Developmental Defects in the Cxcr2 KO FVB/N Mice Cause Decreased Survival, Body Weight, and Body Fat

We backcrossed transgenic Cxcr2 -/- mice from the C57BL/6 background (Cxcr2 KO (C57BL/6)) onto the FVB/N background for more than 10 generations (Fig. 1a). After backcrossing the mice, we observed developmental defects as well as significantly decreased survival of Cxcr2 KO FVB mice compared to Cxcr2 wildtype (WT) and heterozygous (Het) mice (Fig. 1b). Cxcr2 KO animals were often runted with rounded facial features (Fig. 1c).

We measured body weight in 3-, 5-, and 8-week-old virgin male and female mice (Fig. 1d; n = 6/genotype). In male animals, the mean body weight in 3-week-old Cxcr2 KO mice was only 10.98 g, compared to 15.10 g in 3-week-old WT mice (P = 0.0011). Body weight difference was not statistically significant in 5-week-old male Cxcr2 KO compared to WT mice, while 8-week-old Cxcr2 KOs were significantly smaller, with an average weight of 25.45 g in Cxcr2 KO versus 31.19 g in WT mice, respectively, an 18% reduction in normal body weight (P < 0.0001).

The body weight differences were more pronounced in female KO mice than in males. The Cxcr2 KO mice were notably smaller than WT and Het littermates from birth and had delayed growth up to 8 weeks. Significantly, the mean body weight in 3-week-old Cxcr2 KO mice was only 10.85 g, compared to 16.67 g in 3-week-old WT mice. At 5 weeks, Cxcr2 KO body weight was 15.07 g, on average, versus 21.18 g for WT, while at 8 weeks old the average weight was 14.64 g versus 25.28 g in Cxcr2 KO and WT mice, respectively. Thus, by 8 weeks of age, female Cxcr2 KO mice were only 58% of the normal weight of their WT littermates (P < 0.0001). Therefore, Cxcr2 deficiency significantly decreased the ability to grow and thrive in these mice.

To assess body fat and lean content, we used EchoMRI body composition analysis (EchoMRI LLC, Houston, TX) in 3-, 5-, and 8-week-old male and female mice (Fig. 1e; n = 6/genotype). Younger male mice did not significantly differ in fat or lean content, while in 8-week-old males, the fat content decreased by 35% in Cxcr2 KO mice compared to WT mice, with a corresponding increase in lean body content. Like body weight, the body composition differences were more prominent in female mice. At 3 weeks, the body fat content (as a percentage of body weight) in female Cxcr2 KO mice was not significantly different from WT. However, fat content in the KO mice markedly decreased as they aged, with a 40% reduction compared to WT by 5 weeks and a 55% decrease in body fat compared to WT at 8 weeks.

Accordingly, lean content did not differ significantly in 3- or 5-week mice but increased 14% in 8-week-old KO mice versus WT littermates.

Fig. 1

Breeding strategy, survival, and body composition in Cxcr2 KO mice. a Breeding scheme to generate Cxcr2 KO mice on FVB/N background. b Kaplan-Meier plot showing survival of healthy Cxcr2 WT, Het, and KO mice. Significance was determined using the log-rank (Mantel-Cox) test. c Representative photograph of female Cxcr2 WT, Het, and KO FVB/N mice. d Body weight and e EchoMRI body composition analysis of body fat and lean body mass in 3-, 5-, and 8-week-old male (top) and female (bottom) Cxcr2 WT, Het, and KO mice. Statistical analysis by ANOVA. Lines indicate mean with standard deviation. Significance levels for differences are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns: not significant (p ≥ 0.05). N = 1716 mice (469 WT, 908 Het, 339 KO) for survival analysis; N = 6 mice per group for body composition analysis

