Reduced Bone Loss in a Murine Model of Postmenopausal Osteoporosis lacking Complement Component 3

Abstract

The growing field of osteoimmunology seeks to unravel the complex interdependence of the skeletal and immune systems. Notably, we and others have demonstrated that complement signaling influences the differentiation of osteoblasts and osteoclasts, the two primary cell types responsible for maintaining bone homeostasis. However, the net effect of complement on bone homeostasis in vivo was unknown. Our published in vitro mechanistic work led us to hypothesize that absence of complement component 3 (C3), a central protein in the complement activation cascade, protects against bone loss in the ovariectomy-based model of postmenopausal osteoporosis. Indeed, we report here that, when compared to their C57BL/6J (WT) counterparts, ovariectomized C3 deficient mice experienced reduced bone loss at multiple sites and increased stiffness at the femoral neck, the latter potentially improving mechanical function. WT and B6;129S4‐C3^tm1Crr/J (C3^−/−) mice were either ovariectomized or sham-operated at 6 weeks of age and euthanized at 12 weeks. MicroCT on harvested bones revealed that the trabecular bone volume fraction in the metaphyses of both the proximal tibiae and distal femora of ovariectomized C3^−/− mice is significantly greater than that of their WT counterparts. Lumbar vertebrae showed significantly greater osteoid content and mineral apposition rates. Mechanical testing demonstrated significantly greater stiffness in the femoral necks of ovariectomized C3^−/− mice. These results demonstrate that C3 deficiency reduces bone loss at ovariectomy and may improve mechanical properties.

Graphical Abstract

MicroCT on bones harvested from C57BL/6J (WT) and B6;129S4‐C3^tm1Crr/J (C3^−/−) mice revealed that the trabecular bone volume fraction in the metaphyses of both the proximal tibiae and distal femora of ovariectomized C3^−/− mice is significantly greater than that of their WT counterparts. Their lumbar vertebrae showed significantly greater osteoid content and mineral apposition rates, and mechanical testing demonstrated significantly greater stiffness at the femoral neck. These results demonstrate that C3 deficiency reduces ovariectomy-induced bone loss and may improve mechanical properties.

INTRODUCTION

A seminal study by Sato et al. revealed a linkage between the complement cascade and bone homeostasis¹ which helped to launch the discipline of osteoimmunology. As a result, scientists have been driven to reassess immune contributions to conditions and biological events ranging from gingivitis-induced bone resorption to osteoarthritis and T cell-mediated bone anabolic activity^2–4. As part of the innate immune system, the defining function of complement is to mount a rapid and potent defense against pathogens and to clear cellular debris. Three main activation pathways, classical, lectin, and alternative, each with differing triggers, have been identified. C3 is the central piece of all three pathways, and C3^−/− mice have been widely used as a model to explore the potential role of complement in many diseases^5–10.

Osteoimmunological studies point to roles for complement quite disparate from its conventionally defined function. Within skeletal development, C3 and C1s are purported to be involved in the cytolysis of the cartilaginous model central to endochondral ossification, with C3a driving the chemotaxis of angiogenic cells critical to bone formation^{11; 12}. With regard to fracture healing, Ehrnthaller et al. demonstrated that the terminal pathway is crucial to successful osteotomy repair, as femora of C5^−/−, but not C3^−/−, mice exhibited significant reductions in flexural rigidity following a 21-day healing period¹³. Van der Ende et al. argued for a causal link between insufficient levels of mannose-binding lectin and non-healing in a case study of a nonunion fracture in the foot^{14; 15}. Ignatius et al. showed that C5 elicits osteoblast chemotaxis to the fracture callus¹⁶. These roles of complement in promoting skeletal development and repair, with the proper controls in place, stand in stark contrast to the self-destructive activity of complement so well established in rheumatoid arthritis and, more recently, in osteoarthritis³.

Sato et al. first demonstrated C3 production in primary osteoblastic cells¹ and later determined that blocking of C3 in bone marrow cell cultures attenuates osteoclast maturation¹⁷. This paracrine signaling of C3 was central to our previous work, which demonstrated that bone marrow cell cultures derived from C3^−/− mice generate fewer osteoclasts than their WT-derived counterparts. Together, these publications laid the groundwork for the current study. Here, we test the hypothesis that the reduced osteoclastogenesis observed in the absence of C3 is sufficient to effect a measureable protection against bone loss in a murine model of osteoporosis. We present data demonstrating that C3^−/− mice, relative to their WT counterparts, experience a reduction in the bone loss associated with ovariectomy.

