. In Indohyus, these facets are much larger than in Tragulus. The cranial facet of Indohyus lies on the dorsal surface of the cranial aspect of 19

the dorsal arch. It is flattened and dorsomedial-facing, as in Tragulus. The body is wide and short.

Cervical Vertebra 5 – Cervical vertebra 5 of Indohyus (RR 38 Fig. 5E,F; RR 136 Fig. 5G-I) is identified based on the orientation and length of the transverse processes, and size of the vertebral body. This vertebra has delicate processes, as in Tragulus, and is gracile compared to the robust body and processes of the Ce-5 of pakicetids (H-GSP 92082). The steep laminae of the dorsal arch create a triangular-shaped neural canal (RR 136). The cranial facets are dorsal- facing, flat, and circular in shape, while those of Tragulus are near rectangular and gently convex. In most artiodactyls, the cranial facets are convex. Caudal facets of Indohyus are ventral-lateral facing, circular, and slightly concave, while those of Tragulus are triangular. The cranial surface of the vertebral body is triangular, while the caudal surface is circular in Indohyus

(RR 136) and Tragulus, but the cranial surface is circular in pakicetids (H-GSP 92082). The transverse process bears two projections: a small lateral-pointing projection, and an elongated cranial-pointing projection, the scalene processes. The scalene processes are much larger than those of Tragulus. The body is wide and short. Dimensions of Ce5 (RR 136) are: craniocaudal length is 16.8 mm, mediolateral width is 35.8 mm, dorsoventral height is 30.0mm, cranial view neural canal mediolateral width is 6.9 mm, cranial view neural canal dorsoventral height is 7.0 mm. Dimensions of Ce5 (RR 38) are: craniocaudal length is 27.9 mm, mediolateral width is 31.0 mm, dorsoventral height is 12.8 mm, cranial view neural canal mediolateral width is 6.9 mm, cranial view neural canal dorsoventral height is 4.0 mm.

Cervical Vertebra 7 – Cervical vertebra 7 (Ce7, RR 195) of Indohyus is identified based on the absence of a transverse foramen, and tiny cranial surface and large caudal surface of the body. The spinous process projects vertically in both Indohyus and Tragulus, and this spine is flanked by a pair of caudal facets. The facets are caudally oriented in Indohyus, but caudo- medially oriented in Tragulus. Laminae form a steep triangular roof to the vertebral canal while its base is horizontal. Transverse processes project from lateral aspect of the base of the lamina.

The cranial surface of the vertebral body is half the mediolateral width of the caudal surface, as in Tragulus. The body is wide and short. Dimensions of Ce7 (RR 195) are: craniocaudal length is

8.6 mm, mediolateral width is 24.5 mm, dorsoventral height is 26.6 mm, cranial view neural canal mediolateral width is 10.3 mm, cranial view neural canal dorsoventral height is 6.0 mm.

Thoracic Vertebrae –

Cranial Thoracic Vertebrae: The thoracic vertebrae of Indohyus located in the cranial portion of the thoracic column (i.e. RR 20, 21, 247 Fig. 5J-K) display tall and slender spinous processes, as in Tragulus (CMNH 21826), whereas the spinous processes of pakicetids (H-GSP

96516) are robust and extend higher dorsally relative to both taxa. In Indohyus (RR 247) and pakicetids (H-GSP 96248), caudal articular surfaces located at the base of this spinous process face caudolaterally, while in Tragulus they are ventral-facing. Left and right lamina articulate at a steep angle in Indohyus and pakicetids (H-GSP 96248), creating a triangular vertebral canal, while in Tragulus, these laminae maintain a gentle curvature and create an elliptical vertebral canal. The centrum is widest in the mediolateral direction and shortest in its craniocaudal length (RR 247). Mammillary processes (metapophyses) along the superior root of the transverse processes are cranial-projecting, as in Tragulus. The tips of the transverse processes 21

have a hemispherical, arc-shaped and lateral-facing costal foveae with a prominent rim encircling its cranial, caudal and dorsal surfaces (RR 247). In Tragulus, these foveae assume the shape of a cranial-facing arc. In Diacodexis, these foveae are vertical and lateral facing, ovate, and convex (Rose, 1985). Costal foveae along the cranial aspect of the thoracic vertebrae are small, circular, and ventrally-oriented in Indohyus, cranially-oriented in pakicetids (H-GSP 96410,

96248), and craniolaterally-oriented in Tragulus. Vertebral bodies are elliptical in shape while those of Tragulus and pakicetids (H-GSP 96248, 96410) are heart and triangular-shaped.

Dimensions of a cranial thoracic vertebra (RR 247) are: craniocaudal length is 9.2 mm, mediolateral width is 29.7 mm, dorsoventral height is 42.9 mm, cranial view neural canal mediolateral width is 7.4 mm, cranial view neural canal dorsoventral height is 7.0 mm.

Middle Thoracic Vertebrae: Compared to cranial thoracic vertebrae, middle thoracic vertebrae (RR 85, 137, 242, 292) of Indohyus display a relatively tall spinous process (RR 85, Fig.

6A). This process is reduced in height compared to that of pakicetids (H-GSP 96516). The centrum is roughly equivalent in mediolateral width and craniocaudal length, but is short in dorsoventral height, as in pakicetids (H-GSP 96516). The centrum of middle thoracic vertebrae are relatively shorter than that of Tragulus while the spinous processes are similar in height.

Middle thoracic vertebrae have a ventral keel (RR 137, 247), as in Diacodexis and pakicetids, but unlike the smooth arc-shaped ventral surfaces in Tragulus.

Caudal Thoracic Vertebrae: Those thoracic vertebrae of Indohyus located in the caudal portion of the thoracic column (i.e. RR 25 Fig. 5L, two of three vertebra in RR 239 Fig. 5M-O) display short and triangular-shaped spinous processes that are angled caudally in the penultimate thoracic vertebra and vertically in the terminal caudal vertebrae (RR 239), as in 22

Fig. 6. Vertebrae of Indohyus. Middle thoracic vertebra (RR 85) in A, cranial view, and B, caudal view. Lumbar vertebra (RR 215) in C, cranial view and D, caudal view. Two attached lumbar vertebrae (RR 296) with the bottom vertebra in E, left lateral view, and F, right lateral view. Sacrum in G, right lateral view, and H, left lateral view. Terminal caudal vertebra I, RR 249 in dorsal view, and caudal vertebra J, RR 294 in dorsal view. Scale bar 1 cm in length.

Tragulus, and pakicetids. Flanking the spinous processes are cyclindrical caudal articular surfaces that articulate with the revolute zygapophyses of an adjacent vertebra. These caudal articular surfaces are also cylindrical in Tragulus. Cranial articular surfaces (zygapophyses) are dorsoventrally thick and their dorsal lip locks the caudal articular surface of an adjacent vertebra in place. The vertebral bodies are craniocaudally elongated compared to cranial thoracic vertebrae, as in Tragulus. Dimensions of a caudal thoracic vertebra (RR 25) are: craniocaudal length is 24.5 mm, mediolateral width is 8.0 mm, dorsoventral height is 28.2 mm.

Lumbar Vertebrae– Lumbar vertebrae (RR 239 Fig. 5M-O; RR 296 Fig. 6EF) of Indohyus display thin spinous processes, a craniocaudally long body, and cylindrical, laterocaudal-facing caudal facets, as in Tragulus and pakicetids (H-GSP 98154). The spinous process of pakicetids is relatively robust (H-GSP 98154, 98198). The cranial facets (zygapophyses) are revolute such that they cover the entire dorsal surface of an adjacent caudal facet, as in Tragulus. In pakicetids, these zygapophyses are only partially revolute and display a dorsomedial-facing facet (H-GSP

98198, 98154). A tiny, cylindrical transverse process extends craniolaterally from the body of a single specimen of Indohyus (RR 239, Fig. 5N), but in other cranial lumbar vertebrae (RR 296), the preserved bases of the transverse processes are oriented in the cranioventral direction as in

Tragulus and pakicetids (H-GSP 98198, 98154). The cranial lumbar vertebrae display a craniocaudally elongated body, with relatively short spinous processes (RR 239), while more caudal lumbar vertebrae display dorsoventrally tall spinous processes, and longer transverse processes (RR 215, Fig. 6C-D).

Sacral Vertebra– The sacral vertebrae of Indohyus (RR 156, Fig. 6G-H) and pakicetids (H-

GSP 30251) have spinous processes that are fused at the base of each process, while in Tragulus 24

they are fused along their entire dorsoventral height. S3 bears a pair of hemal process along its ventral and caudal surface, as in S4 of pakicetids, but unlike the ventral processes of Tragulus that lack hemal processes. S2 and S3 display small, ovoid and caudolateral-facing facets that make up the intermediate sacral crest. These processes are elongate, cylindrical, and caudal- facing in pakicetids, while they are fused into a single longitudinal crest in all vertebrae but S4 in

Tragulus. The bodies of these sacral vertebrae are roughly equivalent in craniocaudal length in

Indohyus and pakicetids. Dimensions of the sacrum (RR 156) are: craniocaudal length of S1 is

17.3 mm, craniocaudal length of S2 is 17.8 mm, craniocaudal length of S3 is 16.9 mm, total sacrum craniocaudal length is 57.2 mm, S2 dorsoventral height is 22.8 mm, S3 dorsoventral height is 24.1 mm.

Caudal Vertebrae – The caudal vertebrae of Indohyus decrease in size in the craniocaudal direction, as in all artiodactyls. Processes on the caudal vertebrae are gracile and lack sizeable dorsal projections, unlike the caudal vertebrae of pakicetids (H-GSP 96557) that display large and robust processes that extend further dorsally and farther beyond the vertebral bodies compared to Indohyus (RR 294, Fig. 6J). Tragulus also displays relatively larger processes compared to Indohyus, but they extend farther laterally and are mediolaterally wider, and unlike pakicetids, the processes remain close to the vertebral body and lack a dorsal projection.

Indohyus (RR 169, 294) and Diacodexis (USGS 2352) have relatively large cranial transverse processes and hemal processes compared to pakicetids and Tragulus. The cranial articular processes of Indohyus (RR 294) and pakicetids (H-GSP 96557) are medial facing, those of

Tragulus are dorsal-facing. The mammillary processes of Indohyus and pakicetids are robust and prominent compared to the absence of these processes in Tragulus. 25

Terminal caudal vertebrae are smaller in size, and processes also are smaller relative to vertebral size, as in Diacodexis (USGS 2352, Rose, 1985; Thewissen and Hussain, 1990) and pakicetids (Madar, 2007). The caudal-most vertebrae of Tragulus lack these processes.

Dimensions of the middle caudal vertebra (RR 294) are: proximodistal length is 33.0 mm, mediolateral width is 26.5 mm, dorsoventral height is 14.7 mm. Dimensions of the terminal caudal vertebra (RR 249) are: proximodistal length is 26.0 mm, mediolateral width is 11.4 mm, dorsoventral height is 6.4 mm.

Scapula– The glenoid fossa (RR 155, Fig. 7A-B, RR 263, Fig. 7C) is only slightly concave as in pakicetids (H-GSP 96507, 96532, Madar, 2007), but is flattened compared to the more concave fossae of Diacodexis (Rose, 1985) and Tragulus (CMNH 21826). This fossa is triangular in Indohyus, but is spherical in pakicetids, Diacodexis, and Tragulus. The cranial border of the fossa projects ventrally in Indohyus and pakicetids, while in Diacodexis and Tragulus the cranial border is relatively flattened. Both Indohyus and pakicetids display a prominent supraglenoid tubercle, while in Diacodexis and Tragulus the tubercle is reduced. The coracoid process is bulbous and projects in the medial direction, while in both pakicetids and Tragulus this process is hook-shaped with the tip pointing medially. A prominent scapular spine runs the proximodistal length of the lateral aspect of the scapula. The spine bears a large projection, as in pakicetids (Madar 2007). The distal end of the spine terminates into an acromion process that is directed in the cranial direction at its tip. In contrast, the acromion process of pakicetids,

Diacodexis, and Tragulus are oriented closer to a vertical plane with the scapular spine. Only the distal aspect of the supraspinous fossa of Indohyus is preserved. There are no accessory spines. 26

Fig. 7. Forelimb elements of Indohyus. Left scapula (RR 155) in A, lateral view and B, medial view. Glenoid of the scapula (RR 263) in C, ventral view. Left humerus (RR 149) in D, caudal view and E, cranial view, J, proximal view, and K, distal view. Radius (RR 265) in F, cranial view and G, caudal view. Left ulna (RR 39) in H, medial view. Right ulna (RR 149) in I, medial view. Magnum (RR 250) in L, cranial view and M, distal view. Scale bars are 1 cm in length.

The portion of this fossa contributing to the scapular neck faces laterally, and this fossa is probably larger than the infraspinous fossa. Measurements of the glenoid of RR 155 are as follows: craniocaudal width is 14.9 mm, mediolateral width is 11.4 mm.

Humerus– Humerii of Indohyus (RR 149, Fig. 7D-E, J-K) are elongate and gracile relative to pakicetids, and are similar in their relative thickness as Diacodexis (USGS 2352; Rose, 1985) and Tragulus (CMNH 17917). Proximal humeral epiphyses have relatively small tuberosities for attachment of muscles originating from the scapula. The proximal humeral shaft is broad, but the distal aspect of the shaft is narrow. The distal epiphysis is broad with large radial and ulnar facets, including a deep olecranon fossa.

The proximal epiphysis is dominated by a large subspherical head (Fig. 7D, J), as in pakicetids (H-GSP 30128) but unlike the spherical humeral head of Tragulus (CMNH 17917).

The greater tuberosity flanks the cranial surface of the humeral head, and extends well above the humeral head as in Tragulus, but unlike the reduced tuberosity of pakicetids (Madar, 2007).

The lesser tuberosity flanks the medial surface of the humeral head, is level with the humeral head, and shares part of its surface with the humeral head. The lesser tuberosity of pakicetids and Tragulus is also low, but is separate from the humeral head.

The distal humeral shaft is gracile and elliptical in cross-section. This morphology is similar to that of Diacodexis and Tragulus, but is in strong contrast to pakicetids in which the distal humeral shaft is broader than the proximal aspect (Madar, 2007). Indohyus may display a deltopectoral crest as it has been tentatively identified in a single specimen (RR 145). Diacodexis lacks this crest, but it is present in pakicetids (H-GSP 92042; Madar, 2007) and Tragulus. 28

The distal aspect of the humerus (Fig. 7K) is intermediate in size between the large epiphysis of pakicetids and the narrow epiphysis of Tragulus. The epicondyles are small; the medial one is larger than the lateral one, and both being slightly larger than those of Tragulus.

In contrast, pakicetid epicondyles are massive. As in pakicetids and Tragulus, the medial epicondyle of Indohyus (RR 145) has a depression for attachment of the humeral-ulnar ligaments and a caudal projection for attachment of the flexor muscles, and an intercondylar ridge between the capitulum and epicondyle (RR 145, 149). Indohyus displays a flattened entepicondyle and lacks an epicondylar foramen. The deep olecranon fossa perforates the shaft to form a supratrochlear foramen in Indohyus (RR 145, 149) and Tragulus, but not in pakicetids.

The trochlea and capitulum are preserved but slightly deformed. The trochlea bears a central shallow groove and a medial crest that forms the medial boundary of the olecranon fossa. The capitulum bears a sharp crest on its lateral surface that extends proximal to the trochlea in both

Indohyus and pakicetids, but is distal to the trochlea in Tragulus. Dimensions of the humerus

(RR 149) are as follows: proximodistal length is 110.6 mm, proximal epiphysis craniocaudal width is 10.6 mm, proximal epiphysis mediolateral width is 17.5 mm, midshaft mediolateral width is 8.7 mm, midshaft craniocaudal width is 4.9 mm, distal epiphysis mediolateral width is

16.3 mm, distal epiphysis craniocaudal width is 11.2 mm, olecranon fossa mediolateral width is

5.5 mm, olecranon fossa proximodistal length is 6.4 mm, trochlea mediolateral width is 8.9 mm, capitulum mediolateral width is 2.4 mm.

Radius– The radius of Indohyus (RR 265, Fig. 7F-G) is mediolaterally width and dorsoventrally thin. The distal and cranial aspect bears a longitudinal depression for the extensor muscles of the manus (Fig. 7F, RR 36, 265). Pakicetids only display a slight extensor 29

depression, and in Tragulus (CMNH 17917, 21026) this depression is located along the craniomedial aspect of the distal radius and is angled medially. The caudal aspect of the radius is concave to accommodate the deep flexor muscles in Indohyus (RR 36, 265), pakicetids (H-GSP

30324, 96595, 96061, 96228, 98191; Madar 2007), Diacodexis (USGS 2352; Rose, 1985) and

Tragulus (CMNH 17917, 21826). The distal aspect of the radius (RR 265) bears two facets

(scaphoid facet (medial), and lunate facet (lateral)) for articulation with carpal elements (Fig. 4F) that are reduced compared to those of pakicetids, Diacodexis, and Tragulus. Measurements of the radius (RR 265) are as follows: distal epiphysis mediolateral width is 11.2 mm, distal epiphysis craniocaudal width is 5.9 mm, and length of the extensor depression is 18.4 mm.

Ulna – The ulna of Indohyus is separate from the radius along its entire length. The ulna is mediolaterally compressed, its caudal edge is concave, and it distal aspect part is not preserved.

The proximal aspect of the olecranon of Indohyus (Fig. 7H-I) is more robust in the pakicetids Pakicetus (H-GSP 96057) and Nalacetus (H-GSP 30286). The olecranon of Tragulus

(CMNH 17917) extends further in the proximal direction, its caudal surface lacks a distal- projecting slope, and the apex of the process is capped by large, rounded tuberosities. The semilunar notch of Indohyus (RR 39, 144) is shallow and displays small anconeal and coronoid processes compared to that of pakicetids (H-GSP 30286). Two radial facets are present (RR 144) but fragmentary.

Magnum – A single magnum is preserved (Fig. 7L-M, RR 250) in Indohyus, but has not been recovered in pakicetids (Madar, 2007), and is fused in Tragulus. Morphology of the

Indohyus magnum is therefore compared with the magnum of Ambulocetus natans (H-GSP

18507, Thewissen et al., 1996).. 30

The proximal surface of the magnum bears a large, tear-drop shaped facet for articulation with the lunate, as well as a more medial-placed and smaller facet for articulation with metacarpal IV. The lunate facet has a wide cranial surface and tapered caudal surface (Fig.

7M), as in Ambulocetus. The cranial surface is flat and broad (Fig. 7L), while in Ambulocetus this surface is hourglass-shaped and concave. The medial surface of the Indohyus magnum is broad and concave for articulation with the unciform, while the lateral surface is reduced and bears a small facet for articulation with the trapezoid. The unciform and magnum are fused in

Ambulocetus. The caudal aspect of the bone is extremely tapered and bears a caudodistal projection, as in Ambulocetus. The distal aspect of the magnum bears a tear-drop shaped and convex facet for articulation with metacarpal III. Dimensions of the magnum (RR 250) are: proximodistal length is 5.7 mm, cranial surface mediolateral width is 4.6 mm, caudal surface mediolateral width is 0.9 mm.

Metacarpals – Metacarpals are distinguished from metatarsals by their short length, narrow mediolateral width, and relatively deep shaft. Identification of specific elements is based on shape of the proximal epiphysis.

Metacarpals of Indohyus (Fig. 8A-C) and pakicetids (Madar, 2007) are unfused, while those of Tragulus (CMNH 17917) and other ruminants are fused. Metacarpal I (RR 297, Fig. 8A) is the shortest metacarpal and bears a rectangular proximal facet that is convex in the dorsoplantar axis, but is concave in the mediolateral axis as two prominent ridges border the medial and lateral aspects of the facet. The distal facet of metacarpal I is asymmetrical, and its plantar surface bears a sagittal crest. Metacarpals II and V were not preserved. Central metacarpals (III and IV) are robust in both Indohyus (Fig. 8B,C) and pakicetids. The proximal facet 31

Fig. 8. The manus and pes of Indohyus. Elements of the manus include: A, metacarpal I (RR 297), B, left metacarpal III (RR 228), C, metacarpal IV (RR 270), D, proximal phalanx (RR 274), E, proximal phalanx (RR 273), F, proximal phalanx (RR 201), G, intermediate phalanx I (RR 234), H, intermediate phalanx (RR 117). Elements of the pes include: I, peripheral metatarsal (RR 88), J, right metatarsal III (RR 139), K, left metatarsal IV (RR 225), L, peripheral metatarsal (RR 47), M, proximal phalanx (RR 200), N, proximal phalanx (RR 132), O, proximal phalanx (RR 37), P, proximal phalanx (RR 231), Q, intermediate phalanx (RR 276), R, intermediate phalanx (RR 277), S, intermediate phalanx (RR 125), T, intermediate phalanx (RR 230). Enlarged view of left metacarpal III (RR 271) U, in plantar view, left metacarpal III (RR 228) in V, plantar view and W, dorsal view and right metatarsal III (RR 139) in X, plantar view and Y, dorsal view. Roman numerals indicate digit identity. Scale bars are 1 cm in length.

of metacarpals III and IV is rectangular in shape with its longitudinal axis pointing in the craniocaudal plane and the plantar aspect sloping in the distal direction, creating a convex facet in some specimens of Indohyus (RR 228, 271) and pakicetids (H-GSP 92041, 96424).

This facet spans the entire mediolateral width of the bone in pakicetids (H-GSP 92041) and

Tragulus, but in Indohyus (RR 228) it is flanked along its medial and lateral aspects by prominent ridges that limit mediolateral translation of the unciform and trapezoid. Shafts of these metacarpals gently widen along their distal aspect, and the head is only slightly wider than the shaft in both Indohyus (RR 83, 138, 199, 228, 271) and Tragulus. In contrast, the head of pakicetids is bulbous (H-GSP 96315, 98187). In Indohyus, the dorsal surface of the metacarpal head is large and smooth, while the plantar surface bears a sagittal crest. Dimensions of a metacarpal III (RR 271) are: proximodistal length is 34.1 mm, proximal epiphysis mediolateral width is 4.6 mm, proximal epiphysis craniocaudal depth is 5.5 mm, midshaft mediolateral width is 4.0 mm, midshaft craniocaudal depth is 3.3 mm, distal epiphysis mediolateral width is 5.0 mm, distal epiphysis craniocaudal depth is 6.1 mm. Dimensions of metacarpal III (RR 228) are: proximodistal length is 38.1 mm, proximal epiphysis mediolateral width is 5.3 mm, proximal epiphysis craniocaudal depth is 6.9 mm, midshaft mediolateral width is 4.5 mm, midshaft craniocaudal depth is 3.7 mm, distal epiphysis mediolateral width is 7.7 mm, distal epiphysis craniocaudal depth is 4.4 mm. Dimensions of other metacarpals are listed in Table 1.

Phalanges of the Hand – No terminal phalanges of the manus were found. Phalanges of the manus can be distinguished from pedal phalanges based on their small size.

Only proximal phalanges of the central digits are preserved (Fig. 8D-F). Proximal phalanges are conical in shape, smaller than metacarpals, and the head is much smaller than the proximal epiphysis. These phalanges of Indohyus are similar in shape to those of Diacodexis 33

(AMNH 27787) and gracile compared to those of pakicetids (H-GSP 92103, 96592) and Tragulus.

The proximal epiphysis is hemispherical in shape with a groove along its craniocaudal surface for articulation with the sagittal crest of the metacarpal head. Raised medial and lateral edges of this facet cradle the sagittal crest. These raised edges are more pronounced and sharper in pakicetids (H-GSP 96592) and Tragulus (CMNH 21826) compared to Indohyus (RR 273, Fig. 8E).

The distal facet is trochleated with a central groove extending along the dorsopalmar aspect of the facet. Dimensions of a proximal phalanx (RR 281) are: proximodistal length is 21.5 mm, mediolateral width of proximal facet is 5.0 mm, craniocaudal depth of the proximal facet is 6.3 mm, midshaft mediolateral width is 2.6 mm, midshaft craniocaudal width is 3.7 mm, mediolateral width of distal facet is 2.8 mm, craniocaudal depth of the distal facet is 3.2 mm.

Dimensions of other elements of the manus are listed in Table 1.

Only intermediate phalanges of digits I and the central digits are preserved for Indohyus.

Intermediate phalanges are conical in shape, shorter than the proximal phalanges, and the head is considerably smaller than the proximal epiphysis. Compared to the central intermediate phalanges (RR 99, 117), those of digit I (RR 233, 234) are considerably smaller and bear a relatively wider head (Fig. 5G). The central intermediate phalanges of Indohyus bear a proximal epiphysis that is relatively wide compared that of Diacodexis (AMNH 27787). Central intermediate phalanges are also narrow and gracile compared to the robust and rectangular shaped phalanges of a pakicetid (H-GSP 96177) and Tragulus (CMNH 17917). The proximal facet bears two mediolateral concavities separated by a crest for articulation with the trochlea of the proximal phalanx. The distal facet bears a trochlea. Dimensions of the intermediate phalanx (RR 117) are: proximodistal length is 14.5 mm (specimen lacks distal epiphysis), mediolateral width of proximal facet is 5.0 mm, craniocaudal depth of the proximal facet is 34

Table 1. Indohyus Metapodial and Phalangeal Measurements. Lengths Marked with an * Indicate an Incomplete Length.

Bone Specimen # Length (mm) Manus Metacarpal, peripheral RR 69 25.5* Metacarpal, peripheral RR 83 36.7 Metacarpal I RR 297 22.3* Metacarpal III, left RR 138 31.7 Metacarpal III, left RR 228 38.1 Metacarpal III, left RR 271 34.0 Metacarpal IV, left RR 270 35.5 Phalanx, central, 1 RR 273 19.0 Phalanx, central, 1 RR 274 18.0 Phalanx, central, 1 RR 281 21.4 Phalanx, 2 RR 99 12.9* Phalanx, 2 RR 117 14.0* Phalanx, 2 (juvenile) RR 179 14.4* Phalanx, digit I, 2 RR 233 9.3* Phalanx, digit I, 2 RR 234 10.2

Pes Metatarsal, peripheral RR 47 26.4* Metatarsal, central RR 76 22.2* Metatarsal, peripheral RR 88 55.4 Metatarsal, central RR 291 14.1* Metatarsal, central RR 105 31.2* Metatarsal III, right RR 139 57.6 Metatarsal, central RR 158 68.3 Metatarsal, peripheral RR 199 30.1 Metatarsal IV, left RR 225 63.4 Phalanx, central, 1 RR 37 35.1 Phalanx, central, 1 RR 118 26.7* Phalanx, central, 1 RR 126 23.8 Phalanx, central, 1 RR 132 31.2 Phalanx, peripheral 1 RR 91 25.9 Phalanx, peripheral, 1 RR 200 27.6 Phalanx, peripheral, 1 RR 231 26.2 Phalanx, peripheral, 1 RR 236 21.6 Phalanx, peripheral, 1 RR 275 29.2 Phalanx, peripheral, 1 RR 293 18.4* Phalanx, 2 RR 19 20.8 Phalanx, 2 RR 34 14.2 Phalanx, 2 RR 114 17.1 Phalanx, 2 RR 125 17.2 Phalanx, 2 RR 181 19.7 Phalanx, 2 RR 230 17.7 Phalanx, 2 RR 276 18.6 Phalanx, 2 RR 277 18.2 35

3.8 mm, midshaft mediolateral width is 3.4 mm, midshaft craniocaudal width is 2.1 mm.

Dimensions of other manual elements are listed in Table 1.

Pelvis –The innominate of Indohyus (Fig. 9A,B) most closely resembles that of the fossil cetacean

Nalacetus (H-GSP 30395; Madar, 2007). The ilium is generally flat with a robust neck and its cranial extension bears a wide blade (RR 44, 257). The medial surface of this blade bears an irregular surface indicating the presence of a sacroiliac joint (RR 43, 162). The iliac neck of

Indohyus is similar to that of Nalacetus (H-GSP 30395) in that it is flattened in the mediolateral plane (RR 44, 256, 259) while the iliac necks of Diacodexis and Tragulus are flattened in the dorsoventral plane. Indohyus bears a prominent anterior inferior iliac spine, for the origin for the rectus femoris muscle (Rose, 1985) just cranial to the acetabulum (RR 43, 256, 257, 259) that is small compared to that of Nalacetus, similar to that of Diacodexis, and larger than the tiny tuberosity of Tragulus. A depression is found adjacent to the anterior inferior iliac spine (RR

257). The iliac blades are longer than deep and bear concave lateral (gluteal) and convex

(auricular) surfaces (RR 257), as in the fossil cetacean Pakicetus (H-GSP 30213; Madar, 2007).

The ischia of Indohyus (RR 256, Fig. 9A) and Nalacetus (H-GSP 30395) are robust and elliptically-shaped in cross-section, compared to the gracile and triangular cross-sectioned ischium of Tragulus (CMNH 17917). A broad dorsomedial extension of the ischial body is only present in a single specimen of Indohyus (RR 257, Fig. 9B) and displays a series of parallel, caudally-oriented rugosities. The ischial tuberosity is robust and prominent in Indohyus (RR 257) and greater in size compared to that of Tragulus. The dorsocaudal aspect of the ischium displays a sharp dorsal curvature, the tip of which is the ischial tuberosity, in Indohyus (RR 256) and Nalacetus, whereas this curvature is absent in Diacodexis and Tragulus. 36

Fig. 9. Pelvic limb elements of Indohyus. A, Right innominate (RR 256) in lateral view, B, right innominate (RR 257) in dorsal view, C, right femur (RR 101) in cranial view and D, in caudal view, E, right femur (RR 42) in caudal view, F, right femur (RR 133) in cranial view, G, proximal fragment of right tibia (RR 84) in cranial view, H, right tibial plateau (RR 22) in proximal view, I, juvenile tibial shaft (RR 143) in cranial view, J, tibia (RR 301), distal view, K, proximal aspect of the fibula (RR 96), L, distal aspect of the fibula (RR 295) in lateral view, M, patella (RR 269) in cranial view. Scale bar is 1 cm in length. 37

Acetabulae of Indohyus and the fossil cetaceans Nalacetus and Pakicetus are shallow and elliptical in outline with an elongated craniocaudal axis (RR 43, 146, 256, 259) while those of

Diacodexis and Tragulus are deeper and spherical. Indohyus, Nalacetus, and Pakicetus display a prominent dorsocaudal rim, while the ventral rim is flattened, unlike the wholly sharp-rimmed acetabulum of Tragulus.

The obturator foramen (Fig. 9A) is enlarged in Indohyus, Nalacetus, Pakicetus

(Thewissen et al., 2009), and Diacodexis relative to that of Tragulus. Relative to the other compared taxa, the obturator foramen of Tragulus is reduced in the craniocaudal direction as the caudal surface of this foramen is positioned closer to the acetabulum.

Innominates of Indohyus can be separated into gracile (Fig. 9A, RR 43, 44, 162, 256) and robust (Fig., 6B, RR 146, 257, 258, 259) forms. The gracile forms are thinner and display two tuberosities separated by a deep depression for a muscle attachment, located cranial to the acetabulum. Robust forms only display one of these tuberosities, the anterior inferior iliac spine. Gracile forms also display little to no dorsal shelf to the ilium and ischium, while robust forms display medial-projecting, shelf-like, dorsal surface that is thick and marked rugosities consistent with muscle attachment. These differences in morphology could be attributed to sexual dimorphism, but further specimens are needed to confirm this hypothesis.

Dimensions of the innominate (RR 256) are: acetabulum craniocaudal length is 14.6 mm, acetabulum dorsoventral width is 13.9 mm, obturator foramen craniocaudal length is 37.7 mm, obturator foramen dorsoventral width is ~19.0 mm.

Femur –The femoral head (Fig. 9C-E) is oval in shape with an elongated mediolateral axis

(RR 42, 101, 264), as in Tragulus, and Pakicetus (H-GSP 30333, Madar, 2007). The greater 38

trochanter is prominent, extends slightly proximal to the proximal surface of the femoral head, and its intertrochanteric crest conntects with the triangular-shaped lesser trochanter (RR 42, 89,

101, 154, 161, 267), like that of Diacodexis, Tragulus, and Pakicetus (H-GSP 30345). A deep trochanteric fossa is present in Indohyus (RR 42, 101, 154). The greater trochanter of Indohyus is larger than the lesser trochanter (RR 42, 89, 101, 161, 267) as in Diacodexis and Tragulus. The greater trochanter relative size is smaller in Indohyus compared to pakicetids, but is equivalent in size to that of Tragulus. The lesser trochanter projects is a medial projection of the proximal shaft in Indohyus and pakicetids, but a medial-caudal projection in Tragulus. The third trochanter projects from approximately 1/3 of the femoral shaft length and is smaller than that of Pakicetus (H-GSP 30345) but greater than that of Diacodexis and Tragulus.

The femoral shaft (Fig. 9C-E) of Indohyus is thick compared to that of Diacodexis (USGS

2352; Rose, 1985) and Tragulus (CMNH 17917). The shaft is elliptical in transverse section and is wider along its mediolateral aspect rather than its craniocaudal aspect (RR 42, 154, 161, 267), while in Diacodexis the proportions are reversed, and in Tragulus the shaft is circular in cross- section.

The distal end of the femur of Indohyus (RR 101 Fig. 9D,E; RR 133 Fig. 9F) is widened compared to that of Diacodexis and Tragulus. The patellar groove is shallow with slightly raised margins (RR 101), unlike the sharp and deepened grooves of Diacodexis and Tragulus. The medial epicondyle is a raised globular projection in Indohyus (RR 101) and Diacodexis, and a crest in Tragulus. The lateral epicondyle is a raised rugose area in Indohyus (RR 101), Diacodexis, and Tragulus. The distal extent of these condyles are even in a single specimen of Indohyus (RR

101), but the lateral condyle extents distally farther than the medial condyle in another specimen (RR 133), Diacodexis, and Tragulus. The lateral condyle is wider than the medial 39

condyle as in Diacodexis and Tragulus. A distinct round pit, the extensor fossa, is located at the junction of the lateral ridge of the patellar groove and the lateral epicondyle. This fossa is reduced in Indohyus (RR 101), and is much deeper in Diacodexis and Tragulus. The extensor fossa serves as the attachment for the m. extensor digitorum longus. Dimensions of the femur of Indohyus (RR 101) are as follows: proximodistal length is 120.1 mm, midshaft mediolateral width 11.7 mm, midshaft craniocaudal width 7.3 mm, femoral head mediolateral diameter is

15.0 mm, patellar groove length is 26.7 mm, patellar groove width is 9.7 mm.

