The Lyme Disease Pathogen Borrelia burgdorferi Infects Murine Bone and Induces Trabecular Bone Loss

ABSTRACT Lyme disease is caused by members of the Borrelia burgdorferi sensu lato species complex. Arthritis is a well-known late-stage pathology of Lyme disease, but the effects of B. burgdorferi infection on bone at sites other than articular surfaces are largely unknown. In this study, we investigated whether B. burgdorferi infection affects bone health in mice. In mice inoculated with B. burgdorferi or vehicle (mock infection), we measured the presence of B. burgdorferi DNA in bones, bone mineral density (BMD), bone formation rates, biomechanical properties, cellular composition, and two- and three-dimensional features of bone microarchitecture. B. burgdorferi DNA was detected in bone. In the long bones, increasing B. burgdorferi DNA copy number correlated with reductions in areal and trabecular volumetric BMDs. Trabecular regions of femora exhibited significant, copy number-correlated microarchitectural disruption, but BMD, microarchitectural, and biomechanical properties of cortical bone were not affected. Bone loss in tibiae was not due to increased osteoclast numbers or bone-resorbing surface area, but it was associated with reduced osteoblast numbers, implying that bone loss in long bones was due to impaired bone building. Osteoid-producing and mineralization activities of existing osteoblasts were unaffected by infection. Therefore, deterioration of trabecular bone was not dependent on inhibition of osteoblast function but was more likely caused by blockade of osteoblastogenesis, reduced osteoblast survival, and/or induction of osteoblast death. Together, these data represent the first evidence that B. burgdorferi infection induces bone loss in mice and suggest that this phenotype results from inhibition of bone building rather than increased bone resorption.

serum ALP activity was less prominent than in older mice, likely because bones in juvenile animals are still undergoing rapid bone formation and skeletal development (28). These data prompted us to investigate whether B. burgdorferi colonizes bone and/or affects bone health.
Bone colonization by B. burgdorferi. We examined B. burgdorferi colonization and infection effects on bone in 12-week-old adult male C3H/HeN mice. At this age, murine bone development stabilizes, and peak bone mineral density (BMD), trabecular volume, and cortical thickness are attained (28,29). Mice were infected for 4 weeks with 1 ϫ 10 4 B31-derived B. burgdorferi bacteria. Bone morphology, cellular composition, mineralization, growth rate, and biomechanical properties were investigated in long bones (tibiae, femora) and vertebrae, since long bones and vertebrae can exhibit different degrees of response to external stimuli. For example, rosiglitazone, a common antidiabetic drug, increases fracture risks in women in the limbs but not in vertebrae (30).
We first investigated whether B. burgdorferi could colonize murine bone. DNA was extracted from cleaned, crushed L6 vertebrae and distal halves of tibiae from which associated soft tissues and bone marrow had been removed (see Materials and Methods). Distal regions of tibiae were examined to ensure samples were collected as far from joints as possible, since joints are typically readily colonized by B. burgdorferi. It was more difficult to remove all surrounding soft tissues from vertebrae than from tibiae. Although we were confident that any bacterial DNA detected in tibiae likely originated from bone itself, we could not be certain that B. burgdorferi DNA detected in vertebrae was entirely bone derived. Total copies of a portion of the B. burgdorferi flaB gene and mouse nidogen DNA sequence were measured in each sample by quantitative real-time PCR (qPCR), and the flaB copy number/1,000 copies of nidogen was compared in different bones of mock-infected and infected mice (Fig. 2). flaB DNA copy number was significantly greater than background in mock-infected animals in the tibiae but not in vertebrae of infected mice (Fig. 2), indicating that tibiae were colonized more effectively. The flaB:nidogen copy number ratios in tibiae were similar to those previously reported in other tissues of B. burgdorferi-infected mice (31), implying that B. burgdorferi can colonize regions of long bones distal to articular surfaces as efficiently as other tissues.
