Skeletal Growth and the Changing Genetic Landscape during Childhood and Adulthood

Abstract

Growth, development, and decline of the human skeleton are of central importance to physical anthropology. All processes of skeletal growth (longitudinal growth as well as gains and losses of bone mass) are subject to environmental and genetic influences. These influences, and their relative contributions to the phenotype, can be asserted at any stage of life. We present here the gross phenotypic and genetic landscapes of four skeletal traits, and show how they vary across the life span. Phenotypic sex differences are found in bone diameter and cortical index (a ratio of cortical thickness over bone diameter) at a very early age and continue throughout most of life. Sexual dimorphism in summed cortical thickness and bone length, however, is not evident until shortly after the pubertal growth spurt. Genetic contributions (heritability) to these skeletal phenotypes are generally moderate to high. Bone length and bone diameter (which both scale with body size) tend to have the highest heritability, with heritability of bone length fairly stable across ages (with a notable dip in early childhood) and that of bone diameter peaking in early childhood. The bone traits summed cortical thickness and cortical index, that may better reflect bone mass, a more plastic phenomenon, have slightly lower genetic influences, on average. Results from our phenotypic and genetic landscapes serve three key purposes: demonstration of the integrated nature of the genetic and environmental underpinnings of skeletal form; identification of periods of bone’s relative sensitivity to genetic and environmental influences; and stimulation of hypotheses predicting the effects of exposure to environmental variables on the skeleton, given variation in the underlying genetic architecture.

The concept that bone changes over the life course is one of the fundamental paradigms in physical anthropology. Articles dealing with age-related skeletal changes appear early in the history of the AJPA, including Todd’s series on changes in the pubic bone (Todd, 1920; Todd, 1921) and Schultz’s classic paper on fetal growth (Schultz, 1923). The first AAPA meeting in 1930 also had several notable presentations on bone development, including Todd’s paper on the adolescent lag in skeletal growth (Todd, 1930). The study of the skeleton and skeletal growth was facilitated beginning in the early years of the 20^th century by the creation of large scale skeletal collections such as the Hamann-Todd and Terry Collections (now housed at the Cleveland Museum of Natural History and National Museum of Natural History, respectively), which allowed for cross-sectional studies of detailed anatomy. At the same time, radiographic collections formed the core of many growth studies that focused on longitudinal characterization of bone properties in living subjects. One important figure in physical anthropology in terms of radiographic measures of bone growth was Stanley Garn (Garn, 1981). Garn, an influential researcher at the Fels Research Institute from 1952 to 1968, characterized the gains and losses of cortical bone in an early sample of subjects of the Fels Longitudinal Study (Garn et al., 1968; Garn, 1970a).

As in the 20^th century, understanding the complex nature of the composition, modeling, remodeling, and structure of bone remains a lifelong research focus of scientists in a variety of fields. In addition to the study of physical bone and radiographic images available to earlier researchers, the advent of genetic research has expanded the tools available to modern researchers of bone growth and development. This is important because skeletal phenotypes result from a balance between genetic and environmental influences. Components of the skeleton are not wholly controlled by one or the other, but rather represent environmental adaptations that play against a backdrop of genetic constraints. The shape and content of bone will change throughout the life span with fluctuating contributions of these overarching influences. Integrating the study of genetic and environmental components of skeletal form provides insight into the driving forces underlying the observed morphology that serves as the basis for many core inquiries within physical anthropology.

It is well accepted that a bone’s form and content change over the course of life, with a review of the relevant literature appearing recently in the Yearbook of Physical Anthropology (Gosman et al., 2011). Briefly, in humans, long bones lengthen with age from birth to ~18 years (or until skeletal maturity), with fluctuations in growth velocity occurring at various periods therein. Increases in total bone diameter and cortical thickness occur from birth to approximately 30 years of age (Garn, 1970a; Magarey et al., 1999; Baxter-Jones et al., 2011). Bone mass (bone mineral density and bone mineral content) also changes over the life course, with mineral content peaking in the third decade of life and drastically decreasing in women after menopause (Bonjour et al., 1994; Matkovic et al., 1994; Heaney et al., 2000). Additional changes in bone can be seen throughout life with adaptations in shape and content to the surrounding mechanical environment (Trinkaus et al., 1994; Ruff, 2003; Auerbach and Ruff, 2006).

