HAVERSIAN REMODELING IN PRIMATE LIMB BONES: EFFECTS OF LOADING MAGNITUDE, FREQUENCY, AND STRAIN MODE By SUSAN ELIZABETH LAD A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2018
2018 Susan Elizabeth Lad
To my grandmother, Delores T. Lad PhD 1919 2011
4 ACKNOWLEDGMENTS This dissertation was completed with endless support, encouragement, and advice from a number of people to whom I would like to express my gratitude. First and foremost I want to thank my advisor Dr. David Daegling for his guidance and patience these past six years. I always left our m eetings feeling reassured and motivated. I also want to thank my committee Drs. Jon Bloch, Marty Cohn, and Valerie DeLeon, fo r challenging me and always encouraging me to be confident I am fortunate to have several other mentors : Dr. Scott McGraw, who ha s been there since I was a freshman at OSU with no idea what I was getting myself into; Dr. Erin Kane who was my undergrad grad student mentor and has become a wonderful friend and continuous source of support ; and Dr. John Krigbaum, who always makes me fe el like he is on my team. I also want to thank Dr. JD Pampush for always being available as a labmate, friend, mentor, supervisor, and big brother. I will always be appreciative of the time and opportunites he has given to me. Parts of this dissertation wo uld not have been possible without the assistance of Kay Lee Summerville in obtaining mangabey femora from Yerkes National Primate Research Center, and Drs. Carol Dirain and Ashok Kumar, who generously donated guinea pig and rat cadavers, respectively, to my project. Aspects of this projec t were supported financially by the following: NSF BCS 1440278; NSF BCS 1440532 ; ORIP/OD P51OD011132 I want to ac knowledge friends and fellow graduate students who have mad e my time at UF most enjoyable, particularly Kim Le, Dr. Michala Stock, Dr. Paul Morse, Lauren Cirino, Amanda Friend, Elise Geissler, Taylor Polvadore, Jordan Traff, Sarah Zaleski, Andree Cunningha m, Jennifer Massimin, Janet Finlayson and Kylie
5 Williamson I especially want to acknowledge Dr. Katie Bailey for being there since the beginning and doing this whole grad school craziness with me every step of the way. I am forever grateful to Dad Lyle Mom, Larry, and Clare who have for many years made me feel capable of anything I wanted t o do and supported me in many different ways. And finally, so many thanks are owed to Sean Moran for being my support system day in and day out for the past several years ( and also for feeding me and doing all the dishes in the final stretch of the dissertation )
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 9 LIST OF FIGURES ................................ ................................ ................................ ........ 11 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 A Brief Introduction to Bone Remodeling ................................ ................................ 17 Format of the Dissertation ................................ ................................ ....................... 19 2 BACKGROUND ................................ ................................ ................................ ...... 25 Historical Review of Bone Biology and Biomechanical Adaptation ......................... 25 ................................ ................................ ................................ ....... 25 The New Bone Biology ................................ ................................ ..................... 26 The mechanostat hypothesis ................................ ................................ ..... 27 The mechanostat and bone remodeling ................................ ..................... 29 Non mechanical factors ................................ ................................ ............. 31 Haversian Remodeling ................................ ................................ ............................ 32 The ARF Sequence ................................ ................................ .......................... 34 Remodeling Versus Modeling ................................ ................................ ........... 36 3 BONE HISTOMORPHOLOGY METHODS ................................ ............................. 39 Thin Section Preparation ................................ ................................ ........................ 39 Imaging and Data Collection ................................ ................................ ................... 40 Targeted Remodeling ................................ ................................ ............................. 40 4 THE INFLUENCE OF LEAPING FREQUENCY ON SECONDARY BONE IN CERCOPITHECID PRIMATES ................................ ................................ ............... 42 Introduction ................................ ................................ ................................ ............. 42 Methods ................................ ................................ ................................ .................. 49 Results ................................ ................................ ................................ .................... 53 Discussion ................................ ................................ ................................ .............. 53 Null Hypothesis: Mineral Homeostasis ................................ ............................. 54 Load Magnitude ................................ ................................ ................................ 54 Load Frequency ................................ ................................ ............................... 55 Phylogenetic Effects ................................ ................................ ......................... 56
7 General Remarks ................................ ................................ ............................. 58 Conclusion ................................ ................................ ................................ .............. 61 5 SPATIAL DISTRIBUTION AND SIZE OF SECONDARY OSTEONS IN FEMORAL AND HUMERAL MIDSHAFTS OF CERCOPITHECID PRIMATES ...... 75 Methods ................................ ................................ ................................ .................. 79 Results ................................ ................................ ................................ .................... 81 Discussion ................................ ................................ ................................ .............. 82 Load Complexity ................................ ................................ ............................... 82 Age at Death ................................ ................................ ................................ .... 86 Osteon Cross sectional Area ................................ ................................ ............ 87 Additional Considerations ................................ ................................ ................. 87 Conclusion ................................ ................................ ................................ .............. 88 6 BONE REMODELING IN THE MAC AQUE ( Macaca fascicularis ) SKELETON: THE EFFECTS OF STRAIN MODE, MAGNITUDE, AND FREQUENCY ............... 99 Introduction ................................ ................................ ................................ ............. 99 Strain Magnitude and Frequency ................................ ................................ ... 100 Strain Mode ................................ ................................ ................................ .... 102 Methods ................................ ................................ ................................ ................ 104 Sectioning Methods and Data Collection ................................ ........................ 104 Statistical Analysis ................................ ................................ .......................... 106 Result s ................................ ................................ ................................ .................. 106 Among Bone Comparisons ................................ ................................ ............. 106 Femur and Tibia Quadrants ................................ ................................ ............ 107 Discussion ................................ ................................ ................................ ............ 108 Strain Magnitude and Frequency ................................ ................................ ... 108 Strain Mode ................................ ................................ ................................ .... 109 Osteon population density and relative osteonal area ............................. 109 Osteon Cross Sectional Area ................................ ................................ ... 111 Conclusion ................................ ................................ ................................ ............ 112 7 HAVERSIAN REMO DELING IN WILD AND CAPTIVE SOOTY MANGABEYS ( Cercocebus atys ) ................................ ................................ ................................ 127 Methods ................................ ................................ ................................ ................ 129 Results ................................ ................................ ................................ .................. 130 Discussion ................................ ................................ ................................ ............ 131 8 THE ABSENCE OF SECONDARY OSTEONS IN AGED RATS .......................... 137 Introduction ................................ ................................ ................................ ........... 137 Methods ................................ ................................ ................................ ................ 141 Results ................................ ................................ ................................ .................. 142 Discussion ................................ ................................ ................................ ............ 142 Conclusion ................................ ................................ ................................ ............ 146
8 9 DISCUSSION ................................ ................................ ................................ ....... 153 Summary and Recap ................................ ................................ ............................ 153 Implications of Combined Results ................................ ................................ ......... 15 8 Hypotheses ................................ ................................ ................................ .... 158 Main Questions ................................ ................................ .............................. 160 10 CONCLUSION ................................ ................................ ................................ ...... 162 Bone Remodelin g as a Tool for Behavioral Inferences ................................ ......... 162 Future Directions ................................ ................................ ................................ .. 164 LIST OF REFERENCES ................................ ................................ ............................. 169 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 188
9 LIST OF TABLES Table page 4 1 Species locomotor profiles ................................ ................................ ...................... 63 4 2 Activity budgets of Ta monkey species ................................ ................................ .. 63 4 3 Results of Shapiro Wilk normality tests for each variable for the femur and humerus ................................ ................................ ................................ ............. 63 4 4 Heidelberger Welc h results for tests of stationarity and interval halfwidth of the MCMCs for each variable per bone ................................ ................................ .... 64 4 5 Summary of the data for each variable by species for the femur and humerus ...... 64 5 1 Results of the Shapiro Wilk normality tests for each variable within femoral and humeral midshafts of each species ................................ ................................ .... 90 5 2 Summar y of OPD,On.Ar, and %HAV data for each region of the femur for each s pecies sampled. ................................ ................................ ................................ 91 5 3 Summary of OPD,On.Ar, and %HAV d ata for each region of the humerus for each species sampled ................................ ................................ ........................ 92 5 4 Results of the resampled ANOVAs testing for differences in OPD and %HAV between quadrants of the femur and humerus in each species. ......................... 93 5 5 Results of the nested ANOVAs testing for differences in On.Ar between quadrants of the femur and humerus within each indi vidual ............................... 93 6 1 OPD and %HAV in midshaft thin sections of macaque ( Macaca fascicularis ) long bones ................................ ................................ ................................ ........ 113 6 2 On.Ar in midshaft thin sections of macaque ( Macaca fascicularis ) long bones ..... 113 6 3 Shapiro Wilk norm ality test results for all data ................................ ...................... 114 6 4 Pairwi se t test p values for OPD in the femur, tibia, fibula, and rib ....................... 114 6 5 Pairwise t test p values f or On.Ar in the femur, tibia, fibula, and rib ...................... 114 6 6 OPD and %HAV in quadrants of macaque ( Macaca fascicularis ) femora l an d tibial midshaft thin sections ................................ ................................ ............... 115 6 7 On.Ar in quadrants of macaque ( Macaca fascicularis ) femoral and tibial midshaft thin sections. ................................ ................................ ...................... 116 6 8 ANOVA results for OPD and %HAV in femur and tibia quadrants ........................ 117
10 6 9 Pairwise t test p values for OPD and %HAV in quadrants of the tibia ................... 117 6 10 Nested ANOVA results for femoral and tibi al On.Ar ................................ ............ 117 7 1 OPD and %HAV data f or wild and captive Cercocebus atys ................................ 134 7 2 Summar y of OPD and %HAV in femoral midshafts of wild and captive Cercocebus atys ................................ ................................ ............................... 134
11 LIST OF FIGURES Figure page 2 1 Feedback model of the mechanostat modified from Ruff et al. (2006) .................... 38 2 2 Bone remodeling cutting bone formed by the basic multicellular unit (BMU) in two different cross sectio nal views ................................ ................................ ..... 38 4 1 A Piliocolobus badius femoral midshaft thin section at 100x magnification, highlighting some second ary osteons and oste on fragments ............................. 65 4 2 Trace and density plots for the mo del parameters of the femoral OPD MCMC posterior distribution ................................ ................................ ........................... 66 4 3 Trace and density plots for the model parameters of the femoral On.Ar MCMC posterior distribution ................................ ................................ ........................... 67 4 4 Trace and density plots for the model parameters of the femoral %HAV MCMC posterior distribution ................................ ................................ ........................... 68 4 5 Trace and density plots for the mo del parameters of the hume ral OPD MCMC posterior distribution ................................ ................................ ........................... 69 4 6 Trace and density plots for the mo del parameters of the hume ral On.Ar MCMC posterior distribution ................................ ................................ ........................... 70 4 7 Trace and density plots for the mo del parameters of the hume ral %HAV MCMC posterior distribution ................................ ................................ ........................... 71 4 8 O PD, On.Ar, and %HAV in the femur of low and high frequency leapers ............... 72 4 9 OPD, On.Ar, and %HAV in the humerus of low and high frequency leapers .......... 73 4 10 OPD in the femur, humerus, and mandible within the same individuals ................ 74 5 1 S chematic representation of long bone cross section divided into the four quadrants among which OPD, %HA V, and On.Ar data were compared ............ 94 5 2 OPD in the four quadrants of Cercocebus atys humeral midshaft .......................... 95 5 3 %HAV in the four quadra nts of Cercocebus atys humeral midshaft ........................ 96 5 4 On.Ar in the four midshaft quadrants of the femur in Cercocebus atys Cercopithec us diana Colobus polykomos and Piliocolobus badius .................. 97 5 5 On.Ar in the four midshaft quadrants of the humerus in Cercocebus atys Cercopithecus diana Colobus polykomos and Piliocolobus badius .................. 98
12 6 1 Boxplot of OPD in macaque femora, tibiae, ribs, and fibulae ................................ 118 6 2 Boxplot of O n.Ar in macaque femora, tibiae, ribs, and fibulae .............................. 119 6 3 Boxplot of %HAV in macaque femora, tibiae, ribs, and fibulae ............................. 120 6 4 Boxplot of OPD in femoral midshaft quadrants ................................ ..................... 121 6 5 Box plot of %HAV in femoral midshaft quadrants ................................ ................... 122 6 6 Boxplo t of OPD in tibial midshaft quadrants ................................ .......................... 123 6 7 Box plot of %HAV in tibial midshaft quadrants ................................ ....................... 124 6 6 Femoral mean z rank ed On.Ar interaction plot ................................ ..................... 125 6 7 Tibial mean z rank ed On.Ar interaction plot ................................ .......................... 126 7 1 OPD in wild and captive Cercocebus atys femoral midshafts ............................... 135 7 2 %HAV in wild and captive Cercocebus atys femoral midshafts ............................ 136 8 1 Rat bone depicting points on the diaphysis from which thin sections were prepared ................................ ................................ ................................ ........... 148 8 2 Composite image of thin sections prepared from the distal diaphysis of a rat femur ................................ ................................ ................................ ................ 149 8 3 Posteromedial portion of a thin section from the distal diaphysis of a rat femur .... 150 8 4 Pos terolateral portion of a thin section from the distal diaphysis of a rat femur .... 151 8 5 Lateral portion of a thin section from th e midshaft of a guinea pig femur depicting primary osteons ................................ ................................ ................. 152
13 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy HAVERSIAN REMODELING IN PRIMATE LIMB BONES: EFFECTS OF LOADING MAGNITUDE, FREQUENCY, AND STRAIN MODE By Susan Elizabeth Lad December 2018 Chair: David J. Daegl ing Major: Anthropology Haversian remodeling is the resorption and subsequent replacement of cortical bone, resulting in the formation of secondary osteons. Remodeling is a process that allows the skeleton to maintain structural integrity and adapt to me chanical loading throughout life. However, exactly how closely remodeling is tied to loading parameters such as strain magnitude, frequency, and mode is not entirely understood This dissertation address es three main questions about bone remodeling:  Is the majority of bone remodeling throughout the skeleton targeted to microdamage caused by mechanical loading, or is it non targeted, occurring more stochastically to aid in mineral homeostasis?  How closely do the density and distribution of secondary bone reflect the loading history (i.e., strain magnitude, frequency, mod e) ?  Is there variation in secondary bone density and distribution within the skeleton, and can variation be explained by the unique loading parameters of different skeletal elements ? This project used histological methods to asses s osteon population density (OPD), relative osteonal area (%HAV), and osteon cross sectional area (On.Ar) across taxa, within different skeletal elements and within different regions of skeletal e lements
14 The main findings are as follows:  loading frequency has a stronger effect on bone remodeling activity than previously appreciated under strict interpretation of the mechanostat model of bone adaptation ;  the spatial distribution of secondar y osteons corresponds to strain distributions in bones with low load complexi t y, and regions of compression have more secondary bone than regions of tension;  the occurrence of Haversian remodeling might be lifespan dependent, such that mammals with ver y short lifespans do not experience Haversian remodeling. The results suggest that bone remodeling in the examined bones is targeted, is closely related to loading parameters, and that it varies throughout the skeleton according to the unique loading histo ry of individual bones. Together this means that the density and distribution of secondary bone may be a useful tool in some contexts for making behavioral inferences in past populations.
15 CHAPTER 1 INTRODUCTION The following dissertation is concerned with understanding the functional skeleton. To this end, three main questions regarding bone remodeling are addressed.  There is th ought to be a distinction between remodeling that is targeted to regions of microdamage resulting from mechanical loading, and that which is nontargeted (i.e., stochastic) and mediates mineral homeostasis (Martin et al. 2015) but how likely is it that the observed remodeling in a given region of the skeleton is targeted rather than nontargeted? Identifying the relative amount of targeted and nontargeted remodeling throughout the skeleton will determine how reliable remodeling may be for interpreting distri butions of stress and strain, and, ultimately, the behavioral patterns of the animal to which the skeletal materials belong.  To what extent do specific aspects of mechanical loading, including load magnitude, load frequency, and strain mode, each contr ibute to the mediation of targeted remodeling? In other words, how accurately and precisely does remodeling reflect load history? Identifying the specific contexts under which targeted remodeling tends to occur may allow for unambiguous interpretations of remodeling data in terms of load history and behavior.  How much does remodeling loading environments of different skeletal elements? This final question is important for determining if remodeling is equally useful for inferring load history in all parts of the skeleton. The ultimate goal in addressing these issues is to determine the utility of bone remodeling for inferring the loading regimes of bones for which in v ivo strain gage data
16 cannot be obtained. Such cases include analysis of fossil skeletal material, especially that of primates, and skeletal remains from bioarchaeological contexts, and even cases in which it is unethical or methodologically challenging to place strain gages in vivo The capacity to remodel is an evolutionary adaptation, but the degree to which bone is remodeled is a norm of reaction, occurring during life. For this reason, the amount and distribution of secondary bone may be particularly in formative for interpreting load history because it is, presumably, directly reflective of the behaviors an individual was doing during life. Additionally, Rabey et al. (2015) showed that bone microstructure may be a more reliable indicator of locomotor act ivity than gross morphological measures, such as muscle attachment morphology. Histological methods have already been used to examine the microstructure of fossil and ancient bone in order to infer various life history variables. These include age at death (Abbott et al. 1996; Streeter et al. 2010) and developmental patterns of fossil hominins (e.g., Bromage 1989; Rosas and Martinez Maza 2010) and even ontogenetic changes in dinosaurs (e.g., Chinsamy 1990; Chinsamy 1993; Chinsamy 1995; Chinsamy 1997) Vari ations in patterns of bone remodeling have also been documented in past populations of Homo (Abbott et al. 1996; Pfeiffer 1998; Pfeiffer et al. 2006; Pfeiffer and Zehr 1996; Sawada et al. 2004; Streeter et al. 2010) and have even been used to roughly infer activity patterns (Burr et al. 1990; Mulhern and Van Gerven 1997) Resolving questions about the nature of bone remodeling will establish its usefulness as a tool for inferring load history. Such inferences may allow us to extrapolate the behavioral patte rns of extinct animals and past human populations, such as how they moved around their environment or what
17 types of foods they ate. These behavioral inferences are important for reconstructing the past and making hypotheses about evolutionary processes. A Brief Introduction to Bone Remodeling Bone remodeling is a process that requires coordinated activity of osteoclasts and osteoblasts to first resorb existing bone and then replace it with newly formed bone. Remodeling has two known primary functions :  allowing the skeleton to adapt to its mechanical environment and  maintaining mineral homeostasis Mechanical loads are applied to bone through regular activity. This regular loading causes fatigue over time and can result in microdamage in the form of microscopic cracks. Remodeling removes and replaces this damaged bone, serving a repair function that maintains the structural integrity of the skeleton (Burr 2002; Burr and Martin 1993; Burr et al. 1985; Carter 1984; Carter and Hayes 1977a; Mori and Burr 1993) In addition to repairing damaged bone, remodeling also alters the mechanical properties of bone (Carter and Hayes 1977b; Carter et al. 1976; Currey 1959; Reilly and Burstein 1974) and, by (Mar tin et al. 2015; Piekarski 1970; Pope and Murphy 1974) This can reduce the capacity for cracks to spread, preventing or reducing further damage accumulation. Thus, one function of remodeling is to allow bone to adapt to its mechanical environment by repai ring it and preventing further damage. The mineral homeostasis function of remodeling is less biomechanical and more tied to systemic metabolic factors. Remodeling serves as a mechanism for regulating mineral homeostasis by liberating calcium reservoirs in bone (Bouvier and Hylander 1996; Burr 2002; Martin et al. 2015) Remodeling can be considered to be either targeted or nontargeted (Martin et al. 2015) depending on which of these functions it serves in a given area of the skeleton.
18 The mechanically ind uced remodeling is called targeted because it is a localized response to events (i.e., microdamage, osteocyte apoptosis) that signal the activation process of remodeling (Allen and Burr 2014) Targeted remodeling is thought to be efficient because bone is remodeled only when and where necessary, rather than at a systematic rate throughout the skeleton. Remodeling related to mineral homeostasis is nontargeted because it is not site specific but is somewhat stochastic (Burr 2002) potentially occurring anywhe re in the skeleton as long as it is not detrimental to structural integrity. Osteocyte apoptosis, which can be caused by microdamage or hormonal changes, is the most common way of activating osteoclasts and initiating the resorption phase of remodeling. Wh en microdamage causes osteocyte apoptosis, undamaged osteocytes further from the area of microdamage produce antiapoptotic signals, which guide the basic multicellular unit (BMU) to the area of microdamage (Allen and Burr 2014) Burr (2002) estimated that only 30% of remodeling targets microdamage, which leaves 70% to be nontargeted, but a theoretical model by Martin (2002) suggests that all intracortical remodeling, in the absence of trauma, is associated with microdamage. The new bone produced by remodel ing is called secondary bone because it replaces primary bone. It can also be called osteonal bone, because it is laid down in concentric layers (lamellae) that surround a Haversian canal (through which vasculature passes), and these structures are known a s secondary osteons. Secondary osteons are visibly distinguishable from the surrounding lamellar bone because of the presence of the cement or reversal line, which is visible under a microscope. The cement line is the interface between the osteon and the s urrounding matrix, and is a remnant of the
19 reversal from bone resorption to formation. Secondary osteons are not to be confused with primary osteons, which appear similar but lack a cement line because they are produced during bone formation that is not pr eceded by resorption (Crowder and Stout 2012) Blood vessels that have been incorporated into compact bone during growth are also sometimes referred to as primary osteons. Remodeling can be quantified using histological methods in which osteons are viewed in cross section. The size, density, and distribution of osteons in a given region of bone can be measured in this view. Several variables indicative of remodeling activity were measured for the purposes of this dissertation: osteon population density (OP D), osteon cross sectional area (On.Ar), and percent osteonal area (%HAV). OPD is the number of intact secondary osteons plus the number of fragmentary osteons per area of bone. Fragmentary osteons, also referred to as interstitial bone, are the remnants o f previously formed secondary osteons that have been partially resorbed and replaced by newer osteons. A fragmentary osteon that has lost its Haversian canal is only recognizable by the presence of a cement line. On.Ar is the area contained inside the ceme nt line. %HAV is the combined area of all intact and fragmentary osteons (the area of fragmentary osteons is also that which is contained by the cement line) divided by the total area of the bone, and multiplied by 100. Format of the Dissertation W hat follows this introduction is a background chapter, which includes a historical review of the study of bone remodeling and a review of the literature on bone remodeling and osteon morphology that pertains to the hypotheses tested. Then there are five ch apters presented in a modular format, meaning that each presents an independent study but they address interrelated issues regarding bone remodeling.
20 Following that are discussion and conclusion chapters that summarize and combine the results of the disser tation as a whole. Brief descriptions of each chapter are presented below. Chapter 2 This chapter provides the background information for the disse rtation and is divided into two sections. The first is a historical review of bone biology, including the p aradigms of and their theoretical underpinnings, and a discussion of how bone remodeling fits into these paradigms. The second section is a review of Haversian remodeling, the activation resorption formation (ARF) sequence and the cells that make up the BMU and the differences between remodeling and modeling. Chapter 3 The third chapter describes the thin sectioning, imaging, and data collection methods common to subsequent chapters of the dissertation There is a brief discussion of targeted and non targeted remodeling. Chapter 4 This chapter examines OPD, On.Ar, and %HAV in midshaft thin sections of the femur and humerus in four cercopithecid species from Ta Forest, Cte Colobus polykomos Piliocolobus bad ius Cercopithecus diana and Cercocebus atys These species differ in their locomotor behaviors, especially in leaping frequency. The null hypothesis is that measures of bone remodeling do not vary among the species. There are several alternative hypothes es.  If remodeling depends on the magnitude of loading, then the species that leap often ( C. polykomos and P. badius ) should have greater remodeling in their limbs than low frequency leapers since leaping requires high propulsive and braking forces, dur ing takeoff and landing, respectively.  If remodeling depends more on load frequency, then the magnitude of
21 loads is less important than the overall proportion of the energy budget spent moving. Cercopithecus diana is significantly more active than the other species, and is thus expected to have the most secondary bone.  If phylogenetic effects, unrelated to locomotion or activity budget, determine the rate of remodeling, then the amount of secondary bone is likely to be similar among species belongin g to the same subfamily. In this case, Colobus polykomos and Piliocolobus badius should be similar to each other, as they are both colobines, but different from Cercopithecus diana and Cercocebus atys which are both cercopithecines.  If remodeling is p rimarily nontargeted in these species then no taxonomic differences are expected, or at least none that align with behavioral differences. This aim of this chapter is to set up hypotheses to be tested in subsequent chapters, based on the results, and to ge nerally test whether remodeling can be tied to aspects of behavior. Results that align with the third or fourth hypothesis would suggest that further testing for behavioral explanations for remodeling might not be fruitful. Chapter 5 Limbs that are involved in locomotion are generally thought to be loaded in bending, meaning that opposite sides of the cortex are under different modes of strain (e.g., tension versus compression). The null hypothesis is that the amount of secondary bone does not differ between regions of the femoral and humeral midshaft despite any differences in strain mode. The alternative hypothesis is that regional differences in secondary bone (OPD, On.Ar, and %HAV) correspond to differences in local strain mode. C ortical bone is generally weaker under tension than compression, potentially accumulating more microdamage, and it has been proposed that osteons of different sizes are optimal for different strain modes. The sample used here is identical
22 to that describe for the previous chapter; however, the midshaft is divided into quadrants (anterior, posterior, medial, lateral) in order to test for differences among them. The goal of this chapter is to examine how localized remodeling is (i.e., whether remodeling varies within a singl e skeletal element) and if it can be tied to strain mode. If bone remodeling is closely tied to strain mode then more specific behavioral interpretations can be made from it. Chapter 6 This chapter compares OPD, On.Ar, and %HAV in the tibia, fibula, fem ur, and rib of five Macaca fascicularis individuals. The null hypothesis is that there are no differences in the amount of secondary bone in different skeletal elements. Under this hypothesis, there is a systemic rate of remodeling that does not depend on the loading environments of individual bones. This would represent nontargeted remodeling. Th e alternative hypothesis is that there will be variation in secondary bone among skeletal elements because of their unique loading histories. This would be targete d. If the latter hypothesis is supported, then two more hypotheses can be addressed. The first is that load magnitude determines the amount of targeted remodeling. Under this hypothesis, elements loaded predominantly by gravitational forces (i.e., femur, t ibia) will have more secondary bone than those that are not weight bearing (i.e., rib, fibula). The tibia fibula comparison may be particularly informative because the tibia is weight bearing, whereas the fibula is not, but they are presumably loaded with the same frequency; i.e., every time a step is taken. The second hypothesis is that the amount of secondary bone depends on load frequency. Under this hypothesis, the rib should have more secondary bone than any of the limb bones since it is presumably loa ded constantly in synchronization with respiration. Other
23 expectations under this hypothesis are that the tibia and fibula should have fairly similar amounts of secondary bone since the frequency of loading in these two bones should be the same, despite th e difference in load magnitude. Chapter 7 This chapter also tests the hypothesis that load frequency is the primary factor related to remodeling. Whereas the previous chapter tests this notion within an individual skeleton, this chapter will examine OPD, On.Ar, and %HAV in Cercocebus atys femora from wild and captive individuals. The premise is that wild individuals are much more active than their captive conspecifics, so the frequency of loading in their fem ora should differ. The magnitude of loading sho uld be relatively similar given that they are similar in body size and in their locomotor behavior. If the wild individuals have more secondary bone, then it can be interpreted that their more active lifestyle results in more microdamage, and thus more tar geted remodeling. If there is no difference between wild and captive, then secondary bone may not be an ideal measure for inferring activity level. Chapter 8 Chapter 8 examines the bone microstructure of two rodents: Rattus norvegicus and Cavia porcellus It is generally taken for granted that rodents do not have secondary osteons. However, reports on comparative mammal bone microstructure in the literature provide conflicting information about the presence of secondary osteons in rats (cf. Martiniakov e t al. 2006; Singh et al. 1974) These previous studies examined rats that were considered adults but were only 4 6 months of age. The rats examined in this chapter are 2 years of age. Microdamage accumulates with age, so examining older individuals increas es the likelihood of osteon formation. Guinea pigs are slightly larger rodents that have much longer lifespans than rats. The
24 guinea pigs used in this sample, while adults, are much younger than the rats at 3 months of age. Older guinea pigs were not avail able. This chapter is a discussion of why osteons form in some animals and not others. Factors that influence remodeling may include body size, life span, and the size of the bones. Understanding which animals do have osteons, and generally how much Havers ian remodeling occurs in those animals, may help determine the minimum requirements for microdamage to occur and for remodeling to be a necessary response. This may help identify for which fossil animals secondary bone may be a useful tool for behavioral i nference. Chapter 9 Chapter 9 synthesizes the results and findings of chapters 4 8 and bridges topics discussed in each. The implications of the dissertation as a whole are discussed in the context of the three main questions presented at the beginning of this chapter. Chapter 10 The final chapter presents the conclusions and main poi nts of the dissertation. New questions and future directions are proposed.
