Recombinant murine soluble RANKL (315-11) was purchased from Peprotech (Rocky Hill, NJ, United States). CQ phosphate (S4157) was obtained from Selleck (Shanghai, China). ML171 (492002) was acquired from Millipore Corporation (Temecula, CA, United States). Diphenyleneiodonium chloride (DPI) (sc-202584) and 5-O-methyl quercetin (sc-483298) were purchased from Santa Cruz Biotechnology (Dallas, TX, United States). N-acetyl-L-cysteine (NAC) (A9165) and Mito-TEMPO (SML0737) were purchased from Sigma-Aldrich (St. Louis, MO, United States). GSK2795039 (HY-18950) and GSK2606414 (HY-18072) were purchased from MCE (Shanghai, China).

Both the mouse Nox4 short hairpin RNA (shRNA)-containing retrovirus and the corresponding empty vector were designed and synthesized by HanBio (Shanghai, China). The mRFP-GFP-LC3-containing adenovirus (HB-AP210 0001) was provided by HanBio. Primary antibodies against Nox3 (20065-1-AP), Nox4 (14347-1-AP), ATF6 (24169-1-AP), ATF4 (10835-1-AP), XBP1S (24868-1-AP), GAPDH (60004-1-Ig), ERp57/ERp60 (15967-1-AP), and VDAC1/Porin (55259-1-AP) were obtained from Proteintech (Wuhan, Hubei, China). Primary antibodies against LC3A/B (4108S), PERK (C33E10) (3192S), phospho-PERK (Thr980; 16F8) (3179S), eIF2α (D7D3) XP^® (5324T), and phospho-eIF2α (Ser51; D9G8) XP^® (3398T) were purchased from Cell Signaling Technology (Danvers, Massachusetts, United States). Primary antibodies against Nox1 (ab131088) and Nox2/gp91phox (ab129068) were obtained from Abcam (Cambridge, United Kingdom).

The RAW264.7 mouse monocyte/macrophage cell line was purchased from the Cell Culture Center of the Chinese Academy of Sciences (Shanghai, China) and cultured in Dulbecco’s modified Eagle’s medium (11995065, Gibco, Grand Island, NY, United States) supplemented with 10% fetal bovine serum (10091148, Gibco) and 100 U/ml penicillin and 100 mg/ml streptomycin (15070063, Gibco). The cells were incubated in a humidified atmosphere with 95% air and 5% CO₂ at 37°C. To induce osteoclast differentiation, RAW264.7 cells were stimulated with 100 ng/ml RANKL and further cultured for the indicated times.

RAW264.7 cells were plated and cultured in 35 mm dishes. When the confluence reached 50%, the cells were transfected with retrovirus encoding Nox4 shRNAs or scrambled shRNA at a multiplicity of infection (MOI) of 100 for 24 h according to the manufacturer’s instructions (HanBio). The nucleotide sequences were as follows: sh-Nox4-1, 5′-GCA​GGA​GAA​CCA​GGA​GAT​TGT-3′; sh-Nox4-2, 5′-GCA​TGG​TGG​TGG​TGC​TAT​TCC-3′; sh-Nox4-3, 5′-GGT​ATA​CTC​ATA​ACC​TCT​TCT-3′; and sh-NC, 5′-TTC​TCC​GAA​CGT​GTC​ACG​T-3′. RAW264.7 cells with stable knockdown of Nox4 expression were screened by the addition of 2 µg/ml puromycin (HB-PU-1000, HanBio) to the culture medium for 48 h. Then, the stable cells were digested with 0.25% trypsin (T1350, Solarbio, Beijing, China) and seeded on 35 mm dishes at a density of 8 × 10⁴ cells/dish and incubated overnight for attachment. The next day, adherent cells were treated with or without RANKL (100 ng/ml) for 3 days. The knockdown efficiency of the three Nox4 shRNAs was measured using western blotting, and the most effective was selected for use in subsequent experiments.

