Quantitative Trait Loci for BMD and Bone Strength in an Intercross Between Domestic and Wildtype Chickens

Authors


  • The authors state that they have no conflicts of interest.

Abstract

With chicken used as a model species, we used QTL analysis to examine the genetic contribution to bone traits. We report the identification of four QTLs for femoral traits: one for bone strength, one for endosteal circumference, and two affecting mineral density of noncortical bone.

Introduction: BMD is a highly heritable phenotype, governed by elements at numerous loci. In studies examining the genetic contribution to bone traits, many loci have been identified in humans and in other species. The goal of this study was to identify quantitative trait loci (QTLs) controlling BMD and bone strength in an intercross between wildtype and domestic chickens.

Materials and Methods: A set of 164 markers, covering 30 chromosomes (chr.), were used to genotype 337 F2-individuals from an intercross of domesticated white Leghorn and wildtype red junglefowl chicken. DXA and pQCT were used to measure BMD and bone structure. Three-point bending tests and torsional strength tests were performed to determine the biomechanical strength of the bone. QTLs were mapped using forward selection for loci with significant marginal effects.

Results: Four QTLs for femoral bone traits were identified in QTL analysis with body weight included as a covariate. A QTL on chr. 1 affected female noncortical BMD (LOD 4.6) and is syntenic to human 12q21–12q23. Also located on chr. 1, a locus with synteny to human 12q13–14 affected endosteal circumference (LOD 4.6). On chr. 2, a QTL corresponding to human 5p13-p15, 7p12, 18q12, 18q21, and 9q22–9q31 affected BMD in females; noncortical (LOD 4.0) and metaphyseal (LOD 7.0) BMD by pQCT and BMD by DXA (LOD 5.9). A QTL located on chr. 20 (LOD 5.2) affected bone biomechanical strength and had sex-dependent effects. In addition to the significant QTLs, 10 further loci with suggestive linkage to bone traits were identified.

Conclusions: Four QTLs were identified: two for noncortical BMD, one for endosteal circumference, and one affecting bone biomechanical strength. The future identification of genes responsible for these QTLs will increase the understanding of vertebrate skeletal biology.

INTRODUCTION

Low BMD contributing to osteoporotic fractures is a well-known problem in human medicine. Osteoporosis in commercially bred chickens (Gallus gallus) has been recognized as a problem for many years; as early as 1955, skeletal weakness in hens was reported as cage layer fatigue, with fractures found in as many as 90% of hen carcasses.(1) Both in humans and chickens, osteoporosis is defined by the decrease of mineralized bone leading to bone fragility and increased risk of fracture. BMD is known to be a complex phenotype strongly influenced by genetic factors,(2,3) and fracture liability has also been shown to be heritable.(4,5) Because BMD and bone strength are measured on a continuous scale, they are amenable for quantitative genetic analysis.

In chicken and mammals, cortical and trabecular bone provides the structural integrity of the bone. When the hen reaches sexual maturity, the synergistic action of estrogens and androgens stimulate formation of medullary bone, a labile type of endosteal bone with a lower collagen fibril content, more heavily calcified and metabolized at a much faster rate than cortical bone.(6) Medullary and trabecular bone are present in close proximity in the marrow space of the long bones of the hen,(7) making it difficult to differentiate these bone types from each other by standard techniques.

The most widely used technique to assess BMD is DXA, which has been validated for measuring BMD in chickens.(8–10) The DXA technique, however, cannot differentiate between cortical bone and the mix of medullary and trabecular bone in the medullary canal (hereafter noncortical bone), whereas BMD of these fractions can be measured with pQCT, also validated for determining BMD of these bone fractions in chicken.(11)

Chicken is a well-established model for studying bone resorption(12) but has rarely been used as a model species in quantitative trait loci (QTL) studies investigating BMD and bone strength.(13) The large number of phenotypically divergent breeds available and the recently released draft sequence of the chicken genome(14) make chickens an attractive model for genetic dissection of multifactorial traits. Moreover, information derived from avian QTL studies may provide novel insights about the genetic regulation of bone tissue in vertebrates and could support QTLs identified in other species if QTLs are present in regions showing interspecies synteny.(15)

Domestic animals are excellent models for deciphering the genetic basis of multifactorial traits because selective breeding has led to accumulation of mutations favoring certain desirable phenotypes, thus mimicking evolution by natural selection.(16) The domestic chicken breed white Leghorn (WL) and the main ancestor of domestic chicken breeds, the red junglefowl (RJ), have been separated since domestication, which occurred several thousand years ago.(17,18) During domestication, selection for production traits such as growth and egg production has resulted in large phenotypic differences between domestic chickens and the RJ. During the last century, intense selection for egg production traits in the WL has further accelerated these differences in production traits.

