Osteoporos Int (2008) 19:595–606 DOI 10.1007/s00198-007-0492-z
REVIEW
Relationships between fat and bone I. R. Reid
Received: 14 March 2007 / Accepted: 20 August 2007 / Published online: 27 October 2007 # International Osteoporosis Foundation and National Osteoporosis Foundation 2007
Abstract Body weight impacts both bone turnover and bone density, making it, therefore, an important risk factor for vertebral and hip fractures and ranking it alongside age in importance. The effect of body weight is probably contributed to by both fat mass and lean mass, though in postmenopausal women, fat mass has been more consistently demonstrated to be important. A number of mechanisms for the fat-bone relationship exist and include the effect of soft tissue mass on skeletal loading, the association of fat mass with the secretion of bone active hormones from the pancreatic beta cell (including insulin, amylin, and preptin), and the secretion of bone active hormones (e.g., estrogens and leptin) from the adipocyte. These factors alone probably do not fully explain the observed clinical associations, and study of the actions on bone of novel hormones related to nutrition is an important area of further research. An understanding of this aspect of bone biology may open the way for new treatments of osteoporosis. More immediately, the role of weight maintenance in the prevention of osteoporosis is an important public health message that needs to be more widely appreciated. Keywords Fracture . Insulin . Leptin . Obesity . Osteoporosis
Introduction The principal function of the skeleton is to provide a rigid framework to support, protect, and facilitate the function of the soft tissues. The ribs, pelvis and skull provide protection I. R. Reid (*) Faculty of Medical and Health Sciences, University of Auckland, Private Bag 92019, Auckland, New Zealand e-mail:
[email protected]
for their contents, the ribs are also important for breathing, and the long bones are essential for locomotion. Therefore, it is plausible, from an evolutionary perspective, that the strength of the skeletal framework would be closely related to the mass of soft tissue that it is required to support and facilitate the functions of. If all individuals had the same size skeleton whatever their body weight, some would have bones that were inadequate for the task at hand, and others would be at a disadvantage through having a skeleton that was significantly heavier than it needed to be. With respect to muscle and bone, these relationships have been regarded as self-evident for many years and have been confirmed in recent decades through direct measurement of muscle and bone mass [1]. However, it should be noted that simple correlations of lean mass and bone mineral content overestimate the relationship between bone mineral density (BMD) and muscle mass [2–4], since bone mineral content and muscle mass are markedly dependent on height, whereas bone density is not. In simple terms, taller individuals have longer legs and require a larger muscle mass to cover them. Only much more recently has the possibility of a relationship between bone and fat been acknowledged. Again, such a relationship is plausible since variations in fat mass between individuals can potentially double the load that the skeleton is required to bear, and unless there were some regulation of bone density in response to fat mass then it would be expected that fractures would be commonplace in obesity. In fact, the opposite is the case, adiposity being an important protective factor for most osteoporotic fractures (Fig. 1) [5–7]. Thus, the fat-bone connection is important clinically since thinness is a potentially preventable risk factor for fracture. The strength of the fat-bone relationship, however, indicates that powerful mechanisms for the regulation of bone mass are involved. Therefore, this is a critical area of our under-
596
standing of the regulation of bone mass and, consequently, a potentially important pharmaceutical target for the development of osteoporosis therapies.
