Canalization leads to similar whole bone mechanical function at maturity in two inbred strains of mice

Abstract

Previously, we showed that cortical mineralization is coordinately adjusted to mechanically offset external bone size differences between A/J (narrow) and C57BL/6J (wide) mouse femora to achieve whole bone strength equivalence at adulthood. The identity of the genes and their interactions that are responsible for establishing this homeostatic state (i.e., canalization) remain unknown. We hypothesize that these inbred strains, whose inter-individual differences in bone structure and material properties mimic that observed among humans, achieve functional homeostasis by differentially adjusting key molecular pathways regulating external bone size and mineralization throughout growth. The cortices of A/J and C57BL/6J male mouse femora were phenotyped and gene expression levels were assessed across growth (i.e. 2, 4, 6, 8, 12, 16-weeks of age). A difference in total cross-sectional area (p < 0.01) and cortical tissue mineral density were apparent between mouse strains by 2-weeks of age and maintained at adulthood (p < 0.01). These phenotypic dissimilarities corresponded to gene expression level differences among key regulatory pathways throughout growth. A/J mice had a 1.55 – 7.65 fold greater expression among genes inhibitory to Wnt pathway induction while genes involved in cortical mineralization were largely upregulated 1.50 – 3.77 fold to compensate for their narrow diaphysis. Additionally, both mouse strains showed an upregulation among Wnt pathway antagonists corresponding to the onset of adult ambulation (i.e., increased physiological loads). This contrasts with other studies showing an increase in Wnt pathway activation following functionally isolated, experimental in vivo loading regimens. A/J and C57BL/6J long bones provide a model to develop a systems-based approach to identify individual genes and the gene-gene interactions that contribute to trait differences between the strains while being involved in the process by which these traits are coordinately adjusted to establish similar levels of mechanical function; thus providing insight into the process of canalization.

Introduction

A major unresolved challenge for identifying genetic and environmental factors that contribute to skeletal health is understanding how genes work together to establish system-level function through canalization. Canalization in this context is where a population demonstrates a similar phenotype, here a mechanically competent bone, despite inherent environmental and genotype differences. This process is achieved through developmental adaptations that occur despite variation among traits that may have arisen via differing biological or environmental conditions (1). For bone, primary system-level function includes establishing sufficient whole bone stiffness and strength during growth to resist fracturing while minimizing mass for metabolic conservation (2, 3). Targeted perturbation studies, such as gene function alterations, have provided tremendous insight into the identity of genes and molecular pathways that are involved in establishing mechanical homeostasis (4–8). However, these studies are comprised of selective decreases or increases in expression of single genes, which often lead to abnormal function (9–12), if not an overt pathological state (13, 14). Therefore, it is difficult to determine the coordination of multiple genes in a normal physiological state from those that materialize when gene function is compromised. Likewise, genome-wide association studies have successfully located many quantitative trait loci (QTL) associated with select bone traits, but have not defined how these genetic variants cooperate to establish whole bone mechanical function at the system-level (1, 15). Thus, the ways in which genes interact to establish normal, non-pathological mechanical function is not well understood.

Our previous work has shown that in mice (16, 17) and humans (18–20), phenotypic traits interact in a coordinated manner to establish a mechanically competent bone. For a given external bone size, a subset of morphological and compositional traits will predictably and coordinately adjust to offset the overall size of the bone in the interest of establishing and maintaining normal whole bone function. For example, our mouse model is comprised of two inbred mouse strains that have employed different biological pathways to achieve similar whole bone strength at skeletal maturity (21, 22). A/J mice have a narrow (slender) femoral diaphysis compared to the wide (robust) femoral diaphysis of C57BL/6J (B6) mice (Fig. 1). Both strains have a similar body mass and bone length. Therefore, to offset their differences in external bone size, and thus differing moment areas, they spatially distribute their cortical tissue in different ways, with A/J having a thicker cortex (i.e. greater cortical area with respect to external diaphyseal diameter) compared to the thinner cortex of their B6 counterpart. Moreover, A/J mice have a more mineralized cortical matrix than B6 mice, which effectively increases the whole bone stiffness of A/J femora to functionally match the wide, more ductile femora of B6. Therefore, both mouse strains have adapted their bone structure and tissue-level material properties in different ways to maximize their skeletal stiffness and strength while simultaneously minimizing their bone mass. This phenomenon is consistent with the functional objectives observed across many species (2, 23–25), including humans (20, 26). This consistency in coordination among phenotypic traits is established early in postnatal development in mice (~ 2-weeks of age) (21) and humans (~ 2-years of age) (27), and is maintained throughout growth and into adulthood, leading us to hypothesize that this trait adjustment mechanism is canalized across species (28). How this coordination among intermediate physical bone traits is accomplished at the genetic level remains unknown. Understanding this developmental mechanism may be clinically important for characterizing how human bone traits, which naturally vary among individuals, contribute to aging and respond to pharmacological interventions aimed at reducing fracture risk.

Figure 1.

Schematic highlighting the key morphological and compositional differences between male A/J and B6 mice at 16-weeks of age. Note that at this age of skeletal maturity these mouse strains demonstrate equivalent bone strength. Images taken using nanoCT at an 8μm voxel size.

Here, we use the A/J and B6 mouse model described above to begin investigating the molecular pathways between these two mouse strains whose inter-individual differences in bone structure and material properties mimic that observed among humans (29). This is a departure from a traditional targeted genetic approach, as we seek to investigate system-wide differences among unperturbed mice to investigate how genes interact to establish normal level function. This model can be used to determine how the system compensates for allelic and environmental factors that might disrupt homoeostasis, particularly those that influence the marked differences in external bone size, cortical thickness, and tissue mineralization between the two mouse strains.