Increased Serum Cytokines and Bone Resorption Marker Expression in Cxcr2 KO Mice

Given the considerable structural phenotypic abnormalities of the Cxcr2 KO mice, we began to assess the differences in cytokine expression and bone development. We profiled the serum from WT and Cxcr2 KO mice by cytokine array (32-Plex Cytokine Array, Eve Technologies) to assess the consequence of Cxcr2 deficiency on systemic cytokine signaling (Fig. 2a). Cxcr2 KO serum contained significant elevation in multiple cytokines (P < 0.0001), including G-CSF, M-CSF, GM-CSF, VEGF, TNFα, IFNγ, LIF, IL-3, IL-4, IL-6, IL-7, IL-10, IL-12p70, IL-13, IL-15, IL-17, CXCL1, and CXCL9. We also measured cytokine expression of G-CSF and CXCL5 by ELISA (Fig. 2b, left). Increased G-CSF levels in Cxcr2 KO serum were validated by ELISA (P = 0.0003). Due to the nonspecific detection of LIX/Cxcl5 in the Luminex panel, we also quantified CXCL5 levels in WT and KO serum by ELISA (Fig. 2b, right). CXCL5 was significantly elevated in Cxcr2 KO mice (P = 0.0410), which is consistent with a lack of receptor binding, which elevated chemokine levels in circulation. These data support a role for Cxcr2 in the suppression of cytokines in serum. These factors are candidate cytokine regulators of bone remodeling that crosstalk with Cxcr2 signaling. We assessed the systemic levels of bone turnover markers in Cxcr2 KO mice by serum analysis. Type I collagen is the main protein of the bone matrix. During bone formation, type I collagen is cleaved into procollagen type 1 N-terminal propeptide (P1NP) and C-terminal (P1CP). Elevated serum P1NP levels indicate increased bone formation. In contrast, the C-terminal telopeptide of type 1 collagenase (CTX-1), an accepted clinical marker for fracture risk, is a byproduct of osteoclast activity that increases during bone resorption [26].

Because P1NP and CTX-1 are used as standard bone formation and resorption markers, respectively, for assessing bone remodeling activity, we analyzed P1NP and CTX-1 levels by ELISA in serum (Fig. 2c). While P1NP levels were not different in WT and KO mice, CTX-1 was significantly elevated in KO serum, indicating excess bone resorption with Cxcr2 deletion. Specifically, the mean CTX-1 level measured in WT serum was 27.96 ng/mL, compared to 52.04 ng/mL in KO serum (P = 0.0384).

We also measured the levels of RANKL and OPG, critical regulators of osteoclastogenesis, in conditioned media collected in ex vivo cultures from Cxcr2 WT or KO tibia and femur bones. RANKL and OPG levels then were used to calculate the localized RANKL/OPG ratio, an indicator of bone resorption index (Fig. 2d). Both RANKL and OPG decreased significantly in Cxcr2 KO bone versus WT bone. The mean RANKL concentration was 70.32 ± 31.66 pg/mL in WT and 39.74 ± 10.57 pg/mL in KO bone conditioned media (P = 0.0303). OPG concentration was 16,215.57 ± 4349.43 pg/mL in WT and 10,784.21 ± 3763.25 pg/mL in KO bone conditioned media (P = 0.0185). This is a reduction by 33.49% in RANKL and 43.4% in OPG concentration in Cxcr2 KO bone compared to WT. However, the OPG/RANKL ratio did not differ between Cxcr2 WT and KO bone. Thus, we focused on assessing the role of Cxcr2 in bone resorption and bone strength.

Fig. 2

Cxcr2 deletion expands the detected pro-inflammatory cytokines and alters systemic bone turnover. a Log-2 transformed fold change in cytokine expression (KO/WT) in Cxcr2 WT versus KO serum, quantified by Eve Technologies 32-Plex Cytokine Array. N = 6 mice per genotype. b Serum concentration of G-CSF and CXCL5 in Cxcr2 WT and KO mice. N = 3–8 mice per genotype. c Serum concentration of osteoblast and osteoclast activity markers P1NP and CTX-1 in Cxcr2 WT and KO mice. N = 4–5 mice per genotype. d Concentration of RANKL and OPG and OPG/RANKL ratio in Cxcr2 WT and KO bone conditioned media. N = 8 samples per group. Error bars represent standard deviation. Analysis of Cxcr2 WT versus KO using Welch’s t-test. Significance levels for differences are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 4–6 mice per group

Decreased Cortical Bone Area and Bone Strength in Femurs of Cxcr2 KO Mice

To examine the effects of Cxcr2 deficiency on bone structure, we quantitatively evaluated three-dimensional cortical bone in femurs of 8-week-old male and female Cxcr2 KO mice using µCT analysis (Fig. 3a). We quantified cortical bone area, polar moment of inertia, and section moduli (Fig. 3b and c).

