Interleukin-6 (IL-6) family cytokines act via gp130 in the osteoblast lineage to stimulate the formation of osteoclasts (bone resorbing cells) and the activity of osteoblasts (bone forming cells), and to inhibit expression of the osteocyte protein, sclerostin. We report here that a profound reduction in trabecular bone mass occurs both when gp130 is deleted in the entire osteoblast lineage (Osx1Cre gp130 f/f) and when this deletion is restricted to osteocytes (DMP1Cre gp130 f/f). This was caused not by an alteration in osteoclastogenesis, but by a low level of bone formation specific to the trabecular compartment. In contrast, cortical diameter increased to maintain ultimate bone strength, despite a reduction in collagen type 1 production. We conclude that osteocytic gp130 signaling is required for normal trabecular bone mass and proper cortical bone composition. © 2014 American Society for Bone and Mineral Research.
Osteoporosis results from imbalanced bone remodeling in which the level of bone resorption, carried out by osteoclasts, exceeds that of bone formation, carried out by osteoblasts. This leads to reduced bone mass in both the trabecular and cortical bone compartments. Both osteoblast and osteoclast formation are stimulated by cytokines that signal through glycoprotein 130 (gp130), a signal transducer utilized by the many different interleukin-6 (IL-6) family cytokines, including IL-6, IL-11, leukemia inhibitory factor (LIF), oncostatin M (OSM), ciliary neurotrophic factor (CNTF), and cardiotrophin-1 (CT-1). Cytokine-specific knockout mouse models have established unique and necessary roles for each of these cytokines in physiological regulation of longitudinal growth, periosteal expansion, and trabecular structure,[2-9] as well as bone loss associated with inflammation and estrogen deficiency.
Although gp130 is ubiquitously expressed, stimulation of osteoclast formation by IL-6, IL-11, LIF, CT-1, and OSM depends on the presence of osteoblasts in vitro[12, 13] and is mediated by increased osteoblast receptor activator of NF-κB ligand (RANKL) mRNA expression.[12-15] Furthermore, stimulation of osteoclast formation by other agents such as IL-1, parathyroid hormone (PTH), and 1,25-dihydroxyvitamin D3, are also mediated, at least in part, by gp130. However, the role of osteoblast lineage gp130 signaling in osteoclast formation is not well defined, and has been complicated by conflicting effects on osteoclast formation by genetic deletion of individual cytokines that signal via gp130 and phenotypic differences between the sexes in vivo. For example, systemic deletion of OSM receptor (OSMR) inhibited osteoclast formation in male and female mice. A similar phenotype was reported in male mice with systemic deletion of IL-11 receptor (IL-11R), but females showed only a reduction in osteoblastogenesis. In contrast, CT-1 knockout mice had increased formation of osteoclasts with impaired activity in both males and females. Further complicating the data, systemic LIF, LIF receptor (LIFR), or gp130 deletion resulted in increased numbers of large, active osteoclasts clustered near the growth plate in neonate mice. The contrasting nature of these in vivo osteoclast phenotypes are remarkable, given the strong stimulatory effects of OSM, IL-6, CT-1, and IL-11 on osteoclast formation in vitro.[12, 13]
IL-6 family cytokines also act on the osteoblast lineage to stimulate bone formation. IL-6, IL-11, CT-1, and OSM all promote osteoblast differentiation in vitro,[9, 19] and OSM, CT-1, and LIF stimulate bone formation in vivo.[6, 9, 20] OSM, CT-1, LIF, and IL-11 also inhibit adipocyte differentiation,[5, 6, 9] suggesting an influence of these cytokines on early osteoblast precursor commitment. Furthermore, OSM, CT-1, IL-11, and LIF influence terminally differentiated osteoblasts embedded within the bone matrix (osteocytes), by suppressing their production of sclerostin, a potent inhibitor of Wnt signaling and bone formation. This indicates that IL-6 family cytokines act on osteoblasts at all stages of differentiation (eg, early osteoblasts, late osteoblasts, or osteocytes), but the relevant stages at which they support bone formation, bone resorption, or adipogenesis are unknown.
To determine the stage-specific roles of gp130 in the osteoblast lineage in both bone modeling and remodeling, we have generated two mouse models where gp130 was conditionally deleted either from the entire osteoblast lineage (Osx1Cre) or specifically in osteocytes (DMP1Cre). The data reported here indicate that gp130 signaling in the committed osteoblast lineage is not required for osteoclast differentiation in trabecular bone. Rather, the key role of gp130 signaling in the osteoblast lineage is in osteocytes, where it maintains bone formation in trabecular bone and cortical bone strength.
Subjects and Methods
All animal procedures were conducted with approval of the St. Vincent's Health Melbourne Animal Ethics Committee. Osx1Cre mice backcrossed to C57BL/6 were obtained from Carl Walkley (St. Vincent's Institute, Fitzroy, Australia) and DMP1Cre mice (containing the DMP1 10-kb promoter region) were obtained from Lynda Bonewald (University of Kansas, Kansas City, KS, USA). Floxed gp130 mice backcrossed onto C57BL/6 mice were obtained from Rodger McEver (Oklahoma Medical Research Foundation). Mice hemizygous for the Cre transgene of each strain were crossed with a gp130 flox mouse in which the transmembrane domain (exon 15) was flanked by loxP sites, resulting in ablation of intracellular gp130 signaling, as reported. For all experiments, appropriate Cre+ wild-type littermates or cousins were used as controls. Both male and female mice were analyzed, as indicated in Results, below.