We hypothesize that both mouse strains achieve a similar whole bone mechanical homeostatic state through the differential regulation of key molecular pathways contributing to the phenotypic morphological and compositional traits that differentially arise during growth. We tested this hypothesis by conducting three aims: 1) to identify which bone genes are differentially expressed between strains at multiple time points during postnatal growth, 2) to ascertain which bone-relevant pathways are differentially regulated throughout postnatal growth between strains, and 3) to determine how these gene expression profile differences between strains relate to their respective bone phenotypes at multiple time points across skeletal growth. Our goal is to understand how whole bone function is achieved at the molecular level through the differential coordination of traits, while illuminating the manner in which this mechanism may be canalized.

Sample and methods

Prior to the study outlined herein, a pilot study was conducted using both male and female B6 and A/J mice at multiple time points. No statistical sex-specific difference was noted within the phenotypic or gene expression results, therefore the following study only focuses on males. Male B6 and A/J mice were acquired from The Jackson Laboratory (Bar Harbor, ME, USA) at multiple ages that span postnatal skeletal growth. Upon arrival, mice were allowed to acclimate for one week before sacrifice. All mice were provided water and fed a standard rodent diet (Purina Rodent Chow 5001; Purina Mills, Richmond, IN, USA) ad libitum. Mice were housed at a maximum of 5 mice per cage and exposed to a 12-hour light/dark cycle. Left and right femora for phenotypic and genotypic analyses were harvested from 2, 4, 6, 8, 12, and 16-week-old mice. These ages were chosen since it is during early postnatal growth when traits are coordinately adjusted to establish whole bone mechanical function at adulthood (21). The University of Michigan Institutional Animal Care and Use Committee approved all mouse studies.

Phenotypic analysis

Changes in femoral diaphyseal growth were assessed at six time points throughout development (i.e., 2, 4, 6, 8, 12, 16 weeks). Left femora were harvested from male A/J and B6 mice (n=10/age/strain). Body mass (BM) was measured prior to sacrifice. Maximum femur length (Le) was measured from the most proximal tip of the greater trochanter to the distal condyles using digital calipers (0.05 mm resolution). All femora were imaged at an 8 μm voxel size using a nanotom S (phoenix|x-ray, GE Sensing and Inspection Technologies, GmbH, Wunstorf, Germany), converted to Hounsfield units using a phantom of air, water and hydroxyapatite (1.69 mg/cc; Gammex, Middleton, WI, USA) and reconstructed using datos|x reconstruction software (phoenix|x-ray, GE Sensing and Inspection Technologies, GmbH, Wunstorf, Germany) as described previously (22). Imaging parameters were varied slightly per age group to accommodate the different attenuations of the bone across groups (80 – 90 kV, 220 – 375 μA, 750 – 2000 ms, 3 – 4 average, 1 skip, 0.3 mm Al filter). These parameters were validated in house to provide the best visualization of characteristics at specific ages. From each image file, the region of interest (ROI) analyzed consisted of a 15–19% region of the mid-diaphysis taken just distal to the third trochanter. Each ROI was thresholded using Otsu’s method (30). ROIs were defined and variables were quantified using MicroView v2.2 Advanced Bone Analysis Application software (GE Healthcare Pre-Clinical Imaging, London, ON, Canada). Variables quantified included total cross-sectional area (Tt.Ar), cortical area (Ct.Ar), marrow area (Ma.Ar), and tissue mineral density. Cortical tissue mineral density (Ct.TMD) was derived from the mean gray value using calibration phantoms.

Following computed tomography, left femora from mice 4-weeks of age and older were loaded to failure using a custom 4-point bending fixture at a displacement rate of 0.05 mm/sec (MTS 858 MiniBionix; Eden Prairie, MN, USA). Bones were kept moist with 1x phosphate-buffered saline (PBS) and tested at room temperature (RT) with the anterior surface of the femur in tension. Variables calculated from the load-deflection curves included stiffness (S), maximum load (ML), post-yield deflection (PYD) and work-to-fracture (Work), as described previously (31, 32). Two-week-old femora were not tested due to the insufficient lamellae at this growth stage to adequately perform four-point bending tests.

Gene expression analysis

Tissue preparation and RNA extraction

Gene expression profiles were obtained from RNA extracted from the femoral diaphysis of 2, 4, 6, 8, 10, and 12-week-old male A/J and B6 mice. Sixteen-week old mice were not used for gene expression assays RNA concentrations yielded were drastically different between strains and, in total, substantially lower than those of the other six time points. For qPCR, 18 mice per strain and age were analyzed (6 pools of 3 mice/strain/age). For RNA-sequencing 8 mice per strain were analyzed at 6-weeks of age. Mice were anesthetized with isoflurane (Piramal Healthcare, Ltd.; Andhra Pradesh, India) and euthanized via cervical dislocation and a bilateral pneumothorax. For RNA extraction, the left and right femur of each animal was harvested and any adhering periosteum, muscle, tendon, and ligament were removed to detach adhering cells along the periosteal and endosteal surfaces. Next, the proximal and distal epiphyses were excised and the bone marrow was manually flushed using minimum essential medium, alpha (Gibco, ThermoFisher Scientific, Waltham, MA, USA). A histological analysis on a subset of processed bones confirmed that our tissue harvesting protocol was effective at removing most of the adhering cells along the periosteal and endosteal surfaces (data not shown). Complete tissue preparation for each sample occurred within 6 minutes after death. Bone samples were then immediately flash frozen in liquid nitrogen and stored at −80°C.