During any of these life stages (e.g., childhood, puberty, skeletal maturity, menopause), both environmental and genetic influences interact to ultimately produce measureable bone phenotypes. The influence of the environment on the human skeleton is multifactorial, with bones receiving input from external factors such as diet, physical activity, body composition, and others. These factors change with age both in their presentation and their potential influence on the skeleton. Likewise, there is a growing body of evidence that the genetic influence on bone also changes over the life course. The mechanisms for changes in genetic influence can include developmental timing of gene action, epigenetic modification of the bone-related gene product, or even variation in environmental factors influencing gene expression.

One of the most obvious examples of this age-dependence is the genetic susceptibility to certain bone disorders, such as osteoporosis. The evidence for genetic influence on osteoporosis is substantial, with heritabilities as high as 90% (Tse et al., 2009). Genome-wide linkage and association studies have identified several candidate genes, with significant progress in identification of gene variants influencing osteoporosis susceptibility (Richards et al., 2008; Cheung et al., 2010; Ermakov et al., 2011). Although the same genetic code is present throughout life, its effect on bone only appears in later adulthood. Importantly, having “the gene” for osteoporosis is not an inevitable sentence for weak or brittle bones, but rather causes a predisposition to such conditions. Our own work has successfully identified genomic regions influencing bone at specific childhood ages (Duren et al., 2011), with some regions containing attractive candidate genes such as the oxytocin receptor gene that has implications for bone gain and loss throughout life given the receptor’s binding affinity to estrogen. As mentioned above, environmental and functional factors play a large role in the resultant bone morphology. Adequate calcium intake or regular weight-bearing physical activity, for example, could mediate the extent of age-related bone loss, thus thwarting the genetic propensity for osteoporosis.

It is clear that identifying the relative influences of genes and environment at different stages of life is necessary to better understand the underlying causes of observed phenotypic variation. In this paper, we provide the first presentation of changes in the magnitude of genetic influences on bone morphological phenotypes over the course of the entire life span. Although the genetic underpinnings of skeletal variation are our focus, we first document the changes that occur in skeletal phenotypes themselves across life. This produces a phenotypic landscape (a virtual topography created when trait values are viewed graphically across an extended timeframe) that characterizes basic growth and development parameters of the skeleton, and provides a comprehensive view of variation within and across life stages. Upon this phenotypic landscape, we then superimpose genetic data to characterize the relative influence of genes on variation in bone phenotypes continuously across life history stages. The resultant genetic landscape serves three key purposes. First, it demonstrates the integrated nature of the genetic and environmental underpinnings of skeletal form. Second, it identifies potential periods of relative sensitivity of bone to genetic and environmental influences. Third, it stimulates the formulation of hypotheses predicting the phenotypic effects of differential exposure to environmental variables (in terms of both timing and degree), given variation in the underlying genetic architecture.

MATERIALS AND METHODS

Study sample

The study sample is comprised of individuals participating in the Fels Longitudinal Study, the world’s longest-running longitudinal study of human growth, development, and body composition change over the life span (Roche, 1992). Initiated in 1929, the Fels Longitudinal Study followed, and continues to follow, participants from birth (when possible) and throughout their lives. Subjects are examined at our research facility at regular intervals (more frequently during childhood than in adulthood). Subjects are primarily of European ancestry and live (or lived at the time of evaluation) in southwest Ohio or immediately surrounding regions. All subjects represented in this sample are considered normal and healthy and were not selected for any disease or bone related trait. We examined hand-wrist radiographs of these individuals taken between 1.5 years and 93 years of age, and between the years 1930 and 2003. Males and females were analyzed together in all genetic analyses, but we tested for sex differences in our phenotypic analyses and tested for sex effects in our genetic models.

Age groups

Subjects were divided into discrete age groups for the present analysis. Age groups contain between 220 and 750 subjects and represent 17 one-year increments from 2 years to 18 years, 7 five-year increments from 20to 55 years, a single ten-year increment between 55 and 65 years of age, and one group of all subjects 65 years of age and older (note: there was an inadequate number of subjects at age 19 to include in genetic analyses due to the Fels scheduling protocol). Each age group contains a single time-point per individual, and the longitudinal nature of the sample places many of our subjects into multiple age groups. For example, a subject who is seen on a regular basis during childhood may be included in every childhood age group presented in this analysis. Likewise, if a subject is seen on our regular schedule as an adult, they may be included in each of the adulthood age groups presented here. Conversely, some individuals, such as subjects who marry into the families, are only included in the adult age groups.

