Divalent metal cations stimulate skeleton interoception for new bone formation in mouse injury models

Bone formation induced by divalent metal cations has been widely reported; however, the underlying mechanism is unclear. Here we report that these cations stimulate skeleton interoception by promoting prostaglandin E2 secretion from macrophages. This immune response is accompanied by the sprouting and arborization of calcitonin gene-related polypeptide-α+ nerve fibers, which sense the inflammatory cue with PGE2 receptor 4 and convey the interoceptive signals to the central nervous system. Activating skeleton interoception downregulates sympathetic tone for new bone formation. Moreover, either macrophage depletion or knockout of cyclooxygenase-2 in the macrophage abolishes divalent cation-induced skeleton interoception. Furthermore, sensory denervation or knockout of EP4 in the sensory nerves eliminates the osteogenic effects of divalent cations. Thus, our study reveals that divalent cations promote bone formation through the skeleton interoceptive circuit, a finding which could prompt the development of novel biomaterials to elicit the therapeutic power of these divalent cations. Mechanisms underlying bone formation induced by divalent metal cations remain largely unknown. Here the authors show that these cations can activate the skeleton interoceptive circuit through the immune-neural axis to initiate new bone formation.

I t has been known from the late 1990s that various divalent metal cations, such as magnesium ions (Mg 2+ ), zinc ions (Zn 2+ ), and copper ions (Cu 2+ ), play vital roles in bone growth, modeling, and remodeling [1][2][3] . Over the decades, the regulatory effects of these divalent cations on osteogenesis, osteoclastogenesis, angiogenesis, and immune responses have been gradually revealed [4][5][6][7][8][9] . However, it was not until recently that the involvement of nervous system in the new bone formation induced by divalent metal cations has begun to be realized 10 . So far, it remains unclear whether the central nervous system (CNS), which has emerged to play an important role in bone homeostasis [11][12][13][14] , participate in this process. The CNS not only reacts to external stimuli, such as temperature, sound, odor, and taste, as exteroception but also receives signals from many physiological systems inside the body, including the cardiovascular, respiratory, gastrointestinal, genitourinary systems, and nociceptive systems, as interoception. In recent years, the interoceptive processes by which our body senses, interprets, integrates, and regulates signals from peripheral organs have emerged as a key mechanism for the control of internal state of our body by CNS 15 .
The interocpetion system consists of ascending neural pathways that transmit the internal body signals to the brain, the CNS where the input interoceptive information is processed, and the descending neural pathways through which the interoceptive signals are circled back to regulate peripheral organs. In addition to the central and peripheral nervous system, several other components, such as the endocrine and immune systems, are proposed to be involved in the interoception process 16,17 .
Unmyelinated free sensory nerve endings branch throughout tissues to recognize and carry signals related to pain, temperature, and mechanical stimuli 18 . As the largest organ, the skeleton accounts for more than 20% of human body weight. Bone tissues have abundant sensory and sympathetic innervations that connect bones with dorsal root ganglia (DRG) and the CNS [19][20][21] . The upregulation of sympathetic tone through serotonin and cAMP-response element binding protein (CREB) signaling in the hypothalamus contributes to decreased bone formation and increased bone resorption 13,22 . We have recently found that CNS senses bone density through prostaglandin E2 (PGE 2 ) as an ascending interoceptive signal and regulates bone formation through sympathetic nerves as the descending interoceptive pathway 14 . Specifically, PGE 2 secreted by osteoblasts in response to mechanical loading activates PGE 2 receptor4 (EP4) to stimulate phosphorylation of CREB in hypothalamus, where the tyrosine hydroxylase (TH) expression is downregulated for a sympathetic activity to induce commitment of mesenchymal stem/stromal cell (MSCs) to osteoblast lineage. Knockout of the EP4 gene in the sensory nerves or knockout of cyclooxygenase-2 (COX2) in the osteoblastic cells significantly increased sympathetic activity and reduced bone volume in adult mice. Our studies further demonstrate that ascending skeleton interoceptive signaling target hypothalamus to maintain the balance between bone and fat metabolism 23,24 . Moreover, we also showed sensory innervation into porous endplates is responsible for spinal hypersensitivity, thus, lowdose celecoxib maintains skeleton interoception in the endplate, thereby decreasing vertebral endplate porosity and innervation for the treatment of spinal pain 25,26 . Importantly, our most recent studies reveal that Skeleton interoception regulates bone and fat metabolism through hypothalamic neuroendocrine hypothalamic neuropeptide Y (NPY) (ref. 1 ) and physiological PGE2 levels maintain skeleton interoception activity for bone homeostasis to reduce vertebral endplate porosity and spinal pain (ref. 2 ). Therefore, the skeleton interoception could represent an essential circuit of the CNS in the control of bone metabolism and may shed a light on our understanding of divalent metal cation-induced bone formation.
Innate and adaptive immune processes, which become profoundly apparent after bone injury, play important roles in bone biology 27 . The presence of a series of immune cell-derived cytokines in the early inflammatory stage promotes the recruitment of fibroblasts, mesenchymal stem cells, and osteoprogenitor cells from their local niches to initiate bone repair 28,29 . For example, PGE 2 , a well-recognized pro-inflammatory cytokine that is upregulated only during the initial stage of bone healing, is suggested to play an important role in new bone formation. The mutation of the 15-hydroxyprostaglandin dehydrogenase gene (HPGD), which is responsible for the degradation of PGE 2 30 , plays a major role in promoting tissue regeneration 31,32 . Indeed, patients with an HPGD mutation have presented with subperiosteal new bone formation 33 . Since PGE 2 can elicit primary pain and prolong nociceptor sensitization 34,35 , nonsteroidal antiinflammatory drugs (NSAIDs), which inhibit cyclooxygenase (COX), the limiting enzyme of PGE 2 , have been used to manage post-injury pain. However, there is increasing evidence that the use of NSAIDs, especially selective COX2 inhibitors, can affect bone healing 36,37 . These findings suggest that the inflammatory molecules produced by the immune system in response to bone injury may serve as biochemical signals to initiate the interoceptive control of bone regeneration.
