The authors state that they have no conflicts of interest.
Bone ALP and OC are under partial genetic control. This study of 591 pedigreed baboons shows a QTL corresponding to human 6p23–21.3 that accounts for 25% (bone ALP) and 20% (OC) of the genetic variance. A gene affecting osteoblast activity, number, or recruitment likely resides in this area.
Bone ALP and OC are under partial genetic control. This study of 591 pedigreed baboons shows a QTL corresponding to human 6p23–21.3 that accounts for 25% (bone ALP) and 20% (OC) of the genetic variance. A gene affecting osteoblast activity, number, or recruitment likely resides in this area.
Introduction: Serum levels of bone alkaline phosphatase (ALP) and osteocalcin (OC) reflect osteoblast activity. Both of these measures are under partial genetic control. Genetic effects on bone ALP have not been previously localized to chromosomal regions in primates, nor has the degree to which genetic effects are shared (pleiotropic) between bone ALP and OC been studied.
Materials and Methods: We applied variance components methods to a sample of 591 adult pedigreed baboons to detect and quantify effects of genes that influence bone ALP and that have pleiotropic effects on bone ALP and OC. A univariate linkage analysis was conducted for bone ALP. Bivariate linkage analyses were conducted in areas for which the bone ALP results presented here and a previous univariate OC linkage analysis showed evidence for linkage on the same chromosome for both bone ALP and OC.
Results: A quantitative trait locus (QTL) for serum levels of bone ALP is evident on the baboon ortholog of human chromosomal region 6p (LOD 2.93). Thirty-seven percent (genetic correlation [ρG] =0.61) of the genetic variance in bone ALP and OC is caused by pleiotropic effects of the same gene(s). Bivariate linkage analysis revealed a QTL in the region corresponding to human chromosome 6p23–21.3, with the strongest evidence for bivariate linkage near D6S422 (LOD =2.97 at 22 cM from our pter-most marker). D6S422 maps to 20.4 Mb in the human genome. The QTL-specific heritability (h2) is 0.25 and 0.20 for bone ALP and OC, respectively.
Conclusions: This first formal test for shared genetic effects on two serum markers of osteoblast activity indicates that a significant pleiotropic effect on bone ALP and OC levels, and thus on bone formation, is detectible. The fact that this region corresponds to one on mouse chromosome 13 that has repeatedly yielded QTLs for BMD should encourage more intensive study of the effect of genes in this region on bone maintenance and turnover.
Bone-specific alkaline phosphatase (bone ALP) and intact osteocalcin (OC) are serum markers of bone formation of interest to clinicians and bone biologists because they are easy to measure and are potentially useful as predictors of risk for fracture and other complications associated with disorders of bone metabolism. Bone ALP is a noncollagenous protein secreted by osteoblasts that is essential for bone mineralization and is a highly specific marker of osteoblast function.(1 Elevated serum levels of bone ALP are seen in conditions characterized by excessive bone turnover including postmenopause and ovariectomy in women and osteoporosis(2,3 and are associated with rapid bone loss(4 and fracture risk.(5,6 Like bone ALP, intact OC, a hydroxyapatite-binding protein found in bone, reflects osteoblast activity(7–10 and is used as a marker of bone formation. Serum OC levels are higher in individuals undergoing rapid bone loss and lower in those with diminished bone turnover.(11,12 Bone ALP and OC levels in humans are relatively high during growth, are lower after skeletal growth has ceased, remain low in premenopausal women and middle-aged men, and increase in older individuals including elderly men and postmenopausal women.(13–15
Previous studies in humans and in animal models consistently show that both bone ALP and OC are under significant genetic control. Quantitative trait loci (QTLs) have been identified for serum OC levels in primates (humans(16 and baboons(17). Bone ALP QTLs have been identified in inbred mice,(18 but not in other animal models or humans. None of these studies address the question of whether shown genetic effects on these two markers of bone formation can be attributed to the same gene(s).
The specific aims of this study are to (1) search for evidence of QTLs affecting normal population-level variation in serum bone ALP levels, (2) examine the extent to which genetic effects on bone ALP and OC are shared, and (3) identify QTLs that act pleiotropically to affect variation in both of these markers of bone formation. We conducted this study using the baboon, an established nonhuman primate model for skeletal maintenance and turnover. This study makes two important contributions to our understanding of the genetics of bone formation. First, it provides the first significant evidence for linkage in a primate species of serum levels of bone ALP levels to specific chromosomal regions. Second, it tests the hypothesis that some or all of the genetic effects on bone ALP and OC are caused by the same gene or genes and provides significant evidence for incomplete pleiotropic effects that are caused by a gene or genes that reside on the baboon ortholog of human chromosome 6p23–21.3.
