The effects of microgravity on bone structure and function

Humans are spending an increasing amount of time in space, where exposure to conditions of microgravity causes 1–2% bone loss per month in astronauts. Through data collected from astronauts, as well as animal and cellular experiments conducted in space, it is evident that microgravity induces skeletal deconditioning in weight-bearing bones. This review identifies contentions in current literature describing the effect of microgravity on non-weight-bearing bones, different bone compartments, as well as the skeletal recovery process in human and animal spaceflight data. Experiments in space are not readily available, and experimental designs are often limited due to logistical and technical reasons. This review introduces a plethora of on-ground research that elucidate the intricate process of bone loss, utilising technology that simulates microgravity. Observations from these studies are largely congruent to data obtained from spaceflight experiments, while offering more insights behind the molecular mechanisms leading to microgravity-induced bone loss. These insights are discussed herein, as well as how that knowledge has contributed to studies of current therapeutic agents. This review also points out discrepancies in existing data, highlighting knowledge gaps in our current understanding. Further dissection of the exact mechanisms of microgravity-induced bone loss will enable the development of more effective preventative and therapeutic measures to protect against bone loss, both in space and possibly on ground.


INTRODUCTION
Extended human spaceflight was once a distant fantasy; however, it is now almost a tangible reality. With NASA's goal to send humans back to the moon by 2024, then onwards to Mars in the 2030s, it is now more critical than ever to understand the impacts of long-term space travel on human health 1 . Among the many technical, logistical and physiological challenges inherent to extended space exploration, the loss of gravitational force is a major prohibitive environmental factor that adversely affects the body of space travellers. The human body is intrinsically adapted to Earth's gravity (~9.907 m/s 2 ), thus exposure to conditions of reduced gravity, or microgravity (µG) can lead to a plethora of complications in normal bodily functions. µG decreases the effort required for movement, while causing mass fluid redistribution 2 . As a result, muscles in the arms and legs experience atrophy 3 , the cardiovascular system is compromised 4 , the immune system is suppressed 5 , and increased cranial pressure leads to vision problems and neurological impairments 6 . Exposure to µG also results in skeletal deconditioning, where significant reductions in bone mass increases the risk of fractures and osteoporosis, threatening the viability of long-duration missions and astronauts' mobility upon return to Earth 7,8 .
A brief background to bone homeostasis Before examining how bone is affected by microgravity, it is important to understand bone function in 1G. Under normal circumstances, bone remodeling is an adaptive and balanced process where bone resorption and formation are coupled to regulate homeostasis of bone tissue 9 . The overall process relies on osteoblasts and osteoclasts acting in concert to regulate bone formation and resorption, respectively. The inactive bone surface is lined with flat remnants of osteoblasts, where they serve as a membrane capable of detecting hormones and/or mechanical loading to initiate the bone remodeling process [9][10][11][12][13] (Fig. 1). Once circulating osteoclast precursors are recruited to sites of bone remodeling, they differentiate into mature osteoclasts that secrete enzymes such as cathepsin K and metalloproteinase to digest the collagen-rich bone matrix 14 . This degradation process also releases calcium and embedded growth factors, such as bone morphogenic proteins (BMPs) and transforming growth factor-β (TGFβ), which contribute to bone formation 9,13 . Once the cavities beneath resorbing osteoclasts reach a certain size, osteoclasts undergo apoptosis to terminate bone resorption and prevent excess bone loss 15 .
The newly liberated growth factors from bone degradation can recruit and stimulate the differentiation of mesenchymal stem cells (MSCs) to osteoblast lineage cells, including osteoprogenitors, osteoblasts and osteocytes 16 . Maturing osteoprogenitors and preosteoblasts secrete a variety of matrix proteins, such as type 1 collagen, as well as non-collagen proteins (osteocalcin, osteonectin, bone sialoprotein II and osteopontin) and proteoglycans, which are mineralised by mature osteoblasts 17 . In addition, preosteoblasts express alkaline phosphatase (ALP) for bone mineralisation, hence its expression and activity are key markers of osteoblast differentiation and maturation 18,19 . Osteoblasts either undergo apoptosis, differentiate into quiescent bone lining cells, or become embedded in the bone matrix to form osteocytes, which form a canalicular network of branched dendritic processes 20,21 . They communicate with bone lining cells, osteoblasts and other osteocytes, and are suggested to influence bone remodeling in response to mechanical loading 17,22 . Thus, with an understanding of bone homeostasis, what is currently known regarding the effect of microgravity on bone will be reviewed below.