MATERIALS AND METHODS

Overview of study design

In this study, we used only F1 to F3 progeny of mice purchased from The Jackson Laboratory (Bar Harbor, ME); genotyping was performed according to their specifications. Housing conditions included an AAALAC-accredited specific pathogen free facility with a 12 hour light - 12 hour dark cycle. Cage conditions included ad libitum chow (PicoLab® Mouse Diet 20 #5058 for breeders and PicoLab® Rodent Diet 20 #5053 for non-breeders), acidified water, cob bedding (The Andersons Bed-o’Cobs® 1/8″), and 1-4 cagemates. At 6 weeks of age, C57BL/6J (WT) or B6;129S4-C3^tm1Crr/J (C3^−/−) female mice (mass range 15 – 20 g) were ovariectomized to model postmenopausal osteoporosis. Tissues were harvested from mice euthanized at 12 weeks of age. Uterine mass was used to assess whether the ovariectomies were successful. Hindlimbs and L4 vertebrae were analyzed by one or more of the following: micro-computed tomography (microCT), histomorphometry, and mechanical testing. Age-matched, sham-operated animals were included as controls, yielding a total of four cohorts of 12. We arrived at this number through power analysis of results from a pilot study in which we had used microCT to determine trabecular number in distal femora. Due to incomplete ovariectomy, as determined by uterine mass, a total of 2 mice were excluded from all analyses, as reflected by cohort numbers. Animal studies described were performed under an approved protocol in accordance with the guidelines of the Animal Care and Use Committee of the Benaroya Research Institute in Seattle, WA. This manuscript was prepared in compliance with ARRIVE Guidelines.

Surgical and Post-Surgical Treatment

Subcutaneous injections of 0.06 mg/kg buprenorphine were administered in the morning, prior to ovariectomy or sham operations. Isoflurane, at 4%, was used to induce anesthesia, and animals were maintained within a surgical plane of anesthesia throughout the procedure, using 1.5 – 2% isoflurane. Peritoneal closure was performed with chromic gut sutures, and 1-2 wound clips were used to close the skin incision. Bupivacaine was administered over the wound site at 1.3 mg/kg just prior to transferal of the mouse to a recovery cage. A second dose of buprenorphine was administered subcutaneously 10-12 hours after the initial dose, and daily monitoring of animal behavior and appearance continued for 5 days following the procedure. Wound clips were removed one week after surgery, and animal behavior and body condition were assessed on a weekly basis until euthanasia.

MicroCT of Long Bone

All microCT images were obtained ex vivo from mice euthanized at 6 weeks post-surgery. Following fixation in 10% neutral buffered formalin for 72 hours at 4°C with agitation, bones were rinsed in distilled water and stored in 70% ethanol at 4°C until analyzed. Right femora and tibiae were scanned at high-resolution on a Scanco vivaCT 40 (10.5 μm voxel size, 55 kVp, 145 μA). Scans of the femoral distal metaphysis covered a 0.85 mm volume immediately proximal to the proximal termination of the articular surface. Tibiae were analyzed both at the proximal metaphysis (0.85 mm volume immediately distal to the growth plate) and at the diaphysis (1.0 mm volume initiating 1.25 mm proximal to the tibia-fibula junction). These microCT image data were preprocessed using a Gaussian Filter algorithm to remove image noise (Sigma = 1.2, Support = 2.0) followed by bone segmentation within the scan volume using standard image thresholding techniques¹⁸. In scan regions in which trabecular bone tissues were analyzed and/or segmented from the cortical shell (i.e. metaphyses of the distal femora and proximal tibia), a threshold of 476.4 mg HA/cm³ provided the best comparisons between 2-D binarized images and the original gray-scale images. When cortical bone was quantified (i.e. tibial metaphysis), a manufacturer recommended threshold for cortical bone identification was used (700.8 mg HA/cm³). Standard trabecular or cortical microCT analysis measures were determined to examine whole tissue changes between experimental groups¹⁹. Imaging and analysis was performed at the Orthopaedics and Sports Medicine Department of the University of Washington by analysts blinded to sample cohorts.