RNA extraction was achieved by pulverizing the pair of bones in 2 ml of TRIzol (Life Technologies, Grand Island, NY, USA) using a high-speed tissue homogenizer (Model 1000; ThermoFisher Scientific, Waltham, MA, USA). Each sample was subjected to three cycles of 20-second homogenization, placing the sample on ice between cycles to reduce thermal accumulation and subsequent RNA degradation. Total nucleic acid content was then isolated from the bone mineral and protein using 0.2 ml of 24:1 chloroform:isoamyl alcohol per 1 ml of TRIzol and subsequently centrifuged at 12,000 g for 15 minutes at 4°C. The supernatant containing the RNA fraction was then further purified using the RNeasy Mini Kit (Qiagen, Valencia, CA, USA) and the DNA was digested on column using an RNase-Free DNase Set (Qiagen, Valencia, CA, USA), according to the manufacturer. Total RNA was then recovered in 30 μl of RNase-free water. RNA concentration for each sample was determined using NanoDrop 2000 (ThermoFisher Scientific, Waltham, MA, USA). RNA quality was assessed using a bioanalyzer (Model 2100; Agilent Technologies, Santa Clara, CA, USA). All samples with a 260/280 nm absorbance ratio of 1.8 or greater and an RNA integrity number greater than 8.0 were deemed to be of sufficient purity and quality for use in downstream analyses.

RNA-sequencing methodology

Whole transcriptome RNA-sequencing was conducted individually on the messenger RNA (mRNA) of 8 pairs of femora from male A/J mice and 8 pairs of femora from male B6 mice at 6-weeks of age. This time point was chosen since this is the age when the external size of the B6 and A/J femoral diaphysis begins its greatest divergence from one another (as reported in the Results). Prior to library construction each RNA sample was depleted of its ribosomal RNA (RiboGone; Clontech, Mountain View, CA, USA), according to the manufacturer. An mRNA library for each sample was subsequently prepared using the SMARTer Stranded RNA-Seq Kit (Clontech, Mountain View, CA, USA). All buffer volumes, times, and temperature settings used to construct the library were exactly as recommended in the manufacturers protocol. Briefly, mRNA was fragmented to obtain segments of ~200 base pairs in length and converted to single-stranded cDNA that was ligated to barcode adapters. cDNA fragments were then purified using Agencourt AMPure XP magnetic beads (Beckman Coulter, Brea, CA, USA). Following purification, cDNA was amplified into RNA-sequencing libraries using DNA polymerase, a universal forward PCR primer, and reverse PCR primers corresponding to the paired-end Illumina primer index. Once amplified, the RNA-sequencing library was purified using Agencourt AMPure XP magnetic beads and validated with a bioanalyzer. Upon library completion, equal volumes of stranded, bar-coded cDNA libraries generated from the left and right femora of each mouse (A/J, n=8; B6, n=8) were multiplexed and sequenced in quadruplicate (4 lanes of flow cell) using a HiSeq 2500 System (Illumina, San Diego, CA, USA) to obtain a minimum of 25 million base pair reads per library.

Upon completion of sequencing, the quality of the raw reads data for each sample was checked using FastQC v1.10.1 (www.bioinformatics.bbsrc.ac.uk/projects/fastqc). Next, raw reads were aligned to the appropriate reference genomes for the A/J and B6 mouse strains using TopHat v2.0.9 and Bowtie v2.1.0. The reference genomes and the corresponding GTF files for the A/J and B6 strains were downloaded from the Center of Genome Dynamics at The Jackson Laboratory (http://cgd.jax.org/tools/Seqnature.shtml). Alignment quality was checked using FastQC to ensure only high quality data was used for expression quantification and differential gene expression analysis. Expression quantification at the gene level was performed using the htseq-count script embedded in HTseq (33). This python script uses the BAM (TopHat/Bowtie output) and GTF files for generating the counts of aligned reads for each gene that overlapped its exons. The htseq-count script is designed to only count reads that mapped unambiguously to a single gene, thus reads aligned to multiple positions or overlapping with more than one gene were discarded. The counts were then used for gene-level differential expression analyses using edgeR (34). To correct for the false discovery rate (FDR, q) p-values were adjusted for multiple hypotheses testing using a Benjamini-Hochberg procedure (q = 0.05). Genes and transcripts that were differentially expressed between A/J and B6 mice had to not only have a q-value less than or equal to 0.05 but also a fold change (Δ) (up or down) greater than or equal to 1.5. Pathway enrichment analysis was initially performed using Gene Ontology (35). Ingenuity Pathway Analysis software (Qiagen, Valencia, CA, USA) was then used to identify significantly differentially expressed canonical pathways and functional groupings of genes. Pathway analyses were performed using a right-tailed Fisher’s exact test to generate a p-value representing the probability that differentially expressed genes placed in a given pathway were not due to random chance and a Benjamini-Hochberg correction was applied to account for multiple comparisons (q = 0.05).

RNA-sequencing was confirmed and validated using a custom TaqMan microfluidic array card (Applied Biosystems, Foster City, CA, USA) that encoded for 46 Wnt pathway-related genes and two endogenous controls using qPCR (Model 7900HT RT-PCR; Applied Biosystems, Foster City, CA, USA) (Suppl. Table 1). As reported in the results below, an enrichment analysis using Gene Ontology of the RNA-sequencing data showed the canonical Wnt pathway to be significantly differentially expressed (≥ 2 fold) between the A/J and B6 mouse femora.

qPCR analysis

Gene expression changes in both B6 and A/J mouse strains across growth were tracked using cDNA generated from 2, 4, 6, 8, 10, and 12-week-old male mice. RNA samples were combined into 6 pools per strain and age (3 samples per pool) to yield 200ng of RNA per pool. Reverse transcription was performed on each pooled sample using qScript cDNA SuperMix (Quanta Biosciences, Gaithersburg, MD, USA) and was run in a C1000 Thermal Cycler (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer. Each cDNA sample was then combined with TaqMan Gene Expression Master Mix (Applied Biosystems, Foster City, CA, USA) and loaded into custom 384-well microfluidic array cards containing 46 unique primers related to the canonical and non-canonical Wnt pathways and three endogenous control primers (Suppl. Table 1). Each card allowed for 8-pooled samples to be run simultaneously. Cards were then centrifuged at 4°C (Legend XTR (with custom TaqMan array card bucket); Sorvall, Waltham, MA, USA) and sealed. Cards were run in accordance with the manufacturer’s recommendations.