Cortical bone area did not vary by sex but decreased significantly by deletion of Cxcr2 (P < 0.0001). The mean cortical area fraction (% BA/TA) was 74.72% and 75.24% in WT males and females and 75.80% and 73.65% in Het males and females, respectively. Meanwhile, Cxcr2 KO cortical % BA/TA decreased to 70.22% in males and 69.41% in females, which is a 7% reduction in cortical bone area in KO mice compared to WT (Fig. 3b, left).

Cortical polar moment of inertia (pMOI), which is a measure of the distribution of material and directly affects torsional stiffness, decreased significantly in Cxcr2 KO bones regardless of sex (Fig. 3b, right). Specifically, the mean pMOI decreased by 23.6% in male and 30.1% in female KO femurs compared to WT bones.

Section moduli (Imax/Cmax and Imin/Cmin), which are indicators of bone resistance to bending, were similarly significantly decreased in KO mice compared to WT mice (Fig. 3c; P = 0.0002) and were independent of sex. The major axis section modulus (Imax/cmax) was 0.269 ± 0.027 and 0.263 ± 0.028 in WT males and females and 0.259 ± 0.048 and 0.232 ± 0.039 in Het males and females, respectively. Meanwhile, Cxcr2 KO exhibited a section modulus of just 0.214 ± 0.047 in males and 0.192 ± 0.027 in females. The minor axis section modulus (Imin/cmin) was 0.185 ± 0.021 and 0.181 ± 0.022 in WT males and females and 0.173 ± 0.025 and 0.157 ± 0.024 in Het males and females, respectively. Imin/cmin decreased to 0.151 ± 0.024 in KO males and 0.138 ± 0.015 in KO females. This is consistent with the changes in the pMOI and cross-sectional area.

Fig. 3

Decreased cortical bone by µCT. a Regions of interest for microCT analysis for trabecular and cortical analysis (top) and representative 3D reconstructions of cortical bone at the femur midshaft (bottom). b µCT quantification of cortical bone area fraction (left) and polar moment of inertia (pMOI) (right). c µCT quantification of major and minor section moduli (Imax/Cmax and Imin/Cmin). Lines indicate the mean. Significance levels for differences (ANOVA) are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 4–6 mice per group

Decreased Bending Strength of Cxcr2 KO Bone

We next evaluated the consequence of these geometric changes on bone mechanical properties by measuring three-point bending stiffness and strength, and mineral content. First, we conducted three-point bending tests of femurs from 8-week-old male and female Cxcr2 KO mice to assess bone strength and stiffness (Fig. 4). While there were no sex-dependent differences, Cxcr2 KO bones broke at lower ultimate loads and were less stiff than WT and Het femurs in both males and females (Fig. 4a). The ultimate breaking load of male Cxcr2 KO femurs was 19.6 ± 5.8 N compared to 29.1 ± 1.8 N for Het mice (P = 0.0009) and 28.8 ± 0.9 N for WT mice (P = 0.001), representing a 40% reduction. Similarly, the ultimate bending force in female Cxcr2 KO mice was 17.8 ± 2.7 N compared to 27.0 ± 2.8 N for Het mice (P = 0.0008) and 29.1 ± 4.4 N for WT mice (P = 0.0001), a 55% reduction in KO compared to WT bones.

The bending stiffness (i.e., modulus multiplied by moment of inertia) of male KO mice was 260.2 ± 87.7 MPa·mm4 compared to 415.4 ± 41.9 MPa·mm4 for WT mice (P = 0.004) and 433.3 ± 70.4 MPa·mm4 for Het mice (P = 0.001). Similarly, the bending stiffness of female Cxcr2 KO mice was 227.5 ± 50.9 MPa·mm4 compared to 415.4 ± 41.9 MPa·mm4 for WT mice (P = 0.00009) and 416.3 ± 69.2 MPa·mm4 for Het mice (P = 0.0002). This signifies a 49% and 63% reduction in bone stiffness in male and in female Cxcr2 KO mice, respectively, compared to WT mice.