In recent years, the immunomodulatory function of orthopedic biomaterials has been increasingly acknowledged. In fact, the host immune response to implanted biomaterials is now recognized as a determinant for the long-term survival and regenerative function of such biomaterials 38,39 . Owing to the central role of macrophages in the immune reaction to bone biomaterials, as well as their heterogeneity and plasticity, macrophage is one of the most important target cells for immunomodulation in the biomaterial field 39 . Divalent metal cations, such as Mg 2+ , Zn 2+ , and Cu 2+ , have been extensively used to modify orthopedic biomaterials since the discovery of their osteogenic effects 8,[40][41][42][43] . However, despite their well-recognized roles in the regulation of the immune response 44 , the mechanism through which the divalent cation-modulated immune niche contributes to bone regeneration remains largely unclear. Indeed, the in vitro effects of divalent metal cations on osteogenesis starkly contradict our findings in a more complex in vivo model, because their osteogenic effects on the well-orchestrated bone healing process may involve the interplay of multiple systems in our body. We have recently showed the essential role of early inflammatory response in Mg 2+ -induced new bone formation, as Mg 2+ stimulates macrophage via transient receptor potential cation channel member 7 (TRPM7) to create a pro-osteogenic immune microenvironment 7 . Since the effective window of Mg 2+ coincides with the reinnervation phase of bone healing 45 , we hypothesize the communication between the immune system and the neural system can trigger the skeleton interoception for the regulation of new bone formation.
In this study, we sought to characterize the mechanism for divalent cation-induced bone formation. We found that divalent metal cations, including Mg 2+ , Zn 2+ , and Cu 2+ , activated skeletal interoception through the immune-neural axis to initiate CNS regulation of bone formation. During the early stage of bone healing, divalent cation-induced PGE 2 secretion from macrophages stimulated EP4 in the sensory nerves. Importantly, the sprouting and arborization of sensory nerves in response to macrophage-derived PGE 2 transmitted interoceptive signals to the CNS to tune down sympathetic tone through hypothalamic CREB signaling, resulting in increased osteogenesis and decreased osteoclastogenesis in the injured bone. We revealed a previously unknown role of divalent cations in bone formation through skeletal interoception.

Results
Divalent cation-induced bone formation. To investigate the effect of metallic divalent cations including Mg 2+ , Zn 2+ , or Cu 2+ on bone formation, we used an alginate-based hydrogel to facilitate the temporary and localized delivery of these divalent cations in mouse femurs ( Supplementary Fig. 1a) after evaluating their cytotoxicity in vitro ( Supplementary Fig. 1b). Pure alginate (Alg) or divalent cation-releasing alginate (i.e., Mg-Alg, Cu-Alg, and Zn-Alg) was placed in a tunnel defect drilled from the patellofemoral groove of the distal femur along the axis of the femoral shaft. Micro-computed tomography (µCT) analysis showed a substantial increase in bone volume fraction (BV/TV), bone mineral density (BMD of TV), and polar moment of inertia (ρMOI) in the femur 4 weeks after the placement of divalent cation-releasing alginate compared with the pure alginate-treated control mice (Fig. 1a, b and Supplementary Fig. 2a). Particularly, cortical bone area (Ct.Ar), cortical bone thickness (Ct.T), and bone perimeter (B.Pm) were greater in the divalent cation-treated group compared with the controls (Fig. 1a, b and Supplementary  Fig. 2a). We found a similar trend in the increase of trabecular thickness (Tb.T) in the femurs treated with Mg 2+ or Cu 2+ , although it was less pronounced (Supplementary Fig. 2a). Indeed, hematoxylin and eosin (H&E) staining showed significant new bone formation located primarily at the peripheral cortex in the   divalent cation-treated femurs (Fig. 1c). Moreover, immunofluorescent staining revealed that the number of osteocalcin + (OCN + ) osteoblasts increased significantly on the endocortical and periosteal surfaces of femurs treated with Mg 2+ , Zn 2+ , or Cu 2+ (Fig. 1d, f). Meanwhile, the number of tartrate-resistant acid phosphatase + (TRAP + ) cells on the endocortical and periosteal surfaces decreased significantly in the divalent cationtreated femurs compared with the controls (Fig. 1e, f). In particular, the number of osterix + and Runx2 + osteoprogenitors in the periosteum was also significantly higher after the release of Mg 2+ from the hydrogel ( Supplementary Fig. 2b, c). Furthermore, using fluorochrome labeling, we showed that the release of divalent cations, including Mg 2+ , Zn 2+ , and Cu 2+ , contributed to a significantly higher rate of mineral deposition, as manifested by an increased fluorochrome (i.e., calcein and xylenol) intensity and an increased distance between the two fluorochrome labels (Fig. 1g-i and Supplementary Fig. 2d). We also confirmed that the delivery of divalent cations didn't lead to any histological alteration in the spleen, liver, kidney, and heart tissues at either the early (week 1, Supplementary Fig. 3a) and later stage (week 4, Supplementary Fig. 3b) of the bone healing process. Therefore, our results reveal that divalent cations, including Mg 2+ , Zn 2+ , and Cu 2+ , could stimulate periosteal new bone formation.