Following the generation of the amplification plots, the baseline and threshold settings were adjusted to obtain an accurate threshold cycle (C_T) that was the same for both strains at each given age. A comparative C_T method (ΔΔC_T) was used to calculate fold change expression levels by normalizing the data to an endogenous reference (36). Due to the specificity, sensitivity and amplification efficiency of the TaqMan assays (37), fold changes (up or down) of 1.5 or greater that were identified as being statistically significant (p < 0.05) via a paired t-test met our criteria for denoting differences in gene expression levels between the two strains.

Protein analysis

Left and right femora of B6 and A/J 6-week old male mice (n=10/strain) were harvested in the same manner as described for RNA preparation. Whole cell extracts (lysates) were obtained by homogenizing both femora of two mice (5 pools/strain) in 50 mM Tris-base, 150 mM sodium chloride, 2 mM ethylene glycol tetraacetic acid, 10 mM sodium pyrophosphate decahydrate, and 10% nonyl phenoxypolyethoxylethanol for 30 seconds. Equal protein lysate concentrations (13μg per pool) were separated on 4% stacking gels and 12% SDS-PAGE gels (150V, 90 min) and transferred to nitrocellulose membranes (100V, 60 min). Membranes were blocked for 1 hour at RT with 5% bovine serum albumin (BSA) diluted in Tris buffered saline containing 0.05% Tween 20 (TBS-T). Following blocking, the membranes were incubated overnight at 4°C with a rabbit polyclonal anti-sclerostin antibody (sc130258, Santa Cruz Biotechnology, Dallas, TX, USA) diluted at 1:200 in 5% BSA. Membranes were then washed in TBS-T and incubated in a goat anti-rabbit immunoglobulin-G horseradish peroxidase-conjugated secondary antibody (sc2004, Santa Cruz Biotechnology, Dallas, TX, USA) diluted at 1:20000 in 5% BSA for 1 hour at RT. After rinsing in TBS-T the blots were developed on Fuji HR-T X-ray film (Fujifilm Medical Systems, Stamford, CT, USA) using SuperSignal West Femto Maximum Sensitivity Substrate (ThermoFisher Scientific, Waltham, MA, USA).

Statistical analysis

All data were analyzed using Minitab v16 (State College, PA, USA) and Prism v6 (GraphPad Software, La Jolla, CA, USA). A Shapiro-Wilk test confirmed that the data were normally distributed. To test for phenotypic differences between strains at each age, a general linear model (GLM) analysis of variance (ANOVA) was employed. A GLM was used to adjust all traits relative to BM for each age separately, which is a known covariate to these traits. Additionally, Tt.Ar was included as a covariate to Ct.Ar, Ma.Ar, and Ct.TMD since these traits vary relative to external bone size. Significance was taken at a p-value equal to or less than 0.05. An analysis of covariance (ANCOVA) was employed to test for significance in growth trajectories between mouse strains for Tt.Ar and Ct.TMD since these traits have previously been shown to have the greatest divergence between strains. This was done by plotting each trait against the logarithm of age to linearize the data, followed by a comparison of the slopes and y-intercepts to determine differences in phenotypes across growth. Using a Bonferroni correction to adjust for multiple comparisons, significance of the phenotypic data between strains across ages was based on an alpha level of 0.008. Effect sizes were reported as Cohen’s d. These were calculated as the difference between the means between strains for any given trait divided by the pooled standard deviation of those means. Gene expression level differences between mice strains were quantitatively assessed as described in the Gene Expression Analysis sub-section of the Methods outlined above. Western blot data was analyzed using a paired t-test with significance taken at a p-value equal to or less than 0.05.

Results

Femoral phenotypes for both B6 and A/J mice were largely established by 2-weeks of age

The unadjusted and adjusted means along with the standard deviations of all pertinent phenotype data are listed in Supplemental Table 2. By 2-weeks of age, a significant 33% difference in external bone size (i.e. Tt.Ar) between A/J and B6 mice was established and maintained throughout growth to 16-weeks of age (A/J: 0.68 – 1.33 mm²; B6: 1.01 – 2.04 mm²; effect size = 4.39 – 6.77; p < 0.001, GLM ANOVA; slope: p < 0.0001, ANCOVA) (Fig. 2A). Periosteal expansion of the femoral diaphysis was largely complete in both strains by 8-weeks of age. Ma.Ar increased between 2 and 4-weeks of age within both strains, with A/J mice showing significantly less marrow space into adulthood compared to B6 (A/J: 0.57 – 0.97 mm²; B6: 0.64 – 1.04 mm²; effect size = 1.23 – 3.71; p < 0.07 – 0.001, GLM ANOVA) (Fig. 2C). Ct.Ar relative to external diaphyseal size was significantly higher in A/J femora compared to B6 femora across growth (2 – 16-weeks of age), which explains their thicker cortex (A/J: 0.26 – 0.91 mm²; B6: 0.20 – 0.85 mm²; effect size = 1.23 – 3.71; p < 0.07 – 0.0001, GLM ANOVA) (Fig. 2B). For A/J to accommodate a narrower diaphysis while not drastically altering their absolute volume of tissue, they significantly increased their Ct.TMD by 8-weeks of age (A/J: 1321± 32.61 mg/cc; B6: 1257 ± 31.28 mg/cc; effect size = 2.00; p = 0.004, GLM ANOVA; y-intercept: p < 0.0001, ANCOVA) (Fig. 2D). This tissue-level adjustment partially contributed to a 10% significant difference in whole bone stiffness (A/J: 215.4 ± 23.9 N/mm; B6: 195.0 ± 23.4 N/mm; effect size = 0.86; p = 0.02, GLM ANOVA) (Fig. 3A), but only a 3% non-significant difference in ML at 16-weeks of age (A/J: 28.6 ± 2.5 N; B6: 27.7 ± 2.4 N; effect size = 0.35; p = 0.33, GLM ANOVA) (Fig. 3B). However, these adjustments in cortical mineralization, and in turn whole bone stiffness, in A/J femora throughout growth came at a mechanical cost, resulting in a more brittle bone as evidenced by the 62 – 77% significant difference in post-yield displacement between strains throughout growth (A/J: 0.12 – 0.24 mm; B6: 0.33 – 0.90 mm; effect size = 2.13 – 7.28; p < 0.001, GLM ANOVA) (Fig. 3C). This lower post-yield displacement largely contributed to the 41 – 58% significantly less work-to-fracture in A/J during growth (AJ: 4.92 – 7.33 Nmm; B6: 8.35 – 14.21 Nmm; effect size = 1.48 – 4.62; p < 0.005, GLM ANOVA) (Fig. 3D). This phenotypic analysis thus confirms that the primary morphological and compositional differences between B6 and A/J femora were external bone size and tissue mineral density, and that these differences are apparent as early as 2-weeks of age.