25 CHAPTER 2 BACKGROUND Historical Review of Bone Biology and Biomechanical Adaptation This dissertation largely concerns the biomechanical factors that influence bone remodeling in the primate skeleton. The idea that the mechanical environment determines bone form, to some extent, is not new it has long been accepted that the skeleton has biomechanical functions. However, the biomechanical functions were architecture is constrained by non mechanical factors. A new paradigm of bone biology emerged and took hold in the second half of the 20 th century that recognizes that both non mechanical and mechanical factors influence bone architecture, but gives dominance to the biomechanical function of bone. Below is a brief history of how this paradigm came to be and some of its limitations. The notion that bone adapts to its mechanical environment during life dates to at least as early as Galileo (1638) architecture is influenced by mechanical forces acti ng upon it, that its organization optimizes bone strength in relation to the amount of material used (Bell, 1827; Bourgey, 1832) and that excess bone does not form where it does not need to be (Ward, 1838) In 1866, engineer Karl Culmann and anatomist Her mann von Meyer noticed similarities between the principal stresses of a crane and the trabecul ar architecture within a proximal femur, and hypothesized that bone trabeculae are aligned in the direction of principal stresses (Martin et al., 2015) This beca me known as the Culmann Meyer stress trajectory hypothesis. During this time, the concepts of homeostasis and self
26 regulation were new (Cooper, 2008) and Wilhelm Roux (1881) applied them to bone and proposed that bone cells form and resorb bone in respons e to mechanical stimuli, resulting in the trabecular architecture that lines up with the principal stress directions. The implication of this was that bone is a dynamic tissue, rather than just an optimally designed structure. Julius Wolff (1892) synthesiz ed all of these concepts into the idea that bone is a dynamic tissue that responds to its physiological and mechanical environments and is able to modify its structure by sensing mechanical loads. He thought that there were mathematical laws defining bone structure in relation to loading. The mathematical rules were later found to be false because Wolff mo deled bones as solid, homogenous structures under static loads, none of which is true (Cowin, 2001) but the field of bone biology has since been largely based on this general idea, which is concept, but he is credited with integrating them and was a prolific writer on the topic. (Ruff et al., 2006) can be summarized by three main concepts (Roesler, 1981, 1987) based on the works of the early researchers noted above:  bone strength is optimized with respect to its weight,  bone trabeculae are aligned with principal stress directi ons, and  bone structure is regulated by cells within bone that respond to mechanical stimuli. The New Bone Biology A large body of and by the midd le of the of the 20 th century the idea that the primary function of bone is mechanical load bearing and that other functions, such as calcium storage and homeostasis are secondary (Frost, 1998; Jee, 2005; Stout and Crowder, 2012) was taking hold The
27 lite rature revolving around this new paradigm gr e w rapidly in large part because of the work of Harold Frost, an orthopedic surgeon who published over 500 articles on bone methodology and to what are now considered basics of bone histomorphology and characteristics of bone remodeling (e.g., Frost, 1958, 1960b, 1960a, 1964, 1969) He later proposed the mechanostat hypothesis and was a leading figure of what would become known as the Ut ah P aradigm. The mechanostat hypothesis Thompson (1961) first proposed that bone adaptation is a response to mechanical deformation (i.e., strain). Frost claimed that strain magnitude was the signal to which bone cells respond and proposed the mechanost at (Frost, 1987) as an analogy for the mechanism s that determine bone mass through bone growth, modeling, and remodeling. A thermostat is set to an ideal temperature and is programmed to turn on when there is an error, me aning the ambient temperature diffe rs from the set ideal, or off when there is no error (i.e., no departure from the ideal temperature). Similarly, the mechanostat is set to maintain optimum strain level in bone, and is dependent on bone mass relative to applied load. Bone mass is maintained unless strain increases, in which case more bone is deposited and strain returns to the optimum range, or strain decreases, in which case bone is resorbed until strain increases and returns to the optimum range (Figure 2 1). Minimum effective strains (MES) (Frost, 1983) are the thresholds above and below which bone formation and resorption are activated or suppressed A s long as strains are above the MES bone will be retained (Frost, 1997, 1998) meaning t he resorption rate will not exceed the rate of formation. This is called
28 the conservation mode. However, if strains drop below the MES, bone resorption increases, resulting in pathological conditions. Frost did recognize that hormonal factors can affect bone cells but gave primacy to biomechanical loads as moderators of bone mass. According to his hypothesis hormones can alter the MES set points (Frost, 1987) tricking bone cells into thinking there is an error when there is not (e.g., parathyroid hormone increases osteoblast sensitivity, thus causing bone formation at lower strain levels). However, the stimulus for bone formation or resorption is still dependent on strain magnitude. lted in the Utah paradigm (Frost, 1998; Jee, 2005; Stout and Crowder, 2012) The Utah paradigm maintains structural integrity under mechanical loading. It was generally acc epted that the load bearing performance was dependent on mechanical stimuli, and that while non mechanical factors such as hormone and mineral levels, sex, and age can influence bone architecture and strength, these are not primary drivers of bone architec ture (Frost, 2000) Since the mechanostat hypothesis was proposed, there have been several critiques, or amendments to it. Turner (1999) pointed out a problem with the mechanostat: the mechanostat states that when strains are below the MES, resorption oc curs but it does not explain why non weig ht bearing bones, which may lack a strong mechanical stimulus, do not undergo disuse remodeling to the extent that they resorb away completely. Turner proposed the Principle of Cellular Accommodation Theory which st ates that bone cells adapt to their mechanical environment so that MES is site
29 specific dependent on local strain. Under this hypothesis, weight bearing elements will have higher MES than non weight bear ing elements. Evidence from the macaque face (Hyland er and Johnson, 1997) and rat ulnae ( Hsieh et al. 2001) that strain thresholds are greater in areas that habitually experience greater strain supports this hypothesis. Another alteration of the mechanostat was the osteocyte inhibitor hypothesis. Frost ( 1960b) hypothesized that when cracks destroy part of the osteocyte canaliculi network, osteocytes signal the activation phase of remodeling Burr (2002) suggested that the system of intact osteocyte canaliculi actually sends signals that inhibit osteoclast activit y and that osteocyte apoptosis disrupts that signal and allows remodeling to occur. The mechanostat and bone remodeling damaged bone and preventing future dam age to maintain mechanical efficiency (Martin et al., 2015) These functions comply with the general concept of bone functional adaptation but there are some issues with how bone remodeling fits into the mechanostat model. This dissertation aims to better understand the biomechanical factors that incite bone remodeling in order to better understand the role of remodeling in bone functional adaptation. The mechanostat model describes modeling as the overloading response and BMU based remodeling as the under loading response (Frost, 1987) Remodeling is thought to be potentially detrimental in areas of high strain because the presence of resorption cavities increases bone porosity such that the mechanical integrity could be at risk. Under this model remodeling should only occur in areas of low strain, but this does not account for the remodeling that occurs to repair microdamage caused by high
30 strains. The mechanostat does not adequately account for the remodeling that does occur in high strain regions because it does not explain how high strain both inhibits and stimulates remodeling (Martin, 2000) Another critique of the mechanostat is that it gives primacy to strain magnitude as the important signal influencing bone adaptation, whereas other factors may be equally or more important. Several key experimental studies (Hert et al., 1969; Hert et al., 1971; Liskova and Hert, 1971; Hert et al., 1972; Rubin and Lanyon, 1984) demonstrated that dynamic or cyclic loads are necessary to maintain bone mass, meaning tha t one loading event, even if strain is high, is not sufficient. Other studies have demonstrated that strain rate (i.e., the amount of deformation with respect to time) is also critical, and that high strain rate results in increased bone formation (O'Conno r et al., 1982; Turner et al., 1995; Mosley and Lanyon, 1998; Deere et al., 2012) Additionally, there seems to be a point past which bone formation ceases to increase in response to additional loading cycles per day (Rubin and Lanyon, 1984; Umemura et al ., 1997) This is because the mechanical signal saturates quickly and then the bone cells become desensitized after repeated bouts (Burr et al., 2002) Thus, there are many other factors involved in the regulation of bone mass aside from strain magnitude. Finally, Burr (2002) estimates that that 30% of bone remodeling is targeted to areas of microdamage or high strain, leaving the other 70% to be non targeted The n on targeted remodeling occurs for metabolic reasons and is not truly stochastic but is not usually site specific Other estimates (Martin, 2002) suggest that most intracortical bone remodeling is targeted, but if Burr is correct, then there is a lot of remodeling that
31 is unexplained by biomechanical factors, and therefore unaccounted for by the mechanostat. Non mechanical factors Much of modern bone research has been based on the idea that bone morphology is dependent on mechanical loading. Anthropologists in particular are interested in bone biology for what it can reveal about behavior, pathology, and age in past populations (Crowder and Stout, 2012) and tend to heavily emphasize biomechanics and mechanobiology. However, more recent research in cellular biology and evolutionary and developmental bio logy is highlighting the importance of genetic and hormonal determinants of bone morphology. Wallace et al. (2015) have demonstrated that genetic differences in strains of mice can result in differential responses in bone formation, structural geometry, an d mechanical properties to exercise, and Wallace et al. 2012 demon strated that bone morphology is determined by a combination of genetic composition and mechanical loading. Lovejoy et al. (2002) and Lovejoy et al. (2003) have even argued that to assume bon e morphology is largely a reflection of loading history is faulty. They argue that instead it is determined by gene (Lovejoy et al., 2003) Karsenty and Ferron (2012) present a review of recent research into the functions of bone and other organs at the cellular and molecular level They describe how the skeleton can be thought of in terms of whole organism physiology and that the genetic revolution has revealed that many organs have more functions than previo usly thought, and that these can be identified with a whole organism approach A whole organism approach means that (1) a given physiological function is not determined by one organ by itself, (2) homeostasis occurs when mul tiple organs have
32 opposing effects on t hat given function serving to regulate it, and (3) regulatory molecules appear in evolutionary history with the functions they regulate and not after. Bone both affects and is affected by fat accumulation, energy met ab olism, and reproduction. It has long been known that bone is influenced by endocrine factors but in molecular and cellular biology, bone is now thought to be an endocrine organ (Ducy et al., 2000; e.g., Burguera et al., 2001; Thomas and Burguera, 2002; H arada and Rodan, 2003; Elefteriou et al., 2004; Lee et al., 2007; Sato et al., 2007; Fukumoto and Martin, 2009; DiGirolamo et al., 2012; Guntur and Rosen, 2012; Karsenty and Ferron, 2012) It is important to acknowledge factors other than loading history when studying bone morphology, but the fact that genetic and hormonal factors are important does not negate the role of biomechanics in shaping bone. Environmental and genetic factors are both responsible for the full variation in bone morphology (see rev iew by Ruff et al. ) and one of the goals of this dissertation is to determining the extent to which the former influences remodeling. This dissertation tests hypotheses about the effects of loading parameters on bone remodeling, but also addresses n ull hypotheses (i.e., that mechanical loading does not affect bone remodeling). Haversian Remodeling Bone remodeling is the resorption of old bone and subsequent formation of new bone by the coupled activity of osteoblasts and osteoclasts. In intracortica l bone this results in the formation of secondary osteons (i.e., Haversian systems) and in trabecular bone, the formation of hemiosteons (partial osteons) (Parfitt, 1994) It is one way that daptations are feat ures that allow an organism to balance its biological role and the selection pressures imposed by the environment (Bock, 1980) For the purposes of this dissertation adaptation refers to
33 physiological (i.e., somatic) adaptation rather than evolutionary ada ptation, unless otherwise specified. Evolutionary adaptation is shaped by natural selection and entails changes in the genetic makeup of the organism, which are heritable. Physiological adaptation occurs during the lifetime of an individual in response to an environmental stimulus. It has a genetic basis as well. Bateson (1963) argued that physiological adaptation is necessary for survival because evolutionary adaptation happens too slowly to allow for flexibility He used the classic example of altitude acclimation to explain physiological adaptation: People who move from sea level to high altitudes experience increased heart rate and shortness of breath, which restricts behavioral and somatic flexibility. Then physiological adjustments take place (e.g., the proportion of red blood cells increases) and respiration and heart rate return to normal. This physiological adaptation allows for flexibility ; since heart rate and respiration have returned to normal, the organism can survive and the range of flexibi adaptations are norms of reaction such as the development of calluses on the feet or tanning of the skin in the sun. Bone morphology is a good example of the interaction between evolutionary and physiological adaptations. Some aspects of bone morphology have been determined over evolutionary time, i.e., the fibrolamellar structure the range of bone stiffness/toughness (Currey, 2003) and even the ability to repair microdam age via remodeling Othe r aspects of bone morphology devel op during ontogeny as a response to mechanical loading These include modeling to some extent, and remodeling The
34 r emodeling is a physiological adaptatio n, but the ability to remodel is an evolutionary adaptation. A description of bone cells, the basic multicellular unit, the activation resorption formation sequence in cortical bone, and a discussion of remodeling and modeling are provided below. This overvi ew is based on texts by Currey (2002) Crowder and Stout (2012) and Martin et al. (2015) unless otherwise indicated. The ARF Sequence There are four types of bone cells: osteoclasts are responsible for bone resorption; osteoblasts for bone formation; and osteocytes and bone lining cells, which are both responsible for bone maintenance and signaling. Each plays a role in the Activation Re sorption Formation (ARF) sequence of bone remodeling. ARF begins with the initiation of a basic multicellular unit (BMU) at the site of osteocyte apoptosis. Osteoclasts derive from hematopoietic stem cells which differentiate into osteoclast precursors and travel through the vasculature to reach basic multicellular unit initiation (BMU) sites, where they begin to remove bone. The arrival of osteoclasts in a given region of bone is the activation phase. The resorption phase begins when the osteoclasts then t unnel through bone, forming a cutting cone (Figure 2 2), which is cylindrical and roughly circular in cross section. In humans a typical cutting cone is ay radially. The resorption period lasts approximately three weeks. There is a lag between resorption and subsequent refilling (formation phase) of the cutting cone, called the reversal period. The reversal period lasts several days, during which osteoblas t precursors begin to migrate to the BMU and differentiate into osteoblasts.
35 The formation phase consists of osteoblasts lining the cutting cone and laying down bone lamellae from the outside eye in cross section. Osteoclasts and osteoblasts require vascular supply to survive, so the center of the cutting cone is not refilled, and remains open as a canal through which blood vessels pass. This canal is called the Haversian canal and the blood vessels p assing through remain after ARF is complete to supply osteocytes. In humans the Haversian canal is ~40 radially. The formation period lasts about 3 4 months in adult humans. The final result, a tube of concentric lamellae with a central Haversian canal, is called a secondary osteon (usually visualized 2 dimensionally in cross section), Haversian system (conceptualized three dimensionally), or basic structural unit (BSU). Secondary osteon is the term predominantly used throughout this dissertation. (It should be noted that remodeling in cancellous bone involves similar processes and results in hemiosteons, which are also referred to as BSUs, but this is not Haversian remodeling as no canal forms.) The bone laid down by osteoblasts is unmineralized matrix (i.e., osteoid) and mineral is deposited in two phases beginning about 10 days after formation ends. The first phase, primary mineralization, mineralizes 60 70% o f the osteoid and occurs within th ree weeks. Secondary mineralization takes place over about a year at a decreasing rate over time. As a result, newly formed osteons often have very different mechanical properties compared to older osteons, as a result of different mineral content. Once A RF is complete, some osteoblasts die but many osteocytes or bone lining cells. Osteocytes are osteoblasts that have become trapped in the bone matrix, whereas bone lining cells reside on the surfaces of bone. Both function
36 as bone maintenance cells interconnected by a network of canaliculi, though which signals are transmitted to surrounding cells. Continuous signaling among living osteocytes and bone lining cells inhibits the activation phase of remodeling. Osteocyte apopto sis or disruption of the canaliculi between them by either mechanical (e.g., microdamage) or hormonal (e.g., decreased estrogen) factors eliminates the inhibitory signal and allows ARF to begin (Martin, 2000) Remodeling Versus Modeling Remodeling should not be confused with modeling, which also entails bone resorption and formation by osteoclasts and osteoblasts but is distinct in several ways. Modeling involves growth and shaping of bones. During development long bones grow in length and diameter, and ca n also change shape. This requires the addition of new bone via endochrondral ossification (length) and periosteal appositional ossification (diameter), but also the removal of bone by osteoclasts on existing surfaces and addition of bone by osteoblasts in differs from remodeling in several ways.  The actions of osteoclasts and osteoblasts that sculpt bone during modeling are not strictly coupled the way they are in remodeling.  Modeling alters the si ze and shape of bones by removing and/or adding bone de novo. Remodeling does not change gross bone morphology but replaces pre existing bone.  Modeling continues through growth and development but ceases, for the most part, with skeletal maturity, wher eas remodeling continues throughout life.  Modeling is more or less continuous, while remodeling episodes are intermittent with more defined beginnings and ends. Haversian remodeling results in the formation of secondary osteons, whereas modeling does not result in such structures. Primary bone does have longitudinal
37 as blood vessels on the bone surface become incorporated into newly laid bone during the modeling process The canal of a primary osteon is also called a Haversian canal and concentric lamellae can be present. However, secondary osteons can be easily distinguished from primary osteons in histological thin sections by the presence of the cement line. The cemen t line (also referred to as the reversal line) is a remnant of the reversal stage of remodeling; the lag between resorption and formation. The cement line remains fairly unmineralized and lines the outer circumference of an osteon, separating its concentri canals also lack a cement line and tend to run perpendicular to the more longitudinally oriented secondary osteons.
38 Figure 2 1 Feedback model of the mechanostat in which error from the nor mal strain range causes bone deposition or bone resorption when strains are higher or lower, respectively, than the normal range. Modified from Ruff et al. (2006). Figure 2 2. Bone remodeling cutting bone formed by the basic multicellular unit (BMU) in two different cross sectional views. The multinuclear cells (spotted circles) are osteoclasts ( bone resorbing cells), and the single nucleus cells (black circles) are osteoblasts (bone forming cells). The front of the cutting cone resorbs bone at a rate
39 CHAPTER 3 BONE HISTOMORPHOLOGY METHODS Thin Section Preparation Bones and species included in the samples are described in their respective chapters. Small portions of the diaphysis were cut from each bone at selected sites and embedded in Buehler Epoxicure Resin and allowed to set for a minimum of eight hours. The resin protects the bone from damage during cutting. The embedded bo ne was attached to the arm of a Buehler Isomet low speed saw ( Illinois Tool Works, Lake Bluff, I L ) equipped with a diamond wafering blade and a from each one. In cases where the embedded bone was difficult to cut to the appropriate thickness (i.e., too delicate or slipping off the blade ), a thicker section (1mm) was c Buehler MetaServ 250 grinder polisher. Once the thin sections were the appropriate thickness, as determined by digital micrometer, they were polished using the grinder polisher and a diamond suspension fluid in order to remove any debris or striations from the saw Thin sections were stored in a solution of 70% ethanol and 30% distilled water until they were stained. To stain the sections, they were first hydrated in distilled water for 15 minutes and then placed in 10ml water with two drops Toluidine Blue O solution for 10 minutes. Once removed from the stain they were rinsed in distilled water for three minutes and then dried between two microscope slides lined with Kimwipes, with a small weight placed on top to prevent warping. Each thin section was either microscope slides or mounted to a microscope slide using Cytoseal 60 and covered with a Xylol dipped coverslip The sections placed between 2 slides were removed from the slides dipped in Xylol, and quickly placed between the slides again before
40 microscopy to increase clarity For bones in which quadrants were being compared, e ach section was divided into four equal quadrants (anterior, posterior, medial, lateral), which were ma rked on the top slide with a thin permanent marker. After imaging the sections were stored in a microscope slide box. Imaging and Data C ollection Each thin section was photographed under a compound light microscope at 100x magnification, requiring a serie s of photographs to capture the total area of the bone cortex A 1mm scale bar was photographed under the same magnification settings as each thin section Images were imported into ImageJ (Abramoff et al., 2004) and set to scale using the photograph of th e scale bar The total area of the bone cortex in each photograph was measured using the ImageJ tracing tool. The number and cross sectional areas (measured using the tracing tool in ImageJ) for all secondary osteons and fragmentary osteons were recorded. C areful attention was paid to the areas where images overl apped so as not to measure the same area or feature twice. The cortex areas were summed to get total cortical bone area of each thin section and also a total area for each quadrant, when relevant Osteon population density (OPD), osteon cross sectional area (On.Ar), and relative osteonal area (%HAV) were then measured as the data for analysis. OPD is the total number of secondary and fragmentary osteons per square millimeter of cortical bone. On.Ar is the measured area of each complete secondary osteon. %HAV is the total area of secondary bone (both complete and fragmentary osteons) per square millimeter of bone expressed as a percentage. Targeted Remodeling Burr (2002) described remodeling that occurs in response to microdamage and
41 Some percentage of remodeling throughout the skeleton is targeted and some non targeted. Burr estimate d these relative proportions were 30% and 70%, respectively, whereas Martin (2002) suggested that nearly all Haversian remodeling in the skeleton is spatially associated with microcracks with exception to cases of disease or trauma. The exact proportions of targeted and non targeted remodeling are unknown, and this presents a problem for interpreting load history from osteon densities. There is no difference in appearance between secondary osteons that form during targeted versus non targeted remodeling. I n this dissertation, it is assumed that most Haversian remodeling is targeted, however null hypotheses account for the possibility that results will not meet expectations of remodeling patterns if mechanical loading accounts for it. In other words, if remo deling does not correspond to loading parameters then it is likely that it is non targeted.
42 CHAPTER 4 THE INFLUENCE OF LEAPING FREQUENCY ON SECONDARY BONE IN CERCOPITHECID PRIMATES Introduction Bone is a complex, durable, and dynamic tissue capable of m echanically adapting to unique loading histories. One way bone adapts is via intracortical remodeling (i.e., Haversian remodeling), which is the coordinated activity of osteoclasts and osteoblasts to resorb and replace older bone with new bone, resulting i n the formation of secondary osteons. Secondary osteons are concentric layers of bone (i.e., lamellae) surrounding Haversian canals and are roughly circular in cross section. Cement lines which are artifacts of the reversal from bone resorption to bone for mation during remodeling delineate secondary osteons from the surrounding lamellar bone. Cement lines also distinguish secondary osteons from primary osteons. Primary osteons are typically the result of bone formation around vascular canals, but unlike the formation of secondary osteons, primary osteon formation does not follow a resorption phase (Crowder and Stout, 2012) Remodeling can occur as either a targeted or non targeted process (Burr et al., 1985; Burr, 1993; Mori and Burr, 1993; Burr, 2002; Mar tin et al., 2015) Non targeted remodeling is fairly stochastic, regarding its location in the skeleton, and plays a role in maintaining mineral homeostasis by liberating calcium. Targeted remodeling, on the other hand, is site specific; occurring in areas with sufficiently large amounts of microdamage, likely caused by the mechanical deformation of bone due to external loading. Thus, targeted remodeling likely serves a repair function. In addition to targeting microdamage, remodeling may also increase over all bone toughness. Remodeled bone is generally more compliant than primary lamellar bone, because it is
43 less mineralized (Currey, 1959; Reilly and Burstein, 1974; Reilly et al., 1974; Carter et al., 1976; Carter and Hayes, 1977) Increased compliance mean s that less energy is required to initiate a crack, but more energy is necessary to propagate it (Martin et al., 2015) Additionally, cement lines on the boundaries of osteons can stop or deflect a phenomenon in which an osteon separates from the surrounding matrix along the cement line and bridges a crack (Piekarski, 1970) Consequently, densely packed osteons can increase overall bone toughness by reducing crack propagation (Moyle and Bowden, 1984; Gibson et al., 2006; Mohsin et al., 2006; Martin et al., 2015) Remodeling thus functions to maintain and preventing crack propagation. The extent to which bone has been remodeled can be determined by measuring the density and distribution of secondary osteons in histological sections of bone (e.g., Crowder et al., 2012) Investigating these measures of remodeling and the behavioral regimes of various animals may all ow for a better understanding of how closely remodeling reflects the specific load cases applied to bones. Bone microstructure is not the only technique employed to understand the relationships between load history and skeletal form. Others include trabecu lar bone architecture (e.g., Macchiarelli et al., 1999; Ryan and Ketcham, 2005; Pontzer et al., 2006; Tsegai et al., 2013) and bone cross sectional geometry (e.g., Ruff et al., 1999; Trinkaus and Ruff, 1999a, 1999b; Polk et al., 2000; Stock and Pfeiffer, 2 001) ; however, it is recommended that the latter be used with caution (Pearson and Lieberman, 2004) as in vivo strain gage data show that the orientation of maximum stiffness often does not correspond to the primary axis of
44 bending in long bones (Szivek et al., 1992; Demes et al., 1998; Lieberman et al., 2004) Understanding the nature and limitations of the relationships between these aspects of skeletal morphology, including bone microstructure, and behavior in extant taxa may find application to paleonto logy and bioarchaeology in reconstructing the behavior of fossil taxa and past human populations, respectively. Schaffler and Burr (1984) first examined taxonomic differences within primates in bone remodeling as they relate to skeletal function. They cal culated percent osteonal bone in the femoral cortex of twenty species and found that it corresponded to general locomotor categories: arboreal quadrupedalism, terrestrial quadrupedalism, suspensory locomotion, and bipedalism. Bipeds had the greatest percen t osteonal bone, followed by suspensory primates, then terrestrial quadrupeds, and finally arboreal quadrupeds. The authors concluded that differences in percent osteonal bone were the result of the loading regimes specific to each generalized locomotor pa ttern. Schaffler and Burr (1984) suggested that locomotor differences resulted in loading regimes that incite bone remodeling to differing degrees. However, there were three major limitations in their study.  The sample size for each species was only a single individual, which is not ideal because it does not account for within species variation and it is uncertain if the sampled individuals are typical representatives for each taxon.  The locomotor categories used were very coarse and idealized; that is, a wide range of activities is included in each category. For example, there are various types and frequencies of walking, running, leaping, and climbing that fall within the locomotor repertoires of different arboreal quadrupeds (Napier and Napier, 19 67; Hunt et al., 1996) Broad locomotor categories can hinder functional interpretations of morphology because some
45 biologically meaningful locomotor variation is overlooked, meaning that potential morphological implications may be missed (Ripley, 1967; Hu nt et al., 1996)  The non human primates used were all captive zoo animals. Captive primates tend to have lower activity levels than their wild counterparts (Altmann and Muruthi, 1988) and zoo primates may also have other altered behaviors (Hosey, 2005 ) Some studies have suggested that these captivity altering behaviors can result in unusual bone structure (e.g., Fleagle and Meldrum, 1988) contexts has been relatively unde r explored in non human primates (but see Paine and Godfrey, 1997; McFarlin et al., 2008; Warshaw, 2008; Skedros et al., 2011; Lad et al., 2016; Keenan et al., 2017) Given that broad locomotor categories tend to obscure potentially informative locomotor d ifferences between taxa, it is worthwhile to examine secondary bone in the limbs of species traditionally placed within the same locomotor category, but still differing in some dimension of their locomotor repertoire. This naturally enables the isolation o f specific locomotor behaviors in the study of bone loading and remodeling. such an opportunity because their locomotor behaviors are well documented (McGraw, 1998a, 2007) Wh ile all four species examined here are primarily quadrupedal, they differ in the relative proportions of leaping, climbing, running, and walking in their overall locomotor budget (Table 1). Napier and Napier (1967) described the colobine clade as having ev olved a type of arboreal quadrupedalism characterized by leaping and some arm swinging, while cercopithecines are divided into two categories of quadrupedalism:
46 those adapted to ground walking and running, and those that engage in branch walking and runnin g, which also includes climbing. These generalizations are also true of the members of these clades at Ta; the colobines Piliocolobus badius (red colobus) and Colobus polykomos (black and white colobus) are more frequent leapers than the cercopithecines ( McGraw, 1998b, 1998a, 2007) Colobus polykomos is also the only Ta cercopithecid that engages in bounding quadrupedalism, a gait in which the two forelimbs move and contact the substrate simultaneously, followed by the two hindlimbs, which also move and contact the substrate in unison (Mittermeier an d Fleagle, 1976) Among the cercopithecines, Cercopithecus diana (Diana monkey) is an agile monkey that engages in frequent arboreal climbing while Cercocebus atys (sooty mangabey) is primarily a terrestrial quadruped. Although the leaping frequencies repo rted for Colobus polykomos are intermediate between those of Piliocolobus badius and Cercopithecus diana the leaping frequencies in Table 1 do not include bounding. Bounding is thought to require high propulsive forces in the hindlimb compared to running (Morbeck, 1976; McGraw and Daegling, 2009) Thus, we divide the taxa into two categories: the frequent leapers ( Colobus polykomos and Piliocolobus badius ) and the lower frequency leapers ( Cercopithecus diana and Cercocebus atys ). The central question addr essed here is whether bone remodeling activity is functionally linked to species specific locomotor behaviors. If it is, then differences among species in the amount of secondary bone in the limbs should be expected. The main hypothesis addressed is that l oad magnitude is primarily responsible for the microdamage that incites bone remodeling. Leaping generates high magnitude forces in the hindlimbs as they propel the body off the substrate (Kimura et al., 1979; Demes et
47 al., 1994; Franz et al., 2005) High frequency leapers are expected to encounter these high magnitude forces in their hindlimbs repeatedly. Under this hypothesis, high frequency leapers are expected to have more femoral bone remodeling than the low frequency leapers as a result of presumably greater accumulation of bone microdamage. If the forelimbs are used heavily for braking during landing, then remodeling differences among taxa in the humerus are expected to match those predicted for the femur. However, takeoff forces are greater than la nding forces (Demes et al., 1999) and the nature of landing and the frequency with which the forelimbs are used to brake during landing may differ widely among the Ta cercopithecids (McGraw, 1998a, 1998b) As a result remodeling in the humerus might not f ollow the same pattern as the in the femur. An alternative hypothesis is that load frequency (i.e., how often the limbs experience locomotor forces, regardless of the magnitude of those forces) is the ultimate instigator of bone remodeling. In this scena rio, Cercopithecus diana should have the most remodeled bone. Even though the two colobines have more frequent high magnitude forces in their hindlimbs (due to higher leaping frequency), Cercopithecus diana has more frequent loading of both the hind and f orelimbs as a result of spending more time traveling and foraging (Table 2) (McGraw, 2007) Another possibility is that remodeling differences fall along phylogenetic lines and are not due not to locomotor differences. Species that are closely related to each other tend to have higher trait similarity than distantly related species because of common descent, and are thus not statistically independent (Felsenstein, 1985) Similar traits
48 found in closely related taxa are considered to have high phylogenetic signal (Harvey and Pagel, 1991; Blomberg and Garland, 2002) and potentially high phylogenetic signal in the rate of bone remodeling cannot be ruled out. Piliocolobus badius and Colobus polykomos belong to subfamily Colobinae whereas Cercocebus atys and Cer copithecus diana belong to Cercopithecinae. Classification of the four species into two leaping frequency groups (i.e., high or low), as described for the load magnitude hypothesis presents a potential problem because doing so creates groups that correspon d with subfamily affiliation. Thus, there may be three plausible explanations for any observed differences between the two subfamilies: 1) unspecified phylogenetic effects, 2) leaping frequency differences, or 3) some combination of the two. Statistical me thods accounting for phylogeny are therefore appropriate for this analysis. Aside from functioning to maintain the structural integrity of bone, remodeling also facilitates mineral homeostasis (Bouvier and Hylander, 1996; Burr, 2002; Martin et al., 2015) Therefore, some remodeling in the body is not directly related to biomechanical loading. If the results of our analyses do not conform to expectations based on locomotor differences, then it is possible that non mechanical factors have a stronger effect on the mediation of remodeling in the humerus and femur of these species. Following from above, we identify the following hypotheses and alternative explanations.  The null hypothesis states that bone remodeling is primarily random and non targeted, a nd that its primary function is to aid in mineral homeostasis. Under this hypothesis no taxonomic differences in remodeling are expected, as there is no reason to assume the rate of non targeted remodeling should differ among them. 