Osteoclast formation was measured by quantifying cells positively stained with TRAP. Briefly, RAW264.7 cells were incubated at a density of 1 × 10⁴ cells/well in 24-well plates (Corning, New York, NY, United States) overnight. After stimulation with RANKL (100 ng/ml) and various concentrations of different pharmacological reagents for 6 days, the cells were fixed with 4% paraformaldehyde (AR1069, BOSTER, Wuhan, Hubei, China) for 30 min at room temperature and then stained by using a Tartrate Resistant Acid Phosphatase Assay Kit (P0332, Beyotime Biotechnology, Shanghai, China) according to the manufacturer’s instructions. TRAP-positive and multinucleated cells containing three or more nuclei were considered osteoclasts. For each well, the osteoclasts were observed and counted under a light microscope (Leica, Wetzlar, Germany).

To confirm the bone resorption ability of differentiated osteoclasts, RAW264.7 cells were seeded at a density of 2 × 10⁴ cells/well overnight in 24-well Osteo Assay Surface plates (Corning) coated with hydroxyapatite matrix. Then, the cells were incubated with RANKL (100 ng/ml) or in the presence of various concentrations of different pharmacological reagents. The medium was replaced every 3 days. After 7 days of culture, the cells were removed using a 10% sodium hypochlorite solution, and the wells were stained with 1% toluidine blue. The plate was washed twice with distilled water and air dried at room temperature. The bone resorption pits in each well were observed and photographed by a light microscope. The pit formation area was analyzed using the ImageJ software (National Institutes of Health, Bethesda, MD, United States).

Total cellular RNA was extracted using RNAiso plus reagent (9,108, Takara, Kyoto, Japan) according to the manufacturer’s instructions. Subsequently, the total RNA concentration was determined with a NanoDrop 2.0 spectrophotometer (Thermo Fisher Scientific, Pittsburgh, PA, United States) and the RNA was reverse transcribed to cDNA using a PrimeScript™ RT reagent kit with gDNA Eraser (RR047A, Takara) according to the manufacturer’s instructions. Subsequently, qRT-PCR assays were performed by using a SYBR Premix Ex Taq™ II (2×) kit (RR820A, Takara) according to the manufacturer’s instructions and run on an ABI 7500 Real-Time PCR Detection System (Foster City, CA, United States). The reactions were performed using the following parameters: 95°C for 30 s followed by 40 cycles of 95°C for 5 s and 60°C for 30 s. The primer nucleotide sequences used for qRT-PCR are listed in Supplementary Table S1. All primer sets for mRNA amplification were purchased from Sangon Biotech (Shanghai) Co., Ltd. (Shanghai, China). The relative expression levels of the target gene were normalized with respect to the levels of β-actin expression and calculated using the 2^−△△CT method.

RAW264.7 cells were cultured with the indicated treatments for 3 days. Then, the cells were digested with 0.25% trypsin, centrifuged (2000 rpm) for 10 min and fixed with 2.5% glutaraldehyde overnight at 4°C. Subsequently, the cells were postfixed with 1% osmium tetroxide for 1.5 h, washed and stained in 3% aqueous uranyl acetate for 1 h. Thereafter, the samples were washed again, dehydrated with a graded series of increasing ethanol concentrations to 100% and embedded in Epon-Araldite resin. Subsequently, the ultrathin sections were cut using a Reichert ultramicrotome (Reichert, New York, NY, United States) and counterstained with 0.3% lead citrate. Then, the ultrastructure of autophagic vacuoles (autophagosomes and autolysosomes) was observed under a transmission electron microscope (Leica), and images were captured.

After growth to 50% confluence in 35 mm dishes, the cells were transfected with adenovirus expressing mRFP-GFP-LC3 for 24 h using a MOI of 1,000, according to the manufacturer’s instructions (HanBio). Then, the cell growth medium was replaced with fresh complete medium for another 24 h. Afterward, the transfected cells were digested with 0.25% trypsin and seeded on confocal Petri dishes (NEST, Wuxi, Jiangsu, China) at a density of 5 × 10⁴ cells/dish and incubated overnight for attachment. Thereafter, adherent cells were treated with the various indicated treatments for 3 days. The treated cells were washed with phosphate buffer saline (PBS) (AR0032, BOSTER) and viewed with a laser scanning confocal microscope (Leica). GFP loses its fluorescence in acidic lysosomal conditions, whereas mRFP does not. Therefore, yellow (merged GFP signal and RFP signal) puncta represent early autophagosomes, whereas puncta detectable only as red (RFP signal alone) indicate late autolysosomes that are formed by autophagosome fusion with lysosomes. Autophagic flux was ultimately assessed by quantifying the mRFP and GFP puncta per cell. The number of GFP and mRFP puncta was determined by manually counting 30 cells randomly in 5 fields per dish, and the average number of puncta per cell was calculated.