THE EFFECT OF MICROGRAVITY ON BONE Microgravity-induced bone loss in humans
The first observation of µG-induced bone loss was recorded in the mid-1970s, when Skylab crew members demonstrated the loss of 1-2% bone mass per month compared to pre-flight and ground controls [23][24][25] . Since then, despite the implementation of preventative exercises, bone loss in space has been one of the most frequently observed outcomes among astronauts ( Table 1, row 1-7). The weightlessness experienced in microgravity reduces the loading on weight-bearing bones, resulting in adaptive changes that increase bone resorption and inhibit bone formation 26 . Indeed, bone mineral density (BMD) studies of astronauts demonstrate substantial decrease in the mass of weight-bearing bones such as the tibia, but not in non-weight-bearing bones like the distal radius 8,[27][28][29] . Bone resorption is particularly exacerbated in the first 2 weeks of spaceflight, where urinary concentrations of resorption markers such as N-telopeptide and pyridinium crosslinks are increased 26,[30][31][32][33] . On the other hand, urinary calcium levels are increased 24,32,34 , indicating reduced calcium absorption in astronauts 32,34 . As bone formation is reportedly unchanged or decreased 30,32,33 , this results in an overall negative calcium balance that contributes to bone loss in space. Notably, a recent study suggests that circulating biomarkers of bone turnover preflight can predict the severity of in-flight bone loss, where astronauts with elevated bone resorption and formation markers pre-mission experience greater losses in BMD and strength of their distal tibia during spaceflight 29 .
The severity of bone loss also increases with spaceflight duration, and the time required for recovery to pre-flight BMD levels is reportedly longer than the actual mission 27,28 . Another study evaluates the bone mass, microarchitecture and strength of 13 astronauts who spent 4-6 months aboard the International Space Station (ISS) 35 . This study monitored skeletal recovery of each astronaut for up to 12 months post-landing. Although the cortical bone thickness and density of the weight-bearing distal tibia eventually recover upon landing, the cortical porosity and trabecular bone fail to recover, leading to reduction in the ultimate load of the bone 35 . Congruent to previous findings 8, 36,37 , the non-weight-bearing distal radius is preserved at landing 35 Fig. 1 The stages of bone remodelling. Bone remodelling is a process where cycles of bone resorption and formation are separated by periods of quiescence. During quiescence, the relatively inactive bone surface is lined by flat remnants of osteoblasts. Events such as hormone detection and/or mechanical loading can activate the recruitment of circulating osteoclast precursor cells. These precursor cells fuse to form premature osteoclasts and migrate to the bone surface, while bone lining cells retract to enable preosteoclast binding. Once bound to the bone matrix to form a sealing zone in the isolated area, they differentiate into mature osteoclasts for bone resorption. Mature osteoclasts secrete protons to create an acidic environment that dissolves bone mineral, and proteolytic enzymes to digest the bone matrix. The resorption process results in the formation of cavities, also known as Howship's lacunae, beneath active osteoclasts. Osteoclasts undergo apoptosis once these cavities reach a certain size, leading to the termination of bone resorption. The bone degradation process also releases embedded growth factors that reverses bone resorption by recruiting and stimulating the differentiation of mesenchymal stem cells (MSCs) into bone-forming osteoblast lineage cells. Once recruited to the lacunae, preosteoblasts secrete a variety of matrix proteins in the organic bone matrix, or the osteoid, which are then mineralised by mature osteoblasts. Bone formation is terminated upon completion of mineralisation. Osteoblasts either undergo apoptosis or differentiation into quiescent bone lining cells. Alternatively, osteoblasts can become embedded in the bone matrix to form osteocytes, which form a canalicular network of branched dendritic processes to communicate with bone lining cells, osteoblasts, and other osteocytes.
progressive fragility 6 months after landing, which coincided with bone remodeling markers declining to below pre-flight levels between 6 and 12 months upon return 35 . The observation of progressive fragility in non-weight-bearing bones might have escaped other studies due to inadequate length of recovery monitoring. Future studies should increase follow-up duration to validate this phenomenon.

Microgravity-induced bone loss in animals
Despite the physiological relevance, relying solely on astronaut data to understand µG-induced bone loss is limiting for many reasons. One of which is the rarity of human spaceflight and the limited number of astronauts per mission, thus leading to the challenge of small sample sizes in these studies. The smaller physical sizes of animals such as rats, mice and fish allow for more compact storage in space missions, enabling larger sample sizes while maintaining some physiological relevance. Similar to observations in astronauts (Table 1), µG exposure for 8 days decreases overall bone volume and thickness by 6.3% and 11.9%, respectively, in 15 mice 38 . The negative effects of µG on trabecular bone mass are also reportedly discernible in weight-bearing bones such as tibia, femur and vertebrae 39 , complementing observations in human studies. Seven rats aboard the Soviet mission COSMOS 1667 for 7 days demonstrate a reduction in tibial trabecular bone volume by 47-55%, trabecular thickness by 20-24% and density by 40-43% compared to ground controls 36 . In addition, another study observes a 64% decrease in femoral trabecular bone volume and a 140% increase in bone resorption in 16 mice following 30 days of spaceflight compared to control mice on Earth 40 . Congruently, a larger sample size of 40 mice subjected to 22 days of µG exposure exhibit significant reductions in BMD of both left and right femur compared to ground controls, albeit to a smaller extent of 11% and 8% respectively 41 .
On the contrary, there is opposing evidence that suggests µGinduced bone loss may not be strictly related to its weight-bearing nature 41 . The aforementioned study also reports that BMD of the humerus, a weight-bearing bone, is not affected in spaceflight mice compared to ground controls despite changes in both femurs 41 . Similarly, 14 days of spaceflight reduces the BMD of femurs, but has no effect in the humerus of 12 rats compared to controls on Earth 42 . It should be noted that these rats are ovariectomized 42 , which amplifies phenotypes of bone loss as ovarian hormones are crucial regulators of skeletal growth 43 . The relationship between µG and the weight-bearing nature of bone loss remains unclear given the contradicting evidence presented. Nevertheless, these studies collectively highlight the damaging effects of µG on bone health during space travel, as well as the risk of premature osteoporosis from the uncertainties of post-flight skeletal recovery.

