SIRT6-PAI-1 axis is a promising therapeutic target in aging-related bone metabolic disruption

The mechanistic regulation of bone mass in aged animals is poorly understood. In this study, we examined the role of SIRT6, a longevity-associated factor, in osteocytes, using mice lacking Sirt6 in Dmp-1-expressing cells (cKO mice) and the MLO-Y4 osteocyte-like cell line. cKO mice exhibited increased osteocytic expression of Sost, Fgf23 and senescence inducing gene Pai-1 and the senescence markers p16 and Il-6, decreased serum phosphate levels, and low-turnover osteopenia. The cKO phenotype was reversed in mice that were a cross of PAI-1-null mice with cKO mice. Furthermore, senescence induction in MLO-Y4 cells increased the Fgf23 and Sost mRNA expression. Sirt6 knockout and senescence induction increased HIF-1α binding to the Fgf23 enhancer sequence. Bone mass and serum phosphate levels were higher in PAI-1-null aged mice than in wild-type mice. Therefore, SIRT6 agonists or PAI-1 inhibitors may be promising therapeutic options for aging-related bone metabolism disruptions.


Evaluation of bone parameters and serum phosphate levels in cPKO mice.
To elucidate the molecular targets of SIRT6 in osteocytes, we assessed the mRNA expression of potent regulators of bone formation (Sost), bone resorption (Opg and Rankl), and Fgf23 in osteocyte-enriched cortical bones. Quantitative polymerase chain reaction (qPCR) analysis indicated that the mRNA expression of Sirt6 was knocked out in the cortical bone samples of cKO mice (Fig. 2a). mRNA of Sost and Fgf23 was significantly increased in the tibial cortical bone of cKO mice (Fig. 2a), whereas differences between Opg and Rankl mRNA expression were not significant between cKO and the control mice ( Supplementary Fig. 2). We previously revealed that SIRT6 deficiency enhanced the expression of Pai-1, a potent senescence marker, in chondrocytes 3 , so the mRNA level of Pai-1 was also assessed. As expected, Pai-1 mRNA expression also increased in cKO osteocytes (Fig. 2a). We then evaluated the expression of senescence marker genes, such as p53, and p16 and the markers of senescenceassociated secretory phenotype (SASP), interleukin-6 (Il-6), C-X-C motif chemokine ligand 1 (Cxcl1). qPCR analysis indicated that the expression of these genes was increased in cKO mice (Fig. 2a). In humans, FGF23 is a potent regulator of phosphate metabolism 9 . Therefore, serum phosphate levels were evaluated in cKO mice. As expected, Serum phosphorus levels were significantly lower in cKO mice compared to control mice, whereas, serum calcium levels were comparable in control and cKO mice (Fig. 2b). DMP-1 is also expressed in the distal tubules of the kidney 20 . The distal tubules express klotho, an important factor in phosphorus metabolism. To verify the involvement of the variation in klotho expression in the abnormal phosphorus metabolism observed in cKO mice, we compared the expression of klotho between controls and cKO by q-PCR (Fig. 2c). The results showed no significant difference in klotho expression between control and cKO, suggesting that phosphorus metabolism in cKO mice was due to increased FGF23 expression from osteocytes. Immunohistochemistry analysis of cortical bone in cKO mice. Immunohistochemistry of the cortical bone was performed to detect the presence of SOST, FGF23, and PAI-1. The percentage of SOST-, FGF23-, and PAI-1-positive cells relative to total osteocytes was higher in cKO mice than in the control mice ( Fig. 3a,b). These results indicate that SIRT6 deficiency stimulates the expression of SOST, FGF23, and PAI-1 in osteocytes. Osteocyte number was quantified by HE-stained cortical bone samples and showed no significant difference between wild type and cKO (Fig. 3c).
