Ambient ${\text{PM}}_{2.5}$ Exposure and Bone Homeostasis: Analysis of UK Biobank Data and Experimental Studies in Mice and in Vitro

Abstract

Background:

Previous evidence has identified exposure to fine ambient particulate matter (${\mathrm{PM}}_{2.5}$) as a leading risk factor for adverse health outcomes. However, to date, only a few studies have examined the potential association between long-term exposure to ${\mathrm{PM}}_{2.5}$ and bone homeostasis.

Objective:

We sought to examine the relationship between long-term ${\mathrm{PM}}_{2.5}$ exposure and bone health and explore its potential mechanism.

Methods:

This research included both observational and experimental studies. First, based on human data from UK Biobank, linear regression was used to explore the associations between long-term exposure to ${\mathrm{PM}}_{2.5}$ (i.e., annual average ${\mathrm{PM}}_{2.5}$ concentration for 2010) and bone mineral density [BMD; i.e., heel BMD ($n=\mathrm{37,440}$) and femur neck and lumbar spine BMD ($n=\mathrm{29,766}$)], which were measured during 2014–2020. For the experimental animal study, C57BL/6 male mice were assigned to ambient ${\mathrm{PM}}_{2.5}$ or filtered air for 6 months via a whole-body exposure system. Micro-computed tomography analyses were applied to measure BMD and bone microstructures. Biomarkers for bone turnover and inflammation were examined with histological staining, immunohistochemistry staining, and enzyme-linked immunosorbent assay. We also performed tartrate-resistant acid phosphatase (TRAP) staining and bone resorption assay to determine the effect of ${\mathrm{PM}}_{2.5}$ exposure on osteoclast activity in vitro. In addition, the potential downstream regulators were assessed by real-time polymerase chain reaction and western blot.

Results:

We observed that long-term exposure to ${\mathrm{PM}}_{2.5}$ was significantly associated with lower BMD at different anatomical sites, according to the analysis of UK Biobank data. In experimental study, mice exposed long-term to ${\mathrm{PM}}_{2.5}$ exhibited excessive osteoclastogenesis, dysregulated osteogenesis, higher tumor necrosis factor-alpha ($\text{TNF-}\mathrm{\alpha }$) expression, and shorter femur length than control mice, but they demonstrated no significant differences in femur structure or BMD. In vitro, cells stimulated with conditional medium of ${\mathrm{PM}}_{2.5}\text{-stimulated}$ macrophages had aberrant osteoclastogenesis and differences in the protein/mRNA expression of members of the $\text{TNF-}\mathrm{\alpha }/\text{Traf}6/\mathrm{c}\text{-Fos}$ pathway, which could be partially rescued by $\text{TNF-}\mathrm{\alpha }$ inhibition.

Discussion:

Our prospective observational evidence suggested that long-term exposure to ${\mathrm{PM}}_{2.5}$ is associated with lower BMD and further experimental results demonstrated exposure to ${\mathrm{PM}}_{2.5}$ could disrupt bone homeostasis, which may be mediated by inflammation-induced osteoclastogenesis. https://doi.org/10.1289/EHP11646

Introduction

Age-associated skeletal diseases, such as osteoporosis and fragility fracture,¹ have become important public health concerns with high prevalence and health expenditures around the world.² Osteoporosis is a common skeletal disease characterized by decreased bone mineral density (BMD) and higher osteoporotic fracture risk.³ The risk of fracture increases 2.4- and 2-fold for every 1-standard deviation (SD) decrease in femoral bone mineral density in women ($0.13{\mathrm{g}/\mathrm{cm}}^2$) and men ($0.16{\mathrm{g}/\mathrm{cm}}^2$), respectively.⁴ It was estimated that $>25\text{\hspace{0.17em}million}$ people were affected by osteoporosis and 3.5 million incident fragility fracture cases happened in the European Union in 2010, costing $\sim €37.4\text{\hspace{0.17em}billion}$.⁵ Therefore, it is essential to identify the modifiable risk factors associated with osteoporosis and fragility fracture for the prevention of the diseases.

Bone is regarded as a dynamic organ, modeling and remodeling to grow and change shape throughout life.⁶ Bone turnover, consisting of osteoblastic bone formation and osteoclastic bone resorption, is essential to skeletal health by regulating bone metabolism.⁷ Osteoblasts and osteoclasts take part in the process of cartilage-to-bone transition for bone development, as well as in removing old bone tissue and laying down new bone tissue for bone homeostasis.⁸ Although these two kinds of cells demonstrate different forms of action in the bone modeling and remodeling processes,⁹ disturbance of their activities can induce bone disorders, such as impairment of bone growth and osteoporosis.¹⁰

Hormone- and age-related bone loss or failure to achieve optimal peak bone mass during early adulthood have been recognized as the main risk factors for osteoporosis as well as for fragility fractures.¹¹ Epidemiological studies have identified the associations between several risk factors (e.g., smoking, excessive drinking, physical inactivity) and bone mass.¹¹ However, there is still a long way to go to identify risk factors and elucidate the underlying mechanisms.

Fine particulate matter [PM with an aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$ (${\mathrm{PM}}_{2.5}$)] exposure has been shown to be associated with a range of adverse health effects in the respiratory¹² and cardiovascular systems,¹³ with more recent evidence indicating effects in other organ systems, such as metabolic diseases¹⁴ and neurodegenerative diseases.¹⁵ Furthermore, ambient air pollution has been reported as a potential risk factor for osteoporosis.¹⁶ However, recent epidemiological studies have provided inconsistent evidence of the association between ambient air pollution and skeletal health. Although a population study found no significant association of ambient air pollutants, including ${\mathrm{NO}}_2$, ${\mathrm{O}}_3$, and ${\mathrm{PM}}_{2.5}$, with BMD in women and no association with the incidence of osteoporotic fractures,¹⁷ other studies did demonstrate that ${\mathrm{PM}}_{2.5}$ exposure was associated with lower BMD values and increased fracture risk.^18–21 However, it remains unclear whether ${\mathrm{PM}}_{2.5}$ exposure is associated with markers of skeletal health and the potential biological mechanisms by which ${\mathrm{PM}}_{2.5}$ exposure can contribute to skeletal disorders.

Our previous research, as well as studies conducted by other research groups, have demonstrated exposure to ${\mathrm{PM}}_{2.5}$ could induce proinflammatory factor expression and cause systemic inflammation,^22,23 as well as peripheral inflammation, in vivo.²⁴ Meanwhile, inflammation has been regarded as a potent contributor to excessive osteoclastic bone resorption²⁵ and impaired bone formation in mouse models,²⁶ factors that may eventually cause bone homeostasis imbalance. Therefore, we hypothesized that exposure to ${\mathrm{PM}}_{2.5}$ may impair bone homeostasis, which is associated with increased inflammation.

In the present study, we attempted to further elucidate the relationship between long-term ${\mathrm{PM}}_{2.5}$ exposure and bone homeostasis, as well as the potential underlying biological mechanisms by which ${\mathrm{PM}}_{2.5}$ could affect bone health, by conducting both an observational epidemiological study and experimental in vivo and in vitro examination.

Methods

Epidemiological Study

Study population.

The UK Biobank data set (application number: 41376)^27–29 involves a large-scale longitudinal cohort with $\sim \mathrm{500,000}$ middle-aged participants recruited from 17 centers in England, 2 in Scotland, and 3 in Wales since 2006 (original ethics committee approval number: 21/NW/0157). The UK Biobank data set was used to conduct an analysis of prospective observational data to investigate the relationship between long-term ${\mathrm{PM}}_{2.5}$ exposure and BMD/fracture in humans.

Outcome assessment.

The outcomes of this study were BMD, including heel BMD measured by quantitative ultrasound, and femur neck and lumbar spine BMD measured by dual energy X-ray absorptiometry (DXA) during 2014–2020. Here, as in earlier publications,^30,31 to minimize the influence of residual confounding, we excluded participants who had diseases that might affect BMD (i.e., secondary fracture, rheumatoid arthritis, and lupus erythematosus; Table S1). In addition, given that the exposure (i.e., annual average ${\mathrm{PM}}_{2.5}$ concentration) was estimated for the year 2010,³² we included only the information of BMD estimated after 2010 ($n=\mathrm{44,119}$ for heel BMD; $n=\mathrm{36,896}$ for femur neck and lumbar spine BMD). After excluding participants lacking covariate information (covariates are listed in Table S2), 37,440 participants were included in the analyses of the association for heel BMD and 29,766 participants for femur neck and lumbar spine BMD.

$P{M}_{\mathrm{2.5}}$ exposure assessment.

In the UK Biobank, the average ${\mathrm{PM}}_{2.5}$ concentration in 2010 was estimated using land use regression (LUR), a model that was developed as a part of the European Study of Cohorts for Air Pollution Effects (http://www.escapeproject.eu/).³³ In the LUR model, the spatial variations of annual average ${\mathrm{PM}}_{2.5}$ concentration for each address were obtained from Geographic Information System (GIS)–derived predictors, including traffic, land use, and topography.³³ These annual average ${\mathrm{PM}}_{2.5}$ concentrations (available for the year 2010) were linked to each participant in the UK through participants’ residential addresses given at the baseline visit (2006–2010).