49 The working hypothes is is that the magnitude of loading is primarily responsible for the accumulation of microdamage that instigates bone remodeling. In this case, the species with high leaping frequency (i.e., Piliocolobus badius and Colobus polykomos ) are expected to have m ore remodeled bone in the examined long bones than the two that leap less frequently (i.e., Cercopithecus diana and Cercocebus atys ). This prediction follows from the high magnitude of forces leaping generates in both takeoff and landing.  An alternativ e hypothesis is that load frequency regardless of magnitude is primarily responsible for microdamage accumulation and, consequently, bone remodeling. This hypothesis predicts that leaping frequency does not matter, but the proportion of the overall activit y budget spent moving does. In this scenario Cercopithecus diana would be expected to have the most secondary bone, since the other species are significantly more sedentary.  Another possibility is that unspecified phylogenetic effects, which are unrela ted to locomotor behaviors, cause different degrees of bone remodeling. Remodeling differences are expected to fall along subfamily lines if this is the case. Methods that account for phylogeny are described below to avoid conflating phylogenetic effects w ith the behavioral effects expected for hypothesis  Methods The skeletal sample consists of femora and humeri of four species ( Colobus polykomos, Piliocolobus badius, Cercocebus atys, and Cercopithecus diana ) collected from Ta National Forest, Cte natural causes. Sex and age at death are uncertain due to the opportunistic nature of specimen collection but no juveniles were used, as determined by epiphyseal fusion. Midshaft blocks were cut fr om humeri and femora (N = 5 each per species) using a
50 Buehler Isomet low speed saw with diamond wafering blade (Illinois Tool Works, Lake Bluff, IL), then embedded in resin following the methods of Crowder et al. (2012) Thin the embedded bone and polished using a Buehler Meta Serv 250 grinder polisher to remove any dust and markings from the blade. They were then stained in a Toluidine Blue O solution for 10 minutes to improve the visibility of the cement lines under the micr oscope. Each section was dried under a weight to keep it flat and was then mounted on a slide. The entire cross section was photographed at 100x magnification under a compound light microscope ( Figure 4 1). Measurements were taken from the photographs, carefully aligning them so as not to measure the same area twice. Three variables representing the number and size of secondary osteons, as well as the overall amount of remodeled bone, were measured: osteon population density (OPD ) osteon cross sectiona l area (On.Ar), and relative osteonal area (%HAV). OPD is defined as the number of intact and fragmentary secondary osteons per mm 2 of bone. Fragmentary osteons are remnants of older complete osteons and are indicative of iterative remodeling within a part icular section of bone. Some fragmentary osteons have lost their Haversian canals, but can be identified by the presence of a cement line. On.Ar is total area of bone contained within the cement line of individual intact osteons, reported in mm 2 %HAV is the total area of secondary bone (the area of all intact and fragmentary osteons combined) relative to the total area of the thin section, reported as a percentage. These variables were measured from the thin section photographs in ImageJ (Abramoff et al., 2004)
51 Shapiro Wilk normality tests (Shapiro and Wilk, 1965) were performed to assess whether data were normally distributed. The data for all three variables in the femur violated as sumptions of normality ( Table 4 3 ). All data from the humerus were normally distributed. Data that violated assumptions of normality were z rank transformed using (Aulchenko et al., 2007) prior to any other statistical analysis. Since leaping frequency (i.e., high or low) classification coincides with phylogenetic affiliation (i.e., subfamily), statistical methods that account for phylogeny were employed in an attempt to avoid conflating leaping frequency and phylogenetic effects as agents responsible for any observ ed differences between groups. A generalized linear mixed model (glmm) in which the fixed effect was leaping frequency (high or low) and the random effect was phylogenetic distance was fitted to the data for each variable using Markov Chain Monte Carlo (MC MC) methods. The MCMC was performed with the R (Hadfield, 2010) using a phylogenetic tree downloaded from 10kTrees (Arnold et al., 2010) pruned to include only the species in the sample (i.e., Piliocolobus badius, Colobus polykomos, Cer copithecus diana, Cercocebus atys ). 250,000 iterations of an MCMC were run with a burn in of 3,000 and a thinning rate of 50 to obtain a posterior distribution including 4941 models. Each iteration in the Markov chain can be thought of as a unique model, w hose parameters are tweaked using information about model fit derived from the immediately preceding models. Each model explains a mathematical relationship between the projected dependent variable (e.g., OPD) and the independent variables (i.e. leaping fr equency and phylogenetic position). The model is used to predict dependent variables for each individual based on phylogenetic position and leaping frequency. The model derived and empirically
52 measured dependent variables are then compared and models that minimize differences between them are assigned a high likelihood score. The MCMC maximizes the fit between the predicted and empirical measures, so the models with high likelihood scores are favored. The burn in refers to the first portion of the Markov ch ain during which the model parameters are adjusted without much information on previous model model assumptions). To guard against these initial assumptions conditionin g the results, models generated during the burn in are discarded and not included in hypothesis testing. The thinning rate is the number of steps in the Markov chain between sampled models (i.e., thinning at a rate of 50 iterations is to sample every 50 th model). A properly large thinning parameter assures that each sampled model is independent from the previously sampled model. Thus, models in the posterior distribution will be independent of each other, and will not suffer from autocorrelation derived sam pling bias. Heidelberger Welch diagnostics (Heidelberger and Welch, 1983) were performed on the posterior dist ributions of the MCMCs ( Table 4 4 ) and demonstrate that the sampled models came from stationary distributions, meaning that the Markov chain settl ed on a narrow set of models with similar parameters in terms of both the number and the coefficients of those parameters. The number of iterations and burn in were extended on an individual model basis until the model passed the Heidelberger Welch test. M CMC trace and density plots are displayed in Figures 3 2 through 3 7. Two additional analyses were performed post hoc First, paired t tests compared OPD and %HAV in the femur and humerus. Second, a nested Generalized Linear Model (GLM) tested for correla tions in OPD and %HAV between the femur, humerus,
53 and mandible (data from [Lad et al., 2016]) of the same individuals in order to test whether the rate of remodeling is systemic and unique to any individual or species, as opposed to depending on mechanical factors. Results The glmm MCMCs for femoral OPD ( p = 0.049) and femoral %HAV ( p = 0.042) showed significant results ( Figure 4 8 ). Thus, there is more secondary bone in femora of high frequency leapers than in low frequency leapers, even when controlling for phylogenetic affiliation. In the high frequency leapers, mean OPD and %HAV are about twice the mean OPD and %HAV in low frequency leapers. Generally speaking, %HAV is a function of OPD and On.Ar. Here, greater osteon densities, rather than larger indiv idual osteons, drive differences in relative osteonal area in the high frequency leapers. By species, Colobus polykomos has the greatest OPD and %HAV, followed by Piliocolobus badius then Cercocebus atys and Cercopithecus diana ( Table 4 5 ). None of the MC MCs returned significant results for the humerus, meaning any differences in the amount of secondary bone between the two groups are spurious ( Figure 4 9 ). Data are summarized by species in Table 4 5 The paired t tests demonstrated that OPD ( P = 0.009) a nd %HAV ( P = 0.002) are significantly greater in the humerus than femur Results of the nested GLM were significant for OPD ( P < 0.001) and %HAV ( P = 0.009), indicating differences between the femur, humerus, and mandible within the same individuals ( Figure 4 10 ). Discussion What follows is an evaluation of the results organized around the proposed hypotheses and alternative explanations, followed by a discussion of the remodeling measures employed and a summary of the ideas explored.
54 Null Hypothesis : Mineral Homeostasis The null hypothesis stated that mineral homeostasis, rather than biomechanical factors, is the primary cause of bone remodeling in the limbs of the species examined here. This hypothesis seems unlikely given the variation in OPD and %HAV among species in the femur, and between the humerus and femur in all species. Remodeling that is related to the balance of mineral concentrations is not generally thought to be localized in the manner that strain related remodeling is, but rather to b e more or less stochastically encountered throughout the body (Burr, 2002) There is no reason to assume that this type of remodeling would preferentially occur in the femur of colobines, or in the humerus rather than the femur of all species. Load Magnitu de In the femur, OPD and %HAV are correlated with greater leaping frequency even when using methods that account for phylogenetic affiliation. On.Ar is not significantly different between taxa and is discussed in further detail below. The results for OPD and %HAV seem to suggest that leaping, and thus high magnitude forces on the hindlimbs, incites remodeling in the femur and is consistent with the working hypothesis that load magnitude mediates bone remodeling. The results for the humerus do not follow t his narrative. If the forelimbs of the colobines are often involved in braking during landing, then the colobines are expected to also have greater humeral remodeling than the cercopithecines; however, there is no relationship between secondary bone in the humerus and load magnitude. This finding is similar to that of Terranova (1995) who determined that among strepsirhines, cross sectional parameters of the femur but not the humerus were correlated with leaping frequencies. The fact that the humerus do es not match the pattern observed for the femur may reflect varied
55 use of the forelimb for leaping among the species examined here. Data on the braking forces in the forelimbs of the species examined here would clarify whether the pattern of remodeling in the humerus matches locomotor behaviors as related to load magnitude. Another factor to consider is that colobines tend to be clumsy and imprecise in landing, and hindlimbs (McGraw, 1998a) Imprec ision in landing in any of these taxa may result in different or more unpredictable loading patterns in the humerus compared to the femur, which may result in different degrees of microdamage. This could explain the different patterns of remodeling in the two bones. Studies on remodeling as it relates to strain mode in these species are forthcoming. Load Frequency An alternative explanation was that load frequency determines the degree of bone remodeling in limb bones. This hypothesis is not supported by locomotor differences between taxa because Cercopithecus diana should have more bone remodeling than all other taxa, due to being more active overall, yet C. diana has relatively low OPD and %HAV. However, the notion that bone remodeling depends on the fre quency of loading cannot be completely ruled out. The humerus has significantly greater OPD and %HAV than does the femur ( Table 4 5 Figure 4 10 ). The finding that remodeling is greater in the humerus than the femur is consistent with previous reports on q uadrupedal Old World monkeys (Paine and Godfrey, 1997; McFarlin et al., 2008) In quadrupedal primates, the hindlimbs experience relatively higher peak loads than the forelimbs, whereas the opposite is true in quadrupedal non primate mammals (Kimura et al. 1979; Kimura, 1985; Demes et al., 1994; Schmitt, 1999; Schmitt and Lemelin, 2002; Schmitt, 2003; Hanna et al., 2006) Additionally, takeoff forces, for which the
56 hindlimbs are predominantly responsible, are greater than landing forces in leaping strepsir hines (Demes et al., 1999) The finding that the humerus has more secondary bone than the femur of not only the Ta monkeys but also other primate quadrupeds, despite quadrupedal primates being hindlimb driven, is counter to what would be expected if load magnitude controls remodeling, but may be compatible with the idea that load frequency is responsible. In primates, the role of the hindlimb is limited to mostly locomotion, but the forelimb is involved in not only locomotion but also many other activities (e.g., foraging, food processing, grooming, play). Because of this, the humerus may be loaded more frequently than the femur in all species, even though load magnitude is presumably higher in the femur. The more frequent loads in the humerus in all specie s could swamp the load magnitude signal, explaining the disparity between the two limbs (humerus having greater remodeling) and the lack of taxonomic differences in the humerus. Studies on forelimb movement in the Ta monkeys show that differences in non l ocomotor activities between species correspond to differences in scapular and humeral morphology (e.g., Dunham et al., 2015, 2016) Therefore, it is not unreasonable to expect that these non locomotor activities may be manifesting in the bone microstructur e. Further, a comparison of femoral and humeral bone remodeling in forelimb driven quadrupedal mammals would help determine whether hindlimb and forelimb drive are important factors for bone remodeling. Phylogenetic Effects A second alternative explanatio n was that unspecified phylogenetic effects account for differences in remodeling, predicting that differences will fall along subfamily lines. Although the methods used here account for phylogenetic affiliation, they do not completely eliminate the possib ility that factors other than leaping frequency play some
57 role in our observed results. One such scenario could be that the femoral bone of cercopithecines is stiffer than that of colobines. Stiff bone has higher yield strength than more compliant bone, re quiring higher loads to structurally deform it and to initiate cracks. Compliant bone is more inclined to deform and crack but it may have greater toughness because it reduces crack propagation ( see pages 356, 446, and Figure 7 .12 in Martin et al., 2015) Stiffer bone could thus result in a reduced need for remodeling if fewer cracks form. Elastic modulus data from the femoral midshaft would help determine whether this hypothesis has any bearing Data have been collected from one individual each of Piliocol obus badius and Colobus polykomos ( Galluscio et al., 2016 ) but not from either of the cercopithecines Another scenario could be that these two subfamilies have different systemic rates of remodeling, unrelated to behavioral activity. In relation to this idea, the results here mirror those of Lad et al. (2016) who reported greater remodeling in the postc anine mandible of colobines than cercopithecines in a sample comprised of individuals from the same populations measured here. The combined results raise the possibility that there is a systemic rate of remodeling within an individual or within a species ( i.e., if remodeling is high in the femur of one individual, it may be that the remodeling rate is high in all bones throughout the body of that individual). If this were the case, then the observed results could be due to higher systemic rates of remodelin g in colobines dependent on some aspect of colobine this sample are more folivorous than the cercopithecines (Wachter et al., 1997; Korstjens et al., 2002; Korstjens et a l., 2007; McGraw et al., 2011; Kane and McGraw, 2017) and dietary composition may alter mineral homeostasis (see reviews by Setchell
58 and Lydeking Olsen, 2003; Mundy, 2006) and basal metabolic rate (McKnab, 1986) However, osteon population densities for th e mandible, humerus, and femur within each individual included in both the present sample and in Lad et al. (2016) (two Cercocebus atys three Cercopithecus diana and three Colobus polykomos ) suggest this is not the case ( Figure 4 10 ). It appears that rem odeling in any one of these bones is not necessarily correlated with remodeling in another bone from the same individual. In other words, the individual with the greatest OPD in the mandible does not necessarily have the highest OPD in the humerus and/or f emur. This was confirmed by the nested GLM. If there were a systemic rate of remodeling, individuals with high osteon densities in one bone would have correspondingly high densities in others. Furthermore, while there are taxonomic differences for the femu r, the humerus does not match this pattern. Since there is no evidence that the rate of remodeling is consistent throughout the skeleton of the individuals used here, it seems unlikely that phylogenetic effects, outside of behavior, play a significant role in the observed results. General Remarks Analysis of bone remodeling in the present sample does not provide support for the null hypothesis of non targeted remodeling or the alternative explanation of phylogenetic effects, and there is conflicting suppo rt for hypotheses regarding load magnitude and frequency. Bone remodeling must depend on some interplay among load magnitude, load frequency, and the need to maintain mineral homeostasis, but these factors are not easy to parse. For example, both a single loading event of high magnitude, and frequent loading of very low magnitude may be insufficient to incite a remodeling response. Obviously in this particular sample, isolating the effects of load magnitude and load frequency is complicated. Future work nee ds to test the effects of
59 load magnitude and load frequency on bone remodeling independently of each other. One way to test the load frequency hypothesis could be a comparison of remodeling in the bones of captive and wild conspecifics under the assumption that captive primates have much lower activity levels than their wild counterparts because they do not have to forage for food (Altmann and Muruthi, 1988) Load magnitude would presumably be similar if body size and locomotor behaviors are comparable. The load magnitude hypothesis could be tested by comparing remodeling in the tibia and fibula. These two bones should have roughly the same loading frequency, but load magnitudes should be higher in the tibia since it is weight bearing, whereas the fibula is not. Such comparisons may provide some insight on the relative contributions of load magnitude and frequency to remodeling activity. Osteon cross sectional area (On.Ar) is a measure of individual osteon size. On.Ar was not significantly different between colobines and cercopithecines for either bone. Therefore, it does not seem to be related to load frequency or magnitude. Instead, the size of individual osteons may depend more on the mode of strain (i.e., tension versus compression). If the limbs of the s pecies examined here experience roughly similar spatial distributions of tension and compression, then the design of this study was unable to compare bone remodeling in regions of different strain mode. There are several hypotheses about why different size d osteons would be ideal for different strain modes.  A high density of smaller osteons may more effectively resist damage accumulation than a smaller number of large osteons because there are more cement lines to act as barriers to crack propagation (G ibson et al., 2006; Mohsin et al., 2006)  Osteon pullout, a phenomenon in which an individual osteon becomes
60 increasing the amount of energy required to propagate a c rack (Piekarski, 1970) but it only occurs in regions of tension, not compression (Moyle et al., 1978; Hiller et al., 2003) Large osteons are more likely to debond from the surrounding matrix and pull out prior to failure than small osteons are (Pope and M urphy, 1974) but densely packed smaller osteons increase the number of osteons to be pulled out (Moyle et al., 1978; Martin et al., 2015)  The formation of smaller osteons may just be less likely to affect the overall strength of the bone in regions wi th the highest strain (i.e., compression cortices) because they require smaller resorption cavities (van Oers et al., 2008) Studies of osteon size in artiodactyl and perissodactyl bone have found that osteons tend to be smaller in regions of compression, rather than tension (Skedros et al., 1994; Skedros et al., 1997; van Oers et al., 2008) The limbs of the primate species examined here are presumably loaded in bending, and thus may also have consistent compression and tension cortices, however the load h istory of a particular bone is usually more complex than pure bending (Skedros, 2012) Future work will address strain mode and compare osteon size and density among regions of the limb bone diaphysis in these species. Finally, some limitations of this st udy should be acknowledged First, due to the nature of conducting destructive analyses the sample size per species is small. Second, only adults were used in this study but since the skeletal materials were opportunistically collected from the forest floo r, age at death is uncertain. At the time of epiphyseal fusion (the feature used to define skeletal maturity) the diaphysis contains primary bone tissue of differing ages due to the process of modeling drift (Enlow, 1962;
61 Epker and Frost, 1965; Frost, 2001 ) The younger tissue may have very few to no secondary osteons (Maggiano et al., 2011) and consequently there can be a delay before remodeling actually reflects habitual loading (Gocha and Agnew, 2016) If the individuals in this sample had only just rea ched skeletal maturity at the time of death then the observed remodeling may not completely reflect behavioral differences. Additionally, secondary osteons are known to accumulate with age in humans (Robling and Stout, 2000) so significantly older individ uals may have greater OPD and %HAV than young adults. If any of the individuals in this sample are very old, high OPD and %HAV could be interpreted as an effect of behavior, when in reality they are due to advanced age. Finally, Skedros et al. (2013) demon strate that collagen fiber orientation (CFO) and osteon morphotypes may be much more sensitive to load history than the measures used here. It is possible that OPD and %HAV do not provide high enough resolution to pick up on differences in load history, in terms of magnitude and frequency, between the examined taxa. Conclusion The amount of secondary bone in the femoral midshaft varies among Ta Forest cercopithecids. Colobines have greater osteon population densities and relative osteonal area than cerco pithecines. This variation corresponds to frequencies of high magnitude forces applied to the femur as a result of species specific leaping behaviors, even though all species can be characterized as primarily quadrupedal. However, the lack of taxonomic var iation in remodeling of the humerus suggests this interpretation should be accepted with caution. Further, the greater amount of secondary bone in the humerus relative to the femur suggests that the frequency of loading is also important, as the forelimb i s presumably more active than the hindlimb, due to its recruitment into a
62 greater number of behaviors aside from locomotion. Future work is needed to parse out the independent effects of load magnitude and load frequency. The lack of differences in osteon cross sectional area suggest that this variable is unrelated to load frequency or magnitude, but we propose, as others have suggested (Moyle et al., 1978; Skedros et al., 1994; Skedros et al., 1997; Hiller et al., 2003; Gibson et al., 2006; Mohsin et al., 2006; van Oers et al., 2008) that osteon size may be associated with specific modes of strain applied to different regions of limb bone midshafts.
63 Table 4 1. Species locomotor profiles 1 Subfamily Species Leaping Climbing Running 2 Walking Colobinae Piliocolobus badius 17.8 17.0 8.2 53.1 Colobinae Colobus polykomos 14.5 14.3 29.4 41.8 Cercopithecinae Cercopithecus diana 10.4 19.4 10.8 59.3 Cercopithecinae Cercocebus atys 1.2 12.5 5.7 80.7 1 Data are percentages of overall locomotion sourced from Table 9.4 of McGraw (2007). 2 Running includes bounding, the synchronized takeoff of both hindlimbs, and then touchdown of both forelimbs, upon a substrate. Bounding is thought to require higher propulsive forces than running, more similar to l eaping. In this sample, bounding is unique to C. polykomos Table 4 2. Activity budgets 1 of Ta monkey species (from McGraw, 2007) Species Traveling Foraging Resting Social Feeding N Piliocolobus badius 18.9 15.8 29.9 6.3 29.1 4196 Colobus polykomos 15.1 10.8 33.9 5.3 34.9 3538 Cercopithecus diana 28.5 28.3 8.8 1.2 33.2 3461 Cercocebus atys 10.3 24.5 18.5 7.9 38.8 1343 1 Data in each column are percentages of total time spent doing each activity. Traveling is uninterrupted movement, foraging is movement while feeding, and feeding, resting and social are all stationary activities. C. diana has the highest percentages of traveling and foraging, while the other species have high percentages of stationary activities. Table 4 3 Results of Shapiro Wilk normality tests for each variable for the femur and humerus. O P D = osteon density, O n. A r = osteon cross sectional area, %HAV = relative osteonal area. *Indicates data that violate assumptions or normality and were z rank transformed for the MCMC. Variable p value Femur OPD 0.007* O n. A r 0.001* %HAV 0.003* Humerus OPD 0.814 O n. A r 0.385 &HAV 0.543
64 Table 4 4 Heidelberger Welch results for tests of stationarity and interval halfwidth of the MCMCs for each variable per bone. O P D = osteon density, O n. A r = osteon cross sectional area, %HAV = relative osteonal area. Variable Stationary start Interval halfwidth Test p value Test p value Femu r OPD passed 0.335 passed 0.014 O n. A r passed 0.439 passed 0.016 %HAV passed 0.867 passed 0.027 Humerus OD passed 0.540 passed 0.024 O n. A r passed 0.307 passed <0.001 %HAV passed 0.162 passed 0.001 Table 4 5 Summary of the data for each variable by species for the femur and humerus Species OPD On.Ar (in mm 2 ) %HAV Mean S.D. Range Mean S.D. Range Mean S.D. Range Femur Piliocolobus badius 5.293 5.034 1.370 11.142 0.019 0.012 0.003 0.091 9.743 8.974 2.899 19.905 Colobus polykomos 5.587 2.279 1.927 8.107 0.018 0.011 0.003 0.081 10.211 4.528 3.729 16.258 Cercopithecus diana 2.289 3.615 1.114 9.357 0.020 0.013 0.004 0.108 5.481 5.680 2.460 15.620 Cercocebus atys 2.818 2.786 0.527 7.634 0.018 0.012 0.003 0.100 5.057 4.872 1.438 13.590 Humerus Piliocolobus badius 5.838 1.263 4.171 7.501 0.023 0.016 0.002 0.129 13.411 4.621 8.267 19.817 Colobus polykomos 8.147 3.205 4.496 11.724 0.017 0.009 0.002 0.100 13.104 4.026 7.964 17.809 Cercopithecus diana 7.145 2.091 4.914 10.149 0.022 0.013 0.003 0.114 15.155 3.589 10.017 19.911 Cercocebus atys 6.328 2.042 3.346 8.437 0.023 0.013 0.004 0.138 14.178 3.718 8.715 17.133
65 Figure 4 1. A Piliocolobus badius femoral midshaft thin section at 100x magnification, highlightin g some secondary osteons ( shaded blue) and osteon fragments (shaded red).
66 Figure 4 2 Trace (at left) and density plots (at right) for the model parameters of the femoral osteon population density (OPD) MCMC posterior distribution. among the posterior models. Density plots show avoidance of severe skewness, or contacting model space boundary.
67 Figure 4 3 Trace (at left) and density plots (at right) for the model parameters of the femoral osteon cross sectional area (O n. A r ) MCMC posterior distribution. es lack of autocorrelation among the posterior models. Density plots show avoidance of severe skewness, or contacting model space boundary.
68 Figure 4 4 Trace (at left) and density plots (at right) for the model parameters of the fem oral relative osteonal area (%HAV appearance of the trace plots demonstrates lack of autocorrelation among the posterior models. Density plots show avoidance of severe skewness, or contacting model space boundary.
69 Figure 4 5 Trace (at left) and density plots (at right) for the model parameters of the humeral osteon population density (OPD) MCMC posterior distribution. among the posterior models. D ensity plots show avoidance of severe skewness, or contacting model space boundary.
70 Figure 4 6 Trace (at left) and density plots (at right) for the model parameters of the humeral osteon cross sectional area (O n. A r ) MCMC posterior distribution. among the posterior models. Density plots show avoidance of severe skewness, or contacting model space boundary.
71 Figure 4 7 Trace (at left) and density plots (at right) for the model parameters of the hum eral relative osteonal area (%HAV appearance of the trace plots demonstrates lack of autocorrelation among the posterior models. Density plots show avoidance of severe skewness, or co ntacting model space boundary.
72 Figure 4 8 Osteon population density (OPD), osteon cross sectional area (On.Ar), and relative osteonal area (%HAV) in the femur of low ( Cercocebus atys and Cercopithecus diana ) and high frequency ( Colobus polykomos and Piliocolobus badius ) leapers, with p values from the MCMC for each of the four variables. All data are z rank transformed. The bottom and top of the boxes indicate the interquartile range (IQR), the horizontal line represents the median, and the whisk ers represent the minimum and maximum data points. *Indicates significant result.
73 Figure 4 9 Osteon population density (OPD), osteon cross sectional area in (On.Ar), and relative osteonal area (%HAV) in the humerus of low ( Cercocebus atys and Cercopithecus diana ) and high frequency ( Colobus polykomos and Piliocolobus badius ) leapers, with p values from the MCMC for each of the four variables. The bottom and top of the boxes indicate the interquartile range (IQR) and the horizontal lin e represents the median. The whiskers represent the minimum and maximum data points, and the open circles represent outliers. Note that the y axes do not begin at 0.
74 Figure 4 10 Osteon population density (OPD) in the femur, humerus, and mandible within the same individuals. Comparison of the three bones reveals little to no intra individual correlation of OPD among the sampled bones. Mandible data are from Lad et al. (2016). No P. badius individuals were used in both studies.