B. burgdorferi infection induces osteopenia in the trabecular region of long bones. To determine if B. burgdorferi bone colonization induced pathology, we measured areal and volumetric bone mineral density (aBMD and vBMD) in the left femora and L5 vertebrae from mock-infected and infected animals, using dual-energy X-ray absorptiometry (DXA) and microcomputed tomography (CT). The effects of B. burgdorferi infection on trabecular and cortical bone were also compared using three- dimensional (3D) microarchitectural reconstructions from CT. To distinguish between potential effects on trabecular and cortical bone, CT analysis of cortical bone was performed in femora at the midshaft (composed mainly of cortical bone), whereas analysis of trabecular bone in femora was conducted at the distal metaphyses, where trabecular bone is more abundant (36). BMD values are expressed as percentages of the values for corresponding mock-infected animals (Fig. 3A).
To determine if changes in BMD in infected mice were clinically significant, we also calculated T-scores from the BMD values for vertebrae and femora. A T-score represents the number of standard deviations in the individual's BMD below the mean BMD of healthy young adults. T-scores 1 to 2.5 standard deviations lower than normal BMD are indicative of osteopenia, and T-scores more than 2.5 standard deviations lower than normal BMD correspond to osteoporosis (32). Means of individual T-scores were compared between B. burgdorferi-infected and mock-infected mice (Fig. 3B).
These experiments revealed that BMD of femoral trabecular bone was significantly lower in infected than in mock-infected mice (Fig. 3A). Trabecular BMD in vertebrae of infected animals was reduced but not significantly, and cortical BMD was not affected by B. burgdorferi infection (Fig. 3A). Examination of mean T-scores for bones from infected mice indicated that significant osteopenia was present in femora but not vertebrae (Fig. 3B). Representative reconstructed 3D models also showed visibly reduced bone volume and compactness from B. burgdorferi infection in the trabecular bone of femora, but the difference appeared less prominent in vertebral trabecular bone ( Fig. 3C and D). Finally, to assess whether reductions in the BMD and T-scores were a function of increasing bacterial DNA burden, we calculated correlation coefficients for flaB:nidogen copy number ratios and areal and volumetric BMDs and T-scores in long bones and vertebrae ( Table 1). All parameters were significantly negatively correlated with bacterial DNA burden, i.e., BMD and T-scores decreased as flaB:nidogen copy numbers increased. We concluded that B. burgdorferi infection was associated with clinically significant bacterial DNA load-dependent osteopenia in long bones but not vertebrae and that bone loss was confined to trabecular bone.
Effects of B. burgdorferi infection on trabecular and cortical microarchitecture in long bones and vertebrae. To assess if trabecular bone microarchitecture deterioration was caused by infection, we measured the following parameters: BV/TV (the fraction of bone volume [BV] within total measured tissue volume [TV] of interest); BS/BV (the ratio of the bone surface [BS] to total bone volume [BV]), trabecular number (Tb.N, the average number of trabeculae per millimeter), trabecular separation (Tb.Sp, the mean distance between trabeculae), and the trabecular pattern factor (Tb.Pf, the intertrabecular connectivity, determined by assessing the relationship between convex and concave trabecular surfaces) (33). BV/TV, BS/BV, Tb.N, Tb.Sp, and Tb.Pf values were determined in three dimensions for the left femora and L5 vertebrae by using CT. The same parameters, except Tb.Pf, were also measured in two dimensions in the proximal region of left tibiae and L3 vertebrae by histomorphometry performed in trichromestained bone sections. This permitted evaluation of bone structure by multiple methods, as well as comparison of effects of infection in different long bones.