Samples for histomorphometry, micro–computed tomography (µCT), RNA, and serum analyses were collected at 6, 12, and 26 weeks of age, after injection with calcein at 3 and 10 days prior to tissue collection. Twelve-week-old DMP1.gp130 and Osx1Cre.gp130 homozygous for absence of the gp130flox transgene (w/w) or homozygous for its presence (f/f) (n = 6/7 per genotype) were fasted for 12 hours prior to anesthesia with ketamine/xylazine and blood collection via cardiac puncture. Blood samples were centrifuged 10 minutes at 1500 g and serum supernatant was removed to a fresh tube and stored at –80°C until analysis for CTX-1 or P1NP (Immunodiagnostic Systems Limited; Boldon, Tyne & Wear, UK) as per manufacturer's instructions. Flushed femurs from 12-week-old DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice were harvested for RNA extraction as described.
Fluorescence-activated cell sorting
Osx1Cre.gp130w/w and Osx1Cre.gp130f/f neonates were euthanized by decapitation and calvaria were dissected out for digestion in 1:2 type II collagenase/dispase solution (1 × 5-minutes, 4 × 10-minute digestions), shaking at 37°C. Collagenase/dispase solution was removed and cells were resuspended in appropriate volume of fluorescence-activated cell sorting (FACS) buffer (1× PBS, 2% fetal bovine serum [FBS], 0.5 mM EDTA) and sorted on a FACS Aria (BD Biosciences, San Jose, CA, USA) for green fluorescent protein (GFP) fluorescence driven by Cre transgene expression. Sorted cells were harvested for RNA in Trizol (Life Technologies, Carlsbad, CA, USA) and separated and precipitated using chloroform and isopropanol. Extracted RNA was DNase treated using Ambion TURBO DNA-free kit (Life Technologies) and measured on a NanoDrop ND1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA).
Five consecutive injections of 0.2 µg mouse OSM (mOSM) or saline in 25 µL volume were performed over the calvariae at 24-hour intervals in 6-week-old female DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice via subcutaneous injection as described. Calvariae were harvested upon sacrifice, 10 days after the last injection, and analyzed by histomorphometry as described.
Histomorphometry and µCT
Histomorphometry was performed on tibial sections as described. Ex vivo µCT was performed on femoral, vertebral, and calvarial specimens using the SkyScan 1076 system (Bruker-microCT, Kontich, Belgium). Images were acquired using the following settings: 9-µm voxel resolution, 0.5-mm aluminum filter, 50-kV voltage, and 100-µA current, 2600 ms exposure time, rotation 0.5 degrees, frame averaging = 1. Images were reconstructed and analyzed using SkyScan software programs NRecon (version 184.108.40.206), DataViewer (version 1.4.4), and CT Analyzer (version 220.127.116.11). Femoral trabecular analysis region of interest (ROI) was determined by identifying the distal end of the femur and calculating 15% of the total femur length toward the femora mid-shaft, where we then analyzed an ROI of 12.6% of the total femur length. Analysis of bone structure was completed using adaptive thresholding (mean of minimum and maximum values) in CT Analyzer. Thresholds for analysis were determined manually based on grayscale values (0–255, where 0 = black and 255 = white) for each experimental group as follows: Trabecular bone 12-week-old mice: Osx1Cre.gp130, 38 to 255; DMP1Cre.gp130, 42 to 255. Trabecular bone 26-week-old mice: Osx1Cre.gp130 and DMP1Cre.gp130, 42 to 255. Cortical analyses were performed 35% above the distal end of the femur toward the femora mid-shaft, also with a 12.6% ROI with the threshold values set as follows: Cortical bone, 12-week-old mice: Osx1Cre.gp130, 63 to 255; DMP1Cre.gp130, 57 to 255. Cortical bone, 26-week-old mice: Osx1Cre.gp130 and DMP1Cre.gp130, 100 to 255. Vertebral trabecular bone was evaluated in the L6 vertebral body of 12-week-old mice, where the ROI was defined as a cylinder measuring 50% of the total height and 66% of the width of the vertebral column at the midpoint. Thresholds were as follows: Osx1Cre.gp130, 42 to 255; DMP1Cre.gp130, 50 to 255. Polarized light microscopy was performed on 100-µm-thick transverse sections collected approximately 500 µm from the tip of the distal femur using an Isomet Saw (Buehler, Lake Bluff, IL, USA), and measurements included the entire bone interface.
Semiquantitative real-time PCR
cDNA synthesis from 0.1 to 1 µg DNase-treated RNA was performed using AffinityScript (Agilent Technologies, Santa Clara, CA, USA) per the manufacturer's instructions. Stock cDNA was diluted 1:5 to 1:10 and semiquantitative real-time PCR (qPCR) was performed using an in-house master mix of 10× AmpliTaq Gold with SYBR Green nucleic acid gel stain (Life Technologies). Primers were used as published[7, 30] or designed using Primer Blast (NCBI; http://www.ncbi.nlm.nih.gov/tools/primer-blast) and reported in Table 1. Samples were dispensed onto an optically clear 96-well plate (Thermo Scientific) and run on a Stratagene Mx3000P (Agilent Technologies). Cycling conditions were as follows: (95°C for 10 minutes), (95°C for 30 seconds, 58°C for 1 minute, 72°C for 30 seconds) × 40 cycles, followed by dissociation step (95°C for 1 minute, 55°C for 30 seconds, 95°C for 30 seconds). Post-run samples were analyzed using Stratagene software MxPro and reported using linear delta threshold cycle (ΔCt) values normalized to β-2 microglobulin (β2M) or hypoxanthine phosphoribosyltransferase 1 (HPRT1) for primary mouse cells or hydroxymethylbilane synthase (HMBS) for femoral extracts.