SIRT6 suppresses cellular senescence through multiple factors [20][21][22][23][24] , and PAI-1 is not only a marker of cellular senescence but is also sufficient for the induction of senescence 25 . Thus, cKO mice were crossed with PAI-1-null mice (DMP-1cre::Sirt6f /f ::Pai-1−/− mice, cPKO mice) to investigate the role of PAI-1 in the phenotype of cKO mice. The cPKO mice grew normally, and their body weights were comparable with those of cKO and control mice ( Supplementary Fig. 1). qPCR analysis of cortical bone in cPKO mice. qPCR analysis indicated that the mRNA expression of Sirt6 and Pai-1 were knocked out in the cortical bone samples of cPKO mice (Fig. 4a). We used qPCR analysis to investigate the expression of Sost and Fgf23 in the cortical bones from cKO and cPKO mice and found that Sost and Fgf23 mRNA levels were significantly reduced in cPKO compared to cKO mice (Fig. 4a). Thus, PAI-1 deletion clearly reduced the upregulation of these genes cause by SIRT6 deficiency. Similarly, the expression of p53, p16, Il-6 and Cxcl1, was decreased in cPKO mice compared to cKO mice (Fig. 4a).

Evaluation of bone parameters and serum phosphate levels in cPKO mice. Histological analyses
showed higher bone volume in cPKO mice than in cKO mice in the lumber bone (Fig. 4b,c). In contrast, micro-CT analyses of distal femur showed the difference of BV/TV between cKO miceand cPKO mice was not significant. Reduced phosphate levels in cKO mice were mitigated by the deletion of PAI-1 in cPKO mice (Fig. 4d), whereas serum Ca 2+ levels were comparable between cKO mice and cPKO mice (Fig. 4d). www.nature.com/scientificreports/ www.nature.com/scientificreports/   Sost and Fgf23 in osteocytes, Sirt6 was knocked down by siRNA in osteocyte-like MLO-Y4 cells. qPCR analysis indicated that Sost, Fgf23 and Pai-1 expression increased due to SIRT6 deficiency (Fig. 5a). This suggests that SIRT6 regulates Sost, Fgf23 and Pai-1 expression.  www.nature.com/scientificreports/ qPCR analysis of senescent MLO-Y4 cells. To address the role of cell senescence in the expression of Sost and Fgf23, senescence was induced in MLO-Y4 cells using two different methods: long-term (21 days) culturing after confluency (Long-Term Confluent; LTC) 26,27 , and culturing under doxorubicin administration 28 . qPCR analysis revealed that Sost and Fgf23 mRNA expression was stimulated in MLO-Y4 cells by both LTC and doxorubicin treatments, and the increase in Sost and Fgf23 mRNA levels occurred concurrently with the upregulation of senescence marker gene and SASP marker gene, such as p16 and Il-6 ( Fig. 5b,c). MLO-Y4 cells exhibited an increase in senescence-associated β-galactosidase (SAβ-gal) expression after induction (Fig. 5d).
HIF-1α activation enhances Fgf23 expression. We investigated the mechanisms underlying the regulation of Fgf23 transcription. There is a novel enhancer of Fgf23 at -16 kb upstream of the transcriptional start site 29 . Additionally, the transcription factor HIF-1α activates Fgf23 transcription in osteoblast-like cell lines 30 . SIRT6 directly binds to HIF-1α, preventing its activity 31 . Thus, we focused on the role of SIRT6 and cell senescence in HIF-1α-mediated Fgf23 transcription. We examined the effect of SIRT6 deletion on the binding of HIF-1α to the Fgf23 enhancer with a chromatin immunoprecipitation (ChIP) assay. The HRE (Hypoxia Response Element) identified in silico from the enhancer region located 16 kb upstream of the Fgf23 transcription start site was used as a bait (Fig. 6a,b). HIF-1α protein was detected at low levels with the Fgf23 enhancer DNA under basal conditions (Fig. 6a). However, when Sirt6 was knocked out using the CRISPR/Cas9 system in MLO-Y4 cells, HIF-1α binding to the Fgf23 enhancer sequence increased significantly (Fig. 6a). This finding suggests that SIRT6 regulates the binding of HIF-1α to the enhancer region of Fgf23 in a cell-senescence-independent manner. We then investigated the binding of HIF-1α to the Fgf23 enhancer in senescent cells. When senescence was induced in MLO-Y4 cells after 21 days of culture without passage, HIF-1α binding to the Fgf23 enhancer increased significantly compared with that in MLO-Y4 cells under basal conditions (Fig. 6b). These data indicate that HIF-1α binds to the Fgf23 enhancer region of MLO-Y4 cells in a senescence-dependent manner as well.