Patella– The patella of Indohyus (RR 234; RR 269, Fig. 9M) has a convex cranial surface and concave caudal surface as in Diacodexis (USGS 2352; Rose, 1985). The proximal surface is wider than the distal surface along both the craniocaudal and mediolateral dimensions. The caudal surface displays two longitudinal depressions located on its medial and lateral aspects.

Patellar dimensions (RR 269) are as follows: mediolateral width is 11.3 mm, proximodistal length is 16.5 mm, maximum craniocaudal width (along proximal aspect of patella) is 6.7 mm, and minimum craniocaudal width (along distal aspect of patella) is 2.6 mm.

Tibia--The tibia (Fig. 9G-J) is triangular in transverse section (RR 46, 84). The proximal articular surface (Fig. 9G,H) bears a large lateral condyle and a smaller medial condyle for articulation with the femur (RR 22, 84), as in Diacodexis (USGS 2352; Rose, 1985) and Tragulus

(CMNH 17917). In Pakicetus (H-GSP 30357; the lateral condyle is relatively larger than that of

Indohyus. In Indohyus, these condyles are convex in the craniocaudal direction and concave in transverse section (RR 22). Indohyus (RR 22, Fig. 9H) displays small central spines that separate the condyles, unlike the large and prominent spines of Pakicetus (H-GSP 30357) and Tragulus.

The lateral condyle bears a deep sulcus along its cranial and medial-most border that 40

accommodates the long digital extensors in Indohyus (RR 84), Pakicetus, Diacodexis, and

Tragulus. Below the caudal and distal-most aspect of the lateral condyle lies an ovoid depression for the fibular head as in all examined taxa. The tibial tuberosity (Fig. 9G) is more robust in its mediolateral width than in Diacodexis and Tragulus. Continuing distal from this tuberosity is the tibial crest (RR 143, Fig. 9I) which forms the medial border of the fossa for the tibialis cranialis muscle. This fossa is continues approximately 20-25% down the tibial shaft in

Indohyus (RR 143), Diacodexis, Tragulus, but is enlarged to encompass approximately 40% of the shaft in pakicetids (H-GSP 30357). The popliteal fossa is deepest in the tibia of Indohyus (RR 84) and Pakicetus (H-GSP 30357), whereas in Diacodexis and Tragulus this fossa is shallow. The astragalar facet of the tibia is rectangular in shape and is longest in the craniocaudal plane (RR

301, Fig. 9J), while this facet is approximately square in pakicetids (H-GSP 30357), and Tragulus.

A sagittal crest divides a larger lateral trochlear facet from a narrow medial trochlear facet in

Indohyus and pakicetids. Tragulus bears a larger medial facet. The lateral and distal surface of the tibia displays a triangular concavity for articulation with the fibula that is relatively larger in pakicetids (H-GSP 30357) and absent in Tragulus. The medial malleolus is a blunt distal projection that bears a flattened articular surface, whereas in Tragulus, the malleolus curves laterally and the facet is saddle-shaped. The cranial and caudal surfaces of the distal epiphysis display two small, triangular projections. In Tragulus, these projections are comparatively large and curve toward the center of the epiphysis. Dimensions of the proximal aspect of the tibia (RR

84) are as follows: tibial plateau mediolateral width is 27.5 mm, craniocaudal width is 24.0 mm; tibial crest mediolateral width is 5.3 mm; shaft mediolateral width is 12.9 mm; craniocaudal width is 14.3 mm. Dimension of the distal aspect of the tibia are: distal epiphysis mediolateral 41

width is 10.0 mm, craniocaudal length is 13.2 mm, and proximodistal length of the medial malleolus is 4.6 mm from the center of the distal epiphysis.

Fibula -- Based on the presence of several incomplete fibulae and fibular facets on the tibia and astragalus of Indohyus, the fibula was complete, lacked proximal tibial fusion, and was a slender rod relative to the tibia, as in pakicetids (Madar, 2007), and Diacodexis (USGS 2352,

Rose, 1985). Presence of a fibula in Tragulus is variable with some individuals retaining only vestiges that are fused to the tibia (Rose, 1985). The elliptically-shaped proximal articular surface (Fig. 9K) slopes with a distal-projecting medial surface and fits into the fibular facet of the tibia (RR 84). The distal end widens (Fig. 9L) to a large tear-drop shaped facet. Dimensions of the proximal aspect of the ulna (RR 96) are as follows: shaft mediolateral width is 4.1 mm, shaft craniocaudal width is 2.8 mm, proximal facet mediolateral width is 6.6 mm, and proximal facet craniocaudal width is 2.7 mm.

Astragalus– The astragalus (Fig. 10A-C) is gracile, like that of pakicetids (H-GSP 30153, 98148,

Madar, 2007), but unlike the more robust astragalus of Tragulus (CMNH 21826). The astragalar neck of Indohyus and pakicetids is elongated along its proximodistal axis, but reduced in

Tragulus. As in all artiodactyls, the astragalar head is trochleated. Both trochleae are oriented in a vertical plane in Indohyus (RR 129, 213, 224, 246, 290), Tragulus, and some pakicetids (H-

GSP 98148), but trochleae are offset in Diacodexis (Rose, 1985) and another specimen of pakicetids (H-GSP 30153). The proximal aspect of the astragalus is deeply trochleated while the distal surface bears a shallower trochlea (RR 129, 213, 224, 246), similar to pakicetids (Madar,

2007), Tragulus (CMNH 21826), and Diacodexis (Thewissen and Hussain, 1990).

Fig. 10. Tarsal elements of Indohyus. A, right astragalus (RR 224) in cranial view, B, anterior view, and C, caudal view. D, left calcaneus (RR 170) in lateral view, E, cranial view, and F, medial view. G, left cuboid (RR 214) in caudal view, H, proximal view, and I, lateral view. Scale bar is 1 cm in length.

The lateral surface of the lateral condyle of the tibial trochlea is concave but bears two ectal facets for articulation with the calcaneus, and a fibular facet. The ectal facet (Fig. 10A-C) is elliptical in Indohyus (RR 35, 224), pakicetids (H-GSP 31053, 98148), and Diacodexis (Gingerich et al., 2001), but conical in Tragulus (CMNH 21826). The ectal facet is plantar-facing in Indohyus

(RR 35, 224) and pakicetids (H-GSP 30153, 98149, Madar, 2007). The plantar ectal facet (Madar,

2007), lies along the cranial and distal aspect of the lateral condyle and is only found in Indohyus

(RR 35, 224, 246) and pakicetids (Madar, 2007). The lateral surface of the lateral condyle also has a sharpened rim and slightly concave surface for articulation with the distal aspect of the fibula.

The distal trochlea of the astragalus bears a lateral facet for articulation with the cuboid.

This cuboidal facet is narrower in its mediolateral aspect compared to the medial condyle

(navicular facet) in all examined taxa and the facet forms a right angle with the parasagittal plane. A sharp crest separates these facets.

The sustentacular facet (Fig. 10C), located along the caudal aspect of the astragalar neck, is saddle shaped with a slightly raised lateral ridge in all examined taxa. This facet covers

78% of mediolateral width of the astragalus in Tragulus, but is mediolaterally reduced in

Indohyus (RR 224) and Gujaratia pakistanensis (sustentacular facet width is 57% of astragalar width) and pakicetids (43-50% of the astragalar width, H-GSP 30153, 98148). The lateral edge of the sustentacular facet lies in line with the lateral margin of the tibial trochlea, and the distal aspect of this facet is completely separated from the cuboid and navicular facets. Dimensions of the astragalus (RR 224) are: mediolateral width is 10.8 mm, proximodistal length is 20.3 mm, craniocaudal width is 12.3 mm, proximodistal length of neck is 6.2 mm, and the sustentacular facet is 6.8 mm in mediolateral width. 44

Calcaneus – The calcaneus (Fig. 10D-E) of Indohyus is very narrow in its mediolateral aspect, like that of Diacodexis (USGS 2352, Rose, 1985) and pakicetids (Madar, 2007). A deep fossa (for attachment of calcaneofibular ligaments) excavates the lateral surface, and encompasses the distal two-thirds of the lateral surface in Indohyus (RR 164, 170), the distal half in Diacodexis, and may extend the entire length of the lateral surface in pakicetids (Madar,

2007). The tuber calcis, at the proximal apex of the calcaneus, displays a taller medial tuberosity and flattened lateral tuberosity (RR 167, 170) as in Diacodexis, Tragulus, and the pakicetid

Ichthyolestes (H-GSP 92063). The sustentaculum tali, which articulates with the sustentacular facet of the astragalus bears a reduced ovoid facet in Indohyus (RR 164, 170) compared to the robust facet in Tragulus. The fibular facet along the cranial aspect of the calcaneus is rounded and bears a prominent hemispherical facet for articulation with the fibula in Indohyus (RR 164), pakicetids (H-GSP 30246, 92063, 96612). The cuboidal facet, along the caudal and distalmost aspect of the bone, is elliptical in shape as in Tragulus (CMNH 21826), but is reduced in the craniocaudal dimension compared to some pakicetids (H-GSP 30246, 96612). This facet forms a steep angle at the distal aspect of the calcaneus in all examined taxa. Dimensions of the calcaneus (RR 167) are: proximal distal length is 41.0 mm, mediolateral width at tuber calcis is

7.44 mm, mediolateral width at base of calcaneal tuber is 5.8 mm, proximodistal length of calcaneal tuber is 22.3 mm, and the medial extent of sustentaculum tali from calcaneal tuber is

5.3 mm.

Cuboid –The cuboid (RR 214, Fig. 10H-I) is hourglass shaped in dorsal view as in pakicetids (H-GSP 30174, Madar 2007). The proximal surface bears two facets, a astragalar facet, and a larger calcaneal facet. Only the cranial portion of the astrastragar facet is 45

preserved. The calcaneal facet is steeply sloped distally and is enlarged along its mediolateral aspect as in pakicetids and Tragulus (CMNH 21826). The medial aspect bears two facets for articulation with the navicular tarsal element. Along the cranial half of the medial surface is a broken prominence for articulation with the navicular, as in pakicetids (H-GSP 30174, Madar

2007). Along the caudal aspect of the medial surface is an additional convex facet for articulation with the navicular, but this facet is larger than that of pakicetids. Encompassing most of the cranial and distal borders is a large horizontal facet for articulation with the proximal aspect of metatarsal IV, as in all artiodactyls and pakicetids. This facet is roughly triangle-shaped in Indohyus and pakicetids (H-GSP 30174), but is arc-shaped in Tragulus. A smaller facet for metatarsal V lies along the distal and lateral surface of the cuboid, just lateral to the facet for metatarsal I, although it is tiny in Indohyus and Tragulus, but prominent in pakicetids. The caudal and ventral corner of the cuboid extends distally to form a plantar process that articulates with the caudal and proximal aspects of metatarsal IV, as in Tragulus, but this process is reduced compared to the elongated process in pakicetids (Madar, 2007). The caudal aspect of this process is roughly rectangular shaped in Indohyus, but v-shaped in pakicetids and Tragulus. Cuboid (RR 214) dimensions are: proximodistal length is 10.5 mm, mediolateral width is 7.6 mm, craniocaudal width is 10.6 mm.

Metatarsals – Metatarsals are distinguished from metacarpals by their great length, great mediolateral width, narrow dorsoplantar depth, and prominent plantar tubercle.

Metatarsals of Indohyus (Fig. 8I-J), Diacodexis (Rose, 1985), and pakicetids (Madar,

2007) are unfused, whereas these elements are fused in Tragulus (CMNH 21826). Compared to pakicetids and Diacodexis, the metatarsals of Indohyus are more gracile, but are robust in 46

comparison to the metatarsals of Tragulus. Presence of metatarsal I is unknown, although it could be that RR 199 represents a deformed metatarsal I. The head of central metatarsals (RR

105, 108, 139, 225, 291) is symmetrical along the base of the shaft, while that of peripheral metatarsals (RR 47, 88, 199) is asymmetrical and steeply sloped along its mediolateral aspect.

The proximal facets of the central metatarsals (III and IV) are concave and rectangular to triangular in shape (RR 158, 225) like the rectangular facets of Tragulus (CMNH 21826), and triangular facets of pakicetids. The plantar tubercle is reduced in the central metatarsals

Indohyus (RR 139, 225) but enlarged in pakicetids (H-GSP 30405, 30417, 96299). The head of the distal epiphysis is the widest part of the central metatarsals, is rounded and convex, and their plantar surface bears a sagittal crest, as in Diacodexis (USGS 2352) and Tragulus (CMNH

21826). This morphology is in strong contrast to the bulbous distal shafts and heads of pakicetids that are relatively larger in both the mediolateral and craniocaudal dimensions.

Dimensions of metatarsal IV (RR 225) are: proximodistal length is 62.2 mm, proximal epiphysis mediolateral width is 6.1 mm, midshaft mediolateral width is 6.3 mm, midshaft craniocaudal width is 4.7 mm, distal head mediolateral width is 10.5 mm. Dimensions of the a central metatarsal (RR 158) are: proximodistal length is 57.8 mm, proximal epiphysis mediolateral width is 5.9 mm, proximal epiphysis craniocaudal depth is 5.9 mm, midshaft mediolateral width is 6.8 mm, midshaft craniocaudal depth is 3.2 mm, distal head mediolateral width is 9.9 mm, distal head craniocaudal depth is 5.9 mm. Additional lengths of other metatarsals are listed in Table 1.

Pedal Phalanges – No terminal phalanges of the pes were found for Indohyus.

Phalanges are tentatively identified as pedal phalanges based on their large relative size compared to phalanges of the manus. Proximal and intermediate phalanges are distinguished 47

based on morphology of their proximal epiphyses. Central phalanges are distinguished from peripheral phalanges based on their larger relative size, and symmetrical facets.

Proximal phalanges of the pes (Fig. 8M-P) are conical in shape with a wide proximal epiphysis, narrow distal epiphysis, and a midshaft flattened in the dorsoventral plane. The proximal facet is hemispherical in shape and the facet surface is concave with a central groove for articulation with the sagittal crest of the distal metacarpal. The distal facet is trochleated for articulation with intermediate phalanges. In central proximal phalanges (RR 118, 126, 132), both proximal and distal facets are symmetrical about the longitudinal axis of the bone, whereas in phalanges of peripheral digits (RR 91, 200, 231, 236, 275, 293) these facets are asymmetrical.

The plantar surface is bears a deep groove along the proximal shaft. These phalanges are more robust than those of Diacodexis (USGS 2352), but gracile compared to pakicetids (H-GSP 96593) and Tragulus (CMNH 17917, 21826). The distal facet is trochleated. Lengths of these phalanges are listed in Table 1.

Like proximal phalanges, the intermediate phalanges of the pes (Fig. 8Q-T) are conical in shape and have a distal trochlea, but the intermediate phalanges bear two grooves on their proximal facet for articulation with the distal trochleated facet of the proximal phalanx.

Intermediate phalanges are short and narrow compared to proximal phalanges. Lengths of these phalanges are listed in Table 1.

Discussion

Locomotion

From the above description it can be seen that the skeleton of Indohyus exhibits gross anatomical specializations typical of a cursorial artiodactyl (Rose, 1985). The lumbar vertebrae 48

of Indohyus have revolute zygapophyses that prevent torsion. The appendicular elements are thin and elongated with the hindlimbs longer than the forelimbs. The elbow is stabilized as it lacks the ability to supinate and pronate. The ulna is very thin. The cubital joint is also stabilized as it only functions in flexion and extension and does not allow mediolateral translation. Greater extension of the ulna is afforded by the presence of a humeral supratrochlear notch. As in all artiodactyls, the tarsus functions in flexion and extension as mediolateral joint excursions are limited by bony articulations between the distal aspect of the tibia, a double-trochleated astragalus, and a tight astragalus-calcaneus articulation that affords limited mediolateral rotation. Although Indohyus displays these bone morphologies consistent with terrestrial locomotion, analyses of bone cross-sectional geometries indicate the skeleton was loaded with an exceptional amount of mineral. Presence of hyperostotic bones is indicative of an aquatic lifestyle, and we show here that Indohyus was an effective wader, but was a poor swimmer compared to the earliest fossil whales, pakicetids.

Vertebral Mobility

Indohyus had robust axial musculature and habitually held its head well above the thoracolumbar column. Atlas vertebrae of Indohyus and pakicetids have deeply excavated ventral surfaces, indicating attachment of robust axial musculature. Articulation between terminal cervical vertebrae of Tragulus and Indohyus creates a strong vertical curvature in the neck. Accordingly, the occipital condyles of these taxa are caudoventrally oriented and the cranial facets of the atlas suggest a craniodorsal angle. While Indohyus held its head well above the rest of the spine, pakicetids oriented their heads parallel with the thoracolumbar column, as indicated by morphology of the occipital condyles and cranial facets of the atlas. Cervical 49

vertebra thickness in Indohyus and other artiodactyls is thin compared to the thickened cortical layers in the vertebrae of pakicetids (Madar, 2007), indicating that Indohyus had relatively less robust axial musculature compared to pakicetids. Dimensions of thoracic vertebrae indicate that Indohyus displayed greater axial musculature than the terrestrial artiodactyl Tragulus, but less than that of pakicetids. Thoracic vertebrae of pakicetids are massive with elongated and robust processes, indicating the presence of robust axial musculature.

The lumbar vertebrae of Indohyus and most modern artiodactyls have a limited range of mobility in comparison to pakicetids. In most modern terrestrial artiodactyl taxa, the lumbar spine is stiffened by the presence of revolute zygapophyses that interlock vertebrae and impede torsion or rotation of individual vertebrae about the longitudinal axis of the column. These zygapophyses are present in all taxa examined here, but pakicetids have zygapophyses that only partially envelop the caudal articular processes of an adjacent vertebra. The zygapophyses of pakicetids encase the ventral and lateral surfaces of the caudal articular surfaces, but only partially encase the dorsal surface, thus allowing limited rotation and greater flexion and extension of the vertebral column, which may have assisted in swimming maneuvers.

Caudal vertebrae of Indohyus stabilized the body during swimming maneuvers (see

Hindlimb Mobility discussion), unlike the tail of pakicetids that acted as a propulsor via dorsoventral undulations. The caudal vertebrae of the terrestrial artiodactyls Diacodexis (Rose,

1985) and Tragulus are gracile in comparison to those of Indohyus. Tragulus propels itself by paddling its limbs, and does not utilize its tail for generation of thurst; the tail may act to stabilize the body during paddling maneuvers. Although the tail length of pakicetids is unknown, Madar (2007) noted that the robust size and enlarged processes of the caudal vertebrae were indicative of exceptionally strong epaxial musculature, and concluded that the 50

tail acted as a propulsor via dorsoventral undulations, along with the hindlimbs, during swimming manueuvers. Caudal vertebrae of Indohyus are gracile compared to those of pakicetids, but robust compared to those of the terrestrial artiodactyls Tragulus and Diacodexis.

Tail movements of Indohyus were probably supported by a thick layer of epaxial musculature that helped stabilize caudal vertebral joints while paddling of the hindlimbs generated thrust.

Forelimb Mobility

The glenohumeral joint of Indohyus functioned primarily in facilitating flexion and extension of the humerus, with only a minute amount of mediolateral movement, as in

Odocoileus. As in most artiodactyls, the humerus of Indohyus was habitually held in an extended position such that its distal aspect pointed caudally. In comparison, Pakicetus displayed more mobility in flexion-extension and mediolateral rotation. Relative to these taxa, the joint of Tragulus allowed for the greatest mobility in all directions as the humeral head is relatively larger and displays a greater surface area for the glenoid.

The cubital joint of Indohyus functioned primarily in flexion and extension, but lacked the ability to supinate and pronate, as in Diacodexis (Rose, 1985), pakicetids (Madar, 2007),

Tragulus and Odocoileus. Both Indohyus and Diacodexis have a supratrochlear foramen through the distal aspect of the humerus, which allows greater maximum extension of the ulna. This foramen is absent in pakicetids, Tragulus and Odocoileus.

Metacarpals of Indohyus and pakicetids are unfused, but display different ranges of motion in the metacarpophalangeal joints. Most modern ruminants display fused metacarpals

(e.g, Odocoileus and Tragulus), while in primitive artiodactyls (i.e., Gujaratia, Thewissen and

Hussain, 1990) the metacarpals are unfused. Like primitive artiodactyls, Indohyus and pakicetids 51

(Madar, 2007) display unfused metacarpals that were capable of individual movements, which may have allowed greater control over manus movements at the expense of stability in the distal limb. Based on morphology of the distal articular facet of the metacarpals, Indohyus displayed a greater range of mobility in extension of the metacarpophalangeal joint compared to its ability to flex the joint. Pakicetids displayed enlarge enlarged distal facets on the metacarpals, and were able to equally extend and flex the metacarpophalangeal joint such that the joint could rotate 180 degrees (Madar, 2007). Odocoileus displays a similar morphology to pakicetids and Indohyus in their ability to extend the proximal phalanx, but are limited in flexion compared to Indohyus. This ability for extension is lacking in Tragulus as the distal metacarpal mostly facilitates flexion of the proximal phalanx.

In Indohyus, the first interphalangeal joint was habitually flexed, as in pakicetids. The distal trochlea of intermediate phalanges bears only a limited articular surface along the dorsal aspect, but proximodistally elongated facets along the plantar surface, as in pakicetids.

Movement of this joint is restricted in Tragulus and Odocoileus such that only a few degrees of flexion or extension are possible.

Forelimb function of Indohyus was probably restricted to flexion and extension in all joints. In the proximal joints (i.e. glenohumeral and cubital), Indohyus probably moved like a typical cursorial artiodactyl. Rather than stabilizing the metacarpophalangeal and proximal interphalangeal joints along the vertical plane as in modern cursors, Indohyus displays a surprising amount of extension in the metacarpophalangeal joint, but flexion in the first interphalangeal joint. Unlike Indohyus and other taxa compared in this study, pakicetids retained a greater range of mobility in each of the joints of the manus. 52

Pakicetids are thought to be digitigrade based on their flat and broad metacarpals and phalanges (Madar, 2007). Elements of the manus in Indohyus are not flat or broad. Indohyus is also unlike modern unguligrade cursors that it restricts flexion and extension in the joints of the manus. Based on articular facet morphology, Indohyus probably walked with a modified digitigrade stance with the distal aspect of phalanx 1 raised.

Hindlimb Mobility

The innominates of Indohyus and pakicetids display large and flat ischial surfaces that lack the well-developed and prominent ischial spine and tuberosity seen on a standard terrestrial cursor, such as Odocoileus. This suggests that Indohyus and pakicetids had reduced gluteal musculature compared to terrestrial cursors.

Motion at the pelvic-femoral joint of Indohyus is restricted to flexion and extension, as in all examined artiodactyls. Whereas pakicetids and Tragulus are able to extend the femur until it is parallel with the longitudinal axis of the body and flex the femur until it is 45 degrees below this longitudinal axis, both Indohyus and Odocoileus can only extend the femur until 45 degrees below the longitudinal axis, but share an equivalent ability for flexion.

In all taxa, the knee is habitually flexed and motion at the joint is restricted to flexion and extension. Odocoileus displays the greatest range of motion in flexion and extension.

All examined taxa displayed a proximal tarsus (calcaneus, astragalus, and tibia) that functions like that of a typical artiodactyl in flexion and extension with only the proximal aspect of the calcaneus pivoting medially during flexion (see Schaeffer, 1947, Thewissen and Madar,

1999). These taxa differ from another in the presence or absence of fusion between the cuboid and navicular, and the astragalus-cuboid joint morphology. Gujaratia (Thewissen and Hussain, 53

1990), Indohyus, and pakicetids (Madar, 2007) lack cubonavicular fusion, while these tarsals are fused in Tragulus and Odocoileus. Fusion of these joints further stabilizes the joint and facilitates cursorial and saltatorial locomotion. Position of the astragalus-cuboid joint also varies between these taxa. In some diacodexids, pakicetids and Indohyus, the cuboidal facet is positioned proximal to the navicular facet. These taxa also lack metatarsal fusion, and the cuboid and navicular may function independently in facilitating abduction of the metacarpals, corresponding with increasing pedal surface area.

Within the pes, metatarsals, and proximal and intermediate phalanges are preserved.

Unlike the fused metatarsals of Tragulus and Odocoileus, all metatarsals are free to move individually in Indohyus and pakicetids. Metatarsophalangeal joints of Indohyus indicate a greater range of mobility in flexion, rather than extension, as the distal facet of the metatarsal is longer along its dorsal surface as compared to the ventral surface of the facet, as in Tragulus and pakicetids. The distal facet of the metatarsal of Odocoileus displays a sagittal crest along the entire facet, thus stabilizing the joint against mediolateral translation. This facet in

Indohyus, pakicetids, and Tragulus displays a sagittal crest only on its palmar surface and may allow for slight mediolateral movement when the joint is flexed.

The proximal interphalangeal joints of the pes in Indohyus and Diacodexis were typically oriented in flexion, unlike the joints of Tragulus and Odocoileus that restrict joint excursion away from the vertical plane. The distal trochlea of the proximal phalanx of Indohyus and Odocoileus displays facets along its distal and palmar surfaces, while this facet is absent on the palmar surface (Tragulus) or is small (Odocoileus).

Hindlimb function of Indohyus was probably restricted to flexion and extension in the proximal joints (i.e., pelvic and knee), but the joints of the manus were able to extend and 54

abduct to increase pedal surface area. Joint angles indicate the hindlimb was probably digitigrade, as in pakicetids. Although Indohyus was able to locomote on land, its pes morphology suggests locomotion in either soft substrates (e.g., mud) or pelvic paddling. Both repertoires are reported in pakicetids (Madar, 2007). However, a hallmark of pelvic paddling is the presence of elongated metatarsals and phalanges (Thewissen and Fish, 1997), and these elements are only slightly elongated in Indohyus relative to terrestrial cursors (e.g., Odocoileus,

Tragulus).

Interdigital Webbing

Most modern marine mammals encase the autopod in a soft tissue flipper (see review in

Cooper, 2009). In cetaceans, this soft tissue flipper has a dermis thick with dense irregular connective tissue that impedes digital movement and creates a streamlined hydrofoil that generates lift (e.g., Cooper et al., 2008, Weber et al., 2009). Manatees have thick, muscular flippers that function in drag-based propulsion as manatees use them like oars during turning maneuvers (Hartman, 1979). Pinnipeds, walruses, and otters, however, have a muscular flipper encased with thin dermal and epidermal layers that affords large movements of the joints of the digits. Otariid flippers are oscillated like a bird wing to generate lift (English 1976, 1977), while walruses use their flippers as paddles for steering (Gordon, 1981). Otters restrict webbing to their hindlimbs, and this additional tissue increases surface area of the pes, and facilitates greater thrusts from hindlimbs (Tarassof, 1972, Tarassof et al., 1972).

Osteological correlates for this interdigital webbing occur as flanges along metapodials and phalanges where strong digital abductors attach (Gingerich et al., 2001; Madar, 2007).

These flanges are documented in pakicetids (Fig. 11; Madar, 2007) and protocetids (Gingerich et 55

al., 2001), but are lacking in the later occurring basilosaurids (Uhen, 2004) and modern cetaceans. It could be that strong digital abductors and by extension interdigital webbing first evolved in pakicetids, were retained in protocetids, but were reduced in basilosaurids and modern cetaceans. Indeed, digital abductors are retained in some basal odontocete taxa (i.e. sperm whales (Cooper et al., 2007b)) but are lost in all other cetaceans.

The metapodials and phalanges of Indohyus lack osteological correlates for strong digital abductors and webbing. This absence of a correlate suggests that either strong abductors were not required for aquatic locomotion by Indohyus, and/or that they lacked interdigital webbing.

Therefore, webbing probably first evolved in pakicetid cetaceans, and could have been a key limb morphology that increased autopod surface area during mud-based locomotion and swimming maneuvers.

Buoyancy Control

Aquatic tetrapods utilize several strategies to alter their density relative to water (Uhen,

2004). These hydrostatic strategies include (see definitions in de Ricqlès and de Buffrénil, 2001): osteosclerosis (increased cortical bone thickness along the endosteal surfaces, resulting in a reduced medullary cavity), pachyostosis (increased cortical bone thickness along the periosteal surface, resulting in an enlarged or swollen bone), pachy-osteosclerosis (enlarged bones with a reduced medullary cavity), osteoporosis (lightening the skeleton by decreasing cortical bone content and trabecular bone thickness, Wall, 1983), 56

Fig. 11. Phalangeal cross-sections in Indohyus, pakicetids, and Ambulocetus natans. Cross- sections through the A, manual phalanx.1 (RR 273), and B, pedal phalanx.1 (RR 132) of Indohyus, C, manual phalanx I.1 of the pakicetid Ichthyolestes (H-GSP 92103), D, pedal phalanx V.1 of the pakicetid Pakicetus (H-GSP 92104), and E, manual phalanx I.1 and F, pedal phalanx V.1 of Ambulocetus. Arrows indicate the presence of a lateral flange described by Madar (2007). Scale bar is 1 cm in length.

ingestion of gastroliths or swallowing stones (Taylor, 1993; Wings, 2007), increase in adipose tissue (Wall, 1983), and/or lung compression during deep dives (Ridgway and Howard, 1979).

Modern aquatic tetrapods utilize some of these hydrostatic strategies. Relative to terrestrial taxa, all modern marine mammals have an increased amount of stored fat (Ling,

1974), creating a buoyant body. Marine mammals have therefore ulitized different evolutionary strategies to alter their density: alter the skeleton so that it acts as ballast, or employ external materials that will act to weight the skeleton (i.e., stones). Manatees have pachy-osteosclerotic ribs (Kaiser, 1960; Domning and de Buffrénil, 1991). This additional mineral content in their ribs acts to weight the skeleton, but causes the bone to be brittle and fracture easily when compared to the non-thickened limb bones of humans and bovines (Yan et al., 2006a,b). Sea otters (Fish and Stein, 1991), Hippopotamus, beavers, and some pinnipeds (Wall, 1983; Kaiser,

1960) display some osteosclerotic skeletal elements. The skeletons of modern cetaceans are lightened, or osteoporotic, due to a decrease in bone volume and/or density, usually within the ribs and appendicular skeleton, but vertebrae show no alterations in thickness or structure when compared to that of a terrestrial mammal (Felts and Spurrell, 1965, 1966; de Buffrénil et al., 1985, 1986; de Buffrénil and Schoevaert, 1988; de Ricqlès and de Buffrénil, 2001; Butti et al.,

2007). Some cetaceans experience lung compression during deep dives (Ridgway and Howard,

1979). Pinnipeds (otariids, odobenids, and phocids) also swallow stones, with the greatest frequency of gastroliths documented in phocids (Taylor, 1993), but the stones function as ballast remains controversial because buoyancy control by breathing is more effective (see review in

Wings, 2007).

Evidence of some of these hydrostatic strategies can be found in the fossil record. Fossil sirenians had pachy-osteosclerotic ribs (Buffrénil et al., 2008). The earliest fossil cetaceans 58

(archaeocetes) had osteosclerotic (pakicetids (Gray et al., 2007)) and pachy-osteosclerotic ribs

(ambulocetids, remingtonocetids, protocetids (Gray et al., 2007), and basilosaurids (de Buffrénil et al., 1990, de Ricqlès and de Buffrénil, 2001, Gray et al., 2007)). Some archaeocetes also displayed osteosclerotic limb bones (pakicetids (Madar, 2007, Thewissen et al., 2007), ambulocetids, remingtonocetids, protocetids, basilosaurids (Madar, 1998)). Although stomach stones have been preserved in several fossil aquatic reptiles (Wings, 2007), definitive evidence of fossil aquatic mammals with preserved gastroliths is lacking. The morphology of blubber and lungs cannot be easily determined in fossils as they have no osteological correlates, but future work using isotopes may shed light on the chemical makeup of fat, and rib shape might offer clues as to the ability of fossil taxa to undergo thoracic compression.

Indohyus displays some of these hydrostatic strategies. The long bones (i.e., humerus

(RR 157), femur (RR 42), ulna (RR 39), tibia (RR 46), metacarpal (RR 228), and metatarsal (RR

225)), a rib (RR 217) and a caudal vertebra (RR 249) of Indohyus have a thickened cortex and reduced medullary cavity, indicating the presence of osteosclerosis (Fig. 12A-G,M). In terrestrial artiodactyl taxa (i.e., modern taxa Tragulus, Saiga, Sus, and fossil taxa Caenotherium,

Leptomeryx, Poebotherium and Merycoidon) midshaft femoral cross-sectional areas indicate that the medullary cavity takes up 60-75% of the bone diameter, where as the semi-aquatic

Hippopotamus has a medullary cavity that takes up 55% of the bone diameter (Table 2;

Thewissen et al., 2007). These results indicate that within the Order Artiodactyla, a semi- aquatic form displayed a reduced medullary cavity when compared to terrestrial forms.

Indohyus clearly displays a femoral geometry that suggests it is osteosclerotic as the medullary cavity is 42% of the total bone diameter. The earliest cetaceans also displayed osteosclerotic femoral midshafts, although there was variability in the degree of mineralization. The pakicetids 59

Fig. 12. Bone cross-sectional geometries in Indohyus compared to terrestrial artiodactyls with an equivalent body size (Tragulus and Hyemoschus). Paleohistological sections through the A, femur (RR 42), and D, humerus (RR 157) of Indohyus. Cross-sectional CT scans of the B, tibia (RR 46), C, metatarsal (RR 225), E, ulna (RR 39), F, metacarpal (RR 228), G, rib (RR 217), and M, caudal vertebra (RR 249) of Indohyus. Cross-sectional CT scans of the H, femur, I, tibia, J, fused metatarsals (cannon bone), K, humerus, and L, ulna, of Tragulus (CMNH 17917). Cross-sectional CT scan of the N, rib of Hyemoschus (CMNH 17918).

Table 2. Cortical bone thickness in the long bones of artiodactyls. Percentage of the Medullary Cavity Diameter to Total Bone Diameter in Skeletal Elements of Fossil and Extant Artiodactyls.

