Skeletal Interoception in Bone Homeostasis and Pain

Abstract

Accumulating evidence indicates that interoception maintains proper physiological status and orchestrates metabolic homeostasis by regulating feeding behaviors, glucose balance and lipid metabolism. Continuous skeletal remodeling consumes a tremendous amount of energy to provide skeletal scaffolding, support muscle movement, store vital minerals and maintain a niche for hematopoiesis, which are processes that also contribute to overall metabolic balance. While skeletal innervation has been described for centuries, recent work has shown that skeletal metabolism is tightly regulated by the nervous system and that skeletal interoception regulates bone homeostasis. Here, we provide a general discussion of interoception and its effects on the skeleton and whole-body metabolism. We also discuss skeletal interoception-mediated regulation in the context of pathological conditions and skeletal pain, as well as future challenges to our understanding of these process and how they can be leveraged for more effective therapy.

Interoception regulates internal organ physiological activity and maintains body metabolic homeostasis. Skeletal interoception maintains the homeostasis of constant bone remodeling precisely and skeleton metabolic homeostasis. In this review, Lv et al. introduce the concept of interoception and summarize current research progress on skeletal interoception in metabolism and skeletal pain.

Introduction

Our brain receives external environmental information through the various senses, including vision, hearing, taste, touch, pressure and temperature. Such sensing and its response are referred to as exteroception, which is crucial for reflexive and adaptive behaviors (Koch et al., 2018). Our brain also perceives internal organ physiological and metabolic states, such as blood pressure variation, gut motility, energy stores and visceral pain, which is then integrated to maintain whole-body homeostasis. This process is referred to as interoception (Quadt et al., 2018). Both exteroceptive and interoceptive signals are processed and interpreted by the brain to respond to the constant changes in the external and internal environments (Park and Blanke, 2019).

The skeletal system provides the support needed to be a terrestrial organism and to promote locomotion (Doherty et al., 2015). As such, the skeletal system is one of the largest organs, accounting for approximately 20% of our total body weight (Leider, 1947). And while the role of the central nervous system (CNS) in the regulation of bone has been established for decades, the regulation of skeletal interoception is just beginning to be deciphered (Chen et al., 2019; Ducy et al., 2000; Hu et al., 2020; Lv et al., 2021). It is clear that the skeletal system bears the forces of gravity, which is likely perceived through interoception of the skeleton by the brain. Importantly, bone consists of a large quantity of matrix proteins, with deposition of calcium, phosphate, and many different minerals as a reserve for the body (Riddle and Clemens, 2017). Bone also serves as an endocrine organ for the regulation of whole-body metabolism (Karsenty and Oury, 2012; Lee et al., 2007; Zhou et al., 2021). Further, the bone marrow provides the niche for both mesenchymal stem cells and hematopoietic stem cells (Méndez-Ferrer et al., 2010; Morrison and Scadden, 2014). Therefore, it is imperative to better understand CNS-mediated regulation of the bone; particularly, the regulation of bone homeostasis, bioenergetic metabolism, and the maturation of marrow stem cells and immune cells. In addition, skeletal aging is associated with pain and many major disorders, including osteoporosis, arthritis, and spine degeneration, along with accelerated aging of different organs, including the brain (Foster et al., 2018; Mantyh, 2014; Minghetti, 2004). Uncovering the mechanism(s) of skeletal interoception may provide a potential therapy for skeletal pain and effective treatments for different bone diseases. In this Review, we will introduce the concept of interoception and then highlight the progress of current research into skeletal interoception. Furthermore, we will uncover the role of skeletal interoception in the regulation of bone metabolism, the mediation of bone modeling and remodeling by mechanical load, bone-related pain and metal-induced bone formation. Finally, we will offer a prospective on skeletal interoception research and its application to potential treatments of bone diseases.

The concept of interoception

In general, interoception is defined as the representation and monitoring of an organism’s internal state to regulate the complex interactions between the brain and peripheral organs (Craig, 2002; Sherrington, 1906). But this definition has evolved over the decades to reflect a more dynamic representation of the organism’s physiology, including the process by which the CNS senses, integrates and regulates signals related to the the inner state of the body (Khalsa et al., 2018). The key components of interoception comprise interoceptors that sense peripheral interoceptive signals and then communicate these signals via the ascending pathways to the CNS. The CNS, in turn, then integrates and interprets this information before relaying a response back through descending pathways to the periphery to regulate its homeostasis (Figure 1) (Chen et al., 2021; Quadt et al., 2018). Additionally, the endocrine and immune systems have been involved in interoception (Rinaman, 2007; Salvador et al., 2021).

Figure 1. Diagram of current known interoception circuits.

Biochemical, mechanical, thermal and electromagnetic signals are the three major types of interoceptive signals. Interoceptors, located at peripheral sensory nerve endings, can detect interoceptive signals generated by the peripheral organs. Interoceptive signals are transmitted to the CNS via the ascending (afferent) interoceptive pathways. The vagal and spinal nerves are the two arms of the ascending interoceptive pathways, with cell bodies in the nodose ganglion (NG) and dorsal root ganglion (DRG), respectively. The vagal nerve transmits afferent inputs to neurons of the nucleus of the solitary tract (NTS). The DRG neurons send projections into the brain through the dorsal column of the spinal cord. Various brain regions are involved in central interpretation, integration, and regulation of interoceptive information, including the NTS, the parabrachial nucleus (PB), thalamus, hypothalamus, hippocampus, amygdala (AMY) and prefrontal cortex (PFC). The regulatory signals generated by the CNS, in turn, are sent back to the peripheral organs via descending (efferent) interoceptive pathways. The sympathetic ganglia and parasympathetic chain ganglia, receiving signals from the sympathetic preganglionic neurons in the intermediolateral nucleus (IML) act as effector neurons of interoception to maintain the peripheral organ homeostasis.

Ascending and descending interoceptive pathways

Two major ascending (or afferent) peripheral pathways transmit interoceptive signals to the CNS, and they relay through two divergent categories (Figure 1). The first category consists of somatosensory neurons in dorsal root ganglia (DRG) that project to the brain via the dorsal horn, and predominantly transmit temperature, pain, and tissue injury-related signals (Brazill et al., 2019; Mantyh, 2014). Conventionally, the skeleton, skin, and muscle transmit interoceptive signals to the brain through this pathway (Ma, 2022; Wang et al., 2021). Further, the somatosensory neurons can be functionally divided into exteroceptive and interoceptive subtypes, with the former consisting of reflexive-defensive neurons that respond to external threats while the latter consists of recuperative neurons that respond to internal body injury (Ma, 2022).

The second category consists of sensory ganglia in the cranial and vagal pathways, such as the nodose or jugular ganglia, that largely send axons to the brain stem for visceral ascending signals (Holt et al., 2019). Vagal sensory neurons are distributed in the visceral organs from rostral to caudal and innervate the visceral organs from the surface to the lumen (Berthoud and Neuhuber, 2000; Paintal, 1973). Vagal ascending pathways mediate mechanical and chemical signals and are the main ascending pathways for the visceral organs (Chang et al., 2015; Kupari et al., 2019; Paintal, 1973). Whether the skeletal ascending signals share this pathway remains unknown.

