Mesenchymal stem cells (MSC) of bone marrow are the progenitor of osteoblasts and adipocytes. MSC tend to differentiate into adipocytes, instead of osteoblasts, with aging. This favors the loss of bone mass and development of osteoporosis. Hypoxia induces hypoxia inducible factor 1α gene encoding transcription factor, which regulates the expression of genes related to energy metabolism and angiogenesis. That allows a better adaptation to low O₂ conditions. Sustained hypoxia has negative effects on bone metabolism, favoring bone resorption. Yet, surprisingly, cyclic hypoxia (CH), short times of hypoxia followed by long times in normoxia, can modulate MSC differentiation and improve bone health in aging.

To evaluate the CH effect on MSC differentiation, and whether it improves bone mineral density in elderly.

MSC cultures were induced to differentiate into osteoblasts or adipocytes, in CH (3% O₂ for 1, 2 or 4 h, 4 d a week). Extracellular-matrix mineralization and lipid-droplet formation were studied in MSC induced to differentiate into osteoblast or adipocytes, respectively. In addition, gene expression of marker genes, for osteogenesis or adipogenesis, have been quantified by quantitative real time polymerase chain reaction. The in vivo studies with elderly (> 75 years old; n = 10) were carried out in a hypoxia chamber, simulating an altitude of 2500 m above sea level, or in normoxia, for 18 wk (36 CH sessions of 16 min each). Percentages of fat mass and bone mineral density from whole body, trunk and right proximal femur (femoral, femoral neck and trochanter) were assessed, using dual-energy X-ray absorptiometry.

CH (4 h of hypoxic exposure) inhibited extracellular matrix mineralization and lipid-droplet formation in MSC induced to differentiate into osteoblasts or adipocytes, respectively. However, both parameters were not significantly affected by the other shorter hypoxia times assessed. The longest periods of hypoxia downregulated the expression of genes related to extracellular matrix formation, in MSC induced to differentiate into osteoblasts. Interestingly, osteocalcin (associated to energy metabolism) was upregulated. Vascular endothelial growth factor an expression and low-density lipoprotein receptor related protein 5/6/dickkopf Wnt signaling pathway inhibitor 1 (associated to Wnt/β-catenin pathway activation) increased in osteoblasts. Yet, they decreased in adipocytes after CH treatments, mainly with the longest hypoxia times. However, the same CH treatments increased the osteoprotegerin/receptor activator for nuclear factor kappa B ligand ratio in both cell types. An increase in total bone mineral density was observed in elderly people exposed to CH, but not in specific regions. The percentage of fat did not vary between groups.

CH may have positive effects on bone health in the elderly, due to its possible inhibitory effect on bone resorption, by increasing the osteoprotegerin / receptor activator for nuclear factor kappa B ligand ratio.

The rates of diseases and conditions affecting older individuals have noticeably increased since the Age of Enlightenment, due to an extension in human longevity[1]. Old age is associated, among other factors, with a decline in bone mass and strength. It can culminate with osteoporosis, which provokes significant morbidity and mortality, increasing healthcare costs worldwide[2]. In this sense, accumulating evidence indicates a complex relationship between adiposity and osteoporosis[3]. Osteoblasts and adipocytes originate from the common-precursor mesenchymal stem cells (MSC). The balance between differentiation of osteoblasts and adipocytes is regulated by both intra- and extra-cellular factors[1]. Aging alone alters the fate of MSC in bone marrow, by promoting adipogenesis and reducing osteoblastogenesis[4], increasing age-related bone loss[5].

Also, proinflammatory cytokines associated with obesity and aging regulate receptor activator for nuclear factor kappa B ligand/receptor activator for nuclear factor kappa B/osteoprotegerin (RANKL/RANK/OPG) pathways, stimulating osteoclast activity and bone loss[6]. Osteoclastogenesis is induced in osteoclast precursor cells when RANKL, produced by osteocytes and osteoblasts, binds to its RANK receptor. However, osteoblasts also express OPG, which encodes a RANK decoy, and thus inhibits RANKL binding and activation of osteoclastogenesis. Therefore, the relationship between the expression of OPG and RANKL genes regulates bone resorption[7]. Interestingly, oxygen variations have also been established as another factor that could influence tissue remodeling[8]. Bone marrow is maintained in a hypoxic state in the body[9]. Besides, various genes regulated by the hypoxia inducible factor 1α (HIF-1A) mediate different effects[10].

