Switching fat from the periphery to bone marrow: why in Type I diabetes?

Type I (T1) diabetes is characterized by hypoinsulinemia and hyperglycemia as a consequence of decreased glucose uptake by insulin-sensitive cells. When T1 diabetes is poorly controlled, weight loss occurs and is manifested by a decrease in both muscle and peripheral fat mass. This suggests that anabolic, tissue-building processes are suppressed. Correspondingly, bone formation is decreased and can lead to bone loss in T1 diabetic humans and rodent models [1–13]. The decrease in bone anabolic activity can prevent attainment of maximum bone density and increase the risk of fractures, which are often difficult to heal [2]. An unexpected characteristic of T1 diabetic mice is an increased number of visible lipid-loaded adipocytes in bone, which contrasts with their depleted peripheral fat pad stores [14,15]. This is evident in both chemically induced and spontaneous mouse models [15]. The role of bone marrow fat as an adipose storage tissue, mechanisms accounting for the T1 diabetic bone adiposity and, ultimately, the impact of this knowledge on treatments for diabetic bone loss will be the focus of this article.

Fat deposition in bone marrow is well recognized and readily visible inside steak bones at the grocery store. Exactly how or why marrow fat is present to a greater or lesser extent is not known. What is known is that an inverse relationship often (but not always [16–18]) exists between bone marrow adiposity and bone density [19–24]. Selection of adipogenesis at the cost of osteoblastogenesis is a well-described theme in bone, because adipocytes and osteoblasts (bone-forming cells) are derived from a common mesenchymal precursor. Chronic deficiency of factors required for adipocyte lineage selection and early maturation (e.g., PPARγ) results in enhanced bone mineral density [25,26], suggesting that modulation of the adipose/bone reciprocal relationship could be used as a potential therapeutic treatment for some conditions of bone loss. In T1 diabetes, the increase in marrow adiposity (and PPARγ2 expression) is consistent with a decrease in bone density and, potentially, altered mesenchymal cell lineage selection. However, the loss of peripheral fat stores suggests that the deposition of fat into the marrow must be a unique process compared with other adipose sites.

While much focus is placed on bone density/marrow adiposity correlations (and rightly so), less attention has been paid to the role of marrow fat as an adipose store and the relationship between marrow and peripheral adipose stores. At birth, our bones are devoid of fat but, with developmental growth, bone marrow adiposity increases to over 50% by the age of 30 years [27,28]. This increase occurs during the attainment of maximal bone density, indicating that marrow adiposity is probably linked to more than the loss of bone volume. At first glance, increased bone marrow adiposity appears to parallel levels of adiposity in peripheral fat pads and is inversely correlated to bone density. This is supported by studies demonstrating increased marrow adiposity in models of bone loss, including aging, obesity and skeletal unloading [29–31], that are typically associated with increased body fat or decreased lean mass. Likewise, treatment with a PPARγ agonist (rosiglitazone) increases subcutaneous fat pad mass (a sign of drug effectiveness [32]), bone marrow adiposity and bone loss [33]. Similarly, some treatments that decrease peripheral fat pad mass can decrease bone marrow fat [34]. However, there are also studies indicating that bone marrow adiposity does not correlate with visceral or subcutaneous adiposity in humans or mouse models [35,36]. This is consistent with what is seen in the T1 diabetic mouse model and the idea that fat depots throughout the body can function independently. Furthermore, adipose tissue gene expression profiles are depot/location specific (i.e., visceral, peripheral or deep subcutaneous), are inherent to the isolated adipose cells and not lost with long-term culturing, and can be maintained despite transplanting to a new location (at least for subcutaneous stores) [37–40]. Bone adipose tissue, therefore, could represent a fat depot with similar, but also independent, regulation to other fat stores.

Analysis of bone marrow adiposity indicates that skeletal adiposity is location and gender dependent, similar to general body fat depots. Even within a single bone there are variations in adiposity in healthy mice. For example, the distal end of the adult mouse tibia contains nearly 100% adipocytes (very large in size), while the proximal end contains a small percentage of adipocytes (often smaller in size) and a greater percentage of red (hematopoetic) marrow. Gender itself influences, in a location-dependent manner, fat pad adiposity and responses [41]. Males tend to carry more visceral fat, often visible in the abdominal region, while females tend to carry more subcutaneous fat, often visible in the hips and thighs. Similarly, bone marrow adiposity levels and location also exhibit gender specificity. For example, female mice have greater femur adiposity compared with age-matched males, while male mice have greater fat deposition in the skull marrow region compared with females [42]. The significance of these gender-specific depot differences is not clear, since there is no relationship with bone density. Again, this suggests an additional unappreciated function for these adipose stores. When examined in T1 diabetes, fat depots are affected without gender preference. In both males and females peripheral body fat pad mass is decreased markedly and marrow adiposity is increased in long bone and skull marrow [42].

