Poly-ADP-ribosylation-mediated degradation of ARTD1 by the NLRP3 inflammasome is a prerequisite for osteoclast maturation

Evidence implicates ARTD1 in cell differentiation, but its role in skeletal metabolism remains unknown. Osteoclasts (OC), the bone-resorbing cells, differentiate from macrophages under the influence of macrophage colony-stimulating factor (M-CSF) and receptor-activator of NF-κB ligand (RANKL). We found that M-CSF induced ADP-ribosyltransferase diphtheria toxin-like 1 (ARTD1) auto-ADP-ribosylation in macrophages, a modification that marked ARTD1 for cleavage, and subsequently, for degradation upon RANKL exposure. We established that ARTD1 proteolysis was NLRP3 inflammasome-dependent, and occurred via the proteasome pathway. Since ARTD1 is cleaved at aspartate214, we studied the impact of ARTD1 rendered uncleavable by D214N substitution (ARTD1D214N) on skeletal homeostasis. ARTD1D214N, unlike wild-type ARTD1, was resistant to cleavage and degradation during osteoclastogenesis. As a result, ARTD1D214N altered histone modification and promoted the abundance of the repressors of osteoclastogenesis by interfering with the expression of B lymphocyte-induced maturation protein 1 (Blimp1), the master regulator of anti-osteoclastogenic transcription factors. Importantly, ARTD1D214N-expressing mice exhibited higher bone mass compared with controls, owing to decreased osteoclastogenesis while bone formation was unaffected. Thus, unless it is degraded, ARTD1 represses OC development through transcriptional regulation.

facilitating access of transcription factors to DNA sites. 21 In addition, ARTD1 can influence chromatin modification by PARylating histones at residues that are also regulated by methylation or acetylation. 21 Thus, through direct actions on transcription factors and indirect influence through epigenetic mechanisms, ARTD1 may represent a homeostatic mechanism that alters OC differentiation program. ARTD1 is regulated by various post-translational modifications, including caspase-mediated proteolytic cleavage at aspartate 214, 25 auto-PARylation, 26 SUMOylation 27 and ubiquitination, 28 which was linked to ARTD1 degradation in cancer cells. Evidence indicates that activation of the NOD-like receptor (NLR) family, pyrin domain-containing 3 (NLRP3) inflammasome, a protein complex comprising the adapter protein ASC and caspase-1, triggers cascade that leads to ARTD1 cleavage. 29,30 Despite ARTD1's actions in many tissues, its role in skeletal homeostasis remains unknown as only a few studies have explored its function in osteoclastogenesis in vitro. [31][32][33] We found that ARTD1 was degraded during OC differentiation. Conversely, expression of ARTD1 D214N , which was resistant to degradation, caused a high bone mass phenotype owing to increased expression of OC repressors, decreased OC differentiation and bone resorption, while bone formation was unaltered.

Results
ARTD1 is degraded during OC formation through PARylation-dependent mechanisms. M-CSF provides growth and survival signals for cells of the OC lineage through regulation of numerous pathways, including PI3K/Akt and MAPK. 1 Here, we found that M-CSF contained in CMG media 34 induced massive protein PARylation in mouse bone marrow macrophages (BMM), an effect that was time- (Figure 1a, bracket) and concentration-(Supplementary Figure S1A) dependent, and was inhibited by two chemically different inhibitors of ARTD1 and ARTD2, olaparib (olap) and veliparib (velip) (Figure 1b, bracket) or Artd1 happloinsufficiency (Supplementary Figure S1B). RANKL decreased protein PARylation in a time-dependent manner (Figure 1c, bracket). Whereas M-CSF treatment did not affect ARTD1 protein levels ( Figure 1d, and Supplementary Figures S1B, S1C), RANKL treatment reduced ARTD1 protein abundance (Figure 1e and f). In addition, p89 kDa fragment (p89), a product of caspase-mediated cleavage of ARTD1 30,35 was apparent at day 2 during the differentiation of RAW 264.7 cells (Figure 1e), but was not readily detectable in the differentiation of primary BMM (Figure 1f). Thus, ARTD1 is likely responsible for protein PARylation, which inversely correlates with OC differentiation.