Statistical methods.

Multivariate linear regression was used to estimate the association between long-term exposure to ${\mathrm{PM}}_{2.5}$ (in micrograms per meter cubed) and BMD (in grams per centimeter squared). Results are presented as the change in BMD in grams per centimeter squared for a $1\text{-}{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in annual ${\mathrm{PM}}_{2.5}$ concentrations. In the linear regression model, we included potential BMD-related risk factors as covariates, as in our previous work.^30,31 Specifically, two models were used: a) the basic model that adjusting for sex (categorical: male and female), age (continuous: in years), ancestry (categorical: European, Mixed, Asian, Black, Chinese, and Other ethnic groups), and body mass index (BMI; continuous: in kilograms per meter squared) (model 1); and b) the fully adjusted model, that is, the basic model with additional covariates of education level (categorical: yes and no), smoking (categorical: never, previous, and current), alcohol use (categorical: never, previous, and current), physical activity (categorical: yes and no), circulating calcium (continuous: in millimoles per liter), and calcium supplementary status (categorical: yes and no) (model 2) (Table S2). Furthermore, we conducted an analyses stratified by sex to assess the effect in males and females separately. R (version 4.1.0; R Development Core Team) was used for the statistical analyses, and values of $p<0.05$ were regarded as statistically significant.

Experimental Study

Reagents.

All the antibodies used in this study are listed in Table S3. The Alcian blue [Chemical Abstract Service (CAS) #75881-23-1], hematoxylin (CAS #517-28-2), orange G (CAS #1936-15-8), and tartrate-resistant acid phosphatase (TRAP) staining kit (Sigma; #387) were all obtained from Sigma-Aldrich. Cell culture-related reagents purchased from Gibco included alpha-minimum essential medium ($\mathrm{\alpha }\text{-MEM}$ medium; #12561-056), fetal bovine serum (FBS; #10099-141), nonessential amino acid (NEAA; #11140-050), and GlutaMAX (#35050-061). Both macrophage colony-stimulating factor (M-CSF; #315-02) and receptor activator of nuclear factor-kappa ligand (RANKL; #315-11C) were purchased from Pepro Tech. Mouse tumor necrosis factor-alpha ($\text{TNF-}\mathrm{\alpha }$) neutralizing antibody (D2H4; #11969) was purchased from Cell Signaling Technology. The enzyme-linked immunosorbent assay (ELISA) kit for bone alkaline phosphatase (BALP; #KE1424), osteocalcin (OCN; #KE1428), and TRAP 5b (TRACP-5b; #KE1580) were obtained from ImmunoWay Biotechnology, whereas the ELISA kit for $\text{TNF-}\mathrm{\alpha }$ (#MTA00B) was purchased from R&D.

Animals and animal care.

In this study, twelve 5-wk-old male C57BL/6 mice were obtained from Shanghai Laboratory Animal Co., Ltd (SLAC). All the mice were housed at $24\pm 2°\mathrm{C}$ with $60\pm 10\%$ humidity and maintained on a 12-h light/12-h dark cycle. In addition, water and food were freely available. The animals were treated humanely during the experiment to minimize suffering, and the animal experimental protocols were approved by the Zhejiang Chinese Medical University Animal Care and Use Committee.

$P{M}_{2.5}$ inhalational exposure protocol.

The mice were randomly divided into two groups ($n=6$ per group) and then consecutively exposed to filtered air (FA) or ${\mathrm{PM}}_{2.5}$ using the Zhejiang Whole-body Exposure System 1 (ZJ-WES1) for 6 months from May to November in 2018 (6 mice per cage). The ZJ-WES1 is located at the campus of Zhejiang Chinese Medical University, as described in detail previously.³⁴ Briefly, the exposure system consists of two temperature-controlled chambers with the same volume. Specifically, the PM chamber was filled with ambient air except for the particles $>2.5\mathrm{\mu }\mathrm{m}$, which were cleared by a cyclone. The FA chamber is able to eliminate ${\mathrm{PM}}_{2.5}$ from the air stream with a high-efficiency particulate air filter (Shandong JuKang Technology Co., Ltd. H14; #JKKJ-200) placed in the inlet valve.

$P{M}_{\mathrm{2.5}}$ concentration measurement and component analysis.

As in our previous studies,^35,36 the ${\mathrm{PM}}_{2.5}$ concentration in the chambers was reflected in real time by an aerosol monitor model pDR-1500 (Thermo Scientific). To confirm the accurate concentration of ${\mathrm{PM}}_{2.5}$ and analyze the main components, the collection of ${\mathrm{PM}}_{2.5}$ samples from the chambers was implemented with Teflon filter membranes ($37\mathrm{mm}$; GE HealthCare) or a quartz filter ($47\mathrm{mm}$; MTL). The weight of the membranes was measured using an Excellence Plus XP microbalance (Mettler Toledo) in a temperature- and humidity-controlled room.³⁴ Membrane weight before and after sampling was used to calculate the exposure concentration.

To analyze trace metal elements, the sampling filter membrane was cut with ceramic scissors and immersed in $10\mathrm{mL}$ of nitric acid solution (5%) followed by an ultrasonic water bath at 70°C for 3 h. Then, the suspension was centrifuged at 4,500 rpm for 5 min, and the supernatant was filtered with a $0.45\text{-}\mathrm{\mu }\mathrm{m}$ strainer and collected for measurement. The concentrations of corresponding trace metal elements [mercury (Hg), antimony (Sb), aluminum (Al), arsenic (As), beryllium (Be), cadmium (Cd), chromium (Cr), lead (Pb), manganese (Mn), nickel (Ni), selenium (Se), and thallium (Tl)] were determined with inductively coupled plasma mass spectrometry (iCAP Qc ICP-MS; Thermo Scientific) and analyzed with Qtegra Instrument Control software. The stock solution of metal standards [$\mathrm{1,000}\mathrm{\mu }\mathrm{g}/\mathrm{mL}\mathrm{Hg}$, $\mathrm{1,000}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ Se, and $100\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ multielement (Sb, Al, As, Be, Cd, Cr, Pb, Mn, Ni, and Tl)] was diluted in a 95% nitric acid solution and used to draw standard curves. The curves with linear correlation coefficients of $\ge 0.999$ were deemed available for sample concentration calculation. A solution containing scandium (Sc), yttrium (Y), germanium (Ge), and rhodium (Rh) was applied to provide internal standards. In addition, quality control measures included device calibration before analysis, reagent and procedural blanks, and synchronous analysis of 20% samples. The same settings of availability of standard curve and quality control were applied to the analysis of the other two components.

For water-soluble inorganic ions analysis, the sampling filter membrane was cut with ceramic scissors and immersed in ultrapure water ($25\mathrm{mL}$) at 20°C for 40 min. The extract was centrifuged at 11,000 rpm for 1 min, and the supernatant was filtered with a $0.22\text{-}\mathrm{\mu }\mathrm{m}$ filter. Various water-soluble inorganic ions [fluoride ion (${\mathrm{F}}^{-}$), chloride ion (${\mathrm{Cl}}^{-}$), nitrate (${\mathrm{NO}}_3{}^{-}$), sulfate (${\mathrm{SO}}_4{}^{2-}$), and ammonium ion (${\mathrm{NH}}_4{}^{+}$)] were determined by ion chromatography (ICS-600; Thermo Scientific) equipped with a dual-piston pulse infusion pump system. The separations were performed with a chromatographic column ($4\times 250\mathrm{mm}$) at 25°C, and the eluent solution contained $20\mathrm{mM}$ methanesulfonic acid for cations and $4.5\mathrm{mM}$ sodium carbonate and $0.8\mathrm{mM}$ sodium bicarbonate for anions and was used at a flow rate of $1\mathrm{mL}/\min$. The injection volume was $200\mathrm{\mu }\mathrm{L}$. The standard stock solution of ${\mathrm{F}}^{-}$, ${\mathrm{Cl}}^{-}$, ${\mathrm{NO}}_3{}^{-}$, ${\mathrm{SO}}_4{}^{2-}$, and ${\mathrm{NH}}_4{}^{+}$ was diluted to 0.1, 0.2, 0.5, 1, 2, 5, 10, and $20\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ with ultrapure water and used for standard curve drawing. Finally, the data were analyzed using Chromeleon 7 software (version 7.1.2.1478).