Imaging and analysis on a total of 18 necks of femora contralateral to those on which mechanical testing was performed was carried out at Baylor College of Medicine Center for Skeletal Medicine and Biology. Left femora from each cohort (n ≥4) were scanned on a Scanco μCT 40 (10 μm voxel size, 55 kVp, 145 μA). As noted by Brommage et al., orientation of the neck is critical in the microCT analysis of this anatomical site²⁰. To achieve vertical orientation, a 4 mm volume region of interest was selected; Scanco files were then exported as DICOM images and imported into GEHC MicroView, version 2.1.2, where the 3-D data set was reoriented (Fig. S-1). To this end, X and Y planes were rotated to the vertical center of the femoral neck shaft from anterior and posterior views, such that the Z plane lay perpendicular to the vertical axis of the femoral neck. The Z plane was moved proximodistally through the scans until just prior to where the neck and shaft meet. This landmark was used to establish the base of a box ROI sized to encompass the neck and that extended 0.6 mm in the proximal direction. This design excludes from the ROI the femoral head and surrounding cartilage. DICOM files were converted back to the Scanco format and imported into the analysis software database. ROI used for trabecular bone analysis was 58 slices of the ROI generated in GEHC MicroView; cortical analysis was limited to the distal-most 39 slices. The contoured ROI was thresholded for geometry using a setting that provided the best comparisons between 2-D binarized images and the original gray-scale images. In the Scanco software version 6.1, this threshold was 400 for both trabecular and cortical analysis. The analyst was blinded to sample cohorts. These methods leave the femoral neck intact and were designed to allow for subsequent mechanical testing. However, because these bones had been stored in 70% ethanol long-term, mechanical testing yielded stiffness values that were increased relative to those generated in testing of the contralateral femoral necks. Because others have reported on this very effect of long-term storage in ethanol²¹, all mechanical property data reported here is from contralateral (right) femora.

Mechanical Testing of Femora

Mechanical testing analysis was limited to stiffness, due to tissue storage conditions. Previously fixed right proximal femora were prepared for femoral neck testing^{22; 23} by rehydration in PBS following microCT imaging. Cohort numbers reflect that a total of 4 samples lacked femoral heads and were thus excluded from mechanical testing. Testing was performed on a Bose Electroforce 3200 using a custom-designed fixture consisting of a plunger designed to cup the femoral head and a plate designed to fix the distal cortical circumference firmly in place (Fig. 1). The femoral head was first loaded to 1N in compression, held for 5 seconds, and then loaded at 0.2 mm/s until failure. The output files were analyzed using a Matlab script to calculate stiffness. The displacement curve between the 1N preload and the maximum load was divided into 5 equal segments which were fitted with a line. Stiffness was calculated as the slope of the steepest segment. Analysts were blinded to sample cohorts.

Figure 1.

Femoral necks were tested by monotonic loading to failure in a direction parallel to the femoral diaphysis.

Histomorphometry of L4 Vertebrae and Distal Femora

Animals were injected i.p. with 21-25 mg/kg calcein (Sigma-Aldrich) in saline with 2% sodium bicarbonate at 1 week, and again, 3 days, prior to euthanasia at 6 weeks post-surgery; harvested tissue was fixed and stored as detailed above. Undecalcified, methyl methacrylate-embedded 5 μm sections of L4 vertebrae or femora were prepared and analyzed at the Bone Histomorphometry Core facility at MD Anderson by an analyst blinded to sample cohorts. For osteoblast measurements, sections were stained with Harris Hematoxylin to visualize nuclei, Acid Fuchsin-Ponceau to stain osteoid, and finally Toluidine blue to better visualize osteoblast morphology and more clearly distinguish mineralized bone from osteoid. Kinetic measurements were performed on unstained sections. The length of all calcein labels within a given region of interest were measured and determined to be single or double labeled surface. The inter-label distance for double labeled surfaces was measured at an interval of every 5 μm, and data were pooled and averaged to generate the MAR (Mineral Apposition Rate). Inter-label distance was not measured on oblique double labels to ensure the accuracy of inter-label width measurements. For accurate osteoclast measurements, sections were stained for Tartrate Resistant Acid Phosphatase (TRAP) by incubating slides with Naphthol AS-MX phosphate substrate (Sigma-Aldrich) and Fast Red TR Salt hemi(zinc chloride) salt (Sigma-Aldrich) as a developer in a 0.2M acetate buffer, pH 5.0 for 1 hour at 37°C. Slides were then counterstained with Harris Hematoxylin. TRAP positive, multinucleated cells lining the bone surface were determined to be osteoclasts; these counts were performed concurrently with assessments of surface erosion.

