p47phox-Nox2-dependent ROS Signaling Inhibits Early Bone Development in Mice but Protects against Skeletal Aging*

Background: How reduced ROS signaling regulates inflammation and remodeling in bone remains unknown. Results: Age-related switch of bone mass in p47phox-deficient mice occurs through an increased inflammatory milieu in bone. Conclusion: p47phox-Nox2-dependent physiological ROS signaling suppresses inflammation in aging. Significance: p47phox-Nox2 and Nox4 may play different roles during early development and skeletal involution because they serve unique functions on osteoblast differentiation and proliferation. Bone remodeling is age-dependently regulated and changes dramatically during the course of development. Progressive accumulation of reactive oxygen species (ROS) has been suspected to be the leading cause of many inflammatory and degenerative diseases, as well as an important factor underlying many effects of aging. In contrast, how reduced ROS signaling regulates inflammation and remodeling in bone remains unknown. Here, we utilized a p47phox knock-out mouse model, in which an essential cytosolic co-activator of Nox2 is lost, to characterize bone metabolism at 6 weeks and 2 years of age. Compared with their age-matched wild type controls, loss of Nox2 function in p47phox−/− mice resulted in age-related switch of bone mass and strength. Differences in bone mass were associated with increased bone formation in 6-week-old p47phox−/− mice but decreased in 2-year-old p47phox−/− mice. Despite decreases in ROS generation in bone marrow cells and p47phox-Nox2 signaling in osteoblastic cells, 2-year-old p47phox−/− mice showed increased senescence-associated secretory phenotype in bone compared with their wild type controls. These in vivo findings were mechanistically recapitulated in ex vivo cell culture of primary fetal calvarial cells from p47phox−/− mice. These cells showed accelerated cell senescence pathway accompanied by increased inflammation. These data indicate that the observed age-related switch of bone mass in p47phox-deficient mice occurs through an increased inflammatory milieu in bone and that p47phox-Nox2-dependent physiological ROS signaling suppresses inflammation in aging.

Bone remodeling is a process in which osteoblasts and osteoclasts are coupled. It changes dramatically during the course of development as it undergoes age-dependent regulation. It has been suggested that bone loss during aging is largely asso-ciated with decreases in osteoblast number and bone formation rate (1), and this age-dependent loss of bone mass is also accompanied by increases in the generation or accumulation of reactive oxygen species (ROS) 2 (2). Whether such age-dependent accumulation of ROS plays a direct role on impairing bone remodeling is unknown. On the other hand, appropriate concentrations of ROS have also been shown to serve as signaling molecules controlling cell differentiation and proliferation (3). Therefore, maintaining a fine balance between ROS production and elimination of excessive ROS seems to be essential for physiologic bone homeostasis at any age.
During the past several decades, investigations on the potential harmful effects of increased levels of ROS to damage DNA, lipids, and proteins leading to many pathophysiologic conditions, including degenerative diseases, have dominated over research examining the consequences of reduced ROS production (4). However, literature reports suggest that a reduction in ROS levels and signaling may also be detrimental (5,6). For example, in the musculoskeletal system, decreased ROS concentrations have been shown to promote inflammation (6). Further investigations on the anti-inflammatory activity of NADPH-oxidases (Nox), the origins and dynamics of ROS production, and their critical biological targets in bone are required.
Although ROS can be generated by several sources, the tightly controlled and cell-specific Nox represent one of the major sources of ROS signaling molecules, including superoxide, hydrogen peroxide, and hydroxyl radicals in many cell types. Activation mechanisms, tissue distribution, and subcellular localization of different members of the Nox family are markedly different (7). Given that they are macrophage in nature, bone resorptive osteoclasts are expected to express all forms of Nox (8). We have previously identified only the expression of Nox 1, 2, and 4 in nonphagocytic osteoblastic cells, with * This work was supported by United States Department of Agriculture/Agricultural Research Service Grant 6251-51000-003 to the Arkansas Children's Nutrition Center, and in part by National Institutes of Health Grant RO1 AA18282. 1 To whom correspondence should be addressed: Arkansas Children's Nutrition Center, Slot 512-20B, 15 Children's Way, Little Rock, AR 72202. Tel.: 501-364-2707; Fax: 501-364-3161; E-mail: chenjinran@uams.edu. 2 The abbreviations used are: ROS, reactive oxygen species; Nox, NADPH-oxidase; CGD, chronic granulomatous disease; pQCT, peripheral quantitative computerized tomography; BMD, bone mineral density; ALP, alkaline phosphatase; 2,7-DCF-DA, 2,7-dichlorodihydrofluorescein diacetate; SABG, senescence-associated ␤-galactosidase; DPI, diphenyleneiodonium chloride.
isoforms 4 and 2 being the most dominant (9). One of the important roles of the catalytically active Nox complex in neutrophils and macrophages is the killing of microbes by producing ROS (8). The release of such ROS and its downstream products from phagocytic cells is also known as the respiratory burst (8). Reduced respiratory burst occurs in chronic granulomatous disease (CGD). These patients commonly inherit abnormalities of Nox2 or of the essential Nox2 co-factors p22 phox , p47 phox , or p67 phox and display persistent inflammation in many tissues (4).
