Parathyroid hormone signaling in mature osteoblasts/osteocytes protects mice from age-related bone loss

Aging is accompanied by osteopenia, characterized by reduced bone formation and increased bone resorption. Osteocytes, the terminally differentiated osteoblasts, are regulators of bone homeostasis, and parathyroid hormone (PTH) receptor (PPR) signaling in mature osteoblasts/osteocytes is essential for PTH-driven anabolic and catabolic skeletal responses. However, the role of PPR signaling in those cells during aging has not been investigated. The aim of this study was to analyze the role of PTH signaling in mature osteoblasts/osteocytes during aging. Mice lacking PPR in osteocyte (Dmp1-PPRKO) display an age-dependent osteopenia characterized by a significant decrease in osteoblast activity and increase in osteoclast number and activity. At the molecular level, the absence of PPR signaling in mature osteoblasts/osteocytes is associated with an increase in serum sclerostin and a significant increase in osteocytes expressing 4-hydroxy-2-nonenals, a marker of oxidative stress. In Dmp1-PPRKO mice there was an age-dependent increase in p16Ink4a/Cdkn2a expression, whereas it was unchanged in controls. In vitro studies demonstrated that PTH protects osteocytes from oxidative stress-induced cell death. In summary, we reported that PPR signaling in osteocytes is important for protecting the skeleton from age-induced bone loss by restraining osteoclast’s activity and protecting osteocytes from oxidative stresses.


INTRODUCTION
Osteoporosis affects an estimated 200 million people worldwide and it becomes increasingly prevalent in the aging population [1,2]. It is well established that the first 34 amino acids of PTH and PTH-related peptide (PTHrP) are necessary and sufficient to fully activate the PTH/PTHrP receptor (PPR) and both PTH AGING (teriparatide) and PTHrP (Abaloparatide) are approved anabolic agents to treat osteoporosis. PPR is coupled to G-proteins capable of activating multiple pathways, including those signaling through cyclic adenosine monophosphate (cAMP)/protein kinase A (PKA), phospholipase C (PLC)/protein kinase C (PKC), and non-PLC-dependent PKC and Ca i ++ [3]. In the skeleton, PTH and PTHrP exert their anabolic and catabolic effects by binding and activating the PPR expressed on cells of the osteoblast lineage. This lineage comprises a variety of cells, from osteoprogenitors to mature osteoblasts and osteocytes; however, cellular targets of PTH actions are still not completely understood.
Mice with constitutively active PPR in osteocytes display increased trabecular bone mass, increased osteoblast number, and decreased Sost/sclerostin expression [20,21] whereas mice lacking RANKL in osteocytes have high bone mineral density and osteopetrosis [14,15], demonstrating an important role for osteocytes in bone remodeling. We have previously generated mice with conditional knockout (KO) of the PPR predominantly in osteocytes by using the 10-Kb dentin matrix protein 1 (Dmp1) promoter to drive Cre recombinase expression in PPR-floxed mice (Dmp1-PPR KO ) [22]. At 3months of age, Dmp1-PPR KO mice show normal serum calcium, phosphate, and PTH, suggesting that under physiological conditions PPR signaling in osteocytes is not needed to maintain normal mineral homeostasis. These mice display a significant increase in trabecular and cortical bone, indicating that PPR on osteocytes is required for normal bone remodeling. When subjected to intermittent or continuous PTH administration, Dmp1-PPR KO mice generated blunted anabolic and catabolic skeletal responses, indicating that PPR signaling in osteocytes is necessary for full skeletal responses to the hormone [22].
To study the role of PPR signaling in osteocytes in agedependent osteopenia, we analyzed the skeletal phenotype of mice at 4 (adult) and 13 (middle-aged) months of age. As compared with controls, 4-month-old Dmp1-PPR KO animals showed increased trabecular bone and decreased osteoclast number and activity, whereas at 13 months these mutant mice had a significant decrease in trabecular bone associated with increased osteoclast number and activity. In vitro, PTH significantly protected osteocytic cells from hydrogen peroxide (H 2 O 2 ) induced cell death and reactive oxygen species (ROS) production. This effect was lost in cells lacking receptor expression. All together these data highlight an essential role of PPR signaling in osteocytes to protect against age-related bone loss and oxidative stresses.

PPR ablation in mature osteoblasts/osteocytes induces severe osteopenia in 13-month-old male mice
Mice lacking PPR in mature osteoblasts/osteocytes, namely Dmp1-PPR KO , have increased bone mineral density and bone mass by three months of age and they are resistant to both the anabolic and catabolic effects of PTH [22]. To investigate whether PPR signaling in mature osteoblasts/osteocytes is needed to maintain skeletal homeostasis during age-dependent bone loss, we analyzed the skeletons of adult and middle-aged animals. Hematoxylin and eosin (H&E) staining of the tibiae and von Kossa staining of the fifth lumbar (L5) vertebrae and the femora showed significant increase in trabecular bone volume over total tissue volume (BV/TV%) at 4 months, as previously reported for young mice [22]. The increased bone mass was followed by a dramatic bone loss in Dmp1-PPR KO mice at 13 months, as compared to littermate controls ( Figure  1A-1D, Table 1, Supplementary Figure 1A). Cortical thickness was similar between Dmp1-PPR KO and controls at both ages ( Figure 1E). Micro-computed tomography (µCT) analysis further confirmed the histological data ( Figure 1F-1J, Table 2). At 4 months of age, in Dmp1-PPR KO mice there was a significant increase in BV/TV% in L5 and distal femur ( Figure  1H-1I) and a significant increase in trabecular thickness ( Figure 1H) in vertebral bones. In contrast, at 13 months of age, BV/TV% in both sites (L5 and femurs) were significantly decreased in Dmp1-PPR KO mice compared to littermate controls ( Figure 1H, 1I). The significant reduction in these trabecular parameters was still present in KO animals at 16 months of age AGING (Supplementary Figure 1B). In Dmp1-PPR KO mice, bone volume of L5 and the distal femora decreased by 41% and 71%, respectively, from 4 to 13 months, whereas in control animals this decrease was 12% and 42%, respectively ( Figure 1H, 1I). At 13 months of age, trabecular separation was also significantly increased in the L5 of Dmp1-PPR KO mice ( Figure 1H) and trabecular number (Tb.N) was significantly decreased in the distal femora ( Figure 1I) of Dmp1-PPR KO mice compared to controls. Taken together this data revealed a marked age-dependent trabecular bone loss in the absence of PPR signaling in mature osteoblasts/osteocytes. Cortical area, as assessed by µCT analysis of the midshaft femur, was significantly increased at 4 months in Dmp1-PPR KO mice ( Figure 1G, 1J, Table 2).  Tables 1 and 2 for the full list of parameters. Analyses were performed in a blinded fashion. Unpaired Student's t test (C-E) and Two-way ANOVA with Tukey's post hoc test or Mann-Whitney test (H-J) was performed. * p < 0.05, ** p < 0.01, *** p < 0.001. Abbreviations: Tb: Trabecular; Cort: cortical; BV: bone volume; TV: total tissue volume; Th: thickness; N: number; Sp: separation; A: area; pMOI: polar moment of inertia; Por: porosity.  Similarly, other cortical parameters, including polar moment of inertia, were also significantly increased in the mutant mice at 4 months of age, compared to controls whereas they were indistinguishable between the two genotypes at 13 months, indicating a differential temporal regulation of trabecular and cortical bone by PPR signaling in mature osteoblasts/osteocytes with age.