Pull-down studies using Af1521 macrodomains (Figure 1g), which have high specificity and affinity for PARylated proteins, 36 and immunoprecipitation studies using anti-PAR antibody ( Figure 1h) showed a decline in the levels of PARylated ARTD1 during OC differentiation. These results suggest that PARylation regulates ARTD1 protein levels. Indeed, in the 3-day OC cultures treated with olap, not only ARTD1 PARylation was inhibited (Figure 1g, top panel), but ARTD1 degradation was also prevented (Figure 1g, middle panel). ARTD1 was degraded through the proteasome pathway since a brief exposure for 3 h of OC cultures to MG-132, a proteasome inhibitor, was able to attenuate ARTD1 loss ( Figure 1i). Unexpectedly, p89 remained undetectable even in the presence of MG-132, suggesting slow accumulation of this fragment in these experimental conditions. The protective effect of olap was not due to blockade of OC differentiation because if anything, the number of OC in inhibitor-treated cultures scored at day 4 was higher than that of vehicle-treated cultures counted at day 3 (data not shown). Collectively, these data suggest that ARTD1 auto-modification is required for its degradation during osteoclastogenesis.
ARTD1 cleavage at D214 is required for its degradation during OC formation. We hypothesized that caspasemediated cleavage of ARTD1 at aspartate 214 is required for osteoclastogenesis to proceed. Hence, we studied the impact of ARTD1 rendered uncleavable by D214N substitution (ARTD1 D214N ) on skeletal homeostasis. First, to ensure that ARTD1 D214N is indeed uncleavable, we exposed LPS-primed BMM to ATP, an activator of the NLRP3 inflammasome. 30 Inflammasome activation caused a timedependent decline in ARTD1 abundance in WT BMM, a response that inversely correlated with the levels of the cleaved p89 kDa ARTD1 fragment (p89) (Figure 2a). In contrast, activated inflammasome failed to induce the cleavage of ARTD1 D214N (Figure 2b). ARTD1 processing did not occur in Nlrp3-deficient mice (Figure 2c) as expected. Conversely, following LPS treatment, which up-regulated NLRP3 as reported, 29 ARTD1 protein levels were reduced in BMM expressing constitutively activated NLRP3 (NLRP3 ca ) inflammasome in the absence of exogenously added ATP, but not control cells (Figure 2d). Moreover, ARTD1 D214N blocked NLRP3 ca -induced OC formation (Supplementary Figure S2). Thus, ARTD1 D214N is resistant to cleavage in response to various stimuli, including proapoptotic cues as reported 25 and NLRP3 inflammasomeinduced signals.
We previously reported that the NLRP3 inflammasome is activated during RANKL-induced OC formation in the absence of exogenously added secondary inflammasome-activating signals. 37 Although ARTD1 is degraded during osteoclastogenesis, it is still unclear whether ARTD1 cleavage at D214 is a prerequisite for its degradation during this process. To understand the relationship between these two nonmutually exclusive mechanisms, we monitored the fate of WT and ARTD1 D214N during OC formation. WT and ARTD1 D214N mRNA levels were unaltered during OC formation (Figure 2e), and ARTD1 protein levels were diminished by day 3 of cultures in WT cells whereas those of ARTD1 D214N protein remained unchanged during OC formation ( Figure 2f). Occasionally, a fragment of~78 kDa was observed in cells expressing ARTD1 D214N , suggesting that mutant ARTD1 is proteolytically processed to some extent at a different site to enable minimal osteoclastogenesis. Collectively, our results also suggest that WT, but not ARTD1 D214N is efficiently degraded during this process. Consistent with a role for the NLRP3 inflammasome in ARTD1 processing, ARTD1 was degraded at a slower pace in NLRP3-deficient cells compared with WT counterparts (Figure 2g). ARTD1 levels consistently declined by day 3 of OC formation, yet the p89 fragment, which was detectable in RAW 264.7 cells (Figure 1e), remained elusive in primary BMM. Notably, we found that when cells were fed daily (starting at day 1.5) instead of every 2 days as in Figure 2f and g, p89 was readily detected 12 h after media replenishment in WT, but not Artd1 D214N/D214N cells ( Figure 2h). These data suggest that NLRP3 inflammasome-mediated ARTD1 cleavage at D214 is the first step in the processing of ARTD1, an event that generates protein fragments which are subjected to full proteolysis during osteoclastogenesis.  