Family relationships

Our sample contains 1,761 individuals from 235 nuclear and extended families. Contained within these families, there are 7 identical sibling pairs; 1,868 parent-offspring pairs; 1,070 sibling pairs; 1,110 grandparent-grandchild pairs; and 1,638 avuncular pairs, as well as 7,360 other relative pairings, lending the study considerable power for the genetic analyses herein.

Skeletal phenotypes

Quantitative measures of bone size and shape were obtained from the second metacarpal from antero-posterior radiographs of the left hand. The second metacarpal has been widely used in biomedical and anthropological investigations (Garn, 1970a; Mays, 2001; Lazenby, 2002; Duren et al., 2007; Lazenby et al., 2008) as it is the most cylindrical bone in the hand, thus it is often assumed that bone is evenly distributed about its axis (Garn, 1970a; and Lazenby, 1997; for an alternative view see Lazenby, 1998; Duren et al., 2007). Radiographs were digitized using a Vidar DosimetryPRO Advantage Radiographic Scanner, and analyzed with a dedicated image processing program written for this purpose in MatLab v.7.7.0.471 (MATLAB, 2008). The program first determined maximum length (diaphysis + epiphysis) of the metacarpal along the long axis of the bone, then determined external (periosteal) bone diameter and medial and lateral cortical thicknesses and cortical index (summed cortical thickness divided by bone diameter) perpendicular to the long axis at midshaft of the maximum bone length. All measurements were recorded to the 0.01 millimeter. Each radiograph was reviewed for accuracy of measurement placement and adjusted, if necessary, before the data were entered into a database. The radiographs were reviewed by two assessors, and a subset was reviewed by both assessors to assess inter-observer error. Inter-observer repeatability, expressed as Lin’s concordance correlation coefficient (Lin, 1989), ranged from 0.93 to 0.99, with the mean measurement differences between 0.008mm and 0.023mm (N=256). Intra-observer repeatability ranged between 0.95 and 0.99, with the mean differences between 0.001mm and 0.04mm (N1=261, N2=248).

Statistical analyses

Skeletal phenotypes were analyzed separately in each age group. Within-group sex-specific means and standard deviations were calculated for second metacarpal bone length, bone diameter (midshaft diameter), summed cortical thickness (sum of medial and lateral cortical widths), and the cortical index (summed cortical thickness/bone diameter). The significance of sex differences was determined using Student’s T-test and a nominal alpha of 0.05.

Quantitative genetic methods were used to estimate the residual heritability of second metacarpal cortical bone traits. Narrow sense heritability, defined as the proportion of phenotypic variance attributable to additive genetic effects, was estimated using a maximum likelihood-based variance components method that uses all familial information for parameter estimation (Lange and Boehnke, 1983) as implemented in Sequential Oligogenic Linkage Analyses Routines (SOLAR version 2.1.2) (Almasy and Blangero, 1998). In this approach, the covariance matrix for a pedigree is given by Ω = 2Φσ²_a + I_nσ²_e, where Φ is the kinship matrix for the pedigree, σ²_a is the trait variance due to additive genetic effects, I_n is the n x n identity matrix, and σ²_e is the trait variance due to random, individual-specific, environmental effects. The basic parameters estimated include the trait mean, standard deviation, and h², where h² = σ²_a/(σ²_a+ σ²_e). The null hypothesis is that σ_a² equals zero. Significance was tested by comparing the likelihoods of restricted models in which σ_a² was set to zero to those of general models in which σ_a² was estimated (Edwards, 1992). Variables examined for covariate effects included age (or skeletal age, a measure of bone development assessed using the FELS Method (Roche et al., 1988), in the under-20 age groups), sex, age-by-sex interaction, stature, weight, and a measure of radiographic quality (relative pixel brightness). To estimate the effects of these various covariates on the study traits, each covariate was screened in SOLAR by comparing the likelihood for a model that included the covariate to the likelihood for a model that excluded the covariate. Only covariates significant at the p < 0.1 level were retained in the final model for each second metacarpal cortical bone trait.