For each set of parameters, vertebral results were derived from the quantitation of at least 20 mm of trabecular bone surface in the mid-cancellous region of the L4 vertebral body, excluding the 150 μm adjacent to cortical bone and growth plate, thereby avoiding cortical bone spurs and primary spongiosa. This region of interest ranged from 3-4 mm². Measurements of distal femora were restricted to the mid cancellous region of the distal metaphysis. This region of interest starts 150 μm proximal to the growth plate and extends 1.3 mm in the proximal direction; in the lateral and medial directions, it excludes a 150 μm border with the cortical bone. All measurements were performed using Bioquant Osteo 2012 image analysis system (Bioquant Image Analysis Corporation, Nashville, TN) and a Leica DM1000 microscope. Vertebral samples were excluded as reflected when injection error yielded a single calcein label. Six samples from each cohort were chosen at random for the distal femur substudy; one of these samples was excluded by the blinded analyst after sectioning, due to tissue damage.

Statistical Analysis

Each data point represents the value determined for a given animal with the group mean and standard deviation shown. Differences between treatment groups were determined by 2-way ANOVA followed by Sidak’s correction. The resulting adjusted p values are reported left of the solid vertical line. Data points right of the solid vertical line reflect OVX data scaled to the mean of the average sham value for that strain. Differences between treatment groups were determined by unpaired Student’s t-test with or without Welch’s correction, based upon variance of the cohort, or Mann Whitney test, depending upon the distribution of values. The delta is the difference between the mean values of these scaled calculations.

RESULTS

Body and Uterine Mass

On the date of surgery, mice from all cohorts had similar body masses. Successful OVX was confirmed by uterine atrophy. As expected, the body masses of ovariectomized mice were significantly greater relative to those of their sham-operated counterparts; this change did not differ as a function of strain (Table 1).

Table 1. Body and Uterine Mass.

Data are presented as mean ± standard deviation (n = 11 for WT Sham and C3^−/− OVX; n = 12 for WT OVX and C3^−/− Sham). p values were determined by Student’s t-test with or without Welch’s correction, based upon variance of the cohort, or Mann-Whitney test, depending upon the distribution of values, as determined by F‐test with the confidence level set to 95%.

	WT		C3−/−		p value
	Sham	OVX	Sham	OVX	Sham vs. Sham	OVX vs. OVX	WT Sham vs. WT OVX	C3−/− Sham vs. C3−/− OVX
Body Mass at Surgery (g)	17.2 ± 1.3	17.3 ± 0.9	17.7 ± 1.1	17.3 ± 1.6	0.345	0.968	0.884	0.495
Body Mass at Euthanasia (g)	20.2 ± 1.7	22.3 ± 0.8	20.8 ± 1.2	22.2 ± 1.2	0.368	0.868	0.003	0.009
Uterine Mass at Harvest (% body mass)	0.65 ± 0.27	0.11 ± 0.02	0.77 ± 0.31	0.11 ± 0.03	0.497	0.868	< 0.0001	< 0.0001