PPR deletion in mature osteoblasts/osteocytes increases osteoclast, but decreases osteoblast, activity in 13-month-old mice
To delineate the cellular mechanism of the age-related osteopenia in the Dmp1-PPR KO animals, we performed histomorphometric analysis on the L5 and the femora of adult and middle-aged mice. As shown in Figure  1C-1E, histomorphometric analysis confirmed the decrease in trabecular BV/TV% and Tb.N in male Dmp1-PPR KO mice at 13 months, both in axial and appendicular sites. Tartrate resistant acid phosphatase (TRAP) staining on the distal femora of these mice showed relatively fewer osteoclasts per bone perimeter at 4 months, but strikingly more TRAP-positive cells at 13 months in Dmp1-PPR KO animals as compared with controls ( Figure 2A-2B). The significant increase in the number of TRAP-positive osteoclasts was also present in the proximal tibiae of 20-month-old Dmp1-PPR KO female mice compared with control littermates (Supplementary Figure 2A). These findings were further supported by a significant reduction in osteoclast activity, such as erosion and osteoclast surface per bone surface, in 4-month-old Dmp1-PPR KO mice but a significant increase of all these parameters at 13 months of age ( Figure 2B, Table 1). Bone formation rate (BFR) over bone volume (BFR/BV) was significantly decreased in Dmp1-PPR KO at 4 months and mineral apposition rate (MAR) was significantly reduced in 13-month-old Dmp1-PPR KO mice compared to controls ( Figure 2C-2D, Table 1). These results indicate that, in the absence of PPR signaling in mature osteoblasts/osteocytes, there is an age-dependent increase in osteoclast numbers and activity and ageindependent decrease in osteoblast activity resulting in increased bone resorption and bone loss.

Age-dependent changes in serum markers and skeletal genes in Dmp1-PPR KO mice
The main function of PTH is to maintain mineral homeostasis and it is still unclear whether mature osteoblasts/osteocytes directly contribute to mineral-ion homeostasis. To investigate if lack of PPR signaling in mature osteoblasts/osteocytes impaired mineral ions homeostasis, we measured serum levels of calcium, phosphate and PTH in adult and middle-aged mice. Biochemical analysis of male Dmp1-PPR KO and littermate control mice showed normal calcemia and phosphatemia at 4 and 13 months of age ( Figure 3A), demonstrating that PPR in mature osteoblasts/osteocytes is not required to maintain mineral-ion homeostasis. In both control and Dmp1-PPR KO , with age, there was a significant decrease in serum calcium and an increase in serum PTH, similar to what has been observed in older adult mice. PTH levels were similar between Dmp1-PPR KO and controls at 4 months of age ( Figure 3A), whereas they were significantly increased in Dmp1-PPR KO mice at 13 months of age ( Figure 3A), indicating a possible resistance to PTH. This increase in serum PTH in Dmp1-PPR KO was not observed in 13-month-old female mice (Supplementary Figure 3B), suggesting the sexdependent difference. Interestingly, phosphate serum levels significantly increased with age in Dmp1-PPR KO mice but not in controls ( Figure 3A). Serum markers of bone formation, procollagen type 1 N-terminal propeptide (P1NP), were significantly reduced in both male Dmp1-PPR KO and controls at 13 months whereas markers of bone resorption, C-terminal telopeptide of type I collagen (CTX), were unchanged in both mice groups at both ages ( Figure 3B), despite the significant increase in osteoclast numbers and activities present in Dmp1-PPR KO mice at 13 months of age, as shown in Table 1 and Figure 2B.
To investigate the molecular mechanism leading to the increased osteoclasts in 13-month-old Dmp1-PPR KO mice, we assessed the expression of PPR (encoded by Pth1r) and other osteocytic markers in bone marrowdeprived long bones of adult and middle-aged AGING Dmp1-PPR KO and control mice. Receptor ablation in osteocytes was still present in both male and female 13month-old KO animals, as demonstrated by a significant decrease in PPR expression ( Figure 3C, Supplementary Figure 4A). Although middle-aged male Dmp1-PPR KO mice showed a trend of increase (2.0-fold, p = 0.06) in Sost expression compared to controls, overall Sost expression was unaffected by genotypes ( Figure 3C). Interestingly, serum sclerostin was significantly increased in the mutant mice as compared to controls, which can contribute to the suppression of bone formation present in these animals ( Figure 3A). Other osteocytic genes, namely RANKL and osteoprotegerin (OPG), were unchanged, suggesting that other factors might be driving the age-dependent increase in osteoclast numbers and activity present in middle-aged male Dmp1-PPR KO mice. In the female mice, OPG expression was significantly downregulated, while the RANKL/OPG ratio was significantly increased, in KO mice at 13 months of age (Supplementary Figure 4A), demonstrating sex-dependent differences.
Tumor necrosis factor α (TNFα) promotes osteoclastogenesis independently of RANKL [23]; therefore, we performed immunofluorescence staining for TNFα on the tibias of 13-month-old male mice and control littermates. In Dmp1-PPR KO mice, the number of TNFα-expressing osteocytes was significantly decreased as compared to control littermates (Supplementary Figure 5), suggesting that other factors might drive osteoclastogenesis. In addition, we found a trend of increase (1.3-fold, p = 0.06) in M-CSF expression in the bone marrow of Dmp1-PPR KO mice at 13 months of age compared to control littermates Figure 4B), and this upregulation of M-CSF in KO mice was also observed in 16-month-old mice (1.5-fold, p < 0.001, male control and Dmp1-PPR KO mice was measured by ELISA. PTH levels were normalized to 13-month-old control due to a high intra-assay variation (>9.8%). (B) Serum levels of bone formation (P1NP) and resorption marker (CTX) were also measured by ELISA. N = 7-19 per group. Data are presented as mean ± SEM. (C) Gene expression in marrow-removed long bones of male mice was analyzed with qPCR. N = 4-6 per group. Two-way ANOVA with Tukey's post hoc test, unpaired Welch's t test, or Mann-Whitney test was performed. * p < 0.05, ** p < 0.01, *** p < 0.001. Each gene expression was relative to 3 housekeeping genes (Actb, B2m, and Gapdh), which were selected by GeNorm. Expression data are presented as mean ± SEM.
data not shown). Sphingosine kinase (Sphk1) mediates TNFα-induced arthritis and osteoclastogenesis via TNFα receptor activating factor 2 (TRAF2) [24]. We analyzed the expression of Sphk1 in the bone marrow of Dmp1-PPR KO mice and found that Sphk1 was significantly upregulated as compared to control littermates (Supplementary Figure 6B), suggesting a potential involvement of TNFα-expressing osteocytes in the increase in M-CSF and Sphk1 expression in the bone marrow of Dmp1-PPR KO mice. Next we measured serum levels of cytokines, previously reported to be regulated by PTH or by aging, including TNF, monocyte chemoattractant protein 1 (MCP-1/CCL2), and interleukin (IL)-6 and -10 in serum of 13-month-old male mice (Supplementary Figure 3A). While there was a trend of increase (2.2-fold, p = 0.06) in serum IL-10 in Dmp1-PPR KO compared to controls, serum concentration of other cytokines was unchanged.