MATERIALS AND METHODS
Subjects and phenotypes
The sample of Papio hamadryas included 591 pedigreed baboons (423 females, 168 males) at the Southwest Foundation for Biomedical Research/Southwest National Primate Research Center in San Antonio, TX, that are subjects of a large-scale study to map genes influencing bone maintenance and turnover. The animals ranged in age from 5 to 30 years and included two subspecies of Papio hamadryas (P.h. anubis [olive baboons] and P.h. cynocephalus [yellow baboons], and their hybrids). All animals are housed out of doors in social groups and maintained on commercial monkey chow to which they have ad libitum access. There are no housing or dietary differences between the sexes. Animal care personnel and staff veterinarians provide daily maintenance and health care to all animals in accordance with the Guide for the Care and Use of Laboratory Animals.(19 All procedures related to their treatment during the conduct of this study were approved by the Institutional Animal Care and Use Committee in accordance with the established guidelines.
Bone ALP and OC levels were assayed from frozen serum samples (blood originally obtained through femoral venipuncture between the hours of 8:00 a.m. and 11:30 a.m.). Bone ALP levels were detected using the Metra Bone ALP assay (Quidel Corp., San Diego, CA, USA). OC levels were assayed using the Metra Osteocalcin assay (Quidel Corp., San Diego, CA, USA), which measures intact (de novo) osteocalcin. Whereas the specific cross-reactivity relative to human bone ALP and OC was not directly assessed for the baboon, cross-reactivity of the kits used in this study has been validated in numerous other mammalian species, including the cynomolgus macaque, an Old World monkey closely related to the baboon.(20,21 We further infer validity of the human bone ALP and OC assays for the baboon from data obtained in our tests of assay reliability in which samples from 40 animals were assayed in triplicate in samples under multiple experimental conditions to which human samples were also subjected. Baboon bone ALP and OC assay values varied in the same way as those for humans under the same experimental conditions. All fell within the human nonparametric reference range published by the kit manufacturer. The intra-assay and interassay CVs for bone ALP were 2.19% and 5.4%, respectively. The intra-assay and interassay CVs for OC were 2.48% and 7.9%, respectively.
Baboon whole-genome linkage map
Genotype data from many of the same animals for whom serum bone ALP and OC data were obtained for this study were used to develop a first-generation genetic linkage map of the baboon genome.(22 Genotype data for 325 human microsatellite loci (simple tandem repeats [STRs]) and six novel baboon microsatellites were used in marker-to-marker linkage analyses, facilitated by the expert system program Multimap,(23,24 which implements routines of the computer program CRIMAP,(24 to produce a meiotic recombination map covering all 20 baboon autosomes. A more recent version of this map,(25 used for multipoint interval mapping of OC QTLs in this study, contained 275 of these STRs that had been placed in unique positions at 1000:1 odds and 8 more that had been placed in unique positions on the map at 100:1 odds.
Direct comparison among orthologous loci reveals large regions of synteny in which the human marker order is conserved (7 autosomes with no major rearrangement, 15 with one or more rearrangements(22,25). Given the known homologies, we have chosen throughout this paper to refer to baboon chromosomes by the number assigned their human ortholog (with the baboon chromosome number in parentheses). This facilitates comprehension of our results and the application of baboon linkage data to the human genome. For example, “chromosome 6 (PHA 4)” designates the syntenic grouping of microsatellite marker loci that map to human chromosome 6, a syntenic grouping that comprises the loci on chromosome 4 in the baboon (Papio hamadryas, [PHA]). Using similar logic, “chromosome 7/21 (PHA 3)” designates baboon chromosome 3, which consists of a fusion of two different syntenic groups in humans (i.e., chromosomes 7 and 21).