Taxon Specimen ID Bone Medullary Cavity as Reference % Bone Diamater Indohyus RR 157 Humerus 48 Thewissen et al., 2007 Ichthyolestes H-GSP 96227, 96247 Humerus 14-15 Thewissen et al., 2007 Andrewsiphius IITR-2871 Humerus 41 Thewissen et al., 2007 Marmota HT Humerus 56 Thewissen et al., 2007 Caenotherium IVAU Humerus 58 Thewissen et al., 2007 Merycoidodon USNM 2459 Humerus 61 Thewissen et al., 2007 Oreotragus USNM 314959 Humerus 63 Thewissen et al., 2007 Sus barbatus USNM 34891 Humerus 93 Thewissen et al., 2007 Hyemoschus CMNH 17918 Humerus 68 this study Tragulus CMNH 17917 Humerus 67 this study Tragulus CMNH 28126 Humerus 57 this study

Indohyus RR 42 Femur 42 Thewissen et al., 2007 Ichthyolestes H-GSP 30345 Femur 25 Thewissen et al., 2007 Nalacetus H-GSP 98124 Femur 54 Thewissen et al., 2007 Ambulocetus H-GSP 18507 Femur 57 Thewissen et al., 2007 Hippopotamus USNM 162977 Femur 55 Thewissen et al., 2007 Caenotherium IVAU Femur 60 Thewissen et al., 2007 Merycoidodon USNM 2460 Femur 68 Thewissen et al., 2007 Hyemoschus CMNH 17918 Femur 78 this study Sus barbatus USNM 34891 Femur 70 Thewissen et al., 2007 Tragulus USNM 578462 Femur 63 Thewissen et al., 2007 Tragulus CMNH 17917 Femur 71 this study Tragulus CMNH 28126 Femur 67 this study Leptomeryx USNM 362713 Femur 64 Thewissen et al., 2007 Saiga USNM 336264 Femur 68 Thewissen et al., 2007 Capra CMNH B640 Femur 70 this study

Indohyus RR 46 Tibia 39 this study Hyemoschus CMNH 17918 Tibia 68 this study Tragulus CMNH 17917 Tibia 60 this study Tragulus CMNH 28126 Tibia 61 this study Capra CMNH B640 Tibia 51 this study

Indohyus RR 225 Metatarsal 50 this study Hyemoschus CMNH 17918 Metatarsal 58 this study Tragulus CMNH 17917 Metatarsal 62 this study Tragulus CMNH 28126 Metatarsal 62 this study Capra CMNH B640 Metatarsal 59 this study

Indohyus RR 228 Metacarpal 57 this study Indohyus RR 271 Metacarpal 55 this study Hyemoschus CMNH 17918 Metacarpal 55 this study Tragulus CMNH 17917 Metacarpal 43 this study Tragulus CMNH 28126 Metacarpal 36 this study

Indohyus RR 235 Rib 62 this study Indohyus RR 217 Rib 59 this study Hyemoschus CMNH 17918 Rib 58 this study 61

Nalacetus and Ichthyolestes displayed medullary cavities that were 54% and 25% of the bone area respectively, while Ambulocetus displayed a hippo-like bone structure of 57% (Thewissen et al., 2007). Taken together, these results suggest that the femur of Indohyus was osteosclerotic relative to terrestrial and semi-aquatic artiodactyls, but its level of osteosclerosis was less pronounced relative to the extreme osteosclerosis seen in Ichthyolestes.

Similar results were found in an analysis of humeral midshaft dimensions (Fig. 12D,K,

Table 2). Terrestrial artiodactyls (i.e. modern taxon Sus, and fossil taxa Caenotherium,

Merycoidodon, Oreotragus) had large medullary cavities that took up 56-93% of the humeral diameter, while the cavity of Indohyus encompassed only 48% of the total bone diameter. The archaeocete Ichthyolestes had an extremely reduced cortex (14-16%, Thewissen et al., 2007).

Rib cross-sectional dimensions (Fig. 12G,N) of aquatic tetrapods have been studied in detail (e.g, Fawcett, 1942, Domning and Myrick, 1980, de Buffrènil et al., 1990, Domning and de

Buffrénil, 1991, de Ricqlès and de Buffrénil, 2001, Uhen, 2004, Gray et al., 2007). Aquatic mammals typically display greater mineral content in rib cross-sections (e.g., osteosclerosis in otters, pachy-osteosclerosis in manatees) compared to terrestrial taxa (e.g., bovids, horses, Gray et al., 2007). Indohyus displays an osteosclerotic medullary cavity caused by an increase in cortical dimensions, whereas archaeocetes evolved osteosclerosis by only slightly increasing the cortical thickness, but greatly thickening the trabecular struts within the medullary cavity (Gray et al., 2007; Madar, 2007).

Vertebral cross-sectional dimensions are poorly studied. Madar (2007) noted a thick cortical layer to the vertebra of the pakicetid Ichthyolestes. A CT scan through the midshaft of a caudal vertebra (RR 249) indicated that Indohyus had a thin cortical layer, but the center of the vertebra was filled with bone, presumably in the form of trabecular struts (Fig. 12N). 62

Phylogenetic analyses indicate that Indohyus is the closest fossil relative to all cetaceans

(Thewissen et al. 2007; Geisler and Theodor, 2009), and this analysis of its postcranial skeleton indicates Indohyus is a morphological intermediate between primitive terrestrial artiodactyls

(i.e., Diacodexis and Gujaratia) and the earliest cetaceans. Both dichobunid artiodactyls and pakicetids display skeletal morphologies consistent with adaptations associated with cursoriality

(Rose, 1985; Madar, 2007), however, the skeletons of pakicetids are derived in their robust and massive skeletal elements and the presence of hyperostosis within the skeleton. Indohyus retains the gracile skeleton of primitive artiodactyls, but also displays osteological correlates of robust axial musculature and a hyperostotic appendicular skeleton to a lesser degree than seen in pakicetids.

The presence of bone ballast indicates that Indohyus occupied an aquatic niche

(Thewissen et al., 2007). However, the degree of hyperostosis in Indohyus is considerably less than that of pakicetids and modern Hippopotamus (see Chapter 3 of this dissertation). Because hyperostotic bones act as skeletal ballast, Indohyus probably lacked the ability to forage and locomote at depth relative to pakicetids and Hippopotamus. Results of this study indicate that the skeleton of Indohyus probably represents an incipient phase of the Eocene invasion of the aquatic environment. The minimal level of hyperostosis in the skeleton of Indohyus represents an early adaptation of an evolving life in a fluid environment that was later intensified within

Cetacea. Pakicetid cetaceans gave rise to the only fully aquatic lineage of artiodactyls; it could be that this increased skeletal ballast was a key innovation that facilitated the successful invasion into the aquatic environment. CHAPTER III

EVOLUTION OF BONE MICROSTRUCTURE DURING THE AQUATIC INVASION OF CETARTIODACTYLA (MAMMALIA)

Introduction

Secondarily aquatic tetrapods (e.g., marine forms of the classes Mammalia (e.g. sirenians, cetaceans, pinnipeds), Aves (e.g. penguins), and Reptilia (marine turtles and snakes)) transitioned from a terrestrial habitat and successfully invaded the marine environment. In adapting to this fluid medium, these aquatic tetrapods altered their body shape and skeletal mass, relative to terrestrial taxa. Furthermore, to reduce drag during swimming maneuvers, most evolved a streamlined body shape that resulted in less turbulent flow over their body contour. Accordingly, limbs became paddle or aerofoil-shaped (e.g. forelimbs of penguins and marine mammals, hindlimbs of pinnipeds), or were reduced to mere vestiges and encased within the smooth contours of the body (e.g., fore- and hindlimbs of snakes, hindlimbs of cetaceans and sirenians). Beyond modifying their external morphology, most aquatic tetrapods utilized various strategies to alter mineral content in their bones and thereby change their total skeletal mass (e.g., Wall 1983, Stein 1989, Uhen 2004, de Buffrénil et al., 2008).

Variation in Bone Microanatomy

In general, bones of terrestrial taxa have a thin cortex and a large medullary cavity in cross-section, while bones of aquatic tetrapods have a thick cortex and reduced medullary cavity

63 64

(Kriloff et al., 2008). The thick cortex of aquatic tetrapod bones increases skeletal mass

(hyperostosis) and provides ballast that counteracts buoyancy (Wall 1983). Hyperostosis also changes the material properties of these bones, making them more brittle and easier to fracture relative to those of terrestrial taxa (Clifton et al., 2006). In this study, three forms of hyperostosis are discussed (Francillon-Vieillot et al., (1990), de Ricqlès and de Buffrénil (2001),

Gray et al., (2007); Table 3). First, osteosclerosis develops when the endosteal surface of a bone is compact, causing an increase in cortical bone cross-section dimension and a dimensional reduction of the medullary cavity (de Ricqlès and de Buffrénil, 2001). Osteosclerosis differs from other forms of hyperostosis in that bone external dimensions do not change. Osteosclerosis occurs in early whales (de Buffrénil et al., 1990; Madar, 1998; Gray et al., 2007; Madar, 2007), manatees (Kaiser, 1960; Domning and de Buffrénil, 1991), sea otters (Fish and Stein, 1991),

Hippopotamus, beavers, pinnipeds (Wall, 1983), and some Mesozoic marine reptiles (Kaiser,

1960; Taylor, 2000). Second, some aquatic tetrapods display pachyostosis, i.e. an outward thickening of compact bone via increased appositional deposition of cortical bone, leading to wider or swollen-appearing bones (Francillon-Vieillot et al., 1990; de Ricqlès and de Buffrénil,

2001). Pachyostosis has been documented in the ribs of a plagiosaurid amphibian and the limbs of a placodontid (de Ricqlès and de Buffrénil, 2001). Third, pachyostosis commonly occurs in conjunction with osteosclerosis, creating pachy-osteosclerostotic bones (de Ricqlès and de

Buffrénil, 2001) which was observed in the ribs of fossil cetaceans (de Buffrénil et al., 1990; de

Ricqlès and de Buffrénil, 2001; Gray et al., 2007), and the ribs of extant and fossil sirenians

(Fawcett, 1942; de Buffrénil and Schoevaert, 1989; de Ricqlès and de Buffrénil, 2001).

Table 3. Definitions of BonyHhistological Specializations that Alter Skeletal Mass. Based on Francillon-Vieillot et al., (1990), de Ricqlès and de Buffrénil (2001), Gray et al., (2007).

______Bony Morphology Description ______Hypo-ostoses intense thinning of cortical compact bone, fewer trabeculae, Osteoporosis and general bone loss

Hyper-ostoses appositional thickening of cortical compact bone, resulting in a Pachyostosis widened, or swollen bone

Osteosclerosis inner aspect of bone becomes filled by a greater presence of compact bone, and medullary cavity is reduced

Pachy-osteosclerosis combination of pachyostosis and osteosclerosis ______

Rather than increasing skeletal mass, some aquatic tetrapods reduce bone mineral content. Osteoporotic bones are filled with spongy bone, display only a thin cortex of compact bone, and lack a medullary cavity (de Buffrénil and Schoevaert, 1988; de Ricqlès and de

Buffrénil, 2001). This reduction in bone mass minimizes inertia during aquatic locomotion

(Kriloff et al., 2008), and is particularly important for fast swimmers. Osteoporotic bones are present in modern cetaceans (Felts and Spurrell, 1965, 1966; de Buffrénil et al., 1985, 1986;

Buffrénil and Schoevaert, 1988; de Ricqlès and de Buffrénil, 2001, Butti et al., 2007), Mesozoic marine ichthyosaurs (de Buffrénil and Mazin, 1990; de Ricqlès and de Buffrénil, 2001), anurans

(Leclair et al., 1993; Castanet and Caetano, 1995), some marine turtles (Rhodin et al., 1981;

Rhodin, 1985), mesosuchian marine crocodilians (Buffetaut, 1979; Buffetaut et al., 1982; Hua and de Buffrénil, 1996), and plesiosaurs and pliosaurs (Wiffen et al., 1985).

The molecular mechanisms generating this diversity of tetrapod cortical bone thickness are currently poorly understood. Activity rates of osteoblast and osteoclast cells determine bone thickness, and their actions are partially controlled by a cascade of genes within the gut- 66

to-bone axis (Long, 2008; Yadav et al., 2008). Within this pathway, the low-density lipoprotein receptor-related protein 5 (LRP5) gene, among many others, is a key regulator of bone mass

(e.g., Clément-Lacroix et al., 2005; Baron et al., 2006; Bodine and Komm, 2006; Sawakami et al.,

2006; Balemans and Van Hul, 2007; Macsai et al., 2008; Yadav et al., 2008). Experimental results indicate that loss-of-function mutations to the LRP5 gene in mice and humans inhibits osteoblast function and cause osteoporosis pseudoglioma, a disorder characterized by reduced bone mass and strength (Bodine and Komm, 2006; Ralston and de Crombrugghe, 2006; Macsai et al., 2008; Yadav et al., 2008). Alternatively, gain-of-function mutations reduce the rate of programmed cell death in osteoblasts and induces diseases with high bone mass phenotypes with thickened bone cortices (Boyden et al., 2002; Little et al., 2002; Bodine and Komm, 2006;

Ralston and de Crombrugghe, 2006; Balemans and Van Hul, 2007; Yadav et al., 2008). Beyond

LRP5, a plethora of other genes, transcription and endocrine factors have been identified for their potential role in altering bone thickness (e.g. Karsenty, 1999, Ducy et al., 2000, Takeda et al., 2002, Hamrick 2003, Elefteriou et al., 2004, Baron and Rawadi, 2007, Iwaniec et al., 2007,

Cirmanová et al., 2008, Garimella et al., 2008, Hamrick and Ferrari, 2008, Heep et al., 2008, Lee and Karsenty, 2008, Takeda and Karsenty, 2008). Furthermore, most of these studies of molecular mechanisms were conducted on mouse models, and researchers have yet to explore the genetic underpinnings of mammalian ordinal differences in skeletal development.

Bone Microanatomy as an Indicator of Ecological Niche

Cross-sectional geometry of appendicular long bones has been a subject of intense study (e.g., Cubo et al., 2005; Germain and Laurin, 2005; Laurin et al., 2006; Thewissen et al.,

2007; Kriloff et al., 2008). The majority of bone microanatomical studies have focused on the 67

biomechanical or functional attributes (e.g. distributions of stresses and strains during controlled loading events) in extant taxa (e.g., Dehority, et al., 1999; Hiney et al., 2004; Sakata et al., 2004; Pontzer, et al., 2006); however, a small subset of microanatomical studies correlate bone structure with an ecological niche (e.g., Wall, 1983; Stein, 1989; Germain and Laurin, 2005;

Laurin et al., 2006; Habib and Ruff, 2008; Kriloff et al., 2008). These studies quantify bone structural dimensions (cortical and medullary cavity sizes) and offer a substantial data set based on extant taxa that occupy different ecological niches and locomote in associated substrates.

Within extant vertebrates, dimensions of the tibia in cross-section are expected to be the most reliable indicator of habitat as the tibia is the primary weight-bearing bone of the hindlimb zeugopod (Kriloff et al., 2008).

Beyond establishing correlations between bone and habitat or locomotor regime, recent studies illustrate the evolutionary history of bone structure by incorporating phylogenetic data

(e.g., Cubo et al., 2005, Germain and Laurin 2005). These studies track major evolutionary events in cortical bone evolution and allow an estimation of when these evolutionary events occurred in geologic time. However, these studies have only traced the phylogenetic affinities of cortical bone in a taxonomically small sample of mammals including four terrestrial artiodactyls, (even toed ungulates, e.g., Camelus, Sus, Cervus, Capreolus; Germain and Laurin

2005, Kriloff et al., 2008) while no taxon-rich analyses have addressed the evolutionary patterns of cortical bone geometries in artiodactyls and cetaceans (cetartiodactylans).

Cetartiodactylans as an Ideal Group For Testing

Artiodactyls first appeared at the beginning of the Eocene Epoch as terrestrial mammals, but within about 15 million years, a single lineage completed the land-to-sea transition that 68

ultimately led to the most successful and diverse group of marine mammals, the cetaceans

(whales, dolphins and porpoises). Concomitant with changes in their paleoenvironment, major alterations occurred in fossil cetacean long bone anatomy. In the first 10 million years of their evolution, archaic cetaceans (archaeocetes) were waders that foraged in bodies of shallow water (Roe et al., 1998; Thewissen and Hussain, 1998; Thewissen et al., 2001a; Clementz et al.,

2006; Nummela et al., 2007) and displayed long bones with thickened cortices and reduced medullary cavities (osteosclerosis) (Madar, 1998; Madar, 2007; Gray et al., 2007; Thewissen et al., 2007). These thickened bones acted as ballast to weight the skeleton (Taylor, 2000). The closest fossil relative of cetaceans, a middle Eocene raoellid, Indohyus, also had osteosclerotic limbs and occupied occupied a freshwater habitat (Thewissen et al., 2007). By the Late Eocene epoch, archaeocetes lost their ability to bear weight on land and a second phase of changes in bone microanatomy occurred. Modern cetaceans now display osteoporotic long bones that are thought to decrease inertia during aquatic locomotion (Felts and Spurrell, 1965, 1966, de

Buffrénil 1985, de Buffrénil 1986, de Buffrénil and Schoevaert 1988).

Beyond this invasion into the marine environment, several extant artiodactyls occupy a variety of habitats including, marshes (e.g. Tragelaphus spekei) and rivers (i.e., Hippopotamus).

Of those typically terrestrial artiodactyls, several species of forest-dwelling tragulids (Nowak,

1991) flee into small bodies of freshwater to escape predation and some taxa can stay submerged for up to five minutes at a time (Dubost, 1978). Because modern and fossil cetartiodactylans occupy terrestrial, marsh, riverine, and aquatic (cetaceans) habitats, they make an ideal subject for a case study testing correlates between an ecological niche and cortical bone dimensions, as well as tracing cortical bone evolution within Cetartiodactyla. 69

This study surveys cortical bone architecture of the humerus, femur, and tibia in a taxonomically broad sample of modern and fossil artiodactyls. We first tested the hypothesis that cortical bone dimensions are correlated with habitat in modern artiodactyls. Specifically, we hypothesized that those taxa that live primarily in a terrestrial habitat (e.g., Odocoileus,

Cervus, Bos), including those that may flee into the water for refuge (i.e. Hyemoschus), and those that occupy soft-substrate niches such as marshes or swamps (e.g., Alces, and

Tragelaphus spekei) have a large medullary cavity and a reduced cortex. Accordingly, we also hypothesize that those modern taxa that occupy a riverine environment and spend most of their time submerged (e.g. Hippopotamus) will display a comparatively thick cortex and reduced medullary cavity. If a reasonable correlation is established between cortical bone dimension and habitat, this dataset will then be used to reconstruct the habits of fossil artiodactyls (i.e. anthracotheres, raoellids, and archaeocetes), and test for patterns of parallel evolution in a phylogenetic context.

Institutional Abbreviations: AMNH, American Museum of Natural History; CMNH,

Cleveland Museum of Natural History; H-GSP, Howard University-Geological Survey of Pakistan;

IITR-SB, Indian Institute of Technology at Roorkee – Sunil Bajpai; USNM, Smithsonian Museum of

Natural History; YPM, Yale Peabody Museum.

Materials and Methods

Osteological Sample

This study analyzes a taxonomically broad range of femora, tibiae and humeri of modern

(n = 29, representing 15 genera and 17 taxa), and fossil cetartiodactylans (n=20, including four fossil cetaceans (Pakicetus, Ichthyolestes, Andrewsiphius, and Kutchicetus), a raoellid Indohyus 70

Table 4. Taxonomic Identity and Compactness Profile Parameters of the Femur, Humerus and Tibia. Based on Midshaft CT Scans and Histological Sections of Cetartiodactylans. Values of S, P, Min and Max are the Algebraic Mean Values of 60 Radial Values Calculated in the Program Bone Profiler. Taxa Reared in Captivity are Indicated with an *, and Fossil Taxa are Indicated with a †.

______Taxon Specimen ID S P Min Max ______FEMUR Alces americanus americanus USNM 275127 0.01379 0.71618 0.00147 0.99859 Axis axis* USNM 395635 0.02312 0.63854 0.00000 1.00000 Axis axis USNM 122532 0.01567 0.68269 0.02106 1.00000 Bos Taurus USNM 277262 0.01383 0.57454 0.00000 0.99993 Bubalis bubalis AMNH 54765 0.01946 0.52500 0.00000 1.00000 Capra hircus USNM A00720 0.01154 0.68886 0.00000 0.99993 Capra hircus CMNH B640 0.01143 0.69511 0.00000 0.99993 Hexaprotodon liberiensis* AMNH 148452 0.01417 0.62615 0.00000 0.99993 Hexaprotodon liberiensis* USNM 464982 0.21362 0.11420 0.00000 1.02634 Hippopotamus amphibius AMNH 15898 0.04969 0.32927 0.69975 1.00000 Hydropotes inermis USNM 239609 0.01358 0.58855 0.00000 0.99993 Hydropotes inermis* USNM 304664 0.01597 0.59834 0.0000 1.00000 Hyemoschus aquaticus AMNH 53613 0.01575 0.68513 0.00000 1.00000 Hyemoschus aquaticus CMNH 17918 0.01287 0.77202 0.00000 0.99993 Hyemoschus aquaticus* USNM 270111 0.02044 0.72611 0.00391 0.98375 Kobus leche AMNH 70010 0.01603 0.57647 0.00000 1.00000 Kobus leche* USNM 254927 0.02024 0.54742 0.00000 1.00000 Kobus megaceros leucotis AMNH 82135 0.02150 0.59811 0.00000 1.00000 Moschus berezovskii USNM 259384 0.00980 0.77559 0.00000 0.99993 Odocoileus virginianus USNM 254652 0.01855 0.61759 0.00000 1.00000 Odocoileus virginianus USNM 254653 0.01290 0.55719 0.00000 0.99993 Odocoileus virginianus USNM 396283 0.01185 0.65759 0.00000 1.00000 Potamochoerus porcus USNM 164542 0.01349 0.56216 0.00000 0.99998 Potamochoerus porcus* USNM 259174 0.01569 0.63468 0.00000 1.00000 Tragelaphus scriptus USNM 164560 0.01439 0.68532 0.00000 1.00000 Tragelaphus spekei USNM 164558 0.01907 0.65719 0.00000 1.00000 Tragulus javanicus CMNH 17917 0.01435 0.68305 0.00000 1.00000 Tragulus javanicus CMNH 21826 0.01054 0.67813 0.00000 0.99999 Andrewsiphius sloani† IITR-SB 2871.15 0.04002 0.25558 0.13268 0.99998 Ichthyolestes pinfoldi† HGSP 30345 0.01490 0.22536 0.00000 0.99993 Indohyus† RR 42 0.03635 0.50757 0.00000 0.99998 Merycopotamus medioximus† YPM 30913 0.01944 0.66010 0.00000 0.99993 Merycopotamus medioximus† YPM 9013 0.01295 0.68799 0.00000 0.99981 ______

Table 4 (continued). Taxonomic Identity and Compactness Profile Parameters of the Femur, Humerus and Tibia. Based on Midshaft CT Scans and Histological Sections of Cetartiodactylans. Values of S, P, Min and Max are the Algebraic Mean Values of 60 Radial Values Calculated in the Program Bone Profiler. Taxa Reared in Captivity are Indicated with an *, and Fossil Taxa are Indicated with a †.

______Taxon Specimen ID S P Min Max ______HUMERUS Alces americanus americanus USNM 275127 0.02806 0.66787 0.00000 0.99993 Axis axis* USNM 395635 0.02484 0.67037 0.00000 1.00000 Axis axis USNM 122532 0.02325 0.57689 0.00000 1.00000 Bos taurus USNM 277262 0.05828 0.75691 0.00000 1.01680 Bubalis bubalis AMNH 54765 0.06405 0.62670 0.01607 1.00313 Capra hircus USNM A00720 0.01692 0.75888 0.00000 0.99993 Hexaprotodon liberiensis* USNM 464982 0.07007 0.51954 0.01670 1.00111 Hippopotamus amphibius AMNH 15898 0.07250 0.48867 0.00265 1.00092 Hyemoschus aquaticus AMNH 53613 0.01124 0.58767 0.00000 0.99930 Hyemoschus aquaticus CMNH 17918 0.01689 0.68828 0.00000 1.00000 Hydropotes inermis USNM 239609 0.01868 0.51665 0.00000 0.99993 Hydropotes inermis* USNM 304664 0.02452 0.51385 0.00000 1.00000 Kobus leche AMNH 70010 0.02373 0.61877 0.00000 1.00000 Kobus leche* USNM 254927 0.02282 0.64357 0.00000 1.00000 Kobus megaceros leucotis AMNH 82135 0.02162 0.69887 0.00000 1.00000 Moschus berezovskii USNM 259384 0.01657 0.75819 0.00000 1.00000 Odocoileus virginianus USNM 396283 0.01999 0.66042 0.00000 1.00000 Odocoileus virginianus USNM 254652 0.02341 0.60885 0.00000 1.00000 Odocoileus virginianus USNM 254653 0.01661 0.59913 0.00000 1.00000 Potamochoerus porcus* USNM 259174 0.01373 0.72508 0.00000 0.99997 Tragelaphus scriptus USNM 164560 0.01477 0.67618 0.00000 1.00000 Tragelaphus spekei USNM 164558 0.01760 0.72424 0.00000 1.00000 Tragulus javanicus CMNH 17917 0.04338 0.63558 0.00000 1.00025 Tragulus javanicus CMNH 21826 0.01469 0.57419 0.00000 1.00000 Andrewsiphius sloani† IITR-SB 2871.201 0.02936 0.79732 0.27881 1.00075 Ichthyolestes pinfoldi† H-GSP 96227 0.07376 0.20379 0.07260 1.00002 Indohyus† RR 157 0.01474 0.33508 0.00000 0.99993 Kutchicetus minimus† IITR-SB 2647.42 0.23016 0.11678 0.64753 1.00776 Merycopotamus dissimilus† YPM 16911 0.04263 0.38169 0.30035 1.00000 Merycopotamus dissimilus† YPM 49877 0.06081 0.48417 0.00000 1.00020 Merycopotamus medioximus† YPM 17587 0.02504 0.43334 0.00000 1.00000 Merycopotamus medioximus † YPM 17697 0.03980 0.46540 0.01631 1.00000 Merycopotamus sp.† YPM 683 0.01333 0.53975 0.00000 0.99996 ______

______Taxon Specimen ID S P Min Max ______TIBIA Alces americanus americanus USNM 275127 0.01753 0.58940 0.00000 1.00000 Axis axis* USNM 395635 0.01760 0.65972 0.00000 1.00000 Axis axis USNM 122532 0.01406 0.59784 0.00000 0.99993 Bos taurus USNM 277262 0.02188 0.57336 0.00000 1.00000 Bubalis bubalis AMNH 54765 0.02600 0.45547 0.00000 1.00000 Capra hircus USNM A00720 0.00943 0.42493 0.00000 1.00000 Capra hircus CMNH B640 0.01327 0.48303 0.00000 0.99993 Hexaprotodon liberiensis* AMNH 148452 0.07571 0.42062 0.20609 1.00040 Hexaprotodon liberiensis* USNM 464982 0.11693 0.27774 0.00000 1.00229 Hippopotamus amphibius AMNH 15898 0.05069 0.22605 0.00639 1.00000 Hydropotes inermis USNM 239609 0.03246 0.47605 0.01032 1.00000 Hydropotes inermis* USNM 304664 0.02319 0.48511 0.00000 0.99981 Hyemoschus aquaticus AMNH 53613 0.03791 0.58980 0.00000 1.00002 Hyemoschus aquaticus CMNH 17918 0.01891 0.66087 0.01084 1.00000 Hyemoschus aquaticus USNM 270111 0.02215 0.69420 0.00000 0.99990 Kobus leche AMNH 70010 0.01874 0.46328 0.00000 1.00000 Kobus leche* USNM 254927 0.01263 0.48013 0.00000 1.00000 Kobus megaceros leucotis AMNH 82135 0.02483 0.55296 0.00000 1.00000 Kobus megaceros leucotis USNM 164777 0.01299 0.59812 0.00000 0.99993 Moschus berezovskii USNM 259384 0.02216 0.67792 0.00000 1.00000 Odocoileus virginianus USNM 254652 0.02854 0.51422 0.00000 1.00000 Odocoileus virginianus USNM 254653 0.02282 0.48373 0.00000 1.00000 Odocoileus virginianus USNM 396283 0.02056 0.56527 0.00000 1.00000 Potamochoerus porcus USNM 164542 0.02609 0.43955 0.00000 1.00000 Potamochoerus porcus* USNM 259174 0.04960 0.60772 0.01465 1.00040 Tragelaphus scriptus USNM 164560 0.01897 0.55354 0.00000 1.00000 Tragelaphus spekei USNM 164558 0.01586 0.51690 0.00000 1.00000 Tragulus javanicus CMNH 17917 0.01127 0.58625 0.00000 1.00000 Tragulus javanicus CMNH 21826 0.02407 0.55962 0.00000 1.00000 Kutchicetus minimus† IITR-SB 2647.44 0.08773 0.13791 0.00000 0.99428 Indohyus† RR 301 0.02521 0.48576 0.00001 1.00000 Pakicetus attocki† H-GSP 30357 0.03582 0.24822 0.00000 1.00000 Merycopotamus nanus† YPM 31826 0.02215 0.42873 0.00000 0.99993 Merycopotamus sp.† YPM 49533 0.02907 0.40114 0.00000 1.00000 Microbunodon silistrense † YPM 28368 0.02031 0.63849 0.00000 1.00000 ______

and four between the medullary and cortical regions. P is the position of the sigmoid inflection point on the x-axis and it represents the position of the transition zone fossil anthracothere taxa

(Merycopotamus sp., Merycopotamus medioximus, Merycopotamus dissimilus, Merycopotamus nanus, and Microbunodon silistrense; see Table 4). Extant samples were taken from wild-caught and captive-reared specimens. Included in this analysis are humerii, femora, and tibiae of several modern terrestrial artiodactyl taxa, as well as taxa that occupy a terrestrial habitat but flee into the water to escape predators (i.e., tragulids, Nowak, 1991), and an artiodactyl that occupies a riverine habitat (Hippopotamus) (Table 5).

Table 5. Habitats of Cetartiodactylan Taxa.

Taxon Habitat Reference Alces americanus (moose) Terrestrial, spends some time in water Nowak 1991

Axis axis (chital) Terrestrial Nowak 1991

Bos taurus (domestic cow) Terrestrial Nowak 1991

Bubalis bubalis (Asian water buffalo) Terrestrial Nowak 1991

Capra hircus (goat) Terrestrial Nowak 1991

Hippopotamus amphibius Spends most time in water, able to walk on land Nowak 1991

Hydropotes inermis (Chinese water deer) Spends most time in water, able to walk on land Nowak 1991

Hyemoschus aquaticus (water chevrotain) Terrestrial, flees into water to escape predation Dubost 1978, Nowak 1991

Kobus leche (lechwe) Terrestrial, spends some time in water Nowak 1991

Kobus megaceros leucotis (Nile lechwe) Terrestrial, spends some time in water Nowak 1991

Moschus berezovskii (musk deer) Terrestrial Nowak 1991

Odocoileus virginianus (white-tailed deer) Terrestrial Nowak 1991

Potamochoerus porcus (African bush pig) Terrestrial Nowak 1991

Tragelaphus scriptus (bushbuck) Terrestrial Nowak 1991

Tragelaphus spekei (sitatunga) Spends most time in swamps, able to walk on land Nowak 1991

Tragulus javanicus (mouse deer) Terrestrial, flees into water to escape predation Nowak 1991 74

Visualization of Cortical Bone Structure

This study utilized high resolution CT scans for most bones with small midshaft diameters, conducted on a XCT540 Research M scanner located at the Northeastern Ohio

Universities College of Medicine and Pharmacy and a medical scanner was also used at the

Akron City Hospital in Ohio. Large limb elements of some anthracotheres were scanned at the high-resolution x-ray computed tomography facility at the University of Texas at Austin

(www.ctlab.geo.utexas.edu). All scans were taken at bone midshaft.

This study also employed paleohistological thin sections of the long bone midshafts of

Indohyus (femur (RR 42), humerus (RR 157)) and the pakicetid Ichthyolestes pinfoldi (H-GSP

96227) (Thewissen et al., 2007).

Quantification of Bone Cross-Sectional Area

Images of long bone cross-sections were analyzed using Bone Profiler (Girondot and

Laurin, 2003). For detailed instructions on this program, see Girondot and Laurin (2003) and

Laurin et al., (2004). Bone Profiler automatically determines the cross-sectional center of a bone image and then divides the image into 60 radial sections with a width of six degrees each section. The whole bone image is also divided into 51 circumferential sections. The ratio between total bone cross-sectional area and the surface occupied by bone (bone compactness) is measured from each of these sections as well as the whole bone (global compactness). Bone

Profiler then employs a sigmoid equation to describe the compactness profile of the bone image, resulting in four descriptive parameters (Fig. 13). S is the reciprocal of the slope at the inflection point and is proportional to the relative width of the transition between medullary and cortical regions. P is the position of the inflexion 75

Fig. 13. (A) Mathematical model that calculates amount of bone as a function of the distance to the center of the bone (Girondot and Laurin, 2003). S is the reciprocal of the slope at the sigmoid inflection point and it is proportional to the relative width of the transition zone between medullary and cortical regions. Min is the minimal asympotitic value and corresponds to the compactness in the center of the medullary area. Max is the maximum asymptotic value and corresponds to the compactness in the superficial cortex. (B) Femoral cross-section of Alces americanus (USNM 275127) and (C) compactness profile. (D) Femoral cross-section of Hippopotamus amphibius (AMNH 15898) and (E) compactness profile. 76

point on the x-axis and it generally represents the position of the transition zone between medullary and cortical areas. Min is the minimal asymptotic value and corresponds to the compactness at the center of the medullary region. Max is the maximum asymptote and corresponds to the compactness in the superficial bone cortex. Taxa with femoral and humeral

P-values between 0.1 and 0.5 were considered osteosclerotic, while those of the tibia were considered osteosclerotic if they were between 0.1 and 0.4.

All four parameters were calculated for two specimens of Hexaprotodon, but values of cortical thickness varied so greatly between the two individuals (Table 4) that this taxon was excluded from subsequent analyses. Both specimens were captive. Although the cortex of these specimens was visible, the medullary cavity was filled with fine trabeculae, but because of poor CT scan resolution, it was impossible to reliably distinguish between bone and space in the medullary cavity.