To evaluate the role of HIF-1α in the regulation of Fgf23 expression, MLO-Y4 cells were cultured with an HIF-1α activator or inhibitor. The expression of Fgf23, along with that of Pai-1, was significantly increased in MLO-Y4 cells following the administration of N-(2-methoxy-2-oxoacetyl) glycine methyl ester (DMOG), a prolyl 4-hydroxylase (P4H) inhibitor, which increases HIF-1α levels by inhibiting HIF-1α prolyl hydroxylase (HIF-PH) (Fig. 6c). The expression of Sost was also increased by DMOG administration (Fig. 6c). When MLO-Y4 cells were cultured with CAY10585, an HIF-1α inhibitor that suppresses the transcription of HIF-1α target genes, Fgf23 expression was not enhanced even when Sirt6 was knocked down (Fig. 6d). In contrast, when SIRT6 was overexpressed in MLO-Y4 cells, the expression of Fgf23 and Sost was suppressed, and these effects were cancelled by DMOG (Fig. 6e). These data indicate that HIF-1α activation is indispensable for Fgf23 upregulation due to SIRT6 deficiency or senescence.

Evaluation of bone parameters in aged PAI-1−/− mice.
To verify whether PAI-1 deficiency prevents age-related changes in bone tissues and mRNA transcripts, we compared bone tissues from aged PAI-1-deficient (PAI-1KO) and wild-type mice. At 6 months of age, the BV/TV of PAI-1KO was lower than that of WT and the trabecular number, trabecular separation, and trabecular spacing were higher in both the femur and lumbar spine. BMD was comparable in the lumbar spine, but PAI-1KO was lower in the femur. Conversely, at 18 months after birth, PAI-1KO was higher for BV / TV and trabecule number in both femur and lumbar spine, and PAI-1KO was lower for trabecular separation and trabecular spacing. BMD was also higher in PAI-1KO (Fig. 7a). Analysis of the cortical bone tissue by qPCR indicated that the mRNA expression of Fgf23 and Sost decreased in PAI-1-KO mice compared with those in littermates at 18 months after birth (Fig. 7b). Serum phosphate levels were elevated in 18 months-old PAI-1-KO mice compared with those in littermates (Fig. 7c).
Evaluation of the expression of FGF23, SOST and PAI-1 mRNA in human bone. Finally, to analyze the relationship between the expression of FGF23, SOST, and PAI-1 in human bone and age, q-PCR analysis was performed using bone samples from the femoral neck bone taken during hip arthroplasty. The results showed that the expression levels of FGF23, SOST, and PAI-1 were significantly correlated with age of donors (Fig. 8a).

Discussion
The regulatory mechanisms of bone mass in aged animals are poorly understood. We aimed to investigate the role of SIRT6, a longevity-associated factor, in osteocytes. We found that Sost, Fgf23, expression was increased in SIRT6-deficient osteocytes, leading to a decrease in osteoblasts and an increase in osteoclasts, resulting in decreased bone mass. In addition, PAI-1, whose expression was promoted by SIRT6 deficiency, was actively involved in bone metabolism, indicating that PAI-1 deficiency suppresses age-related bone loss. Thus, we showed that induction of senescence by SIRT6 deletion was one of the causes of age-related changes in bone metabolism. Based on the results of this study, we proposed two signaling pathways downstream of SIRT6, a senescencedependent and a senescence-independent pathway involved in the regulation of SOST and FGF23 (Fig. 8b). The relationship between degenerative cellular changes associated with aging and bone loss has been studied. Aging decreases cellular mitochondrial activity and increases oxidative stress. Genetically engineered mice with mitochondrial dysfunction or loss of mitochondrial homeostasis exhibit osteoporosis 32,33 . Mice lacking osteocyte-specific mitochondrial superoxide dismutase 2 (SOD2) exhibit osteocytopenia and osteoporosis with increased ROS levels 34 .