RESULTS

Phenotypic landscape

Phenotypic changes in second metacarpal morphology across the life span are presented in Figure 1. Descriptions of gross age-related changes in each phenotype are described below.

Figure 1.

Phenotypic landscape of second metacarpal a) bone length, b) bone diameter, c) summed cortical thickness, and d) cortical index across the life span. Circles indicate means, whiskers indicate standard deviation. Black circles are male, white circles are female.

Growth in length of the second metacarpal is displayed in Figure 1a. At age two years the second metacarpal is approximately 30 mm in length, and increases by about 2.5 mm per year until after puberty, when growth in length slows. Adult length of the second metacarpal averages 70.5mm in men and 64.3mm in women. The average length of the second metacarpal is almost identical for boys and girls throughout the childhood years (ages 2 to 10 years), but at ages 11, 12, and 13 years girls have slightly longer metacarpals than boys (but not significantly so; p>0.05). By 14 years of age, boys’ metacarpals are significantly longer than girls’ (p<0.05), and this dimorphism persists into adulthood, where men have on average 9.6% longer metacarpals than women.

The diameter of the second metacarpal, unlike its length, is sexually dimorphic by age 2 years (p<0.05), with boys having significantly wider bones than girls by approximately 5%. The relative difference between the sexes remains constant throughout the childhood years (0–13 years of age). Around 14 years of age the sex difference increases and persists throughout adulthood, with men having approximately 9% wider bones than women.

The average total amount of cortex distributed about the midshaft of the second metacarpal is almost identical in girls and boys from age 2 to 10 years, after which girls have more (i.e., on the order of 5–10% thicker) cortex for a short time (11–15 years of age). From age 16, boys have significantly more cortical bone in the second metacarpal than girls (p<0.05), and this dimorphism persists throughout adulthood, with men having 9% thicker cortex than women. There are modest fluctuations in the total amount of cortex in the adult years, and both sexes begin to lose cortical bone after age 50.

The cortical index, a ratio of summed cortical thickness and bone diameter, shows sexual dimorphism by age 2 years, with girls having a higher cortical index than boys (0.36 versus 0.32, respectively; p<0.05). This dimorphism persists through adulthood, peaking between ages 30 and 35, when women have 12% higher cortical index than men. This trend reverses after age 55, where women have lower cortical indices than men. Both sexes increase in cortical index until well into adulthood, with minor plateaus occurring at various points (e.g., mid-childhood, mid-adulthood). At age 45 for men and 40 for women, cortical index begins to decrease rather rapidly, ending with indices equivalent to those seen at age 15 in males, and at ages 10 to 11 in females.

Genetic landscape

Heritabilities for each phenotype at each age group are presented in Figure 2 (and in Supplementary Table 1). The percent of trait variance due to model covariates is also given. The following paragraphs describe the genetic landscape, defined as changes in heritability over time, for the skeletal phenotypes bone length, bone diameter, summed cortical thickness, and cortical index.

Figure 2.

Genetic landscape of second metacarpal a) bone length, b) bone diameter, c) summed cortical thickness, and d) cortical index across the life span. Bars indicate heritability estimates, lines indicate the percent of trait variance due to measured covariates. Gap in x-axis at 16 in bone length and 25 in total cortex is due to convergence failure during genetic analysis.

The heritability of second metacarpal length changes over the life course, and ranges in our sample from 0.61 to 1.0. At 2 years of age the residual heritability is 1.0, or 100%. That is, after removing the variance due to measured covariates (sex and stature), the remaining phenotypic trait variance is entirely due to the genetic relatedness of the sample. In the early childhood years (3–6 years), heritabilities range from 0.61 to 0.73, and the genetic contribution to bone length remains high and is relatively consistent throughout the remainder of childhood and into young adulthood (average heritability of 0.92 from age 7 to 29). In later adulthood, the genetic influence on metacarpal length again fluctuates, with heritabilities ranging between 0.64 and 0.88. The percentage of trait variance due to measured covariates remains fairly constant at approximately 70% throughout the life course, although early childhood values start out lower (48%).