Correlation with Body Mass

To test whether cortical bone thickness was correlated with body mass in fossil and extant cetartiodactylans, double logarithmic plots of P versus body mass were employed, a regression analysis was conducted, and residual values of the regression were plotted for each bone. The ratio of double logarithmic P versus Body Mass was also calculated for each bone.

Body mass values were taken from published observations of extant artiodactyls (Eisenberg,

1981, Nowak, 1991, Brashares et al., 2000) and calculated for fossil cetacean taxa based on bone dimensions (Lancaster et al., in prep). Body mass calculations for fossil cetaceans assume a normal mammalian cortical bone thickness as no measure of normal versus hyperostotic bone mass calculations for cetartiodactylans is known. 77

Ancestral Character State Reconstruction

The program Mesquite version 2.6 (Maddison and Maddison, 2009) was used to trace when, in evolutionary time, a thickened cortex evolved within Cetartiodactyla. For the continuous variable P, ancestral character states for each tree were determined through squared-change parsimony (Maddison, 1991). Ancestral character states were calculated twice, once with all branch lengths equal to one (patristic distances) and once with branch lengths ultrametricized using the arbitrarily ultrametricized option, which extends branch tips to the same position in Mesquite.

Results

Visualization of Cortical Bone Structure

Results of CT scans of long bone midshafts facilitated visualization of the diversity of relative cortical bone thickness in both fossil and extant cetartiodactylans (Figure 13).

Representative pakicetids (Fig. 14 J-L) and Hippopotamus (Fig. 14A-C) displayed the greatest relative cortical thickness as compared to other sample taxa. Conversely, the relatively thinnest cortices were found in Tragulus (Fig. 14M-O), Alces (Fig. 14S-U), and Moschus (Fig. 14Y-AA). The bones of Odocoileus had unusually thick cortices that were intermediate between those of pakicetids and Hippopotamus, but thicker than those of Tragulus, Alces, and Moschus (Fig. 14V-

X).

Although Hippopotamus limb bones exhibited a thickened cortex, this taxon also displayed an abundance of spongy bone within the long bone medullary cavity (Fig. 14A-C). In contrast, most taxa either lack spongy bone in the cortex (e.g., Tragelaphus (Fig. 14HH-MM)), or only display spongy bone in the humerus (i.e., Alces (Fig. 14T), Bos (Fig. 13CC), 78

Fig. 14. Radiographs through the long bone midshafts of cetartiodactylans illustrating diversity in cortical bone thickness. (A-C) Femur, humerus and tibia of Hippopotamus (AMNH 15898). The (D) femur (YPM 30913) and (E) humerus (YPM 17587) of Merycopotamus medioximus. The (F) Tibial cross-section of Merycopotamus nanus (YPM 31826). (G) Femur (RR 42), (H) humerus (RR 157), and (I) tibia (RR 301) of Indohyus. The (J) femur (H-GSP 30345), and (K) humerus (H-GSP 96227) of Ichthyolestes pinfoldi. (L) Tibial cross-section of Pakicetus attocki (H-GSP 30357). (M-O) femoral, humeral and tibial cross-sections of Tragulus (CMNH 17917). (P-R) femoral, humeral, and tibial cross-sections of Hyemoschus (AMNH 53613). (S-U) femoral, humeral, and tibial cross-sections of Alces (USNM 275127). (V-X) femoral, tibial and humeral cross-sections of Odocoileus (USNM 254653). (Y-AA) femoral, tibial, and humeral cross-sections of Moschus (USNM 259384). (BB-DD) femoral, tibial and humeral cross- sections of Bos (USNM 277262). (EE-GG) femoral, tibial, and humeral cross-sections of Bubalis (AMNH 54765). (HH-JJ) femoral, tibial and humeral cross-sections of Tragelaphus scriptus (USNM 164560). (KK-MM) femoral, tibial, and humeral cross-sections of Tragelaphus spekei (USNM 164558). (NN-PP) femoral, tibial, and humeral cross-sections of Kobus leche (USNM 254927). Images scaled to the same height for comparisons.

Fig. 15. Radiographs through the long bone midshafts of fossil cetaceans from the archaeocete family Remingtonocetidae. Both Kutchicetus minimus ((A) femur (IITR-SB 2647.41), (C) humerus (IITR-SB 2647.42), and (E) tibia (IITR-SB 2647.44)) and Andrewsiphius sloani ((B) femur (IITR-SB 2871.201) and (D) humerus (IITR-SB 2871.15)) were recovered from western India and are sister taxa (Thewissen and Bajpai, 2009). Although both taxa display osteosclerosis in their limb bones, those of Kutchicetus bear a larger cortex and a more compact medullary cavity.

Bubalis (Fig. 14FF)). Instead of having a vacuous or spongy bone-filled medullary cavity, pakicetids and Indohyus increased the relative cross-sectional area of the cortex, thus creating a reduced medullary cavity.

In most sampled taxa, the tibial cortex was thickest, while the humerus displayed the relatively thinnest cortex. Only a few taxa deviate from this typical pattern. The tibia of

Indohyus displayed a relatively thin cortex, and its humerus displayed the thickest cortex (Fig.

14G-I), although these specimens were not from a single individual of Indohyus.

Merycopotamus (Fig. 14D-F) also deviated from the typical pattern as its femur displayed the thinnest cortex.

Quantification of Bone Cross-sectional Area

Using the program Bone Profiler (Girondot and Laurin, 2003), P was calculated, along with other parameters S, Min and Max for each bone (Table 4). The parameter P was expressed in relation to body mass for some representative taxa. Results indicate most terrestrial artiodactyls have fairly consistent cortical dimensions with the greatest deviations in cortical thickness occurring in cetaceans and hippopotamids. Results showed that within the femur, both cetaceans (i.e., Ichthyolestes and Andrewsiphius) and Hippopotamus exhibited a greatly thickened cortex relative to other sampled taxa (Fig. 16A,B). However, the humerus of

Hippopotamus was only slightly thickened compared to most artiodactyls. The humeral cortex of cetaceans (i.e., Ichthyolestes and Kutchicetus) displayed the greatest thickness relative to all other sampled taxa (Fig. 16C,D). Finally, tibial cortices were thickest in fossil cetaceans (i.e.,

Pakicetus and Kutchicetus) and Hippopotamus (Fig. 16E,F). 81

Fig. 16. Double natural logarithmic plots of P (measure of cortical thickness) of the (A) femur, (C) humerus, and (E) tibia against body mass (in grams). P is inversely correlated with cortical thickness (e.g., as P becomes more negative, cortical thickness increases). 82

Cladogram of cetartiodactylans consistent with Geisler and Theodor (2009), Marcot (2007), and Boisserie et al., (2005b), with the end points of the branches indicating size increase of the bony cortex expressed as the residuals from the regression for P (compare to A, C, and E) of the (B) femur, (D) humerus, and (F) tibia. Results show that cetaceans and hippopotamids consistently have thicker long bone cortices relative to body mass compared to other cetartiodactylans.

The parameter S calculated in Bone Profiler (see Table 4) is proportional to the relative cross-sectional width of the transition between medullary and cortical regions (Girondot and

Laurin, 2003). Most taxa have a vacant medullary cavity and therefore had low S-values (0.01-

0.03). However, a few taxa displayed medullary cavities that were partially filled with spongy bone. Within the femur, the following taxa displayed the greatest S-values and accordingly had a relatively greater abundance of spongy bone: Hippopotamus (S=0.05), and Andrewsiphius

(S=0.04). Within the humerus, large S-values were calculated in Bos (S=0.058), Bubalis

(S=0.064), Hippopotamus (S=0.073), Ichthyolestes (S=0.074), Kutchicetus (S=0.23), and some specimens of Merycopotamus (YPM 16911 (S=0.043), YPM 49877 (S=0.061), YPM (S=0.04)).

Within the tibia, fewer taxa displayed spongy bone in the medullary cavities: Hippopotamus

(S=0.051), Kutchicetus (S=0.088), Pakicetus (S=0.036).

The parameter Min corresponds to the bone compactness at the center of the medullary region. Min values close to zero indicate a lack of bone in the medullary region, while higher values describe the abundance of bone in this region. Most sampled taxa had Min values near zero (Table 4). Within the femur, Hippopotamus (Min=0.7) and Andrewsiphius (Min=0.13) displayed the greatest Min-values. Humeral Min values were highest in Andrewsiphius

(Min=0.28), Kutchicetus (Min=0.65), and a single specimen of Merycopotamus dissimilus (YPM

16911, Min=0.3). 83

The parameter Max corresponds to the bone compactness at the most superficial layer of the bone cortex. All taxa displayed Max values close to or at 1.000 (Table 4).

Ancestral Character State Reconstruction

Ancestral character state reconstruction revealed little homoplasy in the evolution of cortical bone thickness among cetartiodactylans (Fig. 17). Character reconstruction of femoral

P-values (Fig. 17A) indicated that increased cortical bone thickness (osteosclerosis) probably evolved in the common ancestor of Indohyus and archaeocetes (i.e., Ichthyolestes and

Andrewsiphius). Osteosclerosis was documented in the femur of Hippopotamus, but was lacking in the fossil anthracothere Merycopotamus medioximus. Reconstruction of humeral P-values

(Fig. 17B) indicated that osteosclerosis evolved in the common ancestor of archaeocete cetaceans (i.e., Ichthyolestes and Kutchicetus), Indohyus, Hippopotamus, and anthracotheres

(i.e., Merycopotamus medioximus and M. dissimilus). Similar to the femoral results, character reconstruction of tibial P-values (Fig. 17C) indicated that osteosclerosis also evolved in the common ancestor of archaeocete cetaceans (i.e., Ichthyolestes and Kutchicetus), Indohyus,

Hippopotamus, and anthracotheres (i.e., Merycopotamus medioximus and Microbunodon). Taxa other than cetaceans, Indohyus, anthracotheres, or hippopotamids all displayed consistent P- values, resulting in few changes in cortical bone evolution. Within the femoral reconstruction

(Fig. 17), these taxa displayed P-values between 0.51 and 0.80, with the majority of taxa having values between 0.51 and 0.7. Only two taxa, Alces (P=0.72) and Moschus (P=0.78), displayed P- values greater than 0.7. Compared to the femoral reconstruction, humeral values were more variable as a greater number of taxa displayed values greater than 0.7. Those taxa with P-values greater than 0.7 and therefore the thinnest limb bone cortices include the following: 84

Fig. 17. Phylogenetic tree showing an ancestral character state reconstruction by squared change parsimony of P-values in fossil and extant cetartiodactylans utilizing a composite phylogeny consistent with Geisler and Theodor (2009), Marcot (2007), and Boisserie et al., (2005b). P is inversely correlated with cortical thickness. Results of the squared parsimony reconstruction are shown on each tree, and conflicting results from 85

arbitrarily ultrametricized reconstructions are shown as vertical lines through the main branches of the phylogeny. Average P-values are shown in parentheses for each taxon. Reconstructions are illustrated for the (A) femur, (B) humerus and (C) tibia. All trees indicate that increased cortical bone thickness may have evolved in the common ancestor of Indohyus and the earliest fossil cetaceans (i.e., Pakicetus, Ichthyolestes). Increased tibial and humeral cortical bone thickness probably evolved in the common ancestor of cetaceans, Indohyus, Hippopotamus, and anthracotheres (Merycopotamus, Microbunodon).

Potamochoerus (P=0.73); Moschus, Bos, and Capra (P=0.76); Tragelaphus spekei (P=0.72). In contrast, tibial P-values never exceeded 0.7. Instead, taxa outside the common ancestor of cetaceans, Indohyus, anthracothere, and hippopotamid branch P-values ranged between 0.4-

0.7, with most taxa displaying values between 0.51 and 0.7. Outliers include the following:

Hydropotes (P=0.48), Capreolus (P=0.42), Bubalis (P=0.46), Capra (P=0.46), Kobus leche (P=0.47).

These taxa have slightly thickened bones compared to most terrestrial and some of these taxa frequently invade a watery habitat (i.e., Hydropotes, Bubalis, Kobus).

Bone Thickness as a Predictor of Cetartiodactylan Habitat

Bone thickness among cetartiodactylans is an accurate predictor of ecological habitat for taxa that live the majority of their lives in and around water, or are completely aquatic.

Contrary to this, there appeared to be no consistent difference between the cortices of terrestrial taxa and those are mainly terrestrial but flee to the water as a refuge from predators, or occupy swamps and marshes (Fig. 18).

Within the femur, all taxa occupying a terrestrial habitat, or are known to frequently locomote in water (Table 5), all shared high P-values ranging from 0.53-0.78, while those taxa that spend the majority of their time in water had lower P-values (i.e., Hippopotamus and

Indohyus) (see Fig. 17). Both archaeocetes Ichthyolestes and Andrewsiphius displayed the 86

Fig 18. A summary of P-values calculated in this analysis with modern taxa grouped into “terrestrial,” “sometimes enter water” ranges of values. Because the Hippopotamus is the only semi-aquatic cetartiodactylan, its P-values are indicated with a star. For the (A) femur, (B) humerus, and (C) tibia, all fossil taxa are indicated by a black circle. 87

lowest femoral P-values (0.23 and 0.26 respectively). Ichthyolestes is thought to spend most of its time in water (Clementz et al., 2006), while Andrewsiphius is thought to occupy a marine niche (Thewissen and Bajpai, 2009).

Humeral P-values were also lowest in only those taxa that spend the majority of time in the water, or occupy a marine niche (Table 5). Most terrestrial taxa or those that frequently enter the water displayed humeral cortical values between 0.62 and 0.76 (Fig. 18). Hydropotes along rivers among tall reeds and rushes and is a powerful swimmer, but displayed a low P-value of 0.51. Taxa that spend a majority of their lives in water, had low P-values (i.e., Indohyus

(P=0.34), pakicetids (P=0.2)), except for Hippopotamus which displayed a P-value (0.49), close to that of Hydropotes. Kutchicetus, a taxon thought to inhabit a marine environment, had the lowest P-value (0.12).

Compared to the femoral and humeral P-values, those of the tibia were not as consistent with habitat designations. Most taxa that were either terrestrial or spent some time in water, displayed tibial P-values between 0.4 and 0.65 (Table 5; Fig. 18). However, Indohyus was hypothesized to occupy a freshwater habitat (Thewissen et al., 2007), but its P-value fell within the range of those terrestrial taxa or those that spend some time in the water, indicating that although Indohyus was known to be aquatic, its tibia was surprisingly thin. Hippopotamus and pakicetids spent a majority of their lives in the water, but were still able to locomote on land, and displayed P-values of 0.23 and 0.25. Kutchicetus is a marine archaeocete and displayed the lowest P-value of 0.14.

Discussion

Variation in Cetartiodactylan Long Bone Cortical Dimensions

CT scans of femoral, humeral, and tibial bones of cetartiodactylans facilitated visualization and comparisons of bone cross-sectional geometries. As noted by other studies, the bones of Hippopotamus (e.g., Wall, 1983, Madar, 1998) and pakicetids (Madar, 2007,

Thewissen et al., 2007) had exceptionally thick cortices and this study futhers their work by including a larger sample of these taxa and by utilizing a different methodology to quantify cortical dimensions. In contrast, most terrestrial taxa displayed relatively large medullary cavities, and reduced cortices. Beyond these phenotypic extremes, considerable variation was also documented in the bones of other taxa. The raoellid Indohyus was initially described as having osteosclerosis in femoral and humeral cortices (Thewissen et al., 2007), but because this study included a larger comparative sample, results indicated that a substantial thickening occurred only in the humerus. Most surprising was the thin cortex of the tibia of Indohyus. This study therefore substantiates the hypothesis of Thewissen et al., (2007) that Indohyus displayed osteosclerosis in its long bones, but limits that interpretation to the humerus and femur. Within the fossil family Anthracotheriidae, variation in cortical bone thickness was also documented.

Our sample was skewed with the anthracothere Merycopotamus having the greatest sample size, while only a tibia of Microbunodon was included. As such, within anthracotheres all variation in bone thickness was documented within Merycopotamus. In general these specimens displayed a slightly thickened humerus and tibia relative to terrestrial taxa. In the humerus, both M. dissimilus and M. medioximus displayed similar average P-values, while the 89

tibia of M. silistrense displayed a much thinner cortex than M. nanus. Femoral bone compactness was calculated only for M. medioximus.

An unexpected result was documented in the long bones of two remingtonocetid archaeocetes Andrewsiphius sloani and Kutchicetus minimus (Fig. 15). Although these coeval sister taxa were collected in ancient nearshore marine bays and swamps in close proximity to another, their body proportions differed with both roughly having an equivelant skull length, but the postcranial skeleton of Andrewsiphius was much longer (Thewissen and Bajpai, 2009).

Whereas the long bones of Andrewsiphius are robust, those of Kutchicetus are relatively gracile

(Thewissen and Bajpai, 2009). Osteosclerosis was documented in both specimens, but the medullary cavities of Andrewsiphius lacked an abundance of compact bone, while elements of

Kutchicetus displayed extremely compact medullary cavities (Fig. 18). Kutchicetus displayed the most extreme form of osteosclerosis in its femur, which lacked a medullary cavity. These data therefore indicate that both taxa utilized slightly different hydrostatic strategies to alter their buoyancy in water.

Bone Thickness is Somewhat Correlated with Habitat

P is an accurate indicator of aquatic habitat only in taxa with an extreme form of osteosclerosis (indicated by a low P-value). Hippopotamus is the only extant semi-aquatic cetartiodactylan and it spends its days in rivers, but feeds on land at night. Within extant taxa,

Hippopotamus displayed the lowest bone compactness values in all long bones. Isotopic and lithologic evidence taken from fossil pakicetids and Indohyus are most consistent with these taxa occupying an aquatic habitat (Thewissen et al., 1996, 2007, Roe et al., 1998, Clementz et al.,

2006). All sampled long bones of pakicetids certainly are consistent with an aquatic lifestyle, 90

and Madar (2007) also reported pakicetids displayed osteosclerotic vertebrae and some autopodal elements. However, only the humerus of Indohyus was thickened and is the sole morphological characteristic that suggested it occupied an aquatic lifestyle.

P is not, however, an accurate indicator of habitat for those taxa that occupied marsh or swamp habitats, or fled into shallow bodies of water to escape predation (Fig. 18). Instead, taxa with these habitats all displayed relatively high P-values as in terrestrial taxa. Therefore, bone thickness is effective at differentiating only those taxa that exhibit extremes in behaviors across the two environments (i.e. terrestrial vs. aquatic). As such, this study cannot reliably reconstruct the ancient habitat of anthracotheres, and isotopic evidence and further study are probably required.

Cortical Bone Evolution

Only incipient forms of hyperostosis were present in Indohyus and anthracotheres (Fig.

18). Ancestral character state reconstructions (Fig. 17) indicate that a slightly thickened femoral cortex probably evolved in the common ancestor of Indohyus and cetaceans. However, a thickened cortex probably evolved earlier in the common ancestor of Indohyus, cetaceans, anthracotheres, and Hippopotamus in both the femur and tibia. These results suggest taxa did not evolve uniformly systemic osteosclerosis, but rather that the evolution of osteosclerosis occurred in certain regions of the body before others.

Within the sample of modern taxa, no outstanding differences in cortical dimensions were found between terrestrial taxa (e.g. Odocoileus and Capra) versus those that flee into the water to escape predation (i.e. tragulids) versus those that occupy marsh or swamp habitats

(e.g., Alces and Tragelaphus spekei). Tragulids are known to flee into and remain underwater 91

for several minutes at a time, and have been filmed walking along the bottom, indicating they were not encumbered by positive buoyancy. No respiratory studies have shown if they regulate buoyancy by breathing, but just like humans, it could be that exhalation decreases buoyancy, thus allowing them to locomote at depth with relative ease. Tragelaphus spekei frequently submerges when foraging and it could also employ breath-regulated buoyancy. Alces usually maintains its head above water when swimming and therefore does not require any hydrostatic adaptations for negating its buoyancy. In contrast to both tragulids and Tragelaphus spekei,

Hippopotamus has a thick adipose layer that would increase buoyancy. In this case, exhalation and osteosclerotic bones may act in concert to counteract its buoyancy in water.

By combining results of this analysis with other published studies, it is now possible to obtain a synthetic view of cortical bone evolution documenting the land-to-sea transition of archaeocete cetaceans. Modern phylogenetic studies have identified mesonychid condylarths as distant relative of cetaceans (O’Leary and Gatesy 2008) and Indohyus as the closest fossil relative to archaic cetaceans (Thewissen et al., 2007; Geisler and Theodor, 2009). The skeleton of mesonychids lacks any cortical bone thickening, indicating it probably did not have an aquatic lifestyle (Madar, 1998). Therefore, osteosclerosis is first documented in the humerus of

Indohyus, but analysis of other bones in its skeleton (i.e., femur, tibia, caudal vertebra, metapodials, and ribs; this study, Chapter 2 of this dissertation) indicated a lacked of cortical thickening. Pakicetids, the most basal cetacean family, displayed systemic osteosclerosis as it was documented in the humerus, femur, and tibia (this study, Madar, 2007), as well as a caudal vertebra, ribs, and metapodials (Gray et al., 2007; Madar, 2007). Ambulocetids, the next diverging family of archaeocetes, displayed a modified form of osteosclerosis that was retained through much of the archaeocete lineage. The femur and radius of Ambulocetus displayed only 92

a slightly thickened cortex, but a medullary cavity infiltrated with trabecular bone (Madar, 1998,

Gray et al., 2007). Remingtonocetid cetaceans, the next diverging lineage of archaeocetes, display multiple hydrostatic strategies. Whereas the femur, humerus, and tibia of Kutchicetus and Andrewsiphius displayed extraordinarily thick cortices (this study), the femur of

Remingtonocetus and Dalanistes displayed medullary cavities filled with spongy bone (Madar,

1998). The ribs of Kutchicetus were pachyostotic and filled with thickened trabecular bone

(Gray et al., 2007). Protocetidae, the first cosmopolitan lineage of archaeocetes, and the femur of Rodhocetus displayed a thin cortex, and a medullary cavity filled with fine trabeculae (i.e.,

Rodhocetus, Madar, 1998), while the ribs of Gaviacetus displayed a slightly thickened cortex but were filled with trabecular bone (Gray et al., 2007). Basilosaurids were the last lineage of archaeocetes and multiple analyses of their skeletal dimensions revealed diverse hydrostatic strategies. The femur of Basilosaurus displayed a thick cortex and a medullary cavity filled with trabecular bone, while the radii of both Dorudon and Ancalacetus exhibited a thin cortex and the medullary cavities were filled with diffuse trabeculae (Madar, 1998). Ribs of the basilosaurids Zygorhiza and Dorudon also displayed thickened cortices and medullary trabeculae

(de Buffrénil et al., 1990; Uhen, 2004; Gray et al., 2007). Modern cetacean (whales, dolphins and porpoises) display osteoporotic limb elements, but their vertebrae are similar in density and thickness to terrestrial mammals (Felts and Spurrell, 1965, 1966, de Buffrénil 1985, de Buffrénil

1986, de Buffrénil and Schoevaert 1988).

Taken all together, these results indicate that an aquatic lifestyle probably evolved in the common ancestor of Indohyus, cetaceans, Hippopotamus, and anthracotheres. Concomitant with this change in paleoniche, osteosclerosis first evolved in a few skeletal elements in the sister to cetaceans, Indohyus, but cetaceans developed an initial systemic hyperostosis. The 93

most basal cetaceans, pakicetids, displayed the greatest and most skeletally diffuse bone thickness of any cetartiodactylan. Subsequent lineages of cetaceans decreased the amount of bone in the appendicular skeleton and ribs, relative to pakicetids, but retained enough skeletal ballast to successfully invade the marine environment. By the end of the Eocene epoch, cetaceans were obligatorily aquatic as they had lost their ability to locomote on land, and subsequently lightened their skeleton to reduce inertia while locomoting in a fluid environment.

Therefore, by releasing developmental constraints on bone thickness, cetaceans utilized a variety of hydrostatic strategies to successfully invade shallow water environments, and then reduce skeletal mass to facilitate easier aquatic locomotion. CHAPTER IV

EVOLUTION OF THE APICAL ECTODERM IN THE DEVELOPING VERTEBRATE LIMB

Introduction

Vertebrate limbs display an incredibly diverse array of skeletal morphologies, including the fins of teleosts and lungfish, wings of birds and bats, and flippers of dolphins. Due to this morphological diversity, limbs are a topic of intense study in paleontology, phylogenetic systematics, descriptive embryology and functional morphology. Evolutionary developmental biology (evo-devo) in particular has focused on understanding the developmental pathways that establish diverse limb phenotypes by integrating data from gene expression and protein signaling with transplantation and ablation experiments. As a result of the increase in the number of evo-devo studies on diverse taxa, additional variants in the limb developmental pathway have been discovered.

Limb evo-devo research attempts to explain how the signaling centers, within the developing vertebrate limb, control patterning and how phenotypic and expression variation within these signaling centers shape vertebrate limb morphology. Limb development is controlled by two main signaling centers: (1) the apical ectoderm of the limb, which is a specialized region of cells at the limb tip that controls outgrowth and patterning along the

95 96

proximodistal axis, and (2) the zone of polarizing activity, which regulates patterning along the anteroposterior axis.

This study compares limb apical ectodermal morphologies and the associated signaling patterns across vertebrates. We initially review the morphology and function of the apical ectoderm. Then, secondly, we describe the expression of fibroblast growth factors (FGF) in the apical ectoderm as the primary signaling molecules. Lastly, we provide a taxonomically broad comparison of ectodermal morphologies and associated FGF expression patterns across vertebrates (bony fish, lungfish, amphibians, squamates, birds and mammals) and subsequently present new morphological and protein signaling findings of the pantropical spotted dolphin

(Stenella attenuata).

General Description of the Apical Ectodermal Ridge (AER)

The apical ectodermal ridge (AER) is aptly named as it typically takes on a ridge-like morphology in vertebrates that is nipple-shaped in cross-section (Saunders, 1948), and is usually composed of stratified or pseudostratified columnar epithelium tissue (Richardson et al., 1998).

Ridge morphology is most frequently studied in chicks (e.g., Saunders, 1948; Jurand, 1965; Rubin and Saunders, 1972; Pizette and Niswander, 1999; Talamillo et al., 2005) and mice (e.g., Jurand,

1965; Lee and Chan, 1991; Talamillo et al., 2005). The AER is a specialized thickened epithelium at the distal apex of a developing limb bud (and in lung fish and teleosts, a fin bud) that secretes morphogens necessary for limb outgrowth and patterning. The AER originates from the ectodermal tissues associated with the medial somatopleure (Michaud et al., 1997).

A detailed description of the morphology of the AER and speculations about its function

(based on transplantation and ablation experiments) were first reported in the chick (Gallus 97

gallus; for review see Saunders 1948, 1998), but the earliest mention of different apical thickenings occurred as early as 1879 (for review see footnote in Saunders (1998)). Early embryological descriptions reported the AER as an “ektodermkappe,” (an ectodermal cap;

Köllicker, 1879; Braus, 1906; Fischel, 1929), “epithelfalte,” (an epithelial fold), “randfalte,” (an edge fold), “epithelverdickung,” (an epithelial thickening; Fischel, 1929),

“extremitätenscheitelleiste,” an extremity crest (Peter, 1903), and a ring (Steiner, 1928;

Schmidt, 1898; O’Rahilly and Müller, 1985).

In this study, we conducted a taxonomically broad comparison of developing vertebrate limb morphologies and documented three apical ectodermal (AE) morphologies (Figure 19): (AE-

1) a thick, or prominent, ridge-shaped ectoderm; (AE-2) the apical ectoderm is only slightly thickened; and (AE-3) no thickening of the apical ectoderm (Fig. 19A-F). In both AE-1,2 the apical thickening is limited to the dorsoventral boundary of the developing limb. Therefore in dorsoventral sections, the ectoderm appears ridge-like, or slightly raised (Figure 19A, C), but in anteroposterior sections, the apical ectoderm is thickened along the distal aspect of the limb

(Figure 19B, D). Lastly, the regenerating limb blastema of amphibians develops an apical

“epithelial” cap (AEC, Figure 19G, H), which functions like an ectodermal cap of normal developing limbs to direct limb outgrowth. Within hours of limb amputation, epithelial cells migrate to the wound surface and proliferate to form a multilayered AEC (Christensen and

Tassava, 2000; Han et al., 2005). The AEC is necessary for limb regeneration, and is functionally homologous to the apical ectoderm in the tetrapod limb (Christensen and Tassava, 2000; Han et al., 2005), but displays several morphological differences (Table 6). The AEC covers the entire end of a limb stump (Figure 19D), whereas the AE in normal developing vertebrate limbs is localized to the dorsoventral boundary (Figure 19A; Christensen and Tassava, 2000). 98

Fig. 19. Schematics of apical ectodermal (AE) morphologies. (A,B) The typical ridge-shaped apical ectodermal ridge (AE-1) of most vertebrates as seen in mice and chicks in dorsoventral sections (A), and anteroposterior sections (B). (C,D) The less prevalent low-relief apical ectoderm (AE-2) that displays slight relief in dorsoventral (C) and anteroposterior sections (D) as seen in the pantropical spotted dolphin (Stenella attenuata). (E,F) A flattened apical ectoderm as reported in the normal developing limb of salamanders in dorsoventral (E), and anteroposterior (F) sections. (G,H) The regenerating salamander limb with an elongating blastema made of epithelial and mesenchymal tissues in dorsoventral (G) and anteroposterior (H) sections (modified from Christensen and Tassava, 2000). Abbreviations: apical ectodermal (AE), apical ectodermal cap (AEC), ectoderm (ecto), epithelium (epith), and mesenchyme (mes).

Furthermore, the AEC is made of 4-15 stratified layers of cells, whereas the AE of vertebrates is composed of only a single layer of pseudostratified cells (Christensen and Tassava, 2000) that may be overlain with non-stratified cells.

Table 6. Comparison Between the Apical Ectoderm of Vertebrates and the Apical Epithelial Cap of Regenerating Salamander (Urodele) Limbs.

apical ectoderm apical epithelial cap References (AE) (AEC) Origin of tissue ectoderm associated with the limb epidermis Michaud et al., 1997; medial somatopleure Christensen and Tassava, 2000 Function limb outgrowth regenerating limb outgrowth Christensen and Tassava, 2000 Gross low-to-ridgelike, uniformly smooth, Christensen and Tassava, 2000 morphology localized to dorsoventral broadly covers wound stump boundary Number of single pseudostratified layer of several stratified layers (4-15 Christensen and Tassava, 2000 stratified cell ectodermal cells topped with layers) layers additional non-stratified cells Basement Present absent Christensen and Tassava, 2000 membrane FGF expression throughout cell layers basal-most layer of cells, Han et al., 2001; Sun et al., 2002 underlying mesenchyme

Function of the Apical Ectoderm in the Developing Limb

While several experimental studies in the chick have revealed the function of the apical ectoderm in the developing limb, a specific organization of cells in the ectoderm is not essential for proper limb development and patterning (for review see Saunders, 1998). For example, AER removal causes a cessation of limb outgrowth for distal skeletal elements, resulting in limb truncation (Saunders, 1948). Alternatively, the addition of AER tissue to the terminus of a developing limb resumes distal outgrowth. Similarly, two limbs can be produced by the transplantation of an isolated AER that lacks associated dorsal and ventral ectodermal tissues, onto a limb bud that already possesses a normal AER (Saunders and Gasseling, 1968). Removal and replacement of ectodermal cells with an inverted ectodermal jacket also produces a normal limb (Errick and Saunders, 1974). If AER cells are removed, disassociated, mixed and then 100

placed on the distal limb bud, a normal limb develops (Errick and Saunders, 1974). Taken together, these experimental manipulations show that the apical ectoderm is required for limb development and proximo-distal outgrowth, but the organization of the ectodermal cells within individuals appears to be inconsequential for proper limb development.

Fibroblast Growth Factors (FGF) in the Apical Ectoderm Control Proximodistal Outgrowth and Limb Patterning

Fibroblast growth factors (FGFs) are expressed in the cells of the limb apical ectoderm of developing limbs (Mariani et al., 2008), and are a member of the heparin-binding growth factor family. They function in promoting cell survival and proliferation of undifferentiated mesenchymal cells (Niswander et al., 1994a,b; Hara et al., 1998; Ngo-Muller and Muneoka,

2000; Han et al., 2001; Niswander, 2002; Weatherbee et al., 2006) as well as specifying cell fate during digit formation (Mariani et al., 2008; Lu et al., 2008). Fgf4 (genes indicated by italicized text, whereas proteins are indicate by Roman font) , Fgf8, Fgf9, and Fgf17 are some of the many genes expressed in the apical ectoderm of developing limbs, but Fgf8 appears to be the most important for normal limb outgrowth (Niswander et al., 1994a,b; Mahmood et al., 1995; Vogel et al., 1996; Hara et al., 1998; Moon and Capecchi, 2000; Ngo-Muller and Muneoka, 2000; Sun et al., 2002; Talamillo et al., 2005; Verheyden and Sun, 2008) and is expressed in a higher concentration compared to other FGFs (Mariani et al., 2008). Ancillary FGFs (Fgf4, 9, and 17) are functionally redundant and can rescue limb outgrowth and patterning in the absence of Fgf8

(Hara et al, 1998; Moon and Capecchi, 2000; Niswander, 2002; Mariani et al., 2008; Verheyden and Sun, 2008).

Of the numerous FGFs present in the apical ectoderm, Fgf8 displays the greatest expression, with decreasing levels of Fgf4 (Niswander et al., 1994a) and Fgf9, while Fgf17 101

displayed the lowest expression levels (Mariani et al., 2008). Fgf4 expression begins later and ends earlier than Fgf8 expression and is only active in the presence of a thickened apical ectoderm (Niswander et al., 1994a; Moon and Capecchi, 2000; Boulet et al., 2004; Mariani et al.,

2008). Furthermore, Fgfs4 and 8 have an inverse relationship in that Fgf8 inhibits expression of

Fgf4 via Grem1 (Moon and Capecchi, 2000; Boulet et al., 2004; Verheyden and Sun, 2008).