To examine the mechanism of divalent cation-induced bone formation, we first characterized the immune response around the divalent cation-releasing alginate, because new bone formation in the periosteum can be evidenced by µCT and H&E staining ( Supplementary Fig. 3a, b) at early inflammatory stage of bone healing (i.e., day 7 after the injury). The majority of immune cells present at this key stage of new bone formation were found to be F4/80 + CD68 + macrophages rather than CD11c + dendritic cells, CD19 + B-cells, or CD3 + T-cells ( Supplementary Fig. 4c). We demonstrated that the number of CD68 + macrophages increased significantly in the bone marrow and periosteum of divalent cation-treated femurs compared with controls ( Fig. 2a, b and Supplementary Fig. 5a). Co-immunostaining of COX2 with CD68 demonstrated that COX2 was expressed primarily in CD68 + macrophages in the periosteum (Fig. 2a), and that the stimulation of divalent cations, including Mg 2+ , Zn 2+ , or Cu 2+ , significantly increased expression of COX2 in CD68 + macrophages in the callus (Fig. 2a, b and Supplementary Fig. 5b). Indeed, the release of Mg 2+ significantly elevated PGE 2 concentration in bone (Fig. 2c) and serum (Fig. 2d) during the early stage (i.e., week 1), when the injured site was undergoing acute inflammation. To examine the mechanism of divalent cationinduced COX2 expression, we isolated and cultured the primary mouse bone marrow macrophages (BMM) with the divalent cations. Mg 2+ , Zn 2+ , and Cu 2+ all promoted secretion of PGE 2 (Fig. 2e). Importantly, when used at a specific level (i.e., 0.1 mM Zn 2+ , 0.1 mM Cu 2+ , or 10 mM Mg 2+ ), the divalent cations upregulated prostaglandin E synthase (PTGES) gene expression (Fig. 2f) and protein levels of COX2 in the BMM (Fig. 2g) without affecting cell viability ( Supplementary Fig. 1b). We also verified these findings by sorting macrophages in LysM-YFP mice for RT-qPCR assay after the placement of divalent cation-releasing alginate. Our data showed Mg-Alg, Cu-Alg, and Zn-Alg all contributed to significantly upregulated expression of PTGES and COX2 (Fig. 2h). Furthermore, divalent cations led to the phosphorylation of nuclear factor-κB (NF-κB) p65 and inhibitor of nuclear factor-κB (IκBα), the key cascade proteins in the NF-κB signaling pathway (Fig. 2i). Indeed, Mg 2+ significantly increased the binding of NF-κB p65 to the COX2 gene promoter as shown in chromatin immunoprecipitation (ChIP) assay (Fig. 2j). Thus, we showed that divalent cations stimulate the secretion of PGE 2 from macrophages during the inflammation stage of the bone healing process.
Ablation of COX2 in macrophages eliminates periosteal bone formation. To validate the mechanism in vivo, we established an inducible macrophage-depleted mouse model by crossing LysM-Cre mice with iDTR wt mice. Macrophage depletion was effectively achieved in iDTR LysM +/− mice by injecting diphtheria toxin (DTX). Macrophage depletion diminished the effect of Mg 2+ on bone healing, as there was no significant difference between the Mg-Alg group and the control group in bone formation on µCT (Fig. 3a, b and Supplementary Fig. 6a), whereas, Mg 2+ -releasing hydrogel effectively promoted new bone formation after bone injury in LysM-Cre mice, similar to its effect in WT mice. Interestingly, when bone PGE 2 level was elevated with injection of SW03329 (a PGE 2 degradation enzyme inhibitor), the new bone formation in the injured femur improved significantly in both the control group and the Mg-Alg group (Fig. 3c, d and Supplementary Fig. 6b). Specifically, new bone formation at the peripheral cortex of the injured femur with injection of SW033291 in the control group was similar in the Mg 2+ -treated group on H&E staining ( Supplementary Fig. 6c). Moreover, the numbers of OCN + osteoblasts and TRAP + osteoclasts on the endocortical and periosteal surfaces of femurs were similar regardless of the presence of Mg 2+ in the alginate with injection of SW033291 ( Supplementary Fig. 6d, e). We further generated mice with conditional COX2 knockout in the macrophages (COX2 LysM −/− ) by crossing COX2 wt mice with LysM-Cre mice. Mg 2+ -releasing hydrogel failed to induce thickening of cortical bone in injured femurs of COX2 LysM −/− mice compared with COX2 wt mice (Fig. 3e, f and Supplementary  Fig. 7a). Mg 2+ -induced new bone formation at the peripheral cortex of femurs in COX2 wt mice was not seen in COX2 LysM −/− mice on H&E staining (Fig. 3g). Again, the effect of Mg 2+ on OCN + osteoblasts and TRAP + osteoclasts on both endocortical and periosteal surfaces was diminished in COX2 LysM −/− mice (Fig. 3g, i). These data show that COX2 in macrophages is responsible for PGE 2 production for divalent cation-induced bone formation.
Sensory nerves are essential for divalent cation-induced bone formation. Given that PGE 2 activates EP4 in the sensory nerve to maintain skeletal interoception activity, we examined whether sensory nerves are associated with divalent cation-induced bone formation. Prominent sprouting and arborization of calcitonin gene-related polypeptide-α + (CGRP + ) sensory nerve fibers coursing longitudinally over the outer periphery of the reactive callus were observed in immunostaining of periosteum (Fig. 4a). The intensity of CGRP + nerve fibers in the periosteum was significantly increased with divalent cations (Fig. 4b). Importantly, terminal dendrites of CGRP + sensory nerves were spatially associated with CD68 + macrophages in the reactive periosteum (Fig. 4c). Moreover, the expression of CGRP was significantly increased in the ipsilateral DRG of mice treated with divalent cations (Fig. 4d-f). Interestingly, CGRP + dendrite sprouting observed primarily around COX2 + macrophages in the periosteum of COX2 wt mice treated with divalent cations was missing from COX2 LysM −/− mice ( Supplementary Fig. 7b). We then created a sensory denervation mouse model (TrkA Avil −/− ) by crossing sensory nerve-specific Cre (Advillin-Cre) mice with nerve growth factor receptor tropomyosin receptor kinase A (TrkA) floxed (TrkA wt ) mice to confirm the essential role of sensory nerve in divalent cation-induced bone formation. Indeed, Mg 2+ significantly increased the density of CGRP + sensory nerve fibers in the bone marrow cavity and the endocortical and periosteal surfaces of TrkA wt mice 7 days after surgery, whereas the number of CGRP + sensory nerve fibers in TrkA Avil −/− mice was significantly lower relative to WT mice (Fig. 4g). Similarly, Mg 2+ stimulated bone formation in TrkA wt mice, and such effects were eliminated in TrkA Avil −/− mice (Fig. 4h, i and Supplementary  Fig. 7c). Taken together, divalent cation-induced sensory innervation in injured bone tissue is essential for new bone formation.