MicroCT

At 6 weeks after surgery, the trabecular bone volume fraction (BV/TV) of distal femora in C3^−/− OVX was not significantly different from that of their sham-operated counterparts, while the BV/TV in WT OVX was significantly lower than that of WT sham (Fig. 2A). Data points to the right of the solid vertical line in all graphs reflect values relative to sham (% sham), with the delta (Δ) indicating the difference between mean scaled OVX values. When presented in this way, the delta between C3^−/− OVX and WT OVX was 15% (Fig. 2A). Ovariectomized mice showed a significant reduction in trabecular number (Tb.N) within the distal femora compared to sham in both C3^−/− OVX and WT OVX; however, when presented as % sham, Tb.N is significantly greater in the C3^−/− group, with a delta of 8% (Fig. 2B). Ovariectomy in C3^−/− mice yielded a downward trend in trabecular thickness (Tb.Th) relative to C3^−/− sham; this trend toward a decrease in Tb.Th in WT OVX approached significance (Fig. 2C). Trabecular separation (Tb.Sp) at the distal femur in WT OVX was also significantly greater than that of their sham-operated counterparts with relative Tb.Sp also significantly greater in WT OVX. This is reflected in the delta of 11% between the averages of C3^−/− OVX and WT OVX when scaled to sham values (Fig. 2D). 3-D reconstructions of microCT scans show significant bone loss in OVX groups with a comparative preservation in C3^−/− mice; scans from the mean animal are shown (Fig. 2E).

Figure 2. Trabecular bone loss at the distal femur was greater in WT OVX.

(A) BV/TV at the distal femur was significantly reduced in WT OVX relative to WT sham; however, BV/TV was not significantly different between C3^−/− cohorts. When presented relative to sham values, BV/TV was significantly greater in C3^−/− mice relative to WT mice, with a delta of 15%. (B, D) Tb.N and Tb.Sp were significantly different from sham in WT OVX cohorts; in C3^−/− OVX, significance was only achieved in Tb.N. Relative to sham values, Tb.N and Tb.Sp in the distal femora of C3^−/− mice were significantly greater and significantly decreased, respectively. (C) Tb.Th values of OVX cohorts did not differ significantly from those of sham cohorts. Left of the solid vertical line, data points represents the value determined for a given animal with the group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2-way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test. (E) 3-D reconstructions were made from scans of distal femora from samples nearest the mean.

At the proximal tibia, the BV/TV in both C3^−/− OVX and WT OVX were significantly lower than those of their sham-operated counterparts (Fig. 3A). When scaled to sham, the value for this parameter in C3^−/− mice is significantly greater than that of their WT counterparts, with a delta of 10% between averages (Fig. 3A). Ovariectomized mice from both strains showed comparable changes relative to sham-operated mice, with no significant difference between WT OVX and C3^−/− OVX values for trabecular number, thickness, or separation (Fig. 3B–D).

Figure 3. Trabecular bone loss at the proximal tibia was slightly greater in WT OVX.

Bone loss at the proximal tibia was significantly greater than sham in both ovariectomized cohorts as determined by BV/TV, Tb.N, Tb.Th, and Tb.Sp. (A) When presented relative to sham values, BV/TV was significantly lower, indicating greater bone loss, in WT OVX mice than in their C3^−/− OVX counterparts. (B-D) Scaled OVX values for Tb.N, Tb.Th, and Tb.Sp in the proximal tibia were not significantly different between groups. Left of the solid vertical line, data points represent the value determined for a given animal with group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2-way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test.

The endocortical volume (EV) of the tibial midshaft in WT OVX was significantly greater than that of WT sham, whereas C3^−/− OVX EV was not significantly different from C3^−/− sham values (Fig. 4A), resulting in a statistically significant delta of 12% between WT OVX and C3^−/− OVX averages. Values for tibial midshaft periosteal volume (PV) were not significantly different between ovariectomized and sham cohorts; however, WT OVX and C3^−/− OVX groups trended in opposite directions from their sham counterparts, such that the relative tibial midshaft PV is significantly greater in WT OVX than in their C3^−/− counterparts (Fig. 4B).

Figure 4. Endocortical erosion was increased in WT OVX but unchanged in C3^−/− OVX.

(A) Bone loss at the endocortical surface was significantly greater in WT OVX, both relative to sham-operated counterparts and relative to C3^−/− OVX when scaled. (B) PV of the OVX cohorts were not significantly different from sham values; however, trends were in opposite directions. When scaled to sham, PV at the tibial midshaft was significantly greater in WT OVX when compared to C3^−/− OVX. Left of the solid vertical line, data points represent the value determined for a given animal with the group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2-way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test.