Serum from middle-aged Dmp1-PPR KO male animals increases the number of osteoclasts in vitro
To explore the molecular mechanism(s) driving the osteopenia and the increase in osteoclast numbers and Immunofluorescence staining of perilipin-1, RANKL and DAPI was performed on the tibiae of middle-aged (13 months) male control and Dmp1-PPR KO mice. Representative images of a tibia from Dmp1-PPR KO mouse are shown. In the merged image, DAPI, perilipin-1 and RANKL staining is shown in blue, red, and green, respectively. The orange arrowheads indicate RANKL + BMAT (identified as perilipin-1 + ). The number of BMAT (left) and the number (middle) and frequency (right) of RANKL + BMAT in the BM space were analyzed. N = 4 per group. Unpaired student's t test was performed. * p < 0.05. Data are presented as mean ± SEM. AGING activity in middle-aged male Dmp1-PPR KO mice, we treated bone marrow mononuclear cells (BMMCs) isolated from wild type mice (3-4 months old) with serum obtained from 13-month-old male mice (Dmp1-PPR KO and littermate controls) and analyzed osteoclastogenesis by TRAP staining and activity. BMMCs treated with serum from KO mice significantly increased the number of TRAP + osteoclasts compared to the control group ( Figure 5A), whereas TRAP activity was similar between the two groups (Supplementary Figure 3C). These findings suggest that, in the absence of PTH signaling, osteocytes secrete factors that contribute, in part, to the increase in osteoclasts present in middle-aged Dmp1-PPR KO mice.

Osteoprogenitors are decreased in middle-aged Dmp1-PPR KO animals
Next we assessed if lack of PPR signaling in mature osteoblasts/osteocytes influenced the commitment or frequency of osteoprogenitor cells, and therefore their osteogenic potential. As shown in Figure 5B, colony forming unit osteoblasts (CFU-Ob) were significantly reduced in bone marrow of male Dmp1-PPR KO animals AGING compared to controls at 13 months, demonstrating a progressive reduction in osteoprogenitors. Similarly, middle-aged female KO animals showed a marked reduction in CFU-Ob in the bone marrow (Supplementary Figure 3D).

Serum from middle-aged Dmp1-PPR KO male animals promotes adipogenic differentiation of BMSCs in vitro
Since circulating factors, including bone morphogenic proteins, are involved in osteolineage commitment of bone marrow stromal cells (BMSCs) [25], we examined the effect of serum from middle-aged mice on osteogenic differentiation of BMSCs. We treated BMSCs isolated from wild type mice (3-4 months old) with serum from 13-month-old male Dmp1-PPR KO mice and littermate controls. Treatment with serum from control and KO mice showed no difference in proliferation and alkaline phosphatase (ALP) activity of BMSCs ( Figure 5C), while BMSCs treated with serum from KO mice significantly increased adipogenic differentiation, as assessed by oil-red-O staining ( Figure  5D). These findings demonstrate that circulating factors promote lineage commitment of BMSCs. To investigate if the secreted factors are osteocyte-derived, we treated BMSCs with conditioned medium from ex vivo culture of osteocyte-enrichment bone explants (OEBEs) from 13-month-old control and Dmp1-PPR KO mice. Similarly, there was no difference in both BMSC proliferation and osteogenic differentiation between both treatment groups (Supplementary Figure 3E). These results indicate that factors secreted from osteocytes are not the major contributor to the reduced osteoprogenitors in Dmp1-PPR KO mice.

Marrow adipocytes are the source of RANKL
As demonstrated in Figure 5D, treatment with serum from 13-month-old Dmp1-PPR KO mice promoted adipogenic differentiation of BMSCs in vitro. To examine if there was any change in marrow adiposity in these mice, we performed histological analysis on bone marrow. In Dmp1-PPR KO male animals, at 13 months of age, there was a significant increase in marrow adiposity ( Figures 1A, 4A). Since marrow adipocytes have been reported as a source of RANKL [26,27], we analyzed RANKL expression in marrow adipocytes by immunofluorescence staining. Staining of RANKL along with an adipocyte marker, perilipin-1, revealed that the number of RANKL + marrow adipocytes was markedly increased in Dmp1-PPR KO compared to control male mice ( Figure 4C). However, the ratio of RANKL + adipocytes over total adipocytes was unchanged between control and mutant animals, suggesting that the increase in the number, but not the frequency, of RANKL + adipocytes contributed to the increased osteoclast numbers in KO mice. It has also been reported that bone marrow adipocytes express M-CSF [28]. Interestingly, we found that M-CSF expression was upregulated (1.3-fold, p = 0.06) in the bone marrow of male Dmp1-PPR KO mice as compared to controls ( Figure 4B). This may indicate that the increased marrow adiposity also contributed to the upregulation of M-CSF in the middle-aged mutant mice. Expression of RANKL was unchanged in the bone marrow of 13-month-old male mice (Supplementary Figure 6A). Interestingly, the marrow adiposity was unchanged between female control and Dmp1-PPR KO at 20 months of age, suggesting sexual dimorphism (Supplementary Figure 2B).

PPR signaling protects osteocytes from early onset of oxidative stress in vivo
Aging is accompanied by an accumulation of oxidative stress due to an imbalance between pro-oxidants and antioxidants [29,30]. PTH protects osteoblasts from oxidative stress-induced cell death [31]; therefore we hypothesize that a similar effect was also present in osteocytes. To test the hypothesis, we analyzed the expression of 4-hydroxynonenal (4-HNE), a biomarker for oxidative stress-induced lipid peroxidation, in the L3/4 vertebrae of 4-and 13-month-old Dmp1-PPR KO and control littermates. As shown in Figure 6A, Dmp1-PPR KO mice exhibited a significant increase in 4-HNEpositive osteocytes at 4 months of age, suggesting the protective effect of PPR signaling in osteocytes from early onset of oxidative stress.
Oxidative stress is one of the causes of cellular senescence and it has been previously shown that senescent osteocytes contribute to age-related bone loss [32]. We investigated if the age-dependent bone loss in Dmp1-PPR KO mice was driven by senescent osteocytes. The expression of the senescence marker p16 Ink4a was significantly increased in Dmp1-PPR KO mice at 13 months of age compared to 4 months, whereas it was unchanged in controls ( Figure 6B). Other markers of cellular senescence were unchanged (Supplementary Figure 4C).