Divalent cations stimulate PGE 2 /EP4 skeleton interoception in downregulation of sympathetic activity. We have shown that PGE 2 activates sensory nerves via EP4 to induce phosphorylation of CREB in the VMH of the hypothalamus as a skeletal interoception pathway 22,46 . We, therefore, investigated whether local delivery of divalent cations could trigger CREB phosphorylation in the hypothalamus. Immunostaining of brain sections showed that phosphorylation of CREB was significantly increased in the contralateral VMH of mice at 1 week after the delivery of Mg 2+ , Zn 2+ , or Cu 2+ in the injured femurs (Fig. 5a, b). CREB phosphorylation in the hypothalamus with upregulation in 5-hydroxytryptamine receptor 2C (HTR2C) was confirmed by Western blot analysis (Fig. 5c, d). Moreover, using enzyme-linked immunosorbent assay (ELISA), we showed that Mg 2+ significantly reduced epinephrine levels in serum and urine compared with controls, indicating suppression of sympathetic tone in Mg-Alginjected mice (Fig. 5e). To examine whether divalent cationinduced CREB phosphorylation downregulates tyrosine Ct t Ct Ct P P P P P P P P P P P P P P P P P P P P P hydroxylase (TH) for sympathetic activity, we performed immunostaining of TH for sympathetic nerves. The density of TH + sympathetic nerve fibers on the periosteal surface of divalent cation-treated femurs was significantly less in mice injected with divalent cation-releasing alginate compared with the controls (Fig. 5f). In parallel, spontaneous activity of mice, an indicator of sympathetic tone and postoperative pain 26 , was assessed by spontaneous activity wheels. Compared with controls, the release of Mg 2+ contributed to significantly shorter daily distance and duration, as well as lower maximum speed of running-wheel activity at weeks 1 and 4 after surgery (Fig. 5g). Finally, we examined the effect of divalent cations on the activation of skeletal interoception in TrkA Avil −/− and COX2 LysM −/− mice, as well as their WT littermates. With sensory nerve denervation, the effect of Mg 2+ on phosphorylation of CREB was diminished compared with their WT littermates (Fig. 5h). Daily distance and duration of running-wheel activity in TrkA Avil −/− mice were significantly longer than in their WT littermates (TrkA wt ), whereas such effect of Mg 2+ on decreasing spontaneous activity was abolished in TrkA Avil −/− mice (Fig. 5i). Additionally, the increase of hypothalamic CREB phosphorylation triggered by Mg-Alg was abolished in COX2 LysM −/− mice (Fig. 5j). Similarly, the effect of Mg 2+ on decreasing the daily distance and duration of running-wheel activity was diminished in COX2 LysM −/− mice (Fig. 5k). Taken together, our data show that divalent cations promote PGE 2 from macrophages to activate skeletal interoception, resulting in downregulation of sympathetic tone and new bone formation.
Knockout of EP4 in the sensory nerve inhibits divalent cationinduced activation of skeletal interoception. EP4 is the PGE 2 receptor in the skeletal interoception and is known as the primary receptor in bone remodeling and homeostasis 47 . We first prepared conditional medium from Mg 2+ -treated macrophages to The divalent cations triggered the production of PGE 2 from macrophages, which activated the EP4 at the sensory nerve to tune down sympathetic tones via the CREB signaling in the VMH, resulting in increased osteogenesis and decreased osteoclastogenesis in the periosteum. Data are mean ± s.d. *P < 0.05, **P < 0.01 by one-way ANOVA with Tukey's post hoc test (a, c, e, h, j). Source data are provided as a source data file. test its effect on EP4 expression. EP4 expression in primary DRG neurons was increased more than 10-fold relative to treatment with Mg 2+ alone (Fig. 6a). We then generated EP4 Avil −/− mice by crossing Advillin-Cre mice with EP4 wt mice. Mg 2+ -induced bone formation was abolished in EP4 Avil −/− mice with induction of EP4 ablation in the sensory nerves, as shown by µCT (Fig. 6b, c and Supplementary Fig. 7d). We confirmed with H&E staining that the new bone formation induced by Mg 2+ at the peripheral cortex was diminished with conditional knockout of EP4 in sensory nerve fibers (Fig. 6d). Moreover, the increase of osteoblasts in the periosteum and decrease of osteoclasts were diminished in EP4 Avil −/− mice after postoperative administration of Mg 2+ (Supplementary Fig. 8a, b). Again, as in COX2 ablation mice, the effect of Mg-Alg on the daily distance and duration of running-wheel activity was also abolished in EP4 Avil −/− mice (Fig. 6e). Importantly, phosphorylation of CREB in the hypothalamus was not detectable in EP4 Avil −/− mice treated with Mg-Alg, unlike their EP4 Avil −/− littermates (Fig. 6f).
To validate the hypothesis that downregulation of sympathetic activity mediates Mg 2+ -induced cortical bone thickening, we injected mice with propranolol, a widely used β2-adrenergic antagonist. Propranolol contributed to significantly more periosteal new bone formation in Mg 2+ -Alg-injected femurs compared with the vehicle group, whereas it failed to increase bone formation in Alg-treated femurs relative to the vehicle group (Fig. 6g, h and Supplementary Fig. 8c). The bone that was newly formed in response to the stimulation of Mg 2+ after the administration of propranolol was located primarily at the peripheral cortex, as was previously observed in WT mice (Fig. 6i). Moreover, mice injected with propranolol had a shorter duration of running-wheel activity regardless of Alg or Mg-Alg treatment (Fig. 6j). Interestingly, the activation of CREB phosphorylation observed in the contralateral VMH was not prominent after the injection of SW033291 ( Supplementary  Fig. 8d), suggesting that the local stimulation caused by the Mg 2+ -induced release of PGE 2 from macrophages was largely masked by the systematic administration of SW033291. Similarly, the decrease in spontaneous activity caused by Mg-Alg relative to Alg was diminished as the distance and duration of runningwheel activity in Alg treated mice were lower after the injection of SW033291 ( Supplementary Fig. 8e). Therefore, EP4 in the sensory nerves mediates the activation of skeletal interoception to downregulate sympathetic activity, leading to new bone formation in the injured site.