The descending (or efferent) interoceptive pathways fully interpret the function of interoceptive circuits. Descending interoceptive pathways communicate from the brain to target organs and influence internal homeostasis (Berntson and Khalsa, 2021). The CNS coordinates involuntary input via efferent regulatory interoception through the autonomic nervous system (ANS), which comprises two distinct branches; namely, the sympathetic and parasympathetic nervous systems (SNS and PSNS) (Hanoun et al., 2015). The hypothalamus-derived descending interoceptive pathway may directly reach sympathetic preganglionic neurons, which are located in the spinal intermediolateral nucleus (IML) from the first thoracic to the second lumbar spinal cord segments, or they may relay via brain stem neurons (Elefteriou, 2018). The para- and prevertebral sympathetic ganglia, as well as chromaffin cells in the adrenal medulla, receive the projection from the sympathetic preganglionic neurons in the IML. The latter maintains the homeostasis of peripheral organs through the key neurotransmitter norepinephrine (NE), which mainly regulates vessel contraction and relaxation by communicating with contractile vascular cells, especially pericytes and smooth muscle cells (Rodriguez-Diaz et al., 2011).

Interpretation processes in the brain

Generally, the ascending pathway first projects to the subcortical regions of the brain, especially the nucleus tractus solitarius (NTS), and then interoception information is sent to the parabrachial nucleus, thalamus, hypothalamus and hippocampus (Chiang et al., 2019; Holt et al., 2019; Livneh et al., 2017; Phillips et al., 2019; Ran et al., 2022; Saper and Lowell, 2014; Suarez et al., 2020). The hypothalamus is mainly responsible for metabolism-related interpretation processes, while pain and odor interoception rely on the parabrachial nucleus (Critchley and Harrison, 2013; Saper, 2002). In turn, the signals may project to higher brain regions, such as the anterior cingulate cortex and somatosensory cortex, for further interpretation (Cobos and Seeley, 2015; Wang et al., 2019). Moreover, the insula (or the insular cortex) is recognized as the key area for interpreting visceral sensation and is a relatively understudied brain region (Craig, 2002). Indeed, because of its central role in interoception, the insula has even been viewed as the seat of a “sense of self” in humans (Underwood, 2021).

Owing to the limited focus of this Review on providing an overview of interoception and its role in the skeleton, please find more comprehensive information regarding interoception in previous reviews (Berntson and Khalsa, 2021; Chen et al., 2021; Petzschner et al., 2021; Quigley et al., 2021; Weng et al., 2021).

Skeletal interoception

The skeletal system provides the mechanical support for our daily activity while providing protection for vital organs, such as the brain and bone marrow. The bone also functions as an endocrine organ that regulates the metabolism of minerals, glucose, and fatty acids by interacting with other tissues and organs (Dirckx et al., 2019; Leider, 1947; Zhou et al., 2021). Importantly, the CNS is a key regulator of the skeletal system, and this process has been studied for decades (Chenu, 2004; Hara-Irie et al., 1996; Mach et al., 2002; Serre et al., 1999). There are abundant sensory and sympathetic innervations for both skeletal homeostasis and pain (Elefteriou, 2018; Fukuda et al., 2013; Hu et al., 2022; Mantyh, 2014). Moreover, bone undergoes constant remodeling, and thus, changes of its density, which is also perceived and regulated through nonconscious sensing - similar to gut motility, blood glucose fluctuations, and so forth. When aberrant mechanical loading of the joints occurs, or during aging of the skeleton, osteoarthritis, spine degeneration, osteoporosis and other skeletal diseases arise (Bian et al., 2016; Zhen and Cao, 2014; Zhu et al., 2020). Pain is the most common symptom of skeletal disorders, and there are no disease-modifying treatments for the major skeletal disorders, including osteoarthritis and spine degeneration, and thus treatment is usually limited to pain management (Foster et al., 2018; Koes et al., 2006; Puljak et al., 2017). Key questions are how skeletal sensory nerves perceive the changes in bone density, and what the signal(s) is from the bone that activates sensory nerves to the CNS. In the following sections, we describe novel findings related to skeletal interoception for bone homeostasis.

PGE2 levels are correlated with bone density changes

Prostaglandin E2 (PGE2), which is derived from arachidonic acid, with the rate-limiting enzymes cyclooxygenase (COX)-1 and COX-2, is a multifunctional molecule that mainly induces the inflammatory response and promotes nociceptor sensitization and vasodilatation (Blackwell et al., 2010). In addition, PGE2 has long been known as a potent intrinsic anabolic bone formation factor via the EP4 receptor (Blackwell et al., 2010; Yoshida et al., 2002). The prostaglandin transporter (PGT), encoded by SLCO2A1, and the prostaglandin-degrading enzyme NAD⁺-dependent 15-hydroxyprostaglandin dehydrogenase (15-PGDH), encoded by the 15-hydroxyprostaglandin dehydrogenase gene (HPGD), play critical roles in PGE2 catabolism (Zhang et al., 2015c). Mutations in these genes results in the impairment of the degradation of PGE2, resulting in subperiosteal new bone formation or the rare disease primary hypertrophic osteoarthropathy (PHO) (Uppal et al., 2008; Yüksel-Konuk et al., 2009). Conventionally, local administration of PGE2 induces bone formation and callus development (Weinreb et al., 1997). Local PGE2 levels increase after bone fracture, while PGE2 inhibition impairs bone healing (Park et al., 2019). In addition, mechanical loading elevates bone marrow PGE2 levels in the skeleton (Hino et al., 2006; Thorsen et al., 1996). Moreover, paradoxically, bone density is negatively correlated with PGE2 concentrations, as its levels rise in osteoporotic animal models and aged mice (Burt-Pichat et al., 2005; Strotmeyer et al., 2006). These phenomena indicate that PGE2 levels change with bone density under various conditions. Interestingly, knockout of EP4 in osteoblastic cells does not change the bone density in mice (Gao et al., 2009). Thus, PGE2-induced osteoblastic bone formation is mediated through an alternate pathway from cell autonomous EP4 signaling.

PGE2 activates EP4 in sensory nerves as an ascending pathway

Patients with cognitive sensory malfunction suffer from bone loss and an increased fracture rate (Pérez-López et al., 2015). Skeleton sensory nerves release different types of proteins to regulate skeleton homeostasis, deletion of the diffusible axonal chemorepellent semaphorin 3A (Sema3A) gene in neurons induces bone loss (Fukuda et al., 2013; Sun et al., 2020). But it was still not clear if sensory nerves in the skeleton contribute to the sensation of bone density changes, such as by sensing changes in PGE2 levels, and whether they transmit these signals to the CNS. As EP4 is responsible for PGE2-mediated bone formation (Yoshida et al., 2002), we utilized EP4_Avil^−/− mice, which have a specific ablation of EP4 in sensory nerves, and found that trabecular and cortical bone mass was dramatically decreased in 12-week-old mutant mice. Ablation of EP4 in sensory nerves remarkably abrogated bone formation mediated by the PGE2 degradation enzyme inhibitor SW033291. Thus, PGE2 mediates bone formation through EP4 signaling in skeletal sensory nerves. Moreover, by calcium imaging we detected a greater number of DRG neurons in the PGE2-treated group, while the number of DRG neurons was significantly lower in the EP4_Avil^−/− and vehicle-treated mice (Chen et al., 2019). These data further confirm that PGE2 activates skeletal sensory neurons through EP4 to perceive the changes of skeleton PGE2 concentrations as ascending pathway in regulation of bone formation.