Figure 2.

Mean and standard deviations of bone morphological and compositional traits across multiple time points during growth and development after adjusting for body mass (A) or both body mass and total area (B–D). A) total area (Tt.Ar); B) cortical (Ct.Ar); C) marrow area (Ma.Ar); D) cortical tissue mineral density (Ct.TMD). * denotes significant differences between B6 and A/J mice at the p < 0.05 alpha level.

Figure 3.

Mean and standard deviations of mechanical 4-point bending data across multiple time points during bone growth and development after adjusting for body mass. A) stiffness; B) maximum load; C) post-yield displacement (PYD); D) work-to-fracture (work). * denotes significant differences between B6 and A/J mice at the p < 0.05 alpha level.

Femoral gene expression profiles significantly differed between B6 and A/J mice

RNA-sequencing of the femoral cortical matrices of 6-week-old A/J and B6 mice yielded 4290 genes that showed significant differential expression between strains (q ≤ 0.05, fold ≥ 1.5). From these genes, we identified 9 canonical pathways that largely encompass the morphological (i.e. external bone size) and compositional (i.e. cortical matrix mineralization) development of bone (Suppl. Table 3). The canonical (β-catenin) and non-canonical (PCP, Ca⁺) Wnt pathways were integral to many of the identified pathways and the majority of the recognized genes encompassing these 9 pathways demonstrated at least a 1.5 fold increased expression in A/J mice relative to B6 mice. The enrichments of greatest interest, selected using Gene Ontology, are listed in Table 1. These enrichments were all upregulated in A/J mice relative to B6, and qualitatively correlated well with the phenotypic differences observed between these mice at skeletal maturity (i.e. Tt.Ar and Ct.TMD).

Table 1.

Enrichment of genes playing key roles in bone development identified using Gene Ontology’s Panther Classification System. The number of identified genes included in each network, the −log p-value, and fold enrichment are provided.

Functional Enrichments	# of Genes	−log (p)	Fold Enrichment
Collagen Fibril Organization	21	5.70e-03	3.66
Regulation of Osteoblast Differentiation	46	1.33e-04	2.58
Regulation of Bone Mineralization	31	3.54e-03	2.85
Skeletal System Morphogenesis	71	1.58e-04	2.08
Regulation of Canonical Wnt Signaling Pathway	53	2.57e-02	2.00

The canonical Wnt pathway, which is instrumental in driving mesenchymal stem cells towards the osteoblastic lineage, demonstrated a 2-fold enrichment between mouse strains. Ingenuity Pathway Analysis identified 45 out of 165 known Wnt pathway-related genes in our bone samples. All but six of these genes showed greater expression in A/J relative to B6. Using knockout/knockdown models, others have identified several of these 45 genes to be antagonistic to the canonical Wnt pathway. When looking at the total net accumulated perturbation (a gene’s measured fold change combined with the accumulated perturbation transmitted from any upstream genes) of each of the 45 genes identified in our dataset, the greater antagonism of the canonical Wnt pathway in A/J through disruption of Wnt ligand signaling and binding translated downstream into a reduction in Dvl and β-catenin, as well as reduced Tcf/Lef transcription. Figure 4 demonstrates this outcome on the Wnt/β-catenin KEGG pathway. These RNA sequencing results were validated by our qPCR analysis of Wnt pathway-related genes (below).

Figure 4.

Schematic of the canonical Wnt pathway modeled after the KEGG diagram, highlighting the total perturbation of genes (a gene’s measured fold change combined with the accumulated perturbation transmitted from any upstream genes) identified via RNA sequencing of B6 and A/J mouse femora at 6-weeks of age. Color scale denotes the perturbation level of A/J femoral genes relative to those of B6. Blue denotes a decrease in a gene’s functional role and red denotes an increase in a gene’s functional role.

An additional outcome of the Gene Ontology and Ingenuity Pathway analyses was the identification of several genes that collectively play a substantial role in bone mineralization (−log (p) = 3.54e-03 and 1.86e-09, respectively). The expression levels of the majority of these genes were elevated in A/J femora relative to B6 femora. Of particular note was that A/J showed greater expression among genes recently discovered to comprise the acidic serine aspartate-rich Mepe-associated motif (ASARM) bone-renal mineral pathway, which includes Phex (fold Δ = 2.45), Dmp1 (fold Δ = 2.19), and Mepe (fold Δ = 2.86). Complementing the greater expression of these genes in A/J femora was the elevated gene expression levels of many other genes known to play a functional role in bone mineralization (Suppl. Table 4), relative to B6. The greater expression level of these genes in A/J femora qualitatively correlates with the higher Ct.TMD values observed in the phenotypic analysis of these mice.