However, if the thickened apical ectoderm is excised from a distal limb bud, or if Fgf8 is inactivated, Fgf4 can rescue limb outgrowth and patterning (Niswander et al., 1994a; Hara et al.,

1998; Boulet et al., 2004). If both Fgf4 and Fgf8 expression fails, apoptotic activity will increase, and the forelimbs will not develop, but the scapula will form (Sun et al., 2002; Boulet et al.,

2004). However, if just Fgf4 is inactivated, normal limbs will form and Shh expression levels are unaffected (Sun et al., 2002; Boulet et al., 2004).

Fgf8 expression is normally localized to apical ectodermal cells in order to signal to the underlying mesenchymal cells during limb outgrowth. However, a few notable exceptions have documented Fgf8 expression within limb mesenchyme (Vogel et al., 1996; Moon and Capecchi,

2000; Pizette et al., 2001; Weatherbee et al., 2006). Unlike the limbs of most amniotes, the developing forelimbs of bats express Fgf8 both in the apical ectoderm and the interdigital mesenchyme, presumably to direct outgrowth of the digits, and promote cell survival and proliferation of interdigital mesenchymal cells for generation of a wing membrane (Weatherbee et al., 2006). Furthermore, in the regenerating limbs of salamanders, Fgf8 expression was found in the basal-most layer of the apical ectoderm and the underlying mesenchymal tissues, also presumably to promote cell survival and proliferation (Han et al., 2001; Christensen et al., 2002).

FGFs produced by the limb apical ectoderm are necessary for the initiation and maintenance of Shh expression, and together FGFs and Shh create a positive feedback loop that 102

is essential for limb outgrowth, polarizing, and patterning (Niswander et al., 1994b; Vogel et al.,

1996; Niswander, 2002; Boulet et al., 2004, Panman et al., 2006; Tickle, 2006; Mariani et al.,

2008; Tabin and McMahon, 2008). This positive feedback loop is also connected to an

Fgf/Gremlin1 inhibitory feedback loop that terminates limb bud outgrowth (Verheyden and Sun,

2008). If either Fgf or Shh experiences a cessation in expression, a normal limb will not form. For instance, in primitive snakes the hindlimbs fail to form an apical ectodermal ridge with associated Fgf8 expression, preventing Shh expression, and causing a cessation of limb development (Cohn and Tickle, 1999). Adult snakes lack visible hindlimbs as they are vestigal and are contained in the body wall (Cohn and Tickle, 1999). In pantropical spotted dolphin embryos, both Fgf8 and Shh protein signals were present during incipient limb bud stages, but a later cessation in Fgf8 signaling (presumably concomitant with the hiatus of all ectodermally- derived FGF expression) arrests limb development. Adult dolphins lack external hindlimbs but incomplete hindlimb and pelvic girdle vestiges are encased in the body wall near the vertebral column (Figure 20; Thewissen et al., 2006) to various degrees and in different cetacean species.

By interrupting Fgf8 expression at different times during limb development, both snakes and dolphins convergently evolved a streamlined body with hindlimbs encased in the body wall.

Duration of Fgf8 expression appears to be essential for normal limb development, and is correlated with both phalangeal number and digit length. Increased duration of Fgf8 expression results in polydactyly (Vogel et al., 1996; Talamillo et al., 2005), inhibits terminal phalanx formation, and directs the development of supernumary phalanges (Sanz-Ezquerro and

Tickle, 2003; Richardson et al., 2004). Conversely, experimentally induced decreases in Fgf8 expression results in the generation of deformed limbs (Sun et al., 2002) with fewer skeletal elements (Vogel et al., 1996; Sun et al., 2002; Mariani et al., 2008) and increased apoptotic 103

activity (Moon and Capecchi, 2000; Boulet et al., 2004; Talamillo et al., 2005). If an Fgf8 inhibitor is present, premature formation of the terminal phalanx will occur, in some cases creating fewer numbers of phalanges (Sanz-Ezquerro and Tickle, 2003).

Fig. 20. Morphology of an approximately 110 day old pantropical spotted dolphin (Stenella attenuata, LACM 94285) fetus. (A) whole fetus, and (B) the same fetus clear and stained, revealing pelvic girdle and hindlimb remnants as well as hyperphalangy in the principal digits of the flipper. Clearing and staining completed by Dr. Sirpa Nummela. Scale bar equals 1 cm.

104

Molecular Pathways Determining Apical Ectoderm Morphology

Through study of modern taxa, such as chicks and mice, some of the developmental pathways creating a ridge-shaped ectoderm are well-known. These findings may offer insight into the mechanisms that may inhibit formation of a ridge-like AER, and allow for development of a low or flattened apical ectoderm (AE-2,3).

The AER lies between the dorsal and ventral ectodermal surfaces of the limb bud

(Kimmel, et al., 2000). Cells of the adjacent dorsal ectoderm display Wnt7a and Lmx1, while the homeobox transcription factor EN1 is expressed in the ventral ectoderm (Talamillo et al., 2005).

If EN1 is misexpressed or the dorsoventral border is lost, the distinctive ridge shape of the AER is lost, resulting in a flattened apical ectoderm (Kimmel et al., 2000). Furthermore, if Fgf8 expression was also interrupted along the AER when EN1 was misexpressed, it leads to missing and in some cases ectopic digits (Kimmel et al., 2000).

Bone morphogenic proteins (BMPs) also play a key role in regulating the height of the apical ectoderm (Ahn et al., 2001; Pizette et al., 2001). Inhibition of BMP signaling, through application of the BMP antagonist Noggin, at early stages of chick limb development resulted in an increase in AER height (Pizette and Niswander, 1999). Gremlin1 also regulates AER height by inhibiting BMP (for review see Fernandez-Teran and Ros, 2008).

The Wnt/β-catenin pathway lies upstream of BMP signaling and associated patterning, and is essential to establishing mouse AER morphology (Barrow et al., 2003; Narita et al., 2005;

Lu et al., 2008). A Wnt/β-catenin/Fgf regulatory loop was found to be essential to the establishment and survival of a morphological AER in mice, and Wnt3 was a key signal regulating

AER thickness in this organism (Barrow et al., 2003). Mouse mutants with disrupted Wnt3 expression displayed a 50% reduction in dorsoventral thickness of the AER and variably 105

displayed fewer limb skeletal elements (Barrow et al., 2003). However, Fgf8 expression was only mildly affected and was localized to only those ectodermal cells that were slightly thickened (Barrow et al., 2003). Wnt3a carried out a similar role in chicks (Kengaku et al., 1998;

Niswander, 2002).

Materials and Methods

Embryonic specimens of the pantropical spotted dolphin (Stenella attenuata) were supplied by the Los Angeles Museum of Natural History. Embryos were immersion-fixed, preserved in 70% ethanol and stored without refrigeration for time periods ranging from 15 to

32 years. The embryos were staged according to a modified version of the Carnegie system

(Thewissen and Heyning, 2007). The immunohistochemical data are based on six dolphin embryos (LACM), ranging from Carnegie Stage 13 to Carnegie Stage 19. Each embryo was embedded in paraffin, and sectioned at 6-μm. Protocols were optimized with immersion-fixed, ethanol-preserved mouse embryos. Non-limb embryonic dolphin tissue was then tested and optimized. Because of the variance in fixation and storage times, slightly different procedures were used for different specimens to obtain optimal results. In addition, negative control samples (minus primary antibody) were used to determine the level of background staining for all experiments.

The following antibodies were used in this study: anti-fibroblast growth factor-8 (Fgf-8;

Santa Cruz Biotechnology; sc-6958); anti-fibroblast growth factor-4 (Fgf-4; Santa Cruz

Biotechnology; sc-1361).

106

Results

Comparative Morphology of the Apical Ectoderm Among Vertebrates

The apical ectodermal ridge (AE-1; Figure 19A,B) was fully developed in most vertebrates studied to date (Hanken et al., 2001), although rare exceptions in AE shape have been documented in tetrapods (Table 6). This study conducted a broad literature review, focusing primarily on morphological variation within the distal limb ectoderm of vertebrates

(fish, lungfish, amphibians, squamates, birds, and mammals; see Table 6). Additionally, morphology of the apical ectoderm of the limb, and if possible, patterns of Fgfs 4 and 8 gene expression or protein signal localization were also investigated (Table 6). We first describe a typical vertebrate apical ectodermal ridge (AE-1), and then discuss variations, such as the low- relief apical ectoderm of dolphins (AE-2). For some vertebrates, such as amphibians and teleosts, limb development is quite different from model organisms (chicks, mice), which are subsequently discussed in detail.

An Apical Ectodermal Ridge (AE-1) is the Typical Ectodermal Morphology of Vertebrates Teleosts

The fins of some teleost fish display an apical ectodermal ridge (AE-1) and typical patterns of Fgf8 expression (e.g., zebrafish (Danio rerio), (Reifers et al., 1998; for review see

Mercader, 2007)), even though they possess fins with rays (lepitotrichia) that lack the appendicular elements seen in the vertebrate autopod, and evolved the ability to regenerate parts of pectoral fins (Poss et al., 2000; Galis et al., 2003). The teleosts apical ectodermal ridge

(AE-1) does not undergo apoptosis, as in most tetrapods (Wood, 1982), but instead folds and elongates distally to form a fin fold (Figure 21). The fin fold then grows distally until a 107

semicircular swimming paddle develops (Wood, 1982). Fin fold cells form dorsal and ventral layers, express similar markers, and perform similar functions as the tetrapod AER (Mercader,

2007).

Fig. 21. Schematic of the transition from an apical ectodermal ridge (AE-1) to a fin fold in the killifish (Aphyosemion scheeli, modified from Wood, 1982). (A) Apical ectodermal ridge (AE-1) is present at ~128 hours. (B) At 135 hours, the apical ectoderm is folded, caused by differential mitosis in the dorsal and ventral ectoderms. (C) At 144 hours, a fin fold has formed and is made of a tightly appressed ectodermal bilayer. Not to scale.

Some teleosts have detailed descriptions of their apical ectoderms documenting how these ectoderms differ from those of most tetrapods. For instance, killifish (Aphyosemion scheeli) have a morphologically distinct apical ectoderm (AE-1) that, relative to the fin bud, is larger than the apical ectodermal ridge (AE-1) of most tetrapods (Wood, 1982). The killifish apical ectoderm spans the entire distal margin of the fin bud along the anteroposterior plane

(Wood, 1982; Grandel and Schulte-Merker, 1998). Compared to non-apical ectodermal cells, those cells within the basal layer of the apical ectoderm are elongated and pseudostratified.

The AE-1 of trout taxa Salmo trutta fario and S. gairdneri exhibits an area of elevated ectoderm

(Bouvet, 1968) made of pseudostratified columnar cells (Géraudie and François, 1973; Géraudie,

1978; Wood, 1982). 108

Lungfish

Lungfishes possess an AER and Fgf8 protein signaling similar to most vertebrates.

However, lungfishes also are able to regenerate both the soft tissues and skeletal elements of their fins (Galis et al., 2003). The distal fin bud ectoderm of the lungfish Neoceratodus is initially a stratified bilayer of cuboidal cells covered by a squamous periderm, and after a period of outgrowth, a morphological AER emerges (Hodgkinson et al., 2007). This AER is composed of pseudostratified columnar cells in the basal membrane of the distal fin bud epithelium

(Hodgkinson et al., 2007). Fgf-8 protein signals have also been documented in the Neoceratodus

AER (Hodgkinson et al., 2007), indicating protein signals consistent with most vertebrates during limb development.

Squamata

The garden lizard (Calotes versicolor) has an AER that is nipple-shaped in cross-section, much like that of most tetrapods (Goel and Mathur, 1977).

Chiroptera

Like most vertebrates, the developing limb of bats displays both an AER and associated

Fgf8 expression. The bat AER is initially present over the entire anterior-posterior distal aspect of the developing handplate, but as digits of the forelimb elongate, the AER and associated gene expression become localized over digits II and III (Weatherbee et al., 2006). Compared to that of similar-aged mice, the Fgf8 AER expression domain of bats is approximately three times wider than that of model taxa (Cretekos et al., 2007).

Bats also display a novel domain of Fgf8 expression in the interdigital tissues between digits II-V, presumably to aid in the proliferation and survival of these tissues during wing 109

membrane formation (Weatherbee et al., 2006). By altering the domain of Fgf8 expression both in the AER and interdigital tissues, bats display relatively elongated metacarpals and phalanges

(Sears et al., 2006) that are connected by a thin wing membrane (Cretekos et al., 2001;

Weatherbee et al., 2006; Sears, 2008). Mesenchymal expression of Fgf8 is similar to that of axolotls, but is unique compared to most amniotes (Weatherbee et al., 2006).

Unusual Vertebrates Lacking An Apical Ectodermal Ridge (AE-1)

Pantropical spotted dolphin (Stenella attenuata)

Descriptive embryological studies document that the forelimb of the pantropical spotted dolphin (Stenella attenuata) first begins as a typical mammalian handplate that is as long as it wide until about 28 days gestation (Richardson and Oelschläger, 2002). Five digital condensations form (Sedmera et al., 1997), and in some cases, a low-relief epithelial thickening appeared along the distal aspect of the limb bud (Richardson and Oelschläger, 2002). After 30 days gestation a weakly organized and thickened epithelium is present at the ends of the central digits II and III (Richardson and Oelschläger, 2002). Toward the end of the embryonic period, near 48 days gestation, digits II and III have an increased number of phalanges relative to the other digits, creating a chisel-shaped flipper in lateral view, and the thickened ectoderm is localized to the ends of these developing digits (Richardson and Oelschläger, 2002).

In the developing Stenella forelimb, neither gross anatomical or sectioned fore- or hindlimbs display an apical ectodermal ridge (AE-1) at any examined ontogenetic stage. Instead, the distal apexes of the limb buds are encapsulated by a thickened ectoderm that has a smoothed contour (Fig. 19, AE-2). This ectodermal morphology is consistent with the morphology of the apical ectodermal cap reported in the tree frog Eleuthrodacylus coqui, as well as the modest 110

limb ectoderm of Xenopus. At approximately 26 days gestation (Carnegie Stage 15), the dolphin limb is conical-shaped in lateral view and the entire distal aspect is lined with a transparent ectoderm. In anterior view, a low-relief apical rise is apparent between the dorsal and ventral surfaces of the limb bud. This apical ectoderm is 2-4 cell layers thick. Cells of the basal layer are elongated and columnar, while cells of the second layer from the bottom are rounded and approximately half the height of the basal cells. The one or two apical layers consist of slightly flattened cells. Fgf8 protein signals are localized along the superficial apical layers as well as the dorsal surface of the apical ectoderm.

By ~35 days gestation (Carnegie Stage 17), both digits II and III have elongated considerably creating a blunt-ended limb bud when viewed laterally. In cross-section, a distinct apical ridge is absent; instead only a low-relief rise of epithelium occurs at the border between the dorsal and ventral surfaces (Fig. 22). Cross-sections through the distal aspect of the limb bud revealed a dorsoventrally narrowed and thickened ectoderm, relative to that of previous ontogenetic stages. The apical ectoderm at this stage consists of tightly packed cells that are slightly elongated and elliptical in shape; however, only a dorsoventral narrow portion of this ectoderm displays a tiny additional cell layer.

At approximately 48 days gestation, at the end of the embryonic period (Carnegie Stage

19; Fig. 22), the flipper is almost entirely formed, with digit II the longest, followed closely by digit III. The other digits are considerably shorter. Five digital condensations are clearly visible and each digit displays obvious interphalangeal joints. Although ectodermal tissues are thickened along the ends of digit II, no distinct apical ectodermal ridge is present. In anterior view, this region of ectoderm appears as a slight thickening with little relief as compared to the adjacent dorsal and ventral ectodermal tissues. Cross-sections through the distal end of digit II 111

showed no apical thickening or change in cell shape compared to the non-apical ectoderm. Fgf4 protein signals were localized to the ends of this digit, indicating an expression pattern consistent with other tetrapods.

Fig. 22. Apical ectodermal morphologies and associated protein signaling in the forelimbs of (A- D) pantropical spotted dolphins [LACM 94613, Carnegie Stage 15 (A); LACM 94594, Carnegie Stage 15 (B); LACM 94770, Carnegie Stage 16 (C); LACM 94817, Carnegie Stage 19 (D)]. This taxon is unusual among most tetrapods in it displays a flattened apical ectodermal (A-D), but localizes Fgf8 (B,C) and Fgf4 protein signals (D) to the distal limb ectoderm. Scale bars are 100 μm in length.

112

Marsupalia

Morphology and patterns of gene expression in the short-tailed opossum (Monodelphis domestica) fore- and hindlimb AER are currently under study (Sears, pers. comm.). Preliminary results indicate the developing forelimb of these taxa lacks a typical apical ectodermal ridge characteristic of vertebrates, and instead possess a low-relief (AE-2) and disassociated apical ectoderm with few cell layers (Sear, pers. comm.). Forelimb apical ectoderm expresses Fgf-8 much earlier (stage 25) than that of the hindlimb (stage 31) (Smith, 2003).

Amphibia

Urodeles

The urodele limb is exceptional by undergoing direct growth of the digits as independent buds off the limb bud (Von Dassow and Munro, 1999; Franssen, et al., 2005).

Morphogenic differences in limb development among amphibians suggests polyphyly (Hanken,

1986; Von Dassow and Munro, 1999; Franssen, et al., 2005) because the pattern of urodele limb development could have evolved separately from that of other amniotes including Anura

(Holmgren, 1933; Jarvik, 1965; Franssen, et al., 2005).

Normal urodele limb buds lack an apical ectodermal thickening of the developing limb

(AE-3; Galis et al., 2003; Franssen, et al., 2005; Han et al., 2005), but regenerating limbs possess an apical ectodermal cap (AEC). Both the normal and regenerating limb bud express Fgf8 in the apical ectoderm (Christensen et al., 2002). The basal layer of the AEC functions much like the amniote AER during normal limb development (Onda and Tassava, 1991; Christensen and

Tassava, 2000; Galis et al., 2003; Franssen et al., 2005). 113

In the Mexican axolotl (Ambystoma mexicanum), Fgf8 expression appeared to be different relative to that of amniotes. Fgf8 expression was detected in both the developing limb bud and the AEC of a regenerating limb blastema (Han et al., 2001; Christensen et al., 2002).

Before digit formation, Fgf8 expression in Ambystoma is localized in the epithelium, however a gradual translocation of Fgf8 expression to the underlying mesenchymal tissue occurs. This expression is unlike that of Xenopus, chicks, and mice where Fgf8 expression is isolated to the

AER (Han et al., 2001). Similarly, in the AEC of regenerating limbs of Ambystoma, Fgf8 is expressed in the basal-most layer of the AEC and the underlying mesenchymal tissues, further suggesting that the basal-most layer of the AEC is functionally equivalent to the amniote AER

(Han et al., 2001; Christensen et al., 2002). Fgf4 was found to only be expressed slightly in the developing limb and was absent from the AEC (Christensen et al., 2002).

Anura

During limb development, the metamorphosing African clawed frog (Xenopus laevis) displays a low-relief apical ectoderm (AE-2) that consists of three layers of ectodermal cells

(Tarin and Sturdee, 1971). The Xenopus apical ectoderm, although ridge-shaped as in amniotes, does not display as great relief, and is considered a “modest” apical ectodermal ridge in comparison to amniotes (Tarin and Sturdee, 1971). Fgf8 is expressed in the distal tip of the developing Xenopus hindlimb but in only the epithelium, whereas in a regenerating limb, Fgf8 expression was localized to both the mesenchyme and basal-most layer of the AEC (Han et al.,

2001).

The large neotropical treefrog (Eleutherodactylus coqui) directly develops as a froglet as it does not proceed through a tadpole stage (Richardson, 1995). Its limb buds appear at an 114

earlier ontogenetic stage as compared to metamorphosing species (e.g., indirect developers like

Xenopus) (Richardson, 1995; Richardson et al., 1998; Hanken et al., 2001; Bininda-Emonds et al.,

2007). E. coqui has been the subject of study because it can form a normal vertebrate limb in the absence of a morphological AER (Richardson, et al., 1998; Hanken et al., 2001). The absence of an AER in E. coqui is autapomorphic (Richardson et al., 1998). Instead of the apical ectoderm being ridge-shaped with a high-relief, the ectoderm of E. coqui is a low-relief thickened ectodermal cap along the limb apex (Richardson et al., 1998; Hanken et al., 2001).

Excision of the hind-limb apical ectoderm of E. coqui resulted in loss and/or fusion of the distal limb elements (Richardson et al., 1998), suggesting the apical ectoderm played a role in controlling limb outgrowth, and functioned much like the AER of amniotes. However, truncation of the distal limb elements was not observed (Richardson et al., 1998), possibly because the ectoderm was partially regenerated.

Discussion

Morphology of the Vertebrate Limb Apical Ectoderm

This study documents that a ridge-like apical ectoderm (apical ectodermal ridge, AE-1) along the apex of a developing limb is the most common morphology for vertebrates as it is present in model taxa (e.g., mice and chicks) and several other lineages of vertebrates (Table 7).

Presence of a ridge-like apical ectoderm in teleosts and lungfishes suggests that the ridge morphology is the primitive condition among vertebrates and evolved before the transition from a fin to a limb in the earliest tetrapods. During fin development in teleosts, the apical ectodermal ridge remains active and morphs from a ridge, to a layered fin fold (Fig. 21), and finally into a swimming paddle (Wood, 1982). In contrast, the apical ectodermal ridge of most Table 7. Patterns of Morphological Variation and Fibroblast Growth Factor (Fgf) Expression in the Limb Ectoderm of Developing Vertebrates During Normal Development. Apical Ectodermal (AE), Apical Epithelial Cap (AEC), Regenerating Limb (rg).

Order Taxon Common Apical Ectoderm Apical Ectodermal References Name Morphology Expression Cypriniformes Danio rerio zebrafish AE-1 Fgf8 Grandel and Schulte-Merker, 1998; Reifers et al., 1998; Mercader, 2007 Cyprinodontiformes Aphyosemion scheeli killifish AE-1 Wood, 1982 Ceratodontoiformes Neoceratodus forsteri Australian lungfish AE-1 Fgf8 protein Hodgkinson, et al., 2007 Urodeles Ambyostoma mexicanum Mexican axolotl AE-3 Fgfs4,8 Han et al., 2001; rg rg rg AEC Fgf8 , lacks Fgf-4 Christensen et al., 2002 Anura Eleutherodactylus coqui Treefrog AE-2 DLX Fang and Elinson, 1996; Richardson et al., 1998 Anura Xenopus laevis African clawed frog AE-2 Fgf8 Tarin and Sturdee, 1971; Fang and Elinson, 1996; rg rg AEC Fgf8 Christen and Slack, 1997 Chelonia Chelonia mydas green turtle AE-1 Milaire, 1957; Vasse, 1972; Chelonia depressa flatback turtle AE-1 Miller, 1985 Caretta caretta loggerhead turtle AE-1 Eretmochelys imbricata hawksbill turtle AE-1 Lepidochelys olivacea Pacific ridley turtle AE-1 Dermochelys coriacea leatherback turtle AE-1 Emys orbicularis European pond turtle AE-1 Testudo graeca Greek tortoise AE-1 Pseudemys pond turtle AE-1 Squamata Lacerta vivipara common lizard AE-1 Milaire, 1957; Calotes versicolor garden lizard AE-1 Dufaure and Hubert, 1961; Chamelaeo chameleon AE-1 Goel and Mathur, 1977 Mabuya long-tailed skink AE-1 Crocodilia Alligator mississippiensis American alligator AE-1 Honig, 1984; Ferguson, 1985 Crocodylus porosus saltwater crocodile AE-1 Crocodylus johnsoni freshwater crocodile AE-1 Marsupialia Monodelphis domestica short-tailed opossum AE-2 Fgf8 Smith, 2003; Sears, pers. comm. Rodentia Mus musculus mouse AE-1 Fgfs4, 8 e.g., Sun et al., 2002; Boulet et al., 2004 Galliformes Gallus gallus chick AE-1 Fgfs 4, 8 e.g., Niswander et al., 1994a; Kengaku et al., 1998; Narita et al., 2005 Chiroptera Carollia perspicillata short-tailed fruit bat AE-1 Fgf-8 Weatherbee et al., 2006; Cretekos, et al., 2007; Sears, 2008 Primata Homo sapiens human AE-1 Bardeen and Lewis, 1901; Steiner, 1929; O’Rahilly et al., 1956; Kelley, 1973; O’Rahilly and Müller, 1985;

Hallgrímsson et al., 2002 115 Cetartiodactyla Stenella attenuata pantropical spotted AE-2 Fgfs4, 8 protein this study

dolphin

116

tetrapods is only transitory during embryogenesis. The tetrapod AER will undergo apoptosis first along the interdigital spaces, then at the ends of developing digits (Fernandez-Teran and

Ros, 2008). Therefore, the apical ectodermal ridge is most common among vertebrates, but its function has reduced in the development of tetrapods.

Unrelated lineages of vertebrates (i.e. amphibians, cetaceans, and marsupials), have convergently evolved a low-relief apical ectoderm of the developing limb. In these groups, the limb apical ectoderm either lacks a thickening (AE-3, salamanders), or only displays a slight thickening (AE-2, dolphins, anurans) (Figure 19, Table 7). A low-relief apical ectoderm is also present the marsupial developing forelimb (Table 7, Sears pers. comm.), but unlike the apical ectoderm of most vertebrates, this apical ectoderm is discontinuous, suggesting that this morphology is also an autapomorphy.

Regardless of the gross morphology of the apical ectoderm (high- vs. low-relief), all taxa included in this analysis displayed normal fin or limb development, indicating that gross morphology of the apical ectoderm does not affect its function. Furthermore, correlations between proper apical ectoderm function and its cellular organization are uninformative, as experimental manipulations of cellular distribution show no effect on function (Saunders, 1948,

1998; Saunders and Gasseling, 1968; Errick and Saunders, 1974). Only removal of apical ectodermal cells negatively altered function (Saunders, 1948, 1998; Saunders and Gasseling,

1968; Errick and Saunders, 1974). A morphological definition of the apical ectoderm is useful for comparative studies and tracing evolutionary transformations, but correlations between function and morphology are dubious. A normal limb apical ectoderm is probably best described by a molecular (signaling) criterion. 117

1998; Saunders and Gasseling, 1968; Errick and Saunders, 1974). Only removal of apical ectodermal cells negatively altered function (Saunders, 1948, 1998; Saunders and Gasseling,

The apical ectoderm of a properly developing limb, regardless of its morphology, secretes morphogens (e.g., FGFs) that control limb outgrowth and digital patterning. We chose the presence FGFs as a molecular indicator of an active limb apical ectoderm as they are the foundation of several pathways involved in limb outgrowth, patterning, and arrest of growth, and their expression is consistent among taxa with varying limb apical ectodermal morphologies

(Table 7) (e.g. Barrow et al., 2003; Verheyden and Sun, 2008). This study documented that all taxa expressed FGFs within the apical ectoderm of normal developing limbs (Table 7). Our results therefore indicate that a molecular definition of an active limb apical ectoderm is a conservative and reliable alternative to a morphological definition.

118

Potential Correlates of Limb Apical Ectoderm Height

Thickness of the limb apical ectoderm is directly related to the number of cells populating that region of tissue. Constituent cells of the normal limb apical ectoderm include those signaling for limb development as well as apoptotic cells. These apoptotic cells are distributed throughout the limb apical ectoderm and are absent from adjacent ectodermal tissues (Fernandez-Teran and Ros, 2008). Their presence has been documented in the limb apical ectoderm of the chick and mouse throughout its lifespan (Jurand, 1965; Todt and Fallon,

1984; Fernandez-Teran and Ros, 2008), however little is known of abundance of these cells in non-model taxa, including those presented here. It could be that the ratio of cells signaling for limb growth and patterning versus apoptotic cells is different between taxa with high-relief (i.e., fishes, lungfishes, chelonians, squamates, crocodilians, chicks, mice, chiropterans, and primates) versus low-relief (i.e., marsupials, cetaceans, and amphibians) limb apical ectoderms.

Alternatively, the ratio of different cell types may be equivalent across these taxa, but regulated differently via those genes directly affecting height of the limb ectoderm (e.g., BMP, Noggin,

Gremlin). Activity level of apoptotic cells has been shown to be a chief determinant of the rate of epithelial morphogenesis, and could directly affect rate of limb outgrowth and digital development (Davidson, 2008; Toyama et al., 2008). Indeed, taxa with low-relief apical ectoderms (i.e., marsupials, cetaceans, and amphibians) have become the topics of intense study as their limb development is either significantly delayed or precocial relative to most vertebrates (McCrady, 1938; Richardson, 1995; Richardson and Oelschläger, 2002; Galis et al.,

2003; Smith, 2003; Sears, 2004; Keyte et al., 2006; Bininda-Emonds et al., 2007). It may be that the abundance and/or activity of apoptotic cells in the apical ectoderm plays a significant role not only in shaping the limb apical ectoderm, but also directing the rate of limb development. CHAPTER V

DEVELOPMENT OF DOLPHIN FLIPPERS: MOLECULAR EVOLUTION OF HYPERPHALANGY AND INTERDIGITAL WEBBING

Abstract

The evolution of aquatic cetaceans (whales, dolphins and porpoises) from a terrestrial ancestor involved dramatic changes to the standard mammalian limb. For instance, cetaceans are the only mammals in evolutionary history to increase the number of phalanges per finger

(hyperphalangy). In dolphins, the standard number of three phalanges per finger form during early embryonic development, but this process persists into the fetal period until between nine and thirteen phalanges are generated in some fingers. Our data indicate that dolphin limb development differs from most terrestrial mammals in four ways. First, dolphins continue synthesizing specific proteins that allow for digit elongation (Fgf) and joint formation (Wnt) into the fetal period, while other mammals terminate phalanx formation during the embryonic period. Second, a recapitulation of Fgf8 signaling correlates with the development of more than six phalanges. Third, dolphins recruit two proteins that typically function in digit and limb formation (Fgf, Gremlin) to act within the interdigital tissues and prevent programmed cell death, thereby creating a soft-tissue flipper. Finally, asymmetric signaling in the ectoderm and interdigital mesenchyme generates a lift-producing, cambered aerofoil-shaped limb. Molecular evidence shows dolphins alter the duration (heterochrony) and location (heterotopy) of proteins

119 120

essential to mammalian limb development. Combined molecular and fossil evidence allows us to pinpoint when in geological time these novel developmental patterns evolved, and that the appearance of a soft-tissue flipper may have been a necessary precursor to the origin of cetacean hyperphalangy.

Introduction

A key innovation that enabled the extraordinary aquatic radiation of cetaceans (whales, dolphins, and porpoises) from terrestrial ancestors is the shift of the forelimb from a weight- bearing limb into a lift-generating flipper. The modern cetacean forelimb is composed of digits with as many as thirteen phalanges in a few digits (hyperphalangy, Richardson and Oelschläger,

2002), and is encased in a single, soft tissue flipper that is shaped like a cambered aerofoil (Fig.

23).

Although scientists have speculated about the genetic mechanisms generating cetacean hyperphalangy and the shaping the flipper (Kükenthal, 1889, Howell, 1930, Fedak and Hall,

2004) for over a century, no embryonic tissues have been available for molecular analyses.

Instead, considerable advances were made with descriptive embryological studies of forelimb development in the pantropical spotted dolphin (Stenella attenuata) (Sedmera et al., 1997,

Richardson and Oelschläger, 2002). Richardson and Oelschläger (2002) hypothesized that the genes responsible for digit formation in mammals were expressed over a longer portion of developmental time (heterochrony) in cetaceans compared to model taxa, creating hyperphalangy in digits II and III. Fedak and Hall (2004) hypothesized that the maintenance of interdigital tissues in dolphins could be a result of disrupted signaling within the interdigital tissues. 121

Fig. 23. Forelimb development in embryos of the dolphin S. attenuata. (A) Clear and stained embryo (LACM 94285, C23) with cartilage stained blue and bone red. The flipper at this stage has a phalanx count of (2/10/6/3/1) with hyperphalangy shown in digits II and III. (B) Clear and stained flipper of a dolphin embryo (LACM 95671, C20) with a phalanx count of 2/10/6/3/2 and hyperphalangy in digits II and III. (C) Clear and stained hand of a mouse embryo E15.5 with a primitive mammalian phalanx count of 2/3/3/3/3. Scale bars are 1cm in length.

122

In amniotes, limb outgrowth is directed by a signaling center, the apical ectodermal ridge (AER), located at distal edge of the developing limb bud. Duration of fibroblast growth factors (Fgfs) expressed within the AER control digit length in amniotes (Mariani et al., 2008).

Wingless type (Wnts) morphogens secreted within and adjacent to a developing digit are fundamental to joint formation (Hartmann and Tabin, 2001, Später et al., 2006). In amniotes, mesenchymal tissues between digits undergo apoptosis and regress, resulting in separated digits capable of individual movement. Bone morphogenic proteins (Bmp) expressed in the interdigital regions of mice and chicks induces interdigital apoptosis unless a Bmp inhibitor terminates interdigital cell death (Gañan et al., 1998). Bat wings (Weatherbee et al., 2006) and duck feet (Merino et al., 1999) express Gremlin, a Bmp inhibitor, within the interdigital mesenchyme and thereby maintain these tissues. Bats further hamper interdigital apoptosis by overlapping Fgf expression in the interdigital tissues with that of Gremlin and Bmp (Weatherbee et al., 2006). Fgf expression inhibits apoptosis by promoting the survival and proliferation of interdigital tissues.

We investigated the temporal and spatial pattern of protein signaling during early forelimb development of the pantropical spotted dolphin (Stenella attenuata). First, we tested the hypothesis that Fgf and Wnt signaling is prolonged in dolphins compared to that of a terrestrial mammal, the mouse (Mus) as proposed by Richardson and Oelschläger (2002).

Second, we tested the hypothesis that signaling within the interdigital tissues is disrupted in a developing dolphin as proposed by Fedak and Hall (2004), relative to Mus. Results show that the duration and location of molecular signals controlling digital and interdigital tissue development in a dolphin deviates from that seen in a generalized terrestrial mammal, Mus musculus. This allowed us to identify uniquely derived developmental features of cetacean 123

digits and the soft tissue flipper. Combined with paleontological evidence, we then identified when in geological time these novel developmental signaling patterns in the cetacean forelimb evolved and how the evolution of a soft tissue flipper and hyperphalangy relate to another.