Although data regarding bone inflammation in CGD patients is lacking, age-dependent increases in development of arthritis have been recently shown in a Nox2-deficient mouse model (6). Membrane-associated Nox-p22 phox complex is thought to be activated by subsequently recruiting cytosolic proteins p47 phox , p40 phox , and p67 phox . Thereafter, this complex interacts with the small GTPase Rac1 to generate superoxide by transporting electrons from intracellular NADPH to oxygen (10,11). It is not surprising that the majority of ROS generated in osteoclasts is through a Nox-dependent mechanism (12,13). Recently, RANKL and TNF␣ have been shown to stimulate ROS generation in osteoclasts or their precursors (14,15). It was unexpected that bone resorption in Nox2-deficient mice was apparently normal during early development (4). Although it may be possible that normal bone resorption in Nox2-deficient mice is partially due to compensatory up-regulation of other isoforms of Nox in osteoclasts (16), data from these reports led us to hypothesize that osteoblastic bone formation is disrupted in p47 phox -Nox2 signaling-deficient mice. Moreover, animals with deletion of p47 phox , one of the complex components of Nox2, showed less severe age-related joint destruction and decreased chondrocyte death (17) compared with animals that had Nox2 entirely knocked out. This suggests that Nox2 signaling may be partially preserved in its components in deficient animals. In the present study, we set to answer some of these critical questions in bone and focus on osteoblasts using p47 phox knock-out young and old mice.

Experimental Procedures
Animals-Time-impregnated female C57BL/6J-Ncf1M1J/J (p47 phoxϪ/Ϫ ) mice and their wild type controls were obtained from The Jackson Laboratory (Bar Harbor, ME). Dams and litters were housed in polycarbonate cages in an Association of Laboratory Animal Care-approved animal facility in an environmentally controlled room at 22°C with a 12-h light/dark cycle and fed standard rodent chow ad libitum throughout experimental period including pregnancy and lactation. Male wild type control and p47 phoxϪ/Ϫ pups (n ϭ 10 per group) were fed chow ad libitum as previously described (18), for 6 weeks and 2 years. We have recorded body weights on a weekly basis and food intake daily for 7 days after 1, 4, and 12 weeks on the diet. After the mice were sacrificed, serum, legs, and vertebrae were collected and stored at Ϫ80°C until use.
Bone Analyses-Peripheral quantitative computerized tomography (pQCT) was performed on formalin-fixed left tibia for bone mineral density (BMD) measurement using a method established in our laboratory (19). A STRATEC XCT 960 M unit (XCT Research SA, Norland Medical Systems, Fort Atkins, WI) specifically configured for small bone specimens was uti-lized. Software version 5.4 was used with thresholds of 570 mg/cm 3 to distinguish cortical bone and 214 mg/cm 3 to distinguish trabecular from cortical and subcortical bone. Tibial BMD and bone mineral content were calculated. The position for pQCT scanning was defined at a distance from proximal tibia 1 mm below growth plate corresponding to 7% of the total length of the tibia. The distance between each scanning was 1 mm, and total of five scans (five slices) were carried out. The data are expressed as the means of three contiguous slices with the greatest trabecular bone density.
Sequential dehydration of right rear tibial was carried out using different concentrations of alcohol. Proximal tibial bone samples were embedded, cut, and Masson and TRAP (tartrateresistant acid phosphatase) stained by standard histology special procedures (20,21). For histomorphometric analysis, sections were read in a blinded fashion. Parameters of cancellous bones in the proximal tibia were measured with a digitizing morphometric system, which consists of an epifluorescent microscope (model BH-2; Olympus), a color video camera, and a digitizing pad (Numonics 2206) coupled to a computer (Sony) and a morphometry program from OsteoMetrics, Inc. Total bone area, total bone surface, osteoblast surface, osteoclast surface, and eroded surface were obtained by manual tracing. Vertebrae (L4) histology and immunostaining including above unstained tibia sections using TNF␣ antibody (Cell Signaling) were carried out using standard protocol from VectaStain ABC kit (Vector Laboratory, Burlingame, CA).
Microcomputed tomography measurements of trabecular of the tibial bone after the above pQCT process were evaluated by using Scanco microcomputed tomography scanner (CT-40; Scanco Medical AG, Bassersdorf, Switzerland) at 6-m isotropic voxel size with x-ray source power of 55 kV and 145 A and integration time of 300 ms. The gray scale images were processed using a low pass Gaussian filter ( ϭ 0.8, support ϭ 1) to remove noise, and a fixed threshold of 220 was used to extract the mineralized bone from the soft tissue and marrow phase. Cancellous bone was separated from the cortical regions by semiautomatically drawn contours. A total of 120 slices starting from about 1 mm distal to growth plate, constituting 0.70-mm length, was evaluated for trabecular bone structure by using software provided by Scanco, as described in detail previously (22).
Three-point bending of the left femur was performed at room temperature using a miniature bending apparatus with the posterior femoral surface lying on lower supports (7 mm apart) and the left support immediately proximal to the distal condyles. Load was applied to the anterior femoral surface by an actuator midway between the two supports moving at a constant rate of 3 mm/min to produce a physiological in vivo strain rate of 1% for the average mouse femur. The mechanical properties including ultimate strength/stress and stiffness were recorded with a digital caliper.
Serum Bone Turnover Markers-The serum bone formation marker alkaline phophatase (ALP) and the serum bone resorption marker C-terminal telopeptides of type I collagen (CTX-1) RatLaps were measured by Rat-MID TM ALP ELISA and Rat-Laps TM ELISA, respectively, from Nordic Biosciences Diagnostic (Herlev, Denmark). Serum total osteocalcin levels were mea-sured by an ELISA-based kit from TAKARA (TAKARA Bio Inc.). ELISA for mouse mTRAP5b levels was performed using enzyme immunoassay kits from TSZ ELISA (TSZ Scientific).