The expression of basic genetic underpinnings of second metacarpal midshaft diameter also fluctuates over the life span, and heritabilities range from 0.35 to 0.98 with values highest during childhood. At age 2 years, heritability is 0.52, and thereafter mean heritability during childhood (3–18 years) is 0.74. Adulthood heritabilities for midshaft diameter are lower, with an average value of 0.51. The percentage of trait variance due to covariates begins low during childhood (on the order of 0.24) and rises during late childhood/adolescence, remaining steady at approximately 0.61 throughout adulthood.

The heritability of the total amount of cortical bone at midshaft of the second metacarpal ranges from 0.23 to 0.73 with noteworthy fluctuations throughout life. Childhood heritabilities are generally high (on the order of 0.50), but substantially lower at ages 13 and 14 years. The lowest heritability occurs in the 30-year age group (0.23) and, ignoring the spike of 0.62 in the 35-year age group, gradually increases to its highest value of 0.73 in the 55-year age group. Percent of trait variance due to measured covariates is low in early childhood (~15% between 2 and 10 years of age), rising to 36% at age 13. Variance due to covariates averages 22% for most of adulthood, rising to 43–45% after age 55.

The genetic landscape of the cortical index fluctuates, with heritabilities ranging from 0.24 to 0.78, with an average of 0.59. Heritability is lowest at age 2 years, and fluctuates between 0.33 and 0.77 during adulthood. Percent of trait variance due to covariates is also generally low: it is at its lowest during early childhood, peaks at 13 years of age, and averages 17% across all ages.

DISCUSSION

The current paper pays homage to, and extends, the classic works of Stanley Garn on the lifelong gain and loss of bone mass. Garn described an increase in second metacarpal cortical thickness during childhood up to approximately 25 years of age in an early sample from the Fels Longitudinal Study (Garn et al., 1969; Garn, 1970a; Garn, 1970b; Garn, 1972). He noted that cortical thickness was greater in men than women at all ages, whereas decreases observed after age 45 were less rapid in men. In addition to his work with Fels data, Garn also collected metacarpal bone data on subjects in Central America (Garn et al., 1969; Garn, 1970b). Using these data, he was able to study the effects of environmental factors such as nutrition on the accrual of bone tissue.

The phenotypic landscape

Our current findings on the phenotypic landscape of bone morphology augment those of Garn and other researchers who have presented longitudinal changes in the growing and aging skeleton (Frisancho et al., 1970; Garn et al., 1972; Garn and Clark, 1975; Bailey et al., 1999; Hui et al., 1999; Specker et al., 1999; Nguyen et al., 2001; Ruff, 2003). Our central phenotypic analysis focused on general patterns of change observed in second metacarpal properties from birth to old age. Like previous studies, we identified differences between the sexes in both the gain and loss of cortical bone, and the well-established distinction between male and female bone lengths, which arise around the time of puberty and persist throughout life. We identified a sex difference in several bone measures even in our youngest age group (2-year-olds) within which bone width is greater in males than in females, similar to other recent studies (Ruff, 2003; Pandey et al., 2009). During early childhood years, girls have roughly equal or slightly (but not significantly) greater cortical thickness than boys. For a short time in early adolescence, females have thicker cortices than males, but after age 15 and throughout adulthood the reverse is true, with males consistently exhibiting greater cortical thickness than females. Although females have lower absolute cortical thickness than males, because they also have a much smaller bone diameter females have a greater proportion of cortical bone (higher cortical index). In the adult age groups, females maintain higher cortical indices than males up until roughly the time of menopause (the 55-year age group), after which females have a lower cortical index.

Compared to earlier works (Frisancho et al., 1970; Garn et al., 1972; Garn and Clark, 1975; Bailey et al., 1999; Hui et al., 1999; Specker et al., 1999; Nguyen et al., 2001; Ruff, 2003), the current findings are based on a large sample of longitudinal observations spanning early childhood to late adulthood. It is reassuring that, where sample ages overlap, our findings are in general agreement previous works using similar phenotypes. Given the four decades since Garn’s original work with the Fels Longitudinal Study we now have substantially increased the sample, especially in older ages, and some of our results differ slightly from that earlier work. For example, we note a more rapid decrease in summed cortical thickness after age 50 in men, where Garn noted a more gradual decline.