RAW264.7 cells (5 × 10⁴) were plated on confocal Petri dishes and allowed to attach overnight. Then, the cells were cultured with the indicated treatments for the indicated times. Next, ER-Tracker (E34250, Thermo Fisher Scientific) was added directly to the culture medium at 500 nM and incubated with cells for 30 min in a 37°C humidified incubator containing 5% CO₂. Then, the cells were washed with PBS and immediately observed under a laser scanning confocal microscope.

The colocalization of Nox4 with the ER was detected by double-labeling immunofluorescence. Briefly, RAW264.7 cells were seeded on confocal Petri dishes at a density of 5 × 10⁴ cells/dish overnight. Then, the cells were incubated with or without RANKL (100 ng/ml) for 3 days. Thereafter, the cells were stained with ER-Tracker (the detailed experimental procedure is described above) and fixed in 4% paraformaldehyde for 20 min at 37°C. Then, the cells were permeabilized with 0.1% Triton X-100 (P1080, Solarbio) for 10 min and blocked with 10% normal goat serum (AR0009, BOSTER) for 1 h at 37°C. Subsequently, the cells were incubated with a rabbit polyclonal anti-Nox4 antibody (1:25) in a humidified chamber at 4°C overnight. Then, the cells were washed and incubated with Alexa Fluor 488-labeled goat anti-rabbit IgG (1:500) (A0423, Beyotime Biotechnology) for 1.5 h at 37°C in the dark. DAPI Staining Solution (C1005, Beyotime Biotechnology) was used to counterstain the cell nuclei. Finally, the cells were observed using a laser scanning confocal microscope.

RAW264.7 cells were seeded in 6 cm dishes overnight. Then, the cells were cultured with the indicated treatments for the indicated times. Subsequently, the cells were digested with 0.25% trypsin and centrifuged (1,000 rpm) for 5 min. Mitochondria and the ER were extracted from cells using a Cell Mitochondria Isolation kit (C3601, Beyotime Biotechnology) and Endoplasmic Reticulum Isolation kit (BB-31454-1, BestBio Science, Shanghai, China) according to the manufacturer’s instructions. The fractions of mitochondria and ER were lysed in RIPA buffer (P0013, Beyotime Biotechnology) that contained a protease and phosphatase inhibitor cocktail (P1050, Beyotime Biotechnology) for subsequent western blot analysis.

RAW264.7 cells were lysed in RIPA buffer that contained a protease and phosphatase inhibitor cocktail. After centrifugation at 14,000 g for 5 min at 4°C, the concentrations of protein were measured using a BCA protein assay kit (P0010, Beyotime Biotechnology). Subsequently, equal amounts of protein (40 μg) were separated by 6, 8 or 12%/5% (w/v) sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Bio-Rad, Hercules, CA, United States) and transferred to polyvinylidene difluoride membranes (Millipore). The membranes were incubated with anti-GAPDH, anti-Nox4, anti-LC3A/B, anti-ERp57/ERp60, anti-VDAC1/Porin, anti-Nox3, anti-Nox2/gp91phox, anti-Nox1, anti-PERK (C33E10), anti-phospho-PERK (Thr980) (16F8), anti-ATF6, anti-ATF4, anti-XBP1S, anti-eIF2α (D7D3) XP^®, and anti-phospho-eIF2α (Ser51; D9G8) XP^® antibodies separately overnight at 4°C. Then, the membranes were incubated with horseradish peroxidase-conjugated goat anti-mouse/rabbit IgG (H + L) secondary antibodies (1:5,000) (ZB-2305/ZB-2301, ZSGB-BIO, Beijing, China) for 1 h at room temperature. Finally, the protein bands of interest on the membranes were visualized with a chemiluminescence substrate kit (WBKLS0100, Millipore Corporation) using an Image Quant LAS4000 instrument (GE, Boston, MA, United States). The band intensity was quantified by densitometric analysis using Image J software.