Expression levels of Wnt pathway antagonist genes were higher in A/J femora; however expression levels shifted at a greater magnitude in B6 mice throughout growth

Mapping the canonical and non-canonical Wnt pathway-related gene expression profiles across growth using the custom microfluidic array cards highlighted several differences between A/J and B6 mice. Throughout growth, many of the 46 genes assayed that are involved in either the induction or inhibition of canonical Wnt pathway signaling, demonstrated elevated expression in A/J mice relative to B6 mice (Table 2). Genes that were elevated in expression among A/J mice that suggest pathway induction included Wnt10b, Ctnnb1, Lef1 and Tcf7, while genes that suggest pathway inhibition due to their elevated expression in A/J mice included Wif1, Sost, Dkk1, Dkk2, Sfrp1 and Sfrp4. Despite this juxtaposition in higher expression levels among genes that either promote or suppress canonical Wnt pathway signaling within A/J mice relative to B6 mice, the inhibitors collectively exhibited greater expression levels than their downstream targets involved in an activated pathway (Fig. 5). This suggested that A/J mice have some degree of sustained Wnt pathway inhibition throughout growth with respect to B6 mice, similar to the results obtained from the RNA-sequencing of 6-week-old mice. Sost showed the greatest difference in expression levels among these inhibitory genes, which was confirmed at the protein level (i.e. sclerostin) in 6-week-old mice via Western blots. A/J had a significantly higher concentration (p = 0.03) of sclerostin compared to B6 (Fig. 6).

Table 2.

Gene expression differences of Wnt pathway-related genes using custom array cards. Fold changes represent differential gene expression levels in A/J mice relative to B6 mice. Directional arrows indicate up or down regulation of all genes showing a 1.5 or greater fold change (up or down), regardless of p-value. Bold denotes significance at the p < 0.05 alpha level of A/J gene expression relative to B6 for each time point, regardless of fold change.

Age	2wk		4wk		6wk		8wk		12wk
Gene	Fold Δ		Fold Δ		Fold Δ		Fold Δ		Fold Δ
Wnt1	2.66	↑	1.35		3.08	↑	3.18	↑	3.34	↑
Wnt5a	1.78	↑	0.93		2.06	↑	2.65	↑	3.32	↑
Wnt6	0.47	↓	0.52		0.54		0.77		2.70	↑
Wnt7b	1.62	↑	0.57	↓	1.53	↑	1.98		5.48	↑
Wnt10a	1.31		0.82		1.34		2.06	↑	5.36
Wnt10b	1.84	↑	0.88		2.01	↑	1.99	↑	3.71	↑
Fzd1	1.28		0.88		1.88	↑	1.72	↑	2.72	↑
Fzd2	1.79	↑	1.22		3.75	↑	4.40	↑	7.09	↑
Fzd6	1.07		0.87		1.33		1.80	↑	3.58	↑
Fzd8	1.68	↑	0.95		1.86	↑	2.94	↑	3.50	↑
Lrp4	1.23		0.79		1.74	↑	2.10	↑	3.29	↑
Lrp5	1.18		0.83		2.50	↑	1.82	↑	3.88
Lrp6	1.10		0.96		1.76	↑	1.56		2.40	↑
Dvl1	1.45		0.97		1.78	↑	1.75	↑	2.31	↑
Axin1	1.25		0.87		1.25		1.26		1.60	↑
Axin2	1.28		0.99		2.51	↑	2.34	↑	4.45	↑
Apc	1.12		0.98		1.68	↑	1.74	↑	3.34	↑
Gsk3β	1.07		0.92		1.45		1.49		2.64	↑
Ctnnb1	1.28		0.85		1.84	↑	1.90	↑	3.04	↑
Lef1	0.85		1.00		1.78	↑	1.77	↑	3.48	↑
Tcf7	1.29		1.04		2.38	↑	2.70	↑	2.81	↑
Sfrp1	0.97		1.19		1.44		2.02	↑	2.05	↑
Sfrp4	1.55		0.89		1.67	↑	1.55	↑	2.43	↑
Wif1	0.73		0.74		1.98	↑	2.67	↑	3.77	↑
Sost	7.65	↑	1.70	↑	5.91	↑	5.14	↑	4.66	↑
Dkk1	2.69	↑	1.18		3.70	↑	2.93	↑	4.94	↑
Dkk2	2.13	↑	1.13		2.98	↑	3.12	↑	3.69	↑
Notch1	0.95		0.60		0.94		0.90		1.16
Sostdc1	1.22		2.21	↑	1.63		7.75		6.15	↑
Cepba	1.29		1.23		1.48		1.96	↑	1.78	↑
Pparγ	0.78		1.13		1.00		1.33		1.67	↑
Msx2	0.45	↓	0.52	↓	1.90	↑	1.40		1.81
Tnfrsf11b	1.62	↑	0.88		2.52	↑	2.41	↑	2.88	↑
Tnfsf11	1.05		0.87		2.00	↑	3.25	↑	2.34	↑
Runx2	1.34		0.98		1.91	↑	2.20	↑	3.70	↑
Sp7	1.70	↑	1.20		2.95	↑	2.85	↑	3.73	↑
Calca	0.71		0.57		2.94		1.63		-
Twist1	1.07		0.48	↓	1.08		1.54		2.30	↑
Wisp1	2.00	↑	1.05		2.70	↑	2.62	↑	6.13	↑

Figure 5.

Canonical Wnt gene expression changes in A/J relative to B6 mice throughout growth in A) key pathway inducers, and B) key pathway inhibitors via custom array cards. Note the collectively greater expression of inhibitory genes of the pathway relative to those that promote induction.

Figure 6.

Sclerostin protein expression in 6-week old B6 and A/J femora via Western blot. The upper band at 26kD may be a result of glycosylation, as sclerostin is believed to have N-glycosylation sites (57). The right panel shows the relative difference in sclerostin protein expression based on gray value band intensity at the 23kD molecular weight.