Real Time RT-PCR Analysis-Mouse femur bone RNA and osteoblastic cell RNA were extracted using TRI reagent (MRC Inc., Cincinnati, OH) according to the manufacturer's recommendation followed by DNase digestion and column cleanup using Qiagen mini columns. Briefly, for RNA isolation from bone tissue, at the time of sacrifice, the right femur was taken, and bone marrow cells were flushed with Eagle's minimum essential medium ϩ Hanks' salts after cleaning the surrounding connective tissue. Femoral bones were then frozen in liquid nitrogen and smashed. Femoral bone was placed in 1000 l of TRI reagent and homogenized using a Polytron aggregate (Kinematica, Luzern, Switzerland). 100 l of 1-bromo-3-chloropropane was added, and the mixture was centrifuged for 15 min at speed of 16,000 rpm, 4°C. 450 l of supernatant was taken, and an equal volume of isopropanol was added and centrifuged for an additional 15 min (16,000 rpm, 4°C). After washing the RNA pellet with 75% ethanol, isolated RNA was resuspended in RNase-free water. For RNA isolation from cultured cells, treated cells from 6-well plates were washed twice with PBS, and 1000 l of TRI reagent was added into each well. Cells were scraped into a 1.5-ml Eppendorf tube. RNA preparation was identical to that of isolation of RNA from bone tissue. Reverse transcription was carried out using an iScript cDNA synthesis kit from Bio-Rad. Real time RT-PCR was carried out using SYBR Green and an ABI 7500 fast sequence detection system (Applied Biosystems, Foster City, CA) (23) ( Table 1).
Western Blotting and Antibody Array-Femoral bone tissue proteins were extracted using a cell lysate buffer as described previously (24). Collagen 1, Runx2, RANKL, TNF␣, and MMP9 protein expression in bone tissue was assessed by standard Western immunoblotting using antibodies recognizing these proteins (Santa Cruz Biotechnology, Inc.) followed by incubation with secondary antibodies conjugated with horseradish peroxidase (Santa Cruz). ␤-Actin protein in bone tissue was analyzed by immunoblotting, using mouse monoclonal antibody recognizing ␤-actin (Sigma) followed by incubation with a secondary anti-mouse antibody conjugated with horseradish peroxidase (Santa Cruz). Immunoblots were visualized using SuperSignal West Pico chemiluminescent (Pierce). Quantitation of the intensity of the bands in the autoradiograms was performed using a VersaDoc TM imaging system (Bio-Rad). Mouse inflammation antibody arrays were carried out using proteins isolated either from vertebrae or cultured cells accord-ing to the procedure provided by manufacture in detail (Ray-Biotech, Inc., catalog no. AAM-INF-1-8).
Cell Culture, DNA Constructs, and Transient Transfection-Neonatal wild type and p47 phox knock-out mouse calvarial cells were isolated from 3-day-old pups using sequential collagenase digestion using a method described previously (25). Cells were cultured in ␣-minimum essential medium (Invitrogen) supplemented with 10% FBS (Hyclone, Logan, UT), penicillin (100 units/ml), streptomycin (100 g/ml), and glutamine (4 mM). To generate replicative senescent calvarial cells, we cultured 2 ϫ 10 4 cells/well in a 6-well plate, when cells became ϳ80% confluent after 5-6 days they were passaged. This procedure was repeated up to 20 times for a total of 20 passages. Transfections of Nox4 shRNA (Sigma-Aldrich; NM_015760/trcn0000076586/pLKO.1) into calvarial cells were performed using protocols provided by the manufacturer. Empty vectors of pLKO.1 entry was used as control for Nox4 shRNA. ALP staining for assessing cell differentiation was performed based on method published previously (20).
Flow Cytometric Measurement of ROS-The cell-permeable dye 2,7-dichlorodihydrofluorescein diacetate (2,7-DCF-DA) (Sigma-Aldrich) becomes fluorescent upon reaction with ROS. 2,7-DCF-DA was dissolved in dimethyl sulfoxide and stored as 50 mM stock. Bone marrow cells or calvarial cells were loaded with 10 M 2,7-DCF-DA for 30 min and then washed three times with PBS before they were harvested. Washed cells were resuspended in 500 l of PBS and kept on ice until flow cytometric analysis was started. ROS measurement was immediately carried out by flow cytometry using FACSort (Becton-Dickinson, Rutherford, NJ) with a 488-nm excitation beam. The signals were obtained using a 530-nm band pass filter for DCF. Each determination was based on the mean fluorescence intensity of 5000 cells. Basal and phorbol 12-myristate 13-acetate-stimulated hydrogen peroxide production was measured in neonatal calvarial cells using the Amplex Red hydrogen peroxide/peroxidase assay (Invitrogen, Molecular Probes) as per the manufacturer's instructions and our previous description (26).
Cell Proliferation Assay-Nonradioactive cell proliferation assay was performed following the protocol provided by manufacturer (Promega Corporation, part no. TB169). According to the manufacturer's instructions, this assay measures absorbance at 490 nm, and there is a linear correlation (the correlation coefficient of the line is 0.997) between cell number and absorbance at 490 nm; absorbance was converted to cell number.  TAA CCG CTA CCC GGA TCC TA  TGT CTT GGA CAG AGC CAT GTG  Tnf␣  GAC GTG GAA CTG GCA GAA GAG  GCC ACA AGC AGG AAT GAG AAG  Mmp9  TCT TCT GGC GTG TGA GTT TCC A  TGC ACT GCA CGG TTG AAG CAA A  Osteocalcin  TTG TGC TGG AGT GGT CTC TAT GAC  CAC CCT CTT CCC ACA CTG TAC A  Runx2  AAG GTT CAA CGA TCT GAG ATT TGTG  GTG AAG ACG GTT ATG GTC AAG GT  p16  GTC GTA CCC CGA TAC AGG TGA T  GGG TCC TCG CAG TTC GAA TCT  p53 GGA GAC ATT TTC AGG CTT ATG GA GCC TTC AAA AAA CTC CTC AAC ATC Senescence-associated ␤-Galactosidase (SABG) Staining and Activity Assay-SABG activity assay was performed by ␤-galactosidase enzyme assay kit (Promega) measured the absorbance at 420 nm according to the manufacturer's instructions. Cell ␤-galactosidase staining was also performed according to a method published previously (23). Senescent cells were identified as blue-stained cells by standard light microscopy.