|Gp130 (exon 15)||Forward||5′-AGAAGCCATAGTCGTGCCTGTGT-3′|
Femurs from 26-week-old male and female DMP1Cre.gp130 mice (males n = 8 w/w, n = 8 f/f; females n = 5 w/w, n = 7 f/f) were tested at the mid-shaft by three-point bending at room temperature. Load was applied in the anteroposterior (AP) direction midway between two supports that were 6.0 mm apart. Load-displacement curves were recorded at a crosshead speed of 1.0 mm/s using an Instron 5565A dual column materials testing system, using Bluehill 2 software (Instron, Norwood, MA, USA). Prior to testing they were kept moist in gauze swabs soaked in phosphate buffered saline (PBS). Ultimate force (FU; N), yield force (FY; N), stiffness (S; N/mm), and energy (work) to failure (U; mJ) were calculated from the load-displacement curves as described. The yield point was determined from the load deformation curve at the point at which the curve deviated from linear. Widths of the cortical mid-shaft in the mediolateral (ML) and AP directions were measured using digital calipers, and the average cortical thickness was determined by µCT. Combining the geometric calculations and the biomechanical test results, the material properties of each bone were calculated as described by Schriefer and colleagues to obtain ultimate strength (σ, MPa); elastic modulus (E, MPa), and modulus of toughness (u, MPa). Average load-deformation and stress-strain curves for each sex and genotype were also generated.
Reference point indentation
Local bone material properties at the femur mid-shafts from 12-week-old male DMP1Cre.gp130 mice (males n = 5 w/w, n = 9 f/f; females n = 5 w/w, n = 5 f/f) were examined by reference point indentation (RPI) using a BP2 probe assembly apparatus (Biodent Hfc, Active Life Scientific Inc., Santa Barbara, CA, USA). The BP2 assembly includes a 90-degree cono-spherical test probe with a ≤5-µm radius point and a flat bevel reference probe with ∼5-mm cannula length and friction <0.1 N. To assure consistency between measurements a line 6 mm from the femoral condyles was marked with pencil to indicate the initial probe position. To achieve a maximum indentation force of 2 N, a specific load of 300 g (reference force) was manually applied onto the femur. Two Newtons of force were applied for 10 cycles. Samples were kept partially hydrated with 70% ethanol during measurements. Internal friction, defined as the force resisting motion between the test and reference probe, was identified by the size and shape of the force-distance graph and kept at a constant of ≤0.3 N to ensure that disruptions within the probe assembly would not affect results. Test measurements were taken pre-experiment and post-experiment on polymerized methyl methacrylate (MMA) to ensure that measurements were consistent and that there were no probe faults through the course of the study. Data was discarded if graphs displayed high friction, or disruptions in the curves of the loading and unloading slopes. The distance the probe travels into the bone (total indentation distance [TDI]) is a measure of the bone's resistance to fractures; indentation distance increase (IDI) is the indentation distance in the last cycle relative to the first cycle and is correlated to bone tissue roughness; average unloading slope indicates the compressibility of the bone and can be used as a measure of stiffness.
All graphs are represented as the mean/genotype or the mean of all biological replicates. The number of animals (n) is reported for each graph, on the graph or in the figure legends. For in vitro experiments, three biological replicates were performed and averaged. All error bars are standard error of the mean. Statistical significance was considered p < 0.05. All statistics were calculated using unpaired Student's t test or one-way ANOVA (GraphPad software) or as indicated in the figure legends.
Confirmation of deletion of functional gp130
Knockdown of gp130 was confirmed in FACS-sorted Osx1Cre-GFP-expressing neonate calvarial osteoblasts, which indicated a 70% reduction in Gp130 expression in Osx1Cre.gp130f/f cells compared to controls (Fig. 1A). Because the DMP1Cre construct does not contain a reporter element, Gp130 knockdown of approximately 50% was verified in RNA from whole femurs, flushed of marrow, from both male and female 12-week-old mice (Fig. 1B).
To confirm deletion of functional response of gp130, DMP1Cre.gp130f/f female mice and their controls were injected with OSM over the calvaria to stimulate bone formation. Although DMP1Cre.gp130w/w mice formed additional bone in response to OSM, DMP1Cre.gp130f/f mice did not (Fig. 1C). Furthermore, the increase in calvarial thickness and greater mineral apposition rate (MAR) in response to OSM in DMP1Cre.gp130w/w mice was completely ablated by osteocytic deletion of gp130 (Fig. 1C).