Bone bending stiffness is dependent both on bone geometry and bone tissue modulus, which in turn depends on the mineral content [27, 28]. To test if the observed mechanical weakness in KO bones is due to differences in bone mineral content, we dried and ashed 8-week-old Cxcr2 WT, Het, and KO femurs. The bones did not differ significantly in total weight, water content, or mineral content by ash weight (data not shown and Fig. 4b). The mean percent ash weight in male Cxcr2 KO mice was 55.21 ± 8.27%, compared to 54.42 ± 8.49% in male WT and 54.81 ± 5.50% in male Het mice. Similarly, the mean percent ash weight in female Cxcr2 KO mice was 56.27 ± 3.99%, compared to 64.21 ± 17.18% in female WT and 60.72 ± 4.51% in female Het mice. Thus, it is likely that the altered mechanical behavior of Cxcr2-deficient femurs is attributable to these changes in the cross-sectional geometry. However, the nonuniform geometry and size difference of the bones hinder accurate comparison of mechanical properties [29].

Fig. 4

Reduced bone strength in Cxcr2 KO mice. a Three-point bending test. Femur bending strength was measured by (Left) ultimate load and (Right) stiffness. b Bone mineral content by percent ash weight of dry bone weight determined after bone ashing. Lines indicate mean. Significance levels for differences (ANOVA) are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 6 mice per group

The Trabecular Bone Architecture was Degraded and Volume Fraction Decreased in Cxcr2 KO Mice

We next used µCT to evaluate trabecular architecture in the proximal femur (Figs. 3a and 5a). While the trabecular number and spacing were unchanged based on Cxcr2 expression, the average trabecular thickness decreased 26% in male and 34% in female Cxcr2 KO mice compared to in WT mice (Fig. 5b, P < 0.0001).

Bone volume fraction (BV/TV) decreased significantly in Cxcr2 KO mice (Fig. 5c). BV/TV was 27.2% and 35.3% in WT males and females and 28.2% and 32.0% in Het males and females, respectively. In contrast, Cxcr2 KO BV/TV decreased to 14.9% in males (45% decline) and 17.9% in females (49% decline) (P < 0.0001).

The structural model index (SMI) is a measure of the trabecular structure that distinguishes more robust plate-like from more damage-prone rod-like morphologies [30]. The SMI was sex-dependent (P = 0.0027) and increased significantly in Cxcr2-deficient trabecular bone (P < 0.0001), with a greater difference in male mice (Fig. 5c). In males, SMI was 0.961 ± 0.488 in WT and 0.854 ± 0.375 in Het, while Cxcr2 KO SMI increased to 2.580 ± 1.169 (P < 0.0001). In females, SMI was 0.264 ± 0.207 in WT and 0.650 ± 0.483 in Het mice, with Cxcr2 KO SMI increased to 1.634 ± 0.278 (P < 0.0001), indicating a more rod-like trabecular architecture in Cxcr2 KO compared to in WT mice. While SMI is correlated with BV/TV within anatomic sites, we analyzed the data by linear regression. SMI and BV/TV were significantly correlated in WT mice (both male and female) and in male Cxcr2 KO mice (Fig. 5c). We compared the slope of SMI versus BV/TV within genotypes by sex, and only WT males versus females significantly differed (P = 0.0146), while Het and KO mice did not show dependence on sex. Within sexes, the correlation of SMI and BV/TV did not differ significantly by genotype. Together, the decreased Tr.Th and increased Tr.Sp and SMI in Cxcr2 KO bone demonstrate the deteriorated trabecular architecture in Cxcr2 KO mice.

Fig. 5

Decreased trabecular bone by µCT. a Representative 3D reconstructions of trabecular bone in the proximal femur of 8-week-old male and female Cxcr2 WT, Het, and KO mice. b Quantification of trabecular bone parameters. Lines indicate mean. Significance levels for differences (ANOVA) are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). c Correlation of trabecular BV/TV and structural model index is sex- and genotype-dependent. Statistical analysis by simple linear regression. N = 4–6 mice per group

Cxcr2 Deletion Increases Osteoclast Surface Area

The loss of trabecular bone density and alterations in bone morphology that were noted by µCT in Cxcr2 KO bones compared to WT bones were also detectable by H&E staining (Fig. 6a). We used histological examination of osteoclasts (TRAP) and osteoblasts (alkaline phosphatase, ALP) to evaluate further the bone remodeling activity in Cxcr2 WT and KO bones.