Although little is known about the metabolic pathways used by bone cells, the glucose metabolism may play a prominent role during osteoblast differentiation[11]. Even in the presence of oxygen, bone cells metabolize glucose by a phenomenon known as the Warburg effect or aerobic glycolysis[10]. On the other hand, HIF-1A regulates bone remodeling, involving genes as vascular endothelial growth factor, erythropoietin and OPG[12]. The former represents a key player in coupling angiogenesis and osteogenesis[10], as well as promoting osteogenesis[13]. On the other hand, erythropoietin has also been shown to stimulate bone formation and repair. Similarly, to this mechanism of physiological hypoxic, some evidence has shown the relevance of inhalation of air with reduced oxygen content. Thus, hypoxia is a nonpharmacological method used as prevention and treatment strategy in many clinical conditions[8,14]. However, the pattern (intermittent or chronic) and the severity (duration and intensity) of hypoxia episodes are key determinants of its physiological impact[13,15,16].

Many in vitro experiments have studied the effects of hypoxia on proliferation and differentiation of human MSC, along adipogenic and osteogenic lineages, with varying results[17-19]. Adipogenic and osteogenic differentiation were severely impaired under severe hypoxia, but increasing oxygen levels from 1% to 3% restored the osteogenic differentiation capacity of cells[9,20]. Furthermore, hypoxia regulates the differentiation of MSC along osteogenic lineages in a time-dependent manner, reducing it in continuous hypoxic conditions[20]. According to aforementioned studies, in vivo research on animal and human development suggests that chronic severe hypoxia[21-23] may lead to deleterious consequences on bone system. Likewise, moderate intermittent hypoxia may lead to positive adaptations[13,24]. However, there is currently an imprecise terminology about the various types of intermittent hypoxia exposure[25]. That is usually associated with sleep-disordered breathing. It is different from cyclic hypoxia (CH), defined as periodic exposure to hypoxia, from a few minutes to days. That is repeated over a number of days, being interrupted by normoxic exposure, or lower levels of hypoxia[26].

Many cycles of hypoxia and re-oxygenation activate HIF-1A during intermittent exposure. But HIF-1A-strictly accumulates during periods of hypoxia, in cyclic exposure. That represents an essential aspect of the beneficial effects of cyclic hypoxia[25]. Thus, in contrast to severe of intermittent hypoxia (e.g., cells cultures under < 2% O₂ or 3%-8% inspired O₂), moderate CH protocols (cells cultures under 2%-9% O₂ or 9%-16% inspired O₂) may reinforce the immune system, while suppressing the production of pro-inflammatory mediators. Therefore, CH holds potential for bone remodeling applications. Yet, more research is required to elucidate the mechanisms responsible for such possible adaptations. They should allow to define optimal protocols and practical guidelines.

Therefore, the aim of this study was to investigate the effects of CH exposure on differentiation of human bone-marrow derived MSC into osteoblasts or adipocytes. In addition, we performed a pilot study, in elderly, on how CH exposure may affect bone mineral density (BMD) and fat mass.

Results showed that mineralization and lipid-droplets formation were inhibited in MSC cultures exposed to CH. That was positively correlated to increasing exposure time to hypoxia, during the differentiation into osteoblasts or adipocytes, respectively, in vitro. It has been reported that bone formation decreased in sustained hypoxia[33]. Our data indicate that this may be due, in part, to the effect on MSC differentiation. It has been observed that spontaneous calcification in MSC cultures from rats after 21 d is inhibited when cells are maintained at 2% O₂[34]. Similar results have also been obtained when primary rat osteoblastic cells were induced to differentiate into osteoblasts in hypoxia[35]. However, in vivo experiments have shown that exposure of rats to cyclic hypoxia for 5 h/d, 5 d/wk for 5 wk, increased bone mineral density with respect to rats maintained in normoxia[13].