The intracellular production of ROS was detected by staining cells with a ROS Assay kit (S0033S, Beyotime Biotechnology). Briefly, RAW264.7 cells (5 × 10⁴) were seeded in confocal Petri dishes overnight. Then, the adherent cells were cultured under conditions with various treatments for the indicated times. Subsequently, the cells were stained with ER-Tracker (the detailed experimental procedure is described above) and washed with PBS. Then, 2′, 7′-dichlorodihydrofluorescein diacetate (DCFH-DA), which was added directly to serum-free medium, was diluted to a final concentration of 10 μM and incubated with cells for 30 min at 37°C in a humidified incubator containing 5% CO₂. DCFH-DA diffuses into cells and is deacetylated by cellular esterases to nonfluorescent 2′,7′-dichlorodihydrofluorescein, which can be oxidized by ROS to produce highly fluorescent 2′,7′-dichlorofluorescein (DCF). The green fluorescence intensity is proportional to the levels of ROS within a cell. The cells were then washed three times with PBS, and the fluorescence intensity was observed using a laser scanning confocal microscope.

The levels of mitochondrial ROS were measured by staining cells with MitoSOX™ Red Mitochondrial Superoxide Indicator (M36008, Thermo Fisher Scientific). Briefly, RAW264.7 cells were seeded at a density of 3 × 10³ cells/well on 96-well plates and allowed to attach overnight. Then, the cells were cultured with various treatments for the indicated times. Subsequently, the cells were incubated with MitoSOX at a final concentration of 5 μM for 15 min at 37°C in a humidified incubator containing 5% CO₂. Then, the cells were washed three times with PBS, and the fluorescence intensity was immediately measured with a Varioskan Flash Spectral Scanning Multimode Reader (Thermo Fisher Scientific). The red fluorescence intensity is proportional to the levels of mitochondrial ROS within the cell.

The SPSS software, version 13.0 (SPSS, Inc., Chicago, IL, United States) and GraphPad Prism 7.0 (GraphPad Software) were used for statistical analysis. Data are presented as the mean ± standard deviation (SD). The unpaired Student’s t-test was utilized to calculate the p value of the statistic difference between 2 independent datasets. One-way analysis of variance (ANOVA) was used to analyze the significance among three or more independent datasets, and the Tukey’s post-hoc multiple comparison testing was performed for multiple comparisons when the probability for ANOVA was statistically significant. Methods of non-parametric statistics such as the Mann–Whitney and Kruskal–Wallis tests were used when variances did not pass the Levene test for normality or homogeneity. All the statistical tests were two-sided and a p value of <0.05 was considered statistically significant in all cases.

Consistent with previous studies, we found that RANKL enhanced the proportion of fused multinuclear cells (Supplementary Figure S1A) and the expression levels of osteoclastogenesis-related genes (TRAP, Cath K and MMP-9; Supplementary Figure S1B), inducing the formation of TRAP-positive multinuclear (≥3) osteoclasts and bone resorption pits in RAW264.7 cells (Supplementary Figure S1C). These results indicate that RANKL induced the differentiation and subsequent bone resorption activity of osteoclasts in vitro, which is consistent with previous findings (Asai et al., 2014).

Autophagy has been demonstrated to play critical roles in enhanced osteoclastogenesis under many conditions, such as hypoxia, oxidative stress, and microgravity (Wang et al., 2011; Sambandam et al., 2014; Sun et al., 2015). First, we assayed the level of autophagy after RANKL treatment. Western blot analysis showed that the LC3-II/LC3-I ratio was significantly upregulated from day 3 in a time-dependent manner during RANKL-induced osteoclastogenesis (Figure 1A). The TEM showed that the number of autophagic vacuoles was dramatically increased after 3 days of treatment with RANKL (Figure 1C). Then, the autophagic flux activity was further determined by using the adenovirus-mRFP-GFP-tagged LC3 system. The data showed that the number of yellow and red puncta in merged images was significantly increased after 3 days of RANKL-induced differentiation, indicating the activation of both autophagosome formation and lysosomal degradation in the RANKL-treated group (Figure 1D). Collectively, these observations indicate that autophagy is activated during RANKL-induced osteoclastogenesis. Second, using a pharmacological inhibitor of autophagy (CQ), we explored whether autophagy is essential in RANKL-induced osteoclastogenesis. The results showed that CQ treatment markedly increased the LC3-II/LC3-I ratio (Figure 1B) and inhibited autolysosomal degradation (Figures 1C,D). More importantly, we found that CQ treatment suppressed the RANKL-induced upregulation of osteoclastogenesis-related genes (TRAP, Cath K and MMP-9; Figure 1E), reduced the number of TRAP-positive multinuclear (≥3) osteoclasts and reduced the area of the bone resorption pits (Figure 1F). Taken together, the above data reveal that RANKL induces osteoclastogenesis and bone resorption through autophagy.