Limitations of studying bone loss in space
Spaceflight experiments are certainly the most physiologically relevant method of studying µG-induced bone loss, however, they come with significant limitations. Firstly, mission launches are infrequent, hence opportunities for experiments in space are limited. Secondly, experimental design must comply with cargo weight and space constraints in both the launch and space station. This often leads to small sample sizes and minimal, if any, biological replicates, which reduces statistical power and rigour of the research. Thirdly, the engineering components must be sound, ensuring that the experimental model survives the harsh launch conditions, the duration of data collection in orbit, and in some cases, the return to Earth. The strategies to overcome or bypass these restrictions ultimately feed into the final limitationcost. Experiments requiring astronaut intervention incur even higher financial burdens, hence simplistic autonomous or remotely controlled experiments are favoured. In light of these limitations, many researchers elect to study µG-induced bone loss using simulated µG here on Earth.

THE EFFECT OF SIMULATED MICROGRAVITY ON BONE Current microgravity simulation technologies
To address the logistical problem of studying bone loss directly from humans in space, several ground-based methods have been developed since the 1970s to subject various models, such as cells, plants, animals and humans, to near-µG conditions. The following discussion will be limited to technologies relevant to research on µG-induced bone loss (Table 2); a wider scope of methods is thoroughly reviewed by Ferranti et al. 44 . Horizontal bed rest/head-down tilt. Initially, horizontal bed rest (HBR) was used to mimic the inactivity of the human body in weightlessness. However, this method failed to recapitulate fluid redistribution towards the head as observed in inflight astronauts. As such, researchers experimented with tilting the subjects towards the head to encourage cephalad redistribution from the legs [45][46][47] . Head-down tilt (HDT) angles range from 4°to 15°, but a tilt angle of 6°which approximates to 0.1 G became the analogue for µG simulation in most human bed rest studies 48 . Comparisons between HBR/HDT and µG are thoroughly discussed by Hargens et al. 49 .
Water immersion. While HBR/HDT simulates microgravity by minimising G forces, water immersion achieves microgravity via neutral buoyancy. Neutral buoyancy describes the displacement of a medium which a mass is immersed in, resulting in the balance of gravitational force. During water immersion, the subject usually sits in water with a temperature of 34-35°C. Such facilities have been utilised since the 1960s to prepare astronauts for spaceflight, as well as in experiments involving µG simulation 50 . However, immersing subjects in water for prolonged periods can lead to adverse cutaneous effects 51 . As such, the technique of dry immersion was introduced.
Dry immersion. In keeping with the neutral buoyancy concept of water immersion, subjects are immersed in a bath to neck level, but are "kept dry by the use of a waterproof, highly elastic cloth" 52 . While some localised pressure can be experienced at the seat and/ or feet during the conventional water immersion, the buoyant force from the air between the elastic cloth and skin in dry immersion effectively prevents any localised surface pressures. Furthermore, dry immersion enables longer experiments for up to 56 days, while reproducing µG-induced physiological effects similar to HDT experiments 53 . The history, utilisation and effects of dry immersion is extensively reviewed by Tomilovskaya et al. 53 .
Rodent HLU. The hindlimb unloading (HLU) model is the most common method to simulate spaceflight conditions in rodents, where rats or mice are suspended by their tails, or the use of surgical pins or body harnesses 54 . This causes the removal of  59 Krause et al. 60 Svejgaard et al. 62 Loon 63 Wuest et al. 64