In the osteoblast lineage, gp130 maintains trabecular bone volume
µCT analysis of trabecular bone structure revealed a significantly lower trabecular bone volume (BV/TV) and trabecular number (Tb.N) (Fig. 2A, C), with no change in trabecular thickness (Tb.N) (Fig. 2B), and significantly greater trabecular spacing (Tb.Sp) in the femur (Fig. 2D,E) and vertebrae (Supporting Fig. 1) of male Osx1Cre.gp130f/f mice compared to Osx1Cre.gp130w/w controls at 12 and 26 weeks of age. Histomorphometric analysis of the proximal tibia confirmed this phenotype (Supporting Fig. 2), and detected significantly lower trabecular thickness in Osx1Cre.gp130f/f 12-week-old mice (Supporting Fig. 2B). Osx1Cre.gp130w/f heterozygote mice demonstrated no significant alteration in trabecular structure at 12 weeks of age (data not shown). Six-week-old Osx1Cre.gp130f/f mice did not exhibit a skeletal phenotype (Supporting Fig. 2), suggesting gp130 signaling in the osteoblast lineage is not required for bone growth.
To determine the stage of osteoblast differentiation at which gp130 is most important for maintaining trabecular bone structure, we used an identical approach to assess trabecular bone structure in 12-week-old and 26-week-old DMP1Cre.gp130f/f males (Fig. 3). µCT evaluation of femoral (Fig. 3) and vertebral (Supporting Fig. 3) trabecular bone structure revealed a low BV/TV (Fig. 3A) and Tb.N (Fig. 3C) in DMP1Cre.gp130f/f mice compared to DMP1Cre.gp130w/w controls, and a significant increase in Tb.Sp at both 12 and 26 weeks of age (Fig. 3D,E). This method also detected a modest, but significantly greater trabecular thickness in DMP1Cre.gp130f/f mice compared to DMP1Cre.gp130w/w at both 12 and 26 weeks (Fig. 3B). Histomorphometric assessment of the tibia confirmed the low BV/TV and Tb.N at both 12 and 26 weeks (Supporting Fig. 4). DMP1Cre.gp130w/f heterozygous mice at 12 weeks of age demonstrated a significant increase in tibial Tb.Th (by 12.7%, p < 0.001) and Tb.Sp (by 34.5%, p < 0.01) by µCT (data not shown).
Neonate Osx1Cre.gp130f/f and DMP1Cre.gp130f/f mice did not show any significant skeletal or morphological defects, indicating that gp130 in the osteoblast lineage is not required for normal bone development.
Conditional deletion of gp130 in osteocytes and osteoblasts inhibits trabecular bone formation, but not osteoclast formation
Dynamic histomorphometry revealed that the low bone mass of DMP1Cre.gp130f/f and Osx1Cre.gp130f/f mice was associated with a low rate of bone formation. Both DMP1Cre.gp130f/f (Fig. 4A) and Osx1Cre.gp130f/f (Fig. 4B) male mice demonstrated 30% lower trabecular bone formation rate (BFR) compared to their respective w/w controls. This was attributed to significantly lower mineralizing surface (dLS/BS), not a low MAR (Fig. 4), suggesting impaired osteoblast differentiation in both strains of mice. However, no significant changes in osteoblast number (N.Ob/B.Pm), osteoblast surface/bone surface (Ob.S/BS), osteoid surface (OS/BS), or osteoid volume (OV/BV) were observed in male DMP1Cre.gp130f/f or Osx1Cre.gp130f/f mice (Table 2). This may relate to the already low baseline values of these parameters in 12-week-old male mice, making any further reduction difficult to detect.
|w/w (n = 9)||f/f (n = 11)||w/w (n = 6)||f/f (n = 8)||w/w (n = 15)||f/f (n = 14)||w/w (n = 15)||f/f (n = 8)|
|N.Ob/B.Pm (1/mm)||5.74 ± 1.15||4.60 ± 1.21||16.51 ± 2.73||11.51 ± 2.31*||5.21 ± 0.76||6.86 ± 1.15||14.24 ± 0.97||8.12 ± 1.06***|
|Ob.S/BS (%)||8.83 ± 1.83||7.35 ± 1.87||24.89 ± 4.53||13.99 ± 1.37*||8.86 ± 1.29||10.85 ± 2.05||25.39 ± 1.98||12.41 ± 1.48***|
|OS/BS (%)||8.37 ± 1.83||7.08 ± 1.97||24.10 ± 5.35||13.62 ± 1.55*||7.79 ± 1.16||9.83 ± 1.43||18.6 ± 1.65||10.33 ± 1.71**|
|OV/BV (%)||0.87 ± 0.23||0.91 ± 0.34||3.94 ± 1.02||2.68 ± 0.94||0.87 ± 0.23||1.45 ± 0.29||3.09 ± 0.38||1.57 ± 0.31**|
|N.Oc/B.Pm (1/mm)||1.56 ± 0.25||1.83 ± 0.27||1.91 ± 0.23||2.68 ± 0.46||1.36 ± 0.13||1.38 ± 0.17||2.66 ± 0.21||2.01 ± 0.25|
|Oc.S/BS (%)||5.98 ± 1.04||7.21 ± 1.35||8.56 ± 1.32||10.47 ± 1.72||6.56 ± 0.81||6.46 ± 0.73||13.88 ± 1.11||11.78 ± 1.77|
In 12-week-old female mice, where osteoblast numbers and extent of osteoid are significantly greater than in male mice, we detected significantly lower N.Ob/B.Pm and Ob.S/BS (Table 2) as well as reduced BFR in DMP1Cre.gp130f/f (Fig. 4C) and Osx1Cre.gp130f/f (Fig. 4D) mice compared to controls, confirming reduced osteoblast differentiation in the absence of osteocyte gp130. To determine whether this was associated with diversion toward the adipocyte lineage, marrow adipocyte numbers were evaluated in Osx1Cre.gp130f/f mice, and were at wild-type levels in both male and female 12-week-old mice (data not shown).