The skeleton has abundant sensory innervations that interact with the CNS (Chenu, 2004; Mach et al., 2002). The route from sensory neurons within the skeleton to the CNS has been well elucidated via viral tracing experiments and immunofluorescence (Brazill et al., 2019; Chartier et al., 2018; Mach et al., 2002). Notably, L₁-L₆ DRG send axons to the tibia as early as embryonic day 14.5 in rodents (Tomlinson et al., 2016). And skeletal sensory neurons send central projection via L₁-L₆ DRG, converging with second order neurons or interneurons within the dorsal horn. Further, as shown by retrograde labeling experiment, acute noxious mechanical stimulation of the rat tibial diaphysis activates the superficial dorsal horn, which in turn projects to the lateral parabrachial nucleus (Williams and Ivanusic, 2008). These findings, thus, anatomically define a potential pathway for skeletal pain. Likewise, we showed by anterograde virus tracing that injection of green fluorescent-labeled herpes simplex virus (HSV) into the mouse femur metaphysis results in staining of the arcuate nucleus (ARC) of the hypothalamus (Lv et al., 2021). Collectively, this afferent neuronal pathway connects the skeleton to the CNS chiefly through the DRG. It will be of interest to further investigate if skeletal sensory afferents also convey the signals via visceral or cranial afferent pathways.

The hypothalamus discriminates hormonal, mechanical and metabolic signals, such as glucose, fatty acids and amino acids levels, from peripheral organs to initiate compensatory changes that modulate ANS responses for the maintenance of homeostasis of global energy metabolism (Dhillon et al., 2006). The ventromedial hypothalamic nucleus (VMH) is vital for the regulation of body weight, glucose balance, mood behaviors and reproductive function via the ANS (Choi et al., 2013; Dhillon et al., 2006; Idelevich and Baron, 2018; King, 2006). Skeletal sensory nerves have an anatomical connection with the hypothalamus, as noted above.

Therefore, we investigated whether PGE2-activation of the DRG could transmit the signals to the hypothalamus. CREB signaling in the VMH is closely related to bone mass (Oury et al., 2010), and we found that activation of CREB signaling in VMH neurons occurred after PGE2 injection, which was significantly decreased in EP4_Avil^−/− mice (Chen et al., 2019). These results indicate that EP4 signaling in sensory nerves is essential for PGE2-induced CREB phosphorylation in the VMH.

Downregulation of sympathetic activity promotes osteoblast differentiation as a descending pathway

The hypothalamic-ANS-skeleton efferent neuronal pathway has been well defined (Takeda et al., 2002; Zhu et al., 2018). Different peripheral organs, including the skeleton and adipose tissues, have been confirmed to share a similar efferent neuronal pathway (Dénes et al., 2005; Wee et al., 2019). Specifically, the lumbar and thoracic paravertebral chain ganglia were labeled 4–5 days after fluorescently labeled neurotropic pseudorabies virus was injected into the rat femur, followed by labeling of the intermediolateral column, brain stem, midbrain, and especially the locus coeruleus and hypothalamus (the PVN and lateral hypothalamic area), which are known to regulate SNS activity at 5–6 days after injection.

CNS regulation of bone metabolism via the SNS has been reported (Ducy et al., 2000; Karsenty and Ferron, 2012; Takeda et al., 2002). Further, leptin regulates bone mass through the SNS, and ablation of the adrenergic receptor β2 (Adrb2) leads to a high bone mass phenotype (Takeda et al., 2002), while activation of Adrb2 in osteoblasts by NE inhibits CREB phosphorylation, resulting in decreased osteoblast proliferation (Kajimura et al., 2011). Importantly, postmenopausal women receiving high levels of the Adrb1 selective blockers nebivolol and atenolol showed higher BMD compared with those treated with the non-selective blocker propranolol or placebo (Khosla et al., 2018). This clinical observation indicates that the SNS regulates bone metabolism primarily through Adrb1, not Adrb2 and Adrb3, and that increases in sympathetic tone promotes bone loss in postmenopausal women and aged patients, while pharmacological inhibition of Adrb1 benefits human bone health.

Our previous results showed that PGE2 induced CREB phosphorylation in the VMH via EP4 signaling in sensory nerves, it was still unclear whether sympathetic tone is regulated by this process. Importantly, expression of the uncoupling protein 1 (UCP1) gene in the brown adipose tissue and concentrations of urine epinephrine were significantly increased in EP4_Avil^−/− mice, thus suggesting that PGE2-mediated activation of EP4 on sensory nerves suppresses sympathetic tone, thereby regulating metabolic activity and osteoblastic bone formation (Chen et al., 2019).

Skeletal interoception regulates differentiation of bone marrow mesenchymal stromal cells

The bone marrow provides a unique microenvironment for multiple types of cells, including hematopoietic stem cells (HSCs) and bone marrow mesenchymal stromal cells (bMSCs). Generally, HSCs and bMSCs reside in the bone marrow microenvironment in a quiescent state and can be activated by injury, inflammation, mechanical stimuli and drug intake (Méndez-Ferrer et al., 2010; Morrison and Scadden, 2014; Shen et al., 2021). Sympathetic nerves are essential for the regulation of HSCs and bMSCs (Hanoun et al., 2015; Maryanovich et al., 2018). Recently, Frenette’s group reported that nociceptor neurons drive HSCs mobilization via CGRP (Gao et al., 2021). Our previous study showed that osteoblast-derived PGE2, acting as an interoceptive signal via EP4 on advilin-positive ascending sensory nerves, transmits mechanical signals to the hypothalamus and regulates not only osteoblasts but also the commitment of Lepr⁺ bMSCs by toning down descending sympathetic activity. Furthermore, adipogenesis dramatically increases, whereas osteogenesis decreases after the abrogation of PGE2-EP4 skeletal interoception (Hu et al., 2020). The activation of skeletal interoception by the interoceptive signal of mechanical load or by the administration of a PGE2 degradation enzyme inhibitor promotes osteogenesis of Lepr+ stromal cells and accelerates the bone healing process in bone fracture and bone regeneration animal models (Hu et al., 2020; Zhang et al., 2015c). These findings extended the array of target cells for skeletal interoception, while expanding the horizons for future research on skeletal interoception.

In summary, skeleton interoception regulates bone homeostasis through PGE2-EP4 signaling in sensory nerves (Figure 2). Mechanical load triggers secretion of the interoceptive signal PGE2 from osteoblasts, and PGE2 activates the interoceptor EP4 in ascending sensory nerves, which induces the interpretation process in the hypothalamus and the phosphorylation of CREB signaling in the VMH nucleus. Descending sympathetic tone is downregulated by CREB phosphorylation in the VMH to promote bMSCs osteoblastic differentiation and bone formation, therefore maintaining skeletal homeostasis. In addition, elevated PGE2 levels activates skeletal interoception, which accelerates the bone regeneration process, and this phenotype is abrogated by disruption of skeletal interoception via genetic and pharmacological manipulations (Hu et al., 2020).