RPM (cells)
• Robust body of studiesbut less than RCC/RWV • Near true µG • Programmable to simulate gravity of any planet • Susceptible to shear forces and vibration • Samples further away from centre stage experience centrifugal force • Tendency for multicellular aggregate formation in non-adherent cell samples • Pause in rotation required for media change in long-term studies • Not applicable for examining acute responses to µG Brungs et al. 59 Krause et al. 60 Svejgaard et al. 62 Loon 63 Wuest et al. 64 Freefall machine (cells) • Designed to measure acute responses to µG • Short window of µG • Adverse effects during hyper-gravity window in between falls • Small body of work Schwarzenberg et al. 57 Mesland et al. 56 This is not an exhaustive list of techniques available, only microgravity simulators relevant to bone-loss experiments are discussed here. HBR horizontal bed rest, HDT head-down tilt, µG microgravity, HLU hindlimb unloading, RWV rotating wall vessels, RPM random positioning machine.
J. Man et al. mechanical loading from the hindlimbs, as well as head-ward fluid shifts comparable to the human HDT model 55 . Despite the hindlimb suspension, it should be noted that internal organs remain affected by gravity, thus HLU presents as a limited model for µG.
There are 2 common models of replicating µG in research involving small biological samples like cell cultures: free fall state and clinorotation. When cells reach terminal velocity during a state of free fall, they are unable to respond to gravitational force. As such, the cells are in a state of functional weightlessness. The Free Fall Machine (FFM) involves dropping a sample in a long vacuum tube, resulting in a free fall for 900 milliseconds (ms) before being propelled back to the top by a current of air at~2-20 times normal gravity for 80 ms 56 . This theory relies on the assumption that cells do not have sufficient time to react to the small hyper-gravity windows between free falls 56 , which was later disproved 57,58 .
Clinorotation models µG by rotating a sample with enough speed to disable adaptation of gravity vector, but slow enough to prevent generation of centrifugal shear forces. Depending on the number of rotation axes, clinostats can be classified into 2 major classes: 1D/2D Clinostat -RWV. 1D clinostats rotate the specimen along its vertical axis, while 2D clinostats rotate the sample on a plane perpendicular to the rotation axis. These clinostats come in varying rotation speed and vessel body sizes. Rotating wall vessels (RWVs) are 2D clinostats with a larger body (5-20 cm in diameter) and a slower rotation speed (10-20 rpm). The culture media rotates at the same angular velocity as the rotating vessel wall, creating laminar fluid flow that minimises shear stress. The rotation frequency also prevents particle sedimentation, such that the cells remain at the centre of the vessel in near zero gravity.
3D Clinostat -RPM. Random positioning machines (RPMs) consist of a sample area mounted at the centre of two frames that rotate independently with randomised speed and direction. The sample is therefore constantly reoriented randomly such that the gravity vector is averaged to near-zero. However, this only applies to the intersection of the two rotational axes. Accelerative  forces from the rotation become stronger the further samples stray from centre, hence attention to sample positioning is critical on the RPM. Moreover, it has been shown that the RPM can induce stress responses in gravity-sensitive cell systems from its small vibrations and shear forces 59,60 . This can lead to cell detachment, and promote the formation of multicellular spheroid structures from increased intercellular interaction due to gravitational unloading 61 . Although, these effects were also observed in other clinostat types during multiple comparisons of 2D and 3D clinostats [62][63][64] . As such, studies involving either type of clinostat are widely accepted as models of microgravity.
Human bone loss induced by simulated microgravity So far, bone loss experiments under simulated-µG largely confirm the data collected in space. Bone resorption symptomatic of longduration space missions is widely reproduced in bed rest studies on Earth 65-71 (Table 3). Following 4, 14, or 30 days of 6°HDT, bone resorption is increased in test subjects as reflected by an increase in bone resorption markers, as well as an overall calcium imbalance 72,73 . Interestingly, bone resorption levels fail to be rescued despite increasing dietary calcium intake 74 . As previously mentioned, ALP is responsible for bone mineralisation. Consistent with the reportedly unchanged or decreased bone formation in space travellers 30,32,33 , levels of ALP remain constant 72 or even decreased 73 in HDT test subjects. Together, overall bone loss and alterations in biomarkers of bone turnover induced by simulated µG appears consistent with spaceflight data. On the other hand, there are some discrepancies in current literature describing how simulated-and true-µG affects different bone types and their compartments. One on-ground study subjected participants to a 60-day bed rest with 6°HDT, simulating the inactivity and fluid shift experienced in µG. The cortical area, thickness and density of the distal tibia in test subjects is decreased 75 , supporting the theory that µG negatively affects load-bearing bones more than non-weight-bearing bones 8,[27][28][29] . In contrast to existing spaceflight data that indicates µG has little/ no effect on non-weight-bearing bones 8,[27][28][29] , HDT causes an increase in cortical area, thickness and bone density of the distal radius in subjects, despite a reduction in the trabecular area 75 . It is possible that an increased use of hands during the HDT study could be responsible for alterations in bone loading patterns, and is reflected in bone growth of the non-load-bearing distal radius. However, a meta-analysis of homogenous bone-related datasets from Gemini, Apollo, Soyuz, Skylab, Salyut, STS, Mir and ISS missions suggests a trend of underreporting positive bone density changes in the upper-limb and thorax region 76 . Taking into account data from both spaceflight and simulated-µG, the relationship between µG and the weight-bearing nature of bone or its compartments remains unclear. This calls for further clarification in future investigations. Some advantages of studying bone-loss recovery in simulated-µG over spaceflight observations are that ground-based alternatives offer more accessible measurement timepoints, testing options, and longer post-µG-exposure monitoring. As such, they provide further insights into the limited understanding of the recovery process post-landing obtained from crewmembers alone. Consistent with post-landing data regarding the distal tibia, its cortical compartment generally recovers by 1 year following exposure to simulated-µG 75,77,78 . The recovery time of the tibia cortical thickness appears to be shorter in females (90 days) 78 compared to males (180 days) 75 , despite both cohorts having been subjected to 60 days of 6°HDT. A combination of bed rest and unilateral lower limb suspension (ULLS) studies indicate that both cortical and trabecular compartments of the tibia suffer initial deterioration for the first month post-reambulation, with the cortical compartment to a larger extent 77 . However, unlike the cortical area of the tibia, the trabecular compartment does not recover even after 1-2 years post-ambulation 75,77,78 . Although the trabecular compartment of the tibia appears to begin recovering 3-6 months post-HDT 75,79 , this is followed by a reversal in BMD to −2% from 6 to 12 months in females 78 and −1% from 3 to 24 months in males 75 . It is unlikely that these observations are related to age or biological sex, as biological sex of the two cohorts are balanced (N = 24 males and N = 24 females) with a similar age bracket (20-to 45-year-old males and 25-to 40-yearold females). It should be noted that similar to spaceflight data, recovery of non-weight-bearing bones such as the distal radius following exposure to simulated-µG is relatively underreported compared to weight-bearing-bones such as the hip and the distal tibia. Existing datasets can be reanalysed for radial bone data, and future studies can also direct more focus on the recovery of nonweight-bearing bones.