The aim of this study was to identify QTLs affecting bone strength and BMD in a unique two-generation WL and RJ intercross, in which several QTLs for behavioral, reproductive, and growth traits have previously been identified.(19–23) The identification of bone trait QTLs may also aid in the identification of genes that contribute to phenotypic variation of BMD and bone strength in other species.

MATERIALS AND METHODS

Animals

A three-generation pedigree was generated by crossing one RJ male with three WL females from the SLU13 line.(22) Four male and 37 female F1 individuals were intercrossed to generate 851 F2 individuals, of which 337 (159 females and 178 males) were analyzed in this study. The F2 population was raised in six separate batches as described previously.(22) WL and RJ individuals (five of each sex and line) were phenotyped at the same age as the F2 individuals.

DNA isolation, genetic marker analysis, and construction of linkage maps

Blood samples were collected from all F2 individuals, their parents (F1), and grandparents (F0). DNA was isolated from this blood using the DNeasy 96 Tissue Kit (Qiagen) for mouse tails, with minor modifications compared with the original protocol. All animals were genotyped for 164 markers (Supplementary File 1), of which 159 were microsatellites. The four nonmicrosatellite markers MCR1, MCR3, MCR4, and PMEL (dominant white) were scored as described previously.(21) The marker Legcol, which is a phenotypic trait, was scored by visual inspection of each individual. The sex-averaged map spanned 3356 cM, with an average marker spacing of 20.98 cM. Linkage maps for 30 (29 autosomal and the Z chromosome) linkage groups were constructed using the CRIMAP software.(24)

QTL mapping

Mapping and significance testing for the F2 cross were performed using a standard interval mapping method(25) using a variation based on least squares regression,(26) with these being implemented using the QTL Express software available at www.qtlcap.ed.ac.uk.(27) Both additive and dominance effects are modeled in this analysis. In the standard QTL analysis, sex was included as a fixed factor, with differences between sexes thereby corrected phenotypically before linkage analysis. However, analysis was also performed using a sex interaction term, whereby phenotypes are examined within sex. This allows differences between sexes at a single QTL to be observed and can help identify loci that would potentially be overlooked because of no or a small effect of the QTL in one sex. The effect of sex was therefore controlled for in both sets of analyses.

Analysis of the Z chromosome data were performed using the Qxpak v.2.13 software.(28) Fixed effects included in the model were rearing batch and sex, whereas body weight at 200 days was included as a growth covariate. In the case of females, a confounding effect caused by egg production can be a problem.(29) Egg production over a 1-week period before slaughter was included as a covariate in the QTL analysis and was found to have virtually no effect on any of the bone phenotypes. Because medullary bone is resorbed for deposition of calcium in eggshells, we cannot exclude the possibility that fecundity/reproductive vigor is related to medullary bone traits in the hen and that monitoring egg production over a longer period of time would have been more informative in the analysis. Confidence intervals of QTLs were defined as 1-LOD score drops at both sides of QTL peaks.