Figure S3A) whereas IRF8 expression was decreased ( Figure 3j). These results, which are consistent with our recent findings 38 suggest that ARTD1 activity is an important mechanism in ARTD1 negative regulation of osteoclastogenesis.  Cleavage of ARTD1 at aspartate 214 is required for its degradation during osteoclastogenesis. BMM isolated from Artd1 +/+ mice (a), Artd1 D214N/D214N mice (b), Nlrp3 −/− mice (c), WT mice or mice expressing constitutively activated NLRP3 inflammasome (Nlrp3 ca , d) were treated with vehicle or 100 ng/ml LPS for 3 h, and exposed to vehicle or 5 mM ATP for the indicated times (min, minutes). Western blot analysis was carried out using ARTD1 antibody (top panel) or p89 ARTD1 antibody (middle panel, a). The lanes from the same membranes were cut and pasted in d. (e) WT and Artd1 D214N/D214N BMM were incubated with 2% CMG (BMM) or 2% CMG and 100 ng/ml RANKL for 2 days (pOC, 2d) or 4 days (OC, 4d), and mRNA expression was analyzed by qPCR. (f) Analysis of ARTD1 degradation during OC formation from Artd1 +/+ or Artd1 D214N/D214N BMM. (g) Analysis of ARTD1 degradation during OC formation from Nlrp3 +/+ or Nlrp3 −/− BMM. (f) and (g) BMM were fed with 2% CMG or 2% CMG and 100 ng/ml RANKL every 2 days. (h) Analysis of ARTD1 cleavage and degradation during OC formation from BMM expressing ARTD1 +/+ or ARTD1 D214N . BMM were fed with 2% CMG and 100 ng/ml RANKL every day starting at day 1.5, and samples were analyzed 12 h later (day 2 or 3) or 24 h later (day 2.5 or 3.5). Samples were analyzed by Western blot; a nonspecific faint band around 89 kDa can be seen across samples (f-h). Data are representative of at least two independent experiments ARTD1 PARylation of histones can alter the modification of these proteins by methylation or acetylation. 19 Here, we found that the patterns of global histone3lysine4 trimethylation (H3K4me3), a mark of transcriptionally active chromatin (Figure 4c), but not histone3lysine27 trimethylation (H3K27me3), repressive mark (Figure 4d), was apparently affected by ARTD1 D214N expression at day 2 when we never found evidence of ARTD1 D214N processing (i.e., generation of 78 or 89 kDa fragment). We therefore focused on the former modification to gain insight onto ARTD1 transcriptional regulation during OC formation. PU.1, the transcription factor that affects the early steps of OC formation, 39 binds to Blimp1 promoter and positively regulates Blimp1 transcription. 40 Given the key role of Blimp1 in osteoclastogenesis, 10 we determined the effect of ARTD1 D214N on PU.1 regulation of Blimp1 by focusing on BMM and OC precursors (pOC), which express ARTD1 in contrast to OC. We found that PU.1 expression was increased slightly in WT pOC compared with mutant pOC (Figure 4b Uncleavable ARTD1 causes a high bone mass phenotype in mice. To determine the skeletal impact of constitutive expression of ARTD1 D214N , 25 we analyzed the femora of mice using micro-computed tomography (μCT). Bone mass was significantly higher in Artd1 D214N/D214N male mice at age 2 weeks (Figure 5a and b) and 8 weeks (Figure 5c Figure S5F) were different between the two tested genotypes. Thus, the high bone mass phenotype of Artd1 D214N/D214N mice stems from diminished OC development, but not bone formation.