Our data indicate that 4 h/d, 4 d/wk hypoxia inhibited osteoblast mineralization. This supports existing data on the negative effect of hypoxia on osteoblastogenesis. However, one- and two-h exposures did not significantly affect it. Therefore, the possible positive effect of cyclic hypoxia on bone mineral density, in vivo, may be due more to the inhibitory effect on osteoclastogenesis than to the activating effect on osteoblastogenesis, as previously suggested[13]. Thus, it has been described that osteoclastogenesis is induced under conditions of hypoxia[36]. Nevertheless, recent results have been reported, showing that constant conditions of hypoxia (1% O₂) inhibit osteoclastogenesis[37]. Such data controversy may be related to the fact that induction of HIF-1A in hypoxia may reduce osteoclast differentiation, yet enhancing bone resorption by mature osteoclasts[38], as well as different experimental conditions.

Analyses of gene expression showed that mRNA levels of two osteoblastogenesis-inducing transcription factors were not repressed in CH conditions. There are conflicting data in the literature on RUNX2 expression in hypoxia. Thus, while some authors describe an upregulation and osteogenic differentiation[39], others have shown that hypoxia inhibits osteoblastic differentiation, through RUNX2 downregulation; and that overexpression of HIF-1A in MSC also represses expression of the former[40-42]. Unlike these studies, there was cell culture reoxygenation after hypoxia in our experiments. That could prevent the repression of RUNX2, and consequently also of SP7. However, the expression of genes coding for extracellular matrix proteins (such as COL1A1 and IBSP), decreased with the hypoxia time in our conditions. Interestingly, it has been described that the expression of both COL1A1 and RUNX2 decreased under continuous hypoxia[34].

Therefore, our results suggest that the increase in the hypoxia period, during CH treatment, negatively affected the maturation of osteoblasts. Thus, MSC were maintained in a pre-osteoblast state, without the ability to mineralize. Indeed, it has been reported that pre-osteoblasts subjected to hypoxia did not mature, nor mineralize[43]. Thus, while pre-osteoblasts maintained at 2% O₂ for two weeks, and changed to 20% O₂ during the third week, do not mineralize, mineralization occurs after changing from 20% to 2% O₂[43]. It has also been observed that the induction of HIF-1A upregulates the metastasis-associated gene 1. In turn, this increased the proliferation of mouse osteoblast precursor cells (MC3T3), but inhibited their differentiation and mineralization[44]. In this sense, although the effect of hypoxia on osteoblast differentiation in vitro is not completely defined, it is generally accepted that hypoxia favors the maintenance of cells in an undifferentiated state[45,46].

The glucose metabolism of undifferentiated MSC is mainly based on glycolysis. However, oxidative phosphorylation is activated during mineralization, taking place when differentiating from pre-osteoblasts into osteoblasts[47]. The induction of HIF-1A favors glycolysis. Therefore, MSC induced into osteoblasts in hypoxia remained undifferentiated[47]. Thus, the 4 h of hypoxia in CH of our model seems enough to maintain cells in a pre-osteoblastic state. This effect may be mediated by the increase in the LRP5/DKK1 ratio, observed mainly at day 13 of osteogenic differentiation. Interestingly, LRP5 encodes a membrane co-receptor that interacts with the frizzled-4 receptor and is involved in the activation of Wnt/β-catenin pathway. On the other hand, DKK1 encodes an inhibitor of the pathway, which acts by binding and neutralizing LRP5 and LRP6 proteins. Therefore, the relationship between LRP5 and DKK1 partly modulates the activity of the Wnt/β-catenin pathway[48]. Activation of this pathway is necessary when precursor cells differentiate into pre-osteoblasts. However, the Wnt/β-catenin pathway must be inhibited for maturation of osteoblasts into osteocytes[49]. Also, it has been described that the expression of RUNX2 must decrease in the last phases of differentiation, for correct progression of osteoblastic differentiation[50].