Materials and Methods

32 years. Embryos were staged according to a modified version of the Carnegie system

(Thewissen and Heyning, 2007). Immunohistochemical data were based on dolphin embryos

(n=9), ranging from Carnegie Stage 13 (abbreviated below as C13, C19, etc.) to C19, and a fetus at C20. Each embryo was embedded in paraffin, and sectioned at 6-μm. Protocols were optimized using immersion-fixed, ethanol-preserved mouse embryos. Subsequently, non-limb embryonic dolphin tissue was then tested and optimized. Because of the variance in fixation and storage times, slightly different procedures were used for different specimens to obtain optimal results. In addition, negative control samples (minus primary antibody) were used to determine the level of background staining for all experiments. The following antibodies were used in this study: anti-fibroblast growth factor-8 (Fgf-8; Santa Cruz Biotechnology; sc-6958); anti-fibroblast growth factor-4 (Fgf-4; Santa Cruz Biotechnology; sc-1361); anti-wingless type 14(Wnt-14; Santa

Cruz Biotechnology; sc-20265); anti-bone morphogenic protein-2,4 (Bmp-2,4; Santa Cruz

Biotechnology; sc-6267); and anti-gremlin (Gre; Santa Cruz Biotechnology; sc-18276).

124

Results

Prolonged and Recapitulated Signaling of Fgfs and Wnt in the Dolphin Forelimb

Regarding digit outgrowth, we tested for Fgfs-4, 8 signaling in the AER from Carnegie (C)

Stages 13-20. We observed protein signals of Fgf-8 in the AER while the forelimb was shaped like a handplate (C 14-16, Figure 24). After C16, digits II and III continued the terminal addition of phalanges while a cessation in novel phalanx formation was seen in other digits. During C17,

Fgf-8 signaling was restricted to the basement membrane, while Fgf-4 signaling was comparatively intense along the AER. During C18 and early C19, the presence of Fgf-8 signals was not detected in the distal limb ectoderm. Instead, Fgf-4 signaling was localized in the distal limb ectoderm. At late C19, which marks the end of the embryonic period, and fetal C20, both

Fgfs-4,8 were expressed in the AER . Thus, in dolphins, an active AER continues Fgf signaling into the fetal period to accommodate development of an increased number of phalanges in digits II and III. In contrast, Mus musculus terminates AER activity, and by extension Fgf signaling, during the embryonic period of development, after a maximum of three phalanges are formed.

In chicks and mice, Wnt-9a is a fundamental player in joint formation as it encourages differentiation of joint cells into synovial connective tissue, but suppresses their chondrogenic potential. Wnt-9a is normally expressed in developing joint spaces, as well as tissues adjacent to a developing digit. To test whether molecular signals typical of standard joint formation were prolonged in formation of the supernumerary interphalangeal joints of dolphins, we documented the location of Wnt-9a protein signals in the dolphin forelimb. Wnt-9a was localized to the zeugopod-carpal, and carpometacarpal joints at C17, and all of the interphalangeal joints (normal and supernumary) at C19 (Fig. 25). Connective tissues adjacent 125

Fig. 24. Fibroblast growth factor (Fgf) protein signals in the developing forelimb of Stenella embryos and a fetus. In sections A-J and L presence of the protein of interest is indicated by brown stain against a blue counterstain (Thionin) stain all cells. In section K, the counterstain is light green in color (Fast Green). Fgf-8 signals in the apical and 126

dorsal ectoderm of (A) LACM 94594 and (B) LACM 94747. (C) Fgf-8 signals in the basement membrane and mesenchyme, and (D) Fgf-4 signals in the apical and dorsal ectoderm of LACM 94670. (E) Lack of Fgf-8 signals in LACM 94634. (F) Lack of Fgf-8 signals, but (G) intense Fgf-4 signals in the apical and dorsal ectoderm and mesenchyme in LACM 94818. (H,I) Recapitulation of Fgf-8 signals in the digit II apical ectoderm and interdigital tissues, and (J) intense Fgf-4 staining in the apical ectoderm of digit II and interdigital tissues of LACM 94817. (K) Fetal Fgf-8 and (L) Fgf-4 signals in the apical ectoderm and mesenchyme surrounding the digit tips.

Fig. 25. Wingless type 9a (Wnt-9a) protein signals in the developing forelimb of Stenella embryos. (A) Wnt-9a signals in the zeugopod-carpal joint and connective tissues adjacent to the developing joint in LACM 94670. (B) Wnt-9a signals in the metacarpal- carpal joint of LACM 94670. Wnt-9a signals in the normal and supernumerary interphalangeal joints of LACM 94817 in (C) a section through the whole hand and (D) a longitudinal section through digit III. (E) Inset of box in D with Wnt-9a signals documented in the supernumerary joints between phalanges 3-6. Arrows pinpoint Wnt- 9a signals.

127

to the developing digit also displayed Wnt-9a signaling. Thus, in dolphins, those signals typical of mammalian interphalangeal joint formation were also found in later developing supernumary joints, indicating a prolonged activity of joint formation.

Dolphin Forelimbs Display Gremlin Signaling and Unique Signal Domains of Fgfs

Interdigital apoptosis occurs in all amniotes that have separated digits. In both duck feet (Merino et al., 1999) and bat wings (Weatherbee et al., 2006), the pathway causing interdigital apoptosis is inhibited by expression of Bmp inhibitors. In dolphins, interdigital mesenchymal cells are not pyknotic, and apoptosis failed to occur, resulting in a syndactylous manus. To test whether dolphin forelimbs attempt to initiate interdigital apoptosis, we tested for the presence Bmp-2,4 protein signals within the interdigital mesenchyme. Bmp-2,4 signals were found throughout interdigital mesenchymal tissues at C17 and C19, with the greatest signal intensity localized near the digit tips (Fig. 26). Thus, developing dolphins initiate the molecular cascade leading to interdigital apoptosis, as in all amniotes.

In bat wings and duck feet, cessation of interdigital apoptosis is partially due to expression of Gremlin, a Bmp inhibitor (Merino et al., 1999, Weatherbee et al., 2006). We tested for the presence of Gremlin in the developing dolphin forelimb. Gremlin signals were localized throughout the interdigital tissues of dolphins, with the greatest signal intensity at the distal end of the interdigital spaces (Fig. 26). Thus, dolphins display overlapping Bmp-2,4 and

Gremlin signals in the mesenchymal interdigital tissues during limb development.

The developing bat wing also employs interdigital expression of Fgfs to further impede apoptosis and promote the survival and proliferation of wing membrane tissues (Weatherbee et al., 2006). To test whether the dolphin also utilized mesenchymal Fgf signals in generation of 128

Fig. 26. Gremlin and bone morphogenic protein (Bmp-2,4) signals in the developing forelimb of Stenella embryos. (A) Gremlin signals in the mesenchyme dorsal to the developing digit (LACM 94650, C-17), (B) Gremlins signals in the interdigital tissues (LACM 94817, C-19), (C) Bmp-2,4 signals in the mesenchyme surrounding the digit tip (LACM 94650, C-17), and (D) Bmp-2,4 signals in the interdigital mesenchyme (LACM 94817, C-19). (E) Schematic of a fetal dolphin flipper in cross-section showing regional protein signaling patterns. Fgfs signals are located throughout the flipper mesenchymal tissues, but are present only along the apical and ventral aspects of the ectoderm. Gremlin signals are located throughout the flipper mesenchymal tissues. Gremlin signals are most intense along the dorsal half of the mesenchymal tissue. 129

the soft tissue flipper, we tested for the presence of Fgfs-4,8 signals in the interdigital tissues.

During C17, Fgf-8 was documented in the mesenchymal tissues just deep to the AER. At early

C19, Fgf-4 signaling was observed throughout much of the mesenchyme just deep to the apical ectoderm. At late C19 and C20, both Fgfs-4,8 signals were localized to the mesenchyme surrounding the distal ends of the digits. Thus, dolphins displayed overlapping Bmp, Gremlin, and Fgf signaling along the distal aspect of the mesenchymal the interdigital tissues.

Asymmetrical Dorso-Ventral Signaling in the Dolphin Forelimb

The cetacean flipper is shaped like a cambered aerofoil in that the dorsal surface is convex while the ventral surface is comparatively flattened. During locomotion this hydrofoil generates lift and acts as a control surface for steering and balance. To understand which proteins may play a role in shaping the flipper, we tested for asymmetrical patterns of protein signaling of Fgfs, Bmps and Gremlin. Fgfs-4,8 were expressed along the apical and dorsal ectoderm during most stages of embryonic development, while signaling within the ventral ectoderm was consistently less intense or absent. Bmp expression remained uniform throughout the dorsal and ventral interdigital tissues. Surprisingly, Gremlin signals were asymmetrical within the flipper mesenchyme. During C17 and C19, Gremlin staining was concentrated on the dorsal surface of the flipper mesenchyme, just above a developing digit, while there was a comparative reduction or lack of staining within the ventral mesenchyme.

Thus, Fgf and Gremlin signals within the flipper ectoderm and mesoderm displayed asymmetrical patterns of protein signaling, while Bmp signaling appeared uniform (Fig. 26).

130

Discussion

Extended Duration of Fgf and Wnt Signaling in the Dolphin Forelimb

In this study, we used molecular techniques to elucidate the developmental basis of hyperphalangy in a dolphin forelimb. We found that dolphin phalangeal development occurs in three phases. During phase I, limb outgrowth occurs in a handplate until at least four phalanges are formed (C17). Outgrowth is initially controlled by AER-derived Fgf-8 signals, but Fgf-4 signaling increases at the end of this phase. Phase II occurs after at least five phalanges (C18,

C19) are formed in the principle digits (II and III). Fgf-4 directs limb outgrowth, as Fgf-8 signals were not detected, indicating at least a drastic reduction in Fgf-8 signaling. Phase III occurs after at least 6 phalanges are formed in the principle digits (end of C19, 20). Surprisingly, this phase is marked by a recapitulation of Fgf-8 signaling in the AER and interdigital tissues. Fgfs-4,8 in the ectoderm and mesenchyme act in concert to generate greater than six phalanges in the principle digits. Therefore, a minor form of hyperphalangy (four-five phalanges) forms during phases I and II, whenFgf-8 signaling is reduced. An extreme form of hyperphalangy (greater than six phalanges) occurs only after both Fgfs-4,8 signals are intense in the AER and mesenchyme. These data support the hypothesis of Fedak and Hall (2004) who predicted two forms of hyperphalangy.

Multiple waves of Fgf-8 signaling in the dolphin forelimb are intriguing and suggestive of an evolutionary role in the delicate regulation of phalanx count in cetacean digits. The evolution of cetacean phalanx number display little homoplasy and phalanx count gradually increases in more derived taxa in both suborders of cetaceans (Odontoceti and Mysticeti, Cooper et al.,

2007). The earliest fossil cetacean with hyperphalangy, Imerocetus (Mchedlidze, 1988), and 131

basal families of modern cetaceans (e.g., Physeter and Eubalaena) have only 4-5 phalanges in their principle digits (Cooper et al., 2007; Fig. 27). It could be that a recapitulation of Fgf signaling would not be required for the completion of phalanx formation. If the duration of Fgf and Wnt signaling was progressively extended longer in development, this could generate extreme forms of hyperphalangy, which is now seen in the most derived cetaceans (e.g.

Globicephala, Megaptera). An extreme form of hyperphalangy was first documented in the fossil rorqual whale Balaenoptera siberi, recovered from the Pisco Fm. of Peru (7-8 Ma, Pilleri

1989, 1990, Muizon et al., 2003). Together, molecular and paleontological data confirm that a minor form of hyperphalangy is the predominant morphology in fossil and extant cetaceans.

Development of this basal form of hyperphalangy is correlated with the upregulation of Fgf-4 signaling, and downregulation of Fgf-8 signaling, as seen during phases I and II of dolphin phalanx formation. Only a minority of cetacean lineages display an extreme form of hyperphalangy with greater than six phalanges in the principle digits. This derived morphology evolved at least 8Ma and molecular data suggests its evolution is coincident the recapitulation of Fgf-8 signaling.

Regarding supernumary joint formation, Wnt signaling was uniform, with only the duration of signaling being exceptional. Compared to the embryonic expression of Wnt in Mus musculus during the formation of two interphalangeal joints, the dolphin extended the duration of Wnt signaling until all ten joints formed in digit II. This process began at an early embryonic stage in both the mice and dolphins, but persisted into the fetal period in only the dolphins.

These data indicate that the primitive joint condition in the mouse was also initially utilized by dolphins, but then within the dolphin, a derived developmental feature of heterochrony evolved, and was correlated with an exceptional number of joints. 132

Fig. 27. Simplified phylogeny of cetaceans with key events in forelimb evolution indicated. Metacarpals (white) and phalanges (primitive phalanges are gray, hyperphalangeous phalanges are black) represent the fossil ambulocetid Ambulocetus (Thewissen et al., 1996), protocetid Maiacetus (Gingerich et al., 2009), basilosaurid Dorudon (Uhen, 2004), cetotheriid Imerocetus (Mchedlidze, 1988), balaenopterid Balaenoptera (Pilleri 1989, 1990), modern odontocetes including the sperm whale (Physeter, Cooper et al., 2007b), pilot whale (Globicephala), and modern mysticetes including the right whale (Eubalaena), and humpback whale (Megaptera) (Howell, 1930). Both arrows indicate further increasing of Fgf and Wnt signal duration.

133

Novel Domains of Fgf and Gremlin Facilitate For Soft Tissue Maintenance and the Shaping of the Cetacean Forelimb

Convergent among those taxa that locomote in fluid environments is a limb with an aerofoil-like aspect ratio that functions to generate lift. The lift-generating ability of a cetacean flipper is well documented (Weber et al., 2009), but no studies have identified the molecular signals that generate this morphology. This study found that the aerofoil-shaped flipper was probably created by the initial maintenance of interdigital tissues followed by asymmetrical signaling that directed the formation of a cambered hydrofoil in cross-section (Fig. 24 C,D, F-L;

Fig. 26 F). Overlapping Bmp-Gremlin-Fgf signaling in the interdigital tissues maintains these tissues, but elaboration is required to shape the developing limb into a lift-generating hydrofoil.

Tests indicate that Fgf signals concentrated in the ectoderm lining the dorsal surface of the limb, but a relative lack of staining along the ventral surface is consistent with differential growth of the flipper. Fgf in general promotes survival and proliferation of tissues; hence the dorsal side of the flipper would grow more than the ventral surface, and this signaling could account for the convex dorsal surface of the cambered cetacean flipper. Gremlin staining was more intense along the dorsal surface of the flipper mesenchyme compared to that of the ventral mesenchymal tissue. Gremlin may protect mesenchyme along the dorsal surface of the developing digits from Bmp signaling, while a lack of mesenchymal staining along the ventral surface may leave cells vulnerable to the apoptotic effects of Bmp. Together, these data suggest that patterning of the limb into a hydrofoil occurs during late embryonic development, and that asymmetrical staining of Fgf and Gremlin expression may have been essential to the evolution of the cetacean hydrofoil (Fig. 26F). 134

Fossil evidence indicates that interdigital webbing first evolved in the common ancestor of modern cetaceans, and was present in the fossil Dorudon (Uhen, 2004). This record indicates that a soft tissue flipper evolved at least 10Ma before the onset of a minor form of hyperphalangy (Fig. 27).

Interdigital Tissues as a Necessary Precursor to the Generation of Hyperphalangy

Presence of a soft tissue flipper may have been a key developmental precursor that was a necessary precursor for the evolution of hyperphalangy in cetaceans. Indeed, of those vertebrate taxa that display an extreme form of hyperphalangy, all retain interdigital tissues

(Richardson and Chipman, 2003). The addition of a single phalanx has been documented in isolated individuals of manatees that also bear a flipper, but is a fixed trait in aquatic and soft- shelled turtles that retain at least some interdigital webbing (Richardson and Chipman, 2003).

Interdigital tissues play a role in patterning the number of interphalangeal joints and phalanges

(Dahn and Fallon, 2000), and the presence of a soft-tissue flipper may be a necessary precursor for the evolution of hyperphalangy. In dolphins, interdigital tissues may act as a signaling reservoir for adjacent developing digits. Indeed, Gremlin expression has been shown to be a necessary Bmp inhibitor that allows for the maintenance of the Fgf-Shh feedback loop that regulates duration limb outgrowth (Khokha et al., 2003). This feedback loop ultimately determines phalanx and interphalangeal joint count. It could be that the evolution of a soft tissue flipper, and the signaling within, was a necessary and sufficient precursor that allowed for the generation of hyperphalangy in cetaceans.

CHAPTER VI

A REVIEW OF THE EVOLUTION AND DEVELOPMENT OF CETACEAN APPENDAGES

The central issue addressed in my research is the evolution and development of cetacean appendages across the cetartiodactylan land-to-sea transition. Genetic and developmental processes shape morphology, which in turn determines appendage performance

(Fig. 28). Various methodologies and analyses (i.e., phylogenetic, experimental immunohistochemical, anatomical, functional) were used to address limb evolution and development. My previous, current, and future works are discussed in the framework of the adaptive pathway in order to show that my research focuses on a central topic of limb evolution and development and utilizes a variety of methods to understand the limb as a complex trait

(Fig. 28).

Developmental Processes

Although many authors have speculated about the developmental mechanisms shaping cetacean limbs, no studies have reported on patterns of gene expression or protein signaling in the forelimbs of cetaceans. Chapters 4 and 5 of this dissertation were the first to identify uniquely derived signaling patterns that direct the development of the pantropical spotted dolphin (Stenella attenuata) forelimb.

135 136

Fig. 28. The Adaptive Pathway includes key theoretical steps used to determine if a biological structure is the result of an evolutionary adaptation (modified from Bock and von Wahlert 1965, Bock, 1977). This study utilizes this model of the adaptive pathway as a theoretical framework to review current understanding of the evolution and development of the cetacean forelimb. Black dots refer to the lineages studied. References at the bottom of some boxes are a list of articles and chapters I have published addressing limb evolution and development using different methodologies.

137

Chapter 4 (Cooper et al., in press) reported that the developing pantropical spotted dolphin forelimb bears a flattened apical ectoderm, a morphology that is similar to that of amphibians, and unlike the ridge-shaped ectoderm of other vertebrates. This chapter also documented patterns of fibroblast growth factor (Fgf) signaling within the limb apical ectoderm that are consistent among vertebrates. Taken together, these data suggest that among vertebrates, morphology of the apical ectoderm displays variation, but in most studied vertebrates, molecules secreted by cells in the apical ectoderm are conserved. My future research will expand on this topic by incorporating additional taxa.

Future molecular research will address the morphology and expression of genes within the apical ectoderm of bats (Order Chiroptera, Mammalia). Like cetaceans, bat forelimb development is highly divergent compared to terrestrial mammals, and it could be that this taxon also displays a non-typical apical ectodermal morphology. Previous studies have noted the bat apical ectoderm extends across the entire anteroposterior surface of the developing limb, but no studies have documented whether the apical ectoderm is a ridge or are flattened as in cetaceans. Chapter 5 reported the duration of fibroblast growth factor (Fgf-4,8) protein signals within the apical ectoderm of Stenella were extended over a longer period of developmental time as compared to a typical terrestrial mammal (i.e, Mus musculus). This heterochrony of protein signaling allows for a longer period of phalangeal formation, resulting in digits with an exceptional number of phalanges (hyperphalangy). This study also documented overlapping expression patterns of Fgfs, Bone morphogenic proteins (Bmps) and Gremlin (Gre) within the interdigital tissues. This overlapping signaling stops cell death and promotes survival of cells within the interdigital tissues resulting in the formation of a soft tissue flipper encasing the cetacean digits. Because these results are focused on a single cetacean taxon, future work 138

will compare signaling patterns of Stenella to another cetacean (i.e. Delphinapterus leucas) to identify the molecular mechanisms that are responsible for the morphological diversity of cetacean flippers and phalangeal counts.

Form and Function

The Embryonic Limb

Morphology of the dolphin embryonic forelimb was documented in Chapters 4 and 5 of this dissertation, and the apical ectoderm is flattened but secretes Fgfs (Chapter 4, Cooper et al., in press). During late embryonic stages, cross-sections of the limb reveal a convex dorsal surface and flattened ventral surface (Chapter 5). Correlated with this asymmetrical flipper shape, Fgf signaling was localized to both the apical and dorsal flipper ectoderm, but no Fgf signaling was found along the ventral ectoderm. Asymmetrical Fgf signaling probably generates this asymmetrical embryonic flipper shape. Adult cetaceans, have a cambered (asymmetrical) flipper that is more efficient at generating positive lift than a symmetrical flipper (Cooper et al.,

2008). Therefore, this work documents that the occurrence of asymmetrical signaling during embryogenesis results in an adult limb phenotype that directly affects the organism’s performance during locomotion.

Limb Skeletal Morphology

Forelimb skeletal morphology of extant cetaceans was documented in both extant toothed whales (odontocetes) and baleen whales (mysticetes). Modal phalangeal counts indicated the presence of supernumary phalanges (hyperphalangy) and little homoplasy within the distribution of phalangeal counts (Fig. 29; Cooper et al., 2007a). The number of phalanges 139

Fig. 29. Phylogenetic distribution of modes of phalangeal counts in the digits of extant cetaceans (from Cooper et al., 2007a). (A-E) Odontocete phalangeal counts optimized onto an odontocete phylogeny (Messenger and McGuire, 1998). (F-J) Mysticete phalangeal counts optimized onto a composite mysticete phylogeny (Rychel et al., 2004, Saski et al., 2005, Nikaido et al., 2006). (A,F) digit I, (B,G) digit II, (C,H) digit III, (D,I) digit IV, (E,J) digit V. Roman numerals indicate digit identity. 140

was correlated with flipper shape. Taxa with extreme phalangeal counts (i.e., dolphins and rorqual whales) have elongated flippers, while taxa with only a moderate form of hyperphalangy

(i.e., sperm and bowhead whales) have broad paddle-shaped flippers (Cooper et al., 2007a,

2008, Cooper 2009, Mellor et al., 2009). These skeletal data also provided the evolutionary foundation for Chapter 5, which documented those molecules that generate cetacean hyperphalangy (Chapter 5). The function of cetacean hyperphalangy is currently unknown. It could be that additional numbers of interphalangeal joints function to dissipate joint loadings in a digit. One possible test would be to attach straing gauges to interphalangeal joints of two digits of equal length. In one digit, some joints would be mechanically stabilized to simulate a digit with fewer joints. A control digit with no stabilized joints would be used as a comparison.

We would predict that in the modified digit, those joints adjacent to stabilized joints would experience greater strains compared to the joints of the control digit.

A surprising abundance of digital malformations were also documented in extant cetaceans (Cooper and Dawson, 2009). Phenotypes of most digital malformations in cetaceans bore a striking resemblance to digital malformations in mice, chicks and humans, in which gene expression during development was altered in the interdigital tissues. Because dolphins overlap

Fgf, Bmp, and Gremlin expression to retain interdigital tissues throughout ontogeny, it could be that they have a predisposition towards developing malformations. Furthermore, no malformations were linked with a change in flipper shape, and therefore probably did not change flipper function. It could be that cetacean digital malformations are not subject to selective pressures and therefore may occur without consequence to the animal.

141

Limb Soft-Tissue Morphology

Limb soft tissue morphologies in extant cetaceans are correlated with flipper function.

An analysis of the muscular and neural anatomy of odontocete and mysticete forelimbs showed that modern cetaceans lack many muscles and connective tissue structures that are essential for terrestrial locomotion, but retain sensory innervations of the limb (Cooper et al., 2007b).

Modern artiodactyls use their forelimbs for bearing loads and propulsion, while those of cetaceans are lift-generating hydrofoils.

Cetacean flippers display a startling variety of shapes (Cooper, 2009). During development, a flipper is initially formed by the termination of interdigital cell death, which allows for the maintenance and proliferation of interdigital tissues, and the flipper shape in cross-section is determined by the asymmetrical signaling of Fgf molecules (Chapter 5). Flipper shape is correlated with the width of interdigital spaces and the number of phalanges. Taxa with large interdigital spaces and only moderate forms of hyperphalangy have broad, paddle- shaped flippers that are used like rudders to assist in turning maneuvers. Taxa with small interdigital spaces with closely appressed digits, and who have extreme forms of hyperphalangy, have narrow, sickle-shaped flippers and a relatively longer flipper which creates a large control surface to assist in fast turning maneuvers and increase agility (Cooper, 2009, Cooper et al.,

2007a).

The Limbs of Fossil Cetaceans and Morphology of Fossil Cetartiodactylan Limbs

Artiodactyls (Cetartiodactyla) were studied in order to reconstruct their terrestrial and aquatic locomotor affinities (Chapter 2) as well as correlate bone geometry with ecological habitat (Chapter 3). The middle Eocene raoellid Indohyus is the closest relative to cetaceans and 142

displays a body plan similar to most terrestrial artiodactyls. Limb and joint shape indicates this taxon was probably utilizing a digitigrade stance during terrestrial locomotion, but propelled itself during aquatic locomotion via pelvic paddling (Chapter 2). Based on articulation of vertebrae, Indohyus held its head up above its vertebral column during both terrestrial and aquatic locomotion, similar to the modern-day deer, while the earliest cetaceans (pakicetids) held their head in line with their vertebral column (Chapter 2).

An analysis of cortical bone thickness among extant artiodactyls found that an extremely thickened cortex and reduced medullary cavity (osteosclerosis) was correlated with an aquatic lifestyle (Chapter 3). The addition of bone material of the limbs is thought to act as skeletal ballast to counteract an animal’s buoyancy. Cross-sectional geometries of the ribs, vertebrae, and long bones indicate that Indohyus displayed osteosclerosis (thick limb cortex and reduced medullary cavity) in its humerus and femur (Chapters 2 and 3). The degree of cortical thickness in the bones of Indohyus was greater than that of terrestrial artiodactyls but less than the extremely thickened bones of Hippopotamus and pakicetids (Chapter 3). By combining locomotion (Chapter 2) and cortical bone thickness data (Chapter 3), research showed that

Indohyus represents a critical morphological intermediate between terrestrial artiodactyls and the earliest cetaceans. Indohyus was amphibious and utilized part of its skeleton as ballast to aid in aquatic locomotion, but it lacked the extreme cortical thicknesses seen in cetartiodactylans that were aquatic specialists (e.g., Hippopotamus and the earliest cetaceans).

Future work on the skeletal morphology of Indohyus will address its cranial and dental anatomy. Cranial and dental anatomy will be used to identify uniquely derived morphological features beyond those initially described by Thewissen et al., (2007). Taxonomic assignment is also required, as all specimens were initially described only to the genus level by Thewissen et 143

al., (2007), and identification to the species level is required. Cooper et al., (in press) described multiple new taxa of the earliest cetaceans (pakicetids) based on dentition. These teeth will be used as a comparison with specimens of Indohyus.

Future work on cortical bone thickness in fossil taxa will address an ancient lineage of fossil artiodactyls, called anthracotheres, which are close relatives of hippopotamids. Chapter 3 of this dissertation found that cortical bone thickness of some anthracotheres fell within the range of values calculated for terrestrial taxa, while others were as aquatic as Indohyus. These results are intriguing and analysis of a greater taxonomic sample is required, as well as integration with isotopic evidence of habitat. It could be that data from Chapter 3 indicate an incipient land-to-sea transition in anthracotheres.

Alteration of cortical bone thickness along the land-to-sea transition is clearly an important adaptation, yet genetic control of bone thickness in cetartiodactylans is not understood. Molecular studies of the genetic mechanisms shaping bone cortices in mice and chicks, however have identified a plethora of genes responsible for altering osteoblast and osteoclast activity rates, but no studies have addressed the activity of these genes in cetartiodactylans or other mammals. The focus of my postdoctoral research will be to identify at least some of the genes responsible for the diversity of cortical bone variation in mammals that occupy different niches (i.e., aquatic, aerial, and terrestrial).

Future work will also address limb gross and cross-sectional phenotypes in sub-Saharan archaeocetes from the family Protocetidae. Specimens have been recovered from south Asia, north Africa, and the United States, and a single specimen was recovered from the sub-Saharan

African country of Nigeria, the protocetid Pappocetus. The limbs of Pappocetus are unknown as only cranial and vertebral material has been recovered. On my recent expedition to Nigeria, the 144

original Pappocetus fossil locality was found, and other fossil-rich marine localities were identified. Future work will continue excavating the Pappocetus locality for additional elements, as well as to identify other fossil-bearing localities in the region. More protocetid skeletons with limbs are needed to fully understand limb cortical changes along the land-to-sea transition.

Performance

Observations of limb performance in captive (Cooper et al., 2007b) and wild (Cooper et al., 2007b, 2008) cetaceans have noted stiffness of the cetacean flipper and its orientation during surfacing, turning, and feeding maneuvers. No quantitative studies have addressed flipper performance on live cetaceans. Instead, a life-sized cast of a minke whale (Balaenoptera acutorostrata) flipper was placed in a low-speed wind tunnel to test for lift, drag, and stall behavior over six speed trials, while the model was rotated through -40 and +40 degrees angles of attack (Fig. 30; Cooper et al., 2008). The lift-bearing range of angles of attack was between -

18 and +10 degrees depending on testing speed (Cooper et al., 2008). Observations of swimming and feeding minke whales in the wild have shown they held their flippers within the laboratory-predicted lift-generating angles of attack. Combined with measurements of forces incurred during feeding maneuvers, Cooper et al., (2008) found that the flippers and other control surfaces helped maintain a positive body pitch even when engulfing thousands of gallons of sea water.

Although forelimb performance has been addressed in extant cetaceans, no adequate modern analogue currently exists to reconstruct limb performance in fossil cetaceans. Future work could address limb performance in fossil taxa by testing modern artiodactlys in substrates 145

Fig. 30. Averaged lift data for the minke whale flipper model (Cooper et al., 2008). The x-axis is the angle of attack, or rotation along the leading edge of the flipper, while the y-axis is a measure of lift. Each of the colored line represents a different windtunnel speed the model was tested at. All near-linear parts of these data lines indicate lift was being generated. Beyond this linear region, drag exceeded lift, indicating the flipper stalled.

146

that fossil cetaceans were known to occupy. The earliest cetaceans, pakicetids, were known to walk in fine-grained muds, but their limb performance is unknown. A full National Science

Foundation grant has been written for a kinematic analysis of extant artiodactyls, pigs (Sus) and goats (Capra), as these artiodactyls walk on both stiff and muddy substrates. Specifically, this project aims to identify morphological and kinematic differences in these two living artiodactyls while they locomote on different substrates, and then utilize these data to reconstruct mud- based locomotion in the earliest fossil cetaceans. This project would therefore synthesize kinematic and paleoecological data to analyze limb performance in the earliest cetaceans.

Environment

The limbs of fossil cetaceans were affected by the viscosity of their substrate while on land, and the density of water while swimming. The shape of the artiodactyl manus is a large determinant in limb kinematics on soft-substrates. Indohyus and the earliest archaeocetes (i.e., pakicetids and ambulocetids) had four main weight-bearing digits in their forelimbs, and were digitigrade. Chapter 2 noted that the metapodials and phalanges of Indohyus were intermediate in length when compared to terrestrial artiodactyls, but the metapodials and phalanges of pakicetids and ambulocetids were greatly elongated. The presence of elongated bones in the hands would increase manus surface area and therefore mitigate the penetration of the limb into the sediment. Later diverging archaeocetes (i.e., remingotoncetids, protocetids) were still able to walk on land but engaged in greater aquatic behaviors, until in basilosaurids, the forelimb had evolved into a flipper (Chapter 5).

Water is an incompressible fluid that flows over and under the limb in a swimming cetacean. The behavior of water molecules as they pass the flipper is determined by the shape 147

and orientation of the flipper. Because cetaceans have a flipper that is hairless and shaped like a cambered hydrofoil, water typically flows in a mostly laminar path over the flipper, which creates lift. If limb orientation shifts, some water molecules move in circles as they pass over the flipper, creating turbulent flow. If enough of these molecules move in circles, vortices are shed from the flipper, creating drag. The orientation of the flipper at which drag exceeds lift is called the angle of stall, and the flipper no longer generates lift. A flipper will stall at different orientations as a function of swimming speed (Cooper et al., 2008).

Biological Role

In extant taxa, the flipper acts in concert with other control surfaces to steer the animal, maintain positive pitch during feeding maneuvers, and halt forward motion (Cooper et al.,

2007b, 2008, 2009), and are essential for successful feeding. During mating displays, courtship, and greetings, flippers can be used to slap the water or for touching conspecifics (Cooper et al.,

2007b).

The forelimb of Indohyus, pakicetids, and ambulocetids probably functioned much like the forelimb of the modern Hippopotamus in weight bearing and transporting of the animal in and out of small bodies of water. No data indicates that the forelimb functioned in aquatic propulsion in these taxa. Later-diverging archaeocetes (i.e., remingtonocetids and protocetids) were still able to walk on land, but continued to engage in swimming maneuvers. Lastly, the forelimb of basilosaurids functioned similar to a modern day flipper as the animal propelled itself through the water via tail-generated thrust and the flipper acted to control steering and balance. REFERENCES

Ahn, K., Mishina Y., Hanks M.C., Behringer R.R., and Crenshaw III E.B. 2001. BMPR-IA signaling is

required for the formation of the apical ectodermal ridge and dorsal-ventral patterning

of the limb. Development 128:4449-4461.

Bajpai S., and Thewissen J.G.M. 2000. A new, diminutive whale from Kachchh (Gujarat, India)

and its implications for locomotor evolution of cetaceans. Current Science 79:1478-

1482.

Bajpai S., Kapur V.V., Das D.P., Tiwari B.N., Saravanan N., and Sharma R. 2005. Early Eocene land

mammals from Vastan lignite mine, District Surat (Gujarat), Western India. Journal of

the Paleontological Society of India 50:101-113.

Balemans W., and Van Hul W. 2007. Minireview: The genetics of low-density Lipoprotein

Receptor-Related Protein 5 in bone: a story of extremes. Endocrinology 148(6):2622-

2629.

Bardeen C.R., and Lewis W.H. 1901. The development of the limbs, body-wall and back in man.

American Journal of Anatomy 1:1-36.

Baron R., and Rawadi G. 2007. Minireview: Targeting the Wnt/β-catenin pathway to regulate

bone formation in the adult skeleton. Endocrinology 148(6):2635-2643.