Uncleavable ARTD1 causes defective osteoclast formation. The comparable gene expression levels of RANK, osteoprotegerin and RANKL (Supplementary Figure S4B) in bone samples from Artd1 +/+ and Artd1 D214N/D214N mice suggests that ARTD1 regulates OC differentiation downstream of RANK signals. Thus, to directly test our hypothesis that the high bone mass phenotype of Artd1 D214N/D214N mice was caused by defective OC development, we treated BMM cultures with M-CSF or M-CSF and RANKL. M-CSFinduced cell expansion was comparable between groups (Figure 6a, top panels). In contrast, OC differentiation was decreased 490% in cultures of ARTD1 D214N -expressing cells (Figure 6a, bottom panels, and Figure 6b). TRAP (Figure 6c) and cathepsin K (Figure 6d) mRNA expression was also comparable between genotypes at day 2, but was significantly reduced in Artd1 D214N/D214N relative to WT cells by day 4 of cultures. Artd1 D214N/D214N BMM ultimately formed OC when the cultures were maintained for 2 additional days (data not shown), suggesting that OC formation from Artd1 D214N/D214N BMM was attenuated but not blocked, consistent with the reduced number of OC in vivo in Artd1 D214N/D214N mice ( Figure  5e and f). Thus, ARTD1 D214N impairs the intrinsic ability of BMM to efficiently undergo osteoclastogenesis in vitro.

Discussion
The expression of ARTD1 mRNA is maintained during OC differentiation, yet WT ARTD1 protein is barely detectable in OC, suggesting either mRNA translation inhibition by noncoding RNA or ARTD1 protein degradation during OC formation. Although ARTD1 is a target of microRNA such as miR-223, 41 our results strongly support the latter scenario as ARTD1 D214N protein was consistently present in OC, and loss of WTARTD1 in OC was attenuated upon acute pharmacological blockade of the proteasome pathway. The notion that ARTD1 is degraded in OC implies that this protein functions as an intrinsic inhibitor of OC development, a view that is consistent with the enhanced anti-osteoclastogenic potential of degradation-resistant ARTD1 D214N , and our other observations indicating that OC formation and bone resorption were enhanced in Artd1-deficient mice. 38 Thus, while ARTD1 is dispensable in non-stress states in certain tissues, it plays a non-redundant cell-context-dependent role in skeletal homeostasis. Although ubiquitination-mediated degradation of ARTD1 in cancer cells was reported, 28 our findings unravel a novel concept in ARTD1 biology whereby degradation of this protein is a prerequisite for full execution of OC differentiation program. ARTD1 PARylates itself in pOC in response to M-CSF stimulation, and was subsequently degraded upon RANKL exposure. Although further work is required to explore the link between PARylation and ubiquitination of ARTD1 during OC development, the fact that ARTD1 auto-PARylation occurs early in BMM suggests that this modification may be a switch that targets this protein for proteolysis. Consistent with this concept, inhibition of ARTD1 activity, not only prevented ARTD1 auto-modification, but also stopped its destruction. In addition, a small fraction of WT ARTD1 that escapes degradation is apparently not PARylated (Figure 2f and g, On the other hand, ARTD1 auto-PARylation followed by cleavage and subsequent release from DNA occurs in response to genotoxic insults, 42 and is the presumed mechanism that prevents ARTD1 overreaction and excessive consumption of NAD + , an important source of cellular energy. Thus, it is reasonable to speculate that promotion of ARTD1 PARylation by osteoclastogenic factors triggers the destruction of this enzyme to preserve NAD + for energy metabolism during osteoclastogenesis. ARTD1 cleavage into p24 and p89 is a hallmark of apoptosis, though the underlying mechanisms have not been elucidated. Intriguingly, mice expressing ARTD1 D214N develop normally 25 as do mice lacking this protein, 43 suggesting that ARTD1 cleavage is not essential for cell death. Thus, ARTD1 actions, which include modulation of the function of various transcription factors such as NF-κB 24 and NFATc1 22,23 are more complex than originally thought. Here, we detected p89 during the early steps of osteoclastogenesis when ARTD1 was maximally PAR-conjugated, implying that PARylated ARTD1 may be a high affinity substrate for caspases, including caspase-7, which presumably cleaves ARTD1 in the nucleus in a non-apoptotic manner as proposed previously. 