The genetic landscape

The genetic underpinnings of cortical bone thickness were addressed by Garn and colleagues (Garn et al., 1967), who noted familial (parent-offspring and sibling pair) correlations in bone width and cortical thickness, and hinted at a possible X-linked inheritance. Advances in statistical genetics over the past decade have expanded the scope of anthropological inquiry to include examination of the determinants of skeletal morphology, growth and development. Recent work by anthropological geneticists has focused on craniofacial and dental traits (Hlusko et al., 2002; Hlusko and Mahaney, 2003; Roseman et al., 2010; Sherwood et al., 2011), and non-human primate bone phenotypes (Cheverud and Buikstra, 1981; Kohn et al., 2001; Havill, 2003; Havill et al., 2005; Havill et al., 2010). This work has contributed to our knowledge of the genetic underpinnings of skeletal growth and development, as well as the evolutionary implications of such foundations.

Genetic studies of skeletal health have demonstrated bone measures to be under substantial genetic influence (Christian et al., 1989; Gueguen et al., 1995; Kammerer et al., 1995; Hopper et al., 1998a; Peacock et al., 2002; Livshits et al., 2003; Wynne et al., 2003; Econs et al., 2004; Duren et al., 2004; Duren et al., 2007), with heritability estimates for bone mass from nuclear families between 64% and 84% (Gueguen et al., 1995; Wynne et al., 2003). Several studies of non-human primates have shown similar results, with 40 to 70% of the phenotypic variance in bone mass attributed to genetic variation (Kammerer et al., 1995; Lipkin et al., 2001).

Like age-related changes in our basic skeletal phenotypes, we demonstrated in this paper that the influence of genes on these traits also varies with age. The heritability of bone length, although it exhibits some notable fluctuations in magnitude, is generally stable across ages. The most notable deviation occurs in the early childhood ages, where, although it begins at very high at age 2 years, it is low afterward relative to the older age groups, then increases well before puberty. Other notable dips in heritability estimates occur in the 40- and 55-year age groups. It is unlikely that this fluctuation is due to a true genetic or environmental perturbation, but instead it is more likely that the composition of this age group may be influencing these results. Our measure of bone robusticity (bone diameter) is generally under greatest genetic influence in the early childhood ages, and falls during adulthood (with a slight increase at 45–50 years). The variance due to covariates is low during childhood, increases rapidly in the teen years, when one might predict that functional (activity-related) differences between individuals might also increase, and remains high during adulthood. Sex is exerting an influence here in that, at about the age that we observe the increase in variance due to covariates, sexual dimorphism becomes more pronounced. Traits closely associated with bone mass are expected to fluctuate phenotypically over the life span and may be controlled by different genes at different ages, as illustrated by our osteoporosis example in the introduction. These traits (summed cortical thickness and cortical index) in our study exhibit striking variation in genetic contribution featuring greater fluctuations in magnitude from childhood to senescence. Both measures show an increase in the percent of variance due to covariates starting around the time of puberty, thereafter decreasing until late in life. It is interesting to note that in the later stages of life the variance in summed cortical thickness due to covariates increases, whereas the variance in cortical index due to covariates decreases (except in the oldest age group).

Until now, the nature of the changing landscape of heritability of bone traits over much of the life span in humans was largely unknown, mainly because estimated values from various published studies differed on the basis of factors such as bone site or sex studied. Comparisons of heritabilities of bone mass in younger and older twin pairs have suggested that bone mass may be under tighter genetic control in early life than in later life (Smith et al., 1973; Slemenda et al., 1991; Rubin et al., 2000). Similarly, from residual heritability of a number of other early childhood traits such as birth weight, recumbent length and basicranial skeletal dimensions, it would also be expected that the heritability of bone mass at birth and during early childhood is high (Wilson, 1976; Fischbein and Nordqvist, 1978; Towne et al., 1993; Clausson et al., 2000; Sherwood et al., 2004). Results of our analyses are congruent with these expectations, as we found heritabilities of our skeletal phenotypes to be generally higher in early childhood than later childhood or adulthood, when individual variation in the mechanical environment may be higher.