Significance thresholds were calculated by the permutation testing option available in QTL Express, with these values thereby tailored to the individual data set. Although these varied slightly from trait to trait, a 5% genome-wide significance level was an F-statistic of ∼9.0, whereas a 1% significance level was reached with an F-statistic of ∼10.5. In the case of sex interaction effects, these thresholds were ∼5.5 and ∼7.0, respectively. Suggestive thresholds are somewhat arbitrary, but levels of around one false positive per genome scan have commonly been used.(25) A more stringent suggestive threshold of 20% genome-wide significance has been used in this study, with this being used in several studies previous to this.(23,30,31) This gave F-statistic thresholds of ∼7.0 for the standard analysis and ∼4.5 for the sex interaction analysis. Because a series of different traits were analyzed, it was necessary to perform multiple testing corrections to these levels, with the degree of correction for each trait depending on the number of noncorrelated traits to this that were also tested. Because these noncorrelated traits were often intercorrelated, the final corrections were based on the number of “suites” of noncorrelated traits, with these final corrections typically being either two or three suites for most traits. For each individual trait, standard Bonferroni multiple corrections upwardly adjusted the final significance threshold by ∼0.3 or 0.5 LOD (for two and three multiple tests, respectively). In the case of the suggestive thresholds, the level used is already substantially more stringent than the one outlined previously(25) (∼0.6 LOD higher), and therefore, this level was not adjusted further. All correlations were determined by Spearmans rank correlation tests (two-tailed).

Phenotypic measurements

All 337 F2 birds, 10 WL, and 10 RJ were killed at 230 days of age. The birds were defeathered and frozen immediately after death and were stored at –20°C until phenotypic measurements were performed.

Determination of total body BMD by DXA

At the time of DXA analysis, each carcass was thawed, and BMD of the total body was measured using a DXA scanner (Prodigy; Lunar, Madison, WI, USA), with the small animal mode used to maximize precision. Two sequential total body measurements were performed on each animal, and the mean value of the two measurements was used in the subsequent analyses. The technical quality was continuously checked according to standard clinical procedures and with phantom measurements. The long-term precision of the equipment has been determined to be <1% for a given phantom. DXA measurements were always performed using both the same instrument and the same two technicians during the entire study period. Total body weight of the carcasses was measured before DXA analysis. After DXA analysis, both femurs were removed and stored at –20°C until they were analyzed as specified below.

Analyses of excised right femora for BMD and composition

DXA:

Measurements of BMC and areal BMD (aBMD) of the femur were performed ex vivo with a Norland pDEXA Sabre (Norland, Fort Atkinson, WI, USA) and the Sabre Research software (v3.6), as previously described.(32) Briefly, femurs were placed on a custom-made 15-mm-thick Plexiglas tray and were scanned at a resolution of 0.5 × 0.5 mm.

pQCT:

CT was performed with the Stratec pQCT XCT Research M (Norland; v5.4B) operating at a resolution of 70 μm as previously described.(32) Noncortical BMD, which in the female bird reflects BMD of both trabecular and medullary bone, was determined ex vivo, with one metaphyseal pQCT scan of the region situated at 6% of bone length from the distal end of femur, and the noncortical bone was defined by setting an inner threshold to density mode (400 mg/cm3). In addition to data for noncortical bone, the metaphyseal scan was also used for derivation of data for total bone (including cortical, trabecular, and medullar bone). Cortical bone parameters were determined ex vivo with a mid-diaphyseal pQCT scan of the femur. Femurs from males and females were analyzed after replacing any air inside the medullary cavity with 70% ethanol. Because pQCT measurements of 70% ethanol alone gave values at 60 mg/cm3, it was concluded that values <60 mg/cm3 can not be correctly quantified. Because almost all males had values for noncortical BMD that fell below the 60-mg/cm3 threshold, all males were excluded from subsequent QTL analysis for this parameter. Ten females fell below the 60-mg/cm3 threshold, and the values of these were set to 60 as a conservative approach for subsequent QTL analysis of noncortical BMD.

After the DXA and pQCT analyses, the femora were stored at –20°C until biomechanical tests were performed.

Biomechanical testing

Three-point bending:

The right femurs, which had previously been measured for DXA and pQCT, were tested for biomechanical strength in a three-point bending test on an electromechanical testing machine (Avalon Technologies, Rochester, MN, USA). The specimens were kept frozen until a few hours before testing, when the right femurs were completely thawed at room temperature. The specimens were placed with the posterior cortex resting against two end supports placed with a distance of 40 mm between them. The bones were placed in such a way that the load was applied 6 mm distal from the mid-part of the femoral diaphysis with an anterio-posterior direction. The aim was to apply the load at the level where DXA and pQCT measurements had been performed. An axial load cell (Sensotec, Columbus, OH, USA) with the range 0–500N was used to apply a load of 1 mm/s to the bone. Values for load and displacement were collected 50 times per second until failure using software provided with the testing machine (Testware II). The collected data were stored as data files including the variables time, displacement, and load. Based on the collected data, load at failure, displacement at failure, stiffness energy to failure, and stiffness were calculated. The variable stiffness was calculated using the load and displacement data to define a slope that was based on the 10 highest consecutive stiffness values. The average of these 10 values was defined as the stiffness. Energy to failure was defined by the area under the load-displacement curve.