Despite the known influence of IL-6 family cytokines on osteoclastogenesis, osteoclast number (N.Oc/B.Pm), and surface/bone surface (Oc.S/BS) were not significantly different in DMP1Cre.gp130f/f or Osx1Cre.gp130f/f male or female mice compared to controls (Table 2).
Cortical circumference is increased with gp130 deletion in the osteoblast lineage
The cortical bone phenotypes of DMP1Cre.gp130 12-week-old and 26-week-old male mice (Fig. 5A) were strikingly different to those observed in trabecular bone. Although there was no change in bone length (Supporting Fig. 5) or cortical thickness (data not shown), femoral periosteal perimeter (Ps.Pm), mean cross-sectional moment of inertia (CSMI), and marrow area were all significantly greater in DMP1Cre.gp130f/f male mice (Fig. 5A) compared to controls. This reflected increased bone width in both the AP and lateral-medial directions (Fig. 5B). A similar cortical phenotype was observed in female DMP1Cre.gp130f/f mice at 26 weeks of age (data not shown), in 12-week-old male Osx1Cre.gp130f/f mice (Fig. 5C), and in 12-week-old and 26-week-old female Osx1Cre.gp130f/f mice (data not shown). At 26 weeks of age there was no difference in cortical bone mass between Osx1Cre.gp130f/f and w/w controls mice owing to delayed bone accrual in the w/w group, an effect of the Osx1Cre transgene that has been reported.
In contrast to their low trabecular BFR, DMP1Cre.gp130f/f and Osx1Cre.gp130f/f mice formed periosteal bone at the same rate and along the same extent of surface as their respective w/w controls at 12 weeks (Fig. 5D, E) and at 6 and 26 weeks (data not shown); this was also found in transverse sections of femurs from 12-week-old DMP1Cre.gp130 mice (data not shown). There was also no detectable difference in endocortical mineralizing surface, MAR, BFR, osteoclast number, or osteoclast surface in 12-week-old mice (data not shown). This indicates that the larger cortical diameter is caused by a level of bone formation slightly greater than control throughout bone growth, but at a level too low to be detectably greater at any single time point (ie, 6, 12, and 26 weeks).
In osteocytes, gp130maintains cortical material properties
Because CSMI and marrow area were significantly greater in the absence of gp130 in both Osx1Cre.gp130f/f and DMP1Cre.gp130f/f mice, mechanical properties of the femoral mid-shaft were evaluated in 26-week-old DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice by three-point bending test. Elastic modulus (intrinsic stiffness), ultimate strength, and yield strength (level of stress at which permanent damage is initiated in the bone) were all significantly lower in male DMP1Cre.gp130f/f mice (Fig. 6A–C), and elastic modulus was significantly lower in female DMP1Cre.gp130f/f mice (Fig. 6A). Toughness (amount of energy required to fracture the bone) was slightly lower in the males, but this was not statistically significant (Fig. 6D). Parameters that are not corrected for the altered size and shape of the bone (stiffness, failure force, deformation at failure, and energy absorbed at failure) were not significantly different in DMP1Cre.gp130f/f males or females compared to w/w controls (Table 3), indicating that the structural integrity of the whole bone remained intact due to the increased cortical dimensions. The load-deformation graphs confirm no difference in response to mechanical load in either the male or female mice prior to normalization for bone geometry (Fig. 6E, F). In contrast, the average stress-strain curves demonstrate that after normalization for bone geometry, the DMP1Cre.gp130f/f male bones are more compliant and withstand less stress than their wild-type counterparts (Fig. 6G, H). Postyield stress (the ratio of stress the bone withstands after yield) was also significantly reduced in male DMP1Cre.gp130f/f mice only (Fig. 6I). Because the bones from the DMP1Cre.gp130f/f males had poorer material properties, tissue mineral density was assessed by µCT on bones from 12-week-old mice (Supporting Fig. 6A) but no difference between w/w and f/f mice was detected. Notably, female mice had a significantly higher baseline tissue mineral density than male mice (p < 0.0001, Supporting Fig. 6A).
|w/w (n = 8)||f/f (n = 8)||w/w (n = 5)||f/f (n = 7)|
|Stiffness (N/mm)||117.6 ± 6.92||119.2 ± 5.48||138.5 ± 3.31||116.3 ± 11.81|
|Failure force (N)||19.10 ± 1.51||19.01 ± 0.77||24.38 ± 1.27||25.94 ± 2.23|
|Deformation at failure (mm)||0.164 ± 0.01||0.164 ± 0.01||0.217 ± 0.01||0.238 ± 0.02|
|Energy absorbed at failure (J)||1.48 × 10−3 ± 2 × 10−4||1.47 × 10−3 ± 3 × 10−4||2.83 × 10−3 ± 3 × 10−4||2.81 × 10−3 ± 4 × 10−4|
To determine whether the poor matrix material properties were due to differences in the proportion of woven versus lamellar bone in DMP1Cre.gp130f/f mice, 100-µm transverse sections of 12-week-old male femurs were analyzed by polarized light microscopy. DMP1Cre.gp130f/f mice showed a significantly higher proportion of disorganized woven bone compared to lamellar bone matrix in the proximal femur (Fig. 6J).