The baboons used in this study are assigned to 11 large extended pedigrees ranging in size from 67 to 171 individuals. The 591 individuals in these extended pedigrees comprise the following relative pairs: 381 parent—offspring, 319 siblings, 28 grandparent—grandchild, 58 avuncular, 5174 half-siblings, 1001 half-avuncular, 3 first cousins, 140 half-first cousins, 35 half-first cousins once removed, 47 half-siblings and half-first cousins, 12 half-siblings and half-avuncular, and 6 double half-avuncular. This is possible because most individuals are members of multiple relative pairs (e.g., one individual is part of not only the parent—offspring pairs that include his or her sire and dam but also those that involve his or her offspring, siblings, half-siblings, cousins, aunts, uncles, grandparents, etc. A representative pedigree is depicted in Havill et al.(26 All 11 pedigrees can be viewed online at http://www.snprc.org/baboon/map/BPPpedigrees/bpppeds.htm.
Univariate multipoint linkage analyses
Basic univariate quantitative genetic and multipoint linkage analyses were conducted for bone ALP and OC using a variance decomposition approach implemented in SOLAR (Sequential Oligogenic Linkage Analysis Routines)(27 to test for evidence of QTLs for bone ALP and OC. This method, described in detail elsewhere,(27 entails specification of the genetic covariance between arbitrary relatives as a function of the identity-by-descent (IBD) relationships at a given marker locus and models the covariance matrix for a pedigree as the sum of the additive genetic covariance attributable to the QTL, the additive genetic covariance caused by the effects of loci other than the QTL, and the variance caused by unmeasured environmental factors.
We estimated marker locus-specific IBD probabilities for the pedigrees using a pairwise maximum likelihood based procedure.(27 To permit multipoint analysis for QTL mapping, we used an extension to the method of Fulker et al.(28 to estimate IBD probabilities at 1-cM intervals along each chromosome. A LOD score evaluation was performed every centimorgan along each chromosome.
We tested the hypothesis of linkage by comparing the likelihood of a restricted model in which variance caused by the QTL equaled zero (no linkage) to that of a model in which it did not equal zero (i.e., is estimated). The LOD score of classical linkage analysis was obtained as the quotient of the difference between the two ln likelihoods divided by ln 10.(29
To control for the overall false-positive rate given the finite marker locus density in the baboon genome linkage map, we estimated the LOD score associated with a genome-wide p value of 0.05 by means of a method suggested by Feingold et al.(30 At LOD > 2.69, genome-wide p < 0.05. Therefore, we used LOD =2.69 as the threshold for genome-wide significance (at α =0.05).(30 This corresponds to a false-positive result that would be expected to occur at random once in 20 genome-wide linkage screens. Lander and Kruglyak(31 define “suggestive” evidence for linkage (LOD 1.9) in humans as a result that would be expected to occur once in a genome wide-linkage screen. In our baboon map, this corresponds to a LOD score of 1.46. The 95% CI includes an area 0.834 below the peak LOD score.
Significance of the mean effect of each covariate (age, age2, sex, and weight) was assessed using a likelihood ratio test. This test compares differences in the likelihoods of a restricted polygenic model, in which the mean effect of the covariate to be tested is constrained to zero, and a general polygenic model, in which all parameters are estimated, to the χ2 distribution with one degree of freedom (df =difference in numbers of estimated parameters in two models). The initial h2 estimate is the proportion of the residual phenotypic variance (the part that remains after covariate effects are removed) that is attributable to the additive effects of genes. Departures from multivariate normality in the distribution of residual trait values, especially kurtosis, can affect the accuracy of heritability estimates and LOD scores. In such cases, we ranked individuals according to their residual trait values and applied an inverse Gaussian transformation to impose a normal distribution to the data. We applied this transformation to the residuals for bone ALP and OC. All linkage analyses were performed using these normalized residuals.
Bivariate multipoint linkage analyses
To study the question of whether variation in bone ALP and OC is caused by the same gene(s), bivariate multipoint linkage analyses were conducted for the chromosomal regions that, in the univariate linkage analyses, showed potential evidence for linkage for both traits. The baboon orthologs of human chromosome 1 (PHA 1), 6 (PHA 4), and 19 (PHA19) yielded LOD scores >1.0 for both OC and bone ALP, indicating that there may be a gene or genes in these areas that affect variation in both bone ALP and OC. Our bivariate analyses test whether the covariance between these two phenotypes is caused by one gene with pleiotropic effects or two genes that happen to reside in the same chromosomal region.(32–34
As with the univariate analyses described above, we conducted these bivariate quantitative genetic analyses using a variance decomposition approach implemented in SOLAR(27 to analyze the normalized (inverse Gaussian-transformed) residuals. The basic bivariate quantitative genetic model involves simultaneous estimation of the following parameters for the normalized residuals for each trait: mean (μ); SD (σ); variance caused by the additive effects of genes (h2); variance caused by unmeasured environmental factors (e2); the correlation that is caused by shared genetic effects (ρG); and the correlation that is caused by shared unmeasured environmental effects (ρE).