75 CHAPTER 5 SPATIAL DISTRIBUTION AND SIZE OF SECONDARY OSTEONS IN FEMORAL AND HUMERAL MIDSHAFTS OF CERCOPITHECID PRIMATES Intracortical bone remodeling, or Haversian remodeling, is the resorption and subsequent replacement of older bone with new bone. This serves three functions (Martin et al., 2015) First, it helps maintain mineral homeostasis by redistributing calcium rese rves (Enlow, 1962a) Second, it replaces damaged bone by targeting areas of microdamage caused by mechanical deformation (Burr et al., 1985; Burr, 1993; Mori and Burr, 1993; Burr, 2002; Martin et al., 2015) Third, it increases bone toughness via its mater ial properties and its microstructural composition. Remodeled bone is more compliant than primary bone since it is less mineralized (Currey, 1959; Reilly and Burstein, 1974; Reilly et al., 1974; Carter et al., 1976; Carter and Hayes, 1977b) and greater ene rgy is required to propagate cracks through a compliant material than a stiff one (Martin et al., 2015) Additionally, the accumulation of secondary osteons, which are formed during remodeling, can prevent crack propagation (Moyle and Bowden, 1984; Gibson et al., 2006; Mohsin et al., 2006; Martin et al., 2015) Secondary osteons are cylindrical structures comprised of concentric lamellae surrounding a central (Haversian) canal, which houses a neurovascular bundle. Each secondary osteon can be distinguishe d from the surrounding matrix in cross section by the presence of the cement line on its outer perimeter. The cement line is a remnant of the reversal from the bone resorption phase of remodeling to the bone formation phase. The cement line is also used to distinguish secondary osteons from primary osteons, the formation of which does not follow a resorption phase (Crowder and Stout, 2012) Strain mode and magnitude can differ between regions of the same bone (Lanyon et al., 1979; Biewener et al., 1986) L imb bones are often modeled as
76 cylindrical beams loaded in a combination of axial compression, due to gravitational and ground reaction forces, and simple bending, which results in compressive strain in one aspect of the cortex and tensile strain in the op posing aspect (Lanyon and Baggott, 1976; Biewener and Bertram, 1993; Lieberman et al., 2003; Skedros et al., 2013) The mechanical properties, fracture properties, and microdamage effects of cortical bone are known to differ under tension versus compressio n (Burstein et al., 1972; Reilly and Burstein, 1975; Carter and Hayes, 1977a; Carter et al., 1981; Burr et al., 1998; Reilly and Currey, 1999, 2000; Hiller et al., 2003) For example, microdamage accumulates more rapidly in regions of tension than compress ion (Burr et al., 1998) because bone under tension has a lower yield strength (Burstein et al., 1976; Cezayirlioglu et al., 1985) However, cracks tend to be longer in regions under persistent compression (Burr et al., 1998) Since removing and preventing microdamage are functions of bone remodeling, secondary bone characteristics, such as osteon density and size, can be predicted to also differ under disparate strain modes. Main (2007) and Skedros (2012) warn that osteon density alone may not reliably predict strain mode, especially in bones that have complex load histories, but there is some evidence that the spatial distribution of osteons, in terms of both density and size, corresponds to distributions of tension and co mpression in stereotypically bent limb bones. In the calcaneus, smaller osteon areas and greater osteon densities have been found in the compression (cranial) cortex than in the tension (caudal) cortex of sheep and deer (Skedros et al., 1994) horse, elk, and sheep (Skedros et al., 1997) Smaller osteon areas were also found in the compression region of mule deer calcaneus (Skedros et al., 2001) Mason et al. (1995) found greater osteon density in the
77 compression (caudal) region than in the tension (cranial ) region of the horse radius, but did not find differences in mean osteon area. The relationship between strain and osteon density is theoretically straightforward if osteons form as a response to osteocyte apoptosis caused by crack formation (Allen and Burr, 2014) but the relationship between strain and osteon size may be less direct. Moyle and Bowden (1984) found that work to fracture is minimal for 0 the overall strength of bone, and both small and large osteons have even been proposed to be adaptations for preventing microdamage, although the mechanism by which osteon size is determined is un clear. The cement line can prevent the spreading of cracks through bone, so a large number of small osteons concentrated in a given area can increase the number of barriers to crack propagation (O'Brien et al., 2005; Gibson et al., 2006; Mohsin et al., 200 6) Thus, densely packed, small osteons may reduce the amount of damage a crack can cause and increase the energy required to fracture the bone in areas where microdamage tends to accumulate. However, larger osteons may also increase bone toughness through pull the amount of energy required to propagate it (Piekar ski, 1970) Osteon pull out only occurs under tension, not compression (Moyle et al., 1978; Hiller et al., 2003) and there is some evidence that large osteons are more likely to de bond from the surrounding matrix and pull out prior to failure than small o steons are (Pope and Murphy, 1974)
78 Together, these two lines of evidence suggest that larger osteons should be found in regions of tension. On the other hand, smaller osteons can be more densely packed and could increase the number of osteons available to be pulled out (Moyle et al., 1978; Martin et al., 2015) Another factor to consider is that the overall strength of the bone, especially in regions of high strain (i.e., regions under compression in bones under combined axial and bending loads), may be co mpromised during the formation of large osteons, which require large resorption cavities. van Oers et al. (2008) have even argued that osteocytes may be able to sense strain magnitude and inhibit resorption resulting in smaller osteon diameters when st rains are high, particularly in areas of compression. The present study investigates whether secondary osteon density and size vary among regions of primate long bone midshafts under the assumption that these bones are loaded in habitual, stereotypical be nding. Osteon density, relative osteonal area, and osteon cross sectional area are measured in cercopithecid femoral and humeral midshafts to determine whether these characteristics vary among midshaft regions, and if they do, whether the patterns correspo nd to hypotheses of load distribution. In quadrupedal mammals, long bones are known to be habitually loaded in bending (Carter and Vasu, 1981; Biewener, 1983; Szivek et al., 1992; Demes et al., 1998; Demes et al., 2001; Demes and Carlson, 2009) Theoretica l analyses using free body diagrams predict mediolateral and anteroposterior bending in the chimpanzee (Preuschoft, 1970) and macaque (Badoux, 1974) femur, but there have been no in vivo analyses of the femur or humerus in primates to inform these theoreti cal models. Thus,
79 stresses in the quadrupedal macaque femur, which predicts compression in the posterior cortex and tension in the anterior cortex. Based on results of pre vious studies (Skedros et al., 1994; Mason et al., 1995; Skedros et al., 1996; Skedros et al., 2001) the compression cortex (posterior) of the femur is expected to have greater osteon density and smaller osteons than the tensile (anterior) cortex. Predict ions for the humerus are less clear but it should be safe to assume the humerus is also loaded in bending, as long bones typically are (Martin et al., 2015) Methods Histological thin sections were prepared from one femur and one humerus each from five in Cercocebus atys (sooty mangabey), Cercopithecus diana (Diana monkey), Colobus polykomos (western black and white Colobus), and Piliocolobus badius (western red Colobus). All speci es are quadrupedal and all are predominantly arboreal, with the exception of Cercocebus atys which spends most of its time on the ground. Sex and age at death are uncertain since all skeletal material was collected opportunistically in the field. Bone was only sampled from skeletally mature individuals, as determined by epiphyseal fusion. Thin sections were prepared from femoral and humeral diaphyses at 50% of the length of the whole bone. Blocks were cut from this region and embedded in resin following Cr owder et al. (2012) Thin sections were cut from the embedded bone wafering blade (Illinois Tool Works, Lake Bluff, IL). Each section was polished using a Buehler Meta Serv 250 gri nder polisher and stained with a toluidine blue O solution to improve visibility of osteon cement lines. The sections were dried under weight to keep them flat and then mounted on microscopy slides. Four equal quadrants (anterior,
80 posterior, medial, and la teral) were marked on the slides (Figure 5 1) and the sections were photographed under a compound light microscope at 100x magnification. The thin section photographs were imported into ImageJ software (Abramoff et al., 2004) which was used to measure ost eon population density (OPD), osteon cross sectional area (On.Ar), and relative osteonal area (%HAV) within each quadrant of each thin section. The quadrants were predicted to have different OPD, On.Ar, and %HAV if these bones are loaded in bending at th e midshaft, with the posterior (compression) quadrant containing the greatest OPD and %HAV, and smallest On.Ar. OPD was measured as the number of intact and fragmentary secondary osteons per mm 2 of cortical bone. Fragmentary osteons are the remnants of osteons that have been partially resorbed and replaced by new osteons, and can be identified by the presence of a cement line even if the Haversian canal has been resorbed. On.Ar data were collecte d by tracing the cement line of each intact osteon and measuring the area within. Osteon fragment areas were also measured and the combined area of intact and fragmentary osteons was divided by the total quadrant area and multiplied by 100 to calculate %HA V. The photographs were carefully aligned using unique sets of identifiable landmarks so that no area was measured more than once. Shapiro Wilk tests (Shapiro and Wilk, 1965) (Table 5 1) showed that no femoral data, with the exception of OPD in Colobus w ere normally distributed. In the humerus, %HAV was non normal for all species. All data that violated assumptions of normality were z (Aulchenko et al., 2007) prior to any other statistical analysis. For On.A r, nested ANOVAs were performed for the
81 in order to test for differences among quadrants both within individuals and also among individuals of the same species. Resamp led Analyses of Variance (ANOVAs) were performed to test for differences in OPD and %HAV between midshaft quadrants within the femur and humerus for each species separately. The resampled ANOVAs were implemented by first calculating the mean OPD or %HAV of each quadrant for each individual so that for each species there were 20 means, five for each of the four quadrants. A one way ANOVA was performed using these means as the dependent variable and quadrant as the grouping variable. The F ratio was recorded. The means were then resampled without replacement to create a new sample equivalent in size to the observed sample. The first five values of this new sample were assigned to an anterior quadrant group, then next five to the posterior group, and so on such that the resulting sample simulated the empirical one, but with the regional assignment randomized. A one way ANOVA was then performed on this new sample and the F ratio was recorded. This resampling process was repeated for 10,000 iterations and the prop ortion of F ratios greater than the observed F ratio was reported. If that proportion was 5% or less then the result was considered statistically significant, as the F ratio for the empirical means was not likely due to chance. Results All data are summ arized in Tables 5 2 and 5 3. The resampled ANOVAs returned no significant differences in OPD or %HAV between quadrants of the femur or humerus of any species except Cercocebus atys in which OPD ( P = 0.001) and %HAV ( P = 0.002) were significant for the hu merus (Table 5 4). In this species, the lateral quadrant of the humerus had the greatest mean OPD and %HAV and the medial h ad the least (Figure 5 2 and 5 3).
82 for the Cercoce bus atys humerus (Table 5 5). This means that On.Ar differs among quadrants within individuals, but also that individuals within a species differ from each other. The quadrants with the greatest and least mean On.Ar are not consistent among the four specie s (Tables 5 2 and 5 3; Figure 5 4 and 5 5) or even among individuals within a species. For example, in the humerus, mean On.Ar is greatest in the lateral quadrant and least in the anterior quadrant of Piliocolobus badius but in Colobus polykomos mean On.A r is lowest in the posterior and lateral quadrants and greatest in the medial and anterior quadrants. Similar incongruities occur between individuals of the same species. This indicates that there is no consistent pattern in the distribution of small and l arge osteons among midshaft quadrants in either the femur or humerus. Discussion Load Complexity Osteon density and relative osteonal area are fairly homogenous (i.e., across quadrants) throughout the femoral and humeral midshafts of the quadrupedal cerco pithecids examined here. The only exception to this is the humerus of Cercocebus atys for which OPD and %HAV are significantly different among regions, with the lateral quadrant having the most secondary bone and the medial quadrant the least. Th e lack of differences among quadrants is in contrast to the results of previous studies demonstrating heterogeneous distributions of osteons in the limb bones of other quadrupedal mammals (Skedros et al., 1994; Mason et al., 1995; Skedros et al., 1997; Skedros et a l., 2001) However these other studies all examined long bones of digitigrade or unguligrade taxa that specialize in high speed terrestrial locomotion (e.g.,
83 horse, greyhound), unlike primates. This difference in locomotion type might result in differences in load complexity in the limb bones. Skedros (2012) whereas others have low load complexity. High load complexity is characterized by less stereotypical or less predictable loadin g. This usually entails a combinati on of multi directional bending in which the neutr al axis frequently shifts, unsymmetric bending, and/or torsion. Low load complexity equates to unidirectional bending with distinct tension and compression regions separat ed by a predictable, non shifting neutral axis. Regional variation in osteon density and size was expected in the present sample under the assumption that the femur and humerus have distinct, habitual tension and compression cortices. In other words, that they have low load complexity. Given the lack of variation in remodeling, however, this may not be the case. If the femur and humerus have high load complexity, in terms of overlap of tension and compression regions, then regionally specific manifestations of bone microstructure would not occur. Furthermore, sectional spatial differences in strain mode (Skedros, 2012) There is evidence that bones with high load complexity lack regional adaptation in terms of collagen fiber orientation (Skedros and Hunt, 2004) osteon morphotypes (Skedros et al., 2009) and secondary osteon density (Main, 2007) Thus, one way to interpret the pre sent results is that load complexity is high in femoral and humeral midshafts; the load histories are not characterized by habitual and regionally specific compression and tension cortices as hypothesized above. This means that there is no cause for differ ential remodeling regionally because there is
8 4 presumably uniform microdamage accumulation throughout the cortex. The femoral and humeral midshafts of these species possibly fall into the high load complexity category rt, whereas the limb bones of the digitigrade taxa examined in previous studies have been placed in the low load complexity categories. The Cercocebus atys humerus was an exception in this sample in that there are regional differences in OPD and %HAV. If load complexity explains the lack of regional variation in all other taxa, then the loading environment of the Cercocebus atys humerus may be less complex and more stereotyped, causing the humeral midshaft of this species to show regional variation in bone remodeling. There are several reasons why load complexity may be lower in Cercocebus atys  All taxa in this sample are quadrupedal but Cercocebus atys is unique in that it is primarily terrestrial, spending 85% of travel and 71% of foraging time on the ground, whereas the next most terrestrial species in the sample, Cercopithecus diana spends 1.7% of travel and 2.3% of foraging on the ground (McGraw, 2007) Demes and Carlson (2009) demonstrated that the capuchin forelimb has greater variability of r esultant directions and bending moments in the frontal plane in a simulated arboreal environment than in a terrestrial one, and suggested that arboreal locomotion is associated with greater variation in loading of long bones because arboreal substrates are less stable than the ground. Substrate differences may thus equate to differences in load complexity in the limb bones.  In addition to moving about on a less stable substrate the predominantly arboreal taxa have more variable locomotor repertoires th an Cercocebus atys They all
85 engage in more climbing, running, and leaping than Cercocebus atys for which 80.7% of the locomotor budget is walking (McGraw, 1998, 2007) Employing a variety of locomotor behaviors presumably results in more complex loading of the humerus of the other species.  Cercocebus atys engages in unique terrestrial, manual foraging behaviors involving the forelimb. This species sifts through the leaf litter on the forest floor in search of fallen fruits and seeds, requiring frequ ent inferior food retrieval and raking motions. O f all food retrieval movements 73.9% are inferior retrieval, compared to 15.6% parallel retrieval and 10.5% superior retrieval (Dunham et al., 2015, 2016) The other three species spend larger percentages o n parallel and superior retrieval, and none have been observed to rake, which constitutes 69% of all forelimb behaviors for Cercocebus atys Raking is a motion in which the forelimb repeatedly alternates between flexion and extension while the forearm supi nates and pronates (Dunham et al., 2015) As a result, Cercocebus atys engages in minimal overhead movements and the movements of the forearm during foraging are not a far departure from those during locomotion. In other words, the movements are repetitive and consistent. Cercocebus atys shows a number of forelimb and shoulder adaptations for stability in terrestrial locomotion and manual foraging not seen in other closely related taxa or other Ta species, including a broad deltoid plane and proximally ext ending supinator crest (Fleagle and McGraw, 1999, 2002) and relatively greater projection of the greater tuberosity superior to the humeral head (Dunham et al., 2015, 2016) Thus, it is not surprising that this species has bone microstructural patterns re lated to these stereotypical behaviors.
86 In sum, the regional variation in remodeling in the Cercocebus atys humerus may be a result of restricted, relatively invariant loads, which is the result of utilizing a stable, terrestrial substrate and having a r elatively narrow range of locomotor and positional behaviors. That is, foraging for the primary food source, Sacoglottis is a stereotypical movement resulting in low load complexity in the Cercocebus atys humeral midshaft. Age at Death Another possible explanation for the results has to do with age and accumulated load history. At a given time, the diaphysis may contain primary bone of varying ages (Enlow, 1962b; Epker and Frost, 1965; Frost, 2001) and bone remodeling activity is positiv ely correlated with age (e.g., Robling and Stout, 2000) Young bone tissue may contain very few secondary osteons (Maggiano et al., 2011) because remodeling patterns do not simply reflect the loading parameters at death; rather they are the results of accu there must be time for the effects of load history to accrue. Unfortunately the ages at death of the individuals in the sample are unknown due to the opportunistic nature of collec ting wild specimens. It is certain that all individuals were skeletally mature prior to death, as determined by epiphyseal fusion, but it is possible that these are very young adults. If this is true then even if the femur and humerus are loaded in a very stereotypical fashion in all species, loading patterns will not be revealed in the distribution of osteons because the effects have not had time to accumulate. However, this does not seem to be a likely explanation for the lack of remodeling patterns becau se the osteon densities relative osteonal areas reported here are not lower than those reported in other non human primate samples (Schaffler and Burr, 1984; Paine and Godfrey, 1997; McFarlin et al., 2008)
87 Osteon Cross sectional Area Osteon cross section al area varies among quadrants of the femoral and humeral midshafts of individuals, with the exception of the Cercocebus atys humerus. The natur e of the heterogeneity, however, is not consistent among individuals of each species. The quadrant with the grea test On.Ar is not the same in different individuals. If osteon size is closely tied to load history, this result implies that individuals of the same species are consistently engaging in behaviors different enough to result in disparate load cases. There a re no b ehavioral data to support this notion. While osteons of varying sizes may be ideal for resisting crack propagation under different strain modes or magnitudes, there is no known mechanism regulating osteon size in such a way. The present results sugg est that o steon size may not be a very reliable indicator of strain mode. Additional Considerations The thin sections examined here were prepared from the mid diaphysis, which may have high load complexity, rather than discrete tension and compression regions. Other aspects of these bones, such as the proximal diaphysis or neck, may experience different loading patterns. Human (Lovejoy et al., 2002) and chimpanzee (Skedros et al., 2011) femoral necks have been modeled as cantilevered beams with clear opposing compression and tension regions. Skedros (2012) suggested that the horse radius can be modeled as bending effects. The same could be true of the femoral diaphysis, but the degree of curvature would have to be very large to significantly amplify bending effects Analyses of second ary bone distribution in bones for which the loading environment has been better characterized may shed light on the results presented here. Demes et al. (2001) demonstrated via in vivo strain gage analyses that the
88 macaque tibia is loaded in bending in an anterolateral posteromedial plane, with the anterior and part of the medial cortex in tension, and the posterior and part of the lateral cortex in compression. In vivo analyses of the macaque ulna (Demes et al., 1998) demonstrated mediolateral bending wit h some torsion. The medial cortex was characterized by compressive strain, and the lateral, tensile strain. The macaque tibia and, to a lesser extent, ulna are thus likely to have low load complexity because they have discrete tension and compression regio ns due to bending, and Skedros (2012) even placed the macaque ulna into a low intermediate category in terms of load complexity. The tibia may be considered to have lower load complexity since torsion is minimal. If bone remodeling is tied to strain mode, patterns may be more likely to be present in these bones, especially the tibia, than in the femur or humerus. The presence of such patterns would support a relationship between strain mode and bone remodeling activity, and support the interpretations of l oad complexity in the Ta monkey sample. Conclusion There are no clear patterns in the distribution of secondary osteons, in terms of density or cross sectional area, in the femoral and humeral midshafts of the cercopithecid monkeys examined. The one exce ption was the humerus of Cercocebus atys. The lack of regional variation in secondary osteon distribution may reflect a complex strain environment in which there are no clear tension and compression regions because of the variable behaviors of these taxa, and/or predominant torsion due to the curvature of the femoral midshaft, which results in evenly distributed shear across the femoral cross section. Cercocebus atys may have a less variable range of locomotor and positional behaviors in addition to adherin g to a more stable substrate.
89 These factors may be responsible for the patterns in secondary osteon density and relative osteonal area in this species. Future studies will compare secondary remodeling in more proximal regions of the femur, and also in othe r bones, such as the tibia, which have lower load complexity and for which in vivo strain distributions are known, in hopes that such analyses will shed light on the interpretation of these results in particular and the mechanical role of secondary bone in general.
90 Table 5 1. Results of the Shapiro Wilk normality tests for each variable within femoral and humeral midshafts of each species. OPD = osteon population density, On.Ar = osteon cross sectional area, %HAV = relative osteonal area. Bone Species OPD On.Ar %HAV Femur Colobus polykomos P = 0.062 P < 0.001* P < 0.001* Piliocolobus badius P = 0.004* P < 0.001* P = 0.004* Cercocebus atys P = 0.002* P < 0.001* P = 0.011* Cercopithecus diana P < 0.001* P < 0.001* P < 0.001* Humerus Colobus polykomos P = 0.377 P < 0.001* P = 0.476 Piliocolobus badius P = 0.086 P < 0.001* P = 0.141 Cercocebus atys P = 0.290 P < 0.001* P = 0.289 Cercopithecus diana P = 0.461 P < 0.001* P = 0.431 *Indicates significant result (i.e., data violate assumptions of normality). Non normal data were z rank transformed.
91 Table 5 2. Summary of osteon population density (OPD), osteon cross sectional area in mm 2 (On.Ar), and relative osteonal area (%HAV) data for each region of the femur for each species sampled. Species Region OPD On.Ar (mm 2 ) %HAV Mean S.D. Range Mean S.D. Range Mean S.D. Range Colobus polykomos Anterior 4.57 3.011 0.000 7.593 0.017 0.009 0.004 0.054 7.43 4.516 0.000 11.665 Posterior 6.90 1.367 5.128 8.399 0.019 0.012 0.003 0.081 13.04 2.480 10.126 16.296 Medial 6.06 5.234 1.041 14.678 0.018 0.011 0.004 0.077 12.67 15.274 1.145 38.849 Lateral 4.72 2.359 0.599 6.167 0.020 0.011 0.004 0.059 8.97 4.544 1.438 13.468 Piliocolobus badius Anterior 5.38 6.718 0.244 13.100 0.018 0.012 0.003 0.081 9.53 12.083 0.299 22.975 Posterior 6.14 3.838 2.957 11.083 0.020 0.012 0.004 0.091 12.65 8.183 4.632 21.932 Medial 5.28 6.998 0.200 14.693 0.016 0.012 0.004 0.077 8.52 11.750 0.269 25.562 Lateral 3.85 3.700 0.278 8.903 0.020 0.013 0.004 0.068 7.44 6.877 0.576 17.586 Cercocebus atys Anterior 1.99 2.540 0.000 6.364 0.019 0.010 0.005 0.069 3 .54 5.165 0.000 12.611 Posterior 4.58 3.277 1.735 10.147 0.019 0.012 0.003 0.087 8.90 5.571 4.910 18.617 Medial 2.57 3.848 0.000 9.158 0.014 0.012 0.004 0.100 3.93 5.167 0.000 12.700 Lateral 1.60 1.524 0.135 3.947 0.019 0.012 0.006 0.056 4.17 4.746 1.349 9.516 Cercopithecus diana Anterior 1.29 2.373 0.000 5.509 0.020 0.011 0.006 0.048 2.57 3.856 0.000 8.969 Posterior 4.68 3.862 1.729 11.478 0.021 0.015 0.004 0.108 8.83 4.530 4.816 16.377 Medial 1.59 2.056 0.000 4.993 0.021 0.013 0.004 0.075 3.06 3.556 0.000 8.389 Lateral 2.79 5.369 0.000 12.379 0.020 0.010 0.006 0.059 5.59 10.184 0.000 23.751
92 Table 5 3. Summary of osteon population density (OPD), osteon cross sectional area in mm 2 (On.Ar), and relative osteonal area (%HAV) data for each region of the humerus for each species sampled Species Region OPD On.Ar (mm 2 ) %HAV Mean S.D. Range Mean S.D. Range Mean S.D. Range Colobus polykomos Anterior 9.61 3.067 6.624 13.847 0.017 0.009 0.004 0.067 15.93 6.414 8.364 23.357 Posterior 6.36 6.126 0.832 13.686 0.016 0.011 0.003 0.082 9.67 8.404 1.481 19.289 Medial 7.11 6.577 0.445 16.025 0.017 0.009 0.004 0.064 13.87 12.723 0.913 27.597 Lateral 9.16 3.680 5.459 14.822 0.016 0.009 0.002 0.100 14.26 5.208 6.662 19.823 Piliocolobus badius Anterior 6.53 1.930 4.147 9.485 0.021 0.015 0.004 0.096 14.05 7.414 7.044 23.233 Posterior 4.81 2.622 0.847 8.068 0.022 0.016 0.002 0.129 10.41 5.902 1.377 16.679 Medial 5.23 6.868 0.114 16.980 0.022 0.013 0.005 0.071 8.28 6.627 0.080 16.541 Lateral 7.49 2.553 5.497 11.338 0.025 0.016 0.005 0.109 19.12 10.312 11.160 37.043 Cercocebus atys Anterior 7.42 3.405 3.346 11.176 0.022 0.012 0.004 0.138 17.94 10.767 7.010 31.774 Posterior 4.22 2.968 0.193 7.153 0.025 0.014 0.005 0.094 10.40 6.736 0.413 17.778 Medial 1.73 1.932 0.193 4.958 0.022 0.011 0.007 0.060 3.97 3.920 0.164 9.783 Lateral 11.73 3.479 7.683 17.168 0.022 0.013 0.004 0.085 25.49 5.232 20.866 34.277 Cercopithecus diana Anterior 7.91 4.500 0.307 12.182 0.021 0.011 0.004 0.073 16.38 9.490 0.583 25.858 Posterior 6.14 4.706 0.443 12.058 0.023 0.013 0.004 0.079 13.00 7.719 0.561 19.721 Medial 3.93 1.863 1.451 6.165 0.026 0.017 0.003 0.114 10.06 4.638 6.264 17.485 Lateral 10.22 5.067 4.936 18.435 0.021 0.013 0.005 0.092 20.10 5.913 12.037 26.489
93 Table 5 4. Results of the resampled ANOVAs testing for differences in osteon population density (OPD) and relative osteonal area (%HAV) between quadrants of the femur and humerus in each species. Both OPD and %HAV are significant for the Cercocebus atys humerus. Species Femur Humerus OPD %HAV OPD %HAV Colobus polykomos P = 0.657 P = 0.306 P = 0.687 P = 0.702 Piliocolobus badius P = 0.863 P = 0.734 P = 0.728 P = 0.162 Cercocebus atys P = 0.246 P = 0.351 P = 0.001* P = 0.002* Cercopithecus diana P = 0.318 P = 0.183 P = 0.157 P = 0.181 *Indicates significant P value Table 5 5. Results of the nested ANOVAs testing for differences in osteon cross sectional area (On.Ar) between quadrants of the femur and humerus within each individual. On.Ar is significantly different between individuals of each species and between quadrants wi thin individuals, with the exception of the Cercocebus atys humerus. The quadrants with the greatest and least On.Ar were not consistent among taxa or even among individuals of the same taxon. Species Femur On.Ar Humerus On.Ar Quadrant Individual Quadrant Individual Colobus polykomos P < 0.001* P < 0.001* P < 0.001* P < 0.001* Piliocolobus badius P < 0.001* P < 0.001* P < 0.001* P < 0.001* Cercocebus atys P < 0.001* P = 0.020* P = 0.108 P < 0.001* Cercopithecus diana P = 0.012* P < 0.001* P < 0.001* P < 0.001* *Indicates significant P value
94 Figure 5 1. Schematic representation of long bone cross section divided into the four quadrants among which OPD, %HAV, and On.Ar data were compared.
95 Figure 5 2. Osteon population density (OPD) in the four quadrants of Cercocebus atys humeral midshaft. The medial quadrant had the lowest mean OPD and %HAV while the lateral quadrant had the highest. The horizontal line inside each box represents the median and the diamond represents the mean. *Indicates significant P value from resampled ANOVAs.
96 Figure 5 3. Relative osteonal area (%HAV) in the four quadrants of Cercocebus atys humeral midshaft. The medial quadrant had the lowest mean OPD and %HAV while the lateral quadrant had the highest. The horizontal line inside each box represents the median and the diamond represents the mean. *Indicates significant P value from resampled ANOVAs.
97 Figure 5 4. Osteon cross sectional area (On.Ar) in the four midshaft quadrants of the femur in Cercocebus atys Cercopithecus diana Colobus polykomos and Piliocolobus badius Differences in On.Ar between quadrants were significant for all species, but the quadrants with highest and lowest mean On.Ar are not consistent among the species. All data are non normally distributed and were thus z rank transformed. The horizontal lin e inside each box represents the median and the diamond represents the mean. *Indicates significant P value from nested ANOVAs.
98 Figure 5 5. Osteon cross sectional area (On.Ar) in the four midshaft quadrants of the humerus in Cercocebus atys Cercopithe cus diana Colobus polykomos and Piliocolobus badius Differences in On.Ar between quadrants were significant for all species, with the exception of Cercocebus atys but, as was the case in the femur, the quadrants with highest and lowest mean On.Ar are n ot consistent among the species. Data for all species, except Colobus polykomos were non normally distributed and were thus z rank transformed. *Indicates significant P values from nested ANOVAs.
99 CHAPTER 6 BONE REMODELING IN THE MACAQUE ( Macaca fascicu laris ) SKELETON: THE EFFECTS OF STRAIN MODE, MAGNITUDE, AND FREQUENCY Introduction Haversian bone remodeling is a process by which cortical bone is resorbed and replaced via the coupled activity of osteoblast s and osteoclasts, resulting in the formation o f secondary osteons. This process allows bone to maintain structural integrity throughout life by replacing microdamage resulting from mechanical deformation (Burr et al., 1985; Burr, 1993; Mori and Burr, 1993; Burr, 2002; Martin et al., 2015), and by prev enting further damage. The latter function is a result of increased bone toughness due to the lower mineral content of newly formed bone (Currey, 1959; Reilly and Burstein, 1974; Reilly et al., 1974; Carter et al., 1976; Carter and Hayes, 1977b) making it more compliant (Martin et al., 2015) and the presence of osteon cement lines which can serve as barriers to prevent crack propagation (Moyle and Bowden, 1984; Gibson et al., 2006; Mohsin et al., 2006; Martin et al., 2015). Bone remodeling also aids in mi neral homeostasis by releasing calcium reserves (Enlow, 1962) Strain is the amount of mechanical deformation, or the relative change in dimension of a bone when a load is applied. Microdamage can occur when strain is of great magnitude (Rubin and Lanyon, 1985; Mullender and Huiskes, 1995; Bouvier and Hylander, 1996; Turner and Pavalko, 1998) or when it occurs repeatedly or cyclically (i.e., fatigue damage) (Carter and Hayes, 1977a) Strain can also occur in different modes: tension, compression, and shea r. While each of these factors (strain magnitude, frequency, and mode) are thought to contribute to microdamage accumulation to some extent, and thus the activation of bone remodeling, the relative contributions of each are
100 not entirely understood. This st udy is set up in two parts: the first part tests hypotheses about strain magnitude and frequency by examining secondary bone in several bones from the macaque skeleton, and the second tests hypotheses about strain mode by analyzing secondary bone in differ ent regions within the macaque femur and tibia. Strain Magnitude and Frequency Lad et al. (2016) found that colobines, which have relatively high mastication frequencies, have greater Haversian remodeling than cercopithecines in the postcanine mandible, e ven though one species, the sooty mangabey ( Cercocebus atys ), engages in powerful (high force magnitude) bites while consuming the very hard seed Sacoglottis gabonensis The authors hypothesized that strain frequency may have a stronger mechanical signal t han strain magnitude for activating remodeling. Chapter four of this dissertation presented conflicting evidence about load magnitude and load frequency in the femur and humerus of cercopithecid primates. The femur of leaping species was found to have more secondary bone than in species that leap less, suggesting that high propulsive forces engender more remodeling. However, in all species there was more secondary bone in the humerus than the femur, despite the hindlimbs routinely experiencing greater peak loads than the forelimbs. This was attributed to the more varied use of the forelimbs, suggesting that load frequency, if load magnitude is low, is an important factor for activating remodeling. Here, hypotheses about load magnitude and frequency are tes ted by comparing the amount of secondary bone in the cortices of different bones in the macaque skeleton: femur, tibia, fibula, and rib. The femur and tibia are weight bearing bones and thus endure higher gravitational load s and impact forces than either t he fibula or rib. The fibula bears a fraction of the axial loads applied to the femur and tibia (Lambert,
101 1971; Goh et al., 1992; Wang et al., 1996) but is loaded with the same frequency as those bones with each step taken. The ribs are not weight bearing and do not experience impact loads. Strain amplitudes in the ribs (Cagle, 2011) do not even approach peak strains in the limbs, but they are loaded cyclically with breathing at a rate of 33 times per minute in macaques (Crosfill and Widdicombe, 1961) whi ch is continuous throughout life. Loading of the ribs is thus high frequency and not particularly high magnitude, compared to the limbs. Consequently, with regards to strain, the femur and tibia are categorized as high magnitude bones, whereas the fibula a nd rib are low magnitude. The rib is high frequency, whereas the others are low frequency. The tibia and fibula are assumed to have the same strain frequency, but different strain magnitudes. Both strain magnitude and strain frequency are expected to contr ibute to microdamage formation, and thus bone remodeling is potentially linked to the interaction of these variables, which can be summarized as strain rate (i.e., deformation over time or the change in strain that occur s in a given time ) 1982; Turner et al., 1995; Mosle y and Lanyon, 1998 ) However, the goal here is to illuminate the respective contributions of load magnitude and frequency to remodeling activity. Three hypotheses can be tested by comparing the relative amount of secondary bone in each of these four bones. Null H ypothesis 1. Neither strain magnitude nor strain frequency has an effect on the amount of secondary bone in the macaque skeleton, but instead there is a systemic rate of remodeling throughout the skeleton. If this hypothesis is true, then there will be no differences in secondary bone among the femur, tibia, fibula, and rib.