Statistical Analyses-The data are expressed as means Ϯ S.E. t test or one-way or two-way analysis of variance followed by Student-Newman-Keuls post hoc analysis was used to compare the treatment groups. Values were considered statistically significant at p Ͻ 0.05.

Results
Increased Bone Formation, Mass, and Strength in Young p47 phoxϪ/Ϫ Mice, Is Reversed in Old p47 phoxϪ/Ϫ Mice Compared with Their Age-matched Wild Type Controls-To determine p47 phoxϪ/Ϫ mouse bone phenotypes at both age of 6 weeks and 2 years, we first analyzed tibial bone by pQCT (peripheral quantitative computed tomography scan). At 6 weeks of age, BMD (with exception of trabecular BMD) and bone mineral content were both significantly higher in p47 phoxϪ/Ϫ mice compared with those from wild type animals (Fig. 1A). To our surprise, at 2 years of age, all pQCT parameters measured in the p47 phoxϪ/Ϫ mice were significantly lower relative to their wild type counterparts (Fig. 1A). To determine whether the observed lower bone quantity, i.e. bone mass, is associated with poor bone quality, three-point bone bending tests were performed to evaluate bone strength in femurs. As expected, and consistent with our pQCT analysis, both stiffness and peak load were significantly higher in 6-week-old p47 phoxϪ/Ϫ mice and lower in 2-year-old p47 phoxϪ/Ϫ mice compared with their respective wild type controls (Fig. 1B). The observed differences in bone quantity and quality between p47 phoxϪ/Ϫ and wild type mice at 6 weeks and 2 years were reflected by the bone metabolite levels in the serum. Levels of bone formation markers osteocalcin and ALP were significantly higher in serum of 6 weeks p47 phoxϪ/Ϫ mice, whereas they were lower in 2-year-old p47 phoxϪ/Ϫ mice compared with their respective age-matched wild type controls (Fig. 1C). On the other hand, bone resorption marker CTX-1, reflecting osteoclast activity, was only significantly different between p47 phoxϪ/Ϫ and wild type mice at 6 weeks (Fig. 1C). TRAP5b, FIGURE 1. Increased bone quantity and quality in 6-week-old p47 phox؊/؊ mice but reversed in 2-year-old p47 phox؊/؊ mice compared with their age-matched respective controls. A, quantitative pQCT analysis of the proximal tibial in 6-week-old and 2-year-old p47 phoxϪ/Ϫ mice versus their respective wild type controls (n ϭ 8 mice). Trabecular BMD (mg/cm 3 ), total BMD, cortical BMD, and total bone mineral content (BMC) are presented. B, femur three-point bending test to determine bone strength was performed; parameters of peak load and stiffness are presented. C, bone formation and resorption markers in serum were measured. The data are expressed as means Ϯ S.E. (n ϭ 8 per group). * Ͼ wild type; # Ͻ wild type p Ͻ 0.05. believed to be a marker of osteoclast number, was significantly different between p47 phoxϪ/Ϫ and wild type mice at age 2 years, but not at 6 weeks (Fig. 1C). Bone histomorphometric analytical data further supported these results on bone mass and bone metabolic markers. In 6-week-old mice, with the exception of osteoblast surfaces, increased osteoblast indices including osteoblast number and bone volume were found in p47 phoxϪ/Ϫ mice compared with their respective controls ( Fig. 2A). Decreased osteoclast surface accompanied by erosion surface was also observed in p47 phoxϪ/Ϫ mice compared with their control at 6 weeks of age ( Fig. 2A). In contrast to 6-week-old p47 phoxϪ/Ϫ mice, 2-year-old p47 phoxϪ/Ϫ mice had significantly decreased bone volume and osteoblast number and increased osteoclast number but no differences in osteoclast surface and erosion surface compared with their wild type controls (Fig.  2B). Despite changes in osteoclast indices in both young and aged p47 phoxϪ/Ϫ mice, we could not quantify obvious differences between control and p47 phoxϪ/Ϫ on bone sections using TRAPase staining (data not shown). These data suggest that osteoblast bone formation was largely affected in both rapidly growing and aged p47 phoxϪ/Ϫ mice, although there were some osteoclast bone resorption changes. Moreover, microcomputed tomography analysis confirmed the increased bone volume in 6-week-old p47 phoxϪ/Ϫ mice and decreased bone volume in 2-year-old p47 phoxϪ/Ϫ mice compared with their respective wild type controls (Fig. 2C). The changes of an index of bone strength, polar moment of inertia (J; mm4) ( Fig. 2C), was especially consistent with the femur bending test in p47 phoxϪ/Ϫ mice at 6 weeks and 2 years compared with their controls.