Strong genetic effects on bone mass are also evident post-puberty, and bone mass heritability is said to reach its peak (h² ≈ 0.84) at the time of peak bone mass (Gueguen et al., 1995; Shimizu et al., 1999; Heaney et al., 2000). Subsequent to this peak, genetic influences on bone mass steadily decrease (Heaney et al., 2000). Some researchers have shown that the genetic influence on bone mass in girls may reach its highest childhood value just prior to puberty when environmental influences may be very strong (Hopper et al., 1998b), again consistent with our findings. Puberty in girls, signaled by menarche, is associated with a rapid increase in circulating estrogen. Because estrogen inhibits periosteal expansion of bone (Seeman, 2003), bone diameter at or just after menarche may be the widest the bone will ever be, future remodeling notwithstanding. After cessation of periosteal expansion (in either sex), the majority of bone accrual will occur at the endosteal (marrow-adjacent) surface. The relative plasticity of endosteal remodeling activity, which affects summed cortical thickness and cortical index, is likely to underpin the greater fluctuations we see in heritability of these traits after puberty.

Although estimates of heritability serve an important function in describing the relative contribution of genes to trait variation, it should be noted that heritability is a population-specific parameter. It is, thus, a tenuous assumption that a trait with high heritability in one population will have the same magnitude in a different population. Results from our study, therefore, pertain specifically to the context of our primarily European-descended, middle-class, Mid-Western United States study sample, although the same criticism of heritability could potentially be made of a cross-sectional multi-generational study. The current study, however, is unique, and minimizes many of these issues by examining heritability of the same bone traits in a single group of individuals through their entire life course. The influence of the age-stratified nature of our analysis is evident in changing values of heritability even for the stable adulthood phenotype, bone length. If each age group truly contained the identical set of individuals as all other age groups, then a trait such as bone length, which does not change phenotypically after childhood, would exhibit a stable heritability estimate over the adult life span. But, because of the composition of even this mixed longitudinal sample, each age category is not represented by the identical set of individuals, albeit each containing large overlap of subjects. In our sample, approximately 20% of persons contained in the oldest age groups (starting primarily at age 45) are also represented in the younger groups. This is in stark contrast to the composition of the younger age groups, where groups share greater than 60% of subjects. The shift in sample composition at age 45 is primarily driven by “marry-in” subjects for whom we do not have childhood data, or subjects who are still young (i.e., they have not yet reached the ages to be included in the adult groups). It is important, also, to acknowledge the concomitant changes in environmental factors as a person ages, which can also alter the measure of heritability.

Results presented are also potentially impacted by possible cohort effects. We acknowledge the limitations of our study design with samples containing multiple generations in each age grouping. The age stratification used here, while lending us a life-span view of bone traits and allowing for a mixed longitudinal approach to changes in genetic influences, may mask environmental variation experienced by the generation extremes contained in the sample. Any inter-generational variation will serve to either reduce heritability estimates or increase the standard error of those estimates. Thus, results presented here can be viewed as conservative estimates of the genetic influence on skeletal morphology.

The genetic landscape presented here is based on narrow-sense heritability, which in a way reflects both genetic and environmental influences on our traits. For example, in areas of the landscape where the heritabilities appear lowest, environmental effects may be more influential on variation in skeletal form. The meaning of “environment” in this context encompasses any non-genetic factor. Investigators attempt to minimize the influence of measurable environmental factors on trait variance by controlling for them in statistical models of heritability. There will, however, always be unknown or unmeasured environmental factors influencing bone growth, mass, or strength at any given time during the life span.

Environmental influences

Two environmental factors known to exert large effects on skeletal form are nutrition and biomechanical forces. Along with caloric intake, the most important nutrients for bone growth and health across the life span are protein, calcium and vitamin D, although other micronutrients also play key roles in skeletal development and maintenance. Because bone metabolic profiles vary according to life history period, they also differ in terms of their macro- and micro-nutritional requirements. Physiological changes over the life span can lead to alterations in nutrient absorption, processing and distribution, which can have substantial effects on bone. Likewise, biomechanical strains on the skeleton that lead to modeling and remodeling of bone are particularly active in childhood, but also persist throughout life.