We used TRAP staining to identify and analyze osteoclasts on the cortical and trabecular bone surfaces of female Cxcr2 WT and KO proximal tibiae extending 1 mm from the growth plate (Fig. 6b). Few osteoclasts were seen on the cortical bone surface and did not differ in number or area between WT and KO mice. The total surface area of trabecular bone did not differ significantly by genotype; the average total trabecular bone perimeter (B.Pm) measured was 4.61 mm in WT and 3.80 mm in KO bone sections (data not shown). In trabecular bone, osteoclast number (N.Oc/B.Pm) was not different between WT and KO mice (Fig. 6b). However, the osteoclast surface area (% Oc.Pm/B.Pm) increased significantly in Cxcr2 KO bone compared to in WT bone (Fig. 6b). Thus, osteoclast length in Cxcr2 KO trabecular bone significantly increased (P = 0.0021), covering 26.23 ± 5.39 μm of bone surface on average, versus an average osteoclast size of 20.22 ± 3.03 μm in WT bone (data not shown). While we did not directly analyze the number of nuclei per osteoclast to address precursor number or fusion specifically, a larger osteoclast area is associated with increased multinucleation and resorption capacity [31,32,33]. This is consistent with the elevated levels of CTX-1 and the observation of decreased trabecular thickness by µCT and implicates Cxcr2 in regulating osteoclast differentiation and/or activity.

To assess the impact of Cxcr2 deletion on bone formation, we used alkaline phosphatase (ALP) IHC staining to quantify osteoblasts on the trabecular bone surface in Cxcr2 WT and KO tibias (Fig. 6c). Osteoblasts were quantified by analyzing the ALP-positive (ALP+) surface area in tibial sections in the trabecular region extending 1 mm distal from the tibial growth plate. The ALP + surface perimeter and osteoblast surface (% bone perimeter) did not significantly differ between Cxcr2 WT and KO mice. Specifically, the total osteoblast surface area in Cxcr2 KO mice was 26.04 ± 8.18% of the bone perimeter, whereas the wild-type mice exhibited an average of 25.06 ± 6.73% (P = 0.77). This indicates that Cxcr2 deficiency does not significantly alter the proportion of the bone surface covered by osteoblasts in the tibia, which is consistent with the unchanged levels of P1NP in Cxcr2 KO versus WT serum (Fig. 2b).

Together, these data indicate that Cxcr2 is not required for the maintenance of osteoblast surface area, while Cxcr2 deficiency disrupts the balance of bone remodeling to favor osteoclast-mediated bone resorption by enhancing osteoclast size and activity [33].

Fig. 6

Increased trabecular osteoclast surface area by histological analysis of bone cells in Cxcr2 WT and KO proximal tibia. a Representative images of H&E (top), tartrate-resistant acid phosphatase (TRAP) staining (middle), and alkaline phosphatase (ALP) immunohistochemistry (bottom) of Cxcr2 WT and KO tibia. Scale bar = 200 μm. b Quantification of TRAP + osteoclast number and surface (% Oc.Pm) per cortical (top) and trabecular (bottom) bone perimeter. c Quantification of ALP + bone perimeter and osteoblast surface (%Ob.Pm) per trabecular bone perimeter. Lines indicate median. Statistical analysis of WT versus KO by Welch’s t-test. Significance levels for differences are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 8–13 bone samples per genotype

The Number of Megakaryocytes in Bone Decreases by Cxcr2 Deficiency

During histological examination by H&E of Cxcr2 WT and KO bone marrow, we saw a notable decrease in megakaryocytes in Cxcr2 KO bones. These differences were quantified by immunohistochemistry staining of the megakaryocyte cell surface marker CD61 to analyze megakaryocyte number (Fig. 7a). Indeed, bones from Cxcr2 KO mice had significantly fewer megakaryocytes in the metaphysis and diaphysis of the tibia compared to WT bones (Fig. 7b). Specifically, the average megakaryocyte counts in the metaphysis were 30 ± 10 megakaryocytes in WT and 14 ± 4 megakaryocytes in Cxcr2 KO mice per 500 µm2 field of view (P = 0.0009). In the diaphysis, the average number was 42 ± 12 megakaryocytes in WT mice and 23 ± 7 megakaryocytes in Cxcr2 KO mice per field of view (P = 0.0008).