Torsional strength:

The left femurs were tested until failure in torsion. A few hours before the biomechanical analysis, the femurs were completely thawed at room temperature. Both bone ends were placed in purpose-made aluminum potting fixtures with a height of 15 mm each. The fixtures were filled with plastic padding (Plastic Padding Elastic) ensuring that the bone ends became immovably embedded, while leaving the remaining section of the femur between the fixtures (50–53 mm for all specimens) free for testing. A custom jig was used to ensure correct alignment of the bone axis along the loading axis of the testing machine. All specimens were tested to failure in torsion at room temperature on an electromechanical testing machine (Avalon Technologies) at a rate of 4°/s, with no axial load during the testing. Values for angulation and torque were collected 50 times per second until failure using the software provided with the testing machine (Testware II). Collected data were stored as data files including the variables time, angulation, and torque. Based on the collected data; maximum torque, torsion at failure (angulation at failure), and torsional stiffness were calculated. The variable stiffness was calculated by using the angulation and torque data to define a slope that was based on the 10 highest consecutive stiffness values in the data file. The average of these 10 values was defined as the stiffness.

RESULTS

Results from phenotypings of 337 F2 individuals, 10 RJ, and 10 WL are presented in Table 1. The distributions among F2 individuals for selected phenotypes are presented in Fig. 1. Marked sex differences were observed among F2 individuals for the measurement noncortical BMD, and most males fell below the detection limit of the pQCT instrument (Fig. 1B; female data presented). In contrast, males had a larger cortical thickness than females (Fig. 1D).

Table Table 1.. Mean Values and SDs for Phenotypes Included in the QTL Analysis
original image
Figure Figure 1.

Distributions in the F2 cross for nine selected phenotypic traits. Heights of black and white bars along the y-axis show the number of females and males falling within each phenotypic interval on the x-axis. For noncortical BMD, males are not included because they generally fell below the detection limit of the pQCT instrument.

When only sex and batch were included as covariates in the QTL analysis, many phenotypes under study were found to be strongly affected by two loci on chr. 1 (peaks at 58–76 and 427–431 cM) and one locus on chr. 27 (peaks at 0–32 cM; QTL results obtained in analysis without the body weight covariate are not presented). These three loci coincide with QTLs for body weight and growth previously identified in the same chicken intercross. Together, these growth QTLs explain >50% of the variance in body weight between adult RJ and WL,(21) and apparently also have pleiotropic effects on bone traits. To minimize the risk of detecting QTLs mainly affecting overall size, subsequent QTL analyses had body weight at 200 days of age included as a covariate.

When body weight was included as a covariate, the QTL analysis resulted in four QTLs being significant at the 5% genome-wide level. These QTLs were distributed on chicken chromosomes 1 (n = 2), 2 (n = 1), and 20 (n = 1) and had effects on both structural and biomechanical femoral bone traits (Fig. 2; Table 2). On chr. 1, we identified a QTL for female noncortical BMD (LOD 4.6) and a sex-independent QTL for endosteal circumference (LOD 4.6). A QTL on chr. 2 affected BMD in females: metaphyseal BMD by pQCT (LOD 7.0), BMD by DXA (LOD 5.9), and noncortical BMD by pQCT (LOD 4.0). A QTL on chr. 20 (LOD 5.2) affected structural rigidity of the femur in torsional strength test. In addition to the four significant QTLs, our results revealed additional QTLs that were significant at the 20% genome-wide level (indicating suggestive QTLs; Table 2).

Table Table 2.. QTLs for Bone Traits in the White Leghorn/Red Junglefowl F2 Cross
original image
Figure Figure 2.