USING ANIMAL AND CELL MODELS TO UNDERSTAND MICROGRAVITY-INDUCED BONE LOSS
The establishment of cellular and animal models have greatly improved our understanding of the molecular mechanisms behind µG-induced bone loss. The following sections review our knowledge to date, comparing cell and animal data between spaceflight (Table 4) and simulated-µG (Table 5) experiments.

Osteoblasts
One of the most reported observations is changes in osteoblast cell morphology, which is reflected in actin-related cellular structures. Exposure to µG on a 4-day spaceflight causes collapse of the actin cytoskeleton in quiescent osteoblasts, with some cells exhibiting a spindle shape 80 , and some becoming rounder and covered with microvilli 81 . In addition, µG-exposure results in altered nuclei morphology; one study reports elongated nuclei with a 30% size reduction in spaceflight osteoblasts 82 , and another Summary of observations in animals and bone cells from various studies during spaceflights compared to respective ground controls. GC ground control, TRAP tartrate-resistant acid phosphatase, MMP-9 matrix metallopeptidase 9, BDSC blood-derived stem cells, BMSC bone marrow stem cells, Sox2 sex determining region Y-box 2, Rac1 Ras-related C3 botulinum toxin substrate 1, Cdc42 cell division control protein 42 homolog, µG microgravity, PTHrP parathyroid hormone-related protein, RANKL receptor activator of nuclear factor κB ligand. observes more condensed and fragmented nuclei compared to ground controls 83 . A symptom of altered actin cytoskeleton is the reduction in number of stress fibres and smaller stress fibre area in spaceflight osteoblasts 82,84 , indicating that µG impairs actin polymerisation. Focal adhesions are another mechanosensitive structure composed of polymerised actin, which appears to be destabilised upon exposure to µG. Consistent with the decrease in stress fibre formation, multiple studies observe reductions in the number of focal contacts and focal adhesion area in osteoblasts following spaceflight 80,83,84 . Moreover, focal contacts established Summary bone-related observations in animal and cellular models from various microgravity-simulation studies. HLU hindlimb unloading, IL-6 interleukin 6, BMD bone mineral density, ALP alkaline phosphatase, Runx2 Runt-related transcription factor 2, NFB nuclear factor kappa-light-chain-enhancer of activated B cells, RANKL receptor activator of nuclear factor kΒ ligand, Wnt wingless/integrated, RWV rotating wall vessel, MSC mesenchymal stem cell, PPARγ Peroxisome proliferator-activated receptor γ, RPM random positioning machine, BMP2 bone morphogenic protein 2, RCC rotary cell culture, COL1A1 Collagen type I alpha 1 chain, Sox9 SRY-box transcription factor 9, DMP1 dentin matrix acidic phosphoprotein 1, AP-1 activator protein 1, PthR1 parathyroid hormone 1 receptor, OPG osteoprotegerin, MITF melanocyte inducing transcription factor; CREB cAMP response element-binding protein, SOST gene encoding sclerostin.
during spaceflight appear less mature than those formed in ground control osteoblasts 83 , confirming that µG negatively impacts osteoblast adhesion. Since cytoskeletal integrity is critical in signal transduction and expression of genes that regulate cell cycle, the loss of both actin cytoskeletal structure and cell adhesion can deter osteoblast proliferation in space. Indeed, µG delays cell cycle initiation in quiescent osteoblasts compared to ground-based analogues 82,85 . Human bone marrow stem cells (hBMSCs) aboard the ISS for 14 days encounter cell cycle arrest despite initial osteoblastic differentiation, and the phenotype of terminal differentiation to osteocyte is inhibited 86 . In support of this, spaceflight osteoblasts also demonstrate reduced osteocalcin and type I collagen expression, indicating reduced differentiation and matrix development 87 . However, it should be noted that despite the impairment of osteoblastic proliferation and differentiation under µG, genome-wide and Next Generation Sequencing analysis does not suggest apoptosis or cell senescence 86 . The authors hypothesised that BMSCs or immature osteoblasts respond to µG by reverting to a quiescent state, in line with the observations of increased BMSC differentiation potential seen in µG-exposed mice following reloading on Earth gravity 88,89 .
Congruent to spaceflight observations, most of the current literature consistently suggest that simulated-µG inhibits the proliferation and differentiation of MSC towards osteoblasts, despite using different methods of µG simulation (Table 5). This is indicated by the suppressed gene expression of osteoblast differentiation markers such as bone morphogenic protein (BMP) and Runt-related transcription factor 2 (Runx2) 18,[90][91][92][93][94][95][96][97][98][99][100][101] , in addition to the lowered expression of other osteoblastogenesis-related genes 18,97,99,[101][102][103] . The marker for osteoblast maturation from MSC progenitors -ALP, is also expressed at lower levels in preosteoblasts and osteoblasts exposed to various forms of simulated-µG [95][96][97]99,104 . Subsequently, enzymatic activity of ALP is reduced in these cells 18,99,103 . Indeed, MSCs extracted from femurs of rats subjected to HLU for 28 days show reduced Runx2 and ALP mRNA expression, as well as decreased ALP activity 105 . However, the authors did not mention whether these changes in biomarker expressions correlated to any physical BMD changes in the rats.
The reduced expression and activity of ALP in simulated-µGcells correlates well with decreased secretion of matrix proteins such as type 1 collagen and osteocalcin [95][96][97]99 , which could lead to reduced production of mineralisation crystals 93,94 . It is suggested that simulated-µG prevents extracellular calcium from entering osteoblasts, thus reducing intracellular free calcium levels which impairs calcium deposition and bone formation 102,106 . Furthermore, simulated-µG has been shown to cause cell cycle arrest at G2 and even induce apoptosis in osteoblasts 100,102 . HLUinduced bone loss in mice is consistently accompanied by a reduction in the number of viable osteoblasts and osteocytes in mice 107 . It is of note that some studies observed morphological changes in osteoblasts treated with simulated-µG, where actin cytoskeleton disruptions cause "bulging" or "spheroidal" morphologies 94,106,108 . However, while Hu et al. have reported simulated-µG-induced inhibition on preosteoblast differentiation, no alterations in cell morphology is observed 96 . Of note, studies that report morphological changes subjected bone cells to extended periods of simulated-µG (up to 20 days) 94,106,108 , compared to the 7-dayexposure period by Hu et al. 96 . One study attributes the inhibition of osteogenesis in simulated-µG to the obliteration of primary cilia on osteoblasts cultured on the RPM 93 , suggesting that primary cilia play a sensory role in bone metabolism 109,110 . Since the actin cytoskeleton plays a critical role in osteogenic differentiation 111 and cell cycle processes 112 , it is possible these phenomena could be symptoms of µG-induced disruptions. Collectively, these observations provide strong evidence that impaired differentiation, maturation and proliferation of osteoblasts could be responsible for the µG-induced reduction in bone formation 93,113 .
Conversely, a smaller fraction of the literature suggests an opposing argument, that simulated-µG in fact promotes proliferative and differentiation capabilities in MSCs 114 , and does not directly induce osteoblast cell death 104 . RPM-induced µG causes MSCs to express increased levels of Runx2 and Osterix (Osx) 115 , which are transcription factors essential in osteoblastic differentiation. Runx2 gene expression also remains unaltered in osteoblasts cultured on the RPM compared to 1 G controls 104 . Furthermore, osteoblasts exposed to simulated-µG show elevated mRNA expression of SOX9 108 , which is a transcription factor characteristic of commitment to the osteoblastic lineage from MSC differentiation. Consequently, simulated-µG reportedly increases bone matrix protein production in MSCs and osteoblasts/osteoblastlike cells, such as osteopontin and osteocalcin gene expression, and type 1 collagen secretion 104,108,115 . It should be noted that immortalised osteoblast/osteoblast-like cell lines are used in some of these studies, and the effects of inherent mutations must be considered. Nonetheless, further investigation is required to either confirm or dispute these observations in future studies.