102 Alternative Hypothesis 1a. Haversian remodeling activity depends predominantly on strain magnitude, rather than strain frequency. Under this h ypothesis, the femur and tibia are expected to have more secondary bone than the fibula or rib, which have lower strain magnitude. Alternative Hypothesis 1b. Haversian remodeling activity depends predominantly on strain frequency, rather than strain magni tude. If this hypothesis holds true, then the rib should have more secondary bone than all other bones because of its high strain frequency, and the fibula and tibia should have fairly even amounts of secondary bone, because although the tibia has greater strain magnitude, the two bones are loaded with the same frequency. Strain Mode Cortical bone is known to have different mechanical properties, fracture properties, and microdamage effects under different strain modes, especially in tension versus compression (Burstein et al., 1972; Reilly and Burstein, 1975; Carter and Hayes, 1977a; Carter et al., 1981; Burr et al., 1998; Reilly and Currey, 1999, 2000; Hiller et al., 2003). With regards to microdamage, bone in tension has a lower yield strength compared to bone under compression (Burstein et al., 1976; Cezayirlioglu et al., 1985) and fatigue life is shorter than that of bone under compression at the same load magnitude (Caler and Carter, 1989; Pattin et al., 1996) Microdamage tends to accumulate more rapidly in regions of tension but microcracks that form under compression tend to be longer, are less often obstructed by osteon cement lines, and cross lamellae, whereas cracks in tension are more likely to debond lamellae from each other (Carter and Hayes, 1977a; Burr et al., 1998) Different rates of microdamage
103 accumulation and microcr ack growth may result in correspondingly different rates of bone remodeling. There is some evidence that the spatial distribution of secondary osteons in limb bones corresponds to regional strain modes, with greater secondary osteon densities found in re gions of compression than tension (Skedros et al., 1994; Mason et al., 1995; Skedros et al., 1997) Main (2007) and Skedros (2012) caution that such patterns will comp lexity equates to very stereotyped unidirectional bending with distinct, opposing tension and compression cortices. In bones with high load complexity loading is not as predictable and usually involves multidirectional bending and/or torsion. If strain mo de were not consistent in a given region of bone, then there would be no cause for spatial patterns in remodeling to occur. The bones examined in the above studies have low load complexity, thus patterns in remodeling are apparent. Chapter five of this di ssertation quantified secondary bone in quadrants of cercopithecid femora, finding no spatial differences in the amount of secondary bone. This was attributed to high load complexity. The macaque femur is also expected to lack regional differences in secon dary bone, given that macaques are quadrupedal primates like the cercopithecid primates examined in chapter five The exact load case for the primate femur is unknown given the difficulty of in vivo strain gage placement. However, strain distribution is kn own for the macaque tibia. Demes et al. (2001) reported that the macaque ( Macaca mulatta ) tibia is loaded in bending such that the anterior and part of the medial cortex are under tension and the posterior and part of the lateral cortex are under compressi on. Torsion is minimal. Thus, the tibia is considered to have low load
104 complexity and may be more likely than the femur to exhibit regional differences in remodeling. Null Hypothesis 2. Strain mode has no effect on remodeling activity in the macaque skel eton. Under the null hypothesis, quadrants of both femoral and tibial midshafts are expected to have no differences in amount of secondary bone. Alternative Hypothesis 2. Bone remodeling is influenced by strain mode (i.e., tension versus compression) an d spatial patterns in secondary bone distribution correspond to strain distributions in bones with low load complexity. The tibia has lower load complexity than the femur because it has distinct regions of compression and tension, whereas the femur has hig h load complexity. Thus, quadrants within the tibial midshaft are expected to differ in amounts of secondary bone, and quadrants within the femur are not. Regions of compression (i.e., posterior, and part of lateral, according to Demes et al. ) are e xpected to have the most secondary bone within the tibial midshaft. Methods Sectioning Methods and Data Collection The sample comprise s the left femur, tibia, fibula, and one mid level rib from each of five female macaque ( Macaca fascicularis ) skeletons b elonging to the primate skeletal collection at The Ohio State University Department of Anthropology. Individuals are known to be adults but exact age at death is unknown. Thick sections of bone were cut from each bone at 50% the length of the whole bone fo r the femur, tibia, and fibula, and from the middle third of each rib. Each thick section was embedded in epoxy resin Buehler Isomet low speed saw (Illinois Tool Works, Lake Bluff, IL) and diamond
105 wafering blade. The thin sections were then polished on a Buehler MetaServ 250 grinder polisher to remove debris and markings from the saw blade. The sections were stained with a toluidine blue O solution to increase visibility of s econdary osteon cement lines, and dried under weight to prevent warping. Thin sections were mounted to microscopy slides with a cover slip. Four equal quadrants (anterior, posterior, medial, and lateral) were marked on the femur and tibia slides. Each thi n section was photographed under compound light microscopy at 100x magnification. A series of overlapping photographs was taken to capture the entire thin section cortex. The photographs for each section were compiled into one composite image using PTGui P hoto Stitching Software. A 1mm scale bar was also photographed and used to set the scale for size measurements. The composite images were analyzed using ImageJ (Abramoff et al., 2004) The entire cortex was traced to measure total area, and each secondar y osteon and fragmentary osteon was counted and traced to measure area. For the femur and tibia these measures were taken for each quadrant. Three variables were calculated from these measurements: osteon population density (OPD), relative osteonal area (% HAV), and osteon cross sectional area (On.Ar). OPD is the total number of secondary osteons and osteon fragments divided by total cortex area. %HAV is the total area of secondary bone (area of all bone within the cement lines of secondary osteons and fragm entary osteons) divided by the total bone area, multiplied by 100 and expressed as a percentage. OPD and %HAV were also measured with respect to quadrant area for the purposes of comparing tibial and femoral quadrants. On.Ar is the area of a complete secon dary osteon reported in mm 2
106 Statistical Analysis Shapiro Wilk tests for normality (Shapiro and Wilk, 1965) were performed on all data and those that failed to meet assumptions of normality were z rank transformed using R package GenABEL. Analyses of var iance (ANOVAs) were performed with OPD, %HAV, and On.Ar as dependent variables and bone as the independent variable to test for differences in means among femora, tibiae, fibulae, and rib. Post hoc pairwise t tests were performed to determine which bones d iffered from each other. To test for differences among quadrants of the femur and tibia, ANOVAs were performed for OPD and %HAV with quadrant as the dependent variable. Nested ANOVAs were performed with On.Ar for each quadrant as the dependent variables an d quadrant nested within specimen as the independent variables. Post hoc pairwise t tests were performed for OPD in the tibia. Results Among Bone Comparisons OPD and %HAV data for all bones are presented in Table 6 1, and On.Ar me ans for all bones are in Table 6 2. OPD and On.Ar failed the Shapiro Wilk tests (Table 6 3) and were z rank transformed to normal distributions. The ANOVAs for all bones returned significant results for OPD ( P = 0.010) and On.Ar ( P < 0.001), but not for %HAV ( P = 0.101). Post hoc pairwise t tests revealed that for OPD (Table 6 4) the femur is significantly different from the rib and tibia, and the tibia and rib are significantly different from each other. Fibula rib, and tibia fibula comparisons were not significant. The femur has the lowest mean OPD, followed by the tibia, fibula, and then rib (Figure 6 1). Pairwise t test results for On.Ar (Table 6 5) showed that the femur is significantly different from all other bones and that the rib is significantly different from all other
107 bo nes. The order of mean On.Ar from smallest to greatest is rib, femur, fibula, and tibia (Figure 6 2). The lowest On.Ar is the same in all bones: 0.004mm 2 However, the laregest On.Ar is 0.129mm 2 in the femur, 0. 122mm 2 in the tibia, 0.075mm 2 in the fibula, and 0.059mm 2 in the rib. Femur and Tibia Quadrants A summary of OPD and %HAV data for quadrants of the femur a nd tibia are provided in Table 6 6 Femoral OPD and tibial OPD and %HAV failed the Shapiro Wilk tests (Table 6 3) and were z rank transformed to normal distributions. On.Ar data summary is presented in Table 6 7 The ANOVAs (Table 6 8 ) revealed that OPD ( P = 0.448) and %HAV ( P = 0.544) do not differ among quadrants within the femur (Figures 6 4 and 6 5), but that there are significant differences a mong quadrants in the tibia ( P = 0.004 and P = 0.007, respectively). Pairwise t tests (Table 6 9 ) showed that in the tibia the posterior quadrant is significantly different from all other quadrants in both OPD and %HAV. The other quadrants do not differ fr om each other. The posterior quadrant has the greatest OPD (Figure 6 6) and %HAV (Figure 6 7) in the tibia. The lateral quadrant has greater mean OPD and %HAV than in the anterior and medial quadrants but the difference is not statistically significant. O n.Ar in the tibia and femur failed the Shapiro W ilk tests for normality (Table 6 3) and were z rank transformed. The nested ANOVAs for On.Ar returned significant results for quadrants within specimen but not among specimens in the femur, and significant re sults for both quadrants within specimen and among specimens in the tibia (Table 6 10 ). This means that quadrants within each specimen differ from each other in both femur and tibia, but that tibia specimens also differ from each other. However, the order of quadrants from least to greatest On.Ar in each specimen is different in both femur
108 and tibia (Figures 6 6 and 6 7). Thus, there is no consistent pattern of secondary osteon distribution in regards to cross sectional area, in either bone. Discussion Stra in Magnitude and Frequency Null Hypothesis 1, that strain magnitude and frequency do not affect bone remodeling, can be rejected because there were differences in the amount of secondary bone among the femur, tibia, rib, and fibula. This result suggests t hat the load case of these bones may have some effect on remodeling activity. Alternative Hypothesis 1a, which stated that strain magnitude is predominantly responsible for remodeling activity, can also be rejected because the two weight bearing bones (fem ur and tibia) did not have more secondary bone than the non weight bearing fibula and rib. Rather, the opposite was found: the rib, which has the greatest loading frequency, but does not bear body weight or impact forces has more secondary bone than eithe r the tibia or the femur. The fibula does not differ significantly in OPD from the tibia, to which it has equal loading frequency but endures only a fraction of the gravitational forces experienced by the tibia. The relatively high OPD in the rib and fibul a provides strong support for Alternative Hypothesis 1b: remodeling activity depends predominantly on strain frequency, rather than strain magnitude. The greater OPD in the fibula than the femur is surprising if loading frequency is equal, yet axial load s are greater in the tibia. However, gravitational forces are not the only stimulus for bone adaptation; muscular activity can also cause significant bone deformation (Robling, 2009) It is possible that while the fibula does not experience the same gravit ational loading that the tibia does, it might experience high strains due to soft tissue actions, and this would explain the relatively high OPD.
109 The finding that mean On.Ar and maximum On.Ar w ere lower in the rib than all other bones may be a result of t he much narrower cortex in the rib than other bones. Resorption space dimensions are limited by cortical thickness, because resorption space diameters near to cortical thickness jeopardize the structural integrity of the bone (Currey and Shahar, 2013) The fact that osteons do not get nearly a s big as they do in other bones and that the size range is consistently narrow, suggest s this biomechanical constraint occur s in ribs. Strain Mode Osteon population density and relative osteonal area Null Hypothesis 2, that strain mode does not a ffect bone remodeling within quadrants of long bone midshafts, can be rejected because quadrants within the tibial midshaft differed in OPD and %HAV. This result, combined with the lack of significant differences in the femora l midshaft, support Alternative Hypothesis 2. These results were expected because the midshaft of the tibia has low load complexity, as defined by Skedros (2012) compared to the femoral midshaft because it has distinct and consistent regions of tension an d compression. If strain mode affects bone remodeling, patterns in secondary bone distribution are only expected to emerge if patterns of strain are consistent. In Chapter five the lack of differences in secondary bone within quadrants of the femur was att ributed to high load complexity, and bones known to have lower load complexity were predicted to have regional differences. Here, that interpretation is supported. There is a differential remodeling response dependent on strain mode, but the loading of the femoral midshaft is too complex for patterns in remodeling to be detected.
110 The greater OPD and %HAV in the posterior quadrant of the tibia is consistent with prediction based on strain distributions in the macaque tibia (i.e., compression in the posterior and part of the lateral cortex) (Demes et al., 2001) Demes et al. (2001) also reported that part of the lateral aspect of the tibia was included in the compression region. The results here do not show significant diffe rences between the lateral quadrant and the medial or anterior quadrants but mean OPD and %HAV are greater in the lateral quadrant. The lack of significant differences may be due to the fact that the entire lateral cortex is not part of the compression reg ion. Most of the osteons in the lateral quadrant are located in the posterior portion, but this is also true of the osteons in the medial quadrant. These results are also consistent with previous analyses of secondary bone distribution in long bones that found greater secondary osteon densities in regions of compression (Mason et al., 1995; Skedros et al., 1997) which was attributed to higher strains in compression regions (Su et al., 1999) Although bone is weaker under tension (Caler and Carter, 1989; P attin et al., 1996; Burr et al., 1998) compressive strains tend to be higher in long bones loaded in bending because axial loading causes additional compression across the cortex (see 2 in Lieberman et al. ). However, Demes et al. (2001) found that in the macaque tibia compressive strains were not consistently higher than tensile strains across individuals and gaits, which suggests that factors other than strain magnitude are the cause for more remodeling. There may be more secondary bone in regions of compression because of the way microdamage spreads. Microdamage caused by compression often occurs in the form of long er cr acks that obliquely split lamellae and tends to spread easily, compared
111 to microdamage caused by tension, which is usually less e xtensive and tends to debond cement lines and lamellae, rather than split them (Carter and Hayes, 1977a; Burr et al., 1998) Cracks are less likely to propagate though osteonal bone compared to interstitial bone because osteon cement lines act as barriers to crack propagation (O'Brien et al., 2005; Gibson et al., 2006; Mohsin et al., 2006; Mullins et al., 2009) However, expla ining the higher OPD in regions of compression as an adaptation for preventing the spread of microcracks should be done with caution. While the prevention of crack propagation may be a function of secondary osteons, the benefit is plausibly a side effect o f osteons forming for another reason. If microdamage is more extensive under compression, in terms of crack length and propensity to spread, then cracks in these areas might interrupt more osteocytes and osteocyte canaliculi. This in turn may result in mor e signaling for remodeling activation. Testing this hypothesis versus the hypothesis that osteons are adaptations for barring crack propagation would requi re knowing the proportion of osteons that form as a direct response to microdamage versus osteons tha t form in accidental association with microcracks. This information might not be attainable but Martin (2002) proposed that practically all basic multicellular units (BMUs) target microdamage, even if they originate several millimeters away from the target ed microcrack, suggesting that the crack barrier function is likely a side effect. Osteon Cross Sectional Area The results for On.Ar resemble those of Chapter five in that there is regional variation in osteon size within individuals, but no consistent p attern across individuals in terms of which quadrants have the largest and smallest osteons. This suggests that either On.Ar is directly tied to strain mode and that typical locomotor behavior varies wildly across individuals, resulting in different strain distributions, or that On.Ar is not
112 closely related to strain mode. The latter seems more likely given osteon density results and what is known about primate behavior. Osteon size may only be limited by cortex width, given the smaller cross sectional area s in the rib compared to other bones. Conclusion Bone remodeling in the macaque skeleton appears to depend more on the frequency of mechanical loading, rather than the magnitude. The rib has more secondary bone than bones for which load frequency is lower even among weight bearing bones. There is also evidence that remodeling is tied to regional differences in strain mode in bones with low load complexity, and that more remodeling occurs in regions under compression than tension. Exact reasons for this di sparity are unknown, but it seems likely that longer cracks and greater crack propagation interrupt more osteocyte lacunae in compression regions, causing greater remodeling activation frequency. No evidence was found for a relationship between osteon size and strain magnitude, frequency, or mode. Osteon size may only be limited by the width of the bone cortex, so as not to introduce deleterious effects.
113 Table 6 1. Osteon population density (OPD) and relative osteonal area (%HAV) in midshaft thin se ctions of macaque ( Macaca fascicularis ) long bones. Bone Specimen OPD %HAV Femur 29 5.59 11.66 31 3.86 8.71 32 8.29 17.26 I30 3.70 7.62 I36 5.56 11.89 Tibia 29 7.33 12.31 31 4.64 10.38 32 13.49 31.88 I30 6.00 13.50 I36 9.07 23.16 Fibula 29 9.59 18.28 31 6.42 14.89 32 25.49 51.06 I30 3.87 6.55 I36 18.64 43.21 Rib 29 15.78 28.78 31 15.12 25.33 32 20.09 26.05 I30 16.25 23.10 I36 19.69 36.79 Table 6 2. Osteon cross sectional area (On.Ar) in midshaft thin sections of macaque ( Macaca fascicularis ) long bones. Bone Specimen Mean On.Ar SD Femur 29 0.021 0.0117 31 0.023 0.0128 32 0.022 0.0125 I30 0.021 0.0118 I36 0.022 0.0133 Tibia 29 0.018 0.0092 31 0.023 0.015 32 0.025 0.0119 I30 0.023 0.0153 I36 0.027 0.0169 Fibula 29 0.02 0.0118 31 0.024 0.0154 32 0.023 0.0124 I30 0.019 0.0118 I36 0.025 0.0148 Rib 29 0.019 0.0115 31 0.018 0.007 32 0.014 0.0068 I30 0.014 0.0074 I36 0.019 0.0096
114 Table 6 3. Shapiro Wilk normality test results for all data. Sample Variable Result P value all bones OPD failed 0.033* %HAV passed 0.068 On.Ar failed <0.001* femur quadrants OPD passed 0.144 %HAV passed 0.291 On.Ar failed <0.001* tibia quadrants OPD failed 0.022* %HAV failed 0.038* On.Ar failed <0.001* Table 6 4. Pairwise t test p values for osteon population density (OPD) in the femur, tibia, fibula, and rib. OPD is significantly lower in the femur than in the fibula and rib, but not significantly different from the tibia. The tibia is significantly lower than the rib, but not different from the fibula. The fibula and rib are not significantly different. Femur Fibula Rib Fibula 0.018* Rib 0.002* 0.269 Tibia 0.156 0.265 0.035* Table 6 5. Pairwise t test p values for osteon cross sectional area (On.Ar) in the femur, tibia, fibula, and rib. On.Ar is significantly lower in the rib than in all other bones. The tibia and fibula do not differ from each other but both have significantly greater On.Ar than the femur. Femur F ibula Rib Fibula 0.0292* Rib <0.001* <0.001* Tibia 0.003* 0.880 <0.001*
115 Table 6 6 Osteon population density (OPD) and relative osteonal area (%HAV) in quadrants of macaque ( Macaca fascicularis ) femoral and tibial midshaft thin sections. Bone Quadrant Specimen OPD %HAV Femur Anterior 29 8.49 15.74 31 5.01 11.25 32 9.88 18.95 I30 2.39 5.19 I36 8.55 17.88 Lateral 29 2.24 5.64 31 3.50 7.34 32 8.20 19.17 I30 5.13 7.95 I36 3.00 6.04 Medial 29 8.05 14.96 31 3.35 7.44 32 7.79 16.48 I30 1.96 4.39 I36 6.24 12.53 Posterior 29 3.03 9.40 31 3.69 8.95 32 7.11 14.22 I30 5.47 13.23 I36 4.62 10.98 Tibia Anterior 29 3.73 4.20 31 1.83 2.77 32 8.44 21.11 I30 0.90 1.14 I36 3.68 11.94 Lateral 29 3.60 5.74 31 4.38 8.98 32 14.85 35.63 I30 2.24 4.22 I36 10.64 24.80 Medial 29 3.86 6.97 31 0.00 0.00 32 7.58 18.46 I30 1.89 3.81 I36 3.50 9.57 Posterior 29 16.59 29.94 31 10.06 24.46 32 21.71 49.44 I30 14.81 34.79 I36 16.43 40.62
116 Table 6 7 Osteon cross sectional area (On.Ar) in quadrants of macaque ( Macaca fascicularis ) femoral and tibial midshaft thin sections. Bone Quadrant Specimen Mean On.Ar SD Femur Anterior 29 0.019 0.0111 31 0.024 0.0141 32 0.019 0.0108 I30 0.022 0.0081 I36 0.021 0.01 Lateral 29 0.026 0.0132 31 0.02 0.0098 32 0.025 0.0153 I30 0.015 0.0079 I36 0.022 0.0106 Medial 29 0.018 0.0085 31 0.022 0.0135 32 0.022 0.012 I30 0.022 0.0105 I36 0.021 0.0089 Posterior 29 0.031 0.0151 31 0.025 0.0136 32 0.02 0.0105 I30 0.025 0.0145 I36 0.023 0.02 Tibia Anterior 29 0.011 0.0056 31 0.015 0.0045 32 0.025 0.0127 I30 0.013 0.0052 I36 0.037 0.0277 Lateral 29 0.017 0.0074 31 0.02 0.0127 32 0.026 0.0148 I30 0.019 0.0064 I36 0.025 0.0163 Medial 29 0.019 0.012 31 N/A N/A 32 0.024 0.0115 I30 0.02 0.0086 I36 0.028 0.01 Posterior 29 0.019 0.0092 31 0.026 0.0165 32 0.024 0.01 I30 0.025 0.0164 I36 0.025 0.0145
117 Table 6 8 ANOVA results for osteon population density (OPD) and relative osteonal area (%HAV) in femur and tibia quadrants. Quadrants are significantly different in the tibia, but not in the femur. Variable Bone Degrees of Freedom Sum of Squares Mean Square F value p value OPD Femur Region 3 17.5 5.834 0.933 0.448 Residuals 16 100.1 6.254 Tibia Region 3 10.473 3.491 6.732 0.004* Residuals 16 8.298 0.519 %HAV Femur Region 3 52.7 17.58 0.739 0.544 Residuals 16 380.4 23.77 Tibia Region 3 9.860 3.287 5.901 0.00 7* Residuals 16 8.911 0.557 Table 6 9 Pairwise t test p values for osteon population density (OPD) and relative osteonal area (%HAV) in quadrants of the tibia. OPD and %HAV are significantly greater in the posterior quadrant than in any other quadrant. The anterior, lateral, and medial quadrants do not differ. Variable Anterior Lateral Medial OPD Lateral 0.184 Medial 0.800 0.119 Posterior 0.002* 0.031* 0.001* %HAV Lateral 0.161 Medial 0.959 0.147 Posterior 0.002* 0.049* 0.002* Table 6 1 0 Nested ANOVA results for fem oral and tibia l Osteon cross sectional area (On.Ar). Specimen:Quad has only 14 degrees of freedom because the medial quadrant of specimen 31 has no secondary osteons, and thus no On.Ar data. Bone Degrees of Freedom Sum of Squares Mean Square F value p value Femur Specimen 4 2.5 0.629 0.653 0.625 Specimen:Quadrant 15 36.5 2.431 2.524 0.001* Residuals 742 714.8 0.963 Tibia Specimen 4 47.9 11.965 13.207 <0.001* Specimen:Quadrant 14 36.5 2.608 2.879 <0.001* Residuals 778 704.8 0.906
118 Figure 6 1. Boxplot of osteon population density (OPD) in macaque femora, tibiae, ribs, and fibulae. OPD is significantly different among the bones (see Table 6 4 for pairwise comparisons). OPD is z rank transformed. The bottom and top of the boxes indicate the in terquartile range the horizontal bar is the median, the diamond is the mean, and the whiskers represent minimum and maximum data points. *Indicates significant ANOVA result.
119 Figure 6 2. Boxplot of osteon cross sectional area (On.Ar) in macaque femora, tibiae, ribs, and fibulae. On.Ar is significantly different among the bones (see Table 6 8 for pairwise comparisons). The bottom and top of the boxes indicate the interquartile range th e horizontal bar is the median, the diamond is the mean, and the whiskers represent minimum and maximum data points. *Indicates significant ANOVA result.
120 Figure 6 3. Boxplot of relative osteonal area (%HAV) in macaque femora, tibiae, ribs, and fibulae. %HAV is not significantly different among the bones, although pairwise comparisons returned significant results for femur rib and femur fibula comparisons (see Table 6 7). The bottom and top of the boxes indicate the interquartile range the horizontal bar is the median, the diamond is the mean, and the whiskers represent minimum and maximum data points.
121 Figure 6 4. Boxplot of osteon population density (OPD) in femoral midshaft quadrants. OPD is not significantly different among the quadrants. The bottom and top of the boxes indicate the interquartile range the horizontal bar is the median, the diamond is the mea n, and the whiskers represent minimum and maximum data points.
122 Figure 6 5. Boxplot of relative osteonal area (%HAV) in femoral midshaft quadrants. %HAV is not significantly different among the quadrants. The bottom and top of the boxes ind icate the interquartile range the horizontal bar is the median, the diamond is the mean, and the whiskers represent minimum and maximum data points.
123 Figure 6 6. Boxplot of osteon population density (OPD) in tibial midshaft quadrants. OPD is significantly different among the quadrants. Pairwise t tests revealed that the posterior quadrant is significantly different from all other quadrants ( Table 6 10). The bottom and top of the boxes indicate the interquartile range the horizontal bar is the median, the diamond is the mean, and the whiskers represent minimum and maximum data points. *Indicates significant ANOVA result.
124 Figure 6 7. Boxplot of relative osteonal area (%HAV) in tibial midshaft quadrants. %HAV is significantly di fferent among the quadrants. Pairwise t tests revealed that the posterior quadrant is significantly different from all other quadrants ( Table 6 10). The bottom and top of the boxes indicate the interquartile range the horizontal bar is the median, the dia mond is the mean, and the whiskers represent minimum and maximum data points. *Indicates significant ANOVA result.
125 Figure 6 6. Femoral mean z ranked osteon cross sectional area (On.Ar) interaction plot. The plot depicts the interaction from the nested ANOVA. Quadrants differ in On.Ar but the interaction plot demonstrates that the ordering of quadrants with greatest to least On.Ar depends on specimen. Thus, there is no clear pattern of On.Ar distr ibution in macaque femora.
126 Figure 6 7. Tibial mean z ranked osteon cross sectional area (On.Ar) interaction plot. from the nested ANOVA. Quadrants differ in On.Ar but the interaction plot demonstrates that the ordering of quadrants with greatest to least On.Ar depends on specimen. Thus, like the femur, there is no consistent pattern of On.Ar distribution in macaque tibiae. The medial region for specimen 29 is repres ented by a light blue triangle because a lack of On.Ar data in the medial quadrant of specimen 31 caused a discontinuity in the medial (light blue) line.
127 CHAPTER 7 HAVERSIAN REMODELING IN WILD AND CAPTIVE SOOTY MANGABEYS ( Cercocebus atys ) A number of factors are thought to influence the amount of remodeling that occurs in a bone. These include loading magnitude, loading frequency, and strain mode (i.e., whether bone is loaded in tension, compression, or shear). All of these contribute to the formation of microdamage, which stimulates the activation of remodeling. Haversian remodeling is resorption and replacement of primary cortical bone by new bone, as a way of repairing microdamage (Burr et al., 1985; Burr, 1993; Mori and Burr, 1993; Burr, 2002; Mart in et al., 2015) preventing future damage (Moyle and Bowden, 1984; Gibson et al., 2006; Mohsin et al., 2006; Martin et al., 2015) and aiding mineral homeostasis (Enlow, 1962) It results in the formation of secondary osteons, or Haversian systems, which are cylindrical structures formed by concentric bone lamellae surrounding a neurovascular canal. Measuring the density and area of secondary osteons can be used to quantify how much remodeling has occurred in a given area of bone. Previous work suggested that loading frequency might have a stronger influence on bone remodeling activity than previously appreciated. Lad et al. (2016) reported a greater incidence of secondary bone in the postcanine mandible of colobine monkeys than cercopithecine monkeys. The colobines have high er mastication frequencies than the cercopithecines due to their highly f olivorous diet. Even the sooty mangabey ( Cercocebus atys ), which engages in powerful bites during removal of the hard casing of the Sacoglottis gabonensis seed had a lower incidence of secondary bone than the colobines. This finding implied that loading frequency might be an important regulating factor of remodeling activity. Previous chapters of this dissertation also discussed this
128 hypothesis. Chapter 4 reported t hat among cercopithecoids there is more secondary bone in the humerus than in the femur, despite the fact that higher peak loads are found in the hindlimbs compared to the forelimbs. The forelimbs, however, are used for a wider variety of behaviors than th e hindlimbs, which are mostly restricted to locomotion, and may thus have greater loading frequency. Chapter 6 addressed the loading frequency hypothesis more directly and found supporting evidence by demonstrating that the macaque rib has more secondary b one than the femur or tibia. The femur and tibia are weight bearing bones with presumably greater applied loads than the rib, which has a high loading frequency because it is loaded continuously with breathing. Chapter 6 tested the loading frequency hypot hesis in different bone within the skeleton. Here this hypothesis is tested again, but between two populations of the same species that have different activity levels. The amount of secondary bone in the femoral midshaft is compared between captive and wil d sooty mangabeys ( Cercocebus atys ). The captive population is the low loading frequency group and the wild population is the high loading frequency group. Because all individuals are adults of the same species, loading magnitude is not expected to differ between them. The wild group is expected to have more femoral secondary bone than the captive group if bone remodeling activity depends on loading frequency. In the wild C. atys spends much of its time on the forest floor foraging in the leaf litter. The activity budget of the wild population is 10.3% traveling, 24.5% foraging, 18.5% resting, 7.9% social, and 38.8% feeding (McGraw, 2007) C. atys is a quadruped that predominantly engages in terrestrial walking, with little climbing and leaping compared to sympatric species. Captive primate populations are assumed to have
129 lower activity levels than wild populations, especially for locomotor behaviors, because they have smaller home ranges. Home range size is limited in captive animals because travel related activity is restricted by enclosure size and because foraging related travel is unnecessary since food is provisioned (Altmann and Muruthi, 1988) While the exact activity levels of the captive C. atys population examined here are unknown, it can be assume d that this group follows the general trend among captive primates. Methods The sample comprised one femur each of nine adult sooty mangabeys ( Cercocebus atys ). Five of the individuals were from a wild population from Ta Forest, were from a captive colony at Yerkes National Primate Research Center. Histological thin sections were prepared from the femoral midshaft at 50% the length of the whole bone. Small sections of the midshaft were cut and embedded in epoxy resin, and then thi bone using a Buehler Isomet low speed saw (Illinois Tool Works, Lake Bluff, IL) equipped with a diamond wafering blade. The thin section s were polished using a Buehler MetaServ 250 grinder polisher to remove markings and dust from the sectioning process. Polished thin sections were stained with a Toluidine Blue O solution, which enhances the visibility of osteon cement lines, and dried under weight to prevent warping. The thin sections were mounted to microscopy slides with Cytoseal 60 and a cover slip. The thin sections were photographed at 100x magnification under a compound light microscope. Because the field of view does not include the entire thin section, a series of photographs were taken for each specimen, and then stitched together using PTGui Photo Stitching Software. A 1mm scale bar was photographed under the same
130 settings to set the scale for the photographs. The stitched images were analyzed using ImageJ (Abramoff et al., 2004) The enti re bone cortex was measured to get total cortical area. The number and sizes of secondary osteons, identifiable by the presence of a cement line, were record ed Size was measured as area, by tracing each secondary osteon and measuring the area inside the c ircumference The same was done for fragmentary osteons, which are older secondary osteons over which new remodeling has occurred. While some fragmentary osteons no long have their Haversian canals, they are identifiable by their cement lines. Osteon popul ation density (OPD) was calculated as the total number of secondary osteons and fragmentary osteons divided by total cortical area. Relative osteonal area (%HAV) was calculated as the sum of the area of all secondary and fragmentary osteons divided by tota l cortical area, expressed as a percentage. Shapiro Wilk normality tests were performed to assess whether data violated distributional assumptions. Data that failed these tests were z ranked transformed to a normal distribution using the R package GenABEL ANOVAs were performed for both OPD and %HAV with origin (i.e., captive or wild) as the grouping variable. Results OPD and %HAV for wild and captive Cercocebus atys specimens are presented in Table 7 1. A summary of the data is found in Table 7 2. Both w ild and captive populations have a wide range of variation, each with one individual with very high OPD and %HAV (Figures 7 1 and 7 2), but with low median values. OPD failed the Shapiro Wilk normality tests ( P = 0.020) and was z rank transformed. Results of the ANOVAs were non significant for OPD ( P = 0.444 ) and %HAV ( P = 0.358 ), thus there were no differences between the two populations.