PTH protects osteocytes from oxidative stressinduced cell death and intracellular ROS accumulation in vitro
To study the effects of PTH in oxidative stress, we used an in vitro model in which PPR expression was knocked-out by CRISPR/Cas9 technique. In the absence of PPR expression in Ocy454-12H cells [33] Figure 7C), which is similar to the phenotype present in 3-monthold Dmp1-PPR KO animals [22], whereas Sost expression was unchanged compared to control cells. 12H-PPR KO cells were treated with hPTH  or forskolin for 18-22 hrs and then exposed to a high dose of H 2 O 2 (1 mM). As shown in Figure 6C-6D, PTH treatment significantly suppressed oxidative stress-induced cell death and intracellular accumulation of ROS in control cells, whereas this effect was lost in 12H-PPR KO cells, demonstrating that PTH protects osteocytes from oxidative stress-induced cell death.

PTH protects osteocytes from oxidative stress
ROS accumulation promotes oxidative stress and cellular senescence, therefore we evaluated the effects of PTH during oxidative stress. Control and 12H-PPR KO cells were pretreated with hPTH(1-34) prior to 7 and 14 days of continuous exposure to a low dose of H 2 O 2 (150 µM). In the non-pretreated groups, H 2 O 2 exposure markedly increased the number of senescenceassociated β-galactosidase-positive (SA β-gal + ) cells in both control and 12H-PPR KO cells, while PTHpretreatment significantly reduced the number of SA βgal + cells in control, but not in 12H-PPR KO cells ( Figure  6E, Supplementary Figure 8A). Gene expression in these cells after 7 and 14 days of H 2 O 2 (100-200 µM) exposure ( Figure 6G) demonstrated that PTH had a long-lasting effect on Sost and RANKL expression. Importantly, pretreatment of control, but not 12H-PPR KO , cells with PTH significantly suppressed p21 expression at day 7, demonstrating a potential molecular mechanism by which PTH protects osteocytes from oxidative stress-induced senescence. Expression of p16 Ink4a was significantly upregulated in control cells exposed to H 2 O 2 alone (-PTH) at day 14 ( Figure 6G).  (Figure 6F), demonstrating that, under oxidative stress, osteocytes secrete osteoclastogenic factors, which might contribute to the increased osteoclasts in Dmp1-PPR KO mice. Treatment with H 2 O 2 medium alone did not increase osteoclast numbers compared to CM control (data not shown).
Taken together these results demonstrate that, in the absence of PPR signaling in mature osteoblasts/osteocytes, there is an age-dependent trabecular bone loss associated with increased bone resorption driven by a significant increase in osteoclast numbers and activity and impaired bone formation. Mechanistically, PPR signaling in mature osteoblasts/osteocytes regulates osteoblast formation through serum sclerostin and osteoclastogenesis via secreted factors other than RANKL and OPG. At the molecular level, PPR signaling protects, in vitro, osteocytes from oxidative stress-induced cell death through a cAMP-mediated mechanism. Further studies will be needed to elucidate the downstream effectors.