Figure 2. Skeleton interoception maintains bone homeostasis.

PGE2 secreted by osteoblasts activates EP4 in sensory nerves as the ascending skeletal interoceptive pathway. Activation of CREB signaling by the PGE2/EP4 ascending interoceptive activity in the ventromedial nucleus of the hypothalamus (VMH) downregulates sympathetic nerve activity as the descending interoceptive pathway. The descending sympathetic tone is downregulated via skeleton interoceptive circuit to induce a commitment of mesenchymal stem/stromal cells (MSCs) to an osteoblastic lineage and bone formation.

Interoception regulation of bone metabolism

Bone remodeling

Given terrestrial life and the frequent development of microcracks, the skeleton continuously undergoes remodeling throughout adulthood in mammals, with the activity of osteoblasts and osteoclasts precisely coupled to ensure proper skeletal integrity (Eriksen, 2010; Zaidi, 2007). Osteoblasts, which deposit calcified bone matrix, are derived from bMSCs (Bianco et al., 2013; Méndez-Ferrer et al., 2010), while multinucleated osteoclasts, which degrade bone, are derived from macrophages (Doherty et al., 2015; Tang et al., 2009; Teitelbaum, 2000; Xie et al., 2014). As part of the bone remodeling process, osteoclast-mediated bone resorption results in the release of multiple factors from the matrix of the bone surface, including transforming growth factor beta 1 (TGF-β1) and insulin-like growth factor 1 (IGF-1), that recruit bMSCs to the newly created bone-resorptive surface where they differentiate into osteoblasts and form new bone (Crane and Cao, 2014; Crane et al., 2013; Dallas et al., 2002; Tang et al., 2009; Xian et al., 2012). The bone formation process is also closely associated with the CD31^hiEmcn^hi type H vessels-coupled angiogenesis (Brandi and Collin-Osdoby, 2006; Kusumbe et al., 2014; Ramasamy et al., 2014). Importantly, platelet-derived growth factor-BB (PDGF-BB) secreted by preosteoclasts induces type H vessel formation and stimulates angiogenesis, which is coupled with osteogenesis (Xie et al., 2014). These studies further clarified that angiogenesis and innervation participate in skeletal homeostasis. Further details on these skeletal mechanisms have been well-described in previous reviews (Crane and Cao, 2014; Zaidi et al., 2018), and thus will not be further outlined here.

Energy uptake in the skeleton

Continuous bone remodeling, bMSC proliferation and differentiation, bone matrix synthesis and mineralization consume tremendous amounts of energy to maintain bone and calcium metabolic homeostasis (Karner and Long, 2018; Riddle and Clemens, 2017). Skeletal cells acquire energy almost exclusively from glucose, lipids and amino acids (Dirckx et al., 2019; Kim et al., 2017). Glucose is historically viewed as the primary energy source for skeletal cells at rest and during times of high energy demand, such as skeletal development and high bone turnover after teriparatide (recombinant human parathyroid hormone (1–34)) treatment for osteoporosis, because of facile and rapid ATP production during glycolysis (Esen et al., 2015; Karsenty and Khosla, 2022; Wei et al., 2015). Individuals with anorexia nervosa or metabolic disorders exhibit skeletal growth arrest in adolescence and bone loss in adulthood, and these clinical phenomena increase the credibility of the view of skeletal energy demand (Legroux-Gerot et al., 2005; Misra and Klibanski, 2011). Moreover, disruption of the skeletal energetic balance induces global metabolic diseases (Zaidi, 2007) (for a comprehensive review, see (Karsenty, 2006; Riddle and Clemens, 2017; Zhang et al., 2015b)).

Neuroendocrine regulation of bone metabolism

Hypothalamic regulation is critical for systemic physiological balance, including energy metabolism, feeding and digestion, reproductive and stress-related endocrine responses (Saper and Lowell, 2014). Conventionally, hypothalamic nuclei-derived hormones stimulate an anterior pituitary gland-depended hypothalamic-pituitary-target organ axis neuroendocrine regulation, which is closely related to bone remodeling and metabolic homeostasis (Abe et al., 2003; Khosla et al., 2012; Kim and Mohan, 2013; Zaidi et al., 2016; Zaidi et al., 2010). Together, the hypothalamic neuroendocrine system precisely coordinates the intimate and complicated relationship between the skeleton and other organs, including the reproductive system and adipose tissue. Other hypothalamic neuroendocrine hormones governing skeletal metabolism have been discussed previously (Khosla, 2020; Zaidi et al., 2016).

The ARC is located at the bottom of the 3^rd ventricle in a region with a less restrictive blood brain barrier, and one that is sensitive to afferent signals (Saper and Lowell, 2014). The ARC contributes to skeletal and whole-body energy metabolic homeostasis by releasing multiple neuropeptides. Historically, the ARC is comprised of two main types of neurons: anabolic orexigenic agouti-related protein (AgRP)/ neuropeptide Y (NPY)-positive neurons and catabolic anorexigenic proopiomelanocortin (POMC)/CART-positive neurons, which exert distinct effects on skeletal metabolism (Kim et al., 2015). NPY is highly expressed in AgRP/NPY-positive neurons in the ARC, and NPY global knockout mice show higher bone mass, with increased osteoblast activity and osteogenesis-related gene expression (Baldock et al., 2009). NPY also performs its function via its receptors (Y1R-Y6R). Global or conditional knock outs of Y2R in the hypothalamus replicate the high bone mass of NPY-deficient mice without altering leptin levels (Baldock et al., 2002), suggesting that central NPY regulates bone mass independently from the leptin-bone pathway, but does so through the hypothalamus. However, deletion of hypothalamic Y1R does not affect bone mass, but rather its ablation in osteoblasts promotes osteoblastogenesis (Lee et al., 2011), implying that Y1R regulates the skeleton mainly through local effects. Intriguingly, a recent study from Herzog’s lab showed that the Y1R antagonist BIBO3304 enhances energy expenditure, improves glucose homeostasis and prevents diet-induced obesity (Yan et al., 2021). On this basis, skeletal and whole-body metabolic homeostasis may also be regulated by skeletal interoception.

Skeletal interoception regulates the neuroendocrine system in bone homeostasis

Skeletal interoception contributes to homeostasis of bone remodeling and bMSC differentiation (Hu et al., 2020). But it is not clear whether skeletal metabolic activity is also regulated by skeletal interoception. Knockout of EP4 in sensory nerves causes bone catabolism and fat anabolism (Hu et al., 2020), implicating a role for skeletal interoception in the regulation of fat and bone energy homeostasis (Figure 3). In particular, the activated PGE2-EP4 skeletal ascending interoceptive pathway downregulates NPY gene expression in the ARC, which promotes lipolysis of white adipose tissue and marrow adipose tissue, facilitating the release of free fatty acids that fuel osteoblastic bone formation (Lv et al., 2021). This finding revealed a new mechanism by which the CNS regulates skeletal metabolic activity while uncovering the neuroendocrine regulation of the descending pathway of skeletal interoception. Further studies are needed, however, to clarify the communication between the skeleton and other energy-consuming organs regulated by interoception.