A test for pleiotropy versus coincident linkage was performed in accordance with methods described in Almasy et al.(32 In short, the likelihood of a model in which the genetic correlation between the normalized residuals for the two phenotypes at the QTL (ρQ) is estimated is compared with the likelihood of a model in which ρQ is constrained to zero (coincident linkage) and a model in which ρQ is constrained to one (complete pleiotropy).
Table 1 presents the descriptive statistics for serum bone ALP and OC levels in these baboons by sex. Although the data in this table are from related individuals and tests of significance that do not account for this fact are inappropriate, these summaries are presented to provide a general appreciation of the population of study animals. For bone ALP and OC, the male mean is absolutely higher than, but within 1 SD of, that for females. Males show considerably wider phenotypic variation in serum bone ALP levels than do females. Range of variation for OC is similar for both sexes. Females, on average are smaller and lighter than males, which is to be expected in this sexually dimorphic species. Figure 1 shows the phenotypic correlation between serum levels of bone ALP and OC.
Table Table 1.. Descriptive Statistics for Serum Bone ALP,* OC,† and Covariates by Sex
Univariate quantitative genetic analyses for bone ALP and OC yield heritability and maximized parameter estimates for each phenotype as presented in Table 2. (Maximum likelihood estimates of h2 obtained in the original polygenic model used to characterize covariate effects are included in Table 2 for completeness, but it is important to note that these estimates are not reliable because of the departure from multivariate normality of the distribution of the residuals (see Table 3 for h2 estimates that derive from the appropriately transformed residuals). Figure 2 shows the results of the first genome-wide linkage analysis for bone ALP in primates. The highest LOD score was 2.93 on chromosome 6 (PHA4), with a peak nearest marker D6S259 −16 cM from our pter-most marker. The QTL-specific h2 =0.30, indicating that the QTL accounts for nearly 100% of the additive genetic variance in serum bone ALP levels. Suggestive evidence for another QTL occurs on chromosome 2q at a point 10 cM from our pter-most marker (LOD =1.65). A sequential oligogenic linkage scan that accounts for the variance caused by the highest peak shows a reduction of this second peak to a LOD of 0.80, indicating that, although there may be another QTL in this location that influences serum levels of bone ALP, the evidence for it is not statistically significant.
Table Table 2.. Univariate Quantitative Genetic Analyses of Serum Bone ALP and OC Levels in Pedigreed Baboons. Maximum-Likelihood Estimates (MLEs) of Covariate Effects and Residual Variance* Terms
Table Table 3.. Bivariate Quantitative Genetic Analysis of Serum Bone ALP and OC Levels in Pedigreed Baboons. Maximum-Likelihood Estimates (MLEs) of Variance Terms and Genetic and Environmental Correlations
We previously published a similar linkage analysis for OC.(17 In brief, we reported suggestive evidence for a QTL on chromosome 16q (PHA 20q). Although the analysis we conducted here is slightly different because it is based on transformed residuals, the results are qualitatively the same (i.e., the trait is significantly heritable and there is suggestive evidence for linkage on chromosomes 16q [PHA20q] and 1q [PHA1q]).
The maximum likelihood estimates that result from the bivariate quantitative genetic analysis are shown in Table 3. Significant h2 estimates for both phenotypes indicate that 20–36% (h2 =0.28 ± 0.08) of the variance in transformed bone ALP residuals, and 14–30% of the variance in transformed OC residuals (h2 =0.22 ± 0.08) is caused by the effects of genes. Thirty-seven percent of the genetic variance (calculated as the square of ρG) in the two traits is caused by the same gene or genes.