Among the studies on the association between sirtuin genes and bone metabolism, SIRT1 has been the most studied. SIRT1 promotes mitochondrial production through deacetylation of PGC-1a and inactivation of HIF-1a, and decreased SIRT1 activity leads to mitochondrial depletion 35  www.nature.com/scientificreports/ www.nature.com/scientificreports/ or osteoclasts both exhibit bone loss via increased NF-kB activity 36 . In mice in which SIRT1 is stabilized in osteocytes, Sost transcription is inhibited 37 . Combined with the results of this study, it is clear that both SIRT1 and SIRT6 regulate SOST expression. Rats lacking SIRT6 specifically in bone marrow mesenchymal stem cells exhibited suppressed osteoblast differentiation, whereas rats overexpressing SIRT6 exhibited the opposite phenotype. Both suggested that NF-kB regulation may be at least partially involved in bone metabolism 38 . Although SIRT6 global knockout mice exhibit low-turnover osteoporosis 5 , osteoblast-specific SIRT6-deficient mice driven by the osteocalcin promoter show reduced bone mass due to increased osteoclasts associated with reduced OPG expression 39 . Interestingly, bone formation parameters were not affected in these mice. Mice lacking SIRT6 in hematopoietic cells, including osteoclast precursors, had reduced osteoclast numbers and increased bone mass; SIRT6 formed a complex with B lymphocyte-induced maturation protein-1 (Blimp1) and suppressed the expression of osteoclast differentiation inhibitors such as Mafb 40 . www.nature.com/scientificreports/ SIRT6 deletion induces cell senescence 6,24 SIRT6 binds to telomeres and prevents cell senescence by preserving telomere function through H3K9 deacetylation 24,41 . Furthermore, SIRT6 deficiency enhances Pai-1 expression, which induces replicative senescence downstream of p53 3,4,21,25 . cKO mice showed increased Fgf23 expression and osteocyte senescence, and Pai-1 depletion in cKO mice led to the normalization of Fgf23 expression. Similarly, two different methods for the induction of cell senescence (LTC and doxorubicin treatment) stimulated Fgf23 expression. In cells undergoing senescence, activation of the p53-p21 and p16-RB pathways are induced. p16 and p21 have different target CDKs, and each of them alone has a weak ability to activate RB protein. However, the simultaneous action of p16 and p53-p21 cause the RB protein to be permanently activated and arrest the cell cycle progression 42 . In this study, we showed that both p16 and p53 expression were upregulated in SIRT6-deficient

SIRT6 in osteocytes
Senescence-independent pathway www.nature.com/scientificreports/ osteocytes, and that PAI-1 deletion suppressed these expressions. These data suggested that SIRT6 regulated cell senescence induction mechanisms at least in part through the regulation of PAI-1 expression. Senescent cells express inflammatory cytokines, such as IL-6, PAI-1, and CXCL1, as part of the senescence-associated secretory phenotype (SASP) response 41 . IL-6 stimulates FGF23 promoter activity through STAT3 in osteoblast-like UMR106 cells 43 . NF-κB signaling plays a critical role in the induction of SASP 44 . NF-κB activation results in FGF23 upregulation downstream of PKC or p38 MAPK 45,46 . These data indicate that senescence plays a role in FGF23 regulation. In hepatic cells, SIRT6 directly binds to HIF-1α and inhibits the transcription of its target gene, Glut-1 31 . Through the deletion of Sirt6 in MLO-Y4 cells, we showed that HIF-1α was able to bind to the Fgf23 enhancer via a consensus motif within its sequence, which promotes Fgf23 transcription (Fig. 6a). Senescent cells have a causal role in aging-related osteoporosis, and the depletion of senescent cells in the bone ameliorates aging-related osteoporosis 18 . Senescent cell-depleted mice present a bone tissue phenotype characterized by a reduced number of osteoclasts and an increased number of osteoblasts, both of which have been linked to the suppression of Sost expression 18 . These data, together with the findings of the present study, indicate that cell senescence, regulation of Sost expression, and possibly SASP, play important roles in aging-related osteoporosis. Epigenetic regulation of Sost controls sclerostin levels, which correlates with bone density and fracture rates 47 . The promoter region of Sost has a CpG-rich domain, which contributes to the transcriptional regulation of Sost 48 . A study of postmenopausal women showed lower Sost mRNA levels in patients with osteoporosis and significantly increased CpG methylation in the Sost promotor region 48 . SIRT6 has histone deacetylase (HDAC) activity, and so it may regulate the expression of Sost through deacetylation of the promoter region; however, the role of the HDAC activity of SIRT6 was not elucidated in this study.