As for the detection of polycyclic aromatic hydrocarbons (PAHs), the sampling filter membrane was cut up and immersed in $5\mathrm{mL}$ decafluorobiphenyl solution ($0.2\mathrm{\mu }\mathrm{g}/\mathrm{mL}$, dissolved in acetonitrile). The samples were sonicated for 30 min and centrifuged at 8,000 rpm for 5 min. Next, the supernatant was collected and filtered with a $0.45\text{-}\mathrm{\mu }\mathrm{m}$ strainer for examination of the 16 PAHs, which included naphthalene (NAP), acenaphthylene (ANY), fluorene (FLU), acenaphthene (ANA), phenanthrene (PHE), anthracene (ANT), fluoranthene (FLT), pyrene (PYR), chrysene (CHR), benzo(a)anthracene (BaA), benzo(b)fluoranthene (BbF), benzo(k)fluoranthene (BkF), benzo(a)pyrene (BaP), dibenzo(a,h)anthracene (DBA), benzo(g,h,i)perylene (BPE), and indeno(1,2,3-cd)pyrene (IPY). The separations were conducted with a column (C18; $250\mathrm{mm}\times 4.6\mathrm{mm}$, $3.5\text{-}\mathrm{\mu }\mathrm{m}$ particle size) at 35°C, with a flow rate of $1\mathrm{mL}/\min$ acetonitrile as the mobile phase. The injection volume was $20\mathrm{\mu }\mathrm{L}$. The $20\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ standard stock solution containing the 16 targeted PAHs was diluted to 0.01, 0.02, 0.05, 0.1, 0.2, and $0.3\text{\hspace{0.17em}ng}/\mathrm{mL}$ with acetonitrile for standard curves. Then the content of the total 16 PAHs was determined by liquid chromatography (Agilent1260) and analyzed using OpenLAB CDS software (version C.01.06).

Sample collection.

After long-term ${\mathrm{PM}}_{2.5}$ exposure (6 months), the mice were euthanized with isoflurane and the blood sample was collected by eyeball enucleation. Each blood sample was placed at room temperature for 30 min after centrifugation with 2,000 relative centrifugal force (RCF) for 10 min at 4°C to obtain serum samples, which were stored at $-80°\mathrm{C}$ for ELISA analysis. After blood sample collection, the mice were sacrificed and livers were collected and washed with phosphate buffered saline (PBS) and then stored at $-80°\mathrm{C}$ for further analysis. In addition, intact femurs were uniformly dissected from the right hind limb followed by fixation with 4% paraformaldehyde (PFA) for 3 d at room temperature for further analysis [micro-computed tomography (micro-CT) and all the histological examinations].

Micro-CT analyses.

All the fixed femur samples were evaluated using micro-CT scanning, as described in the previous study.³⁷ In brief, each sample was scanned with a micro-CT scanner (SkyScan; #1176) with the parameters of $\sim 45\text{-kV}$ voltage, $500\text{-}\mathrm{\mu }\mathrm{A}$ current, 780-ms integration time, $\mathrm{4,000}\times \mathrm{2,672}$ resolution, and $9\text{-}\mathrm{\mu }\mathrm{m}$ slice thickness. The scanned images were performed successively by reconstruction software (NRecon; version 1.6, SkyScan), reposition software (Dataviewer; version 1.5; SkyScan), and analysis software (CTAn; version 1.15; SkyScan), in sequence, to obtain the analyzed femoral parameters.

In addition, three-dimensional (3D) reconstruction images were provided by visualization software (CTVolx; version 3.0; SkyScan). Referring to the previous study,³⁸ in brief, the evaluated region of interest (ROI) for analysis was drawn beginning from the distal femoral metaphyseal growth plate and extending proximally for 100 CT slices. For quantitative analyses, bone parameters, such as BMD, trabecular bone volume fraction (BV/TV), trabecular number (Tb. N), trabecular separation (Tb. Sp), and trabecular thickness (Tb. Th) were collected from the analysis software. After micro-CT analyses, the measurements of femur length and width were precisely conducted by reposition software (Dataviewer; version 1.2; SkyScan) rather than manual measurements. According to the anteroposterior radiograph of full-length femur projection, the length was measured from the greater trochanter to the femoral distal lateral condyle. The femoral width was measured at the traverse position, extending proximally for 15% of the total femur length from the distal femoral growth plate. All the metrical data were accurate to within $0.01\mathrm{mm}$.

Histomorphology analyses.

For histological analyses, all the femurs were decalcified using 14% ethylenediaminetetraacetic acid solution at pH 7.2 for 2 wk following micro-CT scanning. Subsequently, the samples were dehydrated with gradient alcohol. After dehydration, the femur samples were embedded in paraffin with a sagittal orientation. Each paraffin block was sectioned with microtome (Thermo Fisher Scientific; #HM 355S) at a $3\text{-}\mathrm{\mu }\mathrm{m}$ thickness for all staining and immunohistochemistry. Then, Alcian blue hematoxylin/orange G (ABH/OG) staining was carried out for the histological evaluation on the distal femoral metaphysis. Images were captured using a microscope (Zeiss AxioCam HRc Scope.A1) and ImageJ software (version 1.5)³⁹ was used for quantification analyses.

TRAP staining.

In addition, TRAP staining was performed to determine the osteoclast activity. For the tissue detection, the staining was conducted on the $3\text{-}\mathrm{\mu }\mathrm{m}$ thickness paraffin sections from femur tissue. Briefly, sections were rehydrated and incubated with basic stock solution containing napthol AS-BI phosphate ($0.2\mathrm{mg}/\mathrm{mL}$) for an hour at 37°C. Then, sample sections were incubated for 10 min at 37°C in a mixture solution that included sodium nitrite ($0.8\mathrm{mg}/\mathrm{mL}$) and pararosaniline dye ($1.0\mathrm{mg}/\mathrm{mL}$). Next, all the sections were counterstained with Fast Green. Ultimately, multinuclear TRAP-positive cells around the distal femoral growth plate, the key site for bone remodeling in mice,^25,40 were calculated using ImageJ software (version 1.5).³⁹

Immunohistochemistry.

A series of markers for osteoblasts, osteoclasts, and inflammation were examined, as previously reported.⁴¹ In brief, deparaffinized sections of femur tissue were treated by sodium citrate (pH 6.0) at 95°C for antigen retrieval followed by eliminating the activity of endogenous peroxidase using 0.3% hydrogen peroxide, and then blocked for 20 min with normal goat serum at room temperature. After that, the samples were incubated with the primary antibody at 4°C overnight and with the corresponding secondary antibody at room temperature for 20 min. The positive signal was then visible with diaminobenzidine solution (Invitrogen). Ultimately, hematoxylin was used for counterstaining.

In the present study, anti-alkaline phosphatase (ALP) antibody (Arigo; #ARG57422; 1:300), anti-OCN antibody (Abcam; #ab13420; 1:300), anti-cathepsin K (Ctsk) antibody (Abcam; #ab188604; 1:200), anti-RANKL antibody (Abcam; #ab45039; 1:300), $\text{anti-interleukin-}1\mathrm{\beta }$ ($\text{IL-}1\mathrm{\beta }$) antibody (Abcam; #ab9722; 1:300), and $\text{anti-TNF-}\mathrm{\alpha }$ antibody (Arigo; #ARG56080; 1:300) were used to detect the expression level of the corresponding proteins. The positively stained area was measured using ImageJ software (version 5.0).³⁹

Cell culture and conditional medium collection from $P{M}_{\mathrm{2.5}}\text{-stimulated}$ macrophages.

Primary bone marrow macrophages (BMMs) or murine monocytic macrophage RAW264.7 cells (Cell Bank of the Chinese Academy of Sciences) were cultured for conditional medium collection. For isolation of primary BMMs, 10-wk-old male C57BL/6 mice exposed to FA were sacrificed and disinfected with 75% alcohol. Next, the femurs and tibias were separated and the bone marrow was flushed with $\mathrm{\alpha }\text{-MEM}$ medium (Gibco; #12561-056) containing 2% FBS (Gibco; #10099-141) and 1% penicillin/streptomycin (P/S; NorthEnd; #15140). The cell suspension was then filtered with a $70\text{-}\mathrm{\mu }\mathrm{m}$ cell sieve (BIOLOGIX; #15-1070) and centrifuged at 1,000 rpm for 5 min. Cells were suspended in complete cell culture medium [$\mathrm{\alpha }\text{-MEM}$ supplemented with 10% FBS, 1% P/S, 1% NEAA (Gibco; #11140-050) and 1% GlutaMAX (Gibco; #35050-061) with $10\text{\hspace{0.17em}ng}/\mathrm{mL}$ of M-CSF (Pepro Tech; #315-02). Two days later, the medium was replaced with the same formula except for the final concentration of M-CSF up to $30\text{\hspace{0.17em}ng}/\mathrm{mL}$ for another 2 d (Figure S1A and Table S4). RAW264.7 cells suspended in $\mathrm{\alpha }\text{-MEM}$ medium containing 10% FBS and 1% P/S were allowed to adhere and form a monolayer for 2 d (Figure S1A and Table S5).