Osteoclasts
Mechanosensitive osteoclasts also contribute to space-related bone loss by disrupting normal bone homeostasis. Although much less studied than osteoblasts, osteoclasts have demonstrated increased resorptive activity in response to µG compared to controls on Earth 83,[116][117][118] . Mature osteoclasts cultured on ivory or bovine bone slices demonstrate an increase in number of resorption pits formed following 5 or 7 days of spaceflight, respectively, compared to their relative ground controls 83,116 . Indeed, µG induced a dramatic increase in the expression of bone resorption-related genes, as well as elevated collagen telopeptide production 116 , which is correlated with bone resorption 119 . The increase in resorptive activity is reflected in the faster differentiation and maturation of spaceflight osteoclasts. Expression of genes involved in osteoclast differentiation, such as integrin β 3 , cathepsin K, MMP-9 and calcitonin receptor are significantly upregulated following µG exposure in comparison to controls on Earth 116,117,120 . Collectively, these observations in spaceflight experiments indicate that osteoclasts play a critical role in µGinduced bone loss.
There are limited reports of osteoclast behaviour in simulated-µG, however, current literature is in consensus that simulated-µG promotes osteoclastogenesis and osteoclast function [121][122][123][124][125][126][127] . Osteoclastogenesis can be influenced by secretion of osteoprotegerin (OPG) and receptor activator of nuclear factor κB ligand (RANKL) from osteoblasts and osteocytes. RANKLan osteoclastogenesis promotor, binds receptor activator of nuclear factor κB (RANK) receptors on the surface of osteoclasts for differentiation/ maturation, while OPG acts as a decoy receptor that also binds RANKLthus serving as a negative regulator of osteoclastogenesis 128,129 . RPM-facilitated µG decreases osteoblast production of OPG, thereby increasing RANKL/OPG ratios, which enhances osteoclastogenesis 121 . Supporting this observation, RANKLstimulated osteoclastogenesis is increased in mice despite 28 days of HLU 126 . Autophagosome production reportedly enhances osteoclast differentiation, and RWV treatment enhances expression of autophagic genes 123 . This in turn promotes differentiation of osteoclasts, while inhibition of autophagy by 3-methyladenine conversely prevents osteoclastogenesis despite simulated-µGexposure 123 . The positive effect of simulated-µG on osteoclastogenesis is also evidenced by the upregulation of cytokines, growth factors, proteases, signalling proteins and transcription factors such as c-Jun, MITF and CREB in osteoclasts cultured in RWV compared to 1 G controls 122 . Mice subjected to HLU displays increased osteoclast number, elevated osteoclast surfaces in hindlimb and vertebral sites, and ultimately reduced spinal BMD and strength 127,130 . µG-stimulated osteoclastogenesis and osteoclast activity are consistently observed in both on-ground and inspace experiments, highlighting the opportunity of these processes as a therapeutic target against bone loss.

Osteocytes
Osteocytes reside in the cavities of the mineralised bone matrix, also known as lacunae (Fig. 1), where they synthesise proteins such as collagen and glycosaminoglycans which contributes to bone mineralisation 131 . µG-exposure from spaceflight is found to impede the differentiation of osteoblasts into osteocytes 131 . This causes underdevelopment in the Golgi complexes of osteocytes responsible for the secretion of matrix proteins, leading to retardation in bone matrix mineralisation 131 . Furthermore, osteocytes are reported to undergo apoptosis as early as three days into µG-exposure from spaceflight, leaving an increased number of empty lacunae 40,131 , or lacunae with reduced volume and altered shape 40 . On top of reduced bone formation/mineralisation, this may trigger further bone resorption to cause deterioration of bone microstructure and loss of bone mass 40 .
Congruent to spaceflight observations, HLU rats demonstrated an increase in apoptotic osteocytes by 66% compared controls 132 , while HLU mice displayed enhanced osteocyte apoptosis in their spine 127 . The death of osteocytes is associated with increased osteoclast activity 133,134 . Osteocytes can promote osteoclastmediated bone loss by secreting proteins such as sclerostinan inhibitor of bone formation, and RANKLa promotor of osteoclastogenesis. Simulated µG exposure increases RANKL and sclerostin gene expression in osteocytes cultured in RWV 135 , suggesting that osteocytes play a role in µG-induced bone resorption. The mechano-sensitive influence of osteocytes on osteoclasts can be demonstrated even within the same animal in a rodent HLU model. Osteocyte proteins (including sclerostin) is elevated in the unloaded femur compared to the weight-bearing forelimb of the same HLU rat, which coincides with decreased osteocyte number and increased osteoclast activity of the unloaded hindlimb compared to the loaded forelimb 136 .