When 12-week-old bones from these mice were analyzed by RPI, there was no difference in total indentation distance, average loading or unloading slopes, or indentation distance increase after loading (Supporting Fig. 6B), suggesting that the poor material properties of DMP1Cre.gp130f/f bones do not relate to a defect at the microindentation level.
Osteocyte density is greater in the trabecular compartment of gp130 osteocyte conditional knockout mice, resulting in greater local production of sclerostin
The trabecular compartment–specific reduction in BFR in the absence of gp130 in osteocytes suggests that osteoblasts in the trabecular bone and periosteum respond differently to gp130 signals controlling bone formation. Because OSM, LIF, IL-11, and CT-1 all inhibit osteocyte production of the bone formation inhibitor sclerostin, the density of total osteocytes and sclerostin-positive osteocytes was assessed in 12-week-old male Osx1Cre.gp130 and DMP1Cre.gp130 cortical and trabecular bone to determine whether region-specific changes in osteocyte number and sclerostin production could explain the region-specific phenotype. Osteocyte density was no different in the trabecular versus cortical bone of the Osx1Cre.gp130f/f mice (Fig. 7A). However, Osx1Cre.gp130f/f mice had significantly more sclerostin-positive osteocytes in both trabecular and cortical bone than their w/w controls (Fig. 7B), suggesting that gp130 deletion early in the osteoblast lineage increases the number of osteocytes that express sclerostin throughout trabecular and cortical bone.
In contrast to the Osx1Cre.gp130f/f mice, there was no significant difference in total osteocyte density in trabecular bone between DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice. However, within DMP1Cre.gp130f/f mice, there was a substantially greater numerical density of osteocytes in the trabecular bone compared to the cortical bone of the same mice (Fig. 7A). Again, this was in contrast to the Osx1Cre.gp130f/f mice. Although the percent of sclerostin-positive osteocytes was not different between DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice in the trabecular or cortical bone (data not shown), the greater total osteocyte density in trabecular bone compared to cortical bone resulted in significantly more sclerostin-positive osteocytes in trabecular bone compared to cortical bone in DMP1Cre.gp130f/f mice (Fig. 7B), a difference that was not observed in control mice. Thus, the increased number of sclerostin-positive osteocytes is only because of an increase in the total number of osteocytes, not an increase in the proportion of sclerostin-producing osteocytes, and suggests other mechanisms are responsible for the differences in bone volume between DMP1Cre.gp130w/w and DMP1Cre.gp130f/f mice.
Because osteocyte sclerostin production was increased in both the trabecular and cortical bone, but only in Osx1Cre.gp130f/f mice, these data also indicate that region-specific patterns of sclerostin production are not responsible for the region-specific phenotype observed in both Osx1Cre.gp130f/f and DMP1Cre.gp130f/f mice.
Osteocyte deletion of gp130 reduces osterix, collagen type 1-α1, and osteocalcin gene expression
To assess the effects of gp130 deletion on gene expression in bone, we determined differences in osteoblast and osteocyte marker genes in flushed femurs of DMP1Cre.gp130f/f male and female mice with verified gp130 knockdown (Fig. 1B). mRNA levels of the early osteoblast marker osterix (Fig. 8A) and mature osteoblast marker osteocalcin (Fig. 8B), as well as collagen type 1-α1 (Fig. 8C), were all significantly lower (∼50%) in bones from DMP1Cre.gp130f/f male mice compared to DMP1Cre.gp130w/w mice. Notably, collagen type 1-α2 mRNA levels were not significantly changed in f/f bones (Fig. 8D). Thus the ratio of collagen type 1-α1:collagen type 1-α2, which normally exists in a 2:1 ratio at the protein level, was reduced (Fig. 8E). Osx, Ocn, Col1-α1, and Col1-α2 were all at wild-type levels in flushed femora from female DMP1Cre.gp130f/f mice (data not shown). Neither receptor activator of NFκB ligand (RANKL; Tnfsf11) nor Sost mRNA levels were significantly different in DMP1Cre.gp130f/f male or female mice compared to controls (Fig. 8F; female data not shown).
This work demonstrates that the key cell through which gp130 controls trabecular bone formation and cortical bone strength is the osteocyte. Genetic deletion of gp130 in osteocytes results in very low trabecular BFR and mass, and larger cortical bone diameter that compensates for significantly degraded cortical bone material properties and low collagen production. This regionally divergent phenotype was confirmed in a second model where gp130 was deleted in the entire osteoblast lineage. In neither model of gp130 deficiency was osteoclastogenesis altered. This leads us to conclude that the increased RANKL production in the osteoblast lineage that occurs in response to IL-6 family cytokines does not play a key role in physiological bone growth and remodeling, but it is the regulation of osteoblast differentiation and matrix production through the osteocyte that is the key role of gp130 in the osteoblast lineage (Fig. 9).
The findings of a low level of trabecular bone formation and increased cortical dimensions elicited by gp130 deletion in osteocytes is unique. It contrasts with osteocyte-specific knockouts of other pathways that stimulate bone formation (β-catenin, polycystin 1 [Pkd1], and insulin-like growth factor 1 [IGF-1]), which result in low bone mass in both cortical and trabecular bone owing to reduced bone formation at both sites.[37-39] These data suggest that stimulating gp130 signaling in osteocytes may increase trabecular bone formation and trabecular bone mass, without increasing osteoclast formation, a phenomenon that could be exploited therapeutically.