The bivariate linkage analyses conducted on chromosomes 1 (PHA1), 6 (PHA4), and 19 (PHA19) showed significant evidence for a QTL with pleiotropic effects on bone ALP and OC on 6p (PHA4p). No evidence for bivariate linkage occurred on chromosomes 1 (PHA1) or 19 (PHA19). Figure 3 shows the univariate linkage results on chromosome 6 (PHA4) for bone ALP and OC separately and the linkage results of the bivariate analysis. The bivariate analysis provides an improved estimate of QTL location and effect size, localizing the pleiotropic effect to an area flanked by markers D6S271 and D6S259 (6p23–6p21.3), with the strongest evidence for bivariate linkage slightly pter of marker D6S422 (LOD score of 2.97 at a location 22 cM from our pter-most marker). This marker maps to 20.4 Mb in the human genome. The QTL-specific h2 =0.25 and 0.20 for bone ALP and OC, respectively.
The hypothesis of coincident linkage is rejected because the genetic correlation between the two phenotypes at the QTL (ρQ) was estimated as 1.0 in the linkage model (so there is no need to compare it to a model in which the QTL [ρQ] was constrained to 1.0). The likelihood of this model differs significantly from one in which ρQ is constrained to zero (p =0.00009).
A second, smaller (suggestive) peak appears around a point 60 cM from the pter-most marker. A sequential oligogenic linkage scan that accounts for the variance caused by the highest peak (that one 22 cM from the pter-most marker) results in a reduction of this second peak to a LOD of 0.56. This indicates that, although there may be another QTL in this location that pleiotropically influences serum levels of bone ALP and OC, the evidence for it is not statistically significant.
The specific aims of this study were to (1) localize genetic effects on variation in bone ALP levels to chromosomal regions, (2) assess the degree to which genetic effects on bone ALP and OC are caused by the same gene(s), and (3) localize these shared genetic effects to chromosomal regions. Our results provide significant evidence that variation in bone ALP levels is caused by a gene or genes residing on the baboon ortholog of human chromosome 6p, that −37% of the additive genetic variance in bone ALP and OC is caused by the same gene or genes, and that the QTL on 6p (PHA4p) accounts for nearly 100% of the shared additive genetic variance in Bone ALP and OC levels.
This first evidence for linkage regarding normal variation in bone ALP levels in a nonhuman primate represents a significant advance in our understanding of the genetics of this marker of bone formation. Despite its status as an accepted and widely used marker of bone formation and osteoblast activity, research into the influence of genetic variation on serum bone ALP levels is in its infancy. Several studies have confirmed a genetic effect on tissue nonspecific alkaline phosphatase (total ALP). Total ALP is a ubiquitous enzyme thought to play an important role in the formation and mineralization of osteoid. Several dimeric isoforms of total ALP originate from various tissues (e.g., liver, kidney, and bone), with the majority coming from liver and bone.(35 Hypophosphatasia, a genetic disorder marked by depressed total ALP and defective skeletal mineralization, ranging in severity from death in utero to elevated fracture frequency in adults,(36 has been shown to be inherited in both an autosomal dominant and an autosomal recessive manner.(36–38 Two different alleles may independently lead to hypophophatasia,(37 and the structural gene for the total ALP enzyme has been assigned to chromosome region 1p36.1-34.(39 Given this clear evidence for genetic mutations that alter total ALP levels, it was expected that the bone-specific form of this enzyme is also under genetic influence. Indeed, two studies of human twins showed heritability (h2) values of 0.62(40 and 0.74(41 for bone ALP. In addition, Barone et al.(42 report on an inherited congenital disorder of glycosylation type IA that includes various skeletal changes, including low bone mass and elevated serum bone ALP levels. A previous study we conducted of serum levels of bone ALP in baboons showed heritability of this phenotype and showed evidence for genotype-by-sex interaction.(43 The subsequent linkage analysis reported here builds on this previous study by localizing the genetic effect to chromosome 6p (PHA4). Specifically, our 95% CI includes four human microsatellite markers that encompass the region 6p23–21.3. The only other published study identifying QTLs for bone ALP, conducted using two inbred mouse strains, reports heritability of 56% and linkage to mouse chromosomes 2, 6, and 14,(18 none of which are homologous to the chromosomal region for which we report linkage here.