To determine the magnitude in which expression profiles of genes involved in the Wnt pathway changed across growth between A/J and B6 mice, the ΔΔC_T values for each gene analyzed were plotted as a function of age by normalizing the values at each time point to the ΔΔC_T values at 2-weeks of age (i.e. their baseline expression level). Normalization of gene expression levels within both strains to their respective 2-week age revealed that although A/J mice have greater overall gene expression with regard to genes involved in the Wnt pathway across growth relative to B6 mice, B6 mice actually showed greater shifts in gene expression among genes compared to A/J mice (Fig. 7). The gene demonstrating the greatest expression level change at all time points across growth in both mouse strains was Sost. From their baseline fold change ratio of 1 at 2-weeks of age, A/J showed a maximum 5-fold change in expression during growth, whereas B6 showed a maximum 7.5-fold change in expression.

Figure 7.

Canonical Wnt pathway gene expression changes in A) key pathway inducers, and B) key pathway inhibitors throughout growth. Fold changes reported are those following normalization of the data to the 2-week time point. Note the greater magnitude of change for Sost in B6 mice compared to A/J mice.

Discussion

We tested the hypothesis that A/J and B6 mice achieve a mechanical homeostatic state through the differential regulation of key molecular pathways that correspond with their phenotypic differences in external femoral bone size and mineralization. RNA-sequencing and qPCR analyses were conducted on cortical bone of the femoral diaphysis. Analyses of femoral bone RNA throughout growth found significantly different gene expression profiles present between A/J and B6 mice as early as 2-weeks of age. This timeframe corresponded to an age when a difference in Tt.Ar and Ct.TMD was already apparent between A/J and B6 male mouse femora, suggesting that the differing biological pathways employed to achieve bone mechanical competence is canalized. Many of the genes identified from RNA-sequencing reflect bone development, with expression levels of these genes being largely greater for A/J mice relative to B6 mice. Collectively, the genes identified comprise pathways involved in bone homeostasis and mineralization (7). The canonical Wnt pathway, which many have shown to drive osteoblastic differentiation, and thus an increase in bone formation (see (7) for a review), was found to play an integral role in many of the enriched pathways resulting from the RNA-sequencing of these 6-week-old mice. An assessment of key genes encompassing this pathway across multiple time points during growth found gene expression levels that were primarily expressed at a higher level among A/J males relative to B6 males. Since A/J mice show a narrower femoral diaphysis, albeit with a relatively thicker, more mineralized cortex, the elevated expression levels among several genes known to either inhibit Wnt signaling or increase cortical mineralization at 6-weeks of age suggested a compensatory mechanism may be in place to adjust tissue-level strength relative to external bone size. This hypothesis is bolstered with the finding of greater expression levels among genes involved in the ASARM bone-renal pathway, which has been shown to play a regulatory role in mineralization, phosphate regulation, soft-tissue calcification, osteoclastogenesis, and mechanotransduction (38).

Also of interest from the phenotype and gene expression data is that despite relatively greater gene expression levels among Wnt pathway inhibitors in A/J mice, gene expression levels, after being normalized to the 2-week time point (i.e. baseline expression level), tended to increase during growth at a greater magnitude (i.e. fold) among B6 mice compared to that of A/J mice. Thus, there are strain and age-specific gene expression differences. This shift in expression levels among Wnt pathway inhibitory genes may provide a mechanism where the periosteal expansion of B6 diaphyses, which are already wide relative to A/J, is held in check, particularly during the rapid growth phase occurring between 4 to 8-weeks of age. Additionally, the lower expression levels of genes involved in cortical mineralization among B6 mice at 6-weeks of age complements the phenotypic data concerning their lower Ct.TMD relative to their A/J counterparts, whom show increased signaling for mineralizing activity to offset their narrow diaphysis.

Collectively, the gene expression profile data support the dichotomy in external bone size and whole bone stiffness observed phenotypically throughout growth between A/J and B6 mice. By differentially inducing or inhibiting molecular pathways known to play a role in bone homeostasis, it would appear that A/J and B6 mice are able to maintain an intricate balance between bone mass and strength in the interest of achieving a mechanically and metabolically functional bone at adulthood. This outcome was not unexpected, since Currey and Alexander (2) hypothesized years ago that the ultimate goal of bone is to achieve mechanical and metabolic function through a balance of stiffness and mass. The phenotypic and gene expression profiles of mice comprising our model support this hypothesis, with A/J and B6 mice potentially reaching the same functional endpoint through the differential regulation of genes encompassing the Wnt pathway and functional mineralization network involved in achieving a homeostatic state between external bone size (i.e. mass) and cortical mineralization (i.e. tissue-stiffness).

Our assessment of gene expression within the Wnt pathway during growth provided new insight into the role of Sost, and other inhibitors of the Wnt pathway (e.g., Dkk1, Dkk2, Sfrp1, Sfrp4, etc.), on the development of external bone size. Sost encodes the protein sclerostin, a major antagonist to the Wnt pathway in bone, which effectively blocks Wnt ligands from binding to the Lrp5/6 receptors thereby inhibiting downstream signaling (39, 40). Outcomes of targeted knockout studies (41–44) and studies suppressing sclerostin expression using monoclonal antibodies (45–48) have consistently shown that activation of the Wnt pathway is associated with increased periosteal expansion and marrow infilling. In our study, A/J mice demonstrated a narrow femoral diaphysis with a 2–8 fold higher Sost expression throughout growth relative to B6 femora. Although periosteal expansion is largely completed by 8-weeks of age in A/J femoral diaphyses, they continue to deposit bone endosteally to offset their narrow diaphysis. This continuation of marrow infilling once periosteal expansion ceases is important as it enables A/J mice to achieve a near similar Ct.Ar as B6 mice in the interest of constructing a mechanically competent bone that is comparable to the whole bone strength and stiffness achieved by B6 upon skeletal maturity. This outcome suggests that the diaphyseal phenotype observed in A/J femora is not solely dependent on Sost expression, but most likely depends on interactions among many genes. Moreover, the roles Sost and other Wnt pathway inhibitors play throughout growth may differ depending on the age of the mouse, as others have previously shown (49). Overall, the outcome of the gene expression analysis narrows the focus of alleles and suggests that allelic differences between A/J and B6 mice lead to suppression of the entire Wnt pathway in A/J mice, while bolstering their functional competence with a greater expression level among genes integral to an increase in cortical mineralization. To further explore how mechanical function is established during growth it would be beneficial to expand analyses to naturally perturbed mouse strains (e.g., recombinant inbred, chromosome substitution, collaborative cross, heterogeneous stock). These strains are largely non-pathologic, and can minimize the interpretive complications that may arise from the presence of flanking genes in targeted gene perturbation models (50).