As previously described,⁴² the ${\mathrm{PM}}_{2.5}$ suspension (Standard Reference Material 1648a, ID160705; National Institute of Standards and Technology) was made fresh at a concentration of $10\mathrm{mg}/\mathrm{mL}$ in $\mathrm{\alpha }\text{-MEM}$ medium and sonicated for 1 h at room temperature. Both BMMs and RAW264.7 cells were stimulated with ${\mathrm{PM}}_{2.5}$ suspensions at final doses of 25 or $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$, based on preliminary experiments. After a 24-h incubation, the supernatant from the two kinds of cells was respectively collected as conditional medium and filtered through $0.22\text{-}\mathrm{\mu }\mathrm{m}$ sieves. The supernatant was temporarily stored at $-80°\mathrm{C}$ for further experiments.

Osteoclastogenic induction and conditional medium incubation.

To determine the indirect effects of ${\mathrm{PM}}_{2.5}$ on osteoclast differentiation, as shown in Figure S1B, primary BMMs extracted from the bones were initially suspended in complete culture medium in the presence of M-CSF ($10\text{\hspace{0.17em}ng}/\mathrm{mL}$) and seeded in 96-well plates ($2\times {10}^4$ cells per well), 24-well plates ($8\times {10}^5$ cells per well), 12-well plates ($4\times {10}^5$ cells per well), and 6-well plates ($1\times {10}^6$ cells per well) for adherence. Two days later, the medium was exchanged with complete culture medium containing $30\text{\hspace{0.17em}ng}/\mathrm{mL}$ M-CSF and $30\text{\hspace{0.17em}ng}/\mathrm{mL}$ RANKL (Pepro Tech; #315-11C), and the cells were cultured for another 2 d.

Then, the culture media were respectively replaced and the cells were stimulated with 20% conditioned medium from PM-treated BMMs or 5% from RAW264.7 cells for 2 d (Figure S1B and Table S6). When the role of $\text{TNF-}\mathrm{\alpha }$ in the ${\mathrm{PM}}_{2.5}\text{-mediated}$ osteoclastogenesis was explored, $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody (D2H4) at a final dose of $1\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ (Cell Signaling Technology; #11969) was simultaneously employed with conditioned medium from BMMs (Figure S1C and Table S7). At the end of the 2-d incubation, cells were collected for further examination, including cell TRAP staining, bone resorption assay, real-time polymerase chain reaction (RT-PCR), and western blotting.

Cell TRAP staining and bone resorption assay.

Following the full process of BMMs culture, osteoclastogenic induction, in presence of or absence of conditional medium/$\text{TNF-}\mathrm{\alpha }$ neutralizing antibody D2H4 incubation, multinucleated cells in 96-well plates were identified as osteoclasts using TRAP staining. First, the medium was aspirated and the cells were washed with PBS and then fixed in 4% PFA for 20 min at room temperature. Then, a TRAP staining kit (Sigma; #387A) was used to assess the TRAP activity of the cells. TRAP-positive cells that had three or more nuclei were deemed to be osteoclasts. The number of osteoclasts were calculated in each well under a light microscope (OLYMPUS; #IX71).

We carried out a bone resorption assay to analyze the effect of ${\mathrm{PM}}_{2.5}$ on the demineralization function mediated by osteoclasts in vitro. Briefly, primary BMMs were isolated and then seeded into 24-well plates ($8\times {10}^4$ cells per well) coated with hydroxyapatite (CLS3989; #Corning) and cultured with complete culture medium containing M-CSF ($10\text{\hspace{0.17em}ng}/\mathrm{mL}$). The cell culture was performed as shown in Table S7. Next, the conditioned medium was removed and the wells were treated with ammonium hydroxide ($1\text{\hspace{0.17em}mol}/\mathrm{L}$) for 5 min to remove the attached cells. Finally, the areas resorbed by osteoclasts were observed by light microscopy (OLYMPUS; #IX71) and calculated using ImageJ software (version 1.5).³⁹

RNA extraction and RT-PCR.

In this study, the total RNA of the cells seeded in 12-well plates ($4\times {10}^5$ cells per well) was extracted with TRIzol Reagent (TaKaRa). The RNA concentrations were measured using a NanoDrop2000 ultra microspectrophotometer (Thermo Fisher), and the absorbance ratio at 260/280 nm was between 1.8 and 2.0 to ensure RNA quality purity. Then the RNA was reverse transcribed into complementary DNA with PrimeScript RT Master Mix (TaKaRa). As in the previous study,³⁵ the program was completed by the QuantstudioQ7 (Applied Biosystems) using SYBR Green mix (Applied Biosystems) with the following cycling program: 50°C for 2 min, 95°C for 2 min (activation), 40 cycles of 95°C for 15 s (denaturation), and 60°C for 1 min (extension). Here, we adopted the ${}^{\Delta \Delta }\mathrm{CT}$ method to analyze the relative mRNA expression. All the primer sequences of target genes are listed in Table S8.

Western blotting.

Proteins from cells in 6-well plates and mouse liver tissue were extracted with radioimmunoprecipitation assay lysis buffer (Boster Biological Technology Co., Ltd.) on ice. Next, total protein was quantified using a bicinchoninic acid protein assay kit (Thermo Fisher Scientific; #23227), and the equivalent amount of harvested protein, $\sim 15\mathrm{\mu }\mathrm{g}$, was separated on 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) gel and then transferred to polyvinylidene fluoride (PVDF) membranes. The membranes were blocked with 5% bovine serum albumin in Tris-buffered saline-Tween (TBST) and incubated at 4°C overnight with primary antibodies and with the corresponding secondary antibodies [Proteintech; #SA00001-1 (1:5,000) and #SA00001-2 (1:5,000)] at room temperature for 2 h.

The primary antibodies used for the western blot included ALP (Arigo; #ARG57422; 1:2,000), c-Fos (Huabio; #ET1701-95; 1:500), and Traf6 (Huabio; #R1311-2; 1:1,000). We used the ChemiDoc Imaging System (Bio-Rad) to visualize the bands through enhanced chemiluminescence (4A Biotech; #4AW011-200) and quantitated the autoradiograph by densitometric analysis with Image Lab software matching the system. $\mathrm{\beta }\text{-actin}$ (CST; #4970; 1:1,000) or glyceraldehyde 3-phosphate dehydrogenase (GAPDH; Abcam; #ab8245; 1:1,000) was used as the control reference.

ELISA.

Serum samples were collected from mouse whole blood and centrifuged at 2,000 RCF for 10 min at 4°C. In the present study, the concentrations of BALP, OCN, and TRACP-5b in serum were determined with the corresponding ELISA kits (ImmunoWay Biotechnology; BALP, #KE1424; OCN, #KE1428; TRACP-5b, #KE1580) according to the manufacturer’s instructions. In addition, PM-treated conditioned medium was filtered through a $0.22\text{-}\mathrm{\mu }\mathrm{m}$ strainer (Millex; #ROAB40604) and the $\text{TNF-}\mathrm{\alpha }$ levels in the medium were examined with the Mouse $\text{TNF-}\mathrm{\alpha }$ Quantikine ELISA kit (R&D; #MTA00B) according to the manufacturer’s instructions.

Statistical analysis.

All experimental data are presented as $\text{means}\pm \text{SEMs}$ and assessed through unpaired $t$-tests or one-way analyses of variance (ANOVAs). In addition, GraphPad Prism software (version 5.0) was used for the statistical analyses, and values of $p<0.05$ were regarded as statistically significant.

Results

Associations between Long-Term $P{M}_{2.5}$ Exposure and BMD in Human

We used UK Biobank data to analyze the association between long-term exposure to ${\mathrm{PM}}_{2.5}$ (estimated at 2010) and heel BMD (measured during 2014–2020). After excluding participants lacking covariate information and those with diseases that might affect BMD, 37,440 participants ($n=\mathrm{18,849}$ for females and $n=\mathrm{18,591}$ for males) remained (Table S9). The $\text{mean}\pm \text{SD}$ age of these participants was $55.80\pm 7.54$ y, ranging from 40 to 73 y. The $\text{mean estimate}\pm \text{SD}$ of ${\mathrm{PM}}_{2.5}$ concentrations was $9.918\pm 1.031{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$. For the femur neck and lumbar spine BMD analyses, 29,766 participants were included (Table S9) (the mean age was $55.05\pm 7.48$ y, range: 40–70 y; the mean concentration was $9.913\pm 1.039{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ for ${\mathrm{PM}}_{2.5}$; $n=\mathrm{14,968}$ for females, $n=\mathrm{14,798}$ for males). We observed that long-term exposure to ${\mathrm{PM}}_{2.5}$ was associated with decreased heel BMD as estimated by quantitative ultrasound in the basic linear regression model adjusting for sex, age, ancestry, and BMI (model 1) [$\mathrm{\beta }\text{-coefficient}\pm \text{standard error\hspace{0.17em}}\left(\text{SE}\right)$ was $-0.003\pm 0.001$, $p=2.70\times {10}^{-8}$] (Table 1). Here the $\mathrm{\beta }\text{-coefficient}$ represents the change in BMD in grams per centimeter squared for a $1\text{-}{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in annual ${\mathrm{PM}}_{2.5}$ concentrations. The associations remained statistically significant in the fully adjusted model, which included additional covariates such as education level, smoking, alcohol use, physical activity, circulating calcium and calcium supplementary status (model 2) ($\mathrm{\beta }\text{-coefficient}\pm \text{SE}$ was $-0.003\pm 0.001$, $p=4.53\times {10}^{-5}$; Table 1). We also assessed the relationship between long-term exposure to ${\mathrm{PM}}_{2.5}$ and BMD assessed by DXA at the lumbar spine ($\mathrm{\beta }\text{-coefficient}\pm \text{SE}$ was $-0.006\pm 0.001$, $p=4.61\times {10}^{-9}$ in model 1, and $\mathrm{\beta }\text{-coefficient}\pm \text{SE}$ was $-0.005\pm 0.001$, $p=4.18\times {10}^{-6}$ in model 2) and at the femur neck ($\mathrm{\beta }\text{-coefficient}\pm \text{SE}$ was $-0.003\pm 0.001$, $p=8.89\times {10}^{-6}$ in model 1, and $\mathrm{\beta }\text{-coefficient}\pm \text{SE}$ was $-0.002\pm 0.001$, $p=5.60\times {10}^{-4}$ in model 2; Table 1). In the stratified analyses by sex, the trends were similar in both males and females (Table 2).