PAI-1 is a novel regulator of FGF23 metabolism 49 . HIF-1 promotes PAI-1 transcription, thereby increasing its expression. In osteocytes, hypoxia or age-related decrease in SIRT6 activity may promote Pai-1 expression through HIF-1 activation, leading to osteocyte senescence. In fact, there was an increase in Pai-1 expression in cKO mice.
The candidate senescence-independent factors for the regulation of Sost downstream of SIRT6 are HIF-1 and NF-κB. Sost expression is induced by hypoxia through HIF-1α activation in MC3T3-E1 osteoblastic cells 50 . HIF-1α activates Sost transcription by binding to its promoter in MC3T3-E1 cells 50 . Additionally, SOST levels are elevated by hypoxia in human Saos-2 osteogenic sarcoma cells 51 . Consistently, we reported that Sost expression was induced by SIRT6 deletion or cell senescence in the MLO-Y4 osteocyte-like cell line in a HIF-1α-dependent manner. Together, these results suggest that SIRT6 may bind to HIF-1α and regulate Sost expression via its HDAC activity. TNFα stimulates Sost expression through an NF-κB-dependent mechanism in MLO-Y4 cells. NF-κB directly binds to NF-κB binding elements on the Sost promoter to induce Sost transcription 52 . SIRT6 is also a negative regulator of NF-κB signaling 23 . Thus, NF-κB activation may be involved in Sost upregulation in cKO mice; however, the role of NF-κB was not elucidated in this study.
In conclusion, we showed that osteocyte SIRT6 regulates bone volume and phosphate metabolism by modulating Sost and Fgf23 expression. This novel pathway explains the association between Fgf23 and Sost and agingrelated disorders. Furthermore, PAI-1 deficiency, along with the normalization of Sost expression, in cKO mice abolished the effects of SIRT6 deletion in bone tissues. A study of the Berne Amish community members, including those that are carriers of the null SERPINE1 gene that encodes PAI-1, revealed that a rare loss-of-function mutation in SERPINE1 prolongs human life and prevents age-related metabolic abnormalities 53 . Moreover, a small molecule PAI-1 inhibitor has been developed, and clinical trials are underway in humans 54 . We previously revealed that PAI-1 inhibition via a small molecule PAI-1 inhibitor prevented ovariectomy-induced bone loss in mice 55 . Nicotinamide mononucleotide (NMN) activates sirtuins, including SIRT1 and SIRT6, and its long-term administration corrects age-related metabolic disorders 56 . NMN administration promotes bone formation by preventing bone marrow mesenchymal stem cells from differentiating into adipocytes in aged mice 57 . NMN is confirmed to be safe in humans, and clinical studies regarding their use are being conducted 58 . Overall, these results support the potential therapeutic application of a SIRT6 agonist, such as NMN or a PAI-1 inhibitor, against aging-related disruptions of bone metabolism.

Methods
All methods were carried out in accordance with ARRIVE guidelines.

Animals.