The search for genes that affect serum OC levels is slightly more advanced than that for bone ALP genes. Puchacz et al.(44 mapped the OC gene to chromosome 1q through mouse—human somatic cell hybridization. Raymond et al.(45 refined the mapping of the gene, characterized allelic variation at the gene in a sample of postmenopausal white women, and found a significant association between one of the alleles and BMD. Morrison et al.(46 reported that vitamin D receptor polymorphisms contribute to OC level variability. Heritability studies in humans indicate that 37–84% of the variation in serum OC stems from genetic effects.(47–49 Linkage analyses show significant evidence for a QTL affecting serum OC levels on chromosome 16q in Mexican Americans,(16 and we previously reported a cross-species replication of this linkage signal on the baboon ortholog of human chromosome 16q(17 (a result that was confirmed by this study despite slight differences in statistical methodology).
Our results also provide the first formal evidence for shared genetic effects on bone ALP and OC. Tests for genetic correlation between these two serum markers of bone formation yielded a ρG =0.61, indicating that 37% of the residual genetic variance in the two phenotypes is attributable to the same gene or genes acting pleiotropically. The product of the squared genetic correlation and the heritability for one of the two correlated phenotypes yields an estimate of the proportion of the residual (i.e., after partitioning out the effects of measured covariates) phenotypic variance for that trait that is attributable to the effects of shared genes. This can be rescaled to the total phenotypic variance by multiplying the result by that proportion of the total phenotypic variance not accounted for by covariates (0.69 for bone ALP; 0.68 for OC; see Table 2). Seven percent of the total phenotypic variance in bone ALP and 6% of total phenotypic variance in OC is attributable to shared genetic effects. In addition to detecting and quantifying this pleiotropic effect, we also localized this common genetic effect to a QTL on chromosome 6 (PHA4). Because both bone ALP and OC are osteoblast products, it is highly likely that a gene affecting either osteoblast activity or osteoblast number and recruitment resides in this area.
Although commonly measured in both research and clinical settings as an acute marker of bone formation, the precise function of bone ALP and OC remain unknown. In vitro cell culture studies indicate that bone ALP marks earlier aspects of the osteoblast's role in bone formation than does OC,(50 and the general consensus is that these two markers reflect unique osteoblastic activities and different aspects of bone formation. The results of our bivariate genetic analysis support that bone ALP and OC are not simply redundant markers of the same process. Although −37% of the variance in the residuals for the two traits is caused by the effects of the same gene or suite of genes, the remaining 63% percent of the additive genetic variance is caused by genetic effects that are unique to each marker, and presumably, to the unique aspects of bone formation reflected by that marker.
Ultimately, bone ALP and OC are of interest because of the relationship of bone formation and turnover to bone mass and osteoporosis risk. Many linkage studies have been carried out for BMD and bone structure-related phenotypes in rodents and humans. We could find no published reports of QTLs in rats or in humans that correspond to the region of linkage we report here. In the mouse, however, QTLs for BMD have been reported for chromosome 13, including one for femoral BMD corresponding to human chromosomal region 6p25–21(51 and one for spine BMD corresponding to 6p24–22.(52 Further, Shimizu et al.(53,54 report that the Pbd2 locus on mouse chromosome 13, which affects peak BMD in senescence-accelerated mice, is homologous to multiple human chromosomal regions, including 6p25.21. Interestingly, in humans, bone morphogenetic protein 6 (BMP6) is located at 6p24–23. The study of Shimizu et al.(53 identified an exon polymorphism in Bmp6 in the mouse using a candidate gene approach, but functionality of this polymorphism was not shown. This candidate gene lies adjacent to the region included in our 95% CI. Whereas it is an interesting potential positional candidate gene, we are not implying that this gene is responsible for our linkage signal.
In this study, we presented the first statistically significant evidence for a QTL affecting serum levels of bone ALP in a primate species. In addition, we showed that this QTL harbors a gene that acts pleiotropically on normal population-level variation in serum levels of bone ALP and OC in baboons. Given the close genetic and physiological similarity of this animal to humans, these results should be highly informative in our search for genes affecting human skeletal maintenance and repair. The fact that this region corresponds to one on mouse chromosome 13 that has repeatedly yielded QTLs for BMD should encourage more intensive investigation of the effect of genes in this region on bone maintenance and turnover.
This study was supported in part in the form of a collaborative research contract with AxyS Pharmaceuticals (formerly Sequana Therapeutics) and research grants from the National Institutes of Health (F32 AR049694, R01 RR00878, and P51 RR013896). The authors gratefully acknowledge the technical contributions and support of the following persons: KD Carey, DE Newman, KS Rice, T Riley, E Rodriguez, E Windhorst, and SM Witte.