Although a comparison of gene expression differences between A/J and B6 was consistent with expectations for the role of Wnt signaling in regulating external bone size, a comparison of gene expression across growth was inconsistent with previous studies. The 2 – 7.5 fold increase in Sost expression between 2 and 4-weeks of age for A/J and B6 mice was in juxtaposition to the findings of others that reported a decrease in Sost expression among mice subjected to artificially applied loads (41–44, 51). This was unexpected since it has been reported that mice transition into full adult ambulatory activities between these time points (52), which presumably places greater physiologically relevant, weight-bearing mechanical loads on the skeletal system. The discrepancy between these outcomes could result from differences in the ages (e.g. 10–20-weeks of age) and skeletal elements (e.g. ulnae, tibiae, fibulae) of the animals investigated in prior studies (41, 42, 44), or from the assumption that tissue-strains would increase during the initial phases of ambulation. Moreover, differences between our findings and those of others may indicate that genes and the proteins they encode potentially play different roles depending on whether they are acting in a physiological or pathophysiological state (53). However, despite the incongruity between studies, the increase in Sost expression in both A/J and B6 mice throughout growth is consistent with the increase in serum sclerostin observed across growth for human boys and girls (54); though the association between SOST expression and serum sclerostin levels is not well understood (55). Finding a qualitative correlation between an increase among Sost and the transition to full ambulation among mice that have different phenotypic trait sets suggests the relationship between this key Wnt pathway antagonist and bone formation is more complex than has been previously shown among mice that are selectively perturbed. Based on previous findings, one would expect that as total diaphyseal area (i.e., bone formation) increases by ~0.5 mm², Sost expression would decrease to allow β-catenin to accumulate in the cytoplasm and translocate to the nucleus to induce Tcf/Lef transcription factors to drive mesenchymal stem cells towards the osteoblastic lineage. Therefore, it remains unclear as to what role Sost may be playing during the shift to full ambulation. Further investigation into this relationship is required.

One limitation of this study is that only the gene expression differences of cells remaining in the cortical matrix following the removal of the bone marrow and musculotendinous tissue were examined. Presumably, bone cells present along the periosteal and endosteal surfaces also play an important mediatory role in external bone size and mineralization. Future work will need to independently assay the fibrous periosteum and bone marrow to determine the role these tissues have in regulating the distribution and composition of bone in space with regards to the differing diaphyseal phenotypes present in the current model. For example, others have shown periostin, an extracellular matrix protein that in bone is localized to the periosteum and expressed by the osteoblastic-lineage, to promote collagen cross-linking via enhancement of lysyl oxidase activation (51). Additionally, assays of the associated bone marrow could inform on whether signaling pathways responsible for osteoblastic and osteoclastic differentiation is differentially induced or inhibited between strains. Another limitation of this study is that we did not determine whether Sost is differentially expressed across the periosteal, intracortical, and endosteal envelopes to stimulate a localized adaptive bone formation response, which may explain why periosteal bone formation, but not endosteal infilling, was impeded in A/J mice.

In conclusion, the current study has confirmed that there are inherent gene expression profile differences between the cortical diaphysis of two inbred mouse strains that have different phenotypic trait sets yet similar whole bone function by skeletal maturity. Therefore, it would appear that the genotypic background of these mice contributes to their phenotypic outcome in a canalized manner, as hypothesized. A/J mice have a narrow femoral diaphysis that is highly mineralized, which coincides with their gene expression profile, suggesting that the Wnt pathway is being inhibited to some degree throughout growth while the ASARM bone–renal pathway and the greater mineralization functional network is being activated. Importantly, though the Wnt pathway appears to be inhibited in A/J femora this does not translate to a meaningful difference in absolute Ct.Ar relative to B6 femora. In addition to this finding, the dichotomy between high expression level changes among antagonists to the Wnt pathway in both strains during a period of greater physiological loads (i.e. adult ambulatory activity) and the decrease in these same antagonists upon the application of artificial loads, as reported by others, suggests that the role these genes play in developing and maintaining bone homeostasis is more complex than previously thought. To increase our understanding of the role genetic and environmental backgrounds have in establishing function during growth, future work should focus on teasing apart the functional networks responsible for coordinating external bone size, bone mass, and mineralization so that the traits together lead to a bone with sufficient stiffness and strength to be functional under physiological loads. Identifying alleles that may contribute to the physiological differences observed within an animal model that mimics the natural variation among complex trait sets known to contribute to the mechanical competence of bone in humans is of primary importance. This requires a more systems-based approach, rather than a reductionist one, to detect interactions both between genes and between genes and their environment, all of which contribute to the development of whole bone function (1, 15). An integrative approach that takes advantage of mouse strains developed from natural genetic recombinatory activities over several generations of random breeding, may allow us to better understand the underlying functional networks responsible for the establishment of whole bone functional competence (56).