35 Caspase-7 can be activated by caspase-1, the catalytic component of the NLRP3 inflammasome, a pathway that not only regulates OC differentiation and bone resorption, 37,44-49 but is also critical in ARTD1 proteolytic processing during this process as demonstrated in this study. Consistent with an important role of the NLRP3 inflammasome-ARTD1 axis in the regulation of osteoclastogenesis, pharmacological inhibition or deletion of caspase-1, which attenuates ARTD1 degradation, inhibits OC formation. 37,38 Thus, despite the lack of the specific details on the enzyme that cleaves ARTD1 in the OC lineage, our results suggest a sequence of events whereby ARTD1 is highly PARylated in BMM in response to M-CSF, cleaved during RANKL-induced BMM lineage commitment, and finally degraded in late pOC. ARTD1 D214N regulates the expression of anti-OC transcription factors (IRF8, Id2, Lhx2 and MafB), but not of pro-OC transcription factors (NFATc1 and Mitf) in BMM. These results suggest that ARTD1 D214N mainly regulates the expression of the repressive molecules in BMM, and its inhibitory effects on the expression of late OC markers such as cathepsin K may be an indirect consequence of decreased OC formation. Mechanistically, ARTD1 may regulate transcription in BMM by binding to response elements or secondary hairpin structures of gene regulatory regions. Besides the fact that the role of ARTD1 gene regulation via binding to response elements is still unclear, it is not conceptually obvious to envision that ARTD1 directly regulates the expression of its numerous targets, which are either pro or anti-osteoclastogenic. A plausible alternative is that ARTD1 affects the function or accessibility to DNA response elements 35 of master transcription factors of OC development through PARylation of these proteins and/or histones. A detailed elucidation of such mechanisms is important, but is outside of the scope of this manuscript. Nonetheless, consistent with this scenario, we found that ARTD1 D214N decreased PU.1 binding to the promoter of the master repressor of antiosteoclastogenic factors, Blimp1, 10 a response that correlated with H3k4me3. Although PU.1 was not consistently induced in WT pOC (data not shown) it cannot be ruled out that lack of PU.1 induction in ARTD1 D214N -expressing cells contributed to the attenuated binding of PU.1 to Blimp1 promoter. Collectively, our findings indicate that ARTD1 functions to antagonize OC formation, an effect that is heightened in ARTD1 D214N , owing to its enhanced stability ( Figure 7). Advanced knowledge on ARTD1 biology revolves around its role in cell survival and death in non-skeletal tissues. We have discovered that ARTD1 impacts bone remodeling through its ability to regulate OC differentiation, hence positioning ARTD1 as an important candidate to regulate bone loss in diseases.

Materials and Methods
Mice. Germline knock-in mice globally expressing an ARTD1 mutant rendered uncleavable by D214N substitution (Artd1 D214N/D214N mice) have been previously described. 25 Nlrp3 −/− and Artd1 − /− mice were purchased from the Jackson Laboratory (Bass Harbor, ME, USA), and mice expressing constitutively activated NLRP3 (NLRP3 ca ) inflammasome have also been previously described. 44 Briefly, Nlrp3 fl(+/D301N) mice were crossed with LysM-Cre mice to obtain Nlrp3 ca mice. Nlrp3 fl(+/D301N) ; Artd1 D214N were mated with LysM-Cre;Artd1 D214N mice to generate Nlrp3 ca ; Artd1 D214N/D214N mice. All mice were on the C57BL6 background, and mouse genotyping was performed by PCR. All procedures were approved by the Institutional Animal Care and Use Committee of Washington University School of Medicine in St. Louis.
Bone mass and microstructure. Femoral bone structure was analyzed by μCT system (μCT 40; Scanco Medical AG, Zurich, Switzerland) as described previously. 37 Briefly, femora from 2-week-old and 8-week-old male mice were stabilized in 2% agarose gel, and μCT scans at 55 kVp were taken along the length of the femur as described previously. 37 The volume of interest analyzed was located just proximal to the growth plate, spanning a height of 350 μm each for the metaphyseal region.