Figure 3. Skeleton intreoception regulates neuroendocrine factor NPY to balance metabolism between bone and fat.

The skeleton interoceptive signal PGE2 activates the interoceptor EP4 in sensory nerves to induce the heterodimerization of phosphorylated cAMP-response element binding protein (pCREB) and transcriptional suppressor small heterodimer partner interacting leucine zipper protein (SMILE) to downregulate the expression of neuroendocrine factor NPY in the arcuate nucleus (ARC). Decreased NPY concentration in the blood circulation secreted from ARC mediates the neuroendocrine descending interoceptive signal to balance metabolism between bone and fat. Downregulation of hypothalamic NPY expression induces adipose tissue lipolysis to supply enough free fatty acids (FFAs) for energy-consuming osteoblastic bone formation, as well as the differentiation from pre-osteoblasts to osteoblasts.

CNS surveillance of whole-body energy homeostasis has been well accepted for decades. As increasing evidence supports organ crosstalk under physiological and pathological conditions, the idea that systemic regulation maintains physiological homeostasis in the body has gradually been accepted (Waterson and Horvath, 2015). The interoception-mediated regulation of the integration of interoceptive signals by different organs, and the subsequent metabolic response(s) generated, have been gradually defined (Quigley et al., 2021). A better understanding of interoception-mediated regulation of metabolic homeostasis will provide novel effective strategies for treating metabolic or degenerative diseases.

Regulation of energy metabolism by interoceptive mechanisms

Interactions between the interoception networks involved in basic physiology and emotional processes maintain whole-body metabolic homeostasis (Quigley et al., 2021). This relationship has driven the field to explore the neural circuitries responsible for inducing behavioral responses and regulating energy utilization. Here, we review recent studies on interoceptive regulation of adipose tissue and gut metabolic homeostasis (Figure 4).

Figure 4. Interoception-mediated regulation of feeding and fat and glucose metabolism.

Mammalian energy needs, appetite, and fat and glucose metabolism are regulated by interoception. The vagal and spinal nerve communicate the energy status from the energy centers to the brain, which induce the integration of interoceptive signals to regulate whole-body metabolic homeostasis through descending sympathetic nerve. The adipose tissue, gut and pancreas are richly innervated by sensory nerve fibers, and distinct afferent sensory neurons have been identified to control appetite and fat and glucose metabolism through different interoception circuits. Bone remodeling consumes tremendous amounts of energy and interoception regulates whole-body energy metabolic homeostasis to balance the necessary energy flow between the bone and different organs to meet these osteogenic needs.

Adipose tissue serves as an energy reservoir and endocrine organ, which has a close relationship with global energy homeostasis (Cannon and Nedergaard, 2004). Sympathetic innervation in adipose tissue changes dynamically under different physiological and pathological states to meet the body’s energy needs (Bartness et al., 2010; Bentsen et al., 2019; Cannon and Nedergaard, 2004; Lin et al., 2021; Wang and Seale, 2016; Zeng et al., 2015). Deletion of sympathetic innervation in the white adipose tissue (WAT) blocks lipolysis and promotes adipocyte proliferation (Bachman et al., 2002; Harris, 2018). Further, sympathetic innervation of the brown adipose tissue (BAT) contributes to whole-body metabolism via coordinating fat and glucose metabolism (Villarroya et al., 2017; Wang and Seale, 2016). Metabolic signals detected by sensory nerves in the adipose tissue are processed in the CNS, consequently affecting metabolic homeostasis by modulating sympathetic tone (Garretson et al., 2016; Guilherme et al., 2019; Nguyen et al., 2018; Shi et al., 2005; Vaughan and Bartness, 2012; Wang et al., 2022).

The gut-brain axis regulates nutrient absorption, gastrointestinal motility, the microbial environment and even hippocampal cognitive function, and thus has attracted much attention in recent years (Furness et al., 2013; Goldstein et al., 2021; Han et al., 2018; Suarez et al., 2018; Travagli and Anselmi, 2016; Williams et al., 2016). Different types of digestive-related interoceptive signals that participate in energy homeostasis control have been uncovered in recent years (Bai et al., 2019; Borgmann et al., 2021; Pocai et al., 2005; Zimmerman et al., 2019). Further studies show the therapeutic potential of regulating the microbiota within the gut-brain interoceptive circuits for Alzheimer’s disease and osteoporosis (Behera et al., 2020; Dinan and Cryan, 2017; Li et al., 2019; Muller et al., 2020; Yang et al., 2018). For more details about interoception regulation of energy metabolism, please see recent reviews (Berthoud et al., 2021; Zhang et al., 2021).

Mechanical load stimulates skeletal interoception-mediated bone homeostasis

Mechanical support of the body is the primary function of bone for terrestrial vertebrates (Doherty et al., 2015). In aquatic vertebrates, as their skeletons do not bear the force of gravity, their bones do not undergo osteoclastic bone remodeling. When osteoclasts and the parathyroid gland first evolved in amphibians, they were later adapted for bone remodeling and calcium metabolism to cope with the forces of gravity while on land and the resulting large energy demand that came with such environments (Riddle and Clemens, 2017; Zhang et al., 2015b). Mechanical load acts upstream of the regulation of bone remodeling and modeling. Mechanical load-activated skeletal interoception coordinates bone and global energy homeostasis as part of this process.

Mechanical signals are considered important interoceptive signals of vascular tone and gut tension for the maintenance of internal organ homeostasis, especially for the skeleton (Bian et al., 2017; Goodman et al., 2015; Hamill and Martinac, 2001; Robling et al., 2006). Indeed, the skeleton requires continuous adaptation to mechanical stimuli to maintain homeostasis. Disuse or overuse accelerates bone turnover as catabolic bone resorption dominates anabolic bone formation (Robling et al., 2006). The CNS integrates different stimuli, including mechanical stimuli, to maintain physiological homeostasis and energy metabolism (Waterson and Horvath, 2015). However, the mechanism by which the CNS integrates skeletal mechanical interoceptive stimuli remains elusive. For decades, bone biologists have demonstrated various types of skeletal mechanosensitive cells and the corresponding mechanoreceptors, and also sought to identify the molecular mechanism by which skeletal cells sense mechanical interoceptive signals and process mechanotransduction for skeletal homeostasis (Riddle and Donahue, 2009; Seeman, 2009). Recently, our group found that skeletal mechanical load stimulates PGE2 secretion from osteoblasts, causing the conversion of mechanical signals to biochemical signals for the activation of skeletal interoception (Chen et al., 2019).