Table 1.

The observational association of long-term exposure to ${\mathrm{PM}}_{2.5}$ with BMD measured after the year 2010 at different anatomical sites ($n=\mathrm{37,440}$ for heel BMD; $n=\mathrm{29,766}$ for lumbar spine 1 to 4 BMD; $n=\mathrm{29,766}$ for femur neck BMD) in the UK Biobank.

	$\mathrm{\beta }\text{-coefficient}$ a	SE	$p$-Valueb
Heel BMD
Model 1	$-0.003$	0.001	$2.70\times {10}^{-8}$
Model 2	$-0.003$	0.001	$4.53\times {10}^{-5}$
Lumbar spine 1 to 4 BMD
Model 1	$-0.006$	0.001	$4.61\times {10}^{-9}$
Model 2	$-0.005$	0.001	$4.18\times {10}^{-6}$
Femur neck BMD (right)
Model 1	$-0.003$	0.001	$8.89\times {10}^{-6}$
Model 2	$-0.002$	0.001	0.001

Note: Model 1 was adjusted for confounders, including sex, age, body mass index, and ancestry; $\text{model\hspace{0.17em}}2=\text{model\hspace{0.17em}}1+\text{education}+\text{smoking}+\text{alcohol}+\text{physical activity}+\text{circulating calcium}+\text{calcium}$ supplementary status. BMD, bone mineral density; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; SE, standard error.

^a The $\mathrm{\beta }\text{-coefficient}$ represents the change in BMD in grams per centimeter squared for a $1\text{-}{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in annual ${\mathrm{PM}}_{2.5}$ concentrations.

^b All $p$-values were derived from multiple linear regression.

Table 2.

The observational association of long-term exposure to ${\mathrm{PM}}_{2.5}$ with BMD measured after the year 2010 in males ($n=\mathrm{18,591}$ for heel BMD; $n=\mathrm{14,798}$ for lumbar spine 1 to 4 BMD; $n=\mathrm{14,798}$ for femur neck BMD) and females ($n=\mathrm{18,849}$ for heel BMD; $n=\mathrm{14,968}$ for lumbar spine 1 to 4 BMD; $n=\mathrm{14,968}$ for femur neck BMD) in the UK Biobank.

	$\mathrm{\beta }\text{-coefficient}$ a	SE	$p$-Valueb
Heel BMD
Male
Model 1	$-0.004$	0.001	$1.68\times {10}^{-5}$
Model 2	$-0.003$	0.001	0.001
Female
Model 1	$-0.002$	0.001	0.001
Model 2	$-0.002$	0.001	0.036
Lumbar spine 1 to 4 BMD
Male
Model 1	$-0.007$	0.001	$1.54\times {10}^{-6}$
Model 2	$-0.006$	0.002	$2.19\times {10}^{-4}$
Female
Model 1	$-0.004$	0.001	0.004
Model 2	$-0.003$	0.001	0.027
Femur neck BMD (right)
Male
Model 1	$-0.004$	0.001	$6.84\times {10}^{-5}$
Model 2	$-0.003$	0.001	0.004
Female
Model 1	$-0.002$	0.001	0.055
Model 2	$-0.002$	0.001	0.113

Note: Model 1 was adjusted for confounders, including age, body mass index, and ancestry; $\text{model\hspace{0.17em}}2=\text{model\hspace{0.17em}}1+\text{education}+\text{smoking}+\text{alcohol}+\text{physical activity}+\text{circulating calcium}+\text{calcium}$ supplementary status. BMD, bone mineral density; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; SE, standard error.

^a The $\mathrm{\beta }\text{-coefficient}$ represents the change in BMD in grams per centimeter squared for a $1\text{-}{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$ increase in annual ${\mathrm{PM}}_{2.5}$ concentrations.

^b All $p$-values were derived from multiple linear regression.

$P{M}_{2.5}$ Concentration Measurement and Component Analyses of Animal Study

As shown in Figure 1A, the mean ${\mathrm{PM}}_{2.5}$ concentration of ambient air at the PM chamber was $29.27{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$, whereas in the FA chamber it was $1.30{\mathrm{\mu }\mathrm{g}/\mathrm{m}}^3$. The lowest ${\mathrm{PM}}_{2.5}$ concentration was in July and August, whereas the highest concentration was in May and October (Figure 1B). As for the component analysis, the major water-soluble inorganic ions in the composition of ${\mathrm{PM}}_{2.5}$ included ${\mathrm{SO}}_4{}^{2-}$, ${\mathrm{NO}}_3{}^{-}$, ${\mathrm{NH}}_4{}^{+}$, and ${\mathrm{Cl}}^{-}$ (Figure 1C), whereas ${\mathrm{F}}^{-}$ was not detected. Regarding the 12 kinds of trace metals in the ${\mathrm{PM}}_{2.5}$, the major trace metals were Al, Pb, and Mn, which were $>10{\text{\hspace{0.17em}ng}/\mathrm{m}}^3$, whereas Ni, As, Se, Sb, and Cr were $1–5{\text{\hspace{0.17em}ng}/\mathrm{m}}^3$ and only tiny amounts of Cd, Tl, Hg, and Be were detected (Figure 1E). In the 16 PAHs tested in the ${\mathrm{PM}}_{2.5}$ (Figure 1G), the most abundant PAHs was BbF, followed by IPY, PYR, DBA, and BPE. In addition, the traces of the ${\mathrm{PM}}_{2.5}$ compositions during the exposure duration are shown in Figure 1D,F,H. Generally, most of the main components remained at lower levels in July and August and increased in the fall, reaching a peak in November.

Figure 1.

${\mathrm{PM}}_{2.5}$ concentration and component analysis during the mice exposure period. (A) Average concentrations and (B) monthly concentrations in the filtered air (FA), ${\mathrm{PM}}_{2.5}$ (PM) groups ($n\ge 10$). (C) Average concentrations and (D) monthly concentrations of water-soluble inorganic ions in the ${\mathrm{PM}}_{2.5}$ group ($n\ge 6$). (E) Average concentrations and (F) monthly concentrations of trace metals in the ${\mathrm{PM}}_{2.5}$ group ($n\ge 6$). (G) Average concentrations and (H) monthly concentrations of PAHs in the ${\mathrm{PM}}_{2.5}$ group ($n\ge 6$). Numeric data can be found in Table S10. Data are presented as $\text{means}\pm \text{SEMs}$. Note: Al, aluminum; ANA, acenaphthene; ANT, anthracene; ANY, acenaphthylene; As, arsenic; Aug, August; BaA, benzo(a)anthracene; BaP, benzo(a)pyrene; BbF, benzo(b)fluoranthene; Be, beryllium; BkF, benzo(k)fluoranthene; BPE, benzo perylene; Cd, cadmium; CHR, chrysene; ${\mathrm{Cl}}^{-}$, chloride ion; Cr, chromium; DBA, dibenzo anthracene; FLT, fluoranthene; FLU, fluorene; Hg, mercury; IPY, indeno pyrene; Jul, July; Jun, June; Mn, manganese; NAP, naphthalene; ${\mathrm{NH}}_4{}^{+}$, ammonium ion; Ni, nickel; ${\mathrm{NO}}_3{}^{-}$, nitrate; Nov, November; Oct, October; PAHs, polycyclic aromatic hydrocarbons; Pb, lead; PHE, phenanthrene; PM, particulate matter; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; PYR, pyrene; Sb, antimony; Se, selenium; SEM, standard error of the mean; Sep, September; ${\mathrm{SO}}_4{}^{2-}$, sulfate; Tl, thallium.