131 Discussion There are no differences in femoral secondary bone between wild and captive po pulations of C. atys Since wild primates have greater activity levels, especially related to travel, this finding seems to suggest that the frequency of loading does not affect bone remodeling. However, Chapter 6 presented convincing results that loading frequency is a very important factor, given the ribs and the hindlimb bones. The most likely explanation for the lack of differences between the wild and captive groups is that the activity levels of the two groups are not as different as assumed. The disp arity in loading frequency between the ribs and the lower limb might be much greater than the disparity between femora from these captive and wild populations. There are some additional factors that should be acknowledged that could have influenced the r esults. The first is that age is unknown for all individuals, and age could be important if all of the captive individuals were older than the wild individuals. Bone remodeling increases with advanced age (Kerley, 1965) so if all captive individuals lived to advanced ages, then this could result in greater osteon densities than would be expected if all individuals were roughly the same age. A number of primate species tend to live longer in captivity than in the wild (Tidire et al., 2016) so it is possibl e that the average age of the captive population was higher than that of the wild population. Another factor is that the health status of the individuals in both populations is unknown. Captive or semi provisioned macaques and baboons are known to have gr eater growth rates than wild primates because food is a limiting factor for wild populations (Altmann and Alberts, 1987) Furthermore, o besity is common among captive primates as a result of a high energy diet combined with little energy expenditure (Schwi tzer and Kaummanns, 2001) and Cercocebus atys in particular, is
132 prone to obesity in captivity (Leigh, 1994) Increased body weight results in higher levels of leptin, which is known to directly affect osteoclast and osteoblast activity, resulting in incre ased bone formation (Ducy et al., 2000; Reid et al., 2006; Hamrick and Ferrari, 2008) Furthermore, 30% of zoos studied by Kuhar et al. (2013) have a t least one diabetic primate and 51% of those diabetic primates are Old World monkeys. Many of the individu als from Yerkes National Primate Research Center are diabetic (personal communication with W. S. McGraw). In the wild, a large portion of the C. atys diet is comprised of Sacoglottis gabonensis (McGraw et al., 2011) an oily seed used to treat diabetes in humans (Schmelzer and Gurib Fakim, 2008) One hypothesis (Eggers et al., 2014) for the high prevalence of diabetes in captive C. atys is that they do not have access to the Sacoglottis plant, which would otherwise pr event the disease. Diabetes has been associated with decreased osteoblast activity and increased osteoclast activity (Karsenty and Ferron, 2012; Wu et al., 2015) Thus, obesity (via increased leptin) is associated with increased bone formation and diabetes (reduced insulin receptors) is connected to increased resorption. The obesity and diabetes status of the individuals in this sample are currently unknown, but those factors could have influenced the results of this study. It seems likely that the lack of differences in secondary bone density and area between a wild and captive population are due to faulty assumptions about the magnitude of differences in activity levels. Future research will ideally test hypotheses of loading frequency in different popula tions in which differences in activity levels are better defined, and in which factors such as age, body mass, and health status are known. Such comparisons could be made using captive animals with more restrictive
133 enclosures, and in species for which wild populations engage in more travel than Cercocebus atys to make for a more extreme comparison. Data are available for wild Cercopithecus diana (see C hapter four of the dissertation), which is a much more active species in the wild than Cercocebus atys sp ending more than half of its energy budget on travel and foraging (McGraw, 2007) A comparison of this more active species to a captive population of conspecifics might be a better test of the loading frequency hypothesis.
134 Table 7 1. Osteon population density (OPD) and relative osteonal area (%HAV) data for wild and captive Cercocebus atys Status Specimen OPD %HAV Captive 1 1.61 4.35 2 3.77 8.45 3 7.84 17.73 4 1.01 2.13 5 2.90 9.48 Wild 22 13 1.86 3.20 94 9 2.51 4.21 94 21 7.63 13.59 2023 1.56 2.85 2138 0.53 1.44 Table 7 2. Summary of osteon population density (OPD) and relative osteonal area (%HAV) in femoral midshafts of wild and captive Cercocebus atys Measure Status Median Mean SD OPD Captive 2.90 3. 43 2.696 Wild 1.86 2.82 2.786 %HAV Captive 8.45 8. 43 5.998 Wild 3.20 5.07 4.872
135 Figure 7 1. Osteon population density (OPD) in wild and captive Cercocebus atys femoral midshafts. There is no significant difference between the two populations ( P = 0.444)
136 Figure 7 2. Relative osteonal area (%HAV) in wild and captive Cercocebus atys femoral midshafts. There is no significant difference between the two populations ( P = 0.358)
137 CHAPTER 8 THE ABSENCE OF SECONDARY OSTEONS IN AGED RATS Introduction Haversian remodeling is the resorption and replacement of primary intracortical bone, r esulting in the formation of secondary osteons (i.e., Haversian systems). Secondary osteons are concentric bone lamellae surrounding a central Haversian canal and are characterized by a cement line, which is a remnant from the reversal from the resorption to formation phase of remodeling. Haversian remodeling serves several functions:  aiding in mineral homeostasis by liberating calcium reserves in bone (Enlow, 1962a) ;  replacing bone damaged via strain (Burr et al., 1985; Burr and Martin, 1993; Mori and Burr, 1993; Burr, 2002; Martin et al., 2015) ;  toughening bone via the formation of secondary osteons, which may increase bone compliance (Currey, 1959; Reilly and Burstein, 1974; Reilly et al., 1974; Carter and Hayes, 1976, 1977) and provide barrie rs to crack propagation (Moyle and Bowden, 1984; Gibson et al., 2006; Mohsin et al., 2006) Because of these latter two functions, Haversian remodeling is thought to be one of the primary methods by which bone adapts to its mechanical loading environment and is considered essential for maintaining bone structural integrity. Despite serving these essential functions, Haversian remodeling does not occur in all animals. Generally speaking, non mammals do not exhibit secondary osteons, although some birds, lar ge reptiles, early amphibians, and dinosaurs have evidence of Haversian remodeling (Currey, 2002) Original histological descriptions of mammal bone (Foote, 1916; Enlow and Brown, 1958) indicate that secondary bone is present in primates, artiodactyls, per issodactyls, carnivores, lagomorphs, xenarthrans,
138 proboscideans, sirenians, and cetaceans but the orders Monotremata, Insectivora, Chiroptera, Cingulata, and Rodentia generally lack Haversian remodeling. While rodents are generally considered to lack intr acortical Haversian remodeling, whether they have the capacity to remodel is uncertain. The answer to this question has been controversial and is of importance because it may have implications for the utility of rodents as models for human skeletal biology and disease (Frost and Jee, 1992; Shipov et al., 2013) Singh et al. (1974) rats aged three months, but these do not appear to be secondary osteons because there are no clear cement lines More recently Chow et al. (19 93) reported that resorption and formation are not coupled in rat trabecular bone in a site specific way meaning that while increased resorption and increased formation are linked, formation does not necessarily follow resorption at the same site the way it does in Haversian remodeling. Martiniakov et al. (2006) found no secondary osteons in rats, aged four to six months. However, Baron et al. (1984) and Erben (1996) claimed that rats do demonstrate coupled resorption and formation in trabecular bone as e videnced by cement lines suggesting formation of hemiosteons, which were more numerous in older individuals A number of studies have demonstrated that trabecular bone remodeling (i.e., resorption followed by formation) can been initiated in rat limbs under extreme conditions, such as in ovariectomied rats injected with prostaglandin E 2 (Jee et al., 1990; Jee et al., 1991) Bentolila et al. (1998) found that supraphysiological loading results in intracortical remodeling as evidenced by the appearance of resorption spaces after microdamage was induced but did not provide evidence of secondary osteons.
139 Dietary calcium deprivation, however, does appea r to induce the formation of intracortical Haversian systems (Ruth, 1953) de Winter and Steendijk (1975) found osteon like structures in the femoral diaphysis of rats on a low calcium diet. These structures differed in shape and the manner in which bone w as deposited after resorption, compared to Haversian systems, but are nevertheless not normally found in rat cortical bone This is evidence that rats may have the capacity for remodeling, even if they do not utilize the process under typical physiological conditions. Furthermore, rat limb bones do accrue microdamage under normal physiological loading (O'Brien et al., 2005) so the stimulus for Haversian remodeling should at least theoretically be present. If rats have the capacity for bone turnover, and the microdamage stimulus for activating remodeling is present, then lack of conclusive evidence for intracortical Haversian remodeling is perplexing. One explanation could be that the rats examined in previous studies were too young for Haversian remodelin g to occur. Rats examined in previous studies were relatively young in age (usually 3 6 months, but 18 months at the oldest). In humans Haversian remodeling activity increases with advanced age as microdamage accumulates and bone mineral density decreases as a result of repeated bouts of remodeling (Robling and Stout, 2000) and it may take time for patterns in secondary bone to reflect habitual loading (Gocha and Agnew, 2016) It is unclear whether rat bone density decreases with age outside of experimenta l conditions (Ruth, 1953; Jee et al., 1990; Jee et al., 1991; Kalu, 1991) but older rats have been subjected to habitual mechanical loading over a longer period of time and may be more likely to accumulate the microdamage that incites Haversian remodeling
140 The presence of Haversian remodeling is also uncertain in guinea pigs, which have been studied far less than rats in the context of bone turnover. Foote (1916) and Enlow a nd Brown (1958) are found that contain longitudinal canals, most of which are in the form of indistinct, Fiala (1978) in the provided in these works do not provide definitive evidence of secondary osteons (i.e., no cement lines) in guinea pig bone so it is difficult to understand what this means exactly. Thus, it is unc lear whether Haversian remodeling occurs in guinea pigs Domesticated guinea pigs tend to be longer lived and are larger in body size than rats, po ssibly making them more likely candidates, among rodents, in which to find secondary osteons. Domesticated guinea pigs have lifespans reaching 8 to 10 years (Grzimek, 1990; Richardson, 2000) compared to ~4 years in rats (Nowak, 1991; Weigl, 2005) and wei gh 600 to 1000 grams (Grzimek, 1990) whereas rats weigh 200 to 400 grams (Nowak, 1991) The present study examines femora of aged laboratory Norway rats ( Rattus norvegicus ), 24 months old, and femora of domesticated guinea pigs ( Cavia porcellus ), aged 3 months (sexually mature), for the presence of secondary osteons in order to determine whether Haversian remodeling occurs in these animals under normal physiological conditions. If rodents have the capacity for intracortical Haversian remodeling and the a ppearance of secondary osteons depends on age, the aged rat
141 femora are expected to contain secondary osteons. Finding clear secondary osteons in guinea pig bone would confirm that remodeling occurs in guinea pigs. The presence of secondary osteons in guine a pig bone but absence in rat bone might suggest that Haversian remodeling is body mass dependent, given the larger size yet younger age of the guinea pigs compared to the rats. Methods Previously frozen partial cadavers of 11 male laboratory Norway rats ( Rattus norvegicus ), aged 24 months, and 11 male domesticated guinea pigs ( Cavia porcellus ), 3 months of age, were thawed and the left femur was dissected out of each. Body mass was not measurable because the cadavers were not whole. The femora were macera ted with dish soap and water, and air dried. Histological thin sections were prepared from the midshaft of each rat and guinea pig femur at ~50% the length of the whole bone, and also from the distal diaphysis of the rat femora at ~75% the length of the wh ole bone (Figure 8 1). To prepare the thin sections, blocks of bone were cut at using a Buehler Isomet low speed saw and diamond wafering blade (Ilinois Tool Works, Lake Bl uff, IL). The thin sections were polished using a Buehler Meta Serv 250 grinder polisher to remove any debris, and then stained with a Toluidine Blue O solution to improve the visibility of secondary osteon cement lines. The stained thin sections were drie d under weight to prevent warping and then mounted to a microscopy slide with C ytoseal 60 under a slipcover. The thin sections were examined for presence of osteons using a compound light microscope at 100x magnification.
142 Result s No secondary osteons were observed in any of the rat femoral thin sections (Figure 8 2). Portions of the cortex were characterized by non va scular bone (Figure 8 endosteal surface, especially in th e distal diaphysis (Figure 8 4). The Guinea pig bone had some primary osteons but no secondary osteons were found (Figure 8 5). There was also some nonvascular bone. These observations are consistent with previous descriptions of rat and guinea pig bone. D iscussion There are no secondary osteons, and thus no Haversian remodeling, in either aged rat or three month old guinea pig femora. Thus, the hypothesis that older rats will have accumulated sufficient microdamage to incite Haversian remodeling is not su pported. Further, it cannot be concluded from the guinea pig results that slightly larger rodents do remodel. The question remains: why do rats not exhibit Haversian remodeling, despite accruing bone microdamage and even in older individuals? The absence of Haversian remodeling in rat intracortical bone may be due to the relatively short lifespan. Rats live up to four years, whereas animals known to have significant Haversian remodeling live much longer lives. Although microdamage does occur in rat bone (O 'Brien et al., 2005) it may not accrue at a rate that would compromise the structural integrity of their limb bones. Thus Haversian remodeling would not be energetically efficient for a short lived animal that will not live long enough to suffer from the fatigue effects of accumulated microdamage. Testing this hypothesis would require examining cortical bone of other small bodied but long lived rodents, such as the naked mole rat ( Heterocephalus glaber ). Naked mole rats weigh ~35 grams
143 (Nowak, 1991) and ca n live for more than 17 years (Weigl, 2005) and up to 28 years in captivity (Buffenstein and Jarvis, 2002) much longer than expected for their body size, and are often used as models for studies of resistance to aging. Another option would be to examine o lder guinea pigs that lived closer to 10 years. Finding secondary osteons in these rodents would suggest that Haversian remodeling is lifespan dependent, rather than dependent on body size or phylogenetic factors. Another factor to consider is that bone m odeling is more or less continuous resorption and formation, but the actions of osteoclasts and osteoblasts are not coupled in the way they are in remodelin g. Modeling changes the size, shape and/or position of bones, whereas remodeling replaces bone locally without altering size or shape. An individual is considered skeletally mature when all long bone epiphyses have fused, which occurs after the cessation of growth/mo deling (Parfitt, 2002) but remodeling continues throughout life. In male rats, long bone epiphyses can remain unfused up to 30 months of age (Dawson, 1925) suggesting that modeling could continue for that duration as well. The Optimization Model (Liebe rman et al., 2003) hypothesizes that the transition of modeling to remodeling as the predominant response to loading is a tradeoff between increasing bone strength relative to bone mass. In other words modeling continues until adding more bone becomes ene rgetically inefficient and then remodeling becomes a more efficient response to loading. Based on this model, the rat lifespan may be so short that modeling continues for most of it, so the transition to predominant remodeling does not occur. Even among an imals that do remodel, in regions of newly laid bone
144 there are very few osteons, compared to older bone (McFarlin et al., 2008) presumably because fatigue damage has not occurred. Therefore, rat bone throughout most of the cortex may be too young for remo deling to occur, even in older individuals, because modeling continues so late into life. Long bone diaphyses contain primary bone of varying ages due to modeling drift (Enlow, 1962b; Epker and Frost, 1965; Frost, 2001) It seems likely that a larger propo rtion of the bone cortex of rats is young compared to that of larger bodied adult mammals because modeling continues late into life in rats, and other animals cease bone growth at a younger age, relative to total lifespan. Younger bone is less mineralized and therefore more compliant than older bone. Compliant bone is tougher than stiff bone, meaning more energy is required for microcracks to propagate within it (Martin et al., 2015) Stiffer bone is more brittle and microcracks are more likely to travel fu rther. If rats have more compliant bone because the cortex is comprised of younger bone then remodeling might not be necessary because microcracks do not spread easily. Body size seems to be a less likely explanation for the absence of Haversian remodeli ng in rats. The principle of Dynamic Strain Similarity (Rubin and Lanyon, 1984) states that regardless of body mass or mode of locomotion the limb bones of all animals have essentially the same peak strains (between 2000 3000 microstrain). Thus, small anim als elicit higher strains relative to their body size than larger animals. These strains should be sufficient for inducing microdamage and resulting in a remodeling response, as occurs in large bodied mammals. However, the fact that Haversian remodeling on ly definitively occurs in larger bodied (and longer lived) mammals suggests that bones of
145 smaller mammals must accommodate high strains in another manner. Possible explanations are related back to life span. Small, short lived mammals might deal with high strain and fatigue loading via modeling, since it continues through most of life. Modeling adds bone mass, extending fatigue life by lowering stress and consequently reducing strain. Additionally, small mammals might not suffer the effects of fatigue loadi ng to the extent that longer lived mammals do because of their short lifespans. Furthermore, o steon cross sectional area in limb bones scales with negative allometry to body size among mammals, meaning that osteons are absolutely larger in big mammals but are larger relative to body size in small mammals (Felder et al., 2017) The size of a resorption space is limited by cortical thickness, because having a resorption area diameter close or equal in size to cortical thickness would have deleterious effects (Currey and Shahar, 2013) The animal with the lowest body mass but that still had secondary osteons examined by Felder et al. (2017) was the slow loris, which had a cortical thickness of 600 midshaft falls wit hin that range (Komatsu et al., 2009; Zhang et al., 2010) S econdary osteons have also been found in other primates (Warshaw, 2008) comparable in body weight to the 200 400 gram range in Norway rats: Galao senegalensis (173 315g) and Mirza coquereli (304 326g), which each have only a few secondary osteons, and Callithrix geoffroyi (330 360g), Nycticebus coucang (626 1100g), and Saimiri sp. (662 1020g), which all have numerous secondary osteons. Therefore, size is unlikely to impose a constraint on se condary osteon formation. To rule out the possibility that the lack of remodeling is due to some phylogenetic factor limiting remodeling activity in all rodents, other rodents with varying lifespans and
146 body sizes should be examined for the presence of se condary osteons. Naked mole rats ( Heterocephalus glaber ) would be ideal because, as mentioned above, they have long lifespans but are similar in body size to rats. Larger bodied rodents, such as beavers ( Castor canadensis ; 19.6kg; 23.4yrs maximum lifespan) capybaras ( Hydrochoerus hydrochaeris ; 55kg; 15.1yrs), and nutrias ( Myocastor coypus ; 7.2kg; 8.4yrs) (Ernest, 2003; Weigl, 2005) should also be studied. However, the dietary calcium deprivation studies (Ruth, 1953; de Winter and Steendijk, 1975) suggest that the capacity for Haversian remodeling is present in rodents but just does not occur under normal physiological conditions. A final hypothesis to consider is that rats, and other animals that lack Haversian remodeling, may have an alternative bone rep air process. Seref Ferlengez et al. (2014) demonstrated that diffuse damage (clusters of very small microcracks, as opposed to larger, linear microcracks) that occurred in creep loaded rat ulnae was reduced within 14 days of initiation. There was no eviden ce of intracortical remodeling activation, suggesting that another repair process was responsible, the mechanism of which is unknown. Whether the same process repairs the larger, linear microcracks observed in rats by O'Brien et al. (2005) is also unknown, but it can be hypothesized that rats, and other animals without Haversian remodeling, may have alternative bone repair mechanisms that are not yet appreciated. Conclusion Twenty four month old rats lack the reparative process of Haversian remodeling unde r normal physiological conditions, despite their advanced age compared to rats examined in previous studies. This is perplexing since the limb bones of small rodents theoretically experience strains as high as those in larger animals (Rubin and Lanyon,
147 198 4) and microdamage is known to occur in rat bones (O'Brien et al., 2005) while similarly sized primates are known to exhibit secondary osteons (Warshaw, 2008; Felder et al., 2017) The following hypotheses related to the short lifespan of rats may be like ly explanations: 1) rats do not live long enough for Haversian remodeling to be energetically efficient; 2) modeling continues until well into adulthood in rats (Dawson, 1925) meaning that there may not be a modeling remodeling transition as proposed by th e Optimization Model (Lieberman et al., 2003) ; 3) continuous modeling may result in a lot of young, less mineralized bone which may be more compliant and resistant to crack propagation than the older, more mineralized bone found in adults of longer lived a nimals. Long lived yet small bodied rodents, such as naked mole rats, should be examined to further test the lifespan hypothesis. However, other rodents should be examined to rule out an order wide lack of remodeling that would suggest phylogenetic constra ins are at play.
148 Figure 8 1. Rat bone depicting points on the diaphysis from which thin sections were prepared. One midshaft (M, ~50% length of the whole bone) thin section was prepared from each rat and guinea pig femur, and one distal diaphysis (D) thin section was prepared from each rat femur.
149 Figure 8 2. Composite image of thin sections prepared from the distal diaphysis of a rat femur. No secondary osteons were present, as determined by the absence of cement lines surrounding concentric lamellae and a Haversian canal, in any rat thin section.
150 Figure 8 3. Posteromedial portion of a thin section from the distal diaphysis of a rat femur depicting some non vascular bone in the upper right corner of the image. The presence of non vascular bone is consistent with previous descriptions of rat bone.
151 Figure 8 4. Posterolateral portion of a thin section from the distal diaphysis of a rat the endosteal surface.
152 Figure 8 5. Lateral portion of a thin section from the midshaft of a guinea pig femur depicting primary osteons. No secondary osteons were present in any guinea pig thin section, as determined by the absence of cement lines.
153 CHAPTER 9 DISCUSSION The goal of this dissertation was to contribute to the understanding of three main questions about Haversian remodeling.  What is the likelihood that observed bone remodeling in the skeleton is targeted (i.e., occu rring in regions affected by microdamage due to mechanical loading) versus non targeted (i.e., stochastic and occurring for non mechanical reasons)?  What are the respective contributions of loading magnitude, loading frequency, and strain mode to the m ediation of bone remodeling? Or, how closely does the distribution of secondary bone reflect the loading history of a bone?  How much variation in bone remodeling is there within the skeleton, and can this variation be explained by the unique loading pa rameters of individual bones? Collectively, the answers to these questions will help determine whether Haversian remodeling patterns are useful tools for making behavioral inferences in past populations, including fossil animals. What follows is a summary of the hypotheses tested and results of the dissertation, and a discussion of how those results shed light on these main questions. Summary and Recap Chapter 4. This chapter test ed the hypothesis that bone remodeling is primarily induced by high magnitude loads, likely encountered during leaping/bounding behaviors. Osteon population density (OPD), osteon cross sectional area (On.Ar), and relative osteonal area (%HAV) were measured from femoral and humeral midshaft thin sections of four cercopithecids: Colo bus polykomos Piliocolobus badius Cercopithecus diana and Cercocebus atys All species are generalized quadrupeds but vary in leaping frequency and overall activity budget. Differences between taxa with high ( C. polykomos
154 and P. badius ) and low leaping frequency ( C. diana and C. atys ) were assessed via a phylogenetically informed generalized linear mixed model using Markov Chain Monte Carlo methods. Femoral OPD and %HAV are greater in the high frequency leapers than in low frequency leapers, suggesting t hat frequent high magnitude loads engender remodeling, however there is no similar pattern in the humerus, which presumably also experiences high magnitude loads during leaping. Additionally, OPD and %HAV are greater in the humerus than the femur, despite load magnitude being presumably higher in the femur. These results provide conflicting support for hypotheses about load magnitude and load frequency as they respectively relate to bone remodeling activity. Chapter 5. In bones stereotypically loaded in b ending, predictable tension and compression cortices exist, and this regional specificity is thought to be reflected by osteon size and density. This hypothesis was tested by examining the spatial distribution of secondary osteons throughout femoral and hu meral midshafts of quadrupedal cercopithecid primates. OPD On.Ar and %HAV were measured from four equal quadrants of femoral and humeral midshaft thin sections of Cercopithecus diana Cercocebus atys Colobus polykomos and Piliocolobus badius individuals Resampled Analyses of Variance (ANOVAs) were performed on OPD and %HAV, and nested ANOVAs were performed on On.Ar to test for differences among quadrants. The only significant results for OPD and %HAV were in the Cercocebus atys humerus, in wh ich both were greatest in lateral and least in medial quadrants. On.Ar was significant among quadrants but also among individuals (except in the Cercocebus atys humerus) indicating that while osteon size is heterogeneous, there was no consistent pattern fr om thin section to thin section within a species. The lack of heterogeneity in OPD and
155 %HAV meaning that given natural variability in locomotion, the load case is not stereotypical The significant regional differences in the Cercocebus atys humerus might reflect lower load complexity (i.e., habitual, predictable bending) resulting from the repetitive and consistent locomotor and foraging behaviors unique to this species. Chapter 6. This chapter examined OPD, % HAV, and On.Ar in the femur, tibia, fibula, and rib in macaques to test two sets of hypotheses. Null H ypothesis 1 stated that neither strain magnitude nor strain frequency influences th e amount of secondary bone in the macaqu e skeleton, with the expectation that the four bones would not have different amounts of secondary bone. The implication of this hypothesis was that bone remodeling is a systemic process that occurs at the same rate throughout the skeleton. There were two alternative hypotheses that both expected there to be differences in remodeling among the bones. The first alternative hypothesis stated that remodeling activity primarily depends on strain magnitude, and predicted that the femur and tibia, being weight be aring bones, would have more secondary bone than the fibula and rib, which are not directly weight bearing. The second stated that remodeling depends predominantly on strain frequency, and predicted that the rib would have the most secondary bone because i t is loaded at a much higher frequency than the lower limb bones. Results showed OPD to be greatest in the rib and lowest in the femur; the fibula also had greater OPD than the tibia. This result supported the second alternative hypothesis, that there is a strong effect of loading frequency on bone remodeling. An additional finding was that On.Ar is lower in the rib than the other bones, possibly because the narrower cortex presents a limit to osteon size.
156 Bone remodeling was also examined in quadrants of the femoral and tibial midshafts to test the second set of hypotheses. The second null hypothesis was that strain mode (i.e., compression and tension) does not influence bone remodeling, predicting that there will be no differences in secondary bone among quadrants of either bone. The alternative hypothesis was that strain mode does influence bone remodeling and that secondary bone accumulates in regions of compression more so than in regions of tension. The expectations under this hypothesis were that the tibia, which has consistent tension and compression cortices, will have spatial differences in secondary bone with greater OPD and %HAV in the compression region, whereas the femur, which has a more complex loading history, will have no spatial differences The results supported the alternative hypothesis: more secondary bone was found in the compression region (i.e., posterior quadrant) in the tibia, whereas there were no differences in the femur. Chapter 7. This chapter also tested the hypothesis th at bo ne remodeling is dependent on loading frequency but by comparing femoral OPD and %HAV between wild and captive populations, rather than different bones within the skeleton that have different loading histories. Captive primates have presumably lower overal l activity levels compared to wild primates because they are restricted in their travel by the size of their enclosures and because they do not need to travel for foraging purposes. Thus, the hindlimbs of captive sooty mangabeys ( Cercocebus atys ) experienc e presumably lower loading frequencies. The femora of captive individuals were expected to have less secondary bone than femora from a wild population. This hypothesis was not supported, as there was no difference in OPD or %HAV between captive and wild ma ngabeys.