DISCUSSION
Over the last two decades, the actions of PTH have expanded to include important effects on skeletal homeostasis and hematopoiesis. In particular, the amino-terminal fragment of PTH and PTHrP were approved by FDA as therapeutic agents capable of restoring bone mass and increasing the number of hematopoietic stem cells [1,35,36]. It has been shown that the anabolic effects of PTH in bone comprise the recruitment of osteoblast progenitors, the suppression of osteoblasts and osteocytes apoptosis, the suppression of Sost/sclerostin expression and the activation of bone lining cells. Although the cellular targets of these actions are still not completely understood, the use of genetically manipulated animals has shed light on some of the hormonal actions. Using transgenic mice in which the PTH receptor is ablated predominantly in mature osteoblasts and osteocytes, we have demonstrated that receptor expression in these cells was required for bone modeling and remodeling and for a full anabolic and catabolic response to PTH administration [22].
Here we report that, with aging, Dmp1-PPR KO mice develop a significant osteopenia characterized by reduced trabecular bone, whereas the cortical compartment is relatively unaffected. The bone loss is driven by an increase in the number of osteoclasts, their surface area and activity and a concomitant reduction in osteoblast activity (Figure 7). In contrast, at 4 months of age, and as previously reported for 3-month-old animals [22], Dmp1-PPR KO mice have increased trabecular bone associated with reduced osteoclast and osteoblast functions. These findings demonstrate that PPR signaling in osteocytes exerts different effects on trabecular and cortical bone and that these effects are age dependent and shift, with age, from maintenance of bone homeostasis to prevention of bone loss. Interestingly, recent transcriptomic profiles of skeletal tissue in male and female mice over their lifespan identified Pth1r as one of the genes highly dependent on both gender and age [37]. We identify a notably sexdependent difference in our study (8.67-fold-increase of RANKL in male vs. female at 13-month-old, p < 0.001) that can be explained in light on the Pth1r differential expression (1.72-fold-increase in male vs. female at 13month-old, p = 0.07).
Considering the classical role of PTH in calcium homeostasis, we investigated whether PPR signaling in mature osteoblasts/osteocytes is required in maintaining mineral ion homeostasis. There was a significant agerelated decrease in serum calcium and increase in serum PTH in both genotypes. Notably, 13-month-old Dmp1-PPR KO male mice had a higher serum PTH level than littermate controls, whereas no difference was found between female Dmp1-PPR KO mice and controls. The significant increase in serum PTH in Dmp1-PPR KO males could be due to an age-dependent resistance to PTH. Elevated serum PTH levels might be indicative of a secondary hyperparathyroidism-like state, which may contribute to the accelerated bone loss in Dmp1-PPR KO mice. Despite the elevated serum PTH, osteoblasts were not increased, most likely due to the concomitant increase in serum sclerostin. In addition, in the Dmp1-PPR KO mice, there is an age-dependent exhaustion of osteoprogenitor cells, as supported by a reduction in CFU-Ob. Phosphate levels were also increased in male Dmp1-PPR KO mice with age, whereas were unchanged in the control mice. Since PTH increases FGF23 in osteocytes [38], Dmp1-PPR KO mice might have a lower FGF23 secretion than control littermates, which can lead to reduced phosphate excretion in the kidneys.
To delineate the molecular mechanisms by which Dmp1-PPR KO animals display severe bone loss with Figure 7. Graphical summary. 13-month-old Dmp1-PPR KO mice showed trabecular bone loss driven by increased osteoclast number and activity and reduced osteoblast function. Mechanistically, the lack of PPR signaling in mature osteoblasts/osteocytes decreases osteoprogenitors and increases serum sclerostin, RANKL-expressing marrow adipocytes and early onset of 4-HNE+ osteocytes and p16 Ink4a upregulation in KO mice. Circulating factor(s) from these mutant mice increases, directly or indirectly, osteoclastogenic and adipogenic differentiation of BMMCs and BMSCs, respectively. Furthermore, in vitro data showed that PPR signaling induces long-lasting suppression of p21 and protects osteocytes from oxidative stress-induced intracellular ROS accumulation, cell death and senescence.
AGING age, we analyzed the expression of osteocytic markers. Despite the important role of osteocytes in regulation of both osteoblasts and osteoclasts, expression of Sost, RANKL, and OPG was similar between middle-aged male Dmp1-PPR KO and littermate controls. Contrarily, in 13-month-old Dmp1-PPR KO females, OPG and RANKL/OPG expression were significantly suppressed and increased, respectively, as compared to littermate controls. In male mice, Sost and RANKL expression at 4 months of age was unchanged between genotypes, whereas we previously reported an increase and decrease in Sost and RANKL expression, respectively, in KO animals at 3 months of age [22]. At the moment we do not have any explanation for this age-dependent difference and further studies will be needed.
RANKL, OPG, and RANKL/OPG expression were unchanged in the long bones of male mice, suggesting that additional cytokines might be involved in the increase in osteoclast numbers and activity. Serum TNFα levels were unchanged between control and mutant mice at 13 months of age, and TNFα expression in osteocytes, as assessed by immunofluorescence, was decreased in long bones of mutant mice compared to controls, indicating that additional factors might be involved. Bone marrow adipocytes are responsive to PTH and are a local source of RANKL; therefore we investigated if increased osteoclast activity in Dmp1-PPR KO mice was dependent on these cells. Indeed, we found that, with aging, there was a significant increase in RANKL+ adipocyte in the bone marrow of mutant mice, as compared to controls. Treatment of BMMCs with serum of Dmp1-PPR KO mice induced a significant increase in osteoclasts as compared to controls, demonstrating the presence of osteoclastogenic factors in the serum of mutant mice. We previously identified osteocyte-derived molecules that modulate osteoclast numbers and activities, such as Nrp1, Sema3a, and Sema3d [39]. RNA-sequencing of osteocytes [40] demonstrated that PTH significantly increased the expression of Nrp1 (2.52-fold increase, p < 0.001) in these cells, suggesting a possible mechanism of action.
Serum CTX is often used as a marker of bone resorption and, despite the severe bone loss and the significant increase in osteoclasts, we did not detect any significant changes in serum CTX in Dmp1-PPR KO mice, compared to controls. It has been reported that systemic CTX levels vary according to circadian rhythm and that food intake reduces the levels of CTX in humans [41,42]. Moreover, CTX measurements can be affected by lipids and proteins present in the serum after food intake. Since mice were not starved before serum collection, it is possible that the levels of CTX measured were not accurate.
PTH reduces the rate of both osteoblast and osteocyte apoptosis, which releases factors capable of promoting osteoclastogenesis [43][44][45]. We investigated whether the osteopenia present in the Dmp1-PPR KO animals at 13 months old was due to increased osteocyte apoptosis. Terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick-end labeling (TUNEL) assay showed no difference in osteocyte apoptosis in both controls and Dmp1-PPR KO mice, both in males and females (Supplementary Figure 2C, 2D), suggesting that mechanisms other than osteocyte apoptosis are responsible for the osteopenia.
We demonstrated that 13-month-old Dmp1-PPR KO mice have reduced CFU-Ob and increased marrow adiposity compared to controls, suggesting that the stem cell population might be depleted. Similarly, we observed the reduced expression of PPR in bone marrow cells in Dmp1-PPR KO male mice, but not in females (Supplementary Figure 6A, 6C). Treatment of BMSCs with serum from Dmp1-PPR KO mice induced adipogenic differentiation but had no effect on osteoblastic differentiation. It has been reported that ceruloplasmin (Cp), a multicopper ferroxidase (also known as an adipokine), increases adipogenic differentiation of MC3T3 cells [46]. Cp is also the one of the osteocyte-secreted proteins that we identified previously [39] and PTH has been reported to reduce mRNA expression of Cp in osteocytes [40], indicating the possible involvement of Cp in the increase in marrow adiposity in KO mice. We also observed a trend of increase in Sost expression in long bone of 13month-old Dmp1-PPR KO male mice compared to controls and a significant increase in serum sclerostin. We can speculate that sclerostin may contribute to the reduction in osteoblast proliferation, possibly in favor of marrow adipocytes, by suppressing the Wnt/-catenin signaling pathway. Indeed, it has also been reported that sclerostin reduces proliferation and differentiation of BMSC, in part by suppressing bone morphogenetic protein activity and by increasing bone marrow adipose tissue (BMAT) [19,47]. Further studies will be needed to identify the molecular mechanism(s) by which osteocytes affect the BMSCs.
Aging is characterized by reduced skeletal mass and increased oxidative stress and recent studies identified senescence and senescent osteocytes as important players in age-dependent bone loss [32,48]. The aging process comprises not only cellular dysfunction and genomic instability but also stem cell exhaustion. Several studies have demonstrated that the decline in bone formation that occurs with age, both in humans and animal models, is invariably associated with reduced proliferation and differentiation of mesenchymal stem cells. Here we report that the AGING absence of PPR expression in mature osteoblasts/osteocytes in mice induces a significant increase in the number of 4-HNE + osteocytes in 4month-old mutant mice compared to control littermates and a marked increase in p16 Ink4a expression at 13 months of age. These findings suggest that PTH protects osteocytes from early onset of oxidative-stress. Further studies are needed to unveil the underlying molecular mechanism.
Our in vitro study showed PTH protects osteocytes from oxidative stress-induced cell death and intracellular accumulation of ROS in a cyclic AMP-dependent manner. These results suggest that signaling through PPR, which is expressed on these cells and some mature osteoblasts, is needed to protect osteocytes from oxidative stress and possibly cellular senescence. One mechanism by which Dmp1-PPR KO animals become severely osteopenic is by increased production of ROS, due to reduced activities of antioxidants during aging [29]. Indeed, the role of ROS and PTH in osteoblastic cells has been documented by the work of Jilka et al. [45], demonstrating that intermittent PTH administration reduces intracellular ROS and p66 shc phosphorylation.
Autophagy has also recently been shown to play an important role in aging and senescence. To investigate a possible relationship between PPR signaling in osteocytes and autophagy, we analyzed expression of transcripts known to regulate autophagy, such as Sirt-1, FOXO-1 and Beclin-1. However, we found no changes in expression of these genes in our mutant mice (Supplementary Figure 4B).
It has been previously reported that, similar to PTH, PTHrP protects osteoblastic cells from oxidative stress. Ardura et al. reported that PTHrP counteracts the proapoptotic actions of ROS by modulating mitogenactivated protein kinases (MAPK) phosphatase 1 (MAPK1) and promoting dephosphorylation of MAPK [49]. Mice with the "knock-in" (KI) of the 1-84 fragment of PTHrP, which lacks both the nuclear localization sequence (NLS) and the C-terminus, display early senescence and defective osteoblast functions. In these animals, ROS levels are increased and antioxidant enzymes are downregulated, demonstrating a role for PTHrP in prevention of oxidative stress [50][51][52]. Since PTH and PTHrP both bind to and activate the PPR, it is plausible to hypothesize that they might exert similar effects. Additional studies will be needed to analyze PTH and PTHrP responses. Interestingly, PTHrP mRNA expression in long bone of control and Dmp1-PPR KO mice was similar, therefore we can speculate that this effect was not dependent on PTHrP.
In summary, we have reported, for the first time, that PPR signaling in mature osteoblasts/osteocytes is needed to protect the skeleton from age-dependent bone loss. In Dmp1-PPR KO mice, there is a striking agedependent trabecular bone loss driven by increased osteoclast number and activity and reduced osteoblast function. Mechanistically, we demonstrated that lack of PPR signaling decreases osteoprogenitors while increasing serum sclerostin, RANKL-expressing marrow adipocytes and 4-HNE + osteocytes. Moreover, there was an age-dependent upregulation of p16 Ink4a in KO mice. We also found that circulating factor(s) from these mutant mice increases, directly or indirectly, osteoclastogenic and adipogenic differentiation of BMMCs and BMSCs, respectively. Furthermore, in vitro data showed that PPR signaling induces longlasting suppression of Sost and p21 and protects osteocytes from oxidative stress-induced intracellular ROS accumulation, cell death and possibly senescence (Figure 7).