Long-Term $P{M}_{2.5}$ Exposure, Femur Bone Growth, and Osteogenesis in C57BL/6 Mice

After 6 months of FA or ${\mathrm{PM}}_{2.5}$ exposure, no apparent difference in appearance was observed between two groups. Initially, we examined the length and width of the mouse femur, the most representative bone in the skeletal system, to explore the relationship between air pollution and skeletal health. To avoid the interference from soft tissue attached to the surface of the bone, anteroposterior radiograph of the full-length femur was performed to measure precisely (Figure 2A). As shown in Figure 2B, femoral length was shorter in the ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice compared with the mice exposed to FA. In addition, femoral width was a bit narrower with long-term ${\mathrm{PM}}_{2.5}$ exposure, although there was no statistical difference (Figure 2C). These data suggest that ${\mathrm{PM}}_{2.5}$ air pollution exposure affected the femur growth in length.

Figure 2.

The effect of ${\mathrm{PM}}_{2.5}$ on bone morphology and osteogenesis in vivo. Mice were exposed to FA or ${\mathrm{PM}}_{2.5}$ for 6 months. (A) Representative coronal images of intact femur. Scale bar: $1\mathrm{mm}$. Quantitative analyses of femur (B) length and (C) width. (D) Representative images of 3D reconstruction of ROI. Scale bar: $500\mathrm{\mu }\mathrm{m}$. Micro-CT quantitative analyses of (E) cortical bone mineral density and (F) trabecular bone mineral density. (G) Representative images of OCN immumohistochemical staining. Red arrows indicate positive expression. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (H) Quantification of OCN positive area. (I) Quantitative analyses of OCN levels in serum. (J) Representative images of ALP immumohistochemical staining. Red arrows indicate positive expression. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (K) Quantification of ALP positive area. (L) Quantitative analysis of BALP levels in serum. (M) Expression of ALP in liver assessed by western blotting. (N) Quantitative analysis of ALP expressions in liver. Numeric data can be found in Table S11. Data are presented as $\text{means}\pm \text{SEMs}$ (unpaired Student’s $t$-test, $n=4–6$), *$p<0.05$, ***$p<0.001$. Note: 2D, two dimensions; 3D, three dimensions; ALP, alkaline phosphatase; BALP, bone alkaline phosphatase; BMD, bone mineral density; FA, filtered air; micro-CT, micro-computed tomography; OCN, osteocalcin; PM, particulate matter; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; ROI, region of interest; SEM, standard error of the mean.

Based on the finding that ${\mathrm{PM}}_{2.5}$ air pollution had a slight effect on the femur length, we performed micro-CT evaluation on the distal femoral metaphysis, the most representative site reflecting BMD and bone mass in mice,⁴³ to determine whether the skeletal structure was altered by ${\mathrm{PM}}_{2.5}$ exposure. As illustrated in Figure 2D–F, there was no visual (Figure 2D) or statistical (Figure 2E,F) difference in either cortical or trabecular BMD between the FA and PM groups. Parameters reflecting bony architecture, including BV/TV, Tb.N, Tb.Sp, and Tb.Th, were evaluated, and no significant difference was observed with ${\mathrm{PM}}_{2.5}$ exposure (Figure S2).

To determine whether air pollution modulates osteogenesis, levels of the most representative osteogenic markers, OCN and BALP,⁴⁴ were examined. Immunohistochemical analysis showed the OCN expression was significantly lower in response to ${\mathrm{PM}}_{2.5}$ exposure in the distal femur than that in FA controls (Figure 2G,H), which was in accord with the circulating levels of OCN in the serum (Figure 2I). In addition, ALP expression around the distal femur was examined with immunohistochemistry. However, no significant discrepancy of ALP expression around the growth plate was observed between the two groups (Figure 2J,K). Surprisingly, levels of BALP were significantly higher in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice than in mice exposed to FA (Figure 2L). Because the liver is an attributable source of ALP synthesis, we also examined ALP expression in the liver. Interestingly, protein levels of ALP in the liver of ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice were significantly higher than those of the FA group (Figure 2M,N).

Long-Term $P{M}_{2.5}$ Exposure, Osteoclastogenesis, and Inflammation around the Femoral Distal Metaphysis in Mice

Because the femoral length was highly relevant to bone metabolism, we detected the osteoclastogenesis around the femoral growth plate. Immunohistochemical analysis revealed that mice exposed to ${\mathrm{PM}}_{2.5}$ expressed a higher RANKL level in the distal femur; given that RANKL is obligatory for bone resorption,⁴⁵ this suggests its expression may promote osteoclastogenesis (Figure 3A,B). Then, we performed TRAP staining to confirm the osteoclast formation around growth plate. As expected, more TRAP-positive areas were observed around the growth plate of ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice compared with that of the control group (Figure 3C,D), suggesting the number of osteoclast precursors or osteoclasts increased after ${\mathrm{PM}}_{2.5}$ exposure. In addition, the circulating levels of TRACP-5b in serum were higher in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice than those in FA-exposed mice as well (Figure 3E).

Figure 3.

The effect of ${\mathrm{PM}}_{2.5}$ on osteoclastogenesis and inflammation in vivo. Mice were exposed to FA or ${\mathrm{PM}}_{2.5}$ for 6 months. (A) Representative images of RANKL immunohistochemical staining around the growth plate. Red arrows indicate positive expression. Scale bar: $100\mathrm{\mu }\mathrm{m}$. (B) Quantification of RANKL-positive area around the growth plate. (C) Representative images of TRAP staining around the femoral distal growth plate. Images in the right panel are higher magnifications of images in the respective left panel. Scale bar: $200\mathrm{\mu }\mathrm{m}$. (D) Quantitative analysis of TRAP-positive cells around the femoral distal growth plate. (E) Quantitative analysis of TRACP-5b levels in serum. (F) Representative images of ABH/OG staining around the femoral distal growth plate. Red arrows indicate growth plate destruction. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (G) Representative images of Ctsk immunohistochemical staining. Red arrows indicate positive expression. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (H) Quantification of the Ctsk-positive area. (I) Representative images of $\text{TNF-}\mathrm{\alpha }$ immumohistochemical staining. Red arrows indicate positive expression. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (J) Quantification of $\text{TNF-}\mathrm{\alpha }$–positive area. (K) Representative images of $\text{IL-}1\mathrm{\beta }$ immumohistochemical staining. Red arrows indicate positive expression. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (L) Quantification of $\text{IL-}1\mathrm{\beta }$–positive area. Numeric data can be found in Table S12. Data are presented as $\text{means}\pm \text{SEMs}$ (unpaired Student’s $t$-test, $n=4–6$), *$p<0.05$, **$p<0.01$, ***$p<0.001$. Note: ABH/OG, Alcian blue hematoxylin/orange G; Ctsk, cathepsin K; FA, filtered air; $\text{IL-}1\mathrm{\beta }$, interleukin-1beta; PM, particulate matter; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; RANKL, receptor activator of nuclear factor-kappa ligand; SEM, standard error of the mean; $\text{TNF-}\mathrm{\alpha }$, tumor necrosis factor-alpha; TRACP-5b, tartrate-resistant acid phosphatase 5b; TRAP, tartrate-resistant acid phosphatase.

Histological analysis of the distal femur was performed by ABH/OG staining. Intriguingly, much more lacunae occupied by chondroclast-like cells around the growth plate were detected, suggesting that the structure of calcified cartilaginous template may have been impaired in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice (Figure 3F). In concurrence with ABH/OG staining, immunohistochemical analysis of Ctsk, a pivotal indicator of mature osteoclasts, showed a higher expression in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice (Figure 3G,H).

Considering that inflammation is one of the key factors mediating osteoclast formation,⁴⁶ immunohistochemistry was performed to analyze the changes of proinflammatory cytokine ($\text{TNF-}\mathrm{\alpha }$ and $\text{IL-}1\mathrm{\beta }$) expression around the femoral growth plate. Our data showed that more positive staining areas of $\text{TNF-}\mathrm{\alpha }$ were observed around the growth plate in mice exposed to ${\mathrm{PM}}_{2.5}$ (Figure 3I,J), whereas no difference was shown with $\text{IL-}1\mathrm{\beta }$ expression between the FA and PM groups (Figure 3K,L).

$P{M}_{2.5}$ Stimulation and Osteoclast Differentiation in Vitro

In view of our findings suggesting more pronounced osteoclastogenesis in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice, we next sought to gain further insight into the osteoclast differentiation from macrophages in vitro. In the present study, we treated RAW264.7 cell and BMMs with ${\mathrm{PM}}_{2.5}$ particles. We then employed murine primary BMMs to induce preosteoclasts/osteoclasts. We treated RAW264.7 macrophages with ${\mathrm{PM}}_{2.5}$ particles and the culture medium supernatant was collected for $\text{TNF-}\mathrm{\alpha }$ detection by ELISA. As shown in Figure 4A, we found that there was a dose-dependent difference in $\text{TNF-}\mathrm{\alpha }$ expression in cells exposed to ${\mathrm{PM}}_{2.5}$ compared with control cells. Then, the conditional media supernatant containing $\text{TNF-}\mathrm{\alpha }$ was applied to the osteoclasts differentiated from BMMs and osteoclastogenesis in vitro. Cells exposed to ${\mathrm{PM}}_{2.5}$ exhibited markedly higher gene expression of osteoclast differentiation genes, including acid phosphatase 5 (Acp5), Ctsk, nuclear factor of activated T cells cytoplasmic 1 (Nfatc1), c-Fos, dendritic cell-specific transmembrane protein (Dcstamp), Atp6v0d2, and TNF receptor-associated factor 6 (Traf6) in a dose-dependent manner (Figure 4B). In addition, cells exposed to conditional media from RAW264.7 cells stimulated by ${\mathrm{PM}}_{2.5}$ particles demonstrated greater osteoclast differentiation from BMMs as demonstrated by TRAP staining and quantification of osteoclasts (Figure 4C,D).