Clinical studies of fat mass and bone We first explored these relationships in a cohort of healthy postmenopausal women and were intrigued to find that areal BMD was comparably related to weight, body mass index and fat mass (≅ 0.5), and less closely related to lean mass (r ≅ 0.2) [8]. This effect tended to be more marked in the total body scans (which include non-weight bearing bone), in comparison with the weight-bearing regions of the lumbar spine and proximal femur, indicating that it was not accounted for by skeletal load. In this cohort, when fat and lean masses were entered as independent variables into a multiple regression analysis, fat mass was found to be independently related to BMD at all sites. To determine whether the same relationships obtained for volumetric BMD, further studies were done using antero-posterior and lateral DXA scans of the lumbar spine, and produced similar findings [3]. We subsequently showed that baseline fat mass and changes in fat mass were predictive of the change in BMD over both two years [9] and 10 years [10]. Many other groups have confirmed this cluster of findings [2, 11, 12], and body weight appears to predict BMD better than does BMI [13]. The precise relationships found are dependent on gender (the fat-bone relationship is weaker in men) [4], menopausal status (larger fat effect postmenopause) and exercise status (the relationship is stronger in sedentary populations [14]). Work by others has extended the modalities of bone measurement used to include metacarpal cortical thickness [15], quantitative CT scanning of the radius [16] and tibia [17], and ultrasound [18, 19]. The confirmation of the relationship between soft tissue mass and bone mass in the metacarpals and in the radius, emphasizes that it is not simply a function of the skeleton responding to load bearing. The dynamic nature of the relationship between bone and soft tissue is emphasized by a study showing an effect of weight change on metacarpal area over a period of 21 years, indicating that this is not an artifactual effect of weight on bone density measurement [20]. Some studies appear to contradict this body of data by showing an inverse relationship between fat mass and bone density [21]. Such findings are usually accounted for by the inappropriate treatment of highly collinear variables (fat mass and weight) as independent variables. Fewer studies have been carried out in children, but their findings are generally similar. Goulding has shown that areal BMD and bone mineral apparent density of the spine are both increased in obese girls in comparison with agematched controls. Boys showed similar trends which were
Osteoporos Int (2008) 19:595–606
not significant, possibly because of a trend for obese boys to be less sexually mature than boys of normal weight [22]. However, total body bone mineral content was higher in boys with higher fat mass [23]. A further cross-sectional study in several thousand 10-year-olds found that there was a strong positive relationship between total body fat mass and both total body bone mass and area in boys and girls, even after adjustment for height and/or lean mass [24]. This indicates a positive effect of adiposity on linear skeletal growth, as well as on its density. There was a similar positive association between baseline total body fat mass and the increase in bone mass and area over the following 2 years in boys and pre-pubertal girls, but not in those girls who had entered puberty, presumably because of the potent effects of sex hormones on skeletal growth. Recently, Janicka et al. have reported a similar study of normal postpubertal adolescents [25]. They found positive correlations between a variety of bone measures (made using DXA or computed tomography) and fat mass and, to a greater extent, lean mass. However, when they carried out multiple regression analysis, including fat, lean and skeletal size, the fat effect ceased to be significant. It is not clear why skeletal size was included as a co-variable in these analyses. Since adiposity may stimulate skeletal growth (e.g., via leptin production), this may be the mechanism by which fat mass increases bone mass, so adjusting it out of the equation is not necessarily appropriate.
Fat mass and fractures Many authors have shown a relationship between body weight and fracture risk, and these data have recently been brought together in a meta-analysis (Fig. 1) [5]. The metaanalyzed data clearly indicate that high body mass index is protective against total fractures, osteoporotic fractures and hip fractures and is seen equally in men and women. Most of the effect for non-hip fractures is probably mediated by the effects of weight on BMD, but at the hip there is a component which is BMD-independent. A similar effect of BMI on vertebral fractures was found in a pan-European study in which the prevalence of vertebral deformity in a given country was inversely related to the mean BMI in the study population in that country (r=−0.66) [26]. The Study of Osteoporotic Fractures found that body weight in the lowest quartile doubled hip fracture risk, and that lean and fat masses (assessed by bioelectrical impedance) contributed equally to this effect [27]. However, when fat and lean masses are measured with DXA, fat mass alone was found to be a significant risk factor for hip fractures in French women [6] and in Chinese men [7]. Forearm fractures are not so markedly weight-dependent [28], possibly because greater weight increases the skeletal load following a fall.
Osteoporos Int (2008) 19:595–606
597
Fig. 1 Meta-analysis of the effect of body mass index on fracture risk, with and without adjustment for bone mineral density. From de Laet et al. [5], used with permission
In children, high weight actually appears to be a risk factor for fracture [29], possibly suggesting that soft tissue acquisition is running ahead of the appropriate skeletal response during growth.