Nutrition and mechanical strain each can influence fetal bone morphology. Poor maternal nutrition can lead to an under-mineralized and misshapen infant skeleton (Lanham et al., 2011). Biomechanical influences on bone growth, development and mineralization depend on the muscular activity of the fetus, as well as the activity of the mother (Carter et al., 1987; Rodriguez et al., 1988; Carter et al., 1991). In our study, we show the genetic influence on bone length in toddlers (2-year-olds) is very high, even after stature is accounted for. In fact, according to our results, the residual heritability for metacarpal length in our youngest age group (2 years of age) is 1.0, indicating that, other than the effects of the measured covariates, all of the remaining trait variance is due to genetic influences in our healthy, relatively homogenous population. At this early age, variation in the biomechanical environment is not expected to be high. As activity levels increase and muscles develop one would expect trait variation to increase, which may be reflected in reduced heritability estimates (i.e., increased environmental component).

With the onset of rapid pubertal growth, nutritional requirements for bone growth and development increase (Hoch et al., 2008; Rizzoli, 2008; Gat-Yablonski et al., 2009; Mesias et al., 2011; Thatcher and Clarke, 2011). In comparison to dietary needs during infant and childhood development, rapid pubertal growth requires much higher dietary intake of calories, protein, vitamin D, and calcium (Thatcher and Clarke, 2011), which, in turn, affects peak bone mass (Rizzoli, 2008; Mesias et al., 2011). While our sample population is considered normal and in no way deprived of nutrients, we nevertheless observe an increase in environmental influence during the pubertal years in the bone mass related traits. A similar trend is not observed for bone length during puberty. Physical activity during childhood and adolescence also influences bone, with more active individuals tending to have stronger bones (Ruff et al., 2006; Teran-Garcia et al., 2008). Similarly, while early-to-middle-adulthood is largely a period of stasis in terms of bone morphology or mass, intake of calcium and vitamin D during this stage of life can affect the amount of bone available in older age (Peters and Martini, 2010; Gille, 2010).

After menopause in women, and to a lesser extent after ~50 years of age in men, bone remodeling increases and bone loss becomes common (Peters and Martini, 2010), as is evident from our sample (see Figure 1c and d). Bone health in later life is partly a culmination of earlier experiences, but various acute nutritional deficiencies in older age can exacerbate extant maladies or increase trait variance. Adult bone also appears to have less capacity to add new bone than growing bone does, thus the primary goal in older adults is to preserve bone mass and minimize bone loss (Forwood and Burr, 1993; Kohrt et al., 2004; Martyn-St James and Carroll, 2009; Warden and Fuchs, 2009). This is especially important at older ages, when increased endocortical bone resorption (apparent in our data as decreasing summed cortical thickness and cortical index) renders bone more susceptible to fracture in both sexes, but particularly in postmenopausal women (Beck et al., 2001; Ahlborg et al., 2003; Duan et al., 2003; Szulc et al., 2006; Zebaze et al., 2007; Martin and Seeman, 2008).

CONCLUSION

Longitudinal studies, and the Fels Longitudinal Study in particular, provide a unique opportunity to explore changes associated with any number of traits across the life span. While anthropologists are typically interested in change in morphological phenotypes across time, we have presented, for the first time, a comparison of changes in both the phenotype and a corresponding measure of the underlying genetic influences on those phenotypes. The current study adds to the ongoing discussion of genetic and environmental influences on bone traits over the life course. Our results are particularly important because they provide evidence of a change in the genetic contribution to variation in gross bone morphology while bones are actively growing, as well as throughout the adult life span where the biology of the skeleton transitions from growth to maintenance.

Although this study focuses on a few key skeletal traits, the concepts addressed herein are widely applicable. Every phenotypic character, be it skeletal, somatic, or behavioral, has its own genetic landscape. Rather than identifying a global, overarching pattern of age-specific genetic influence, we have noted that each of our traits, even closely-related ones, can exhibit distinctive genetic landscapes. The consideration of trait-specific fluctuations in the genetic contribution to phenotypic variability over the lifespan provides a novel approach for interpreting population variation, adaptation, and evolution. Concepts and data presented here can be used as a resource for future inquiries in physical anthropology, such as issues of modern bone remodeling or susceptibility to environmental perturbations, or life history reconstructions and applications to the human fossil record.