Mechanical Testing and MicroCT of the Femoral Neck

Femoral neck testing revealed that stiffness values of the ovariectomized WT mice (WT OVX) and ovariectomized C3^−/− mice (C3^−/− OVX) trended in opposite directions from associated sham cohorts (Fig. 5). When presented relative to sham, the average C3^−/− OVX stiffness values were significantly greater than those of their WT counterparts, with a delta of 36%. Because stiffness measurements have been shown to rely largely on the mineral phase of bone²⁴, and because this parameter has been reported to be consistent between fresh and formalin-fixed femora²⁵, reported data were limited to stiffness. MicroCT analyses of trabecular bone of contralateral femoral necks did not predict these differences in mechanical properties; however, cortical analysis revealed a small but significant bone volume decrease evident in the WT OVX but not the C3^−/− OVX cohort (Figs. S-2 & S-3).

Figure 5. Femoral neck stiffness is greater in C3^−/− OVX.

Mechanical testing demonstrated that stiffness at the femoral neck, when presented relative to sham values, was significantly greater in C3^−/− OVX, with a delta of 36%. No significant differences in stiffness were observed between sham and OVX cohorts; however, values of OVX cohorts trended in opposite directions relative to sham, and scaled calculations show a significant increase in stiffness for C3^−/− OVX over that of WT OVX, with a p value of 0.002. Left of the solid vertical line, data points represent the value determined for a given animal with group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2-way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test.

Histomorphometry

Histomorphometric analysis of the L4 vertebrae yielded a large delta between scaled OVX groups that did not quite meet significance for % osteoid surface (OS/BS); however, group averages for % bone surface undergoing mineralization during interlabel time (MS/BS), and MAR, when presented scaled to sham, were significantly greater in C3^−/− OVX relative to WT OVX (Fig. 6A–C). OVX cohort values trended in opposite directions compared to sham for MS/BS and MAR; C3^−/− OVX MAR values were significantly greater than those of C3^−/− sham (Fig. 6C). Relative to their sham-operated counterparts, the number of osteoblasts per tissue area (N.Ob/T.Ar) of OVX groups trended downward (Fig. 6D). The delta between group averages presented relative to sham was only 2% and not significantly different.

Figure 6. Osteoblast activity increases in C3^−/− OVX.

(A–B) Relative to sham-operated mice, neither of the ovariectomized cohorts demonstrated a significant difference in either OS/BS or MS/BS. Because trends were opposite in direction, when presented relative to sham, MS/BS was significantly greater in C3^−/− OVX when compared to WT OVX. (C) MAR was significantly greater in C3^−/− OVX, while MAR trended down in WT OVX relative to sham-operated counterparts; the delta between % sham groups was 17%. (D) Despite this increased activity, N.Ob/T.Ar trended down in both OVX groups relative to their sham-operated counterparts. Left of the solid vertical line, data points represent the value determined for a given animal with group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2‐way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test. (E-H) Images of sections stained with Hematoxylin, Acid Fuchsin-Ponceau, and Toluidine blue show the variable levels of osteoid per bone surface between cohorts. (I-L) The difference in MAR between cohorts is qualitatively evident in epifluorescence microscopy of sections of representative L4 vertebrae.

In osteoclast measurements, both the percent bone surface eroded by osteoclasts (ES/BS) and number of osteoclasts per tissue area (N.Oc/T.Ar) were significantly decreased in the OVX cohorts, and relative change was not significantly different between scaled OVX groups (Fig. 7A, B). A subset of femora (n=6) was analyzed histomorphometrically, and the results reflect a similar pattern, with a significant decrease in ES/BS of both OVX cohorts and N.Oc/T.Ar of only C3^−/− OVX relative to C3^−/− sham attaining significance (Fig. 7C, D).

Figure 7. Osteoclast number and activity are reduced in OVX cohorts.