157 Given the strong support for the frequency hypothesis in Chapter 6, the assumptions of loading frequency disparities between wild and captive populations might be faulty. Chapter 8. Haversian remodeling is localized bone resorption coupled with formation, resulting in secondary osteons. This process is known to play roles in mineral homeostasis and in repair of microdamage caused by mechanical loading. Rodents can be useful models for human bone biology, but rodent bone is generally thought to la ck intracortical Haversian remodeling despite its presumed mechanical and metabolic benefits. One reason definitive secondary osteons have not been observed in rodent bone, except in dietary calcium deprived rats, may be that examined individuals were too young. Rat limb bones accrue microdamage and theoretically experience peak strains similar to those of larger mammals, but secondary osteons typically form in older bone and osteon density is known to increase with advanced age in humans. Older individuals are more likely to have older bone and have experienced repeated mechanical loading, which can induce microdamage. To test whether Haversian remodeling depends on age in rodents, histological thin sections were prepared from the femoral midshaft and dista l diaphysis of aged rats (24 months) and young guinea pigs ( three months) and examined for secondary osteons. None were found, suggesting that although rats may have the capacity for Haversian remodeling, it does not occur under normal physiological condit ions, regardless of age. A likely explanation is that modeling continues through most of the short rat lifespan. Thus, even in older individuals, the limb bone cortices likely contain mostly young primary bone. The short lifespan of rats is a likely explan ation for the lack of Haversian remodeling.
158 Implications of Combined Results Hypotheses The combined results of this dissertation provide convincing support of several hypotheses about Haversian remodeling. The first is that the frequency of loading has a strong influence over how much remodeling will occur in a bone. The finding that the rib, which has a high loading frequency, has more secondary bone than bones of the lower limb, which have lower loading frequency but greater loading magnitude (Chapter 6 ) was the most convincing evidence of this. But other chapters provided additional supporting evidence. The greater incidence of secondary bone in the humerus than the femur in all cercopithecid species examined in Chapter 4 also supported this hypothesis, because even though the hindlimbs experience greater peak propulsive forces than the forelimbs (i.e., greater loading magnitude), the forelimbs are involved in a wide variety of behaviors compared to the hindlimbs, which are restricted to locomotor and po sitional functions. This means that the humerus might have greater loading frequency than the femur, due to more varied use. This dissertation finds evidence that loading frequency is more important than previously appreciated. There was some evidence tha t loading magnitude also affects bone remodeling, which was apparent in the greater incidence of remodeled bone of the leaping species than in the species that leap less often, as described in Chapter 4. Chapter 7 provided some seemingly conflicting result s regarding loading frequency in wild and captive populations. Perhaps differences in wild and captive populations should not have been expected based on presumed activity levels, given that Cercopithecus diana did not have the higher osteon density than a ll of the other species in the Ta monkey sample of Chapter 3, despite being more active than the others
159 (McGraw, 2007) The disparity in loading frequency between ribs and lower limb bones might be much greater than the disparity in loading frequency due to different activity levels between closely related species, or wild and captive populations of the same species. This seems to be a likely explanation for these combined results. This dissertation also found support for the hypothesis than bone remodeli ng is dependent on strain mode. In Chapter 5 more secondary bone was found in the compression cortex than in the opposing tension cortex of the macaque tibia. This finding was consistent with previous reports in lower limb bones of sheep and deer (Skedros et al., 1994) horses (Mason et al., 1995) and horses, elk, and sheep (Skedros et al., 1997) The greater amount of remodeled bone in compression regions might be related to the way that microdamage manifests under compression compared to tension. This also prompts the question of whether the function of osteons as barriers to crack propagation is a byproduct of osteon formation, or an adaptation. An important caveat about strain mode and remodeling is that it appears that spatial patterns in osteon dis tribution only occur in bones that have low load complexity. This explains why there were differences among quadrants of the tibia but not the femur. High load complexity equates to a shifting neutral axis, meaning that there are no consistent tension and compression regions, and often significant torsion (Main, 2007; Skedros, 2012) Skedros (2012) has begun to compile of list of high, intermediate, and low complexity bones and this dissertation supports his idea that load complexity is important to know wh en attempting to use bone microstructure to infer load history. Research aiming to characterize load distributions in primate bones to add to this list will
160 provide new candidates to test hypotheses about strain mode and bone remodeling in other parts of t he primate skeleton. There was also some preliminary evidence presented in Chapter 8 that the presence of Haversian remodeling in mammals depends on lifespan. Rats lack secondary osteons, even aged rats, and this might be because modeling continues throug hout most of their short life. In animals that live longer and cease modeling relatively earlier in life, secondary remodeling might be more likely to occur. This is a hypothesis that will require further testing in other mammals of varying lifespans. Main Questions The three main questions addressed by the dissertation can now be evaluated in light of the results.  What is the likelihood that observed bone remodeling is targeted versus non targeted? Burr (2002) estimated that only 30% of bone remodelin g is targeted to areas affected by microdamage due to mechanical loading, but Martin ( 2002) hypothesized that virtually all intracortical remodeling is targeted, in the absence of trauma. The results of this dissertation overall suggest that Haversian remo deling, in the elements of the primate skeleton examined, is tied to mechanical loading parameters. Thus, this dissertation does not support the notion that very little remodeling is mechanically mediated There was no evidence presented in this dissertati on that remodeling occurs at a systemic rate throughout the skeleton. Instead, it seems to depend very much on strain frequency and strain mode, with high incidences of remodeling found in parts of the skeleton with high loading frequencies, and in within regions of bone loaded in compression.  How closely does the distribution of secondary bone reflect the loading history (i.e., strain magnitude, frequency, mode) of a bone? This question was partly answered
161 by the dissertation. There were several lines of evidence to suggest that strain frequency has a stronger influence on bone remodeling than was previously appreciated. Strain mode was also found to be important, and patterns of secondary bone distribution were shown to reflect strain mode in bones wi th low loading complexity. Such patterns may not be detectable in bones with more complex loading histories. The contribution of strain magnitude to the mediation of bone remodeling is still somewhat unclear. Results of the dissertation suggest that high s trains do not generate the extent of remodeling that low strain but high frequency loading does, and bones loaded with the same frequency but different strain magnitudes do not seem to have predictable amounts of remodeling (e.g., the fibula has more secon dary bone than (Frost, 1987) emphasis on minimum effective strains as the primary impetus for remodeling activation.  How much variation in bone remodeling is there within the skeleton, and can this variation be explained by the unique loading parameters of individual bones? The answer to this question is that there is considerable variation in remodeling throughout the skeleton, and this variation is, for reasons discussed above, reflective of the unique loading histories of individual bones, especially of loading frequency.
162 CHAPTER 10 CONCLUSION This dissertation addressed several questions about Haversian remodeling in primate cortical bone. As discussed in the previous chapter these q uestions have been partially answered. Below is a reflection on the ultimate goal of the dissertation within the scope of biological anthropology: determining whether the density and distribution of secondary bone are useful tools for making behavioral inf erences in past populations, particularly in the primate fossil record. Future directions are also discussed, as this dissertation resulted in some incompletely answered questions and generated some new questions to be addressed post dissertation. Bone Rem odeling as a Tool for Behavioral Inferences The ultimate goal of this dissertation was to be able to assess the utility of secondary bone density and distribution as tools for inferring load history in cases in which it cannot be measured directly. Such c ases might be in bones of extant animals for which it is difficult or unethical to obtain in vivo strain gage data, but this is primarily of interest in the context of interpreting behavior in past populations. This goal is of interest to biological anthro pologists, paleontologists, and bioarchaeologists alike. Bone remodeling is a good candidate for such a tool because it is theoretically directly tied to the behaviors of an individual during life, and because measures of cross sectional geometry (Pearson and Lieberman, 2004) and muscle attachment size (Rabey et al., 2015) might not be as reliable as once assumed. Regarding this goal, loading frequency appears to be a strong mediator of bone remodeling, perhaps even more so than strain magnitude. This fact might help illuminate answers to debates about behavior in the fossil record that have been
163 unanswerable based on other measures of morphology. For example, there has been much discussion over interpretations of the hyper robust feeding apparatus of Paran thropus (Jolly, 1970; Du Brul, 1977; Kay, 1985; Ungar et al., 2008; Constantino et al., 2009; Cerling et al., 2011; Pampush et al., 2013) predominantly revolving around whether robust australopiths were adapted to hard object consumption or a tough diet. Hard and tough diets are both mechanically challenging. Tough diets require more prolonged mastication than hard diets, but hard diets require more forceful mastication. Lad et al. (2016) pointed out, in light of the finding that folivorous monkeys have mo re secondary bone in the mandible than frugivorous monkeys, including the durophagous sooty mangabey, that the low incidence of remodeling in the mandible of Paranthropus robustus (Daegling and Grine, 2007) might reflect a low mastication frequency. They a lso hypothesized that if microwear analysis (Ungar et al., 2008) correctly suggested that Paranthropus boisei was not dur o phagous but instead consumed a fibrous diet, then this species would be expected to have a high secondary osteon density. This dissertation provides support for this hypothesis and a histological analysis of the mandible of P. boisei would shed light on th e issue of diet and feeding behavior in this species. Similarly, in an analysis of shearing quotients among Miocene apes Ungar and Kay (1995) found that most taxa fell within the diversity of extant apes. Ouranopithecus macedoniensis was an outlier, fall ing below this range. They interpreted this as evidence of hard object feeding in Ouranopithecus and used the presence of thick enamel as supp orting evidence of their claim. Thick enamel has since been demonstrate d to be homoplas tic in primates (Pampush et al., 2013) having evolved as
164 an adaptation for both hard object feeding (to prevent cracking), and tough or abrasive diets (to accommodate a lot of wear). Thus, thick enamel might not be the best feature to use as additional support for an interpretation of hard object feeding. An analysis of mandibular secondary bone could be an additional line of evidence supporting the durophagy hypothesis if Ouranopithecus has low secondary osteon density. The results with regard to the strain mode hypothesis also have implications for use in the fossil record. Secondary bone density is greater in regions of compression than tension in bones with low load complexity. This information might also be useful for making inferences about behavior. For example, loading is more varied in arboreal locomotion than terrestrial locomotion (Demes and Carlson, 2009) because arboreal substrates are less stable and more variable than the ground. Thus, load complexity may be lower in predominantly terrestrial species than arboreal species, as demonstrated in the sooty mangabey humerus compared to the humerus of sympatric species that spend most of their time in the trees. This could serve as one line of evidence in determinin g whether quadrupedal fossil primates were predominantly arboreal or terrestrial. Future Directions The completion of this dissertation has resulted in more new questions than answered questions. Outlined below are several lines of future work based on n ew hypotheses prompted by results of the dissertation. Compressive strain and microdamage. One result of this dissertation was a high incidence of secondary bone in regions of consistent compressive strain compared to regions of tension. The result was co nsistent with previous comparisons of opposing cortices (Skedros et al., 1994; Mason et al., 1995; Skedros et al., 1997) that also
165 suggested a relationship between osteon density and strain mode. According to Demes et al. (2001) strains are not consistent ly higher in compression regions than tension regions in the macaque tibia, which was the bone in question. Thus, why the compression region has more secondary bone is unknown. It might be related to the way that microdamage manifests under compression com pared to tension. Although bone is weaker under tension (i.e., it fails more quickly than bone under tension and microdamage accumulates more rapidly) (Reilly and Burstein, 1974,1975; Caler and Carter, 1989; Burr et al., 1998) m icrodamage formed by compre ssion tends to be more extensive, more likely to spread, with oblique cracking and longitudinal splitting of lamellae. Cracks that form under tension tend to be shorter and cause debonding (separation) of lamellae or at cement lines (Carter and Hayes, 1977 ; Burr et al., 1998) The higher incidence of secondary bone in compression regions might be a response to the more extensive microdamage that occurs there. Furthermore, the function of secondary osteons as barriers to crack propagation is often discussed with language implying it is an adaptation, rather than a by product of Haversian remodeling. An examination of remodeling activity in response to different types of microdamage might shed light on this issue. Strain magnitude. The mechanostat model depe nds on mediation of bone mass, including remodeling, by minimum effective strains (Frost, 1987) but this dissertation provided evidence that loading frequency is a stronger mediator of remodeling activity. Even in two bones with similar loading frequencie s (tibia and fibula), the one with presumably lower strains had more remodeling, which is in direct contrast to what should be predicted by the mechanostat. This contributes to building skepticism
166 about the mechanostat bone rem odeling (Martin, 2000) and bone formation and mineralization ( Rubin et al., 2001 ). The role of strain magnitude needs to be further investigated. Lifespan. Femora of rats aged 24 months were shown to be devoid of secondary osteons, despite their advanced age, the theoretical assumption that the rat skeleton experiences high strains (Rubin and Lanyon, 1984) and does accumulate microdamage (O'Brien et al., 2005 ) Reasons for this are most likely related to the short rat lifespan combined with the fact that modeling continues throughout most of life. Furthermore, small bodied primates are known to exhibit secondary osteons (Warshaw, 2008; Felder et al., 2017) O ther small bodied rodents with longer lifespans, such as naked mole rats, should be examined for evidence of Haversian remodeling to further test the lifespan hypothesis. Other rodents should be examined also, to rule out a phylogenetic explanation (i.e., all rodents lack remodeling). Age and bone remodeling patterns. Age was repeatedly mentioned throughout this dissertation as a possible confounding factor in the interpretation of results. Unknown age at death is a limitation of studies using wild populat ions from which skeletal materials are opportunistically collected. While most studies on age and bone histomorphology have focused on humans, McFarlin et al. (2008) demonstrated that among primates there are different patterns of secondary bone distribut ion in the femoral and humeral midshafts across taxa and age groups, and between bones, largely due to the effects of modeling and cortical drift. The outer portion of the cortex is younger (i.e., more recently deposited) than the middle and inner portion, thus younger individuals have fewer secondary osteons in this region. If younger individuals, even if
167 adults, have fewer osteons than older individuals, it is possible that patterns that reflect loading history do not become apparent until later in life. On the other end of the spectrum, remodeling is known to increase with advanced age in humans (Kerley, 1965; Stout and Paine, 1992) Since age is known to affect bone remodeling patterns on both ends of the age spectrum, the likelihood that age is an impor tant factor in a sample of non human primates is good information to have. An analysis of secondary bone in non human primates in an ontogenetic series ranging from juvenile to senescent, for which ages at death are known (e.g., the Cayo Santiago macaque s keletal collection), might illuminate whether having young individuals or very old individuals is likely to a ffect interpretations of loading history. Remodeling across taxa. Secondary osteon density ranges and patterns of spatial distribution should be d ocumented for many skeletal elements in a variety of primate taxa that engage in different behaviors (and should thus have different strain distributions, loading frequencies, and load magnitudes) in order to characterize how different behaviors manifest i n bone microstructure. This would provide a range of patterns against which to compare patterns in fossil taxa, allowing for behavioral interpretations through the use of modern analogs. Studies on loading patterns and strain distributions should be consul ted to determine which bones are likely to have low load complexity versus high. Micro CT imaging of bone histomorphology. Micro computed tomography (micro CT) methods for visualizing secondary bone should be used for two purposes.  Histological metho ds are destructive and applying them to rare specimens (fossils, especially) is often met with resistance. Micro CT provides and opportunity to view bone
168 microstructure without causing any damage to the specimen. There should be an investigation into wheth er this is possible in fossil bone.  Dechow et al. (2008) used nano CT to measure length and splitting patterns in osteon networks and Maggiano et al. (2016) used synchrotron nano CT to create three dimensional reconstructions of Haversian systems. Cont inuing to use this method of visualizing Haversian systems will provide a more efficient way to characterize of osteon branching, trajectories, lengths, and would provide a new perspective from which to assess patterns of remodeling that cannot be achieved with two dimensional histological methods.
169 LIST OF REFERENCES Abbott S, Trinkaus E, Burr DB. 1996. Dynamic bone remodeling in later Pleistocene fossil hominids. Am J Phys Anthropol 99:585 601. Abramoff MD, Magalhaes PJ, Ram SJ. 2004. Image processing wi th ImageJ. Biophotonics Int 11:36 42. Allen MR, Burr DB. 2014. Bone modeling and remodeling. In: Burr DB, Allen MR, editors. Basic and Applied Bone Biology. New York: Academic Press. p 75 90. Altmann J, Alberts S. 1987. Body mass and growth rates in a wild primate population. Oecologia 72:15 20. Altmann J, Muruthi P. 1988. Differences in daily life between semiprovisioned and wild feeding baboons. Am J Primatol 15:213 221. Arnold C, Matthews LJ, Nunn CL. 2010. The 10kTrees website: A new online resource for primate phylogeny. Evol Anthropol 19:114 118. Aulchenko YS, Ripke S, Isaacs A, van Duijn CM. 2007. GenABEL: an R package for genome wide associatio analysis. Bioinformatics 23:1294 1296. Badoux DM. 1974. An introduction to biomechanical principles in prim ate locomotion and structure. In: Jenkins FA, editor. Primate Locomotion. New York: Academic Press. p 1 43. Baron R, Tross R, Vignery A. 1984. Evidence of sequential remodeling in rat trabecular bone: morphology, dynamic histomorphometry, and changes durin g skeletal maturation. Anat Rec 208:137 145. Bateson G. 1963. The role of somatic change in evolution. Evolution 17:529 539. Bell C. 1827. Animal Mechanics, or Proofs of Design in the Animal Frame. Cambridge, MA: Morrill Wyman. Bentolila V, Boyce TM, Fyhri e DP, Drumb R, Skerry TM, Schaffler MB. 1998. Intracortical remodeling in adult rat long bones after fatigue loading. Bone 23:7275 7281. Biewener AA. 1983. Locomotory stresses in the limb bones of two small mammals: the ground squirrel and chipmunk. J Exp Biol 103:131 154. Biewener AA, Bertram JE. 1993. Mechanical loading and bone growth in vivo In: Hall BK, editor. Bone Growth B. Bone. Boca Raton: CRC Press. p 1 36. Biewener AA, Swartz SM, Bertram JEA. 1986. Bone modeling during growth: Dynamic strain equilibrium in the chick tibiotarsus. Calcif Tissue Int 39:390 395.
170 Blomberg SP, Garland T. 2002. Tempo and mode in evolution: phylogenetic inertia, adaptation and comparative methods. J Evol Biol 15:899 910. Bock WJ. 1980. The definition and recognition o f biological adaptation. Am Zool 20:217 227. Bourgey JM. 1832. Traite Complet de l'Anatomie de l'Homme. I. Osteologie. Paris. Bouvier M, Hylander WL. 1996. The mechanical or metabolic function of secondary osteonal bone in the monkey Macaca fascicularis A rchs Oral Biol 41:941 950. Bromage TG. 1989. Ontogeny of the early hominid face. J Hum Evol 18:751 773. Buffenstein R, Jarvis JUM. 2002. The Naked Mole Rat -A New Record for the Oldest Living Rodent. Sci Aging Knowl Environ 2002:pe7 Burguera B, Hofbauer L, Thomas T, Gori F, Lassam J, Lassko K, Evans G, Khosla S, Riggs BL, Turner RT. 2001. Leptin reduces ovariectomy induced bone loss in rats. Endrocinology 142:3546 3553. Burr DB. 1993. Remodeling and the repair of fatigue damage. Calcif Tissue Int 53:S75 S 81. Burr DB. 2002. Targeted and nontargeted remodeling. Bone 30:2 4. Burr DB, Martin RB. 1993. Calculating probability that microcracks initiate resorption spaces. J Biomech 26. Burr DB, Martin RB, Schaffler MB, Radin EL. 1985. Bone remodeling in response to in vivo microdamage. J Biomech 18:189 200. Burr DB, Robling AG, Turner CH. 2002. Effects of biomechanical stress on bone in animals. Bone 30:781 786. Burr DB, Ruff CB, Thompson DD. 1990. Patterns of skeletal histologic change through time: comparison o f an archaic native American population with modern populations. Anat Rec 226:307 313. Burr DB, Turner CH, Naick P, Forwood MR, Ambrosius WT, Hasan MS, Pidaparti R. 1998. Does microdamage accumulation affect the mechanical properties of bone? J Biomech 31: 337 345. Burstein AH, Currey JD, Frankel VH, Reilly DT. 1972. The ultimate properties of bone tissue: the effects of yielding. J Biomech 5:35 44. Burstein AH, Reilly DT, Martens M. 1976. Aging of bone tissue: Mechanical properties. J Bone Joint Surg [Am] 5 8:82 86.
171 Cagle DJ. 2011. Investigation of respiration induced strain caused on the rib. In: Mechanical Engineering. Columbus, OH: The Ohio State University. Caler WE, Carter DR. 1989. Bone creep fatigue damage accumulation. J Biomech 22:625 635. Carter DR. 1984. Mechanical loading histories and cortical bone remodeling. Calcif Tissue Int 36:S19 S24. Carter DR, Caler WE, Spengler DM, Frankel VH. 1981. Fatigue behavior of adult cortical bone: The influence of mean strain and strain range. Acta Orthop Scand 52 :481 490. Carter DR, Hayes WC. 1976. Fatigue life of compact bone. I. Effects of stress amplitude, temperature and density. J Biomech 9:27 34. Carter DR, Hayes WC. 1977a. Compact bone fatigue damage a microscopic examination. Clin Orthop Relat Res 127:26 5 274. Carter DR, Hayes WC. 1977b. Compact bone fatigue damage. I. Residual strength and stiffness. J Biomech 10:325 337. Carter DR, Hayes WC, Schurman DJ. 1976. Fatigue life of compact bone. II. Effects of microstructure and density. J Biomech 9:27 34. Ca rter DR, Vasu R. 1981. Stress fields in the unplated and plated canine femur calculated form in vivo strain measurements. J Biomech 14:63 70. Cerling TE, Mbua E, Kirera FM, manthi FK, Grine FE, Leakey MG, Sponheimer M, Uno KT. 2011. Diet of Paranthropus boisei in the early Pleistocene of East Africa. Proc Natl Acad Sci USA 108:9337 9341. Cezayirlioglu H, Bahniuk E, Davy DT, Heiple KG. 1985. Anisotropic yield behavior of bone under combined axial force and torque. J Biomech 18:61 69. Chinsamy A. 1990. Phys iological implications of the bone histology of Syntarsus rhodesiensis (Saurischia: Theropoda). Palaeontol africana 27:77 82. Chinsamy A. 1993. Dicynodont (Therapsida) bone histology: phylogenetic and physiological implications. Palaeontol africana 30:97 1 02. Chinsamy A. 1995. Ontogenetic changes in the bone histology of the late jurassic onithopod Dryosaurus lettowvorbecki J Vertebr Paleontol 15:96 104. Chinsamy A. 1997. Assessing the biology of fossil vertebrates through bone histology. Palaeontol africa na 33:29 35.
172 Chow JWM, Badve S, Chambers TJ. 1993. Bone formation is not coupled to bone resorption in a site specific manner in adult rats. Anat Rec 236:366 372. Constantino PJ, Lucas PW, Lee JJ W, Lawn BR. 2009. The influence of fallback foods on great a pe tooth enamel. Am J Phys Anthropol 140:653 660. Cooper SJ. 2008. From Claude Bernard to Walter Cannon. Emergence of the concept of homeostasis. Appetite 51:419 427. Cowin S. 2001. The false premise in Wolff's law. In: Cowin S, editor. Bone Biomechanics H andbook, 2 ed. Boca Raton: CRC Press. p 30 31 30 15. Crosfill ML, Widdicombe JG. 1961. Physical characteristics of the chest and lungs and the work of breathing in different mammalian species. J Physiol 158:1 14. Crowder C, Heinrich J, Stout S. 2012. Rib histomorphometry for adult age estimation. Method Mol Biol 915:109 127. Crowder C, Stout S, editors. 2012. Bone Histology: An Anthropological Perspective. Boca Raton: CRC Press Taylor & Francis Group. Currey JD. 1959. Differences in the tensile strength of bone of different histological types. J Anat 98:87 95. Currey JD. 2002. Bones: Structure and Mechanics. Princeton, NJ: Princeton University Press. Currey JD. 2003. The many adaptations of bone. J Biomech 36:1487 1495. Currey JD, Shahar R. 2013. Cavities i n the compact bone in tetrapods an fish and their effect on mechanical properties. J Struct Biol 183:107 122. Daegling DJ, Grine FE. 2007. Mandibular biomechanics and the paleontological evidence for the evolution of human diet. In: Ungar PS, editor. Evolu tion of the Human Diet: The Known, the Unknown, and the Unknowable. Oxford: Oxford University Press. p 77 105. Dawson A. 1925. The age order of epiphyseal union in the long bones of the albino rat. Anat Rec 31:1 17. de Winter FR, Steendijk R. 1975. The eff ect of a low calcium diet on lactating rats; Observations on the rapid development and repair of osteoporosis. Calcif Tissue Res 17:303 316. Dechow PC, Chung DH, Bolouri M. 2008. Relationship between three dimensional microstructure and elastic properties of cortical bone in the human mandible and femur. In: Vinyard CJ, J RM, Wall CE, editors. Primate Craniofacial Function and Biology. New York: Springer. p 265 292.
173 Deere K, Sayers A, Rittweger J, Tobias JH. 2012. Habitual levels of high, but not moderate o r low, impact activity are positively corrrelated to hip BMD and geometry: results from a population based study of adolescents. J Bone Miner Res 27:1887 1895. Demes B, Carlson KJ. 2009. Locomotor variation and bending regimes of capuchin limb bones. Am J Phys Anthropol 139:558 571. Demes B, Fleagle JG, Jungers WL. 1999. Takeoff and landing forces of leaping in strepsirhine primates. J Hum Evol 37:279 292. Demes B, Larson SG, Stern JT, Jungers WL, Biknevicius AR, Schmitt D. 1994. The kinetics of primate qua drupedalism: "hindlimb drive" reconsidered. J Hum Evol 26:353 374. Demes B, Qin YX, Stern JT, Larson SG, Rubin CT. 2001. Patterns of strain in the macaque tibia during functional activity. American Journal of Physical Anthropology 116:257 265. Demes B, Ste rn JT, Hausman MR, Larson SG, McLeod KJ, Rubin CT. 1998. Patterns of strain in the macaque ulna during functional activity. American Journal of Physical Anthropology 106:87 100. DiGirolamo DJ, Clemens TL, Kousteni S. 2012. The skeleton as an endocrine orga n. Nat Rev Rheumatol 8:674 683. Du Brul EL. 1977. Early hominid feeding mechanisms. Am J Phys Anthropol 47:305 320. Ducy P, Amling M, Takeda S, Matthias P, Schilling AF, Beil FT, Shen J, Vinson C, Rueger JM, Karsenty G. 2000. Leptin inhibits bone formation through a hypothalamic relay: a central control of bone mass. Cell 100:197 207. Dunham N, Kane E, McGraw WS. 2015. Forelimb use and scapular morphology in four sympatric cercopithecids. Folia Primatol 86:474 489. Dunham N, Kane E, McGraw WS. 2016. Humeral correlates of forelimb elevation in four West African cercopithecid monkeys. Am J Phys Anthropol 162:337 349. Eggers NA, Muoz Acua U, Kane E, McGraw WS, Voldovotz Y, Carcache de Blanco EJ. 2014. Nutritional profile, biological screening, and fingerprint characterization of Sacoglottis gabonensis In: The Ohio State University College of Pharmacology Research Day. p 23. Elefteriou F, Takeda S, Ebihara K, Magre J, Patano N, Kim CA, Ogawa Y, Liu X, Ware SM, Craigen WJ, Robert JJ, Vinson C, Nakao K, Capeau J Karsenty G. 2004. Serum leptin level is a regulator of bone mass. Proc Natl Acad Sci USA 101:3258 3263.
174 Enlow D. 1962a. Functions of the Haversian system. Am J Anat 110:269 305. Enlow D. 1962b. A study of postnatal growth and remodeling of bone. Am J Ana t 110:79 101. Enlow DH, Brown SO. 1958. A comparative histological study of fossil and recent bone tissues. Part III. Texas J Sci 10:187 214. Epker BN, Frost HM. 1965. A histological study of remodeling at the periosteal, Haversian canal, cortical endostea l, and trabecular endosteal surfaces in human rib. Anat Rec 152:129 135. Erben RG. 1996. Trabecular and endocortical bone surfaces in the rat: modeling or remodeling? Anat Rec 246:39 46. Ernest SKM. 2003. Life history characteristics of placental non volan t mammals. Ecology 84:3402 3402. Felder AA, Phillips C, Cornish H, Cooke M, Hutchinson JR, Doube M. 2017. Secondary osteons scale allometrically in mammalian humerus and femur. R Soc open sci 4:170431. Felsenstein J. 1985. Phylogenies and the comparative m ethod. Am Nat 125:1 15. Fiala P. 1978. Age related changes in the substantial compacta of the long limb bones. Folia Morphologica 26:316 320. Fleagle JG, McGraw WS. 1999. Skeletal and dental morphology supports diphyletic origin of baboons and mandrills. P roc Natl Acad Sci USA 96:1157 1161. Fleagle JG, McGraw WS. 2002. Skeletal and dental morphology of African paionins: unmasking a cryptic clade. J Hum Evol 42:267 292. Fleagle JG, Meldrum DJ. 1988. Locomotor behavior and skeletal morphology of two sympatric pitheciine monkeys, Pithecia pithecia and Chiropotes satanas Am J Primatol 16:227 249. Foote JS. 1916. A contribution to the comparative histology of the femur. Smithsonian Contributions to Knowledge 35. Franz TM, Demes B, Carlson KJ. 2005. Gait mechanic s of lemurid primates on terrestrial and arboreal substrates. J Hum Evol 48:199 217. Frost HM. 1958. Preparation of thin undecalcified bone sections by rapid method. Stain Technol 33:273 277. Frost HM. 1960a. In vivo osteocyte death. J Bone Joint Surg [Am] 42:138 143.
175 Frost HM. 1960b. Presence of microscopic cracks in vivo in bone. Henry Ford Hosp Med BUll 8:25 35. Frost HM. 1964. The Laws of Bone Structure. Springfield: Thomas. Frost HM. 1969. Tetracycline based hist ological analyses of bone remodeling. Calcif Tissue Res 3:211 237. Frost HM. 1983. A determinant of bone architecture: the minimum effective strain. Clin Orthop 175:286 292. Frost HM. 1987. Bone "mass" and the "mechanostat": a proposal. Anat Rec 219:1 9. F rost HM. 1997. Why do marathon runners have less bone than weight lifters? A vital biomechanical view and explanation. Bone 20:183 189. Frost HM. 1998. Changing concepts in skeletal physiology: Wolff's law, the mechanostat, and the "Utah paradigm". Am J Hum Biol 10:599 605. Frost HM. 2000. The Utah paradigm of skeletal physiology: an overview of its insights for bone, cartilage and collangeous tissue organs. J Bone Miner Metab 18:305 316. Frost HM. 2001. From Wolff's Law to the Utah Paradigm: insights abo ut bone physiology and its clinical applications. Anat Rec 262:398 419. Frost HM, Jee WSS. 1992. On the rat model of human osteopenias and osteoporoses. Bone Miner 18:227 236. Fukumoto S, Martin TJ. 2009. Bone as an endocrine organ. Trends Endocrin Met 20: 230 236. Galileo G. 1638. Discorsi e dimonstrazioni matematiche, intorno a due nuove scienze attentanti alla meccanica ed a muovementi locali. Madison: University of Wisconsin Press. Galluscio RM, McGraw WS, Rapoff AJ, Daegling D J. 2016. Hierarchical varia tion of bone hardness in the colobine femoral midshaft. Am J Phys Anthropol S62:150. Gibson VA, Stover SM, Gibeling JC, Hazelwood SJ, Martin LB. 2006. Osteonal effects on elastic modulus and fatigue life in equine bone. J Biomech 39:217 225. Gocha TP, Agne w AM. 2016. Spatial variation in osteon population density at the human femoral midshaft: histomorphometric adaptations to habitual load environment. J Anat 228:733 745.