All IL-6 family members, apart from those that signal through CNTF receptor (CNTFR), stimulate osteoclast differentiation. This influence has been understood for many years to depend on the ability of these cytokines to stimulate RANKL production by osteoblast-lineage cells.[12-15] This is supported by in vitro coculture studies, where osteoclast formation in response to IL-6 family cytokines depended on the presence of osteoblasts,[12, 13] and osteoclast formation in response to other cytokines and hormones such as IL-1, PTH, and 1,25-dihydroxyvitamin-D3 was partially dependent on gp130 signaling. It was surprising, then, that we observed no change in osteoclast differentiation or RANKL expression when gp130 was deleted in the osteoblast lineage, or in osteocytes alone. In previous work, adult mice with global deletion of either OSMR or IL-11R demonstrated low levels of osteoclast formation,[6, 16] whereas increased numbers of osteoclasts with impaired activity were observed in mice lacking CT-1. OSMR-null osteoblasts were less supportive of osteoclast formation when stimulated with 1,25-dihydroxyvitamin-D3, but supported enhanced osteoclastogenesis in response to PTH, suggesting stimulus-dependent roles of osteoblastic gp130 signaling in support of osteoclastogenesis. In the cases of IL-11R and CT-1 deletion, altered osteoclast formation levels were intrinsic to the hemopoietic lineage.[9, 16] Importantly, in all of these mouse knockout models, the IL-6 family cytokine was deleted systemically, and therefore it is unclear whether the effects on osteoclastogenesis in these mice was because of the absence of gp130 signaling in the osteoblast lineage. The data presented here indicate that osteoclast formation in physiological bone remodeling does not require gp130 signaling in the committed osteoblast lineage (Fig. 9A), and this is the first instance in which the absence of gp130 signaling was restricted to the osteoblast lineage. This supports the concept that RANKL expression induced by IL-6 family cytokines in the osteoblast lineage may be most important in specific conditions of elevated osteoclast formation such as estrogen deficiency or inflammatory arthritis. In addition, during bone development and growth, control of osteoclast formation by the IL-6 family may be restricted to the growth plate, as observed in the neonate-lethal LIFR and gp130-null mice[17, 18] and adult LIF-deficient mice. In this region, gp130-dependent osteoclastogenesis may be more directly controlled by hypertrophic chondrocytes than by the osteoblast lineage. This is further supported by a lack of bone phenotype in neonate or 6-week-old Osx1Cre.gp130f/f mice, indicating that gp130 signaling in the osteoblast lineage is not required for normal bone development; rather, its key role is to maintain bone formation in the physiological process of bone remodeling in the adult skeleton.
The low level of trabecular bone formation, lower osteoblast number, and reduced osterix, type 1 collagen, and osteocalcin mRNA levels in DMP1Cre.gp130f/f bones confirms a critical role for osteocyte gp130 in promoting osteoblast differentiation (Fig. 9B). This confirms a physiological role for gp130 cytokines that stimulate bone formation in vivo and modify osteocyte gp130 signaling, such as OSM, CT-1, and LIF.[6, 9, 40] This is further supported by the complete absence of an anabolic response to supracalvarial injections of OSM in DMP1Cre.gp130f/f mice. This is also consistent with the known stimulatory effects of the IL-6 family cytokines on bone formation,[6, 41] and low levels of trabecular bone formation reported in mice with global deletion of LIF, CT-1, OSMR, or IL-11R.[5, 6, 9, 16] Although IL-6 family members that stimulate bone formation also inhibit adipogenesis in vitro,[5, 6, 9, 16] no change in marrow adipogenesis was observed in the osteoblast-specific or osteocyte-specific knockouts (data not shown). This indicates that the anti-adipogenic action of IL-6 family cytokines is mediated by gp130 signaling in non-committed osteoblast precursors prior to osterix expression (Fig. 9A).
Deletion of gp130 in osteocytes led to decreased intrinsic bone strength (ie, ultimate strength) in DMP1Cre.gp130f/f mice and increased the diaphyseal dimensions such that ultimate bending load remained unchanged. This means that the fracture resistance of the bones was retained despite a significant reduction in their material properties. This mechanical maintenance was achieved by an increase in periosteal dimensions, hence increased moment of inertia, to compensate for the decline in bone material quality. The modest increase in trabecular thickness, detected only by µCT, may also reflect this compensatory mechanism. Alternatively, this slight increase in trabecular thickness was specific for the DMP1Cre model, which suggests that gp130 signaling in the osteocyte may limit the thickness of new trabeculae formed at the growth plate, and that this limiting effect is negated when gp130 signaling is deleted throughout the osteoblast lineage. Accordingly, the low trabecular BFR reported here and measured in the secondary spongiosa reflects a lower rate of bone remodeling, not trabecular formation at the growth plate. Furthermore, the reduced material properties of the cortical bone in osteocytic gp130 deficiency suggests that gp130 signaling in osteocytes may contribute to collagen deposition during the osteoid production phase, or mineral deposition during matrix maturation (Fig. 9B,C). Because no significant alterations in tissue mineral density or periosteal MAR were detected in 12-week-old DMP1Cre.gp130f/f mice, the poor material properties of the bone matrix likely reflect a defect in collagen deposition rather than altered mineralization. Indeed, this was supported by our findings that the ratio of collagen 1 type α1 to type α2 mRNA levels was altered in male mice, and there was a higher proportion of woven bone in cortical samples of the male mice. However, we cannot rule out an alteration in the distribution of mineral in the bones of DMP1Cre.gp130f/f male mice, particularly given the lower osteocalcin mRNA levels.