F-values for selected traits are presented along chromosomes 1, 2, 13, 20, 27, and Z. The 5% genome-wide significance threshold is presented as an upper dashed line, whereas the lower dashed line represents the 20% genome-wide significance threshold (suggestive QTL threshold). On the x-axis, positions in centimorgans are presented for the five chromosomes. (A) F-values for selected QTLs identified in QTL analysis with sex interactions modeled. (B) F-values for selected QTLs identified in QTL analysis with no sex interactions modeled.

DISCUSSION

Genes controlling body weight and size often have pleiotropic effects on skeletal phenotypes. Large individuals typically have more bone tissue and bones that withstand greater biomechanical stress than small individuals do. Thus, genes affecting body weight or body size can be important indirect regulators of skeletal phenotypes. The QTL ecirc1 on chr. 1 and the suggestive QTLs on chr. 27 (Table 2) overlap QTLs for body weight and are present even after including the covariate body weight at 200 days in the QTL analysis. These loci may contain pleiotropic genes for which the covariate body weight at 200 days could not correct for all the effect on bone size.

All F2 individuals included in this study were killed at 230 days of age, approximately the age when chickens have their peak bone mass.(33) Therefore, it is likely that herein identified QTLs for BMD will harbor factors that affect the acquirement or maintenance of peak bone mass.

In the femur of the hen, trabecular bone resides in close proximity to the medullary bone, and female noncortical bone therefore likely represents a mix of these bone types. Medullary bone is formed in the long bones of female birds shortly before onset of egg laying, and acts, throughout the egg-laying period, as a labile calcium store mobilized to form eggshells.(6) It is a nonstructural type of woven bone(34) derived from endosteal surfaces and can sometimes completely fill the marrow space. Compared with cortical bone, medullary bone contains more noncollagenous proteins, proteoglycans, and carbohydrates, whereas cortical bone has a higher collagen fibril content and a more well-structured deposition of apatite crystal in the organic matrix.(6) Medullary bone is well vascularized, has a large surface area, and can be metabolized at a rate at least 10–15 times faster than cortical bone.(35) Furthermore, medullary bone has a high number of osteoclasts in comparison with cortical bone, indicating that the medullary bone represents a very active bone remodeling system.(36) Numbers of osteoclasts and osteoblasts have been found to be similar during both resorption and formation during eggshell cycles in birds, indicating that the mechanism of very rapid bone resorption of medullary bone during egg-laying depends rather on activation of present cells than the recruitment of new cells.(36–39) The amount of medullary bone in chicken humerus has been positively correlated with breaking strength,(40) which suggests that medullary bone may contribute to bone strength.

The process by which medullary bone turnover is regulated remains to be fully elucidated. Medullary bone is probably the most estrogen-sensitive of all known vertebrate bone types, and its formation can be rapidly induced in adult male birds by exogenous administration of estrogen.(39) Also, the classical calcium-regulating hormone PTH probably plays an important role, as does the vitamin D3 system. Local factors such as calcium and inorganic phosphate and prostaglandins likely also have important regulatory roles.

At the nc-bmd1 locus, female individuals homozygous for the RJ allele have higher noncortical BMD than their WL counterparts (additive effect = 29 ± 7 mg/cm3, dominance effect = −21 ± 11 mg/cm3), whereas the opposite is true for the bmd1 QTL, which acts in the WL direction (additive effect = 28 ± 7 mg/cm3, dominance effect = 7 ± 9 mg/cm3). Because the effects of bmd1 on BMD measured by DXA and metaphyseal BMD measured by pQCT seem to be female specific (Table 2), we speculate that bmd1 mainly affects medullary but not trabecular bone. The large additive effect sizes of bmd1 and nc-bmd1 suggest that F2 males generally should have had detectable amounts of noncortical bone if the QTLs had affected solely trabecular BMD. However, because trabecular and medullary bone can not be differentiated from each other by pQCT, it can not be excluded that bmd1 and/or nc-bmd1 have sex-dependent effects restricted to either trabecular or medullary bone, nor that these QTLs may affect both bone types.