(A) At 6 weeks post-surgery, ES/BS of L4 vertebrae was significantly reduced in both OVX cohorts relative to their sham-operated counterparts. When scaled to sham, average ES/BS values for WT OVX and C3^−/− OVX are similar with a delta of only 2%. (B) L4 vertebral N.Oc/T.Ar is also significantly reduced in the OVX cohorts relative to their sham-operated counterparts. Scaled values yield a delta of only 6% between groups. (C, D) In distal femora, at 6 weeks post-surgery, both ES/BS and N.Oc/T.Ar were greater in the sham cohorts relative to their OVX counterparts; these trends were significant in all cases but in N.Oc/T.Ar of WT cohorts. Scaled values were not significantly different for either parameter. The delta for ES/BS was 1%, and the delta for N.Oc/T.Ar was 24%. Left of the solid vertical line, data points represent the value determined for a given animal with group mean and standard deviation shown. p values indicate only significant differences between treatment groups, as determined by 2‐way ANOVA. Data points right of the solid vertical line reflect OVX data scaled to the mean sham value, with the delta indicating the difference between means of these two groups. p values indicate only significant differences as determined by t-test or nonparametric test. (E-H) As indicated in 7B, representative images of TRAP stained sections of L4 vertebrae reveal far fewer osteoclasts in OVX cohorts than in sham cohorts. Black arrowheads highlight some of the red-stained osteoclasts in each field.

DISCUSSION

Our previously published work used cells derived from a global C3 knockout to corroborate the findings of Sato et al., which first demonstrated reduced osteoclastogenesis in the absence of C3 signaling^{17; 26}. In these in vitro studies, we went on to demonstrate the mechanistic involvement of the alternative complement pathway and IL-6, and to implicate local, rather than systemic, complement sources²⁶. It was then hypothesized that this complement deficiency would attenuate bone loss in an accepted model of postmenopausal osteoporosis. MicroCT analysis showed that at 6 weeks after ovariectomy, bone loss at the distal femur was reduced in C3^−/− OVX as reflected by trabecular bone volume fraction, trabecular number, and trabecular separation. Overall, C3 deficiency offered greater protection of xtrabecular bone at the distal femur compared to the proximal tibia. Trabecular bone volume fraction at the proximal tibia, when presented relative to sham, is 10% greater in C3^−/− (versus 15% in distal femur) than in their WT counterparts. Values for other parameters measured in the proximal tibia did not prove to be significantly different between strains. MicroCT analysis of the tibial midshaft showed evidence of resorption modeling at the endocortical surface in ovariectomized WT mice, reminiscent of what has been reported to occur in the femoral midshaft of osteoporotic humans^{27; 28}. The pattern was not observed in ovariectomized C3^−/− mice. In addition to endocortical erosion, estrogen deprivation in females produces transient periosteal formation modeling²⁹. Interestingly, while this formation modeling is arguably responsible for the upward trend in the periosteal volume of the WT OVX cohort, no such trend was observed in the C3^−/− OVX cohort. The net effect on cortical volume is neutral, with no significant difference among the four groups. While this deposition of bone away from the central axis in WT OVX mice would lend greater structural stability over time, the anticipation is that the formation modeling would slow, but that the erosion at the endocortical surface would continue, resulting in the cortical thinning observed in ovariectomized C57BL/6 mice. Because the tibial midshaft of C3^−/− OVX mice did not exhibit the predicted periosteal formation modeling in response to estrogen deprivation, it is arguable that the static cortical volume at this site is a consequence of reduced resorption in the absence of C3. An alternative explanation would be that the same acute post-OVX response occurs, but that the C3^−/− OVX animals undergo a partial recovery by 6 weeks post-OVX.

To assess bone function, mechanical testing of the highly trabecular region of the femoral neck was performed. C3^−/− OVX bone proved to have significantly greater stiffness than that of WT OVX, with a delta of 36%. Because these bones had been previously fixed, we limited our report to stiffness, which others have shown to be consistent between formalin-fixed and fresh bone²⁵. The downward trend in WT OVX femoral neck stiffness recalls the decreased stiffness seen in idiopathic osteoporosis³⁰, and the absence of this trend in C3^−/− OVX suggests a protective effect of the complement deficiency. Certainly, further studies with a broader assessment of mechanical properties, including brittleness, are warranted. We additionally performed microCT on a subset of femoral necks from contralateral legs and found a small but statistically significant decrease in cortical bone volume in WT OVX but not C3^−/− OVX. This intriguing finding must be repeated with larger cohorts.