Although the DMP1Cre.gp130f/f female bones were less stiff than their wild-type littermates, only male DMP1Cre.gp130f/f bones had lower ultimate strength, yield strength, and postyield stress. These data indicate the DMP1Cre.gp130f/f long bones are fragile, which is consistent with an altered collagen phenotype, because it is the collagen matrix that provides the postyield fracture resistance. Previous work has demonstrated that slender bones have a higher degree of mineralization and tissue mineral density to compensate for their smaller bone size, as we have observed in the smaller diameter and significantly higher tissue mineral density of the female DMP1Cre.gp130f/f and control bones compared to males. The higher mineral content in the female bones may therefore provide some protection against the effect of gp130 deletion on bone stiffness.
Increased periosteal circumference has not been observed in any global gp130 knockout mouse models. In fact, IL-6, IL-11R, CT-1, and OSMR knockout models each show reduced periosteal circumference.[2, 3, 6, 9] This suggests that maintenance of cortical bone strength by increased periosteal apposition depends on influences of IL-6 family cytokines on other cell types that stimulate increased activity of periosteal osteoblasts. The increased periosteal expansion in DMP1Cre.gp130f/f mice is likely to be a response to altered mechanical loading resulting from the poor material integrity. We were unable to detect an increase in periosteal bone formation on transverse and coronal sections at any single time point, which suggests that the periosteal expansion in these bones occurred slowly and over a long period of time, which may be indicative of a cumulative mechanical response over the life of the mouse. This may also indicate that, even though OSM, OSMR, and IL-11 mRNA levels in bone are increased in response to mechanical load, periosteal bone formation that is likely induced by mechanical forces may not require gp130 signaling in osteocytes.
More sclerostin-positive osteocytes were observed in both trabecular and cortical bone of Osx1Cre.gp130f/f mice compared to controls, but not in DMP1Cre.gp130f/f mice. This suggests that the inhibitory paracrine action of endogenous gp130 cytokines on sclerostin expression, previously described by pharmacological studies in vivo and in vitro, is mediated indirectly by an action on early osteoblasts. This increase in sclerostin-positive osteocytes was only observed in Osx1Cre.gp130f/f mice, suggesting it may contribute to the normal trabecular thickness in these mice, a phenotype that was not observed in the DMP1Cre.gp130f/f mice. Because both Osx1Cre.gp130f/f and DMP1Cre.gp130f/f mice exhibited low trabecular bone formation and large periosteal circumference, regulation of sclerostin production can be excluded as the primary driving force behind the change in bone structure observed in these mice.
The greater osteocyte density and low BFR found specifically within the trabecular bone of the DMP1Cre.gp130f/f mice is consistent with reports in human bone, where trabecular osteocyte density and BFR were negatively correlated. This suggests that IL-6 family cytokines, in addition to reducing sclerostin, may also determine the rate at which osteocytes become incorporated within the trabecular network. It is also possible that in the absence of osteocytic gp130 osteoblast apoptosis is impaired, and more osteoblasts survive through to osteocyte differentiation, consistent with previous in vitro reports of an antiapoptotic role of IL-6 family cytokines. However, it remains unclear why this was observed only in DMP1Cre.gp130f/f bones when the Osx1Cre.gp130f/f mice also exhibited low trabecular bone mass.
Therapeutic use of IL-6 inhibition for inflammation is widespread, and new IL-6 family inhibitors are actively being pursued for a wide range of disorders. Clinical trials have begun to test the efficacy of a soluble form of gp130 (sgp130-Fc) that selectively blocks IL-6 trans-signaling by sequestering the IL-6:sIL-6R complex and preventing it from binding membrane-bound IL-6R.[49, 50] These pathologic conditions already exhibit increased fracture risk, and our findings of detrimental effects of osteocyte gp130 inhibition on trabecular bone mass and cortical bone material properties indicates that careful skeletal monitoring of patients enrolled in future clinical trials is warranted.
In conclusion, we report here an essential role for osteocyte gp130 signaling to support trabecular bone formation and healthy composition of the cortical bone matrix. These data also suggest that stimulation of gp130 signaling targeted to the osteocyte may provide therapeutic benefit by stimulating trabecular bone formation and preserving material properties in the cortical bone.
All authors state that they have no conflicts of interest.
This work was completed with financial support from the National Health and Medical Research Council of Australia. The Victorian State Government Operational Infrastructure Support Scheme provides support to St. Vincent's Institute of Medical Research. We thank Joshua Johnson for technical expertise in the processing of bone samples for histomorphometric analysis.
Authors' roles: Study design: NAS and RWJ. Study conduct: RWJ, HJB, CV, IJP, NEM, TS, ECW, TTK, and HN. Data collection: RWJ, HJB, CV, IJP, NEM, TS, ECW, TTK, HN, and NAS. Data analysis: RWJ, HJB, CV, IJP, HN, NCW, MRF, and NAS. Data interpretation: RWJ, HJB, CV, NCW, MRF, TJM, and NAS. Drafting manuscript: RWJ and NAS. Revising manuscript content: RWJ, TS, NCW, MRF, TJM, and NAS. Approving final version of manuscript: RWJ, HJB, CV, IJP, NEM, TS, ECW, TTK, HN, NCW, MRF, TJM, and NAS. NAS takes responsibility for the integrity of the data analysis.