Males being homozygous for the WL allele at the suggestive QTL on chr. Z have a higher femoral BMD by DXA than RJ allele homozygotes, although the RJ allele effect is dominant in heterozygous individuals (additive effect = 24 ± 6 mg/cm3, dominance effect = −17 ± 6 mg/cm3). The QTL interval also harbors a suggestive QTL for metaphyseal BMD by pQCT, for which the possible dominance of the RJ allele is more apparent than for BMD measured by DXA. Male additive and dominance estimates indicate that the RJ allele confers similar metaphyseal BMD effect regardless of being present in one or two copies (additive effect = 110 ± 10 mg/cm3, dominance effect = −100 ± 10 mg/cm3). In birds, males are the homogametic sex with a Z:Z karyotype, whereas females have one Z- and one W-chromosome. Because chr. Z is present in a single copy in females, it is not possible to estimate dominance effects. However, the female additive effect indicates that this QTL affects BMD in both sexes (Table 2).

Tors1 on chr. 20 affected the displacement (in degrees) at which the female femur broke in a torsion test, with higher displacement at failure in RJ/RJ homozygotes, therefore indicating that greater femoral elasticity is conferred by this genotype. The suggestive QTL tpb1 on chr. 13 acted in the direction of the WL line, with male WL homozygotes requiring a greater breaking load. However, male heterozygotes require a substantially greater force to reach breaking point (additive effect = −10 ± 8N, dominance effect = 60 ± 13N), indicating strong overdominance (and therefore heterozygote advantage) at this locus.

In a recent study, the genetic contribution to bone traits was examined in an intercross between broiler and WL chickens.(13) For tibia and humerus, the authors measured biomechanical traits and quantified BMD and BMC by DXA. In contrast to our study, they did not identify any genome-wide significant QTLs for BMD or bone strength, despite examining a large number of F2 individuals.

Comparative mapping using mice and human maps is a useful tool for identifying broad chromosomal regions that are homologous between species. Using the ECR browser (http://ecrbrowser.dcode.org/), human and murine genomic regions homologous to our QTL regions were identified. The Feb. 2004 chicken genome assembly featured a poorly assembled Z-chromosome. When defining QTL regions on chromosome Z, the May 2006 assembly of the chicken genome was consulted in addition to the ECR browser. QTLs nc-BMD1 and ecirc1 on chicken chr. 1 are syntenic to human 12q21–12q23 and 12q13–14, respectively. tors1 on chicken chr. 20 is syntenic to human chr. 20 (20q11.23–20q13.3) and to mouse chr. 2 (2H1–2H4). Interestingly, several murine QTLs for bone traits including BMD and periosteal circumference have been mapped to this region on mouse chr. 2.(41–44) The confidence intervals of QTL bmd1 on chicken chr. 2 are quite large, and the corresponding regions are dispersed on numerous human chromosomes (5p13-p15, 7p12, 18q12, 18q21, and 9q22–9q31). The suggestive QTL region on chicken chr. Z corresponds to human 5q12.1-q13.3 and to 13D1-D2.1 in the mouse genome.

In summary, in an intercross between wildtype RJ and domestic WL chickens, we identified four loci with effects on bone traits. One locus affected bone strength, one affected endosteal circumference, and two influenced noncortical BMD of the female femur. The main aims of this study were thus achieved, and our future work will be directed toward the underlying molecular mechanisms explaining the identified QTLs. Studies are underway for fine-mapping of identified QTL regions. In addition, microarray-based gene expression analyses studying RNA expression in femoral bone from WL and RJ individuals are ongoing, as well as genotyping of the F2 pedigree for single nucleotide polymorphism (SNPs) segregating between the WL and RJ and creating advanced intercross lines. Searching for selective sweeps where regions of a chromosome have been fixed for certain haplotypes is another tool that will be used to identify genes and regions that differ between WL and RJ, and that may affect the phenotypes under study.

Interestingly, in addition to a QTL affecting metaphyseal BMD and BMD measured by DXA, we also identified QTLs affecting bone strength and noncortical BMD. Unraveling the mechanisms behind herein identified QTLs would further our knowledge of vertebrate bone metabolism and could provide valuable avenues for further studies of fracture and bone metabolism in humans.

Acknowledgements

The study was supported by grants from Wallenberg Consortium North and the Swedish Research Council. The authors thank Örjan Carlborg for highly valuable support in the initial QTL analysis.

Ancillary