It is well documented that proprioception receptors located in the musculoskeletal system and skin sense force and gravity, and afferent proprioceptive signals are transmitted to the brain during body movement and distinct postures (Delmas et al., 2011; Proske and Gandevia, 2012). Emerging studies have suggested that skeletal tissue communicates with the nerves and blood vessels in response to mechanical load (Bleedorn et al., 2018; Sample et al., 2008; Tomlinson et al., 2017; Wu et al., 2009). Sample et al. found that axial compression on the ulna induced bone formation in the loaded limb as well as the remote control side; interestingly, local blockade of sensory nerves in the loaded limb with bupivacaine not only disrupted bone formation on the directly loaded area but also in the distant unloaded bones (Sample et al., 2008). It is easy to speculate that the mechanical load-induced anabolic response on the contralateral unloaded side must be facilitated through systemic regulation, such as skeletal interoception, rather than local effects. Further, few corresponding effects at the DRG level are observed after temporal mechanical loading in the rat ulna, suggesting that mechanical load-induced bone formation are not affected by local neuronal active factors and that an advanced neural center, such as the brain, may participate in mechanical signal processing (Bleedorn et al., 2018). Recent studies still mainly focus on the local relationship of the nervous system and skeletal cells rather than from the interoception perspective. Tomlinson et al. revealed the bidirectional communication between osteoblasts and sensory nerves in bone via tropomyosin-related kinase A (TrkA) signaling activated by osteoblast-derived nerve growth factor (NGF). Loading induces osteoblast proliferation and is accompanied by increased NGF expression, which guides nerve sprouting and innervation into the new bone formation area (Tomlinson et al., 2017). Moreover, disruption of TrkA signaling in osteocytes by tyrosine kinase inhibitor 1NMPP1 decreases bone formation after ulnar compression, and these effects are rescued by exogenous NGF (Tomlinson et al., 2016).

The relationship between the SNS and mechanical load is still unclear. Pharmacological blockade of sympathetic tone does not affect load-induced bone formation or disuse-induced bone resorption, whereas sciatic neurectomy promotes compression-induced tibia cortical bone formation in mice (de Souza et al., 2005; Marenzana et al., 2007). Increased bone density and cortical thickness are observed after tibial compression in WT and Adrb2 knockout mice, but not in β1-adrenergic receptor (Adrb1) knockout mice, indicating Adrb1 plays a role in mechanoadaptation during load-induced bone formation (Pierroz et al., 2012). These findings clarify the role of descending sympathetic nerves in mechanical load-induced skeletal interoception.

Reduction of PGE2 production as a treatment for skeletal pain

The global burden of diseases, injuries, and risk factors study found that back pain and osteoarthritis (OA) were the leading causes of disability worldwide in 2015 and led to the greatest number of years lived with disability (YLD) (2016). With an aging population, the burden of back pain and arthritis of the joints on the quality of life and on health care systems has continued to rise (Barnett, 2018; Foster et al., 2018). Indeed, the socioeconomic burden of these conditions is enormous: the cost of spinal degeneration (and associated back pain) is over $100 billion annually in the US alone, more than cardiovascular and cerebrovascular diseases combined, respiratory infections, as well as endocrine and rheumatoid disease (Boos et al., 2002). Approximately 67 million Americans will suffer from OA by 2030 (Hootman and Helmick, 2006), leaving surgical treatments, such as joint replacement or spinal fusion as the only therapeutic options for end-stage OA (Koes et al., 2006; Mueller et al., 2017).

Elevated PGE2 levels in the porous endplates causes low back pain

Low back pain (LBP) is an extremely common health problem that leads to limited mobility and absenteeism in most parts of the world (Samartzis and Grivas, 2017). Unfortunately, the cause of LBP remains unclear and there is no effective disease-modifying treatment. Magnetic Resonance Imaging (MRI) examination reveals vertebral endplate changes, defined as Modic changes, in most patients with non-specific LBP, and these changes correlate with disease progression (Jensen et al., 2008; Rahme and Moussa, 2008). Also, there is a greater degree of innervation found in the degenerative endplate than in normal endplate or degenerative nucleus pulposus (Fields et al., 2014). However, how the degenerative endplate is innervated is still unclear. Bisphosphonates and RANKL inhibitors that inhibit osteoclast activity have shown analgesic effects and alternations in the Modic changes associated with LBP. Further, calcification of the endplate with a narrowed intervertebral disc (IVD) space are characteristics of spinal degeneration (Figure 5) (Bian et al., 2016; Bian et al., 2017; Zheng et al., 2018). The calcified endplates are remodeled by osteoclast resorption, which secrete factors, including PDGF-BB and netrin-1, to induce angiogenesis and sensory innervation, leading to LBP (Ni et al., 2019).

Figure 5. Skeleton interoception in bone homeostasis and disorders.

Metal divalent cations released from implants induce the production of the interoceptive signal PGE2 from macrophages, and this molecule activates the ascending interoceptive pathway to reduce sympathetic tone in the hypothalamus, resulting in increased osteogenesis and decreased osteoclastogenesis in the periosteum. Abnormal concentrations of PGE2 in the endplate and in subchondral bone induce hypersensitivity of sensory nerves by activating EP4, which leads to LBP and osteoarthritis pain. Netrin1 secreted by osteoclasts in subchondral bone induces sensory nerves innervation in subchondral bone, which also contribute to OA pain. Restoration of physiological PGE2 concentrations in the endplate activates skeleton interoception to induce osteoblastic bone formation in porous endplates, thus relieving LBP, reducing vertebral endplate porosity and modifying disease progression.

Importantly, the porous bony endplates mimic the extreme low bone mass of osteoporosis. The levels of PGE2 are significantly elevated in the porous endplates. The high levels of PGE2 sensitize Nav1.8 sodium channels and TRPV1 on sensory nerves, thereby inducing spinal hypersensitivity (Figure 5). Knockout of EP4 in the sensory nerves relieves LBP (i.e., spinal hypersensitivity) (Ni et al., 2019). As an interoceptive signal, physiological PGE2 concentrations activate skeletal interoception for the maintenance of skeletal metabolic homeostasis (Lv et al., 2021), and lose-dose celecoxib, a selective inhibitor of COX2, normalizes PGE2 concentrations and not only relieves spinal hypersensitivity but also activates skeletal interoception to promote bone formation and block sensory innervation in the porous endplates (Xue et al., 2021). In effect, then, low-dose celecoxib kills two birds with one stone by relieving spinal pain and modifying disease progression.

The PVN plays critical roles in the regulation of sympathetic outflow to peripheral organs. Tyrosine hydroxylase (TH) is the rate-limiting enzyme for NE synthesis, and TH levels in the PVN correlate with sympathetic outflow. Immunostaining has shown that TH levels significantly decrease after treatment with low-dose celecoxib (Xue et al., 2021). Therefore, the PVN is another critical region for the interpretation of skeletal interoception.