Reduced ROS Generation and p47 phox -Nox2 Signaling Is Associated with Increased Osteoblast Differentiation in Early Life but Enhanced Inflammation in Aged p47 phoxϪ/Ϫ Mice-To explore mechanistic underpinnings of observed bone mass and quality changes with age in p47 phoxϪ/Ϫ mice compared with their wild type controls, we isolated primary bone marrow cells and neonatal osteoblastic calvarial cells. Flow cytometric analysis for total ROS production in bone marrow cells using DFC fluorescence revealed that ROS levels in mice increased with age (p Ͻ 0.001, between wild type), but that ROS production in p47 phoxϪ/Ϫ mice was significantly lower at 6 weeks and 2 years than their corresponding wild type controls (p Ͻ 0.001) (Fig.  3A). Interestingly, according to this measurement, ROS levels in 2-year-old p47 phoxϪ/Ϫ mice were similar to those from 6-week-old wild type mice. Next, isolated neonatal calvarial cells were utilized to determine whether p47 phox -Nox2-dependent ROS production and Nox2 signaling were abolished in osteoblastic cells from p47 phoxϪ/Ϫ mice. Calvarial cells from neonatal p47 phoxϪ/Ϫ mice had lower basal production of hydrogen peroxide resulting from the action of superoxide dismutase on Nox-generated superoxide, as measured using Amplex red, compared with cells isolated from wild type mice (Fig. 3B) (p Ͻ 0.05). In addition, hydrogen peroxide production in response to treatment with the Nox2 activator phorbol 12-myristate 13-acetate was significantly lower in osteoblastic calvarial cells from p47 phoxϪ/Ϫ mice compared with cells from wild type controls (Fig. 3C). These data indicate that p47 phox -Nox2-dependent ROS generating signaling in osteoblastic cells from p47 phoxϪ/Ϫ mice is indeed impaired. To determine osteoblastic cell differentiation potential of calvarial cells in the absence of functioning Nox2 signaling from p47 phoxϪ/Ϫ mice, we cultured these cells in the presence of osteoblastogenic medium for 12 days. Evaluated by ALP staining, we found that cells from p47 phoxϪ/Ϫ mice have increased differentiation potential compared with cells isolated from wild type neonates (Fig. 3D) (p Ͻ 0.05). Expression of Nox4, another major isoform of Nox in nonphagocytic osteoblasts, was increased in cells from p47 phoxϪ/Ϫ mice (Fig. 3E); however, knocking down Nox4 gene expression using Nox4 shRNA (Fig. 3E) was associated with decreased ALP gene expression compared with cells from p47 phoxϪ/Ϫ mice with control shRNA (Fig. 3F). Interestingly, ALP gene expression was also significantly lower in cells from wild type control animals with Nox4 shRNA compared with without Nox4 shRNA (Fig. 3F). These data suggest that p47 phox -Nox2-dependent ROS suppresses osteoblast differentiation, but Nox4 is at least required for increased osteoblast differentiation in p47 phoxϪ/Ϫ mice.
The above observed results may explain bone phenotype and metabolism in p47 phoxϪ/Ϫ mice during early development (age 6 weeks) compared with their wild type controls. However, it is generation was measured in wild type and p47 phoxϪ/Ϫ mouse neonatal calvarial cell cultures using the Amplex Red hydrogen peroxide/peroxidase assay kit. D, neonatal calvarial cells from wild type or p47 phoxϪ/Ϫ mice were cultured in 6-well plates for 12 days, and ALP staining was performed. E, Nox4 shRNA was transfected into calvarial cells from either wild type or p47 phoxϪ/Ϫ mice. Nox4 expression in these transfected cells were determined using Western blots. F, after 24 h, ALP gene expression was examined in calvarial cells from either wild type or p47 phoxϪ/Ϫ mice with or without Nox4 shRNA using real time PCR. Significant differences (p Ͻ 0.05) are indicated by * and # compared with either wild type mice or control.
still not clear why ROS production is reduced but both bone quantity and quality are reversed in p47 phoxϪ/Ϫ mice at age 2 years compared with their wild type control. Although p47 phox and Nox2 themselves are different genes, deletion of p47 phox or Nox2 in mice are animal models for CGD, and they all produce less ROS. Age-dependent increases in inflammation, such as rheumatoid arthritis, have been reported in Nox2-deficient mice (6). We therefore next explored bone inflammation in p47 phoxϪ/Ϫ mice to determine whether this was associated with their lower bone mass compared with wild type controls at age 2 years. Proteins were isolated from vertebrae, and Western blots showed a decrease in RANKL expression with no obvious differences in the other two inflammatory markers, TNF␣ and MMP9, between 6-week-old p47 phoxϪ/Ϫ and wild type animals (Fig. 4A). Col-1 (type 1 collagen) and Runx2 expression were higher in bone from p47 phoxϪ/Ϫ mice compared with wild type controls at age 6 weeks (Fig. 4A). In 2-year-old mice, however, RANKL, TNF␣, and MMP9 were all expressed at significantly higher levels in bone from p47 phoxϪ/Ϫ mice compared with those from wild type controls (Fig. 4A). On the other hand, Runx2 expression was lower in bone from p47 phoxϪ/Ϫ mice (Fig. 4A). Although specific cell types were not defined, the increased TNF␣ expression in bone from 2-year-old p47 phoxϪ/Ϫ mice was further confirmed using antibody immunostaining of decalcified vertebrae bone sections, and TNF␣ was intensively stained red (Fig. 4B). In accordance with changes of protein levels, mRNA expression of these pro-inflammatory cytokines were also greater in bone of p47 phoxϪ/Ϫ mice at 2 years compared with their age-matched controls (Fig.  4C) (p Ͻ 0.05). However, there were no differences in mRNA expression of these pro-inflammatory cytokines between wild type and p47 phoxϪ/Ϫ mice at age 6 weeks (Fig. 4C). Gene expression of osteoblast differentiation and activity markers revealed significantly higher mRNA expression of ALP, osteocalcin, and Runx2 in bone of p47 phoxϪ/Ϫ mice at 6 weeks compared with their wild type controls (Fig. 4C). Among these markers, only osteocalcin was found to be significantly decreased in bone from p47 phoxϪ/Ϫ mice at 2 years compared with their controls (Fig. 4C). Proteins isolated from vertebrae were further analyzed using inflammation antibody array (Fig. 4D). Although FIGURE 4. Increased inflammation in bone from 2-year-old p47 phox؊/؊ mice. A, proteins were isolated from femur and Western blots showing expression of RANKL, TNF␣ and MMP9, and Col-1 (collagen type 1) and Runx2 in 6-week-old and 2-year-old p47 phoxϪ/Ϫ mice compared with those from wild type mice. B, decalcified vertebrae section antibody immunostaining for TNF␣ (stained red). C, real time PCR for the expression of TNF␣, MMP9, ALP, osteocalcin (OC), and Runx2 in bone from 6-week-old and 2-year-old wild type and p47 phoxϪ/Ϫ mice. D, heat map analysis of mouse inflammation antibody array C1 using proteins isolated from mouse vertebrae. Three pooled samples per group. Significant differences (p Ͻ 0.05) are indicated by * and # compared with respective wild type and p47 phoxϪ/Ϫ mice. the signature of inflammation was slightly different between p47 phoxϪ/Ϫ and control mice at 2 weeks of age, significant amounts of inflammatory factors were found increased in their expression from p47 phoxϪ/Ϫ compared with their control mice at 2 years of age.