CT analyses of trabecular bone from distal metaphyses of femora showed that bone comprised a significantly smaller proportion of total measured tissue volume in infected mice than in mock-infected controls (Fig. 4A, BV/TV), and BV/TV ratios decreased as B. burgdorferi DNA burden increased (Table 1). Infection was also associated with significantly increased bone surface per bone volume ratios in the femora (Fig. 4B, BS/BV), which also increased with increasing bacterial DNA load (Table 1). Trabecular number (Tb.N) was significantly lower in femora of infected animals ( Fig. 4C) and was negatively correlated with B. burgdorferi DNA burden (Table 1). Moreover, the trabecular pattern factor (Tb.Pf) parameter was greater in infected femora ( Fig. 4D) and increased with bacterial DNA load (Table 1), indicating more poorly connected spongy lattices (33). A corresponding increase in trabecular separation (Tb.Sp) that was positively correlated with flaB copy number was observed in infected femora (Table 1; Fig.  4E). Histomorphometry analyses of trabecular bone from trichrome-stained sections of proximal tibiae also revealed significant reductions in BV/TV in infected mice (Fig. 4A). BS/BV (Fig. 4B) and Tb.Sp (Fig. 4E) values increased by a similar percentage in femora (CT) and tibiae (histomorphometry) of infected animals, but the changes observed in tibiae were not significant, possibly due to inherently greater variations in histomorphometry sectioning and measurements. Tb.N did not differ in tibiae and femora of mock-infected and infected mice (Fig. 4C). Collectively, these analyses indicated that trabecular bone in the long bones of B. burgdorferi-infected mice was more porous and was composed of smaller, more poorly connected trabeculae, consistent with the observed osteopenia.
In contrast to the findings with long bones, neither CT nor histomorphometry nor flaB copy number correlation analyses of vertebrae revealed significant infectiondependent changes in BV/TV, BS/BV, Tb.Sp, or Tb.Pf (Fig. 4A, B, and D; Table 1). Representative sections also showed that bone volume was reduced more dramatically by infection in tibiae than in vertebrae ( Fig. 4F and G). Additionally, Tb.N was similar in CT-analyzed vertebrae from both experimental groups but was significantly greater in the vertebrae of infected than mock-infected mice when analysis was performed by histomorphometry (Fig. 4C). The discrepancy in these results suggests that the effect of infection on trabecular number was likely mild and was possibly less frequently observed in some vertebrae than others, likely due to a low B. burgdorferi DNA load in the vertebrae compared (Fig. 2). Together, these data showed that B. burgdorferi infection did not affect vertebral trabecular bone microarchitecture as strongly as it affected long bone microarchitecture, consistent with the results of BMD assays.
To assess the effects of B. burgdorferi infection on cortical bone microarchitecture, we measured cortical bone perimeter (B.Pm), thickness, cross-sectional area, and anteroposterior (AP) and mediolateral (ML) diameters in femoral midshafts via CT. None of these properties differed significantly in infected versus mock-infected groups (see Table S1 in the supplemental material). Furthermore, three-point bending at the midshafts of the same femora revealed that infection did not affect cortical bone biomechanical properties (Table 2). Together with the results of BMD assays, these results implied that B. burgdorferi infection did not affect the structure or function of cortical bone.
Collectively, these data indicated that B. burgdorferi infection significantly disrupted the trabecular microarchitecture of long bones without affecting the vertebrae or cortical bone.
Bone loss in long bones of B. burgdorferi-infected mice is associated with reduced numbers of osteoblasts but does not affect osteoblast activity or osteoclast numbers. Healthy bone remodeling in response to stressors results from balance between bone-degrading activities of osteoclasts and the bone-building activities of osteoblasts (34). Increased osteoclastogenesis and osteoclast activity can result in increased number and size of osteoclasts at the bone surface, which lead to net bone resorption. This is frequently observed in bone loss due to local or systemic inflammation (8). In contrast, osteoblast apoptosis (i.e., decreased osteoblast numbers) and inhibition of osteoblast function (i.e., lower osteoid production, slower mineralization rate) are more prominent in bone loss associated with direct infection of bone (i.e., osteomyelitis) (9).