PGE2 levels are increased in the porous subchondral bone to induce joint pain

Knee joint subchondral bone homeostasis is maintained by temporal-spatial activation of TGFβ to couple bone resorption with proper degrees of angiogenesis and innervation (Zhen et al., 2013). Aberrant mechanical loading on the joints, however, induces uncoupled remodeling that generates porosity of subchondral bone, leading to the development of OA (Zhen and Cao, 2014; Zhen et al., 2013). Specifically, high levels of active TGFβ recruits bMSCs and osteoprogenitors, leading to abnormal bone formation and angiogenesis, which induces excessive sensory innervation of the subchondral bone, resulting in OA-related pain (Glasson et al., 2005; Goldring, 2009). It was shown that inhibition of elevated TGFβ activity rescues proper coupling of subchondral bone remodeling to reduce its porosity structure and excessive sensory innervation and thus reducing OA-related pain (Figure 5) (Zhen et al., 2013). Further, abnormally high PGE2 concentrations in porous subchondral bone induce hyperpathia and OA-related pain (Zhu et al., 2020). Moreover, osteoclastic resorption also creates an acidic environment in subchondral bone that may indirectly sensitize nociceptive neurons (Yoneda et al., 2015).

OA is a complicated syndrome consisting of several subtypes, which often leads to ineffective treatment (Barnett, 2018). One subtype of OA that has been defined exhibits a high concentration of PGE2 and overgrown subchondral bone with impaired articular cartilage. Use of celecoxib was shown to effectively attenuate the progression of OA in relevant mouse models (Tu et al., 2019). Moreover, knockout of COX2 in osteoblasts or EP4 in Nav1.8-positive sensory nerves efficiently attenuates OA-related pain (Zhu et al., 2020). These results suggest that EP4 mediates OA-related pain and skeletal interoception may also participate in the regulation of OA progression and its related pain.

Aberrant mechanical load stimulates the osteoclast bone resorption in spine endplates, joint subchondral bone or different bones in pathological conditions to generate porous structures (Bian et al., 2016; Zhen et al., 2013). Importantly, many factors are released from the bone matrix or secreted by osteoclasts; in particular, Netrin-1 in the porous bone, which promotes CGRP+ neuron innervation (Wu et al., 2016), which in turn is known to promote OA-related pain (Aso et al., 2016). Netrin-1 is an axon guidance factor and induces nociceptive nerve fibers to grow into the porous bone via its receptor, deleted in colorectal cancer (DCC), on sensory nerve fibers, resulting in OA-related pain (Zhu et al., 2019). DCC signaling induces axonal growth during neural development and is required for the development of nociceptive topognosis in both mice and humans, which is essential for the ability to localize painful stimuli (da Silva et al., 2018; Glasgow et al., 2021). Multiple genome-wide association studies (GWASs) have determined that DCC is the primary gene responsible for chronic pain and psychiatric disorders in humans (2019; Johnston et al., 2021). DCC is also a modulator of maladaptive responses to chronic morphine administration and affects tolerance, dependence, and opioid-induced hyperalgesia following chronic opioid exposure (Liang et al., 2014). Moreover, netrin-1-DCC-mediated signaling regulates synaptic plasticity by increasing AMPA receptors (AMPARs), providing a potential mechanism for the increased excitability of DRG neurons (Glasgow et al., 2018). The expression of DCC co-localized with CGRP+ sensory fibers in porous endplates in LBP mouse models where PGE2 concentration is also elevated (Ni et al., 2019). The sensory nerve fiber growth in the porous bone is the basis of bone-related pain, but it is unclear whether DCC is directly involved in elevated PGE2-induced pain.

Cartilage degeneration is also a major contributor to the development of OA (Felson, 2006). Continuous chronic and low-grade inflammation is observed in OA, resulting in gradual cartilage loss and progressive joint dysfunction (Robinson et al., 2016). Increased levels of prostaglandins, especially PGE2, have been detected in the OA joint and induce cartilage catabolism, chondrocyte apoptosis, and angiogenesis (Li et al., 2005; Martel-Pelletier et al., 2003). Recently, studies reported that EP4 antagonism could inhibit cartilage degeneration and osteoclast activity in subchondral bone, which reduces OA pain and prevents the progression of OA (Jiang et al., 2022; Jin et al., 2022; Phillips, 2022).

As a component of traditional Chinese medicine, the electroacupuncture technique has long been used to modulate internal organ function by activating neuronal networks. One of the most frequently stimulated acupoints, Zusanli (ST36), has the potential to correct severe inflammation under pathological conditions. However, the underlying mechanism was not clearly known until recent work from Ma’s group. They determined that Prokr2-positive neurons at the ST36 acupoint mediate low-intensity-induced activation of the vagal-adrenal axis, which activates the SNS under acute inflammation (Liu et al., 2021). It would be interesting to investigate whether acupuncture could modulate activity of PGE2-EP4-mediated skeletal interoception.

Skeletal pain such as OA-related pain and LBP have protective effects and alert the body to the occurrence of joint and spinal dysfunction, and thus pain essentially serves as an interoceptor. In the early stages of skeletal diseases, the transmission of interoceptive signals to the brain promotes protective posture and actions (Ma, 2022). On the other hand, the brain sends signals through the descending interoceptive pathway, the ANS and the neuroendocrine system, to activate a protective negative feedback loop to maintain local metabolic balance and skeletal homeostasis (Xue et al., 2021). Further understanding of the skeletal interoception projection(s) in the brain and pain regulation may provide opportunities to develop novel treatments for skeletal pain, in particular for those associated with OA and LBP.

Metal-induced bone formation through skeletal interoception

The demand for advanced clinical metal implants for the treatment of orthopedic diseases has been increasing (Wang et al., 2020). The development of metal-based orthopedic implants has occurred since the discovery of their osteogenic effects (Lin et al., 2018; Liu et al., 2018; Zhang et al., 2015a), although the underlying mechanism by which they promote bone growth remains elusive. Previous findings from Qin’s lab showed that implant-derived Mg²⁺ diffused toward the periosteum, where CGRP+ sensory fibers are abundantly distributed, promoting the secretion of CGRP+ vesicles from DRG neurons and triggering osteogenic differentiation in periosteal stem cells via CALCRL-RAMP1-pCREB signaling (Zhang et al., 2016). This study suggests that Mg²⁺-based implant-induced osteogenic effects rely on sensory innervation.

Recent advances in the development of neurovascularized bone implant materials have been expected to yield materials that better mimic natural skeletal tissue, but the detailed mechanisms by which the CNS promotes bone regeneration remains unclear (Marrella et al., 2018). Even so, the uncovering of skeletal interoception has shed light on our understanding of the mechanism of metal-induced bone formation and its translational application. We recently found that divalent metal cation-based biomaterials induce PGE2 secretion from macrophages and sprouting of CGRP+ fibers in bone, which activates skeletal interoception to promote new bone formation (Figure 5) (Qiao et al., 2022). Further, sensory denervation or knockout of EP4 in sensory nerves dramatically eliminated divalent cation-induced bone formation through skeletal interoception (Qiao et al., 2022). These findings reveal that divalent metal cations activate skeletal interoception for bone formation via the skeletal immune-neural ascending pathway. Given that, the use of controlled release of divalent metal cations could be developed as part of novel implants with bone regeneration potential.