176 Goh JCH, Lee EH, Ang EJ, Bayon P, Pho RWH. 1992. Biomechanical study of the load beari ng characteristics of the fibula and the effects of fibular resection. Clin Orthop Relat Res 279:223 228. Grzimek B. 1990. Grzimek's Encyclopedia of Mammals. New York: McGraw Hill. Guntur AR, Rosen CJ. 2012. Bone as an endocrine organ. Endocr Pract 18:758 762. Hadfield JD. 2010. MCMC methods for multi response generalised linear mixed models: The MCMCglmm R package. J Stat Softw 33:1 22. Hamrick WM, Ferrari SL. 2008. Leptin and the sympathetic connection of fat to bone. Osteoporos Int 19:905 912. Hanna JB, Polk JD, Schmitt D. 2006. Forelimb and hindlimb forces in walking and galloping. Am J Phys Anthropol 130:529 535. Harada S, Rodan GA. 2003. Control of osteoblast function and regulation of bone mass. Nature 423:349 355. Harvey PH, Pagel MD. 1991. The Compa rative Method in Evolutionary Biology. Oxford, UK: Oxford University Press. Heidelberger P, Welch PD. 1983. Simulation run length control in the presence of an initial transient. Oper Res 31:1109 1144. Hert J, Liskova M, Landa J. 1971. Reaction of bone to mechanical stimuli. Part 1. Continuous and intermittent loading of tibia in rabbit. Folia Morphol 19:290 300. Hert J, Liskova M, Landrgot B. 1969. Influence of the long term continuous bending on bone. Folia Morphol 17:389 399. Hert J, Pribylova E, Liskova M. 1972. Reaction of bone to mechanical stimuli. Part 3. Microstructure of compact bone of rabbit tibia after intermittent loading. Acta Anat 82:218 230. Hiller LP, Stover SM, Gibson VA, Gibeling JC, Prater CS, Hazelwood SJ, Yeh OC, Martin RB. 2003. Osteo n pullout in the equine third metacarpal bone: effects of ex vivo fatigue. J Orthop Res 21:481 488. Hosey GR. 2005. How does the zoo environment affect the behavior of captive primates? Appl Anim Behav Sci 90:107 129. Hsieh Y F, Robling AG, Ambrosius WT, Burr DB, Turner CH. 2001. Mechanical loading of diaphyseal bone in vivo: the strain threshold for an osteogenic response varies with location. J Bone Miner Res 16:2291 2297.
177 Hunt KD, Cant JGH, Gebo DL, Rose MD, Walker SE, Youlatos D. 1996. Standardized des criptions of primate locomotor and postural modes. Primates 37:363 387. Hylander WL, Johnson KR. 1997. In vivo bone strain patterns in the zygomatic arch of macaques and the significance of these patterns for functional interpretations of craniofacial form Am J Phys Anthropol 102:203 232. Jee WSS. 2005. The past, present, and future of bone morphometry: its contribution to an improved understanding of bone biology. J Bone Miner Res 23[Suppl]:1 10. Jee WSS, Ke HZ, Li XJ. 1991. Long term anabolic effects of PGE 2 on tibial diaphyseal bone in male rats. Bone Miner 15:33 55. Jee WSS, Mori S, Li XJ, Chan S. 1990. Prostaglandin E 2 enhances cortical bone mass and activates intracortical bone remodeling in intact and ovariectomied female rats. Bone 11:253 266. Jolly CJ. 1970. The seed eaters: A new model of Hominid differentiation based on a baboon analogy. Man 5:5 26. Kalu DN. 1991. The ovariectomied rat model of postmenopausal bone loss. Bone Miner 15:175 192. Kane EE, McGraw WS. 2017. Dietary variation in Diana mo nkeys ( Cercopithecus diana ): the effects of polyspecific associations. Folia Primatol 88:455 482. Karsenty G, Ferron M. 2012. The contribution of bone to whole organism physiology. Nature 481:314 320. Kay RF. 1985. Dental evidence for the diet of Australop ithecus Ann Rev Anthropol 14:315 341. Keenan KE, Mears CS, Skedros JG. 2017. Utitilty of osteon circularity for determining species and interpreting load history in primates and nonprimates. Am J Phys Anthropol 162:657 681. Kerley ER. 1965. The microscopi c determination of age in human bone. Am J Phys Anthropol 23:149 164. Kimura T. 1985. Bipedal and quadrupedal walking of primates: comparative dynamics. In: Kondo O, Ishida H, Kimura T, Okada M, Yamazaki N, Prost JH, editors. Primate Morphophysiology, Loco motor Analyses and Human Bipedalism. Tokyo: University of Tokyo Press. p 81 104. Kimura T, Okada M, Ishida H. 1979. Kinesiological characteristics of primate walking: its significance in human walking. In: Morbeck M, Preuschoft H, Gomberg N, editors. Envir onment, Behaviour, Morphology: Dynamic Interactions in Primates. New York: Gustav Fischer. p 297 311.
178 Komatsu DE, Brune KA, Liu H, Schmidt AL, Han B, Q ZQ, Yang X, Nunes JS, Lu Y, Geiser AG, Ma YL, Wolos JA, Westmore MS, Sato M. 2009. Longintudinal in vivo analysis of the region specific efficacy of parathyroid hormone in a rat cortical defect model. Endocrinology 150:1570 1579. Korstjens AH, Bergmann K, Deffernez C, Krebs M, Nijssen EC, van Oirschot BAM, Paukert C, Schippers EP. 2007. How small scale diffe rences in food competition lead to different social systems in three closely related sympatric colobines. In: McGraw WS, Zuberbhler K, No R, editors. Monkeys of the Ta Forest: An African Primate Community. Cambridge: Cambridge University Press. p 72 108 Korstjens AH, Sterck EHM, No R. 2002. How adaptive or phylogenetically inert is primate social behaviour? A test with two sympatric colobines. Behaviour 139:203 225. Kuhar CW, Fuller GA, Dennison PM. 2013. A survey of diabetes prevalence in zoo housed p rimates. Zoo Biol 32:63 69. Lad SE, Daegling DJ, McGraw WS. 2016. Bone remodeling is reduced in high stress regions of the cercopithecoid mandible. Am J Phys Anthropol 161:426 435. Lambert KL. 1971. The weight bearing function of the fibula. J Bone Joint S urg 53:507 513. Lanyon LE, Baggott DG. 1976. Mechanical function as an influence on the structure and form of bone. J Bone Joint Surg 58 B:436 443. Lanyon LE, Magee PT, Baggott DG. 1979. The relationship of functional stress and strain to the processes of bone remodelling. An experimental study on the sheep radius. J Biomech 12:593 600. Lee NK, Sowa H, Hinoi E, Ferron M, Ahn JD, Confavreux C, Dacquin R, Mee PJ, McKee MD, Jung DY, Zhang Z, Kim JK, Mauvais Jarvis F, Ducy P, Karsenty G. 2007. Endocrine regulat ion of energy metabolism by the skeleton. Cell 130:456 469. Leigh SR. 1994. Relations between captive and noncaptive weights in anthropoid primates. Zoo Biol 13:21 43. Lieberman DE, Pearson OM, Polk JD, Demes B, Crompton AW. 2003. Optimization of bone grow th and remodeling in repsonse to loading in tapered mammalian limbs. J Exp Biol 206:3125 3138. Lieberman DE, Polk JD, Demes B. 2004. Predicting long bone loading from cross sectional geometry. Am J Phys Anthropol 123:156 171.
179 Liskova M, Hert J. 1971. React ion of bone to mechanical stimuli. Part 2. Periosteal and endosteal reaction of tibial diaphysis in rabbit to intermittent loading. Folia Morphol 19:301 317. Lovejoy CO, McCollum MA, Reno PL, Rosenman BA. 2003. Developmental biology and human evolution. An nu Rev Anthropol 32:85 109. Lovejoy CO, Meindl RS, Ohman JC, Keiple KG, White TD. 2002. The Maka Femur and its bearing on the antiquity of human walking: applying contemporary concepts of morphogenesis to the human fossil record. Am J Phys Anthropol 119:97 133. Macchiarelli R, Bandioli L, Galichon V, Tobias PV. 1999. Hip bone trabecular architecture show uniquely distinctive locomotor behavior in South African australopithecines. J Hum Evol 36:211 232. Maggiano IS, Maggiano CM, Clement JG, Thomas CDL, Carte r Y, Cooper DML. 2016. Three dimensioanl reconstruction of Haversian systems in human cortical bone using synchrotron radiation based micro CT: morphology and quantification of branching and transverse connections across age. J Anat 228:719 732. Maggiano I S, Maggiano CM, Tiesler V, Kierdorf H, Stout SD, Schultz M. 2011. A distinct region of microarchitectural variation in femoral compact bone: histomorphoogy of the endosteal lamellar pocket. Int J Osteoarchaeol 21:743 750. Main RP. 2007. Ontogenetic relationships between in vivo strain environment, bone histomorphometry and growth in the goat radius. J Anat 210:272 293. Martin RB. 2000. Toward a unifying theory of bone remodeling. Bone 26:1 6. Martin RB. 2002. Is all cortical bone remodeling initiated by microdamage? Bone 30:8 13. Martin RB, Burr DB, Sharkey NA, Fyhrie DP. 2015. Skeletal Tissue Mechanics. New York: Springer. Martiniakov M, Grosskopf B, R O, Vondrkov M, Bauerov M. 2006. Differences among species in compact bone tissue microstructure of mammalian skeleton: use of a discriminant function analysis for species identification. J Forensic Sci 51:1235 1239. Mason MW, Skedros JG, Bloebaum RD. 1995. Evidence of strain mode related cortical adaptation in the d iaphysis of the horse radius. Bone 17:229 237. McFarlin SC, Terranova CJ, Zihlman AL, Enlow DH, Bromage TG. 2008. Regional variability in secondary remodeling within long bone cortices of catarrhine primates: the influence of bone growth history. J Anat 21 3:308 324.
180 McGraw WS. 1998a. Comparative locomotion and habitat use of six monkeys in the Tai Forest, Ivory Coast. Am J Phys Anthropol 105:493 510. McGraw WS. 1998b. Locomotion, support use, maintenance activities, and habitat structure: the case of the Ta Forest cercopithecids. In: Strasser E, Fleagle JG, Rosenberger A, McHenry H, editors. Primate Locomotion: Recent Advances. New York: Plenum Press. p 79 94. McGraw WS. 2007. Positional behavior and habitat use of Ta forest monkeys. In: McGraw WS, Zuberb hler K, No R, editors. Monkeys of the Ta Forest: An African Primate Community. Cambridge: Cambridge University Press. p 223 253. McGraw WS, Daegling DJ. 2009. Associations of gait, support use and limb morphology in West African colobines: the effects of bounding on locomotor biomechanics. Am J Phys Anthropol S48:188. McGraw WS, Vick AE, Daegling DJ. 2011. Sex and age differences in the diet and ingestive behaviors of sooty mangabeys ( Cercocebus atys ) in the Tai Forest, Ivory Coast. Am J Phys Anthropol 14 4:140 153. McKnab BK. 1986. The influence of food habits on the energetics of eutherian mammals. Ecol Monogr 56:1 19. Mittermeier RA, Fleagle JG. 1976. The locomotor and postural repertoires of Ateles geoffroyi and Colobus guereza and a reevaluation of th e locomotor category semibrachiation. Am J Phys Anthropol 45:235 256. Mohsin S, O'Brien FJ, Lee TC. 2006. Osteonal crack barriers in ovine compact bone. J Anat 208:81 89. Morbeck M. 1976. Leaping, bounding, and bipedalism in Colobus guereza : a spectrum of positional behavior. Yrbk Phys Anthropol 20:408 420. Mori S, Burr DB. 1993. Increased intracortical remodeling following fatigue damage. Bone 14:103 109. Mosley JR, Lanyon LE. 1998. Strain rate as a controlling influence on adaptive modeling in response to dynamic loading of the ulna in growing male rats. Bone 23:313 318. Moyle DD, Bowden RW. 1984. Fracture of human femoral bone. J Biomech 17:203 213. Moyle DD, Welborn JW, Cooke FW. 1978. Work to fracture of canine femoral bone. J Biomech 11:435 440.
181 Mulher n DM, Van Gerven DP. 1997. Patterns of femoral bone remodeling dynamics in a medieval Nubian population. Am J Phys Anthropol 104:133 146. Mullender MG, Huiskes R. 1995. A proposal for the regulatory mechanism of Wolff's Law. J Orthop Res 13:503 512. Mullin s LP, Sassi V, McHugh PE, Bruzzi MS. 2009. Differences in the crack resistant of interstitial, osteonal and trabecular bone tissue. Ann Biomed Eng 37:2574 2482. Mundy GR. 2006. Nutritional modulators of bone remodeling during aging. Am J Clin Nutr 83:427S 430S. Napier JR, Napier PH. 1967. Handbook of Living Primates. New York: Academic Press. Nowak RM. 1991. Walker's Mammals of the World, 5th ed. Baltimore: The Johns Hopkins University Press. O'Brien FJ, Hardiman DA, Hazenberg JG, Mercy MV, Mohsin S, Taylor D, Lee TC. 2005. The behaviour of microcracks in compact bone. Eur J Morphol 42:71 79. O'Connor JA, Lanyon LE, MacFie H. 1982. The influence of strain rate on adaptive bone remodeling. J Biomech 15:767 781. van Oers RF, Riumerman R, van Rietbergen B, Hilbers PAJ, Huiskes R. 2008. Relating osteon diameter to strain. Bone 43:476 482. Paine RR, Godfrey LR. 1997. The scaling of skeletal microanatomy in non human primates. J Zool Lond 241:803 821. Pampush JD, Duque AC, Burrows BR, Daegling DJ, Kenney WF, Mc Graw WS. 2013. Homoplasy and thick enamel in primates. J Hum Evol 64:216 224. Parfitt AM. 1994. Osteon and hemi osteonal remodeling: the spatial and temporal framework for signal traffic in adult human bone. J Cell Biochem 55:273 286. Parfitt AM. 2002. Mis conceptions (1): Epiphyseal fusion causes cessation of growth. Bone 30:337 339. Pattin CA, Caler WE, Carter DR. 1996. Cyclic mechanical property degradation during fatigue loading of cortical bone. J Biomech 29:69 79. Pearson OM, Lieberman DE. 2004. The ag ing of Wolff's "Law": ontogeny and responses to mechanical loading in cortical bone. Yearb Phys Anthropol 47:63 99. Pfeiffer S. 1998. Variability in osteon size in recent human populations. Am J Phys Anthropol 106:219 227.
182 Pfeiffer S, Crowder C, Harrington L, Brown M. 2006. Secondary osteons and Haversian canal dimensions as behavioral indicators. Am J Phys Anthropol 131:460 468. Pfeiffer S, Zehr MK. 1996. A morphological and histological study of the human humerus from Border Cave. J Hum Evol 31:49 59. Piekarski K. 1970. Fracture of bone. J Appl Phys 41:215 223. Polk JD, Demes B, Jungers WL, Biknevicius AR, Heinrich RE, Runestad JA. 2000. A comparison of primate, carnivoran, and rodent limb bone cross sectional properties: are primates really unique? J H um Evol 39:297 326. Pontzer H, Lieberman DE, Momin E, Devlin MJ, Polk JD, Hallgrmsson B, Cooper DML. 2006. Trabecular bone in the knee responds with high sensitivity to changes in load orientation. J Exp Biol 209:57 65. Pope MH, Murphy MC. 1974. Fracture energy of bone in a shear mode. Med Biol Eng 12:763 767. Preuschoft H. 1970. Functional anatomy of the lower extremity. In: Bourne GH, editor. The Chimpanzee. Basel: Karger. p 221 294. Rabey KN, Green DJ, Taylor AB, Begun DR, Richmond BG, McFarlin SC. 2015 Locomotor activity influences muscle architecture and bone growth but not mucle attachment site morphology. J Hum Evol 78:91 102. Reid IR, Cornish J, Baldock PA. 2006. Nutrition related peptides and bone homeostasis. J Bone Miner Res 21:495 500. Reilly D T, Burstein AH. 1974. The mechanical properties of cortical bone. J Bone Joint Surg Am 56A:1001 1021. Reilly DT, Burstein AH. 1975. The elastic and ultimate properties of compact bone tissue. J Biomech 8:393 405. Reilly DT, Burstein AH, Frankel VH. 1974. T he elastic modulus for bone. J Biomech 7:271 275. Reilly GC, Currey JD. 1999. The development of microcracking and failure in bone depends on the loading mode to which it is adapted. J Exp Biol 202:543 552. Reilly GC, Currey JD. 2000. The effects of damage and microcracking on the impact strength of bone. J Biomech 33:337 343. Richardson VCG. 2000. Diseases of Domestic Guinea Pigs, 2nd ed. Malden, MA: Blackwell Science.
183 Ripley S. 1967. The leaping of langurs: A problem in the study of locomotor adaptation. Am J Phys Anthropol 26:149 170. Robling AG. 2009. Is bone's response to mechanical signals dominated by muscle forces? Med Sci Sports Exerc 41:2044 2049. Robling AG, Stout S. 2000. Histomorphometry of human cortical bone: Applications to age estimation. In : Katzenberg MA, Saunders SR, editors. Biological Anthropology of the Human Skeleton. New York: Wiley Liss. p 187 213. Roesler H. 1981. Some historical remarks on the theory of cancellous bone structure (Wolff's law). In: Cowin SC, editor. Mechanical Prope rties of Bone. New York: American Society of Mechanical Engineers. p 27 42. Roesler H. 1987. The history of some fundamental concepts in bone biomechanics. J Biomech 20:1025 1034. Rosas A, Martinez Maza C. 2010. Bone remodeling of the Homo heidlebergensis mandible; the Atapuerca SH sample. J Hum Evol 58:127 137. Roux W. 1881. Der zuchtende Kampf der Teile, oder die "Teilauslese" im Organisms (Theorie der 'functionellen Anpassung'). Leipzig: Wilhelm Engelmann. Rubin CT, Lanyon LE. 1984a. Dynamic strain simi larity in vertebrates; an alternative to allometric limb bone scaling. J Theor Biol 107:321 327. Rubin CT, Lanyon LE. 1984b. Regulation of bone formation by applied dynamic loads. J Bone Joint Surg 66:397 402. Rubin CT, Lanyon LE. 1985. Regulation of bone mass by mechanical strain magnitude. Calcif Tissue Int 37:411 417. Rubin CT, Turner SA, Bain S, Mallinckrodt C, McLeod, K. 2001. Low mechanical signals strengthen long bones. Nature 412:603 604. Ruff CB, Holt B, Trinkaus E. 2006. Who's afraid of the big ba d Wolff?: "Wolff's Law" and bone functional adaptation. Am J Phys Anthropol 129:484 498. Ruff CB, McHenry HM, Thackeray JF. 1999. Cross sectional morphology of the SK 82 and 97 proximal femora. Am J Phys Anthropol 109. Ruth EB. 1953. Bone studies. II. An e xperimental study of Haversian type vascular canals. Am J Anat 93:429 455. Ryan TM, Ketcham RA. 2005. Angular orientation of trabecular bone in the femoral head and its relationship to hip joint loads in leaping primates. J Morphol 265.
184 Sato S, Hanada R, K imura A, Abe T, Matsumoto T, Iwasaki M, Inose H, Ida T, Mieda M, Takeuchi Y, Fukumoto S, Fujita T, Kato S, Kangawa K, Kojima M, Shinomiya K, Takeda S. 2007. Central control of bone remodeling by neuromedin U. Nat Med 13:1234 1240. Sawada J, Kondo O, Nara T Dodo Y, Akazawa T. 2004. Bone histomorphology of the Dederiyeh Neanderthal child. Anthropol Sci 112:247 256. Schaffler MB, Burr DB. 1984. Primate cortical bone microstructure: Relationship to locomotion. Am J Phys Anthropol 65:191 197. Schmelzer GH, Guri b Fakim A, editors. 2008. Plant Resources of Tropical Africa 11(1). Medicinal Plants 1. Leiden, Netherlands: Backhuys Publishers. Schmitt D. 1999. Compliant walking primates. J Zool 248:149 160. Schmitt D. 2003. Mediolateral reaction forces in forelimb anatomy in quadrupedal primates: implications for interpreting locomotor behavior in fossil primates. J Hum Evol 44:47 58. Schmitt D, Lemelin P. 2002. Origins of primate locomotion: gait mechanics of the wolly opossum. Am J Phys Anthropol 118:231 238. Schw itzer C, Kaummanns W. 2001. Body weights of ruffed lemurs ( Varecia variegata ) in European zoos with reference to the problem of obesity. Zoo Biol 20:261 269. Seref Ferlengez Z, Basta Pljakic J, Kennedy OD, Philemon CJ, Schaffler MB. 2014. Structural and me chanical repair of diffuse damage in cortical bone in vivo J Bone Miner Res 29:2537 2544. Setchell KDR, Lydeking Olsen E. 2003. Dietary phytoestrogens and their effect on bone: evidence from in vitro and in vivo, human observational, and dietary intervent ion studies. Am J Clin Nutr 78:593S 609S. Shapiro SS, Wilk MB. 1965. An analysis of variance test for normality (complete samples). Biometrika 52:591 611. Shipov A, Zaslansky P, Riesmeier H, Segev G, Atkins A, Shahar R. 2013. Unremodeled endochondral bone is a major architectural component of the cortical bone of the rat ( Rattus norvegicus ). J Struct Biol 183:132 140. Singh IJ, Tonna EA, P GC. 1974. A comparative histological study of mammalian bone. J Morphol 144:421 438. Skedros JG. 2012. Interpreting loa d history in limb bone diaphyses: important considerations and their biomechanical foundations. In: Crowder C, Stout S, editors. Bone Histology: An Anthropological Perspective. Boca Raton: CRC Press Taylor & Francis Group. p 153 220.
185 Skedros JG, Hunt KJ. 2 004. Does the degree of laminarity mediate site specific differences in collagen fiber orientation in primary bone? An evaluation in the turkey ulna diaphysis. J Anat 205:121 134. Skedros JG, Keenan KE, Williams TJ, Kiser CJ. 2013. Secondary osteon size an d collagen/llamelar organization ("osteon morphotypes") are not coupled, but potentially adapt independently for local strain mode or magnitude. J Struct Biol 181:95 107. Skedros JG, Kiser CJ, Keenan KE, Thomas SC. 2011. Analysis of osteon morphotype scori ng schemes for interpreting load history: evaluation in the chimpanzee femur. J Anat 218:480 499. Skedros JG, Mason MW, Bloebaum RD. 1994. Differences in osteonal micromorphology between tensile and compressive cortices of a bending skeletal system: Indica tions of potential strain specific differences in bone microstructure. Anat Rec 239:405 413. Skedros JG, Mason MW, Bloebaum RD. 2001. Modeling and remodeling in a developing artiodactyl calcaneous: A model for evaluating Frost's mechanostat hypothesis and its corollaries. Anat Rec 263:167 185. Skedros JG, Mason MW, Nelson MC, Bloebaum RD. 1996. Evidence of structural and material adaptation to specific strain features in cortical bone. Anat Rec 246:47 63. Skedros JG, Mendenhall SD, Kiser CJ, Winet H. 2009. Interpreting cortical bone adaptation and load history by quantifying osteon morphotypes in circularly polarized light images. Bone 44:392 403. Skedros JG, Su SC, Bloebaum RD. 1997. Biomechanical implications of mineral content and microstructural variatio ns in cortical bone of horse, elk, and sheep calcanei. Anat Rec 249:297 316. Stock J, Pfeiffer S. 2001. Linking structural variability in long bone diaphyses to habitual behaviors: foragers from the southern African Later Stone Age and the Andaman Islands. Am J Phys Anthropol 115:337 348. Stout S, Crowder C. 2012. Bone remodeling, histomorphology, and histomorphometry. In: Crowder C, Stout S, editors. Bone Histology: An Anthropological Perspective. Boca Raton: Taylor & Francis Group. p 1 21. Stout S, Paine RR. 1992. Histological age estimation using rib and clavicle. Am J Phys Anthropol 87:111 115.
186 Streeter M, Stout S, Trinkaus E, Burr D. 2010. Brief communication: Bone remodeling rates in Pleistocene humans are not slower than the rates observed in modern p opulations: A reexamination of Abbott et al. (1996). Am J Phys Anthropol 141:315 318. Su SC, Skedros JG, Bachus KN, Bloebaum RD. 1999. Loading conditions and cortical bone construction of an artiodactyl calcaneus. J Exp Biol 202:3239 3254. Szivek JA, Johns on EM, Magee FP. 1992. In vivo strain analysis of the greyhound femoral diaphysis. J Invest Surg 5:91 108. Terranova CJ. 1995. Leaping behaviors and the functional morphology of strepsirhine primate long bones. Folia Primatol 65:181 201. Thomas T, Burguera B. 2002. Is leptin the link between fat and bone mass? J Bone Miner Res 17:1563 1569. Thompson DW, Bonner JT. 1961. On Growth and Form. Cambridge: Cambridge University Press. Tidire M, Gaillard J M, Berger V, Mller DWH, Lackey LB, Gimenez O, Clauss M, L ematre J F. 2016. Comparative analyses of longevity and senescence reveal variable survival benefits of living in zoos across mammals. Sci Rep 6:36361. Trinkaus E, Ruff CB. 1999a. Diaphyseal cross sectional geometry of Near Eastern Middle Paleolithic huma ns: the femur. J Archaeol Sci 26:409 424. Trinkaus E, Ruff CB. 1999b. Diaphyseal cross sectional geometry of Near Eastern Middle Paleolithic humans: the tibia. J Archaeol Sci 26:1289 1300. Tsegai ZJ, Kivell TL, Gross T, Nguyen NH, Pahr DH, Smaers JB, Skinn er MM. 2013. Trabelcular bone structure correlates with hand posture and use in hominoids. PLoS One 8:e78781. Turner CH. 1999. Toward a mathematical description of bone biology: the principal of cellular accomodation. Calcif Tissue Int 65:466 471. Turner Ch, Owan I, Takano Y. 1995. Mechanotransduction in bone: Do bone cells act as sensors of fluid flow? FASEB J 8:875 878. Turner CH, Pavalko FM. 1998. Mechanotransduction and functional response of the skeleton to physical stress: the mechanisms and mechanic s of bone adaptation. J Orthop Sci 3. Umemura Y, Ishiko T, Yamauchi T, Kurono M, Mashiko S. 1997. Five jumps per day increase bone mass and breaking force in rats. J Bone Miner Res 12:1480 1485.
187 Ungar PS, Grine FE, Teaford MF. 2008. Dental microwear and di et of the Plio Pleistocene hominin Paranthropus boisei PLoS One 3:e2044. Ungar PS, Kay RF. 1995. The dietary adaptations of European Miocene catarrhines. Proc Natl Acad Sci USA 92:5479 5481. Wachter B, Schabel M, No R. 1997. Diet overlap and polyspecific associations of red colobus and Diana monkeys in the Ta National Park, Ivory Coast. Ethology 103:514 526. Wallace IJ, Judex S, Demes B. 2015. Effects of load bearing exercise on skeletal structure and mechanics differ between outbred populations of mice. Bone 72:1 8. Wallace IJ, Tommasini SM, Judex S, Garland TJ, Demes B. 2012. Genetic variations and physical activity as determinants of limb bone morphology: an experiemental approach using a mouse model. Am J Phys Anthropol 148:24 35. Wang Q, Whittle M, Cunningham J, Kenwright J. 1996. Fibula and its ligaments in load transmission and ankle joint stability. Clin Orthop Relat Res 330:261 270. Ward FO. 1838. Outlines of Human Osteology. London. Warshaw J. 2008. Comparative primate bone microstructure: records of life history, function, and phylogeny. In: Sargis EJ, Dagosto M, editors. Mammalian Evolutionary Morphology: A Tribute to Frederick S. Szalay. The Netherlands: Springer. Weigl R. 2005. Longevity of Mammals in Captivity; From the Living Collectio ns of the World. Stuttgart: Schweizerbart. Wolff J. 1892. The Law of Bone Remodeling (Das Gesertz der Transformation der Knochen). (Translated by P. Maquet and R. Furlong). Berlin: Springer Verlag. Wu YY, Xiao E, Graves DT. 2015. Diabetes mellitus related bone metabolism and periodontal disease. Int J Oral Sci 7:63 72. Zhang Q, Carlson J, Ke HZ, Li J, Kim M, Murphy K, Mehta N, Gilligan J, Vignery A. 2010. Dramatic increase in cortical thickness induced by femoral marrow ablation followed by a 3 month treatm ent with PTH in rats. J Bone Miner Res 25:1350 1359.
188 BIOGRAPHICAL SKETCH Susan Lad re ceived her Bachelor of Arts in a nthropology at The Ohio State University in 2012 She then came to the University of Florida to study biological a nthropology under the mentorship of Dr. David Daegling. S he completed her Master of Arts in 2014 and continued on to earn her PhD in 2018 At UF Susan has worked as a research assistant in Dr. ist ant and lecturer for courses, including Introduction to Biological Anthropology and The Primat es Her work has been published in the American Journal of Physical Anthropology, The Anatomical Record, and The American Journal of Primatology, and s he has r outinely presented at the American Association of Physical Anthropologists meetings. She will at the University of Notre Dame as a postdoc toral researcher starting late f all 2018.