Mice
Dmp1-PPR KO (Dmp1-Cre + ; PPR fl/fl ) mice under the C57BL/6 genetic background were generated as described previously [22]. Littermates homozygous for floxed PPR gene, but lacking Cre-recombinase expression, were used as control (Dmp1-Cre -;PPR fl/fl ). Details are provided in Supplementary Methods. Institutional Animal Care and Use Committee and the Subcommittee on Research Animal Care at Massachusetts General Hospital and Boston University Medical Center approved all animal protocols.

Cell lines
Ocy454-12H (or 12H) cells, a derivative of the conditionally immortalized osteocytic cell line Ocy454, were used for all the in vitro experiments [33,53,54]. CRISPR/Cas9 genome editing technique was used to knockout PPR expression in these cells (12H-PPR KO ). Both control and 12H-PPR KO cells were routinely maintained at 33°C (permissive temperature) with 5% CO 2 and cultured in growth medium (α minimum essential medium (Gibco, Thermo Fisher Scientific) containing 10% heat-inactivated fetal bovine serum (Gibco) and 1% antibiotic-antimycotic (Gibco)). Upon proliferation, cells were transferred to 37°C with 5% CO 2 to be fully differentiated into mature osteocytes and incubated for the number of days required for each experiment.

Senescence-associated β-gal staining
Cells were plated at a density of 1.0 × 10 5 cells/ml in growth medium on a 6-or 12-well plate (Biolite™) under 37°C with 5% CO 2 followed by treatment, on the next day, with 10 nM hPTH(1-34) for 18-22 hrs. Cells were then exposed to oxidative stress by adding an equal volume of 300 μM H 2 O 2 medium (final concentration: 150 μM) on top of the well and cultured for an additional 7 or 14 days. Medium was replaced once on day 7. On day 7 and 14, conditioned medium was collected and the cells were either harvested for RNA isolation or fixed with 4% paraformaldehyde (Acros Organics, Fair Lawn, NJ, USA) in PBS for 15 min at room temperature for SA β-gal staining. Cells were then washed with PBS and incubated with SA βgal staining solution (pH 6.0) overnight at 37°C without CO 2 injection. The SA β-gal staining solution was prepared by following the protocol from Chen et al. [55]. Cell viability was measured using PrestoBlue cell viability reagent (Invitrogen) as per manufacturer's instruction. Bright field images were acquired under the microscope with a 4X objective (Keyence). The number of SA β-gal positive osteocytes was counted in a blinded fashion using ImageJ.

Oxidative stress-induced cell death and reactive oxygen species accumulation
Cells were plated at a density of 2.6 × 10 4 cells/cm 2 in growth medium on a 24-well plate and grown for 3 days at 33°C. The growth medium was replaced and the cells were cultured for an additional 2 days under 37°C and treated with either 10 nM hPTH(1-34) or 10 µM forskolin for 18-22 hrs followed by an exposure to 1 mM H 2 O 2 for 4 hrs or overnight. Cell viability was measured using PrestoBlue cell viability reagent as per manufacturer's instruction. Intracellular levels of ROS were determined upon cell staining with 14 µM of 2ʹ, 7ʹ-dichlorofluorescin diacetate (DCFDA) (Sigma) for 30 min at 37°C with 5% CO 2 . The cells were then trypsinized and transferred into wells of a 96-well plate and fluorescence (485 nm excitation and 535 nm emission) was measured using a spectrophotometer (TriStar LB941, Berthold Technologies, Oak Ridge, TN, USA). The fluorescent intensity was normalized with cell viability.

Statistics
Normal distribution and heteroscedasticity of the data was tested using D'Agostino-Pearson or Anderson-Darling test and Spearman's test, respectively. Equality of variance between two datasets was analyzed with the F test. Parametrical statistics were used for the data that follow normal distribution and/or equal variance. Otherwise, non-parametrical statistics were used. Statistical significance was defined as p values less than 0.05 (two-tailed). Statistical test used for each analysis is described in the legends of each figure. All statistical analysis was performed on GraphPad Prism (GraphPad Software, San Diego, CA, USA). All mice were euthanized by cervical dislocation following isoflurane (Henry Schein, Melville, NY, USA) anesthesia. Femora and tibiae from both sides and vertebrae were dissected and cleaned of soft tissue. Left femora were used for µCT analysis and subsequently used for either histomorphometric analysis (described below). Right femora were used for histological analysis (described below). The L5 vertebrae were used for µCT analysis and subsequently used for histomorphometric analysis. The third and fourth lumbar (L3 and L4) vertebrae were used for histological analysis (described below). The right femora and/or tibiae were used for RNA isolation and gene expression analysis (described below).