Bone and Osteoblastic Cell Senescence Is Accelerated in Old p47 phoxϪ/Ϫ Mice-Progressive accumulation of excess ROS has been suspected to be the leading cause of many tissue inflammation and degenerative diseases and underlies the effects of aging (2). This is somehow paradoxical to our current findings showing that reduced ROS production was associated with increased inflammation in bone from aged p47 phoxϪ/Ϫ mice, and this was further associated with decreased bone mass. We therefore asked whether physiological levels of p47 phox -Nox2dependent ROS signaling protects against the skeletal aging process. SABG activity is the most widely used assay and marker to detect aging or senescent cells both in vivo and in vitro, and the SABG activity in bone was measured in p47 phoxϪ/Ϫ and wild type mice at 6 weeks and 2 years. Although there was no significant difference in SABG activity in p47 phoxϪ/Ϫ and wild type mice at 6 weeks, SABG activity was significantly higher in p47 phoxϪ/Ϫ mice at age 2 years (p Ͻ 0.05) (Fig. 5A). It needs to be mentioned that this assay was consistent and generated very small standard deviation (n ϭ 9 was used for in vivo samples), including in ex vivo assay below; error bars are difficult to determine. Each individual sample from p47 phoxϪ/Ϫ mice at age 2 years was measured for SABG showed higher than any samples from control mice. In accordance with these results, p16 and p53 gene expression were also significantly higher in bone from 2-year-old p47 phoxϪ/Ϫ mice compared with wild type controls (Fig. 5B). There was no difference in expression of these genes between p47 phoxϪ/Ϫ and wild type mice at 6 weeks (Fig. 5B). Cell division progression regulator anaphase-promoting complex substrates cyclin A and B were also examined. We found that both cyclin A and B expression in bone were significantly higher in p47 phoxϪ/Ϫ mice at 6 weeks but lower in p47 phoxϪ/Ϫ mice at 2 years compared with their respective wild type controls (Fig. 5C). This is interesting, because the results implicated that without sufficient p47 phox -Nox2-dependent ROS production, skeletal aging may be accelerated.
Finally, the in vivo results demonstrating accelerated cell senescence were recapitulated using isolated calvarial cell cultures. We tried to establish replicative senescent osteoblastic cells in vitro; we usually need to passage calvarial cell cultures at least up to 20 times as described in detail under "Experimental Procedures" and in our previous study (23). Cells isolated from p47 phoxϪ/Ϫ mice showed significantly decreased proliferation and differentiation capacities after only eight passages compared with cells isolated from wild type mice passaged the same number of times (Fig. 6A). Gene expression of ALP, Runx2, p16, and p53 from passage eight calvarial cells from p47 phoxϪ/Ϫ mice were further compared with passage 8 cells from wild type controls. Gene expression of ALP and Runx2 gene was significantly FIGURE 5. Increased senescence pathway in bone from 2-year-old p47 phox؊/؊ mice. A, proteins were isolated from bone from 6-week-old and 2-year-old wild type and p47 phoxϪ/Ϫ mice, and SABG activity was measured (see "Experimental Procedures"). B, RNA was isolated from bone from 6-week-old and 2-year-old wild type and p47 phoxϪ/Ϫ mice; real time PCR was performed for p16 and p53 gene expression. C, real time PCR was performed for cyclin A and B gene expression. Significant differences (p Ͻ 0.05) are indicated by * compared with respective wild type control mice.