To determine whether osteopenia in trabecular regions of long bones in B. burgdorferi-infected mice was associated with changes in bone cell numbers, we measured the osteoclast number per bone surface (Oc.N/BS) in tartrate-resistant acid phosphatase (TRAP)-stained sections of right proximal tibial metaphyses and L4 vertebrae, and also osteoblast numbers per bone surface (Ob.N/BS) in trichrome-stained sections of the left proximal tibial metaphyses and L3 vertebrae (Fig. 5A and B). We also measured the ratio of total osteoclast surface to total bone surface (Oc.S/BS) and osteoid surface to bone surface (OS/BS) from the same sections ( Fig. 5C and D).
Since B. burgdorferi-induced pathology in most tissues is primarily due to host inflammatory responses to these bacteria (3), we expected that bone loss in infected mice would be accompanied by significant osteoclast-driven bone resorption. However, we found that osteoclast numbers (Fig. 5A) and coverage of bone surfaces (Fig. 5C) did not significantly increase in either the tibiae or vertebrae of infected mice. Instead, we observed that bone loss was accompanied by reduced osteoblast numbers at bone surfaces in tibiae (Fig. 5B), although reductions in osteoblast numbers did not appear to depend on bacterial DNA burden (Table S3). Osteoblast numbers in vertebrae were significantly greater than in tibiae (Fig. 5B), suggesting that differences in osteoblast numbers possibly contributed to differences in the bone loss phenotypes in these bones.  Table S2 in the supplemental material. Statistics used included a two-way analysis of variance with the Holm-Sidak posttest (A and B) and two-tailed unpaired t tests (C to E). *, P Ͻ 0.05 versus mock controls within bone and measurement type; #, P Ͻ 0.05 for long bones versus vertebrae.
Despite reduced osteoblast numbers in the tibiae (Fig. 5B), osteoid production in bones of infected mice was not impaired (Fig. 5D). This implied that infection did not inhibit the activity of the osteoblasts present in bone. To further test this conclusion, rates of mineralized bone formation over time were examined via dynamic histomorphometry in calcein-labeled bone (Fig. 5E to G). Mineralization apposition rates (MARs) in infected and mock-infected animals were similar (Fig. 5E), as was the ratio of mineralization surface area to osteoid surface area (MS/OS) (Fig. 5F). Furthermore, infection did not reduce bone formation rate/osteoid surface ratios (BFR/OS) (Fig. 5G), implying that growth rates for mineralized bone were not affected when adjusted for osteoid volume. Thus, we concluded that although osteoblast numbers were reduced, osteoid production and mineralization were not affected by infection. Collectively, these data indicated that B. burgdorferi infection-dependent bone loss was associated with reductions in osteoblast numbers in long bones, but not with increased osteoclastogenesis, increased osteoclast activity, or impaired osteoblast activity.

DISCUSSION
In this study, we found that B. burgdorferi infection in mice is associated with infectious dose-and time-dependent reductions in ALP, a serum biomarker of bone formation in mice. Infection is also associated with B. burgdorferi DNA burdendependent osteopenia and disrupted trabecular microarchitecture in long bones but not in vertebrae. These effects were not associated with increased osteoclastdependent bone resorption, inhibition of osteoblast synthesis of osteoid cells, or bone mineralization. Instead, trabecular bone loss was associated with osteoblast number reductions in long bones but not in vertebrae. It is likely that differences in the effect of B. burgdorferi infection in long bones and vertebrae were due at least in part to differences in bacterial DNA burdens in these bones. Many bone loss parameters were significantly correlated with flaB copy number in long bones but not in vertebrae, suggesting that B. burgdorferi did not appear to colonize and/or persist as effectively in vertebral bone. However, other differences between long bones and vertebrae may have also contributed to differences in susceptibility to B. burgdorferi infection and infection outcomes, since long bones and vertebrae have been observed to be affected differently by systemic factors (e.g., rosiglitozone treatment) for reasons that remain unknown (30).