Mechanisms It would be expected that if fat mass were to impact bone mass, then it would do so by modulating activity of bone cells. There is evidence for this in the inverse relationship between BMI and osteoclast activity in normal postmenopausal women [30] and the increase in bone resorption which occurs following weight loss [31]. The latter study emphasizes the dynamic nature of the fat-bone connection and suggests that much shorter-term studies looking at feeding effects on bone metabolism might also be relevant to understanding this area. Hormones responsive to feeding Glucose ingestion results in an increase in calcitonin secretion [32] and a decrease in parathyroid hormone [33, 34], both changes apparently being mediated by hyperinsulinemia. This results in decreased bone turnover, though Clowes et al. failed to find an acute effect of hyperinsulinemia itself on bone turnover [34]. Amylin is co-secreted with insulin, and potently inhibits bone resorption [35], so might contribute to the effect of glucose. There is also evidence of reduced bone resorption following feeding of fat or protein in humans [36– 38]. This could also be mediated by changes in parathyroid hormone, amylin and calcitonin, but there are other potential players since these nutrients stimulate secretion of the incretin hormones, glucagon-like peptide-1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP), which act to enhance postprandial insulin secretion. GLP-1 and GIP do not acutely influence bone resorption [37], but GIP stimulates osteoblast proliferation, can attenuate post-
ovarectomy bone loss [39, 40], and the GIP-receptor knockout mouse shows decreased bone size, mass, and formation rate [41]. Parenteral administration of the related peptide, GLP-2, produces a dose-dependent reduction in serum C-terminal telopeptide of type I collagen [37], and GLP-2 administration over a five-week period decreased bone resorption and increased bone density in a small uncontrolled study [42]. Consistent with these findings, deletion of GLP-2 receptor leads to marked skeletal deficits in growing mice [43]. Ihle has studied the impact on bone of short-term dietary energy restriction in more detail, finding that markers of bone formation are suppressed by even modest restrictions in dietary energy, whereas more marked energy deficits also cause increased bone resorption [44]. They noted that the suppression of osteoblast markers is associated with suppression of insulin, IGF1 and thyroid hormones, and that the increase in bone resorption mirrors a decline in estrogen, raising the possibility that these hormonal changes might be causal. IGF1 is an important regulator of bone growth. Its secretion is reduced in states of poor nutrition, possibly because of resistance to the actions of growth hormone. Long-term randomized, controlled trials of milk [45] or protein supplements [46] have demonstrated increases in both circulating IGF1 and BMD. These findings point to a close relationship between bone metabolism and energy availability and could certainly contribute to the relationship between fat and bone. Ghrelin is a further candidate hormone that might mediate feeding effects on bone. It is synthesized in the stomach and released in response to fasting, acting as an appetite stimulant. Its receptor is expressed in osteoblastic cells, and ghrelin stimulates osteoblast proliferation and differentiation [47–50], as well as osteoclastogenesis and the bone-resorbing activity of mature osteoclasts (Cornish et al, manuscript in preparation). The latter data suggest that ghrelin could contribute to the increased bone resorption that accompanies fasting. However, its anabolic effects appear to predominate in vivo,
598
since it increases BMD when administered to rats [47]. Clinical studies are contradictory, however, overnight ghrelin levels being closely related to BMD in adolescent women [51], but no consistent relationship being seen between fasting levels and BMD in older men and women [52]. Beta cell hormones At the time that the connection between fat mass and bone density was originally made, the understanding of the endocrine changes in obesity was relatively unsophisticated. However, it was well established that hyperinsulinemia was a hallmark of this condition. Insulin is a potential regulator of bone growth, since osteoblasts have insulin receptors [53] as well as IGF1 receptors, which can also mediate the effects of insulin. In vitro, insulin directly stimulates osteoblast proliferation [54] and when administered locally over the calvariae of adult male mice, it produces two- to three-fold increases in histomorphometric indices of bone