MicroCT and histomorphometry of long bones and vertebrae, respectively, suggest one or both of the following: 1) that the absence of C3 in OVX yields diminished overall bone loss relative to WT 2) that C3 deficiency enables a robust anabolic reversal of the initial period of bone loss seen in both C3^−/− and WT. In light of the similar N.Ob/T.Ar across cohorts, the increased osteoblast activity (MS/BS and MAR) in C3^−/− OVX may be interpreted to be a level of anabolic mobilization not apparent in WT OVX. Histomorphometric calculations of BV/TV of L4 vertebrae did not reflect the same relative patterns of protection against bone loss established by microCT analysis of long bone. Further study is needed to determine whether this indicates a temporal distinction, or differing responses, as a function of anatomical site.

To our knowledge, this is one of only two studies reporting on the in vivo modulation of bone homeostasis by C3. Matsuoka et al., in a rigorous in vitro study, found medium conditioned by osteoclasts to promote osteoblast differentiation via C3aR signaling. They went on to show that two weeks post-ovariectomy, ddy mice treated with a C3aR antagonist suffered greater bone loss at the tibial metaphysis than those left untreated³¹. While C3a is irrefutably involved in bone homeostasis, it appears that temporal control may be a key determinant of the direction in which the shift will occur. In vivo, C3aRs on various cell types compete for C3a binding; taken together, our previous work and the Matsuoka paper demonstrate competing outcomes, based upon whether C3a signaling occurs at the osteoblast or the osteoclast^{26; 31}. The in vivo portion of the study by Matsuoka et al. limited analysis to a single anatomical site and did not include mechanical testing. Future studies to further our understanding of these two intriguing but disparate outcomes should address differences attributable to strain, age, and time point, as well as how histomorphometric and microCT data translate into mechanical data. In addition, studies using SB 290157 as a pharmacologic approach to dampen or abrogate signaling at C3aR must take in to account the agonist activity of this antagonist reported by Mathieu et al³². The present study demonstrates a remarkable difference in stiffness of a highly trabecular site in C3^−/− OVX relative to WT OVX; these data are supported by the microCT and histomorphometric analyses presented here.

Studies in both humans and rodents show estrogen deprivation yields an acute skeletal response of bone resorption concomitant with an increase in bone deposition, with the net effect being one of bone loss^{33; 34}. Histomorphometric analyses of L4 vertebrae yielded % osteoid surface and MAR values significantly greater in ovariectomized C3^−/− mice than in their WT counterparts, thus supporting the in vitro data-derived hypothesis. However, the osteoclast measurements appear to contradict the hypothesis, with data indicating similar osteoclast numbers and activity between WT OVX and C3^−/− OVX. This was shown not to be site-specific, as the same pattern was observed in a subsequent substudy of the distal femora of legs contralateral to those from which the initial microCT data were collected. Because most of the data reveal diminished microarchitectural quality in ovariectomized relative to sham-operated bone, a likely explanation for these collective observations is that an acute period of bone resorption in C3^−/− OVX precedes the 6-week post-surgery time point. In support of this, Sakai et al. published on a 2-phase development of osteopenia following ovariectomy³⁵. Here, mice (ddy) were assessed at 2-, 4‐, and 6-weeks post-surgery. The authors found enhanced bone turnover due to increases in both osteoblastogenesis and osteoclastogenesis at the first two time points; however, between days 28 and 42, both osteoblast and osteoclast formation were equal to levels found in sham-operated controls. In the current study, while histomorphometry of the vertebrae demonstrated an increase in osteoid surface, the histomorphometric substudy of distal femora corroborated the findings of Sakai et al., yielding similar values between ovariectomized and sham-operated mice. This ostensible incongruence between cellular- and tissue-level femoral bone formation suggests that earlier events account for improved trabecular structure in C3^−/− mice over their WT counterparts at the 6-week post-surgery time point. Ongoing temporal studies will clarify this.

Collectively, this work makes abundantly clear that C3 is involved in murine bone homeostasis following ovariectomy. The absence of C3 diminishes bone loss and improves trabecular microarchitecture, lending bone potentially improved mechanical properties. These positive effects are, in part, attributable to increased osteoblast activity.