Although bone pain and bone involvement have been described in patients with Lyme disease (5, 12-18, 20, 21) and B. burgdorferi has been detected in human bone and bone marrow (10)(11)(12)(13), the possibility that B. burgdorferi may infect and cause pathology in bone at sites other than articular surfaces has not been widely investigated. We found that B. burgdorferi infection in mice causes a level of osteopenia in the trabecular regions of long bones that would be considered a clinically significant finding in humans. However, structural and biomechanical properties of cortical bone were not affected by infection. It will be important to determine if conditions such as prolonged B. burgdorferi infection or infection in the context of preexisting osteopenia or osteoporosis would eventually result in a clinically symptomatic phenotype in cortical bone. It may also be useful to determine if B. burgdorferi infection in humans is associated with increased rates and severity of osteopenia and osteoporosis, and/or increased risk of fracture outcomes.
Bone loss results from an imbalance between the bone-resorbing activities of osteoclasts and bone-building functions of osteoblasts. Our data suggest that bone loss at 4 weeks postinfection was not primarily due to increased osteoclastogenesis or osteoclast activation, although it is possible that osteoclast-driven bone loss occurred earlier in infection. This is plausible, because tumor necrosis factor alpha (TNF-␣), interleukin-1 (IL-1), IL-6, and IL-17, which are systemically upregulated in response to B. burgdorferi infection (35)(36)(37)(38), are some of the major inflammatory cytokines that stimulate osteoclastogenesis and osteoclast activation (8,39). TNF-␣, IL-1, and IL-6 also suppress differentiation of osteoblasts (8), which is consistent with our observation that osteoblast numbers were reduced in bones which exhibited osteopenia. However, regardless of whether increased osteoclast-dependent resorption was responsible for early bone loss in infection, data from the present study suggest that osteoblast deficiency makes important, long-lasting contributions to bone loss.
The major causes of osteoblast-related abnormalities in bone disease are loss or reduction of signals required for osteoblast differentiation (e.g., hormones or local regulatory proteins, such as bone morphogenetic proteins), differentiation of bone marrow mesenchymal stem cells into cells of nonosteoblastic lineage, premature osteoblast cell death, and inhibition of osteoblast activities (i.e., osteoid synthesis and mineralization) (40). We did not find that osteoid production or mineralization by the osteoblasts present in bones of infected mice were impaired, implying that osteoblast abnormalities occurred instead due to suppression of osteoblastogenesis or reduced osteoblast viability. It is unknown if B. burgdorferi infection affects mesenchymal stem cell differentiation or other major regulators of osteoblastogenesis, such as growth hormones, growth factors, estrogen, and parathyroid hormone (40). However, one of B. burgdorferi's complement-regulating surface proteins, CRASP-1, is known to bind bone morphogenetic protein-2 (BMP-2) (41), a potent inducer of osteoblast differentiation (40). Treatment with vitamin D, another key regulator of osteoblastogenesis, has also been shown to reduce the severity of Lyme arthritis in mouse models (42). Additionally, direct or indirect bacterial induction of cell death in osteoblast lineage cells is a known important mediator of bone loss seen in osteomyelitis caused by other bacterial species (9). B. burgdorferi is known to induce cell death in a number of cell types, including glia, neurons, oligodendrocytes, lymphocytes, monocytes, and macrophages (43)(44)(45)(46)(47)(48), but the effect of this bacterium on osteoblast biology and cell death have not yet been examined. These literature observations, together with our findings, warrant future investigations on the effect of B. burgdorferi infection on osteoblastogenesis and osteoblast viability.
In summary, this paper presents the first animal study directly demonstrating an effect of B. burgdorferi infection on bone tissue deterioration at sites other than articular surfaces involved in Lyme arthritis. Since these observations may have implications for our understanding of Lyme disease in humans, further investigation of the mechanisms underlying bone loss in response to B. burgdorferi infection is warranted. B. burgdorferi cultivation and mouse infections. The B. burgdorferi strain used in this study was B31 5A4-derived GCB726 (49). As described elsewhere (50), this strain was cultivated to log phase in Barbour-Stoenner-Kelly II (BSK-II) medium before infection experiments. At 12 weeks of age, randomly assigned mice were inoculated subcutaneously at the dorsal lumbar midline with B. burgdorferi or BSK-II medium alone (mock infection). With the exception of the experiments presented in Fig. 1 and also Fig.  S1 in the supplemental material, mice were infected with 10 4 B. burgdorferi for a period of 4 weeks. Each of these mice was also intraperitoneally injected with freshly prepared, 30 mg/kg 0.6% calcein green (Sigma-Aldrich Canada, Oakville, ON) at 9 and 2 days prior to sacrifice, to label newly formed bone.