Discussion
In recent years, rapid progress has been made in the development of novel biodegradable metal implants 10,48 . With controlled degradation kinetics and gradual integration with bone tissue, biodegradable metal implants have been demonstrated to be superior to traditionally used bioinert metal implants for the treatment of musculoskeletal injuries. Particularly, these biodegradable implants can induce new bone formation through the release of various divalent metal cations 10,48 . Moreover, given the promising osteogenic properties of these divalent metal cations, they are also widely used for the modification of various kinds of orthopedic biomaterials 8,41,49 . Indeed, compared with expensive biological therapeutic agents and complex surgical procedures, the intervention using divalent cations appears to be a costeffective way to achieve bone regeneration. However, it remains largely unclear how divalent cations induce new bone formation and whether these divalent cation-releasing biomaterials also induce bone formation through similar molecular and cellular mechanisms. In this study, we discovered that the divalent cations, Mg 2+ , Zn 2+ , and Cu 2+ serve as interoceptive signals to initiate CNS regulation of new bone formation after injury, which suggests the complexity of the underlying mechanism for divalent cation-induced bone regeneration has been greatly underestimated. Therefore, we believe that a clear understanding of the mechanism through which these divalent cations contribute to new bone formation through interoception will greatly benefit the development of novel biomaterials to fully elicit the therapeutic potential of these divalent cations.
The immune system and the nervous system are paramount to sense and respond to changes in our body, as they possess unique qualities that enable them to counter deviations in the internal environment 16 . Upon the delivery of divalent cations, which disrupts the homeostasis in bone microenvironment, the acute immune response contributes to a series of inflammatory chemokines and cytokines that relay information to the nervous system to initiate interoceptive control of bone formation. As a hallmark of inflammatory response after bone injury, COX2/ PGE 2 plays a key role in cortical bone repair 33,36,37 . Subcutaneous administration of PGE 2 contributes to new bone formation on the endocortical and periosteal surfaces of ovariectomized and intact rats 50 , whereas global knockout of COX2 almost eliminates of periosteal bone formation during bone healing 51 . In this study, we showed divalent cations, including Mg 2+ , Zn 2+ , and Cu 2+ , significantly increased PGE 2 production from CD68 + macrophages in the reactive periosteum. Interestingly, PGE 2 activated EP4 signaling in the CGRP + sensory nerve endings as the biochemical interoceptive signal. Moreover, our data showed that the injection of SW03329 elevated PGE 2 concentration in the control group to support cortical bone formation but did not further significantly promote Mg 2+ -induced periosteal new bone formation. This implies that PGE 2 -EP4 signaling in skeletal interoception is not entirely concentration-dependent, and there might be a specific window of PGE 2 levels for the activation of the immune-neural axis during bone healing-if so, this warrants further investigation. This observation suggests that secretion of PGE 2 specifically by macrophages stimulates divalent cationsinduced bone formation through skeleton interoception.
Although the immunomodulatory effects of divalent cations have been reported, the inflammatory responses induced by these divalent cations don't always produce specific inflammatory cytokines (e.g., PGE 2 ) to facilitate new bone formation [52][53][54][55] . Instead, the immunomodulatory effects of these divalent cations are highly concentration-dependent and tissue-specific. Moreover, as bone tissue is the major reservoir for minerals in human body, the inflammation in this microenvironment induced by high levels of divalent cations may initiate the deposition of these trace minerals in hard tissue to alleviate their potential long-term toxicity. Therefore, the storage exogenous divalent cations in newly formed bone tissue could be an effective and efficient strategy to maintain the physiological condition in bone microenvironment. In this study, by utilizing the different cross-linking potential of Mg 2+ , Zn 2+ , and Cu 2+ on alginate 56 , Mg-Alg, Zn-Alg, and Cu-Alg were designed exquisitely to harness the immunomodulatory effects of these divalent cations on macrophages. For instance, Zn 2+ and Cu 2+ released at a very low level (i.e., approximately 0.1 mM), which is also seen in either zinc or copper implants with slow degradation rate, produced PGE 2 without eliciting cytotoxicity. In comparison, 10 mM of Mg 2+ released from Mg-alginate, similar to that from degradable magnesium-based implants 57 , stimulated the production of PGE 2 from macrophages, though Mg 2+ has been recognized as an antiinflammatory agent due to its suppressive effects on proinflammatory molecules 58,59 . The immunomodulatory effects of the divalent cations were also shown manifested by the activation of the NF-κB signaling pathway in BMM. Given that Mg 2+ promotes the nuclear translocation of NF-κB p65 in binding to COX2 promoter, it is anticipated that the NF-κB signaling pathway, which is crucial in inflammation responses 60 , may play a central role in divalent cation-induced regulation of PGE 2 . Although many of the inflammatory cytokines, such as IL-1β, IL-6, and TNF-α, can be recognized by their receptors expressed in the sensory nerve to modulate the nervous system 61 , PGE 2 , among all inflammatory mediators, tends to receive the most attention due to its role in mediating peripheral pain pathway 62 . By using a primary culture of DRG neurons and the mice with conditional knockout COX2 in macrophage, we demonstrated that the activation of sensory neurons is actually triggered by the macrophage-derived PGE 2 in response to the stimulation of divalent cations. Moreover, the ablation of EP4 in sensory nerves abolished the osteogenic effect of Mg 2+ in the periosteum, thus showing that the crosstalk between immune and neural system is essential in skeletal interoception-mediated new bone formation.