Baron R., Rawadi G., and Roman-Roman S. 2006. Wnt signaling: a key regulator of bone mass.

Current Topics in Developmental Biology 76:103-127.

148 149

Barrow J.R., Thomas K.R., Boussadia-Zahui O., Moore R., Kemler R., Capecchi M.R., and

McMahon A.P. 2003. Ectodermal Wnt3/β-catenin signaling is required for the

establishment and maintenance of the apical ectodermal ridge. Genes and Development

17:394-409.

Behrensmeyer A.K., Deino A.L., Hill A., Kingston J.D., and Saunders J.J. 2002. Geology and

geochronology of the middle Miocene Kipsaramon site complex, Muruyur Beds, Tugen

Hills, Kenya. Journal of Human Evolution 42:11-38.

Benke H. 1993. Investigations on the osteology and the functional morphology of the flipper of

whales and dolphins (Cetacea). Investigations on Cetacea 24:9-252.

Bininda-Emonds O.R.P., Jeffery J.E., Sánchez-Villagra M.R., Hanken J., Colbert M., Pieau C.,

Selwood L., Cate C.T., Raynaud A., Osabutey, C.K., Richardson, M.K. 2007. Forelimb-

hindlimb developmental timing changes across tetrapod phylogeny. BMC Evolutionary

Biology 7:182.

Bodine P.V.N., and Komm B.S. 2006. Wnt signaling and osteoblastogenesis. Reviews in Endocrine

and Metabolic Disorders 7:33-39.

Boisserie J.-R., Lihoreau F., Brunet M. 2005a. Origins of Hippopotamidae (Mammalia,

Cetartiodactyla): towards resolution. Zoologica Scripta 34:119–143.

Boisserie J.-R., Lihoreau F., Brunet M. 2005b. The position of Hippopotamidae within

Cetartiodactyla. Proceedings of the National Academy of Sciences, USA. 102(5):1537-

1541.

Boulet A.M., Moon A.M., Arenkiel B.R., Capecchi M.R. 2004. The roles of Fgf4 and Fgf8 in limb

bud initiation and outgrowth. Developmental Biology 273(2):361-372. 150

Bouvet J. 1968. Histogenèse précoce et morphogenèse du squelette cartilagineux des ceintures

primaires et des nageoires paires chez la truite (Salmo trutta fario L.) Arch. d’Anat.

Microsc. 57, 35-52.

Boyden L.M., Mao J., Belsky J., Mitzner L., Farhi A., Mitnick M.A., Wu D., Insogna K., Lifton R.P.

2002. High bone density due to a mutation in LDL-receptor-related protein 5. New

England Journal of Medicine 346:1513–1521

Brashares J.S., Garland Jr T., and Arcese P. 2000. Phylogenetic analysis of coadaptation in

behavior, diet, and body size in the African antelope. Behavioral Ecology 11, 452-463.

Braus H. 1906. Die Entwickelulng der form der Extremitäten und des Extremitätenskeletts.

Hertwig’s Hbh. Der Entwicklungslehre der Wirbeltiere 3:167-338.

Buchholtz E.A. 1998. Implications of vertebral morphology for locomotor evolution in early

Cetacea (pp. 325-351). In: J.G.M. Thewissen (ed.) The Emergence of Whales:

Evolutionary Patterns in the Origin of Cetacea, Plenum Press, New York.

Buffetaut E. 1979. A propos d’un crâne de Metriorhynchus (Crocodylia, Mesosuchia) de Bavent

(Calvados): I’allègement des os crâniens chez les Metriorhynchidae et sa signification.

Bulletin de la Société Géologique de Normandie 66: 77-83.

Buffetaut E., de Buffrénil V., de Ricqlès A., Spinar Z.V. 1982. Remarques anatomiques et

paléohistologiques sur Dyrosaurus phosphaticus, crocodilien mésosuchien des

phosphates yprésiens de Tunisie. Annals de Paléontologie (Vertébrés) 68: 327-341. de Buffrénil V., Collet A., Pascal M. 1985. Ontogenetic development of skeletal weight in a small

delphinid, Delphinus delphis (Cetacea: Odontoceti). Zoomorphology 105:336-344. 151

de Buffrénil V., de Sire J.-Y., Schoevaert D. 1986. Comparison de la structure et du volume

squelettiques entre un delphinidé (Delphinus delphis L.) et un mammifère terrestre

(Panthera leo L.). Canadian Journal of Zoology 64:1750-1756. de Buffrénil V., Ricqlès A. de, Ray C.E., and Domning D.P. 1990. Bone histology of the ribs of the

archaeocetes (Mammalia: Cetacea). Journal of Vertebrate Paleontology 10:455-466. de Buffrénil V., and Schoevaert D. 1988. On how the periosteal bone of the delphinid humerus

becomes cancellous: ontogeny of a histological specialization. Journal of Morphology

198:149-164. de Buffrenil V., Astibia H., Pereda Suberbiola X., Berreteaga A., and Bardet N. 2008. Variation in

bone histology of middle Eocene sirenians from Western Europe. Geodiversitas

30(2): 425-432.

Butti C., Corain L., Cozzi B., Podestà M., Pirone A., Affronte M., Zotti A. 2007. Age estimation in

the Mediterranean bottlenose dolphin Tursiops truncatus (Montagu 1821) by bone

density of thoracic limb. Journal of Anatomy 211: 639–646.

Caldwell M.W. 2002. From fins to limbs to fins: limb evolution in fossil marine reptiles. American

Journal of Medical Genetics 112:236–249.

Castanet J., and Caetano M.-H. 1995. Influence du mode vie sur les caractéristiques pondérales

et la structure du squelette chez les amphibiens anoures. Canadian Journal of Zoology

73: 234-242.

Christen B., and Slack J.M.W. 1997. FGF-8 is associated with anteroposterior patterning and limb

regeneration in Xenopus. Developmental Biology 192:455-466.

Christensen R.N., and Tassava R.A. 2000. Apical epithelial cap morphology and fibronectin gene

expression in regenerating axolotl limbs. Developmental Dynamics 217:216-224. 152

Christensen R.N., Weinstein M., and Tassava R.A. 2002. Expression of fibroblast growth factors

4, 8, and 10 in limbs, flanks, and blastemas of Ambystoma. Developmental Dynamics

223:193-203.

Cirmanová V., Bayer M., Stárka L., and Zajíčková K. 2008. The effect of leptin on bone – an

evolving concept of action. Physiological Research 57(Suppl. 1):S143-S151.

Clément-Lacroix P., Ai M., Morvan F., Roman-Roman S., Vayssière B., Belleville C., Estrera K.,

Warman M.L., Baron R., and Rawadi G. 2005. Lrp-5 independent activation of Wnt

signaling by lithium chloride increases bone formation and bone mass in mice. PNAS

102(48):17406-17411.

Clementz M. T., Goswami A., Gingerich P. D., and Koch P.L. 2006. Isotopic records from early

whales and sea cows: contrasting patterns of ecological transition. Journal of Vertebrate

Paleontology 26(2):355–370.

Cohn M.J. and Tickle C. 1999. Developmental basis of limblessness and axial patterning in

snakes. Nature 399:474-479.

Cooper L.N. 2009. Forelimb anatomy. (pp. 449-452) In: W.F. Perrin, B. Würsig, and J.G.M.

Thewissen (eds.). Encyclopedia of Marine Mammals, 2nd Ed., Elsevier, San Diego.

Cooper L.N. and Thewissen J.G.M. In press. Indohyus and the origin of whales. 2010 Yearbook of

Science and Technology.

Cooper L.N., Dawson S.D., Reidenberg J.S., Berta A. 2007a. Neuromuscular anatomy and

evolution of the cetacean forelimb. Anatomical Record: Advances in Integrative

Anatomy and Evolutionary Biology 209(9): 1121-1137.

Cooper L.N., Berta A., Dawson S.D., Reidenberg J. 2007b. Evolution of digit reduction and

hyperphalangy in the cetacean manus. Anatomical Record 290:654–672. 153

Cooper L.N., Sedano N., Johannson S., May B., Brown J., Holliday C.M., and Fish F.E. 2008.

Hydrodynamic performance of the minke whale (Balaenoptera acutorostrata) flipper.

Journal of Experimental Biology 211: 1859-1867.

Cretekos C.J., Rasweiler J.J. IV, Behringer R.R. 2001. Comparative limb morphogenesis in mice

and bats: a functional genetic approach towards a molecular understanding of diversity

in organ formation. Reproduction Fertility and Development 13:691-695.

Cretekos C.J., Deng J.M., Green E.D., Rasweiler J.J., Behringer R.R. 2007. Isolation, genomic

structure and developmental expression of Fgf8 in the short-tailed fruit bat, Carollia

perspicillata. International Journal of Developmental Biology 51:333-338.

Dahn R.D., and Fallon J.F. 2000. Interdigital regulation of digit identity and homeotic

transformation by modulated BMP signaling. Science 289(5478): 438-441.

Dawson S.D. 2003. Patterns of ossification in the manus of the harbor porpoise (Phocoena

phocoena): hyperphalangy and delta-shaped bones. Journal of Morphology 258:200-

206.

Dehm R., and Oettingen-Spielberg T. zu. 1958. Paläontologische und geologische

Untersuchungen im Tertiär von Pakistan. 2. Die mitteleocänen Säugetiere von Ganda

Kas bei Basal in Nordwest-Pakistan. Bayerische Akademie der Wissenschaften,

Mathematisch-Naturwissenschaftliche Klasse. Abhandlungen Neue Folge Heft 91:1−54.

Dehority W., Halloran B.P., Bikle D.D., Curren T., Kostenkuik P.J., Wronski T.J., Shen Y., Rabkin B.,

Bouraoui A., Morey-Holton E. 1999. Bone and hormonal changes induced by skeletal

unloading in the mature male rat. American Journal of Physiology Endocrinology and

Metabolism 276:62-69. 154

Domning D.P., and de Buffrénil V. 1991. Hydrostasis in the Sirenia: quantitative data and

functional interpretations. Marine Mammal Science 7:331-368.

Domning D.P. and Myrick A.C. Jr. (1980). Tetracycline marking and the possible layering rate of

bone in an Amazonian manatee (Trichechus inunguis). In Age determination of toothed

whales and sirenians: 203–207. Perrin, W.F. & Myrick, A.C. Jr. (Eds)., Special Issue 3,

Reports of the International Whaling Commission Cambridge: International Whaling

Commission.

Dubost G. 1978. Un apercu sur l’écologie du chevrotain africain Hyemoschus aquaticus

Ogilby, artiodactyle tragulide. Mammalia 42, 1–62.

Ducy P., Amling M., Takeda S., Priemel M., Schilling A.F., Beil F.T., Shen J., Vinson C., Rueger J.M.,

Karsenty G. 2000. Leptin inhibits bone formation through a hypothalamic relay: a central

control of bone mass. Cell 100:197-207.

Dufaure J.P., and Hubert J. 1961. Table de développment du lézard vivipare: Lacerta (Zootoca)

vivipara Jacquin. Archives d'Anatomie Microscopique et de Morphologie Experimentale

50: 309-327.

Eisenberg J.F. 1981. The Mammalian Radiations: An Analysis of Trends in Evolution, Adaptation,

and Behavior. Chicago: University of Chicago Press. p.610.

Elefteriou F., Takeda S., Ebihara K., Magre J., Patano N., Kim C.A., Ogawa Y., Liu X., Ware S.M.,

Craigen W.J., Robert J.J, Vinson C., Nakao K., Capeau J., Karsenty G. 2004. Serum leptin

level is a regulator of bone mass. PNAS 101(9):3258-63.

Endo H., Sasaki M., Kimura J., and Fukuta K. 2006. Functional strategies of the hindlimb muscles

in the mouse deer. Mammal Study 31:73-78.

155

English A.W.M. 1976. Functional anatomy of the hands of fur seals and sea lions. American

Journal of Anatomy 147:1-17.

English A.W.M. 1977. Structural correlates of forelimb function in fur seals and sea lions. Journal

of Morphology. 151:325-352.

Errick J., and Saunders J.W. 1974. Effects of an “inside-out” limb bud ectoderm on development

of the avian limb. Developmental Biology 41:338-351.

Fang H., and Elinson R.P. 1996. Patterns of distal-less gene expression and inductive interactions

in the head of the direct developing frog Eleutherodactylus coqui. Developmental

Biology 179:160-172.

Fawcett D.W. 1942. The amedullary bones of the Florida manatee (Trichechus latirostris).

American Journal of Anatomy 71:271-309.

Fedak T.J., and Hall B.K. 2004. Perspectives on hyperphalangy: patterns and processes. Journal

of Anatomy 204:151–163.

Felts W.J.L., and F.A. 1965. Structural orientation and density in cetacean humeri. American

Journal of Anatomy 116:171-204.

Felts W.J.L., and Spurrell F.A. 1966. Some structural and developmental characteristics of

cetacean (odontocete) radii. A study of adaptive osteogenesis. American Journal of

Anatomy 118:103-134.

Ferguson M.W.J. 1985. Reproductive biology and embryology of the Crocodilians. In: Gans C.,

Billett F., Maderson P.F.A. (eds.) Biology of the Reptilia. New York: John Wiley & Sons.

pp. 329-491.

Fernandez-Teran M., and Ros M.A. 2008. The apical ectodermal ridge: morphological aspects

and signaling pathways. International Journal of Developmental Biology 52:857-871. 156

Fischel, A. 1929. Lerbuch der Entwicklung des Menschen. Springer, Berlin.

Fish F.E., and Rohr J. 1999. Review of dolphin hydrodynamics and swimming performance.

SPAWARS System Center Technical Report 1801, San Diego, CA.

Fish F.E., and Stein B.R. 1991. Functional correlates of differences in bone density among

terrestrial and aquatic genera in the family Mustelidae (Mammalia). Zoomorphology

110:339-345.

Flower W.H. 1885. Introduction to the Osteology of Mammalia. 3rd ed. London: MacMillan and

Co.382 p.

Fordyce R.E., and de Muizon C. 2001 Evolutionary history of cetaceans: a review. pp. 169-233. In

Mazin J.-M. and de Buffrénil V. (eds.). Secondary Adaptation of Tetrapods to Life in

Water. München, Germany: Verlag Dr. Friedrich Pfeil.

Francillon-Vieillot H., de Buffrénil V., Castanet J., Meunier F.J., Sire J.-Y., Zylberger L., de Ricqlès

A. 1990. Microstructure and mineralization of vertebrate skeletal tissues. In: Carter J.G.

(ed.). Skeletal Biomineralization Patterns, Processes and Evolutionary Trends. Vol. 1.

New York: Van Nostrand Reinhold. p. 471-530.

Franssen R.A., Marks S., Wake D., and Shubin N. 2005. Limb chondrogenesis of the seepage

salamander, Desmognathus aeneus (Amphibia:Plethodontidae). Journal of Morphology

265:87-101.

Galis F., Wagner G.P., and Jockusch E.L.. 2003. Why is limb regeneration possible in amphibians

but not in reptiles, birds, and mammals? Evolution and Development 5(2):208-220.

Galatius A., Andersen M.E.R., Haugan B., Langhoff H.E., and Jespersen Å. 2006. Timing of

epiphyseal development in the flipper skeleton of the harbour porpoise (Phocoena

phocoena) as an indicator of paedomorphosis. Acta Zoologica (Stockholm) 87:77-82. 157

Gañan Y., Macias D., Basco R.D., Merino R., Hurle J.M. 1998. Morphological diversity of the avian

foot is related with the pattern of msx gene expression in the developing autopod.

Developmental Biology 196:33-41.

Garimella R., Tague S.E., Zhang J., Belibi F., Nahar N., Sun B.H., Insogna K., Wang J., Anderson

H.C. 2008. Expression and synthesis of bone morphogenic proteins by osteoclasts: a

possible path to anabolic bone remodeling. Journal of Histochemistry and

Cytochemistry 56(6):569-577.

Geisler J.H. 2001. New morphological evidence for the phylogeny of Artiodactyla, Cetacea, and

Mesonychidae. American Museum Novitates 3344:1-53.

Geisler J.H., and Theodor J.M. 2009. Hippopotamus and whale phylogeny. Nature 458, (E1-E4

on-line only, doi:10.1038/nature07776).

Geisler J.H., and Uhen M.D. 2003. Morphological support for a close relationship between

hippos and whales. Journal of Vertebrate Paleontology 23:991-996.

Geisler J.H., and Uhen M.D. 2005. Phylogenetic relationships of extinct Cetartiodactyls: results of

simultaneous analyses of molecular, morphological, and stratigraphic data. Journal of

Mammalian Evolution 12:145-160.

Geisler J.H., Theodor J.M., Uhen M.D., and Foss S.E. 2007. Phylogenetic relationships of

cetaceans to terrestrial artiodactyls (pp. 19-31). In: D.R. Prothero and S.E. Foss (eds.) The

Evolution of Artiodactyls. Baltimore: Johns Hopkins University Press.

Géraudie J., and François Y. 1973. Les premiers stades de la formation de l’ébauche de nagoire

pelvienne de truite (Salmo fario et Salmo gairdneri). I. Etude anatomique. Journal of

Embryology and Experimental Morphology 29:221-237. 158

Géraudie J. 1978. Scanning electron microscope study of the developing trout pelvic fin bud.

Anatomical Record 191:391-396.

Germain D., and Laurin M. 2005. Microanatomy of the radius and lifestyle in amniotes

(Vertebrata, Tetrapoda). Zoologica Scripta 34:335-350.

Gilbert F.S. 2006. Developmental Biology, 8th Ed. Sunderland: Sinauer Associataes. 817 p.

Gingerich P.D. 1977. A small collection of fossil vertebrates from the middle Eocene Kuldana and

Kohat formations of Punjab (Pakistan). Contributions from the Museum of Paleontology,

University of Michigan 24:190-203.

Gingerich P.D. 1989. New earliest Wasatchian mammalian fauna from the Eocene of

Northwestern Wyoming: composition and diversity in a rarely sampled high-floodplain

assemblage. University of Michigan Papers on Paleontology 28. pp. 97.

Gingerich P.D. 2003. Land-to-sea transition in early whales: evolution of Eocene Archaeoceti

(Cetacea) in relation to skeletal proportions and locomotion of living semiaquatic

mammals. Paleobiology 29:429-454.

Gingerich P.D., and Bajpai S. 1998. New early Eocene cetacean from India and the timing of the

origin of whales. PNAS, USA 95:15464-15468.

Gingerich P.D., and Russell D.E. 1981. Pakicetus inachus, a new archaeocete (Mammalia,

Cetacea) from the early-middle Eocene Kuldana Formation of Kohat (Pakistan).

Contributions from the Museum of Paleontology, University of Michigan

25(11):235−246.

Gingerich P.D., and Russell D.E. 1990. Dentition of early Eocene Pakicetus (Mammalia, Cetacea).

Contributions from the Museum of Paleontology, University of Michigan 28(1):1−20.

159

Gingerich P.D., Raza S.M., Arif M., Anwar M., and Zhou X. 1994. New whale from the

Eocene of Pakistan and the origin of cetacean swimming. Nature 368:844-847.

Gingerich P.D., Haq M. ul, Zalmout I.S., Khan I.H., Malkani M.S. 2001. Origin of whales from early

artiodactyls: hands and feet of Eocene Protocetidae from Pakistan. Science 293:2239-

2242.

Gingerich P.D., Haq M. ul, Koenigswald W.V., Sanders W.J., Smith H.B., and Zalmout I.S. 2009.

New protocetid whale from the middle Eocene of Pakistan: birth on land, precocial

development, and sexual dimorphism. PLoS One 4(2): 1-20.

Girondot M., and Laurin M. 2003. Bone Profiler: a tool to quantify, model and statistically

compare bone section compactness profiles. Journal of Vertebrate Paleontology 23:458-

461.

Goel S.C., and Mathur J.K. 1977. Morphogenesis in reptilian limbs. In: Ede, D.A.,

Hinchliffe, J.R., Balls, M. (eds.) Vertebrate Limb and Somite Morphogenesis.

Cambridge: Cambridge University Press. pp. 387-404.

Gol’din P.E. 2005. Body proportions of harbour porpoise Phocoena phocoena (Cetacea,

Phocoenidae) in the Sea of Azov and the Black Sea. Vestnik Zoologii 39(5):59-65.

Gordon K.R. 1981. Locomotor behavior of the walrus (Odobenus). Journal of Zoology, London

195:349-357.

Grandel H., and Schulte-Merker S. 1998. The development of paired fins in the Zebrafish (Danio

rerio). Mechanisms of Development 79:99-120.

Gray N.M., Kainec K., Madar S., Tomko L. and Wolfe S. (2007) Sink or swim? Bone density as a

mechanism for buoyancy control in early cetaceans. Anatomical Record 290: 638–653. 160

Gunnell G.F., and Bartels W.S. 2001. Basin margins, biodiversity, evolutionary innovation, and

the origin of new taxa (pp. 404-430) In: G.F. Gunnell (ed.). Eocene Biodiversity: Unusual

Occurrences and Rarely Sampled Habitats. New York: Kluwer Academic/Plenum

Publishers.

Habib M., and Ruff C.B. 2008. The effects of locomotion on the structural characteristics of avian

limb bones. Zoogical Journal of the Linnean Society 153:601-624.

Hallgrímsson B., Willmore K., and Hall B.K. 2002. Canalization, developmental stability, and

morphological integration in primate limbs. Yearbook of Physical Anthropology 45:131-

158.

Hamrick M.W. 2001. Primate origins: evolutionary change in digital ray patterning and

segmentation. Journal of Human Evolution 40:339-351.

Hamrick M.W. 2002. Developmental mechanisms of digit reduction. Evolution and

Development 4(4):247-248.

Hamrick M.W. 2003. Increased bone mineral density in the femora of GDF8 knockout mice. The

Anatomical Record Part A 272A:388-391.

Hamrick M.W., and Ferrari S.L. 2008. Leptin and the sympathetic connection of fat to bone.

Osteoporosis International 19:905-912.

Han M.-J., An J.-Y., and Kim W.-S. 2001. Expression patterns of Fgf-8 during development and

limb regeneration of the axolotl. Developmental Dynamics 220:40-48.

Han M., Yang X., Taylor G., Burdsal C.A., Anderson R.A., and Muneoka K. 2005. Limb

regeneration in higher vertebrates: developing a roadmap. Anatomical Record 287B:14-

24.

161

Hanken J. 1986. Developmental evidence for amphibian origins. Evolutionary Biology 20:389-

417.

Hanken J., Carl T.F., Richardson M.K., Olsson L., Schlosser G., Osabutey C.K., Klymkowsky M.W.

2001. Limb development in a “non-model” vertebrate, the direct-developing frog

Eleutherodactylus coqui. Journal of Experimental Zoology (Mol Dev Evol) 291:375-388.

Hara K., Kimura J., and Ide H. 1998. Effects of FGFs on the morphogenetic potency and AER-

maintenance activity of cultured progress zone cells of chick limb bud. International

Journal of Developmental Biology 42:591-599.

Hartman D.S. 1979. Ecology and behavior of the manatee (Trichechus manatus). American

Society of Mammalogists, Special Publication No. 5.

Hartmann C., and Tabin C.J. 2001. Wnt-14 plays a pivotal role in inducing synovial joint formati

on in the developing appendicular skeleton. Cell 104: 341-351.

Heep H., Wedemeyer C., Wegner A., Hofmeister S., von Knoch M. 2008. Differences in

trabecular bone of leptin-deficient ob/ob mice in response to biomechanical loading.

International Journal of Biological Sciences 4(3):169-175.

Hiney K.M., Nielsen B.D., Rosenstein D., Orth M.W., Marks B.P. 2004. High-intensity exercise of

short duration alters bovine bone density and shape. Journal of Animal Science

82:1612-1620.

Hodgkinson V.S., Johanson Z., Ericsson R., and Joss J.M.P. 2007. Apical ectodermal ridge (AER)

development in the pectoral fin of the Australian lungfish (Neoceratodus forsteri).

Journal of Morphology 268(12):1085.

Holmgren N. 1933. On the origin of the tetrapod limb. Acta Zoologica 14:185-295. 162

Honig L.S. 1984. Pattern formation during development of the amniote limb. pp. 197-221. In:

Ferguson, M.J.W. (ed.) The Structure, Development, and Evolution of Reptiles. Academic

Press: London.

Howell AB. 1930. Aquatic Mammals: Their Adaptations to Life in The Water. Springfield: Charles

C. Thomas. 338 p.

Hua S., and de Buffrénil V. 1996. Bone histology as a clue for the interpretation fo the functional

adaptations in Thalattosuchia (Reptilia: Crocodylia). Journal of Vertebrate Paleontology

16:703-717.

Hulbert R.C. 1998. Postcranial osteology of the North American middle Eocene protocetid

Georgiacetus (pp. 235-267). In: J.G.M. Thewissen (ed.), Emergence of Whales:

Evolutionary Patterns in the Origin of Cetacea, Plenum Press, New York.

Hulbert R.C., Petkewich R.M., Bishop G.A., Bukry D., and Aleshire D.P. 1998. A new middle

Eocene protocetid whale (Mammalia:Cetacea:Archaeoceti) and associated biota from

Georgia. Journal of Paleontology 72:905-925.

Iwaniec U.T., Boghossian S., Lapke P.D., Turner R.T., Kalra S.P. 2007. Central leptin gene therapy

corrects skeletal abnormalities in Leptin-deficient ob/ob mice. Peptides 28(5):1012-

1019.

Jarvik E. 1965. On the origin of girdles and paired fins. Israel J Zool 14:141-172.

Jurand A. 1965. Ultrastructural aspects of early development of the fore-limb buds in the chick

and the mouse. Proceedings of the Royal Society of London, Series B, Biological Sciences

162(988):387-405.

Kaiser H.E. 1960. Untersuchungen zur vergleichenden osteologie der fossilen und rezenten

pachyostosen. Palaeontographica A 114:113-196. 163

Karsenty G. 1999. The genetic transformation of bone biology. Genes and Development 13:3037-

3051.

Kelley R.O. 1973. Fine structure of the apical rim-mesenchyme complex during limb

morphogenesis in man. Journal of Embryology and Experimental Morphology 29:117-

131.

Kengaku M., Capdevila J., Rodriguez-Esteban C., de la Peña J., Johnson R.L., Belmonte J.C.I., and

Tabin C.J. 1998. Distinct WNT pathways regulating AER formation and dorsoventral

polarity in the chick limb bud. Science 280:1274-1277.

Keyte A.L., Imam T., and Smith K.K. 2006. Limb heterochrony in a marsupial , M. domestica.

Developmental Biology 295:414-422.

Khokha M.K., Hsu D., Brunet L.J., Dionne M.S., and Harland R.M. 2003. Gremlin is the BMP

antagonist required for maintenance of Shh and Fgf signals during limb patterning.

Nature Genetics 34(3): 303-307.

Kimmel R.A., Turnbull D.H., Blanquet V., Wurst W., Loomis C.A., Joyner A.L. 2000. Two lineage

boundaries coordinate vertebrate apical ectodermal ridge formation. Genes and

Development 14(11):1377-1389.

Köllicker A. 1879. Entwicklungsgeschichte des Menschen und der höhren Thiere, Zweite Auflage.

W., Englemann, Leipzig.

Kriloff A., Germain D., Canoville A., Vincent P., Sache M., and Laurin M. 2008. Evolution of bone

microanatomy of the tetrapod tibia and its use in palaeobiological inference. Journal of

Evolutionary Biology 21:807-826. 164

Krishtalka L., and Stucky R.K. 1985. Revision of the Wind River faunas, early Eocene of Central

Wyoming. Part 7. Revision of Diacodexis (Mammalia, Artiodactlya). Annals of Carnegie

Museum, Carnegie Museum of Natural History 54(14):413-486.

Krishtalka L., and Stucky R.K. 1986. Early Eocene artiodactyls from the San Juan Basin, New

Mexico, and the Piceance Basin, Colorado. Contributions to Geology, University of

Wyoming, Special Paper 3:183-196.

Kükenthal W. 1889. Die Hand der Cetaceen. Denkshr D Med Naturwiss Gesell Jena 3:23–69

Kumar K. and Sahni A. 1985. Eocene mammals from the Upper Subathu Group, Kashmir

Himalaya, India. Journal of Vertebrate Paleontology 5(2):153-168.

Laurin M., Girondot M., and Loth M.-M. 2004. The evolution of long bone microanatomy and

lifestyle in lissamphibians. Paleobiology 30:589-613.

Laurin M., Germain D., Steyer J.-S., and Girondot M. 2006. Données microanatomiques sur la

conquête de l’environnement terrestre par les vertébrés. Comptes Rendus Palevol

5:603-618.

Leclair R., Lamontagne C., and Aubin A. 1993. Allométrie de la masse du squelette chez les

amphibiens anoures. Canadian Journal of Zoology 71:352-357.

Lee K.K.H., and Chan W.Y. 1991. A study of the regenerative potential of partially excised mouse

embryonic fore-limb bud. Anatomy and Embryology 184:153-157.

Lee N.K., and Karsenty G. 2008. Reciprocal regulation of bone and energy metabolism. Trends in

Endocrinology and Metabolism 19(5):161-166.

Ling J.K. 1974. The integument of marine mammals. pp. 1-44. In: R.J. Harrison (ed.). Functional

Anatomy of Marine Mammals. Academic Press: London. 165

Little R.D., Carulli J.P., Del M., Dupuis J., Osborne M., Folz C., Manning S.P., et al. 2002. A

mutation in the LDL receptor-related protein 5 gene results in the autosomal dominant

high-bone-mass trait. American Journal of Human Genetics 70:11–19

Long F. 2008. When the gut talks to bone. Cell 135:795-796.

Lu P., Yu Y., Perdue Y., and Werb Z. 2008. The apical ectodermal ridge is a timer for generating

distal limb progenitors. Development 135:1395-1405.

Macsai C.E., Foster B.K., and Xian C.J. 2008. Roles of Wnt signaling in bone growth, remodeling,

skeletal repair disorders and fracture repair. Journal of Cellular Physiology 215:578-587.

Madar S.I. 1998. Structural adaptations of early archaeocete long bones (pp. 353-378). In: J.G.M.

Thewissen (ed) The Emergence of Whales. New York: Plenum Press.

Madar S.I. 2007. The postcranial skeleton of early Eocene pakicetid cetaceans. Journal of

Paleontology 81:176-200.

Madar S.I., Thewissen J.G.M., and Hussain S.T. 2002. Additional holotype remains of

Ambulocetus natans (Cetacea, Ambulocetidae), and their implications for

locomotion in early whales. Journal of Vertebrate Paleontology 22(2):405-422.

Maddison W.P. 1991. Squared-change parsimony reconstructions of ancestral states for

continuous-valued characters on a phylogenetic tree. Systematic Zoology 40:304-314.

Maddison W.P., and Maddison D.R. 2009. Mesquite: A modular system for evolutionary

analysis. Version 2.6. http://mesquiteproject.org

Mahmood R., Bresnick J., Hornbruch A., Mahony C., Morton N., Colquhoun K., Martin P.,

Lumsden A., Dickson C., Mason I. 1995. A role for FGF-8 in the initiation and

maintenance of vertebrate limb outgrowth. Current Biology 5(7):797-806. 166

Marcot J.D. 2007. Molecular phylogeny of terrestrial artiodactyls: conflicts and resolution (pp. 4-

18). In: D.R. Prothero and S.E. Foss (eds.) The Evolution of Artiodactyls. Baltimore: Johns

Hopkins University Press.

Mariani F.V., Ahn C.P., and Martin G.R. 2008. Genetic evidence that FGFs have an instructive role

in limb proximal-distal patterning. Nature 453:401-405.

McCrady E. 1938. Embryology of the opossum. American Anatomical Memoires 16:1-233.

Mchedlidze G.A. 1988. Fossil cetacea of the Caucasus. (pp. 49-75). In: J.F.C. Whitmore (ed.)

Rotterdam: AA Balkema.

Mercader N. 2007. Early steps of paired fin development in zebrafish compared with tetrapod

limb development. Development, Growth, and Differentiation 49:421-437.

Merino R., Rodriguez-Leon J., Macias D., Gañan Y., Economides A.N., and Hurle J.M. 1999. The

BMP antagonist Gremlin regulates outgrowth, chondrogenesis and programmed cell

death in the developing limb. Development 126: 5515-5522.

Michaud J.L., Lapointe F., Le Douarin N.M. 1997. The dorsoventral polarity of the presumptive

limb is determined by signals produced by the somites and by the lateral somatopleure.

Development 124:1443-1452.

Milaire J. 1957. Contribution à la connaissance morphologique et cytochimique des bourgeons

de membres chez quelques Reptiles. Archives de Biologie 68:429-512.

Miller J.D. 1985. Embryology of marine turtles. (pp. 269-328). In: Gans C., Billett F., Maderson

P.F.A. Biology of the Reptilia. New York: John Wiley & Sons.

Moon A.M. and Capecchi M.R. 2000. Fgf8 is required for outgrowth and patterning of limbs.

Nature Genetics 26:455-459. 167

Muijres F.T., Johansson L.C., Barfield R., Wolf M., Spedding G.R., Hedenström A. 2008. Leading-

edge vortex improves lift in slow-flying bats. Science 319:1250-1253. de Muizon C., McDonald H.G., Salas R., and Urbina M. 2003. A new early species of the aquatic

sloth Thalassocnus (Mammalia, Xenarthra) from the Late Miocene of Peru. Journal of

Vertebrate Paleontology 23:886–894.

Narita T., Sasaoka S., Udagawa K., Ohyama T., Wada N., Nishimatsu S.I., Takada S., and Nohno T.

2005. Wnt10a is involved in AER formation during chick limb development.

Developmental Dynamics 233:282-287.

Ngo-Muller V., and Muneoka K. 2000. Influence of FGF4 on digit morphogenesis during limb

development in the mouse. Developmental Biology 219:224-236.

Niswander L. 2002. Interplay between the molecular signals that control vertebrate limb

development. Int. J. Dev. Biol. 46:877-881.

Niswander L., Tickle C., Vogel A., and Martin G. 1994a. Function of FGF-4 in limb development.

Molecular Reproduction and Development 39(1):83-89.

Niswander L., Jeffrey S., Martin G.R., and Tickle C.. 1994b. A positive feedback loop coordinates

growth and patterning in the vertebrate limb. Nature 371:609-612.