MATERIALS AND METHODS
Harvesting and preparation of sera and bones. Blood was collected by cardiac puncture from anesthetized mice before euthanasia by cervical dislocation, as described previously (50). Serum was harvested after overnight incubation at 4°C by centrifugation at 10,000 ϫ g for 10 min.
The following bones were collected for all other analyses in this study: femora, tibiae, and L3 to L6 vertebrae. The left femora and L5 vertebral columns were cleaned of adherent soft tissue, wrapped in saline-soaked gauze, and stored at Ϫ20°C. These bones were thawed overnight at 4°C before DXA, CT, and biomechanical analyses. Tibiae were excised and clipped at the midshaft into proximal and distal halves. Proximal metaphyses of the right tibiae and L4 vertebral columns were fixed immediately in 10% buffered formalin (Sigma) and decalcified in 14% EDTA (pH 7.4) at 4°C for 21 days, with solutions replaced every 2 days. These samples were then processed by the Toronto Centre for Phenogenomics (see the "Bone histomorphometry" section). Proximal metaphyses of the left tibiae and L3 vertebral columns were fixed in 70% ethanol and stored at 4°C until submission to the Toronto Centre for Phenogenomics for processing (see details in the "Bone histomorphometry" section). Distal tibial halves and L6 vertebral bodies cleaned of adherent soft tissue were stored at Ϫ80°C until DNA extraction for qPCR analysis. All extractions were performed using bone distal from joints to ensure that any bacterial DNA detected was not joint derived.
Serum ALP activity assays. Serum ALP activity was measured in technical duplicates using a colorimetric ALP assay kit (Abcam, Cambridge, UK) following the manufacturer's instructions.
DNA extraction and qPCR measurement of B. burgdorferi DNA copy number. For DNA extraction, bones were thoroughly cleaned of peripheral soft tissue, and bone marrow was removed by flushing with a syringe. Cleaned, rinsed bones were then pulverized with a mortar and pestle on dry ice. DNA was extracted from pulverized bone using Qiagen DNeasy tissue extraction kits (Toronto, ON, Canada) as per the manufacturer's instruction. Concentration of extracted DNA were measured using a Nanodrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Measurement of numbers of copies of B. burgdorferi flaB DNA by qPCR was performed in technical sextuplicates as described previously (50), using triplicate logarithmic dilutions (10 1 to 10 6 copies) of plasmid pTM222 (which contained a copy of the flaB DNA sequence) as a standard curve. The mean of technical replicates was used for subsequent analyses, and findings were expressed as the ratio per 1,000 copies of mouse nidogen DNA measured by qPCR from the same sample. For each bone type, the median flaB:nidogen ratio from mock-infected mice was subtracted from flaB:nidogen ratios from infected and mock-infected animals.
qPCR for mouse nidogen was performed with logarithmic dilutions of plasmid standard pTM221 containing the nidogen target amplification sequence (standard range, 1 to 10 8 copies/reaction mixture). pTM221 (Escherichia coli strain GCE1844) was cloned by PCR amplifying the nidogen target sequence from mouse genomic DNA with primers nido.F (5=-CGAGCCACAGAATACCATCC-3=) and nido.R (5=-GGA CATACTCTGCTGCCATC-3=) into the pJET1.2/Blunt cloning vector using the CloneJet PCR cloning kit (Thermo-Fisher Scientific), followed by sequencing of the insert. qPCR was performed as described for flaB (50), using primers nido.F and nido.R (51). qPCR conditions were 95°C for 5 min, followed by 50 cycles of 98°C for 5.75 s, 60°C for 2.75 s, and 86°C for 5 s, and then melt curve analysis ranging from 60°C to 95°C. Average nidogen copy numbers obtained from triplicate reaction mixtures were used for normalization of flaB copy numbers.