Cell lines
CRISPR/Cas9 genome editing technique was used to knockout PPR expression in Ocy454-12H (or 12H) cells. Briefly, three independent single-guided RNAs (sgRNAs) were designed using both sgRNA Design Tool ( Figure 7). An empty PX458 plasmid (without any sgRNA insertion) was used to generate control cells (12H-PPR Cont ). Both control and 12H-PPR KO cells were routinely maintained at 33°C (permissive temperature) with 5% CO 2 and cultured in growth medium (α minimum essential medium (αMEM) (Gibco, Thermo Fisher Scientific) containing 10% heat-inactivated fetal bovine serum (FBS) (Gibco) and 1% antibiotic-antimycotic (Gibco)). Upon proliferation, cells were transferred to 37°C with 5% CO 2 to be fully differentiated into mature osteocytes and incubated for the number of days required for each experiment. Fresh growth medium was added every three-four days.

Colony forming unit assay
Upon sacrifice, BMSCs were isolated from long bones (tibiae and/or femora) of 13-month-old control and Dmp1-PPR KO animals (both male and female) by either a brief centrifugation or flushing out bones using a 25G needle (BD, Franklin Lakes, NJ, USA) with αMEM containing 0.1% bovine serum albumin (BSA) (Sigma) and 25 mM HEPES buffer (Fisher Scientific, Hampton, NH, USA). Cells were hemolyzed with red blood cell lysing buffer (Sigma) and plated on either 6-or 24-well plates (Biolite ™ , Thermo Fisher Scientific, or Primaria ™ , Corning, Corning, NY, USA) at 1.0-2.0 × 10 6 cells/well under growth medium. Cells were cultured at 37°C with 5% CO 2 for 2 days and nonadherent cells were removed by replacing the growth medium with osteogenic differentiation medium (growth medium supplemented with 10 mM βglycerophosphate (Calbiochem, Millipore, Burlington, MA, USA), and 50 g/ml of ascorbic acid (Sigma)). The differentiation medium was replaced every 2-3 days. After 6-12 days in culture, cells were fixed with 10% phosphate-buffered formalin (Fisher Scientific) and stained for alkaline phosphatase (ALP) by incubating with 0.1 mg/ml naphthol AS-MX phosphate (#N5000, Sigma) and 0.6 mg/ml Fast Blue BB salt AGING (#151112, MP Biomedicals, Santa Ana, CA, USA) dissolved in 0.1 M Tris buffer (pH 8.5) (Fisher Scientific). CFU-Ob assay was performed by quantifying the number or the area of colonies that are positive for ALP as previously reported [2,3]. Cells were subsequently stained with 0.05% crystal violet (CV) for the fibroblasts CFU (CFU-F) assay and the total area of CV-positive colonies was quantified using ImageJ (NIH, Rockville, MD, USA).

Histology
The left tibiae and/or the L3 and L4 vertebrae were dissected from adult (4 months old) and middle-aged male (13-16 months old) or female (13-20 months old) mice and were fixed in 10% phosphate-buffered formalin (pH 6.9-7.2) for at least 24 and 48 hrs, respectively. The bones were further decalcified in 20% ethylenediaminetetraacetic acid (EDTA) (Fisher Scientific or Alfa Aesar, Haverhill, MA, USA), processed, embedded in paraffin, and sectioned as described previously (12). Following deparaffinization, sections were stained with hematoxylin (Thermo Fisher Scientific) and eosin (Fisher Scientific) or used for immunofluorescent staining, TUNEL assay, or TRAP staining (each described below).

Immunohistochemistry
Immunohistochemistry for 4-HNE was performed on deparaffinized sections of the L3 and L4 vertebrae from adult (4 months old) and middle-aged (13 months old) male mice. Endogenous peroxidase activity was inhibited by 3% H2O2 treatment for 10 min. Antigen retrieval was performed with Tris-EDTA buffer (10 mM Tris, 1 mM EDTA, 0.05% Tween 20, pH 9.0) in a boiling water bath at 95°C for 12 min. Slides were cooled to room temperature and blocked with TNB blocking buffer for 30 min at room temperature and incubated with anit-4-HNE antibody (#GTX40953, GeneTex, Irvine, CA, USA) in TNB blocking buffer overnight at 4°C. Slides were washed and incubated with rabbit anti-goat IgG (#305-065-003, Jackson ImmunoResearch Laboratories) for 30 min at room temperature. Subsequently, slides were washed and incubated for 30 min with streptavidin-conjugated horseradish peroxidase (#43-8323, Zymed Laboratories) and tyramide following the manufacturer's protocol (TSA Biotin Kit, PerkinElmer). The signal was developed with a 3,3'-diaminobenzidine (DAB) substrate-chromogen system (Vector Laboratories). Cell nuclei were counterstained with hematoxylin. Bright field images were acquired under the microscope with a 40× objective (Nikon Eclipse E800, Nikon, Tokyo, Japan) or with a 20× objective (EVOS XL, Thermo Fisher Scientific). For quantitative analysis, using ImageJ, 4-AGING HNE positive osteocytes were counted with a 20× objective in three-five adjacent fields per vertebra and normalized to the total number of osteocytes per field.

Marrow adiposity analysis
Bone marrow adiposity was quantitatively analyzed, using ImageJ, as the ratio of adipocyte area over the marrow area within 300 or 1,200 µm from the proximal epiphyseal plate of the tibiae of middle-aged male (13 months old) or aging female (20 months old) animals, respectively. Briefly, bright-field images of H&E-stained proximal tibiae were acquired under the microscope with a 4x objective (Keyence). Using the segmented line tool (width = 300 or 1,200 µm) on ImageJ, the marrow area of interest was defined by tracing the epiphyseal plate manually. The area of marrow adipocytes (represented as white patches with a round or oval shape) was manually traced by using the polygon selection tool.

TUNEL assay
TUNEL staining was performed on deparaffinized tibiae sections of Dmp1-PPR KO and control male (13 months old) and female (20 months old) mice following the manufacturer's protocol (DeadEnd colorimetric TUNEL, Promega, or TUNEL chromogenic apoptosis detection kit, GeneCopoeia, Rockville, MD, USA). For the TUNEL chromogenic kit (GeneCopoeia), modifications were made on the antigen retrieval and permeabilization step where the deparaffinized and rehydrated tibia sections were exposed to 10 mM citrate buffer (pH 6.0) for 5 min at 48°C and 0.05% pepsin (Roche, Basel, Switzerland) in 0.1 N HCl (Fisher Scientific) for 4 min at 37°C, respectively as reported previously [30]. Cell nuclei were counterstained with either hematoxylin or 0.5% methyl green (Sigma). Images were acquired under the microscope with a 40X objective (EVOS XL). The number of TUNEL-positive apoptotic osteocytes in the cortical and trabecular bone was counted using ImageJ either manually or unbiasedly. For the unbiased quantitative analysis, images were first color-deconvoluted into three singlecolored images (methyl green, DAB, and background color) by manually selecting regions of interest that represent each color. The intensity of DAB signal on each osteocyte (identified as a methyl green positive cell within the bone matrix) was quantified and osteocytes with high and homogeneous DAB signal (threshold: mean >100, modal >60, standard deviation >60 per cell) were defined as TUNEL-positive.