Increasing the period of hypoxia in CH suppresses mineralization of MSC induced to differentiate into osteoblasts, in our experiments. That can be explained by the maintenance of the expression of RUNX2 and activation of the Wnt/β-catenin pathway, preventing progression from pre-osteoblasts to mature osteoblasts. It has been reported that induction of HIF-1A in MC3T3 cells decreases the activity of the Wnt signaling pathway. That happens through the upregulation of sclerostin (mainly expressed by osteocytes), which encodes an inhibitor of such pathway[51]. It has also been reported that hypoxia and the stabilization of HIF-1A downregulate the expression of sclerostin and DKK1, using a different model in vitro and in vivo. That increases the activity of the Wnt/β-catenin pathway[52,53]. The results in our CH model, although using a different exposure to hypoxia, are partially in agreement with the data of such latest studies.

Surprisingly, the osteocalcin-encoding gene increased its expression in cultures exposed to longer hypoxia, unlike COL1A1 and IBSP. Osteocalcin is the most abundant non-collagenous bone matrix protein. It modulates the osteogenic differentiation of MSC, playing an essential role in bone mineralization[54]. But it is also considered a multifunctional hormone produced by bone, modulating the regulation of the glucose metabolism. Thus, it has been described to enhance the glucose uptake in muscle, insulin production in pancreas, insulin sensitivity in liver, promotion of β-cell proliferation in pancreas and adipose tissue upregulation of adiponectin expression[55,56]. Recently, however, new findings have questioned such putative hormonal function of circulating osteocalcin, and thus its role on the energy metabolism, mainly in humans[57,58].

The higher expression of the osteocalcin gene in CH, with the period of hypoxia, cannot be related to an increase in mineralization in our model. The rationale is that a significant decrease of extracellular matrix mineralization was observed in these cultures. Therefore, a possible explanation is that the observed increase in expression is part of the metabolic response of osteoblasts to CH. Indeed, the glucose metabolism by osteoblasts is primarily glycolytic. That is less efficient than oxidative phosphorylation: 2 vs 38 ATP moles per metabolized glucose mole, respectively[59]. The decrease in O₂ levels in hypoxia favors glycolysis through HIF-1A, which is further stabilized by osteocalcin, leading to an increased rate of glucose uptake. In other words, osteocalcin favors the glycolytic pathway against the mitochondrial oxidative phosphorylation[60]. Therefore, the interaction between HIF-1A and osteocalcin could explain our results, although further research would be needed to obtain more conclusive results in this regard.

Additionally, CH with the longest period of hypoxia decreased the formation of lipid droplets, downregulating the expression of genes involved in lipid metabolism, in MSC induced to differentiate into adipocytes. This suggests that hypoxic treatments may keep cells in undifferentiated stages, as in MSC induced to differentiate into osteoblasts. Hypoxia induces genes involved in glucose metabolism in adipocytes, including the ones encoding transporters (like GLUT1, GLUT3 and GLUT5). Additionally, it represses the expression of genes related to adipogenesis, mitochondrial biogenesis (such as PPARG2 and PPARGC1A) and oxidative phosphorylation[61]. During adipogenesis, changes in gene expression allow to adapt the cell metabolism to oxidative phosphorylation, involving structural and functional changes in mitochondrial content[62-64]. This adaptation occurs within the first seven days of adipogenic differentiation of human adipose tissue-derived stromal cells[65].

Therefore, our results indicate that 4 h of hypoxia during CH treatment produces changes in gene expression and metabolism, in MSC induced to differentiate into adipocytes. That includes downregulating the PPARGC1A gene. The rationale is that the glycolytic metabolism predominates during the hypoxia period. This inhibits the maturation of adipocytes, because it negatively interferes with the change towards an oxidative metabolism, based on greater mitochondrial activity, as well as an important role of oxidative phosphorylation. Our results agree with those observed in murine C3H/10T1/2 MSC. Thus, such cells lose capacity to differentiate into adipocytes after being pre-conditioned for 24 h with 0.1 mmol/L CoCl₂, mimicking low-oxygen conditions[66].