BMD and morphometry measurements. Measurement of areal bone mineral density (aBMD, in grams per centimeter squared) in left femora and L5 vertebral bodies was performed by DXA using a PIXImus mouse bone densitometer (GE Medical Systems, Madison, WI). Measurements of volumetric bone mineral density (vBMD, in grams per cubic centimeter) and morphometry from trabecular and cortical bones of left femora, as well as from L5 vertebral bodies, were performed after DXA analysis by CT, using a SkyScan 1174 CT scanner (Bruker microCT, Kontich, Belgium). Images were acquired at 50 kV and 800 A, and a 0.25-mm aluminum filter was applied for noise removal. Two hydroxyapatite (HA) phantoms were scanned and used to calculate vBMD. Structural indices were computed using the SkyScan CT Analyzer software. Following image acquisition, three-dimensional reconstruction was performed to calculate bone microarchitecture parameters. Morphometry measurements were performed with global thresholding to exclude soft tissue (trabecular bone, 0.9 g HA/cm 3 ; gray value, 65; cortical bone, 1.1 g HA/cm 3 ; gray value, 100).
For trabecular bone analyses, distal femoral metaphyses and L5 vertebrae were scanned at a voxel size of 6.1 m 3 . The volume of interest (VOI) for femora consisted of a 400-slice (2.4-mm) section extending 50 slices (0.3 mm) from the first cartilage bridge of femur metaphyses. The VOI for L5 was between the growth plates, with a 50-slice (0.3-mm) offset from each. For cortical bone analyses, femoral middiaphyses were scanned at 11.6 m 3 . The VOI for cortical bone was a 100-slice (1.16-mm) section at the middiaphysis.
Biomechanical testing. After CT analysis, biomechanical properties of the left femorae were assessed by three-point bending at the midshaft, using an ElectroPuls E1000 machine (Instron Corp., Canton, MA, USA) with a 100-N load cell. Femora were placed anteriorly face up on two supports, 7 mm apart. A pretest of 10 cycles (maximum, 5-N load; minimum, 2-N load) was applied to each femur. A destructive load was then vertically applied to the midshaft at 2.0 mm/min until failure. Load deformation curves derived from biomechanical testing were generated using the accompanying Bluehill 3 software (Instron). Parameters were calculated using geometry measurements for cortical bone from the CT data.
Bone histomorphometry. Tartrate-resistant acid phosphatase (TRAP) staining for identification of osteoclasts was performed on 5-m-thick sections cut from formalin-fixed, decalcified right tibiae and L4 vertebrae embedded in paraffin. Ethanol-fixed, undecalcified left tibiae and L3 vertebrae were embedded in methylmethacrylate, sectioned, and stained with Goldner's trichrome (4-m-thick sections) or left unstained for measurement of calcein fluorescence (7-m-thick sections). All histology services were performed by the Toronto Phenogenomics Centre.
Static and dynamic histomorphometry analyses of TRAP-stained, trichrome-stained, and calcein green-labeled sections were performed manually using BioQuantOsteo v11.2.6. All histomorphometric measurements were performed and analyzed following guidelines from the American Society of Bone and Mineral Research (52).
Statistical analyses. Statistical analyses were performed in Prism v.6.0 (GraphPad Software, La Jolla, CA, USA). Specific tests used for analyses are indicated in the figure legends and table footnotes. Pearson correlation analyses were performed for log-transformed flaB copy numbers per 1,000 nidogen copies and various bone parameters from each mouse. Data were considered statistically significant at a confidence level of 95% (P Ͻ 0.05).