Nowak R.M. 1991. Walker’s Mammals of the World, 5th Edition, Vol. II. Baltimore:

John’s Hopkins University Press. pp. 1629.

Nummela S., Thewissen J.G.M., Bajpai S., Hussain T., and Kumar K. 2007. Sound transmission in

archaic and modern whales: anatomical adaptations for underwater hearing.

Anatomical Record 290(6): 716-733. 168

O’Leary M.A. 1998. Phylogenetic and morphometric reassessment of the dental evidence for a

mesonychian and cetacean clade (pp. 133-161). In: J.G.M. Thewissen (ed) The

Emergence of Whales. New York: Plenum Press.

O’Leary M.A., and Gatesy J. 2008. Impact of increased character sampling on the

phylogeny of Cetartiodactyla (Mammalia): combined analysis including fossils.

Cladistics 24(4):397-442.

Onda H., and Tassava R.A. 1991. Expression of the 9G1 antigen in the apical cap of axolotl

regenerates requires nerves and mesenchyme. Journal of Experimental Zoology

257:336-349.

O’Rahilly R. ,and Müller F. 1985. The origin of the ectodermal ring in staged human embryos of

the first 5 weeks. Acta Anatomica 122:145-157.

O’Rahilly R., Gardner E., Gray D.J. 1956. The ectodermal thickening and ridge in the limbs of

staged human embryos. Journal of Embryology and Experimental Morphology 4:254-

264.

Panman L., Galli A., Lagarde N., Michos O., Soete G., Zuniga A., and Zeller R. 2006. Differential

regulation of gene expression in the digit forming area of the mouse limb bud by SHH

and gremlin 1/FGF-mediated epithelial-mesenchymal signaling. Development 133:3419-

3428.

Peter K. 1903. Mitteilungen zur Entwicklungsgeschichte der Eidechse. IV. Die

Extremitätenscheitelleiste der Amnioten. Arch. Mikr. Anat. 61:509-521.

Pickford M. 1983. On the origins of Hippopotamidae together with descriptions of two species, a

new genus and a new submfamily from the Miocene of Kenya. Géobios 16:193-217.

169

Pilgrim G.E. 1940. Middle Eocene mammals from north-west India. Proceedings of

the Zoological Society of London, Series B 110:127-152.

Pilleri G. 1989. Balaenoptera siberi, ein neuer spatmiozaner bartenwal aus der Pisco-Formation

Perus. (pp. 65–84). In: Pilleri G. (ed.) Beiträge zur paläontologie der cetaceen Perus.

Waldau Berne, Switzerland: Hirnanatomisches Istitut der Universitat Ostermundigen

(Bern).

Pilleri G. 1990. Paratypus von Balaenoptera siberi (Cetacea: Mysticeti) aus der Pisco Formation

Perus. (pp. 205–215). In: Pilleri G, editor. Beiträge zur paläontologie der cetaceen und

pinnipedier der Pisco Formation Perus II. Waldau-Berne, Switzerland: Hirnanatomisches

Istitut der Universitat Bern Ostermundigen (Schweiz).

Pizette S., and Niswander L. 1999. BMPs negatively regulate structure and function of the limb

apical ectodermal ridge. Development 126:883-894.

Pizette S., Abate-Shen C., and Niswander L. 2001. BMP controls proximodistal outgrowth, via

induction of the apical ectodermal ridge, and dorsoventral patterning in the vertebrate

limb. Development 128:4463-4474.

Pontzer H., Lieberman D.E., Momin E., Devlin M., Polk J.D., Hallgrimmson B., Cooper D.M.L.

2006. Trabecular bone in the knee responds with high sensitivity to changes in load

orientation. Journal of Experimental Biology. 209:57 – 65.

Poss K.D., Shen J., Nechiporuk A., McMahon G., Thisse B., Thisse C., Keating M.T. 2000. Roles for

Fgf signaling during zebrafish fin regeneration. Developmental Biology 222:347-358.

Price S.A., Bininda-Edmonds O.R.P., and Gitttleman J.L. 2005. A complete phylogeny of the

whales, dolphins, and even-toed hoofed mammals (Cetartiodactyla). Biological Reviews

80:445-473. 170

Ralston S.H., and de Crombrugghe B. 2006. Genetic regulation of bone mass and susceptibility to

osteoporosis. Genes and Development 20:2492-2506.

Ranga Rao A. 1971. New mammals from Murree (Kalakot zone) of the Himalayan foot hills near

Kalakot, Jammu and Kashmir State, India. Journal of the Geological Society of India

12(2): 125–134.

Ranga Rao A. 1972. Further studies on the vertebrate fauna of Kalakot, India. New mammalian

genera and species from the Kalakot zone of Himalayan foot hills near Kalakot, Jammu

and Kashmir State, India. Directorate of Geology, Oil and Natural Gas Commission,

Dehra Dun, Special Papers 1:1-22.

Ranga Rao A., and Misra V.N. 1981. On the new Eocene mammal localities in the Himalayan

foot-hills. Himalayan Geology 11:422–428.

Reifers F., Böhli H., Walsh E.C., Crossley P.H., Stainier D.Y.R., Brand M. 1998. Fgf8 is mutated in

zebrafish acerebellar (ace) mutants and is required for maintenance of midbrain-

hindbrain boundary development and somitogenesis. Development 125:2381-2395.

Rhodin A.G.J. 1985. Comparative chondro-osseous development and growth of marine turtles.

Copeia 3:752-771.

Rhodin A.G.J., Ogden J.A., and Conlogue G.J. 1981. Chondro-osseous morphology of

Dermochelys coriacea, a marine reptile with mammalian skeletal features. Nature

290:244-246.

Richardson M.K. 1995. Heterochrony and the phylotypic period. Developmental Biology

172:412-421.

171

Richardson M.K., and Chipman A.D. 2003. Developmental constraints in a comparative

framework: a test case using variations in phalanx number during amniote evolution.

Journal of Experimental Zoology 296B: 8-22.

Richardson M.K., and Oelschläger H.A. 2002. Time, pattern, and heterochrony: a study of

hyperphalangy in the dolphin embryo flipper. Evolution and Development 4:435–444

Richardson M.K., Jeffery J.E., Tabin C.J. 2004. Proximodistal patterning of the limb: insights from

evolutionary morphology. Evolution and Development 6(1):1-5.

Richardson M.K., Carl T.F., Hanken J., Elinson R.P., Cope C., and Bagley P. 1998. Limb

development and evolution: a frog embryo with no apical ectodermal ridge (AER).

Journal of Anatomy 192:379-390.

Ridgway S.H., and Howard R. 1979. Dolphin lung collapse and intramuscular circulation during

free diving: evidence from nitrogen washout. Science 206:1182-1183. de Ricqlès A., and de Buffrénil V. 2001. Bone histology, heterochronies, and the return of

tetrapods to life in water: where are we? (pp. 289-310) In: J.-M. Mazin and V. de

Buffrénil (eds.). Secondary Adaptation of Tetrapods to Life in Water. München:Verlag.

Roe L. J., Thewissen J.G.M., Quade J., O’Neil J.R., Bajpai S., Sahni A., and Hussain S.T. 1998.

Isotopic approaches to understanding the terrestrial-to-marine transition of the earliest

cetaceans. (pp. 399−422). In: J.G.M. Thewissen (ed.), The Emergence of Whales. Plenum

Press, New York.

Rose K.D. 1985. Comparative osteology of North American dichobunid artiodactyls.

Journal of Paleontology 59(5):1203-1226. 172

Rubin L., and Saunders J.W. 1972. Ectodermal-mesodermal interactions in the growth of limb

buds in the chick embryo: constancy and temporal limits of the ectodermal induction.

Developmental Biology 28:94-112.

Russell D.E., Zhai R.-J. 1987. The Paleogene of Asia: mammals and stratigraphy. Mémoires du

Muséum National D’Histoire Naturelle, Serie C, Sciences de la Terre, 52. pp. 1-488.

Sahni A., and Khare S.K. 1971. Three new Eocene mammals from Rajauri District, Jammu and

Kashmir. Journal of the Palaeontological Society of India 16:41-53.

Sahni A., Bhatia S.B., Hartenberger J.-L., Jaeger J.-J., Kumar K., Sudre J., Vianey-

Liaud M. 1981. Vertebrates from the Subathu Formation and comments on the

biogeography of Indian subcontinent during the early Paleogene. Bull. Soc. Géol.

France XXIII(6):689-695.

Sakata T., Wang Y., Halloran B.P., Elalieh H.Z., Cao J., Bikle D.D. 2004. Skeletal unloading induces

resistance to Insulin-like Growth Factor-1 (IGF-1) by inhibiting activation of the IGF-I

signaling pathways. Journal of Bone and Mineral Research 19(3):436-446.

Sanz-Ezquerro J.J. and C. Tickle. 2003. Fgf signaling controls the number of phalanges and tip

formation in developing digits. Current Biology 13(20):1830-1836.

Saunders J.W., Jr. 1948. The proximodistal sequence of origin of the parts of the chick wing and

the role of the ectoderm. Journal of Experimental Zoology 108:363-403.

Saunders J.W., Jr. 1998. Apical ectodermal ridge in retrospect. Journal of Experimental Zoology

282:669-676.

Saunders J.W., and Gasseling M.T. 1968. Ectodermal-mesodermal interactions in the origin of

limb symmetry. (pp. 78-97). In: Epithelial-Mesenchymal Interactions. R.E. Fleischmajer

and R. Billingham, eds. Williams & Wilkins, Baltimore. 173

Sawakami K., Robling A.G., Ai M., Pitner N.D., Liu D., Warden S.J., Li J., Maye P., Rowe D.W.,

Duncan R.L., Warman M.L., and Turner C.H. 2006. The Wnt co-receptor LRP5 is essential

for skeletal mechanotransduction but not for the anabolic bone response to parathyroid

hormone treatment. Journal of Biological Chemistry 281(33): 23698-23711.

Schaeffer B. 1947. Notes on the origin and function of the artiodactyl tarsus. American

Museum Novitates 1356:1-24.

Schmidt H. 1898. Über die Entwicklung der Milchdruse und die Hyperthelie menschlicher

Embryogen. Morph. Arb., Jena 8:157-193.

Schlosser G. 2001. Using heterochrony plots to detect dissociated coevolution of characters.

Journal of Experimental Zoology (Mol Dev Evol) 291:282-304.

Sears K.E. 2004. Constraints on the morphological evolution of marsupial shoulder girdles.

Evolution 58(10):2353-2370.

Sears K.E. 2008. Molecular determinants of bat wing development. Cells Tissues Organs 187:6-

12.

Sears K.E., Behringer R.R., Rasweiler J.J., IV, Niswander L.A. 2006. Development of bat flight:

morphologic and molecular evidence of bat wing digits. PNAS 103(17):6581-6586.

Sedmera D., Míšek I., and Klima M. 1997. On the development of cetacean extremities: II.

Morphogenesis and histogenesis of the flippers in the spotted dolphin (Stenella

attenuata). European Journal of Morphology 35:117-123.

Shedlock A.M., Milinkovitch M.C., and Okada N. 2000. SINE evolution, missing data and the

origin of whales. Systematic Biology 49:808-817.

Smith K.K. 2003. Time’s arrow: heterochrony and the evolution of development. International

Journal of Developmental Biology 47:613-621. 174

Später D., Hill T.P., O’Sullivan R.J., Gruber M., Conner D.A., and Hartmann C. 2006. Wnt9a

signaling is required for joint integrity and regulation of Ihh during chondrogenesis.

Development and Disease 133:3039-3049.

Stein B.R. 1989. Bone density and adaptation in semiaquatic mammals. Journal of

Mammalogy70:467–476.

Steiner K. 1928. Entwicklungsmechanische Untersuchungen über die Bedeutung des

ektodermalen Epithels der Extremitätenknospe von Amphibienlarven. Wilhelm Roux’

Arch. Entwm. Org. 113:1-11.

Steiner K. 1929. Über die Entwicklung und Differenzierungsweise der menschlichen Haut. I. Über

die fruhembryonale Entwicklung der menschlichen Haut. Z. Zellforsch. mikrosk. Anat.

8:691-720.

Sudre J., Russell D.E., Louis P., and Savage D.E. 1983. Les artiodactyles de l’Éocène inférieur

d’Europe. Bull. Mus. Natn. Hist. Nat., Paris 5(3):281-333.

Sun X., Mariani F.V., and Martin G.R. 2002. Functions of FGF signaling from the apical

ectodermal ridge in limb development. Nature 418:501-508.

Tabin C.J., and McMahon A.P. 2008. Grasping limb patterning. Science 321:350-352.

Takeda S., Elefteriou F., Levasseur R., Liu X., Zhao L., Parker K.L., et al. 2002. Leptin regulates

bone formation via the sympathetic nervous system. Cell 111:305-317.

Takeda S., and Karsenty G. 2008. Molecular bases of the sympathetic regulation of bone mass.

Bone 42:837-840.

Talamillo A., Bastida M.F., Fernandez-Teran M., Ros M.A. 2005. The developing limb and the

control of the number of digits. Clinical Genetics 67:143-153. 175

Tarassof F.J. 1972. Comparative aspects of the hind limbs of the river otter, sea otter, and seals.

(pp. 333-359). In: R.J. Harrison (ed.) Functional Anatomy of Marine Mammals,Vol. 1.

Plenum Press: New York.

Tarassof F.J., Bisaillon A., Pièrard J., and Whitt A.P. 1972. Locomotory patterns and external

morphology of the river otter, sea otter, and harp seal (Mammalia). Canadian Journal

of Zoology 50: 915-929.

Tarin D. and Sturdee A.P. 1971. Early limb development of Xenopus laevis. Journal of Embryology

and Experimental Morphology 26(2):169-179.

Taylor M.A. 1993. Stomach stones for feeding or buoyancy? The occurrence and function of

gastroliths in marine tetrapods. Philosophical Transactions of the Royal Society of

London B 341: 163-175.

Taylor M.A. 2000. Functional significance of bone ballast in the evolution of buoyancy control

strategies by aquatic tetrapods. Historical Biology 14: 15-31.

Theodor J.M. and Foss S.E. 2005. Deciduous dentitions of Eocene cebochoerid artiodactyls and

cetartiodactyl relationships. Journal of Mammalian Evolution 12:161-181.

Theodor J.M, Erfurt J., and Métais G. 2007. The earliest artiodactyls: Diacodexeidae,

Dichobunidae, Homacodontidae, Leptochoeridae, and Raoellidae. (pp. 32-58). In: D.R.

Prothero and S.E. Foss (eds.) The Evolution of Artiodactyls. Baltimore: Johns Hopkins

University Press.

Thewissen J.G.M., and Bajpai S. 2001. Dental morphology of the Remingtonocetidae (Cetacea,

Mammalia). Journal of Paleontology 75:463-465.

176

Thewissen J.G.M. and Fish F.E. 1997. Locomotor evolution in the earliest cetaceans:

functional model, modern analogues, and paleontological evidence. Paleobiology

23(4):482-490.

Thewissen J.G.M., and Heyning J. 2007. Embryogenesis and development in Stenella attenuata

and other cetaceans. (pp: 307-329). In: D.L. Miller (ed.) Reproductive Biology and

Thewissen J.G.M., and Hussain S.T. 1990. Postcranial osteology of the most primitive artiodactyl

Diacodexis pakistanensis (Dichobunidae). Anatomia, Histologia, Embryologia 19:37-48.

Thewissen J.G.M, and Hussain S.T. 1998. Systematic review of the Pakicetidae, early and middle

Eocene Cetacea (Mammalia) from Pakistan and India (pp. 220−238). In: K.C. Beard and

M.R. Dawson (eds.), Dawn of the Age of Mammals in Asia. Bulletin of Carnegie Museum

of Natural History 34.

Thewissen J.G.M., and Hussain S.T. 1990. Postcranial osteology of the most primitive artiodactyl,

Diacodexis pakistanensis (Dichobunidae). Anatomia, Histologia, Embryologia (Berlin),

19:37-48.

Thewissen J.G.M., and Heyning J. 2007. Embryogenesis and development in Stenella attenuata

and other cetaceans. (pp. 307-329). In Miller, D.L. (ed.) Reproductive biology and

phylogeny of cetacea: whales, dolphins, and porpoises. Vol. 7.

Thewissen J.G.M., and Madar S.I. 1999. Ankle morphology of the earliest cetaceans and its

implications for the phylogenetic relations among ungulates. Systematic Biology 48:21-

30.

Thewissen J.G.M., Madar S.I., and Hussain S.T. 1996. Ambulocetus natans, an Eocene cetacean

(Mammalia) from Pakistan. Courier Forschungsinstitut Senckenberg 191:1-86.

177

Thewissen J.G.M., Russell D.E., Gingerich P.D., and Hussain S.T. 1983. A new dichobunid

artiodactyl (Mammalia) from the Eocene of North-West Pakistan. Part I: Dentition and

classification. Proceedings of the Koninklike Nederlandse Akademie Van Wetenschapen,

Serie B 86:153-180.

Thewissen J.G.M., Gingerich P.D., Russell D.E. 1987. Artiodactyla and perissodactyla (Mammalia)

From the early-middle Eocene Kuldana Formation of Kohat (Pakistan). Contributions

from the Museum of Paleontology, The University of Michigan. 27(10):247-274.

Thewissen J.G.M., Hussain S.T., and Arif M. 1994. Fossil evidence for the origin of aquatic

locomotion in archaeocete whales. Science 263:210-212.

Thewissen J.G.M., Williams E.M., and Hussain S.T. 2001a. Eocene mammal faunas from northern

Indo-Pakistan. Journal of Vertebrate Paleontology 21(2):347–366.

Thewissen J.G.M., Williams E.M., Roe L.J., Hussain S.T. 2001b. Skeletons of terrestrial

cetaceans and the relationships of whales to artiodactyls. Nature 413:277-281.

Thewissen J.G.M., Cohn M.J., Stevens L.S., Bajpai S., Heyning J., and Horton W.E., Jr. 2006.

Developmental basis for hind-limb loss in dolphins and origin of the cetacean body plan.

PNAS 103:8414-8418.

Thewissen J.G.M., Cooper L.N., Clementz M.T., Bajpai S., and Tiwari B.N. 2007. Whales

originated from artiodactyls in the Eocene epoch of India. Nature 450:1190-1194.

Thewissen J.G.M., Cooper, L.N., George, C., and Bajpai, S. 2009. From land to water: the origin of

whales, dolphins and porpoises. Evolution: Education and Outreach 2:272–288.

Thewissen J.G.M., Madar S.I., and Hussain S.T. 1996. Ambulocetus natans, an Eocene cetacean

(Mammalia) from Pakistan. Courier Forschungsinstitut Senckenberg 191. Frankfurt a.M.

28(6):1-86. 178

Tickle C. 2006. Making digit patterns in the vertebrate limb. Nature Reviews: Molecular Cellular

Biology 7:45-53.

Todt W.L. and Fallon J.F. 1984. Development of the apical ectodermal ridge in the chick wing

bud. Journal of Embryology and Experimental Morphology 80:21-41.

Toyama Y., Peralta X.G., Wells A.R., Kiehart D.P., Edwards G.S. 2008. Apoptotic force and tissue

dynamics during Drosophila embryogenesis. Science 321(5896):1683--1686.

Uhen M.D. 2004. Form, function and anatomy of Dorudon atrox (Mammalia, Cetacea):

an archaeocete from Middle to Late Eocene of Egypt. University of Michigan

Papers on Paleontology 34:1-222.

Uhen M.D. 2007. Evolution of marine mammals: Back to the sea after 300 million years.

Anatomical

Record 290: 514-522.

Vasse J. 1972. Sur les activités de synthèse dans la crête épiblastique apicale de l’ébauche du

membrane antérieur chez les embryons de tortue (Testudo graeca L. et Emys orbicularis

L.); étude histologique et autoradiographique. C. r. hebd. Séanc. Acad. Sci., Paris, D.

274:284-287.

Verheyden J.M., and Sun X. 2008. An Fgf/Gremlin inhibitory feedback loop triggers termination

of limb bud outgrowth. Nature 454:638-641.

Vogel A., Rodriguez C., and Izpisúa-Belmonte J.-C. 1996. Involvement of FGF-8 in initiation,

outgrowth and patterning of the vertebrate limb. Development 122:1737-1750.

Von Dassow G., and Munro E. 1999. Modularity in animal development and evolution: elements

of a conceptual framework for EvoDevo. Journal of Experimental Zoology (Mol Dev Evol)

285:307-325. 179

Wall W.P. 1983. The correlation between high limb-bone density and aquatic habits in recent

mammals. Journal of Paleontology 57(2):197-207.

Weatherbee S.D., Behringer R.R., Rasweiler J.J. IV, Niswander L.A. 2006. Interdigital webbing

retention in bat wings illustrates genetic changes underlying amniote diversification.

PNAS 103(41):15103-15107.

Weber P.W., Howle L.W., Murray M.M., and Fish F.E. 2009. Lift and drag performance of

odontocete flippers. Journal of Experimental Biology 212(14):2149-2158.

West R.M. 1980. Middle Eocene large mammal assemblage with Tethyan affinities, Ganda Kas

region, Pakistan. Journal of Paleontology 54(3):508−533.

Wiffen J., de Buffrénil V., de Ricqlès A., and Mazin J.-M. 1995. Ontogenetic evolution of bone

structure in Late Cretaceous Plesiosauria from New Zealand. Geobios 28:625-640.

Wings O. 2007. A review of gastrolith function with implications for fossil vertebrates and a

revised classification. Acta Palaentologica Polonica 52(1):1-16.

Wood A. 1982. Early pectoral fin development and morphogenesis of the apical ectodermal

ridge in the Killifish, Aphyosemion scheeli. Anatomical Record 204:349-356.

Woodward B.L.,Winn J.P., andFish F.E. 2006. Morphological specializations of baleen whales

associated with hydrodynamic performance and ecological niche. Journal of Morphology

276:1284-1294.

Yadav V.K., Ryu J.-H., Suda N., Tanaka K.F., Gingrich J.A., Schütz G., Glorieux F.H., Chiang C.Y.,

Zajac J.D., Insogna K.L., Mann J.J., Hen R., Ducy P, Karsenty G. 2008. Lrp5 controls bone

formation by inhibiting serotonin synthesis in the duodenum. Cell 135: 825–837.

Yan J., Clifton K.B., Reep R.L., and Mecholsky J.J., Jr. 2006a. Application of fracture mechanics to

failure in manatee rib bone. Journal of Biomechanical Engineering 128:281-289. 180

Yan J., Clifton K.B., Mecholsky J.J., Jr., Reep R.L.. 2006b. Fracture toughness of manatee rib and

bovine femur using a chevron-notched beam test. Journal of Biomechanics 39: 1066-

1074.

APPENDIX

CATALOGUE OF INDOHYUS POSTCRANIAL ELEMENTS FROM THE A. RANGA RAO COLLECTION USED IN THIS ANALYSIS.

Specimen Number Skeletal Material

Ribs

RR 168 Cranial rib, crushed, with head, neck, and proximal and midshaft, lacking distal end RR 217 Three fragments of rib shaft, lacking head, neck and end RR 221 Rib shaft, crushed, lacking proximal and distal ends RR 222 Rib shaft, crushed, lacking proximal and distal ends RR 235 Two fragments of rib shaft, lacking proximal and distal ends RR 243 Cranial rib with head, neck and partial shaft, lacking distal end RR 244 Rib with crushed head and shaft, lacking distal end RR 245 Rib shaft, crushed, lacking proximal and distal ends

Vertebrae

RR 7 Two articulated vertebrae, lacking ventral half of each vertebra, and body of smallest vertebra RR 220 Vertebra, dorsoventrally crushed, lacking caudal articular facet, caudal portion of body, and crushed neural arch Cervical Vertebrae RR 198 Cervical vertebra, crushed lacking spinous process, cranial facets, tips of transverse processes RR 86 Atlas lacking transverse processes, lacking part of left articular facet for axis RR 165 Atlas with right transverse process missing most of its cranial portion, dorsal arch lacking its cranial surface and dorsal portion of occipital condyle facet RR 33 Cervical vertebra 3 lacking all but base of spinous process, left transverse foramen, cranial and caudal surface of vertebral body RR 38 Cervical vertebra 5, lacking tip of spinous process, articular facets on left side RR 136 Cervical vertebra 5 lacks spinous process, distal half of right transverse process, cranial epiphysis RR 195 Cervical vertebra 7 lacking cranial half of bone, tips of transverse processes, left caudal facet

181

182

Cranial Thoracic Vertebrae

RR 20 Cranial thoracic vertebra, crushed, lacking cranial facets, transverse processes, and had unfused epiphyses RR 21 Cranial thoracic vertebra, crushed, spinous process lacks tip, transverse processes lack tips RR 247 Cranial thoracic cranial and caudal surface of body incomplete, lacking spinous process

Middle Thoracic Vertebrae

RR 85 Middle thoracic vertebra, unfused cranial epiphysis and lacking cranial articular facets RR 137 Middle thoracic vertebra, crushed, spinous process lacking tip, cranial articular surfaces broken tips, deformed body lacking part of caudal face RR 242 Middle Thoracic vertebra, crushed, spinous process lacks tip, broken right transverse process and caudal articular surfaces RR 292 Middle thoracic vertebra lacks tip of spinous process, transverse processes and ventral half of body, dorsal half of body is crushed

Terminal Thoracic Vertebrae

RR 25 Thoracic vertebra tentatively identified as T13, crushed mediolaterally axis, tips of cranial articular facets, transverse processes, left accessory process, left cranial costal fovea RR 239 Two articulated caudal thoracic vertebra articulated with L1, crushed mediolaterally, spinous process broken, lacks ventral surface of each vertebra (see also description in “Lumbar Vertebrae”)

Lumbar Vertebrae

RR 215 Lumbar vertebra lacking right transverse process and cranial half of vertebra RR 218 Body of a lumbar vertebra lacking all processes, cranial and caudal epiphyses unfused RR 219 Middle lumbar vertebra, dorsoventrally crushed, lacking spinous and transverse processes RR 239 Cranial lumbar vertebra in articulation with two caudal thoracic vertebrae, lacking tip of spinous process, ventral surface of vertebra (see also description in “Terminal Thoracic Vertebrae”) RR 296 Two attached, but not articulated, lumbar vertebrae, mediolaterally crushed, lacking tips of spinous processes and transverse processes, smallest vertebra lacks caudal articular facets

Sacrum

RR 156 Three fused sacral vertebrae, S1-S3, cannot determine if S4 is also fused. S1 with complete body, lacking spinous and transverse processes, and the left auricular surface. S2, S3 lacking tips of left transverse processes, right surface of median sacral crest, right intermediate sacral crest.

183

Caudal Vertebrae

RR 45 Middle caudal vertebra lacking left mammillary process, left cranial transverse process, caudal transverse processes,and ventrolateral surfaces of caudal aspect of vertebra RR 65 Terminal caudal vertebra with body dorsoventrally crushed, lacking left cranial transverse process RR 75 Middle caudal vertebra lacking some transverse processes RR 169 Caudal vertebra, dorsoventrally crushed, lacking right cranial articular process, left caudal articular processes RR 184 Middle caudal vertebra mediolaterally crushed, lacking distal half RR 249 Terminal caudal vertebra lacking left cranial transverse process RR 294 Caudal vertebra, asymmetrical, crushed right cranial transverse and transverse processes, lacks tip of caudal articular processes

Scapula

RR 131 Scapula fragment with partial spine, damaged acromion process RR 155 Left scapula lacking cranial and dorsal aspects RR 171 Right scapula fragment, distal aspect of blade, spine and coracoid process RR 263 Glenoid fossa

Humerus

RR 145 Humerus, left RR 149 Humerus, left RR 31 Humerus, right, lacking distal half RR 90 Humerus, fragmentary proximal epiphysis RR 140 Humerus, lacking distal half RR 157 Humerus, left, lacking distal half

Radius

RR 2 Radius, two shaft fragments, lacking proximal epiphysis RR 36 Radius, fragment of distal shaft, epiphysis RR 79 Radius, shaft lacking proximal epiphysis, crushed distal epiphysis RR 122 Radius, proximal epiphysis RR 265 Radius, right, lacks proximal aspect of bone

Ulna

RR 39 Ulna, left, adult, crushed shaft and proximal epiphysis, lacks distal epiphysis RR 142 Ulna, crushed shaft and proximal epiphysis, lacks anconeal and coracoid processes and distal epiphysis RR 144 Ulna, right, adult, crushed shaft and proximal epiphysis, lacks olecranon process and distal epiphysis

184

Magnum

RR 250 Magnum

Metacarpals

RR 69 Distal metacarpal shaft with head, crushed in mediolateral aspect RR 83 Metacarpal, sheared with offset shaft, proximal epiphysis crushed RR 138 Metacarpal III, left, proximal epiphysis crushed RR 228 Metacarpal III, left RR 270 Metacarpal IV, left, proximal shaft missing pieces of cortex RR 271 Metacarpal III, left RR 297 Metacarpal I, proximal shaft missing pieces of cortex, shaft broken at steep angle

Phalanges of the Hand

RR 273 Proximal phalanx of the manus, central RR 274 Proximal phalanx of the manus, central RR 281 Proximal phalanx of the manus, central RR 99 Intermediate phalanx of the manus, lacking distal epiphysis, crushed along dorsoventral plane RR 117 Intermediate phalanx of the manus lacking distal epiphysis RR 179 Intermediate phalanx of the manus, juvenile, lacking proximal epiphysis RR 233 Intermediate phalanx of the manus, digit I, lacking proximal epiphysis RR 234 Intermediate phalanx of the manus, peripheral

Pelvis

RR 43 Innominate, left, broken ischium and pubis, lacks tip of ilium RR 44 Innominate, left, missing iliac blade, caudal aspect of the pubis and ischium RR 146 Acetabulum RR 162 Acetabulum, left, fragmentary ilium, ischium RR 187 Innominate fragment RR 256 Innominate, right lacking tip of ilium and pubis RR 257 Innominate, right RR 258 Ilium fragment, left, partial acetabulum RR 259 Ilium, left, lacks tip of iliac blade, pubis, and caudal aspect of ishium

Femur

RR 5 Femur, juvenile, distal shaft lacking epiphysis RR 41 Femur, adult, left, head, proximal half of shaft, lacks trochanters RR 42 Femur, right, adult, proximal epiphysis, proximal half of shaft RR 89 Femur, right, adult, epiphysis, fragmentary proximal shaft RR 101 Femur, right, adult RR 133 Femur, right, adult, lacks proximal half RR 141 Femur, left, crushed proximal epiphysis, lacks trochanters and distal half of shaft RR 154 Femur, left, adult, lacking distal half of shaft 185

RR 161 Femur, left, adult, lacking distal half of shaft RR 175 Femur, distal epiphysis only RR 176 Femur, distal epiphysis only RR 203 Femur, crushed distal epiphysis RR 237 Femur, left, adult, fragmentary head, separate greater trochanter and intertrochanteric crest RR 264 Femur, adult, head only RR 267 Femur left, adult, lacks humeral head and part of distal shaft, broken midshaft, crushed distal epiphysis RR 268 Femur, right, juvenile, crushed, lacks greater tuberosity and distal epiphysis, broken distal to proximal epiphysis

Tibia

RR 22 Tibia plateau, left RR 23 Tibia plateau, crushed RR 46 Tibia shaft, lacking epiphyses RR 84 Tibia right, adult, lacks distal half of shaft and distal epiphysis RR 134 Tentatively identified, crushed tibial plateau RR 143 Tibia shaft, crushed distal epiphysis, lacks tibial plateau and tibial tuberosity RR 148 Tibia, crushed, lacks distal half of shaft and distal epiphysis RR 301 Tibia, right, crushed shaft, lacking distal epiphysis

Fibula

RR 96 Fibula proximal fragment RR 216 Fibula shaft, crushed proximal epiphysis, lacking distal epiphysis. RR 295 Fibula shaft with crushed distal end, sheared midshaft, lacking proximal end

Patella

RR 234 Patella RR 269 Patella

Astragali

RR 35 Astragalus, left, lacking medial half RR 129 Astragalus, left, lacks part of proximal trochlea RR 213 Astragalus, left RR 224 Astragalus, right RR 246 Astragalus, right RR 290 Astragalus, left

Calcaneus

RR 164 Calcaneus, right RR 167 Calcaneus, left RR 170 Calcaneus, left RR 290 Calcaneus, left, attached to astragalus

186

Cuboid

RR 178 Cuboid fragment with astragalar facet RR 191 Cuboid, left, fragment with plantar process, fragmentary calcaneal and astragalar facets RR 214 Cuboid, left RR 240 Cuboid, right, fragment with plantar process, calcaneal facet, incomplete astragalar facet

Metatarsals

RR 47 Peripheral metatarsal, crushed epiphysis and plantar tubercle, lacks distal half RR 76 Central metatarsal, broken plantar tubercle, lacks distal half of bone RR 88 Peripheral metatarsal with fragmentary proximal epiphysis RR 105 Central metatarsal, lacks proximal aspect and shaft RR 139 Right metatarsal III, proximal cortex damaged, plantar tubercle crushed RR 158 Central metatarsal, crushed, lacking distal part of midshaft and plantar tubercle RR 199 Peripheral metatarsal with fragmentary plantar tubercle,dorsoventrally crushed shaft RR 225 Left metatarsal IV with fragmentary plantar tubercle RR 291 Distal end of central metatarsal, lacking proximal half

Pedal phalanges

RR 37 Phalanx proximal, central RR 118 Phalanx proximal, central lacking distal epiphysis RR 132 Phalanx proximal, central, with sheared shaft RR 91 Proximal pedal phalanx, peripheral, sheared shaft RR 126 Proximal pedal phalanx, peripheral, sheared shaft RR 200 Proximal pedal phalanx, peripheral RR 231 Proximal pedal phalanx, peripheral RR 236 Proximal pedal phalanx, peripheral RR 275 Proximal pedal phalanx, peripheral RR 293 Proximal pedal phalanx, peripheral, 2, lacks distal end RR 19 Pedal intermediate phalanx RR 34 Pedal intermediate phalanx RR 114 Pedal intermediate phalanx RR 125 Pedal intermediate phalanx RR 181 Pedal intermediate phalanx RR 230 Pedal intermediate phalanx RR 276 Pedal intermediate phalanx RR 277 Pedal intermediate phalanx