lower, whereas that of p16 and p53 genes was significantly higher in passage 8 cells from p47 phoxϪ/Ϫ mice compared with those from wild type mice (Fig. 6B). Consistent with cell senescence-associated gene expression, SABG activity was also significantly higher in passage 8 calvarial cells from p47 phoxϪ/Ϫ mice compared with their control cells as evaluated by SABG ELISA measurement and SABG blue staining (Fig. 6, C and D). Furthermore, substantial increases in TNF␣ expression were also found in passage 20 calvarial cells from p47 phoxϪ/Ϫ mice compared with wild type control cells (Fig. 6E). As we published previously, to generate replicative senescent primary cells, cells need to be passaged at least 20 times. In current experiment, we only passaged cells from p47 phoxϪ/Ϫ mice eight times, and they were quicker to enter senescence program than cells from wild type mice, suggesting that the p47 phox gene exerts roles on antiosteoblastic cell senescence or skeletal aging. Passage 20 cells from both p47 phoxϪ/Ϫ and wild type control mice were further analyzed using inflammation antibody array (Fig. 6F). Expres-FIGURE 6. Passage 8 neonatal calvarial cells from p47 phox؊/؊ mice are senescent. Neonatal calvarial cells were isolated from either wild type or p47 phoxϪ/Ϫ mice and were passaged into new 6-well plates after cells reached 80% confluent, and cell passaging was continued 20 times. A, at passage 8, calvarial cells from p47 phoxϪ/Ϫ mice significantly decreased their proliferation (left, proliferation assay) and differentiation (right, ALP staining) compared with calvarial cells from wild type control mice. B, RNA was collected from calvarial cells at passages 2 and 8; real time PCR showed that passage 8 calvarial cells from p47 phoxϪ/Ϫ mice decreased osteoblastic cell differentiation markers ALP and Runx2 but increased cell senescent markers p16 and p53 compared with cells from wild type control mice. C, SABG activity measurement from collected protein samples showed similar changes to p16 and p53 gene expression. D, cell passage was continued until 20, SABG blue staining showed remarkable changes in cells from p47 phoxϪ/Ϫ mice. E, Western blot for TNF␣ from protein isolated from passage 8 and 20 calvarial cells from either wild type or p47 phoxϪ/Ϫ mice. F, heat map analysis of mouse inflammation antibody array C1 using proteins isolated from passage 20 cells in triplicate. Significant differences (p Ͻ 0.05) are indicated by * compared with respective cells from wild type control mice.
sion of large amounts of inflammatory factors was significantly increased in passage 20 cells from p47 phoxϪ/Ϫ mice compared with those cells from wild type controls.
Different effects of transient versus sustained inhibition of Nox activity on osteoblastic cell senescence signaling and inflammation from control mice were examined. Fig. 7A shows increased expression of cyclin A and B in passage 2 calvarial cells from p47 phoxϪ/Ϫ mice but decreased expression of both cyclin A and B in passage 8 and 20 cells from p47 phoxϪ/Ϫ mice compared with cavarial cells from control mice. Calvarial cells from control mice were treated with Nox pan inhibitor diphenyleneiodonium chloride (DPI) for 4, 24, and 72 h with three different doses. We found cell senescence pathway (SABG activity measurement) was significantly lower after 4 h of DPI treatment (50 and 100 nM of DPI) (Fig. 7B). However, SABG activity was increased after 24 and 72 h of 50 nM DPI treatment (Fig. 7B). Similarly, TNF␣ mRNA expression was decreased after 4 h of 50 nM DPI treatment but significantly increased after 72 h of treatment (Fig. 7C).

Discussion
In the present study, we investigated bone metabolism in p47 phox knock-out mice. We have shown that Nox2 activity and signaling in the bone was markedly reduced in p47 phoxϪ/Ϫ mice, as indicated by decreased ROS generation in both bone marrow cells and primary osteoblastic cells, and osteoblastic cells from p47 phoxϪ/Ϫ mice failed to respond Nox2 activator phorbol 12-myristate 13-acetate. We observed increased bone mass and strength during early development of p47 phoxϪ/Ϫ mice that was reversed in 2-year-old p47 phoxϪ/Ϫ mice compared with their age-matched respective wild type controls. Therefore, p47 phox -Nox2-mediated reduction of ROS generation may be only temporally beneficial for early bone development. The age-related switch of bone mass in p47 phoxϪ/Ϫ mice appears to occur through an increased inflammatory milieu in bone accompanied by accelerated senescent signaling pathway in osteoblasts and that p47 phox -Nox2-dependent physiological ROS signaling suppresses inflammation in aging. Our data in aged p47 phoxϪ/Ϫ mice may conflict with a previous expectation that age-associated bone loss and inflammation are associated with increased ROS production, but they are consistent with recent evidence that showed age-dependent increases in development of inflammatory arthritis in an entirely Nox2-deficient mouse model (6). Our data are also consistent with clinical observations where data indicated increased inflammatory diseases in CGD patients (27). Although the use of a global knock- out for p47 phox may have limitations to discriminate a specific bone effect from a generalized endocrine effect, there is no evidence for such a generalized endocrine effect on bone.
ROS have previously been thought to be toxic by-products of cellular function, but specific Nox-dependent ROS production is not recognized as a by-product. More recent evidence suggests that Nox-dependent ROS signaling may play an important role on cell cycle progression and proliferation (28,29). This may be true, particularly in bone cells from 2-year-old p47 phoxϪ/Ϫ mice, which showed less activity to secrete osteocalcin. Similarly, passage 8 osteoblastic cells from neonatal calvarias of p47 phoxϪ/Ϫ mice showed less potential to proliferate compared with cells from wild type animals. We could not definitively address whether p47 phox -Nox2-dependent ROS generation plays an age-dependent role to promote osteoblastic cell proliferation in vivo; however, increased cell senescent signaling pathway in aged p47 phoxϪ/Ϫ mice may be pivotal in suppression of cell proliferative capacity.