The density of sensory and sympathetic fibers in the periosteum remains the highest in the skeletal system 21 . The periosteum covers almost the entire bone surface and is one of the most regenerative tissues for skeletal osteogenesis 63 . Damage of the periosteum severely impairs cortical bone homeostasis 64,65 and bone fracture healing 10,65 . Moreover, an increase in the density of sensory and sympathetic nerve fibers in injured bone tissue has been reported to contribute to the bone healing process 45,66,67 . The sprouting and arborization of sensory nerve fibers in the periosteum in response to bone injury 45 or mechanical stimulation 68 may facilitate the sensation of stimuli and the initiation of osteogenesis. CGRP + sensory nerves that emanate from the DRG of the spinal cord detect multiple stimuli in bone tissues (including inflammatory cytokines) and process and relay these signals to higher CNS levels 69 . In addition to its major role as a neurotransmitter and neuromodulator, CGRP is also considered a peptide that can be released from the peripheral nerve terminals to regulate osteoclast and osteoblasts 70,71 , which was recently reported to be implicated in the new bone formation induced by pure magnesium implant 10 . In our study, a conspicuous increase in the number of CGRP + nerve fibers in divalent cation-treated femurs suggests an association between the activation of sensory afferents and the inflammatory microenvironment, which is indispensable in skeleton interoceptive circuit. Therefore, the upregulation of CGRP in response to divalent cations may possess multiple functions, including the nociceptive transmission contributing to central sensitization and the direct control of bone cells through the receptors they shared. The spiral-like TH + sympathetic nerve fibers typically wrap around the major blood vessels in the periosteum and penetrate the cortical bone via Volkmann's and Haversian canals 72,73 . They are known to be capable of regulating osteoblast activity through β2-adrenergic receptors (β2AR) 13,74 . Meanwhile, by increasing the secretion of the receptor activator of nuclear factor kappa-B ligand (Rankl) from osteoblasts, the sympathetic nervous system can also stimulate osteoclastic bone resorption 12,75 . We found that PGE 2 produced by divalent cation-stimulated macrophages tunes down sympathetic activity for the differentiation of OCN + osteoblasts on the endocortical and periosteal surfaces. Moreover, in both the macrophage depletion and sensory denervation models, elevated PGE 2 in the periosteum failed to trigger the activation of CREB signaling in the VMH or periosteal new bone formation, indicating that the macrophage-mediated immune response and the sensation of afferent nerves are indispensable in the CNS recognition of stimulation from Mg 2+ , Zn 2+ , or Cu 2+ after bone injury. Importantly, CREB signaling in the ipsilateral DRG and the contralateral VMH by periosteal delivery of divalent cations downregulates sympathetic activity as a precise temporalspatial feedback to the injured site for bone regeneration, as the bone homeostasis of the contralateral femur was not altered after the activation of skeletal interoception (Supplementary Fig. 9a-c).
The discovery of the skeleton interoception-mediated new bone formation triggered by Mg 2+ , Zn 2+ , and Cu 2+ indicates that the complexity of the underlying mechanism for the osteogenic effects of these divalent cations has been greatly underestimated. Nevertheless, it is important to note that there have been multiple approaches reported in the past few decades through which these divalent cations may modulate bone homeostasis [6][7][8][9] . And this may explain the difference among the three tested divalent cations in term of their bone formation outcome. For instance, in our study, the highest bone volume was observed in the Cu-Alg group, as the old bone was barely resorbed from the endosteal surface by osteoclasts following the new bone formation on periosteal surface. In fact, from as early as 1981, Cu 2+ has been reported to have a direct dose-dependent inhibitory effect on osteoclastic activity 76 . Cu 2+ has also been used in biomaterials to shift the equilibrium between bone formation and bone resorption because the tolerance of osteoclasts and osteoblasts to Cu 2+ differs 77,78 . In addition to the bone cells, some divalent cations like Mg 2+ and Zn 2+ have also been shown to facilitate type H vessel formation by targeting endothelial cells, which couples angiogenesis with osteogenesis during bone healing 6 . This subtype of vessels, characterized by the coexpression of CD31 and endomucin, can provide niche signals for perivascular osteoprogenitors to promote osteogenesis 79,80 . Recent studies showed the sensory innervation following bone injury is an essential upstream mediator for vasculature 45 , and electrical stimulation at DRG could promote type-H vessel formation to enhance bone regeneration 81 . Therefore, it would be interesting to further explore to which extend are angiogenesis and vasculogenesis involved in skeleton interoception and what their specific role may be in divalent cation-induced new bone formation.
In summary, we found that skeletal interoception mediates divalent cation-induced bone formation (Fig. 6K). The controlled delivery of Mg 2+ , Zn 2+ , or Cu 2+ stimulates to the production of PGE 2 from CD68 + macrophages, which is followed by rapid reinnervation of CGRP + sensory nerve fibers in the reactive periosteum. Activation of EP4 receptor in sensory nerves by local PGE 2 induces CREB signaling in the VMH as an ascending interoceptive signal, which downregulates sympathetic activity as a descending interoceptive signal to induce osteogenesis at the injured bone treated with divalent cations. The discovery of divalent cation-induced bone formation through skeletal interoception could revolutionize the current understanding of bone regeneration and inspire innovative orthopedic biomaterials for bone tissue engineering. The offspring were intercrossed to generate the following genotypes: wild type (referred to as WT in the text), Avil-Cre (Cre recombinase expressed driven by advillin promoter), TrkA fl/fl (mice homozygous for TrkA flox allele, referred to as TrkA wt in the text), EP4 fl/fl (mice homozygous for EP4 flox allele, referred to as EP4 wt in the text), Avil-Cre::TrkA fl/fl (conditional deletion of TrkA receptor in Advillin lineage cells, referred to as TrkA Avil −/− in the text), Avil-Cre::EP4 fl/fl (conditional deletion of EP4 receptor in Advillin lineage cells, referred to as EP4 Avil −/− in the text). Heterozygous male LysM-Cre mice were crossed with a female iDTR fl/fl mouse or a COX2 fl/fl mouse. The offspring were intercrossed to generate the following genotypes: WT, LysM-Cre, iDTR fl/fl , COX2 fl/fl mice (mice homozygous for COX2 flox allele, referred to as COX2 wt in the text), LysM-Cre::iDTR fl/-(referred to as iDTR LysM +/− in the text), LysM-Cre::COX2 fl/fl (conditional deletion of COX2 in monocyte-macrophage lineage, referred to as COX2 LysM −/− in the text). Homozygous male LysM-Cre mice were crossed with a female Rosa26YFP mouse to generate LysM-YFP mice.