formation [55]. In clinical studies, circulating insulin levels, both fasting and following a glucose load, are related to bone density. We have shown this in normal postmenopausal women and found the effect to be partly independent from that of adiposity [56]. Similar effects have been demonstrated in both men and women in the Rotterdam study, where again they were partly independent of the effects of BMI [57]. The San Antonio Heart Study has produced comparable findings in women [58]. Abrahamsen [59] studied these relationships in men, and found that insulin sensitivity (measured from an intravenous glucose tolerance test) was inversely related to bone density independently of weight and fat mass. In addition, they found that the dependence of bone density on fat mass was lost when insulin sensitivity was entered into the multiple regression analysis, suggesting that this relationship was mediated through insulin sensitivity. Recently, the Tromso study has found that non-vertebral fracture risk progressively declines with increasing insulin resistance (inferred from the severity of the metabolic syndrome), being reduced by 50% in those with greatest insulin resistance [60]. The terms ‘insulin resistance/sensitivity’ refer to the hypoglycemic effects of insulin. The other actions of the hormone, including those on bone, appear to be intact. Thus, hyperinsulinemic patients develop a cluster of abnormalities, including androgen and estrogen overproduction in the ovary, and reduced production of sex hormone-binding globulin, in the liver. These phenomena result in increased free concentrations of sex hormones which would be expected to reduce osteoclast activity and probably have positive effects on osteoblasts. As a result, high bone density is a very consistent finding across a wide range of hyperinsulinemic states, including obesity, poly-
Osteoporos Int (2008) 19:595–606
cystic ovary syndrome [61] and congenital generalized lipodystrophy [62]. The latter is particularly significant because it represents a dissociation of fat mass and insulin levels. These findings are consistent with the burgeoning literature relating to bone density and fractures in diabetes. In general, densities tend to be lower than normal in insulinopenic diabetes, and supranormal in those with diabetes associated with insulin resistance [63, 64]. As a result, fracture risk is elevated in type 1 diabetes [65]. The direct effects of insulin on bone are likely to be reinforced by those of two other hormones which are cosecreted with insulin. Amylin, a 37-amino acid peptide that belongs to the calcitonin family and has evolutionary links with insulin, directly stimulates osteoblast proliferation [66] and has a calcitonin-like effect on the osteoclast [35]. Its systemic administration substantially increases bone volume in mice [67] and rats [68, 69], and the amylin knockout mouse has reduced bone density [70]. Recently, preptin has been identified as a further product of the beta cell. It corresponds to Asp69 - Leu102 of pro-insulin-like growth factor2 and is anabolic to the osteoblast in vitro. Its local administration in adult mice increases bone formation and bone mass [71]. Thus, the beta pancreatic cell exerts a cluster of anabolic effects on bone, all of which will be accentuated in individuals with high fat mass. These are summarized in Fig. 2. One way to assess the effect of these hormones on bone is to reduce their levels through the use of an insulin sensitizing agent. We have recently done this using rosiglitazone, in a 3month randomized controlled trial in normal postmenopausal women [72]. This study demonstrated reductions in bone formation markers within weeks of initiating treatment, and reductions in hip BMD of 2% at 3 months, suggesting that increased insulin sensitivity in humans has profound effects on bone within a short period. Glitazones are also likely to influence bone via their stimulation of peroxisome-proliferator-activated receptor-gamma, which directs precursor cells
Fig. 2 Summary of the principal mechanisms by which the hyperinsulinemia associated with obesity leads to increased bone mass. © IR Reid, used with permission
Osteoporos Int (2008) 19:595–606
599