Histology and histomorphometry. Mice were labeled twice by injection of calcein (15 mg/kg i.p.; Sigma-Aldrich, St. Louis, MO, USA) 5 and 2 days before euthanasia, which was performed under light anesthesia by exsanguination through dorsal aortic puncture. Blood was collected and the serum stored at − 80°C for later assays. Tissue samples were processed as described previously. 37 Briefly, long bones were fixed in 10% formalin, decalcified in 14% (w/v) EDTA pH 7.2 for 10-14 days at room temperature (RT), embedded in paraffin, sectioned at 5 μm thickness and mounted on glass slides. Stained sections with H&E or TRAP were used for the analysis of osteoblasts and OC, respectively, as described previously. 37 For dynamic histomorphometric analysis, bones were fixed in 70% ethanol for 24 h, left undecalcified and embedded in methyl methacrylate. Photographs were taken using nanozoomer (Hamamatsu, Hamamatsu City, Japan).
OC formation. Bone marrow macrophages (BMM) were obtained by culturing mouse bone marrow cells in culture media containing a 1 : 25 dilution of supernatant from the fibroblastic cell line, CMG 14-12, as a source of M-CSF, 34 a mitogenic factor for BMM, for~5 days in a 10-cm dish as described previously. 37 Nonadherent cells were removed by vigorous washes with PBS, and adherent BMM were detached with trypsin-EDTA, plated at 5-10 × 10 3 /well in a 96-well plate in culture media containing a 1 : 50 dilution of CMG and 100 ng/ml RANKL, a required cytokine for OC differentiation. For OC formation from RAW 264.7 cells, 3 × 10 3 cells/well in a 96-well plate were treated with 100 ng/ml RANKL. Media with supplements were changed every other day, and maintained at 37°C in a humidified atmosphere of 5% CO 2 in air.
TRAP staining. Cytochemical staining for TRAP was used to identify OC as described previously. 37 Briefly, cells in a 96 well plate were fixed with 3.7% formaldehyde and 0.1% Triton X-100 for 10 min at RT. The cells were rinsed with water and incubated with the TRAP staining solution (leukocyte acid phosphatase kit, Sigma-Aldrich) at RT for 30 min. Under light microscopy, multinuclear TRAP positive cells with at least 3 nuclei were scored as OC.
mRNA expression analysis. Total RNA was harvested from cells using RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). Complementary DNA was then synthesized with iScript reverse transcription kit (Bio-Rad, Hercules, CA, USA) and quantified using primers listed in Supplementary Table 1. Gene expression was analyzed by SYBR Green gene expression assay (Applied Biosystems, Waltham, MA, USA).
Chromatin immunoprecipitation. Chromatin immunoprecipitation (ChIP) was carried out using standard procedures. Briefly, cells were washed, scraped with ice-cold PBS, centrifuged, and the pellets were sonicated to generate ChIP DNA fragments (200-600 bp), which were cross-linked using standard protocols. Samples were incubated with either normal rabbit IgG, H3 antibody (Abcam, Cambridge, UK), H3K4me3 (Millipore, Billerica, MA, USA) or PU.1 antibody (Santa Cruz, Dallas, TX, USA) for overnight at 4°C under rotation, followed by incubation with protein A/G plus agarose beads for 2-3 h at 4°C. After several washes, precipitated chromatin complexes were eluted, and uncrosslinked overnight at 65°C with in buffer containing 5 M NaCl, followed by treatment with RNase A and proteinase K. DNA was extracted with QiaQuick PCR purification kit (Qiagen), and quantified by qPCR using primers listed in Supplementary Table S1.

Figure 7
A model of ARTD1 regulation of OC formation. In the absence of M-CSF and RANKL, the NLRP3 inflammasome is minimally active in Artd1 +/+ and Artd1 D214N/D214N cells (a and a'). As a result, the epigenetic action of ARTD1 D214N (ARTD1 D ) and WT ARTD1 to a lesser extent, promotes the expression of the repressors of OC differentiation, thereby inhibiting this process. In WT cells, NLRP3 inflammasome activation by M-CSF and RANKL cues leads to ARTD1 auto-PARylation, cleavage and subsequent degradation; this restrains ARTD1-dependent chromatin alterations, and enables the expression of activators and promotion of OC maturation (b). In contrast, ARTD1 D is resistant to cleavage and degradation (b'). Its sustained epigenetic action maintains the expression of OC repressors, thereby, inhibiting OC formation