Methods
The genotypes of the mice were determined by polymerase chain reaction (PCR) analyses of the genomic DNA, which was extracted from mouse tails. The primers used for genotyping were Avil-Cre: In vivo treatment. Twenty percent alginate gel (Alg) was prepared by mixing sodium alginate powder (Sigma-Aldrich, St. Louis, MO, USA,180947) in deionized water, while 10% magnesium chloride (MgCl 2 , Sigma-Aldrich, M8266), 10% zinc chloride (ZnCl 2 , Sigma-Aldrich, 208086), and 10% copper chloride (CuCl 2 , Sigma-Aldrich, 307483) were used for the preparation of magnesium cross-linked alginate (Mg-Alg), zinc cross-linked alginate (Zn-Alg), and copper cross-linked alginate (Cu-Alg). Twelve-week-old male mice were anesthetized by intraperitoneal injection with ketamine (Vetalar, Ketaset, Ketalar; 75 mg/kg, intraperitoneally) and xylazine (Rompun, Sedazine, AnaSed; 10 mg/kg, intraperitoneally). A longitudinal incision was made at the left knee and the patella was dislocated to expose the femoral condyle. Using a 20-gauge needle (Becton, Dickinson and Company, BD Syringe, 309644), we created a tunnel with a diameter of 1 mm from the patellofemoral groove of the distal femur along the axis of the femoral shaft. After thorough irrigation with saline (Quality Biological Inc, 114055101), either pure alginate or divalent cation-releasing alginate (0.01 mL) was injected into the femoral canal. Meanwhile, a same tunnel defect was created in the right femur, but no material was injected. The wounds were sutured layer-by-layer, and the mice were housed in a specific-pathogen-free facility after the surgery. A monocytemacrophage lineage depletion mouse model was induced by intraperitoneal injection of diphtheria toxin (DTX, Sigma-Aldrich, D0564) in iDTR LysM +/− mice every other day during the week before and the week after the surgery. A PGE 2 degradation enzyme inhibitor, SW033291 (Selleck Chemicals, Houston, TX, USA), was administered by intraperitoneal injection at 10 mg/kg every day for 1 week after the surgery. A low-dose β-adrenergic receptor blocker, propranolol (PROP, Sigma-Aldrich, 1576005), was administered by intraperitoneal injection at 0.5 mg/ kg every day for 1 week after the surgery.
µCT analysis. At designated time points, the femurs were harvested from the mice and fixed overnight using 4% paraformaldehyde. Analysis was performed using a high-resolution µCT scanner (SkyScan 1275, Bruker, Kontich, Belgium). The voltage of the scanning procedure was 65 kv with a 153-μA current. The resolution was set to 8.7 μm per pixel. Two phantom-contained rods with a standard density of 0.25 and 0.75 g/cm 3 were scanned with each sample for calibration. Data reconstruction was completed using NRecon software (v1.6, SkyScan), data analysis was accomplished using CTAn software (v1.9, SkyScan), and 3D model visualization was performed using CTvox software (v3.2, SkyScan). Bone volume fraction (BV/TV), bone mineral density (BMD of TV), trabecular number (Tb. N), trabecular thickness (Tb. Th), cortical thickness (Ct. Th), cortical area (Ct. Ar), bone perimeter (B. Pm), and ρ-moment of inertia (ρMOI) were measured via µCT data.
Fluorochrome labeling. Two fluorochrome labels were used sequentially to evaluate periosteal new bone formation rate after the placement of divalent cation releasing alginate. In brief, calcein green (5 mg/kg, Sigma-Aldrich, C0875) was subcutaneously injected into mouse femora one week after the surgery, while xylenol orange (90 mg/kg, Sigma-Aldrich, 52097) was injected two weeks after the surgery. The fluorochrome labels were visualized under a fluorescence microscopy (Niko ECL IPSE 80i, Japan). The intensity of fluorescence and the distance between two labeling were measured and analyzed by ImageJ software (NIH, USA).
Immunofluorescence and histomorphometric analysis. At designated time points, mice femurs were fixed with 4% paraformaldehyde (Fisher Scientific, SF100-4) overnight and decalcified with 10% ethylenediaminetetraacetic acid (pH = 7.4) for 21 days. For immunostaining, the samples were dehydrated in 20% sucrose solution with 2% polyvinylpyrrolidone (PVP, Sigma-Aldrich, PVP40) for 24 h and embedded in 8% gelatin (Sigma-Aldrich, 1288485) supplemented with 20% sucrose (Fisher Scientific, S5-500) and 2% PVP. 40 μm-thick coronal-oriented sections of the femurs were obtained using a cryostat microtome. For histomorphometry, the samples were dehydrated in ethanol, embedded in paraffin, and prepared into 5 µm-thick coronal-oriented sections using a rotary microtome. The brain and DRG tissues harvested from the mice were fixed with 4% paraformaldehyde, dehydrated with 30% sucrose, and embedded in an optimal cutting temperature compound (OCT, Sakura Finetek, Torrance, CA, USA, 4583). 10 μmthick coronal-oriented sections of the brain and DRG were obtained using a cryostat microtome.