into the adipocyte lineage in preference to becoming osteoblasts [73]. Some fatty acids are converted into agonists for this receptor, indicating another potential route for adipose tissue to impact bone [74]. Adipocyte hormones The first adipocyte-derived hormone implicated in the fatbone relationship was estrogen, which is produced in adipocytes from circulating androgens. Over the last decade, however, leptin has come to be regarded as the classic adipocyte hormone, and it is now clear that it can directly modulate the activity of bone cells, as well as indirectly influencing skeletal metabolism through its actions on the central nervous system. Leptin Direct skeletal effects The signaling form of the leptin receptor has now been shown by many groups to be expressed in both osteoblasts and chondrocytes [75, 76]. It increases proliferation and differentiation in osteoblasts in vitro [76–78] and promotes mineralized nodule formation [79, 80]. It has similar stimulatory effects on chondrocytes, both in vitro and in vivo [76, 81], so may increase linear growth. Leptin also directly regulates osteoclast development, reducing production of RANK and RANK-ligand, and increasing osteoprotegerin in cultures of human bone marrow stromal cells or peripheral blood mononuclear cells [82, 83], with a resultant inhibition of osteoclastogenesis [76, 82]. Thus, its local effects are all directed towards skeletal preservation and would be consistent with a need for individuals with high fat mass to have stronger skeletons to support the greater soft tissue mass. Centrally mediated effects Outside of the bone field, the hypothalamus is regarded as the principal target of leptin. The arcuate nucleus (in the hypothalamus) is at the center of a system which balances food intake against energy expenditure, thus maintaining energy homeostasis (Fig. 3). In simple terms, this nucleus contains anabolic neurons which express both neuropeptide Y and agouti-related protein, the activity of which is inhibited by leptin, and neurons expressing pro-opiomelanocortin (POMC) which are activated by leptin. Interestingly, insulin acts on both types of neuron in the same way as leptin, suggesting that these hormones reinforce each others’ actions centrally, as well as peripherally [84]. In a series of elegant studies, the Karsenty group have demonstrated that intracerebroventricular infusion of leptin causes bone loss in leptindeficient and wild-type mice, through inhibition of bone formation [85] and stimulation of bone resorption [86]. Blockade of the sympathetic nervous system abrogates
Fig. 3 Control of energy homeostasis by arcuate nucleus neurons. There are two sets of neurons in the arcuate nucleus (Agrp/Npy and Pomc/Cart neurons) that are regulated by circulating hormones. Agrp (agouti-related protein) and Npy (neuropeptide Y) are neuropeptides that stimulate food intake and decrease energy expenditure, whereas α-melanocyte stimulating hormone (a post-translational derivative of proopiomelanocortin, Pomc) and Cart (cocaine- and amphetamineregulated transcript) are neuropeptides that inhibit food intake and increase energy expenditure. Insulin and leptin are hormones that circulate in proportion to body adipose stores; they inhibit Agrp/Npy neurons and stimulate adjacent Pomc/Cart neurons. Lower insulin and leptin levels are, therefore, predicted to activate Agrp/Npy neurons, while inhibiting Pomc/Cart neurons. Ghrelin is a circulating peptide secreted from the stomach that can activate Agrp/Npy neurons, thereby stimulating food intake; this provides a potential molecular mechanism for integrating long-term energy balance signals with short-term meal pattern signals. Ghsr, growth hormone secretagogue receptor; Lepr, leptin receptor; Mc3r/Mc4r, melanocortin 3/4 receptor; Y1r, neuropeptide Y1 receptor. From Barsh and Schwartz [84], reprinted with permission from Macmillan Publishers Ltd: Nature Reviews Genetics 3:589–600, copyright (2002)
these effects, which appear to be mediated by the βadrenoreceptor on the osteoblast [87]. This observation led to speculation that beta adrenergic blockade might be important clinically in assessing fracture risk and as a therapeutic intervention, but this has not been borne out in clinical studies which do not show consistent relationships between the use of beta blockers on the one hand, and bone density, rates of bone loss or fracture on the other [88]. Indeed, in a prospective randomized controlled trial, we found that propranolol reduced markers of bone formation, implying that the sympathetic nervous system impacts bone via other pathways also [89].
600
Osteoporos Int (2008) 19:595–606
Fig. 4 Integrated levels of circulating hormones in rats in which an adeno-associated virus vector-encoding leptin was introduced into the third ventricle (LEP). Comparison is made with untreated (UT) animals, those with a vector encoding green fluorescent protein (GFP), and rats pair-fed with the leptin group. Body weights from vector placement to 42 days are shown in the right-hand panel. Bars with different superscripts are significantly different from each other (p