Figure 4.

The effect of PM stimulation on osteoclastogenesis in vitro. Preosteoclasts were induced and treated with conditional medium from ${\mathrm{PM}}_{2.5}\text{-stimulated}$ RAW264.7 macrophages and primary BMMs. (A) $\text{TNF-}\mathrm{\alpha }$ release from RAW264.7 macrophages in the supernatant after 24-h incubation with ${\mathrm{PM}}_{2.5}$ particles dosed at 0, 25, and $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$. (B) Relative mRNA expression levels of molecules related to osteoclastogenesis in BMMs after treatment with 5% conditioned medium collected from supernatant of RAW264.7 macrophages. (C) Representative images of TRAP-stained cells treated with 5% conditioned medium collected from supernatant of 0, 25, and $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ ${\mathrm{PM}}_{2.5}$ particles–incubated RAW264.7 macrophages. Scale bar: $100\mathrm{\mu }\mathrm{m}$. (D) Quantification of TRAP-positive multinucleated cells with more than three nuclei on each well. (E) $\text{TNF-}\mathrm{\alpha }$ release from BMMs in the supernatant after 24-h incubation with ${\mathrm{PM}}_{2.5}$ particles dosed at 0, 25, and $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$. (F) Relative mRNA expression levels of molecules related to osteoclastogenesis in BMMs after treatment with 20% conditioned medium collected from supernatant of BMMs induced by ${\mathrm{PM}}_{2.5}$ particles. (G) Representative images of TRAP-stained cells treated with 20% conditioned medium collected from supernatant of 0, 25, and $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ ${\mathrm{PM}}_{2.5}$ particles–incubated BMMs. Scale bar: $100\mathrm{\mu }\mathrm{m}$. (H) Quantification of TRAP-positive multinucleated cells with more than three nuclei on each well. Numeric data can be found in Table S13. Data are presented as $\text{means}\pm \text{SEMs}$ ($n=3–4$) and assessed through one-way ANOVA. *$p<0.05$, **$p<0.01$, ***$p<0.001$ vs. the $0\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ group. ^#$p<0.05$, ^##$p<0.01$, ^###$p<0.001$ vs. the $25\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ group. Note: ACP5, acid phosphatase 5; ANOVA, analysis of variance; Atp6v0d2, ATPase, ${\mathrm{H}}^{+}$ transporting, lysosomal V0 subunit D2; BMMs, bone marrow macrophages; c-Fos, FBJ murine osteosarcoma viral oncogene homolog; Ctsk, cathepsin K; Dcstamp, dendritic cell-specific transmembrane protein; Nfatc1, nuclear factor of activated T cells cytoplasmic 1; PM, particulate matter; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; RAW264.7, a cell line of murine macrophages; SEM, standard error of the mean; TNF-α, tumor necrosis factor-alpha; Traf6, tumor necrosis factor receptor-associated factor 6; TRAP, tartrate-resistant acid phosphatase; UD, undetectable.

We also consistently found that BMMs exposed to ${\mathrm{PM}}_{2.5}$ had higher levels of $\text{TNF-}\mathrm{\alpha }$ (Figure 4E). In addition, BMMs exposed to ${\mathrm{PM}}_{2.5}$ also had greater osteoclast differentiation as demonstrated by TRAP staining and quantification of osteoclasts (Figure 4G,H). The mRNA expression of Traf6, Nfatc1, c-Fos, and Atp6v0d2 in BMMs was higher in cells exposed to ${\mathrm{PM}}_{2.5}$ at $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ but not at $25\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ (Figure 4F).

$\text{TNF-}\mathrm{\alpha }$ Inhibition and $P{M}_{2.5}$-stimulated Osteoclastogenesis in Vitro

We then further explored whether inhibition of $\text{TNF-}\mathrm{\alpha }$ could attenuate the osteoclastic phenotypes induced by ${\mathrm{PM}}_{2.5}$ stimulation. First, we tested whether $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody (D2H4) could affect osteoclastogenesis in cells exposed to ${\mathrm{PM}}_{2.5}$. We found that cells treated with ${\mathrm{PM}}_{2.5}$ and D2H4 had fewer osteoclasts per well than those exposed to ${\mathrm{PM}}_{2.5}$ alone as determined by TRAP staining and subsequent quantification of osteoclasts (Figure 5A,B). Likewise, results from the bone resorption assay revealed a much greater resorption area formed in the ${\mathrm{PM}}_{2.5}$ group when compared with the control group. However, treatment with $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody showed a rescue in bone resorption (Figure 5C,D).

Figure 5.

The effect of $\text{TNF-}\mathrm{\alpha }$ inhibition on ${\mathrm{PM}}_{2.5}\text{-stimulated}$ osteoclastogenesis. Osteoclasts treated with conditional medium from ${\mathrm{PM}}_{2.5}\text{-stimulated}$ primary BMMs were incubated with $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody D2H4. (A) Representative images of TRAP-stained BMMs treated with 20% conditioned medium and additional $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody D2H4. Scale bar: $100\mathrm{\mu }\mathrm{m}$. (B) Quantification of TRAP-positive multinucleated cells with more than three nuclei on each well. (C) Representative images of bone resorption after treatment with $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ ${\mathrm{PM}}_{2.5}$ particles–incubated BMMs with or without D2H4. Scale bar: $50\mathrm{\mu }\mathrm{m}$. (D) Quantification of bone resorbed area in plates coated with hydroxyapatite. (E) Relative mRNA expression levels of molecules related to osteoclastogenesis in BMMs after treatment with conditioned medium collected from supernatant of $50\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ ${\mathrm{PM}}_{2.5}$ particles incubation or plus $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody D2H4. (F) Representative bands of Traf6, c-Fos, and GAPDH in BMMs. (G) Quantification of Traf6 and c-Fos protein expression in PM-induced BMMs with or without D2H4. (H) Schematic diagram depicting the effect of ${\mathrm{PM}}_{2.5}$ in bone homeostasis based on our findings. Numeric data can be found in Table S14. Data are presented as $\text{means}\pm \text{SEMs}$ ($n=3–4$) and assessed through ANOVA. *$p<0.05$, **$p<0.01$, ***$p<0.001$ vs. the $0\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ group. ^#$p<0.05$, ^##$p<0.01$ vs. the $25\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ group. ^$$p<0.05$, ^$$$p<0.01$ vs. the $50\text{-}\mathrm{\mu }\mathrm{g}/\mathrm{mL}$ group. Note: ANOVA, one-way analysis of variance; Atp6v0d2, ATPase, ${\mathrm{H}}^{+}$ transporting, lysosomal V0 subunit D2; BMMs, bone marrow macrophages; c-Fos, FBJ murine osteosarcoma viral oncogene homolog; D2H4, $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; Nfatc1, nuclear factor of activated T cells cytoplasmic 1; OCN, osteocalcin; PM, particulate matter; ${\mathrm{PM}}_{2.5}$, particulate matter with aerodynamic diameter of $\le 2.5\mathrm{\mu }\mathrm{m}$; SEM, standard error of the mean; $\text{TNF-}\mathrm{\alpha }$, tumor necrosis factor-alpha; Traf6, tumor necrosis factor receptor-associated factor 6; TRAP, tartrate-resistant acid phosphatase.

Next, we performed western blot and RT-PCR to detect the expression of downstream molecules expression. RT-PCR results showed that the higher mRNA expression of Nfact1, c-Fos, Atp6v0d2, and Traf6 associated with ${\mathrm{PM}}_{2.5}$ exposure were suppressed by D2H4 treatment (Figure 5E). In addition, both Traf6 and c-Fos protein expressions were significantly higher in cells exposed to conditional medium from RAW cells stimulated by ${\mathrm{PM}}_{2.5}$ in a dose-dependent manner. As expected, these elevated expressions were inhibited by $\text{TNF-}\mathrm{\alpha }$ neutralizing antibody treatment as well (Figure 5F,G).