Cell sorting. All cell sorting was performed on a BD FACSAria TM SORP Cell Sorter (BD Biosciences, USA). At week 1 after the placement of divalent cationreleasing alginate, the femurs were harvested and rinsed with 1X PBS (Thermo Fisher Scientific, 10010023). After the removal of muscle tissue, bone marrow was flushed out with 1X PBS. The cells on the cortical bone were isolated by digesting the crushed bone chips in PBS containing 1 mg/mL Collagenase I (Sigma-Aldrich, SCR103) for 20 min at 37°C. Cells were thoroughly washed, underwent red blood cell lysis (Invitrogen, Thermo Fisher Scientific, 00-4333-57), and resuspended in PBS containing 2% fetal bovine serum (FBS, Invitrogen, Thermo Fisher Scientific, 26140079). For cell sorting, YFP positive cells were sorted into chilled 15 mL centrifuge tubes containing PBS supplemented with 2% FBS. The harvested cells were washed and centrifuged before the addition of RNeasy lysis buffer (Qiagen, Germantown, MD, USA, 79216) for the extraction of total RNA. Cell culture. Primary BMM and DRG neurons from 4-week-old mice were isolated. For the isolation of BMM, the mice were euthanized and both femurs were dissected to remove soft tissue. The femurs were then crushed into pieces and digested with α-MEM (Minimum Essential Medium Eagle α Modification, Fisher Scientific, MT10022CV) containing 3 mg/mL Collagenase I (Worthington Biochemical Corp, Lakewood, NJ, USA, LS004194), 4 mg/mL dispase (Sigma-Aldrich, D4818), and 1 U/mL deoxyribonuclease-I (Invitrogen, Thermo Fisher Scientific, 18047019). The single-cell suspension was achieved by passing the solution through a cell strainer to remove tissue fragments. After incubation in a humidified incubator with 5% CO 2 at 37°C, the non-adherent cells were harvested and cultured in α-MEM supplemented with 20 ng/mL macrophage colony-stimulating factor (M-CSF, R&D Systems, Minneapolis, MN, USA, 416-ML) for 7 days. For the isolation of DRG neurons, DRGs from the L2-L5 spinal levels were isolated in cold DMEM/F12 (Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12) medium (Invitrogen, Thermo Fisher Scientific, 11320033) and then digested with 1 mg/mL collagenase type A (Roche, Basel, Switzerland, 05349907103) at 37°C. After trituration and centrifugation, cells were resuspended and seeded on glass coverslips coated with ploy-D-lysine and laminin. The culture medium was replaced 6 h after seeding, and the adherent cells were further cultured at 37°C with 5% CO 2 for 3 days before use.
ELISA. Whole blood samples were collected by cardiac puncture immediately after the mice were euthanized. Serum was collected by centrifuging at 2000 × g for 15 min and then stored at −80°C before analysis. The total bone protein was harvested from the femurs grafted with pure alginate or Mg-Alg. The mid-shaft of the femur, approximately 1 cm long, was ground into mud using a ceramic mortar and pestle under cooling. The mud of the bone tissue was then homogenized in pre-cooled radioimmunoprecipitation assay (RIPA) lysis and extraction buffer (Thermo Fisher Scientific, 89900) for 1 h. The buffer solution was centrifuged at 15,000 × g for 20 min at 4°C. The supernatant was collected for protein concentration quantification with the BCA Protein Assay Kit (Thermo Fisher Scientific, A53225). An equal amount of protein from each sample was subjected to quantitative analysis using a specific ELISA kit per the manufacturer's instruction. The PGE 2 concentrations in the serum and bone marrow were determined by the PGE 2 ELISA kit (Cayman Chemical, Ann Arbor, MI, USA, 514010). The OCN level was determined by an OCN ELISA kit (Biomedical Technologies Inc, Tewksbury, MA, USA, BT470). The serum and urine epinephrine levels were determined by an epinephrine ELISA kit (ALPCO, Salem, NH, USA, 17-EPIHU-E01.1).
Quantitative real-time polymerase chain reaction (qPCR). The total RNA of the cells was extracted and purified using the RNeasy Plus kit (Qiagen, Germantown, MD, USA, 74034) per the manufacturer's instructions. For the reverse transcription, complementary DNA was synthesized using the SuperScript First-Strand Synthesis System (Invitrogen, Thermo Fisher Scientific, 18091050). The primers used in the RT-qPCR assay were synthesized by Life Technologies (Thermo Fisher Scientific) based on sequences retrieved from Primer Bank (http:// pga.mgh.harvard.edu/primerbank, Supplementary Table 1). SYBR Green-Master Mix (Qiagen, Germantown, MD, USA, A46113) was used for the amplification and detection of complementary DNA on a C1000 Thermal Cycler (Bio-Rad Laboratories, Hercules, CA). The mean cycle threshold (Ct) value of each target gene was normalized to the housekeeping gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The results were shown as a fold change using the ΔΔCt method.
Behavioral analysis. The spontaneous activity of mice after surgery was assessed using spontaneous activity wheels (BIO-ACTIVW-M, Bioseb, Boulogne, France). Mice were housed in polycarbonate cages with free access to stainless steel activity wheels (diameter 23 cm; width 5 cm), which were connected to an analyzer that automatically recorded the distance traveled, mean speed, maximum speed, and total active time. The mice had ad libitum access to food and water during the test. They were allowed to acclimatize to the environment for at least 24 h before data were recorded.
Chromatin immunoprecipitation (ChIP) assay. After the stimulation, the ChIP assay was performed using an Agarose ChIP Kit (Thermo Fisher Scientific, 26156) according to the manufacturer's instructions. In brief, the chromatin was crosslinked by 1% formaldehyde and digested by micrococcal nuclease. The lysate was incubated with rabbit anti-NF-κB p65 at 4°C overnight followed by incubation with ChIP Grade Protein A/G Plus Agarose. The purified DNA was analyzed by PCR assay using primers targeting mouse COX2 promoters: sense 5′-CCCGGAG GGTAGTTCCATGAAAGACTTCAAC-3′ and antisense 5′-GGTGGAGCTGGC AGGATGCAGTCCTG-3′. The primers targeting the GAPDH promoter served as a positive control. PCR products obtained after 40 cycles were separated on 2% agarose gels.
Statistical analysis. All data analyses were performed and illustrated using Prism software (v. 7, GraphPad Software, San Diego, CA, USA). Data are presented as means ± standard deviations (SD). For comparisons between two groups, twotailed Student's T-tests were used. For comparisons among multiple groups, oneway or two-way analysis of variance (ANOVA) was used, followed by Tukey's post hoc test. Significant differences among groups were defined and noted as *P < 0.05 or **P < 0.01. The sample size was based on preliminary data, as well as on observed effect sizes.