Interestingly, CH increased the expression of the DKK1 gene and decreased the LRP6/DKK1 ratio, proportionally to the time of hypoxia, in MSC induced to differentiate into adipocytes. That should inhibit the Wnt/β-Catenin pathway, consequently enhancing adipogenesis[67]. However, contrary to expectations, the cultures with a higher induction of DKK1 were those with lower adipogenesis. Interestingly, DKK1 is mainly expressed in the early stages of adipogenesis, decreasing during the final maturation of adipocytes[68,69]. It has been recently described that mature adipocytes express various components of the Wnt/β-Catenin pathway. Additionally, they intervene in the regulation of lipogenesis, as well as the expansion of white adipose tissue[70,71]. Therefore, a possible explanation for the results is that the continuous upregulation of DKK1, with the greatest periods of hypoxia, prevents the correct maturation of adipocytes. In this regard, other authors have shown that MSC exposed to hypoxia increased the expression of DKK1[72], and that such gene modulates signaling pathways, promoting cell proliferation and migration[73]. Therefore, it would be interesting to further explore how its overexpression may affect both preadipocytes and mature adipocytes.

CH also differently affected the expression of VEGFA in osteoblastic differentiation, in relation to the adipocytic one. While its expression increased in the former, it decreased in the latter, mainly with longer hypoxia times. Generally, hypoxia stimulates the expression of VEGF, through HIF-1A, working as an angiogenic factor in response to low O₂ levels. Additionally, RUNX2 and HIF-1A interact to induce VEGF expression in osteoblasts[74]. That is in agreement with our results. On the other hand, exposure to hypoxia should induce VEGF in adipocytes[61]. Yet, treatment of cyclic hypoxia repressed it. One possible explanation is that the increase of DKK1 expression negatively affect the Wnt/β-Catenin pathway. As this pathway may directly (or by interacting with HIF-1A) induce VEGF transcription, its inhibition may affect VEGF expression[75,76].

One of the most interesting effects of CH on both osteoblastogenesis and adipogenesis was the increase of the OPG/RANKL ratio. Such genes encode factors that are part of the regulation of bone remodeling. The RANKL protein binds to its RANK receptor in cells of the osteoclastic lineage, activating osteoclastogenesis and bone resorption. On the other hand, OPG is a decoy receptor of RANKL, preventing its binding to RANK, and therefore osteoclastic activation[7]. Interestingly, HIF-1A can induce OPG expression[12]. Thus, overexpression of HIF-1A in osteoblasts increased bone mass in mice, through a decrease in the resorptive activity of the osteoclasts. That is a consequence of the upregulation of OPG[77]. Our data showed that CH in osteoblasts did not induce the expression of OPG but increased the OPG/RANKL ratio. It is described that the expression of OPG increase with the maturation of osteoblasts but it is repressed in pre-osteoblasts that constitutively express RUNX2, which prevents their correct maturation[78,79]. Therefore, our data suggest that MSC induced to differentiate into osteoblasts remain in a more undifferentiated state in CH, with hypoxia times of 4 h. This may negatively affect the expression of OPG. However, our data did not necessarily mean that CH activated osteoclastogenesis. The rationale is that RANKL expression also decreased, as described by some authors for longer or sustained periods of hypoxia[80,81].

On the other hand, OPG and RANKL expressions increased and decreased, respectively, in MSC induced to differentiate into adipocytes, with CH, mainly with longer periods of hypoxia. The expression of such genes in adipose tissue has been described[82-84]. In addition, RANKL expression by bone marrow adipocytes is involved in the regulation of bone remodeling. Thus, bone resorption increased with bone-marrow adipocytes that secrete RANKL, in type I diabetic mice[85]. Ovariectomy also increased the expression of RANKL in bone-marrow adipocytes. This may contribute to the loss of bone mass[86].