Fig. 7

Histological analysis of megakaryocytes in Cxcr2 WT and KO tibia. a Representative images of immunohistochemistry for CD61 + megakaryocytes in metaphysis (top) and diaphysis (bottom) of Cxcr2 WT and KO tibia. Scale bar = 200 μm. b Quantification of CD61 + megakaryocyte number counted in bone marrow in metaphysis (top) and diaphysis (bottom). Lines indicate median. Statistical analysis used Welch’s t-test. Significance levels for differences are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 7–9 bone samples per genotype

Increased Neutrophils and RBC Heterogeneity in the Serum of Cxcr2 KO Mice

Cxcr2 is a known critical regulator of neutrophil homeostasis. To determine the impact of Cxcr2 deletion on the systemic immune environment, we analyzed the circulating immune cell profiles in Cxcr2 WT and KO mice using complete blood count with differential analysis via Element HT5 hematology analyzer (Fig. 8).

Consistent with previous reports of granulocyte expansion in Cxcr2 KO mice [18], the circulating neutrophil population expanded significantly in Cxcr2 KO (FVB/N) mice compared to WT mice (Fig. 8a). Specifically, the average neutrophil count in WT animals was 1.05 × 103/µL, compared to 4.06 × 103/µL in KO mice. Similarly, the percentage of neutrophils of total white blood cells was 20.25% in WT mice and 56.75% in Cxcr2 KO mice.

While lymphocyte numbers were unchanged with Cxcr2 deletion (3.50 ± 1.71 × 103/µL in WT compared to 2.23 ± 1.25 × 103/µL in KO), the expanded neutrophil population coincided with a significant decrease in lymphocyte percentage of total white blood cells, from 64.98% in WT mice to only 29.48% in Cxcr2 KO mice. Monocytes, basophils, and eosinophils were unaffected by Cxcr2 deficiency, and there was no difference in total white blood cell counts (Fig. 8a and b).

Fig. 8

Cxcr2 deletion alters the immune microenvironment. Complete blood count with differential analysis via Element HT5 Hematology Analyzer of cardiac blood from Cxcr2 WT and KO mice. a Quantification of count (Top) and percent of total white blood cells (WBCs) (Bottom) of neutrophils, lymphocytes, monocytes, basophils, and eosinophils. Percentages were compared by two-way ANOVA and expressed as mean ± standard deviation. Quantification of b total white blood cell (WBC) count and c red blood cell (RBC) count, hematocrit (HCT) (% red blood cells of total cells), mean corpuscular volume, and RBC width distribution (percent coefficient of variation). Quantification of d hemoglobin parameters, including total hemoglobin (HGB, g/dL), mean corpuscular hemoglobin (MCH), and mean corpuscular hemoglobin concentration (MCHC); and e platelet count and volume. Lines depict the median. Statistical analysis of WT vs. KO by Welch’s t-test. Significance levels for differences are indicated: * p < 0.05, **p < 0.01, ***p < 0.001, ns: not significant (p ≥ 0.05). N = 4 per genotype

Red blood cell count, percent, and size as well as hemoglobin concentration were unchanged by Cxcr2 deletion (Fig. 8c and d). However, the RBC width distribution significantly increased in Cxcr2 KO animals (Fig. 8c), indicating anisocytosis, or increased variation in red blood cell size.

Megakaryocytes are precursor cells of platelets. Despite the decreased megakaryocytes observed in KO bone marrow, platelet count and volume were normal in Cxcr2 KO blood (Fig. 8e).

Together, these blood cell data and the serum cytokine analyses (Fig. 2) confirm roles for Cxcr2 in regulating granulocyte populations that may ultimately regulate bone remodeling.