Prospects for interoception research and therapeutic potential for bone disorders

PGE2 is recognized as the primary inflammatory factor in skeletal pain. PGE2 concentrations in bone is perceived by sensory nerves, which in turn regulate the activity of PGE2-EP4 signaling-mediated skeletal interoception for homeostasis. In pathological conditions, uncoupled bone remodeling often generates low bone density with high PGE2 concentrations. Therefore, PGE2 concentrations within the physiological range regulates bone homeostasis whereas pathology-related high PGE2 concentrations generate a pain signal to alert the body of a possible fracture or other damage to the bone or joints (Chen et al., 2019; Ni et al., 2019; Xue et al., 2021).

PGE2 levels in porous endplates increase in a LBP mouse model. And as the main effect of nonsteroidal anti-inflammatory drugs (NSAIDs), such as celecoxib, is to reduce the production of PGE2, they are currently recommended as the first-line treatment for chronic skeletal pain, and have been extensively used for decades, with annual sales of more than $7 billion in the U.S. alone (Qaseem et al., 2017). In a recent study, we found that high-dose celecoxib (80 mg/kg per day) treatment reduced PGE2 levels and LBP in a mouse model, whereas a very low dose (5 mg/kg per day) had no effect. However, high-dose treatments impaired bone formation in porous endplates, and LBP recurred after discontinuing the treatment. Interestingly, low-dose celecoxib (20 mg/kg per day) reduced both the endplate porosity and LBP. Importantly, LBP did not recur even after discontinuing treatment. Most importantly, the PGE2 concentration in the endplates 4 weeks after the low dose was equal to that of sham control mice. However, the PGE2 concentrations in both the high-dose and the very low-dose groups 4 weeks after the treatment were similar to the vehicle-treated LBP mice (Xue et al., 2021). These results indicate that a therapeutic treatment that leads to local physiologically-relevant PGE2 concentrations activates skeletal interoception-induced bone formation to modify disease progression.

Osteoblasts are the primary source of PGE2 in the bone microenvironment, especially upon mechanical stress (Bakker et al., 2003; Ponik and Pavalko, 2004). Thus, a relatively easy and effective therapeutic means to increase local bone PGE2 levels is exercise (Zhao et al., 2022a). Indeed, exercise is recommended as a potential strategy to manage osteoporosis (Howe et al., 2011; Pagnotti et al., 2019). Studies have shown that mechanical loads with high-magnitude strains of weight-bearing loading at high rates or frequencies are critical for bone responses (Kohrt et al., 1995; Rubin and Lanyon, 1985). A clinical trial involving high-intensity, progressive resistance and impact weight-bearing training (HiRIT) demonstrated that such a form of exercise enhanced indices of bone strength and functional performance in postmenopausal women with low bone mass (Watson et al., 2018). Further, eight months of HiRIT training had no adverse effects in 101 women aged 65±5 years with, or at risk of osteoporosis. More importantly, this type of weight-bearing exercise also improved functional and neuromuscular performance measures. These results further suggest a possible projection of the signals in the hypothalamus from skeletal interoception to the higher levels of the brain. Thus, a deeper mechanistic understanding of skeletal interception may provide insight into novel therapeutic approaches for musculoskeletal-related diseases, including improving the outcome of weight-bearing exercise in the management of such diseases.

Besides bone and joints, other components of the musculoskeletal system, including muscle, tendon and ligament, also interact with the CNS. There is growing evidence that suggests that skeletal muscles can act as an endocrine organ to influence the function of the brain and other organs (Pedersen, 2019; Severinsen and Pedersen, 2020). Cathepsin B derived from skeletal muscle cells passes through the blood–brain barrier to stimulate brain-derived neurotrophic factor (BDNF) production, resulting in hippocampal neurogenesis and enhanced cognitive functions (Moon et al., 2016). Exercise induces FDNC5 gene expression in skeletal muscle and increased serum irisin levels, which stimulates BDNF in the hippocampus (Wrann et al., 2013). Upon skeletal muscle contraction, IL-6 is released into the blood, impinging on the central regulation of appetite (Febbraio and Pedersen, 2002; Timper et al., 2017). Clinical studies have shown that tendon or ligament injuries can affect mood and pain perception through the corticospinal pathway (Baez et al., 2021; Rio et al., 2016; Rodriguez et al., 2021). The underlying mechanism are expected to identified to provide sufficient evidence to support the interoception circuit of muscle, tendon and ligament.

It is believed that locomotion improves memory and cognition (Mortimer and Stern, 2019). Skeletal interoceptive signals arrive at the hypothalamus, where they then project to different nuclei in the brain for integration and interpretation to coordinate physiological activity. However, we still know little about the central projection pattern in skeletal interoception. Recently, Minhas’s group found increased production of PGE2 by macrophages in both the brain and periphery, especially in the bone marrow, and that activation of the receptor EP2 induces suppression of oxidative phosphorylation and glycolysis in ageing macrophages and microglia, resulting in cognitive impairment (Minhas et al., 2021). Although PGE2 was speculated to influence brain function via the circulatory system, skeletal interoception may also be involved in this process. This notion dovetails with our belief that skeletal interoception will become a potential target for the treatment of metabolic disorders and age-related diseases, including osteoporosis and Alzheimer’s disease.

Due to the complex interactions between bone and the peripheral nervous system, specific skeletal cells beyond osteoblasts and macrophages that may communicate with peripheral nerve fibers to activate skeletal interoception need to be identified. Likewise, interoceptors other than EP4 that mediate skeletal interoception, and whether they go awry in pathology, require further investigation. Indeed, the COX2 gene and different types of PGE2 receptors are widely expressed in the CNS (Hoffmann, 2000; McCullough et al., 2004; Minghetti, 2004). It will be interesting to investigate whether their expression is associated with skeletal interoception. For example, the PGE2 receptor EP2 is expressed in the DRG, dorsal horn and brain, and EP2 has been shown to mediate PGE2-induced pain (Kras et al., 2013; Liu et al., 2019). But whether EP2 is activated by high levels of PGE2 to mediate skeletal interoception has not been explored.

Most recent research shows the multidimensional coding systems of interoception, which provides new ideas for further investigation (Choi et al., 2018; Ran et al., 2022; Zhao et al., 2022b). New technologies, such as single-cell sequencing, whole-tissue imaging, chemical genetics and optogenetics, will help to elucidate novel skeleton-related neurocircuits in the CNS (Moffitt et al., 2018; Prescott et al., 2020; Wang et al., 2021). Further, dynamic monitoring of neuronal activity with calcium imaging or fiber photometry can identify the subtypes of afferent neurons responsible for different skeletal interceptive signals. Such approaches could help explore if neuropathy under pathological conditions, such as aging, obesity and diabetes, affect skeletal interoception.

Along these lines, and as compared with the progress made in exteroception, research in interoception is still in the early stage. A Blueprint for Neuroscience Research Workshop entitled “The Science of Interoception and Its Roles in Nervous System Disorders” was hosted by the US National Institutes of Health (NIH) in 2019. In particular, a request for application “Notice of Special Interest (NOSI): Promoting Research on Interoception and Its Impact on Health and Disease” was released by the NIH in 2021. There is no doubt, then, that interoception will certainly become one of the mainstream research efforts in the physiology and pathology fields, as well as in neuroscience. Such research should also open new avenues in the skeletal biology field, hopefully leading to new and effective therapies for bone-related diseases and pain.