The elevated sclerostin secretion by osteocytes also negatively regulates oasteoblast-mediated bone formation by antagonising BMP/Wnt signalling [137][138][139] . Simulated µG has been shown to depress the Wnt signalling pathway and downregulate cell cycle related genes such as Cyclin D1 in osteocyte-like cells, which could be partially restored by the administration of antibodies against sclerostin 140 . Indeed, the ablation of sclerostin or RANKL appears to be protective over unloading-induced bone loss in HLU mice models 107,141 . Interestingly, osteocyte-deficient mice demonstrated resistance to HLU-induced bone loss, where bone loss and microstructural deterioration is prevented despite unloading 142 . Thus, osteocytes appear to play a vital role in promoting osteoclastogenesis and bone resorption upon mechanical unloading or µG-exposure. Future studies should leverage the unique ability of osteocytes to influence both osteoblasts and osteoclasts when exploring therapeutic interventions against bone loss.
Despite being the most abundant type of bone cells, osteocytes represent the least studied cell type due to their inaccessibility from being embedded in the bone matrix. The current literature landscape also points to a general lack of methods for osteocyte isolation and culture. In addition, isolated osteocytes may not be physiologically relevant to osteocytes in vivo, as their native environment is three dimensional and complex. As such, there is a need to establish more robust protocols for osteocyte culture and characterisation.
Molecular therapies against microgravity-induced bone loss Research using cell and animal models of µG-induced bone loss has revealed various possible therapeutic targets as described above, while providing accessible and ethical platforms for the testing of molecular therapeutic agents. Multiple agents have been identified to prevent µG-induced bone loss by promoting osteoblast differentiation, maturation and activity. Rapamycin treatment of human blood-derived stem cells (hBDSCs) aboard the ISS for 3 days induces earlier differentiation towards osteogenic lineage cells compared to analogous ground controls in 1G 143 . hBDSCs exposed to µG exhibit reduced expression of embryonic markers Sox2, Oct3/4, Nanog and E-cadherin to a larger extent compared to ground controls, indicating that rapamycin induces an earlier loss of pluripotency in space than on Earth 143 . Furthermore, expression of differentiation-related transcription factors in µG-exposed hBDSCs are altered in favour towards osteogenesis 143 . Downregulation of the transcription factors Otx2 and Snail in µG inhibits osteogenic differentiation; while GATA4 and SOX17, which promote differentiation towards osteogenesis are upregulated 143 . In another study, IL-6 neutralisation prevented the reduction in Runx2 and ALP mRNA expression, as well as ALP activity induced by simulated-µG treatment 103 . Similarly, IL-6neutralisation successfully alleviated HLU-induced bone loss in mice tibia 103 . Irisin administration has also been shown to prevent µG-induced downregulation of transcription factors and proteins critical to osteoblast differentiation and activity 120 . Similar to IL6neutralisation, irisin treatment elevated Runx2 and ALP expression, and ALP activity, as well as increased osteoblast numbers and bone formation in mice [144][145][146] . These observations suggest that the administration of rapamycin, irisin and IL-6 neutralisation can prevent µG-induced downregulation of osteoblast differentiation, maturation and mineralisation activity.
Irisin has been demonstrated to prevent bone loss induced by HLU in mice 147 . As previously discussed, osteocytes can secrete sclerostin to promote osteoclast-mediated bone loss. The number of sclerostin-positive osteocytes is higher in the hindlimb of HLU mice compared to controls 130 . Upon irisin treatment, the sclerostin-positive population was significantly lowered, indicating a negative effect on osteoclast activity as an increase in bone formation rate was observed 130 . Irisin has also been demonstrated to downregulate osteoclastogenesis. Its administration reduced RANKL-induced osteoclastogenesis 146 , and decreased osteoclastcovered bone surfaces in HLU mice compared to ambulatory controls 130 . Analogously, melatonin treatment upregulated osteoclast inhibitor calcitonin and downregulated RANKL in goldfish scales 117 . However, it is unknown whether this translates into suppression of osteoclast activity in this study. Nevertheless, these molecular compounds all serve as promising therapeutic agents against µG-induced bone loss. However, further animal studies will be required prior to phase 1 human clinical safety trials.