The differentiation of MSC from bone marrow into adipocytes increased with age. As a consequence, the formation of new osteoblasts decreased, which favors the loss of bone mass and the development of osteoporosis[4]. In this context, it has been described that pre-adipocytes express RANKL in the adipose tissue of bone marrow. That directly contributes to the osteoclastic activity and bone resorption, which is associated to aging[84,87]. Our results showed that CH exposure can reverse the induction of RANKL expression, in the early stages of adipogenesis, while inducing OPG expression, in MSC induced to differentiate into adipocytes. This suggests that exposure to CH may raise the OPG/RANKL ratio in bone marrow, partly preventing age-related bone mass loss. This is particularly relevant in older people, who have higher bone marrow adiposity, and therefore higher RANKL secretion by adipocytes.

The aim of this pilot study in humans was to investigate the effects of CH, during 18 wk, on BMD and fat mass of elderly adults. Cyclic exposure to moderate hypoxia was sufficient to increase whole body BMD. Yet, it did not alter femur proximal BMD, nor fat mass percentage. Results of this investigation showed that the fat mass percentage in healthy elderly adults did not decrease, using eighteen 16-min sessions of CH exposure (16.1% O₂). Yet, previous studies have showed decreases in body fat, after hypoxia exposure in different conditions. This might be due to the profile of body composition, varying with the hypoxia duration or severity[88]. Natural high altitude or artificial hypoxic conditions enhanced the oxygen transporting ability of muscles. That improved fat oxidation, reducing body fat[88,89].

Some studies have shown that moderate hypoxia could involve higher fat mass reduction[90]. However, such effects could be due to differences between the studied populations[91] or experimental design. It has been reported that obese subjects are three times more susceptible to the hypoxia effects that people with normal BMI[92]. Furthermore, inter-individual variability and confounding variables of fit or normal-weight participants may bias results. Among them are exercise or fitness, which might have interfered with hypoxic exposure results[89]. Therefore, a moderate CH exposure of 18 wk could not be sufficient to provoke changes in fat mass percentages, in subjects with normal BMI. Longer lasting or more severe passive hypoxia (until 12% O₂) could be necessary to improve the body composition, such as our in vitro results suggest on the effect of CH on MSC differentiating into adipocytes. Total BMD significantly increased compared to the control group, whose levels decreased after exposure to CH (twice a week, 16 min per session, for 18 wk), simulating 2500 m above sea level. Although the results indicated that it might cause an individualized response after 18 wk of treatment, such exposure was not sufficient to improve the proximal femoral BMD.

Few studies have investigated the in vivo effect of CH on BMD. Different findings have been reported, after applying passive CH protocols in rats. For instance, healthy rats were exposed to CH during 2 wk (4 h per day, simulating 3000 to 5000 m above sea level), maintaining BMD throughout the body[80]. In another study, increases in dorsal spine BMD were observed after 5 wk of exposure to CH (5 d per week, 5 h per day) at 4500 meters above sea level[13]. In this way, the number of episodes of hypoxia, severity and total duration of exposure can lead to different physiological responses. Therefore, a certain number of short episodes with a modest level of hypoxia (9%-16% of O₂) could generate benefits, if repeatedly administered for weeks[93].

Exposure to CH may have an impact on bone health because it affects the osteogenic and adipogenic differentiation of MSC. Thus, CH tends to inhibit differentiation to both cell types, increasing the OPG/RANKL expression ratio, although requires longer hypoxic periods. Therefore, our results suggest that CH could be applied to improve bone health, decreasing resorption mainly in the elderly, as has been suggested by other authors[13,33]. The in vitro results obtained may partly explain those obtained in our pilot study, conducted on elderly people. Our in vivo results indicate that exposure to CH (16.1% O₂ for 18 wk) is sufficient to significantly increase total BMD in older adults, but not in specific regions, such as the proximal femoral. This passive protocol is also not enough to decrease the percentage of fat mass. In summary, CH exposure is a potential tool for the prevention and treatment of age-related dysfunctions once an optimal protocol is defined. Indeed, the available evidence is promising, albeit limited. Therefore, larger randomized controlled trials are required to ascertain the significance of current results.