Discussion

In the present study, with the combination of an observational epidemiological study and an animal toxicological study, we assessed the association among long-term exposure to ${\mathrm{PM}}_{2.5}$, bone metabolism, and the pathogenesis of osteoporosis. Our analysis suggests that long-term exposure to ${\mathrm{PM}}_{2.5}$ was associated with lower BMD. Animal studies further presented shorter femur length, excessive osteoclast activity, inhibited osteoblastic bone formation, and higher $\text{TNF-}\mathrm{\alpha }$ expression around the growth plate in response to ${\mathrm{PM}}_{2.5}$ challenge. In vitro experiments suggested that ${\mathrm{PM}}_{2.5}$ exposure promoted osteoclastogenesis through regulating the $\text{TNF-}\mathrm{\alpha }$/Traf6/c-Fos pathway by the activation of bone resorption. These data, taken together, suggest that long-term exposure to ${\mathrm{PM}}_{2.5}$ was associated with the dysfunction of bone metabolism and the pathogenesis of osteoporosis, which may be mediated by inflammation-induced osteoclastogensis around the growth plate.

Previous epidemiological studies reported inconsistent associations between ${\mathrm{PM}}_{2.5}$ exposure and BMD. For example, a cross-sectional study including 3,717 participants did not show a significant association between air pollution and lumbar spine BMD (mean difference of $-0.018$; 95% confidence interval: $-0.038$, 0.001).⁴⁷ However, results from the population-based Oslo Health Study of 590 men revealed an inverse association between ${\mathrm{PM}}_{2.5}$ and BMD.²⁰ The reasons for these conflicting results were complicated, such as different study designs (i.e., cross-sectional vs. cohort), or the age difference of study samples, or confounding factors that were not considered. In our study, we found that the association between ${\mathrm{PM}}_{2.5}$ and BMD were consistently observed at different anatomical sites in the human (such as heel, femur neck, and lumbar spine) even when we adjusted multiple BMD-related factors, including sex, age, BMI, ancestry, education, smoking, alcohol, physical activity, circulating calcium, and calcium supplementary status.

One limitation of our epidemiological study is that only a single measurement of air pollution was available. These annual average ${\mathrm{PM}}_{2.5}$ concentrations (available for the year 2010) were linked to each participant in the UK through participants’ residential addresses given at the baseline visit (2006–2010). Considering that air pollution emissions remained relatively stable from 2010 to 2019 in the UK,^48,49 we used the annual mean ${\mathrm{PM}}_{2.5}$ concentrations at baseline (2010) as exposure, assuming that the pattern of ${\mathrm{PM}}_{2.5}$ concentrations would not have changed greatly during the period in this study. Previous studies, which used the same ambient air pollution data set from UK Biobank, have explored the influence of long-term exposure to ambient air pollution on mortality⁵⁰ and mental health.⁴⁹

To assess the potential mechanism for the association observed in humans, we further conducted an animal study on bone health in a setting with ambient levels of ${\mathrm{PM}}_{2.5}$ exposure. Mice exposed to ${\mathrm{PM}}_{2.5}$ for 6 months exhibited shorter average femoral length. Mice incorporated into the study at 5 wk of age were exposed to ${\mathrm{PM}}_{2.5}$ for 6 months, roughly spanning the life stages from childhood to adolescence and middle age. The life stages of childhood and adolescence are particularly essential for bone modeling, including the formation of bone shape and growth of bone size.⁵¹ Several population studies have demonstrated that air pollution in childhood may be associated with transiently shorter height, as well as slower growth rate, but the underlying mechanism is unknown.^52–54

Bone remodeling predominating in adulthood is pivotal to maintain bone homeostasis, which is achieved by the balance of bone formation and bone resorption.⁵¹ Bone remodeling is also crucial to the process of bone growth and shape alteration. As to bone formation, the most classic and sensitive markers are OCN and BALP.⁵⁵ In our study, lower levels of OCN in serum and the femur in ${\mathrm{PM}}_{2.5}\text{-exposed}$ animals suggest an inhibition of activation of osteogenesis by ${\mathrm{PM}}_{2.5}$ exposure. However, ALP, another bone formation marker, showed no difference in the femur but did show a higher level in serum in ${\mathrm{PM}}_{2.5}\text{-exposed}$ mice. Bone-type ALP is a kind of tissue-nonspecific alkaline phosphatase and is expressed in both bone and liver.⁵⁶ Hepatic levels of ALP were examined to address the source of BALP in serum. Although significantly higher ALP expression was detected in the liver, suggesting that the liver contributes to the accumulation of circulating BALP, the function of liver-derived ALP in the mineralization process and whether it is adaptive to long-term ${\mathrm{PM}}_{2.5}$ exposure require further study. The lower OCN expression may indicate that osteogenesis could be inhibited by ${\mathrm{PM}}_{2.5}$ exposure. Hence, even though no difference in bone structure or BMD was observed in the animal studies, the combination of our population analysis results and some experimental evidence consistently supported the effects of ${\mathrm{PM}}_{2.5}$ exposure on bone health.

In terms of bone resorption, RANKL is essential for osteoclast formation in osteocytes,⁵⁷ TRAP staining around the growth plate and circulating TRACP-5b level are the most important biomarkers reflecting the amount and activity of osteoclasts on bone surface.⁵⁸ In addition, Ctsk is highly expressed in activated osteoclasts as a biomarker of osteoclastogenesis.⁵⁹ Convincingly and first, the elevated levels of all these bone resorption biomarkers were demonstrated in the present study. Moreover, the calcified cartilaginous template was destroyed with signs of osteoclast activation around the growth plate in mice exposed to ${\mathrm{PM}}_{2.5}$ as well, which may lead to damage to the growth plate structure and consequential aberrant long bone growth during the growing stage.⁶⁰ In addition, results from cell TRAP staining and bone resorption assay also confirmed that cells exposed to ${\mathrm{PM}}_{2.5}$ had abnormal osteoclast formation and function in vitro. It is important to note that the excessive osteoclast activity shown in vivo and in vitro did not associate with differences in trabecular bone mass. Assessment of an earlier time point in the study should be evaluated in future work given that by the end of this study, the trabecular bone was largely gone. Unlike the human population results, our animal results did not show a significant difference in bone microarchitecture or BMD in the ${\mathrm{PM}}_{2.5}\text{-exposed}$ individuals; this could be explained by the incomplete consistence in exposure concentration or exposure time between UK Biobank data and animal data. Taken together, the results of this study suggest that osteoclast activity could be prominently enhanced by ${\mathrm{PM}}_{2.5}$ exposure.

According to our experimental results and previous studies, an imbalance between bone resorption and bone formation could be induced by ${\mathrm{PM}}_{2.5}$ exposure through various ways, including inflammation,⁶¹ liver toxicity,⁶² and immunoreaction.^63,64 However, the potential mechanism by which ${\mathrm{PM}}_{2.5}$ exposure disturbed bone metabolism remains unknown. In light of the close association of the immune system in ${\mathrm{PM}}_{2.5}$ effects,⁶⁴ we hypothesized that inflammation may be involved in the adverse effects of ${\mathrm{PM}}_{2.5}$ exposure on bone health. Recent studies have indicated a role of inflammation in regulating proliferation and differentiation of osteoclasts and osteoblasts^46,65,66 Therefore, we examined the expression of proinflammatory cytokines specifically around the growth plate area. $\text{TNF-}\mathrm{\alpha }$ has been shown to be involved in bone homeostasis by directly potentiating the expression of RANKL to enhance the bone resorption process and impairing the osteoblast performance.^67,68 Thus, the higher expression of $\text{TNF-}\mathrm{\alpha }$ in ${\mathrm{PM}}_{2.5}\text{-exposed}$ cells/animals may disturb the balance by activating osteoclastogenesis as well as by inhibiting osteogenesis.

The expression of Traf6 is closely related to TNF-α.⁶⁹ In addition, Traf6 is one of the key molecules for osteoclastogenesis, as demonstrated by a study in which Traf6-knockout mice developed severe osteopetrosis.⁷⁰ Briefly, during the osteoclast differentiation upon RANKL induction, the binding of Traf6 to RANK results in the activation of downstream regulators, such as Nfatc1 and c-Fos, and then induces expression of osteoclast-specific genes.^71–73 Here, our data show that the wells treated with ${\mathrm{PM}}_{2.5}$ had a higher number of osteoclasts, higher expression of Traf6 and c-Fos, as well as more bone resorption. Most importantly, neutralizing $\text{TNF-}\mathrm{\alpha }$ could partially or completely block all these altered markers or molecules for osteoclast formation observed with ${\mathrm{PM}}_{2.5}$ stimulation. Hence, our in vitro and in vivo results suggest that ${\mathrm{PM}}_{2.5}$ stimulation can facilitate osteoclastogenesis through the $\text{TNF-}\mathrm{\alpha }$/Traf6/c-Fos pathway.

In conclusion, we found that long-term exposure to ${\mathrm{PM}}_{2.5}$ exposure was associated with an increased risk of osteoporosis with analysis of UK Biobank data, and we further demonstrated ${\mathrm{PM}}_{2.5}$ exposure may promote osteoclastogenesis by the $\text{TNF-}\mathrm{\alpha }$-mediated Traf6/c-Fos pathway in vivo (Figure 5H). These findings provide a novel insight into bone homeostasis in response to long-tern ${\mathrm{PM}}_{2.5}$ exposure.