Histomorphometric analysis
Adult (4 months old) and middle-aged (13 months old) male mice were injected intraperitoneally with 20 mg/kg calcein (Sigma) on days 9 and 2 before sacrifice. The left femora and the L5 vertebrae were isolated, cleaned of adherent tissues, and used for µCT analysis (described above) prior to histomorphometric analysis. Both the femora and vertebrae were dehydrated, infiltrated, and embedded in methyl methacrylate. Undecalcified 4-μm-thick sections were cut using a microtome (Leica RM 2255, Heidelberg, Germany). Consecutive sections were stained by von Kossa with van Gieson counterstain and toluidine blue to determine structural and cellular parameters, respectively. Dynamic parameters were quantified using unstained and undecalcified sections of the femora. Undecalcified sections of the femora were also stained for TRAP. Histomorphometric parameters (i.e., formation and resorption) were measured using the Osteomeasure image analysis system (OsteoMetrics, Decatur, GA, USA) coupled to a microscope (BX 50 combined with a DP72 digital camera, Olympus, Hamburg, Germany). A sampling site of about 2 mm 2 was established in the cancellous bone at 400 µm below the growth plate. Analysis was done in a standardized fashion by an experienced scientist and results were expressed according to the updated standardized American Society for Bone and Mineral Research (ASBMR) nomenclature [4]. In detail, cancellous bone volume in the distal femora and the L5 vertebrae was assessed as BV/TV (%). Tb.Th, Tb.N, and Tb.Sp were also calculated both in the distal femora and in the L5 vertebrae. Additional parameters included: mineralizing surface over bone surface, MS/BS (%); mineral apposition rate, MAR (µm/day); bone formation rate (BFR) over bone surface, BFR/BS; over bone volume, BFR/BV (%/year); and over total volume, BFR/TV (%/year); osteoid volume (OV) over bone volume, OV/BV (%); and osteoid surface (OS) over bone surface, OS/BS (%). Formation parameters were expressed as osteoblast surface per bone surface, Ob.S/BS (%); osteoblast number per bone perimeter, N.Ob/B.Pm (/mm); and osteoblast number per tissue area, N.Ob/T.Ar (/mm 2 ). Resorption parameters were expressed as osteoclast surface per bone surface, Oc.S/BS (%); osteoclast number per bone perimeter, N.Oc/B.Pm (/mm); and osteoclast number per tissue area, N.Oc/T.Ar (/mm 2 ). We also quantified the number of osteocytes per bone volume, N.Ot/BV (/mm 2 ), in the cancellous bone. Cortical bone parameters, such as Cort.Th (mm); endocortical mineral apposition rate, End.Cort MAR (µm/day); and osteocyte density, Ot density (/mm 2 ), were analyzed at the femoral midshaft. See Supplementary Table 1 for the full list of analyzed parameters.

RNA isolation and purification
The right femora and/or tibiae were dissected from adult male (4 months old) and middle-aged (13-16 months old) male or middle-aged and aging (13 and 20 months old) female mice. The bones were cut at both proximal and distal end, centrifuged briefly at 8.0-9.0 × g for 7 sec (at start) to remove bone marrow, placed in 1 ml of TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) with a grinding stainless steel bead (5-mm diameter, OPS Diagnostics, Lebanon, NJ, USA), snapfrozen in liquid nitrogen, and subsequently stored at -80°C for RNA isolation. The bone marrow removed from the femora and/or tibiae was lysed in 600 µl of lysis buffer (RLT Plus (Qiagen) supplemented with 1% β-mercaptoethanol (Sigma)) by pipetting and stored at -80°C for RNA isolation. Cells under in vitro culture were harvested in 350 µl of lysis buffer and either frozen at -80°C or directly processed for RNA isolation.
Prior to RNA isolation, the frozen bones were homogenized with TissueLyser II (Qiagen) at 30 Hz at room temperature for 8-12 min (including the time for thawing), refrozen in liquid nitrogen, and subsequently restored at -80°C. Bone marrow and in vitro cell samples were both homogenized with QIAshredder (Qiagen). The total RNA from long bones or bone marrow and in vitro cells was isolated using PureLink RNA mini kit (Invitrogen) or RNeasy Plus mini kit (Qiagen), respectively, as per manufacturer's instructions. Some of the bones from middle-aged male (13 months old) mice were subjected to sequential collagenase (Worthington Biochemical, Lakewood, NJ, USA) and EDTA digestions to remove endosteal and periosteal osteoblasts and bone marrow cells and snapfrozen in liquid nitrogen followed by homogenization and RNA isolation. The quality and quantity of the total RNA was measured by UV spectrophotometry (NanoDrop 2000c, Thermo Fisher Scientific).

Quantitative real time PCR
RNA was reverse transcribed into complementary DNA (cDNA) using PrimeScript RT reagent kit (Clontech, Takara Bio, Kusatsu, Shiga, Japan) with a genomic DNA elimination step or qScript cDNA SuperMix (QuantaBio, Beverly, MA, USA). Real time qPCR assay was subsequently performed with StepOne Plus (Applied Biosystems) using SYBR Green (Power SYBR, Thermo Fisher Scientific). The PCR conditions were as follows: one cycle of 95°C (10 min), 50 cycles of 95°C, 58°C, and 72°C (15 sec, 15 sec, and 30 sec, respectively), and one cycle of 95°C (30 s), followed by a melting curve step (58°C to 95°C with an increment of +0.5°C, 1 min for each temperature). See Supplementary Table 2 for the primer sequences. Data analysis was based on the comparative cycle threshold (ΔCT) method and expression of each gene was normalized to Actb expression or the geometric mean of the three gene expression (Actb, B2m, and Gapdh), which were selected as the most stable housekeeping Unlike control cells, the sequence of PPR KO after the cleavage site was unreadable due to random repair via nonhomologous end joining (NHEJ). (C) Expression of PPR and osteocytic genes in control and PPR KO cells treated with 10 nM PTH or 10 µM forskolin was analyzed by qPCR. N = 3 per group. Data are presented as mean ± SD. Two-way ANOVA with Tukey's post hoc test was performed. * p < 0.05, ** p < 0.01, *** p < 0.001.

Supplementary Tables
Supplementary Table 1. Abbreviations of bone parameters for histomorphometrical and µCT analysis.