The above experiments showed that RANKL induced osteoclastogenesis and bone resorption through autophagy. It is well known that NADPH oxidase (Nox) family proteins promote the activation of autophagy in many cell types by generating ROS (Sciarretta et al., 2013). Recent evidence indicated that RANKL increases the generation of intracellular ROS by promoting the expression and activity of intracellular Nox family proteins during osteoclastogenesis (Lee et al., 2005; Goettsch et al., 2013; Kang and Kim, 2016). Therefore, we explored whether Nox family proteins are involved in RANKL-induced autophagy.

Western blot analysis showed that RANKL time-dependently upregulated the levels of Nox1 and Nox4 proteins and decreased the levels of Nox2 protein but had no significant influence on the levels of Nox3 protein (Figure 2A). Nox pharmacological inhibitor DPI treatment obviously downregulated the RANKL-induced increase in the LC3-II/LC3-I ratio (Figure 2B). To further investigate which Nox isoforms are involved in RANKL-induced autophagy and osteoclastogenesis, pharmacological inhibitors targeting specific Nox isoforms were used. The data showed that the inhibition of Nox1 or Nox4 (not Nox2) separately by ML171 and 5-O-methyl quercetin significantly inhibited the RANKL-induced increase in the LC3-II/LC3-I ratio and osteoclastogenesis-related gene (TRAP, Cath K and MMP-9) expression and reduced the number of TRAP-positive multinuclear (≥3) osteoclasts and the area of the bone resorption pits (Figures 2C–F). Importantly, the inhibitory effect of 5-O-methyl quercetin on RANKL-induced autophagy, osteoclastogenesis and bone resorption was significantly greater than that of ML171. Therefore, compared with Nox1, Nox4 may play a leading role in autophagy activation induced by RANKL. We selected Nox4 for the subsequent experiments.

To further determine the functional significance of Nox4 in RANKL-induced autophagy and osteoclastogenesis, retroviruses encoding three different Nox4 shRNAs or scrambled shRNA were utilized. The results showed that the expression levels of Nox4 were dramatically decreased after the transfection of sh-Nox4-1, sh-Nox4-2, and sh-Nox4-3 (Figure 3A). Importantly, sh-Nox4-2 had the greatest Nox4 silencing effect. Therefore, sh-Nox4-2 was selected for the subsequent experiments. The knockdown of Nox4 markedly inhibited the RANKL-induced increase in the LC3-II/LC3-I ratio, autophagic flux activity, and expression of osteoclastogenesis-related genes (TRAP, Cath K and MMP-9) and decreased the number of TRAP-positive multinuclear (≥3) osteoclasts and bone resorption pit area (Figures 3B–E). Collectively, the above results indicate that knockdown of Nox4 suppresses RANKL-induced autophagy and osteoclastogenesis.

Recent studies have indicated that Nox4 is localized on intracellular membranes, mainly in mitochondria and the ER (Lassegue et al., 2012). Western blot analysis showed that RANKL treatment markedly increased the protein level of Nox4 in the ER (not in the mitochondria) of RAW264.7 cells (Figures 4A,B). As shown in Figure 4C, Nox4 shRNA treatment significantly downregulated the RANKL-induced elevation of Nox4 protein in the ER of RAW264.7 cells. To further ascertain the ER localization of Nox4 protein induced by RANKL, an immunofluorescence staining assay was utilized. The results showed that RANKL treatment significantly enhanced the localization of Nox4 in the ER, which was markedly suppressed by Nox4 silencing (Figure 4D). Collectively, the above data reveal that RANKL markedly upregulates the protein level of Nox4 in the ER. These results are consistent with the fact that Nox4, mainly synthesized in the ER, contains multiple ER-specific signal sequences in the N-terminal portion of Nox4. (Chen et al., 2008; Stephens and Nicchitta, 2008).