The p47 phox deficiency in mouse is one of the animal models for the human disease CGD in which patients commonly inherits abnormalities of Nox2, p22 phox , or p67 phox and display persistent inflammation in many tissues. Patients usually die early in life because of a variety of infections; bone metabolism in these patients has never been documented. However, according to Nox2-deficient data (16), bone resorption was not suspected to be significantly changed in CGD patients. This prompted us to focus on examining bone formation in p47 phox -deficient mice. We have previously demonstrated that although Nox2 was highly expressed in osteoblastic cells, Nox4 expression was also abundant. In contrast, very low expression of Nox1 and no expression of Nox3 were observed in these cells (9). These results were confirmed in RNA from the bone tissue after aspiration of bone marrow cells in the current study. Our results suggest that without fully functioning p47 phox -Nox2, osteoblast differentiation is increased during early life bone development. We therefore suspect that before inflammatory factors have a significant impact, Nox4 is up-regulated in p47 phoxϪ/Ϫ mice to compensate for a decrease in Nox2 activity, and this up-regulation of Nox4 may be essential for increasing osteoblast differentiation in young mice (Fig. 4). Although there are disagreements from a recent study using global Nox4 knock-out animal (30), cell type-specific conditional knock-out animal models may be used to untangle this possibility. However, in the current study, we used an ex vivo cell culture model and determined that knocking down Nox4 in cells from p47 phoxϪ/Ϫ mice blocked temporal osteoblastic cell differentiation signaling. These data are consistent with recent findings demonstrating that Nox4 is required for cardiomyocyte and adipocyte differentiation in vitro via redox activation (31), and ROS-dependent signaling pathways are also known to play an important role in osteoblast differentiation (32). Our data therefore suggest that Nox2 and Nox4 function differently during early development and skeletal involution to preserve different roles on osteoblast differentiation and proliferation, respectively.
Early life bone development and skeletal remodeling during aging are believed to be two completely different processes. The current understanding on bone development throughout life is that increases in bone formation relative to bone resorption in rapidly growing rodents result in bone accrual. On the other hand, decreases in bone formation and osteoblast function in addition to increases in bone resorption are considered features of the aging skeleton (1,2). Although osteocytes were suspected to be involved in bone remodeling (33), bone loss during aging is still recognized largely as a result of decreased osteoblast number and bone formation rate (1). Decreased osteoblast number in 2-year-old p47 phoxϪ/Ϫ mice could explain diminished osteoblast proliferation. Moreover, it is possible that increased TNF␣ in 2-year-old p47 phoxϪ/Ϫ mice suppresses Runx2 expression. Therefore, we believe that, in contrast to the redundant role of p47 phox -Nox2 in osteoblast differentiation in early development, p47 phox -dependent Nox2 signaling or constitutive p47 phox -dependent Nox2-derived ROS production is indispensable for osteoblast proliferation in aged animals. To promote cell proliferation, a certain level of p47 phox -Nox-associated ROS production may be required. A recent study in a primary acute myeloid leukemia cell line and primary acute myeloid leukemia blasts showed that Nox-derived ROS promoted cell proliferation and was associated with defective oxidative stress signaling (34). In our hands, ROS levels in bone marrow cells of 2-year-old p47 phoxϪ/Ϫ mice were similar to 6-week-old wild type mice. This ROS levels in 2-year-old p47 phoxϪ/Ϫ mice may be far lower to promote osteoblast proliferation.
The root cause of increased inflammation in aged p47 phoxϪ/Ϫ mice characterized by elevation of numbers of inflammatory factors in bone is still not clear, but it appears to be associated with accelerated activation of senescence pathway in osteoblasts. Based on our previous in vitro and in ex vivo cell culture experiments (23), osteoblastic cells (including primary cells and cell line) need to be passaged 20 -30 times to generate replicative senescent and aging osteoblasts. However, neonatal calvarial cells isolated from p47 phoxϪ/Ϫ mice showed rapidly increased cell senescent pathway after only eight passages. We therefore speculate that physiological p47 phox -dependent Nox2-derived ROS levels may suppress not only inflammation but also osteoblast aging. Our speculation may be contradictory to the accepted idea of the universal role of ROS accumulation plays in the aging process; however, it agrees with evidence provided recently in which yeast cells were grown anaerobically without ROS. The lifespan of these cells was shorter than that of cells grown aerobically and was further shortened by caloric restriction (35). Therefore, the current understanding of the free radical theory in aging process has been challenged (36). An increasing number of studies contradicting the free radical theory of aging have been published recently and have been very useful in defining the contribution of oxidative damage to the aging process (37)(38)(39). Moreover, evidence has shown that increased antioxidant protection may even lead to shortened lifespan, and decreased antioxidant function may extend it (40). It is not known whether increased inflammation up-regulates tissue senescence pathway or whether accelerated cell senescence signaling causes tissue inflammation; nonetheless, they seem to be always accompanied by each other. Most of the time, increased cell senescent signaling is also associated with increased tissue inflammation (41). It has been known that cellular senescence is a stable form of cell cycle arrest that may limit the proliferative potential of a cell and promotes aging (42). Our understanding of how ROS and cellular senescence contributes to noncancer tissue pathophysiology is extremely limited. However, as we previously hypothesized, increased osteoblastic cell senescence may play a central role in bone pathology or osteoblastic cellular senescence may be a fundamental factor contributing to bone loss associated with, not only aged p47 phoxϪ/Ϫ mice, but also high fat Western diet-induced obesity, sex steroid deficiency, and other conditions, such as systemic insulin resistance and chronic alcohol abuse (43).
In conclusion, based on our data, we believe that greater bone mass and strength in young p47 phoxϪ/Ϫ mice are due to increased osteoblastic cell differentiation leading to increased bone formation. Reversal of bone mass and strength in older p47 phoxϪ/Ϫ mice is due to decreased bone forming cell proliferation and accelerated cell senescence accompanied by increased inflammation. Our data indicate that p47 phox -Nox2 and Nox4 may play different roles during early development and skeletal involution because they serve unique functions on osteoblast differentiation and proliferation.