All animal experiments were approved by the Animal Care and Use Committee of Tokyo Medical and Dental University and were carried out in accordance with the approved guidelines (approval number: A2018316). All mice were allowed unrestricted activity and given access to standard rodent food pellets (Labo MR Stock, Nosan, Tokyo, Japan) and tap water ad libitum. Sirt6f /+ mice (FVB.129S6(Cg)-Sirt6tm1.1Cxd/J) were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Dmp1-Cre (B6N.FVB-Tg (Dmp-1-Cre)1Jqfe/ BwdJ) mice were kindly supplied by Dr. Shu Takeda from the Endocrine Center, Toranomon Hospital. Sirt6f /f mice were backcrossed for at least ten generations with C57/BL6J mice (CLEA Japan Inc., Tokyo, Japan) and then crossed with Dmp-1-Cre mice to generate Dmp-1-Cre::Sirt6f /+ mice, and their progeny were intercrossed to obtain cKO mice. PAI-1 ± mice (C57BL6/J background) were obtained from the Jackson Laboratory. PAI-1 ± mice were crossed with Dmp-1-Cre::Sirt6f /+ mice to obtain Dmp-1-Cre Sirt6f /+ ::Pai-1 ± mice. cPKO mice were obtained by crossing Dmp-1-Cre::Sirt6f /+ ::PAI-1 ± mice. Male mice were euthanized at 20 weeks of age using an overdose of isoflurane inhalation followed by cervical dislocation. Male PAI-1−/− mice and their lit- www.nature.com/scientificreports/ termates were grown for 72 weeks, and their samples were analyzed by micro-CT and qPCR analysis following euthanasia. None of the mice died during the experimental period.
Histological and histomorphometric analysis. Calcein (Sigma-Aldrich, St. Louis, MO, USA) was injected subcutaneously for bone labeling five and two days prior to euthanasia. Blood and bone samples were collected at the time of euthanasia. The undecalcified samples of the spinal bone were sectioned, and the third and fourth lumbar vertebrae were stained using von Kossa and tartrate-resistant acid phosphatase (TRAP) staining, as previously described 59 . The OsteoMeasure Analysis System (OsteoMetrics, Decatur, GA, USA) was used for static and dynamic histomorphometric analyses following the nomenclature defined by the American Society for Bone and Mineral Research, as previously described 60 .
Immunohistological analysis. Cortical bone sections were fixed in 4% paraformaldehyde on ice for 10 min. After washing with 1% T-PBS buffer for three times (5 min each), the sections were blocked with normal serum at room temperature for 1. Micro-CT analysis. Two-dimensional images of the distal femur and lumbar spine were obtained through micro-CT analysis (Comscan, Yokohama, Japan). The following three-dimensional morphometric parameters were determined using TRI/3D-BON software (RATOC, Tokyo, Japan): bone morphometric analysis of femoral bones performed at a region of 0.2 to 1 mm above the distal growth plates of the femora, bone volume/tissue volume (BV/TV), trabecular bone thickness (Tb.Th), Tb.N, Tb.Sp, and Tb.Spac for trabecular bone analyses 61 . The mineralized tissue volumes were measured using a calibration curve obtained from the BMD phantom.
Chromatin immunoprecipitation assay. The EpiQuik ChIP kit (Epigentek Group Inc., NY, USA) was used for the ChIP assay, and it was performed according to the manufacturer's instructions. To evaluate the binding of HIF-1α to the Fgf23 enhancer 29 , PCR was carried out using a Thermal Cycler 9700 (Applied Biosystems, Foster City, CA, USA) according to a standard procedure. The primer sequences used were 5′-GTC AAG TGA GTC CGG CTT CA-3′ (forward) and 5′-CCG AGC CAG GAC TTT CCT TT-3′ (reverse).
Human samples. The Ethics Committee of Tokyo Medical and Dental University approved this study (approval number: M2000-2121). All methods were carried out in accordance with the guidelines and regulations of the Ethics Committee of Tokyo Medical and Dental University. All the donors provided written informed consent to participate in the study. Bone samples were harvested under informed consent from the femoral neck of patients (n = 19) with hip osteoarthritis or femoral hip fracture scheduled to undergo total hip arthroplasty or surgical fixation. Exclusion criteria included a history of steroid administration, inflammatory arthritis, autoimmune disease, metastatic cancer, osteonecrosis of femoral head, endocrine disorders, or bone-related disorders.
Statistical analysis. Data are presented as mean ± standard deviation. Statistical analyses of the quantitative measures among the three groups were performed using one-way ANOVA. To assess the significance of differences between groups, a two-tailed Student's t-test was employed. Statistical significance was set at P < 0.05. Pearson linear regression was used to determine the degree of association between mRNA expression of FGF23, SOST, PAI-1 and age of donors. The linear regression coefficient R were reported. Values of P < 0.05 were accepted as significant.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.