It has been reported that Nox family proteins can promote the production of intracellular ROS (Finkel, 2011). As shown in Figure 5A, RANKL treatment markedly enhanced the level of intracellular ROS and ER ROS in RAW264.7 cells, which was reduced by Nox4 silencing. As shown in Figure 5C, the level of mitochondrial ROS was increased in RAW264.7 cells during RANKL-induced osteoclastogenesis. As ROS have been reported to play an important role in autophagy regulation (Kongara and Karantza, 2012), we explored whether ROS are involved in RANKL-induced autophagy and osteoclastogenesis. Intracellular ROS scavenger (NAC) treatment significantly inhibited the RANKL-induced accumulation of intracellular ROS and ER ROS (Figure 5B). Mitochondrial-targeted antioxidant (Mito-TEMPO) treatment significantly inhibited RANKL-induced mitochondrial ROS accumulation (Figure 5C). Importantly, Mito-TEMPO treatment did not affect the RANKL-induced increase in the LC3-II/LC3-I ratio (Figure 5D), whereas NAC treatment obviously reduced the RANKL-induced increase in the LC3-II/LC3-I ratio (Figure 5D) and the number of yellow and red puncta in merged images, which implies an impairment in autophagic flux activity (Figure 5E). Additionally, NAC treatment also significantly downregulated the RANKL-induced upregulation of the expression of osteoclastogenesis-related genes (TRAP, Cath K and MMP-9; Figure 5F) and reduced the number of TRAP- positive multinuclear (≥3) osteoclasts and bone resorption pit area (Figure 5G). In summary, these data indicate that Nox4 promotes RANKL-induced autophagy activation and osteoclastogenesis by generating nonmitochondrial ROS. Furthermore, we also found that the majority of Nox4-derived ROS colocalize with ER-Tracker (Figure 5A). These results suggest that Nox4 may promote the activation of autophagy via the generation of ER-derived ROS during RANKL-induced osteoclastogenesis.

Previous studies have demonstrated that UPR-related signaling pathways (ATF6, PERK/eIF-2α/ATF4, and IRE-1α/XBP-1) are involved in ROS-induced autophagy (Avivar-Valderas et al., 2011; Walter and Ron, 2011). Therefore, we further explored whether UPR-related signaling pathways mediate ROS-induced autophagy during RANKL-induced osteoclastogenesis. Western blot analysis showed that RANKL treatment time-dependently increased the phosphorylation of PERK (Thr980) and eIF-2α (Ser51) and upregulated the expression levels of ATF4 and XBP-1S in RAW264.7 cells but had little influence on the level of ATF6 (Figure 6A). Moreover, knockdown of Nox4 significantly repressed the RANKL-induced phosphorylation of PERK (Thr980) and eIF-2α (Ser51) and the upregulation of ATF4 expression but had little influence on XBP-1S (Figure 6B). Furthermore, NAC treatment significantly inhibited the RANKL-induced activation of the PERK/eIF-2α/ATF4 pathway (Figure 6C). These results reveal that RANKL activates the PERK/eIF-2α/ATF4 pathway by elevating Nox4-mediated ROS production in RAW264.7 cells. To further confirm the involvement of the PERK/eIF-2α/ATF4 pathway in the activation of autophagy, osteoclastogenesis and bone resorption, a pharmacological inhibitor of PERK (GSK2606414) was utilized. The data revealed that GSK2606414 treatment dose-dependently reduced the RANKL-induced phosphorylation of eIF-2α (Ser51) and the upregulation of ATF4 expression in RAW264.7 cells (Figure 6D). More importantly, GSK2606414 treatment obviously inhibited the RANKL-induced increase in the LC3-II/LC3-I ratio (Figure 6E) and the number of yellow and red puncta in merged images, which implies that autophagic flux is impaired (Figure 6F). Additionally, GSK2606414 treatment also significantly abrogated the RANKL-induced upregulation of osteoclastogenesis-related genes (TRAP, Cath K and MMP-9; Figure 6G) and reduced the number of TRAP-positive multinuclear (≥3) osteoclasts and bone resorption pit area (Figure 6H). These results indicate that Nox4-derived ROS promote autophagy via the PERK/eIF-2α/ATF4 pathway during RANKL-induced osteoclastogenesis.