CONCLUSION
Through the combination of spaceflight and simulation data on humans, animals and cellular models, we are continuously improving our knowledge of µG-induced bone loss. Our current understanding suggests µG affects weight-bearing bones to a larger degree than non-weight-bearing bones. However, this relationship warrants further investigation, especially with the underreporting of changes in non-weight-bearing bones in astronauts. Moreover, there are inherent biases in HDT/HBR studies, where the use of non-weight-bearing bones, such as the distal radius, may be increased in subjects. Together, these observations might mask the true effect of µG on non-weightbearing bones. Future studies can also clarify the discrepancies in the effect of µG on different bone compartments, as well as how µG influences recovery processes.
The use of cellular models has shed light into the molecular mechanisms behind µG-induced bone loss, which in turn provides potential targets for therapeutic intervention. Although bone loss stems from both reduced bone formation by osteoblasts and elevated resorption by osteoclasts, the current literature landscape is largely focused on osteoblasts. Preliminary observations of increased osteoclast activity in µG present promising therapeutic targets, as such, the role of osteoclasts in bone loss requires further elucidation. In addition, as osteocytes are capable of influencing both osteoblasts and osteoclasts, this unique position should be leveraged for potential therapeutic options. Thus, the role of osteocytes in balancing bone resorption and formation under µG also deserves further clarification. Collectively, these efforts should enable the development of more effective preventative and therapeutic measures against µG-induced bone loss, thus paving the way for safer journeys as we venture further away from Earth.

Future directions
• Extend monitoring of astronauts post-mission for at least 12 months to investigate the possibility of progressive fragility, particularly in non-weight-bearing bones such as the distal radius.
• Clarify the relationship between µG and the weight-bearing nature of bone loss, particularly whether µG affects nonweight-bearing bones.
• Clarify how µG affects different bone compartments in humans and animals.
• Investigate how the bone-loss recovery process differs between weight-bearing and non-weight-bearing bones, as well as between different bone compartments.
• Address discrepancies in the literature that points to µG promoting osteogenic proliferation and differentiation.
• Establish robust differentiation, isolation and culturing protocols for osteoclasts and osteocytes, as well as characterisation and validation of existing osteoclast/osteocyte-like cell lines.
• Dissect the mechanisms behind how osteoclasts and osteocytes contribute to µG-induced bone loss to identify possible therapeutic targets.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

DATA AVAILABILITY
No datasets were analyzed or generated in the writing of this review.

CODE AVAILABILITY
No codes were used in the writing of this review.