As noted above, the location of fat depots in the body determines their response (i.e., lipolysis or lipogenesis) to different stimuli [41]. Bone marrow adipose stores also exhibit variation in responses, and this is seen in T1 diabetes. Specifically, while T1 diabetes is associated with increased marrow adiposity in the long bones and skull, vertebrae do not show an increase in marrow adiposity despite a significant decrease in bone volume fraction. This indicates a potential inherent location difference in bone marrow adipocyte (or stem cell) responses to metabolic alterations and/or differences in local environmental factors (e.g., more nerves in the vertebrae). Interestingly, leptin deficiency causes a somewhat similar response, marked by increased long bone marrow adiposity (and decreased bone density) but no change in vertebral adiposity (although density is increased) [43]. This is relevant to the T1 diabetes model, because adipocytes secrete leptin, peripheral adiposity is reduced in T1 diabetes and, hence, leptin levels are reduced in T1 diabetic mice [42]. Preventing the reduction in serum leptin levels in T1 diabetic mice by chronic leptin treatment inhibits the increase in long bone marrow adiposity but not bone loss [44]. This suggests a role for leptin in the regulation of bone marrow adiposity, but not bone loss, in T1 diabetes. Leptin treatment also caused a reduction in bodyweight and peripheral adiposity (probably as a result of decreased food intake and possibly increased metabolism) [44]. Taken together, the data indicate that larger peripheral fat depots should cause an elevation in serum leptin levels (secreted by adipocytes), which would decrease bone marrow adiposity (and in some cases perhaps increase bone density). Certainly there are reports indicating a link between increased adiposity and increased bone density, but there are a similar number suggesting that increased adiposity is associated with decreased bone mass and increased marrow adiposity. Thus, relationships between adipose tissue, adipokine secretion (and brain vs systemic effects), bone density and bone marrow adiposity are complex.

This brings up the question “why does long bone marrow adiposity increase with T1 diabetes while other fat stores are depleted?”. T1 diabetes is marked by several key physiologic changes, any of which could be involved in stimulating marrow adiposity. If the loss of peripheral body fat somehow contributes to the hording of fat in the marrow under metabolic stress then other conditions of peripheral fat loss could cause a similar bone phenotype. However, conditions that cause significant weight and fat loss, such as anorexia nervosa, cancer and HIV infection, lead to a decrease in marrow adiposity [45–48]. When T1 diabetes onset is rapid and severe, marked by high blood glucose levels and weight loss, we observe a greater increase in marrow adiposity compared with mice with a less-severe onset [49]. One difference between these conditions is the more dramatic loss of insulin-producing β-cells with severe diabetes onset. This implicates decreased insulin signaling in the mediation of T1 diabetic bone loss and marrow adiposity. However, euglycemic insulin receptor knockout/knock-in mice that are deficient in insulin receptor expression in bone do not lose bone and actually have fewer marrow adipocytes compared with wild-type mice [50]. This suggests that bone can develop normally despite the loss of insulin receptor signaling, and that insulin receptor signaling is involved in maintaining basal levels of marrow adiposity. This does not fit the T1 diabetes model, where low insulin is associated with increased marrow adiposity, but it should be noted that the insulin receptor-deficient bone may have a higher level of IGF-1 receptor signaling (due to hyperinsulinemia and elevated receptor expression) that may enable the bone to overcome any effects of reduced insulin signaling. We have not been able to observe a similar adaptive change in IGF-1 receptor expression in diabetic mouse bone; however, the diabetic mice have only switched to insulin-deficient conditions for a short period (weeks) compared with the insulin receptor-knockout mice, which developed and matured without insulin receptors in bone.

Another metabolic abnormality that T1 diabetic mice and patients can also experience is hyperlipidemia, which could activate PPARγ and promote increased marrow adiposity and bone loss. However, treatment of T1 diabetic mice with an antagonist to PPARγ is capable of preventing T1 diabetic marrow adiposity (as expected) but not bone loss [51], suggesting that the increase in adiposity alone does not account for the decrease in bone density (although osteoblast differentiation may also be affected by PPARγ inhibition). This treatment also prevented hyperlipidemia from occurring in T1 diabetic mice, indicating that hyperlipidemia is not a major contributor to diabetic bone loss.

In vitro studies indicate the possibility of hyperglycemia causing an increase in marrow adiposity. Chronic treatment of osteoblasts or mesenchymal precursor cells with elevated glucose levels leads to decreased expression of osteoblast markers, reduced osteoblast responsiveness and increased expression of adipocyte markers [52–56]. Similarly, hypoxic stress of osteoblast cultures can suppress osteoblast markers and enhance adipocyte markers. Taken together, these studies suggest that metabolic stress is capable of enhancing adiposity but, at the same time, suppressing osteoblast maturation. Something similar may be occurring in vivo, such that suppression of adipogenesis alone cannot promote osteogenesis since osteogenesis is simultaneously and independently being inhibited. This would mean that the increase in marrow adiposity in T1 diabetes could be a marker of potential bone loss but would not be a good therapeutic target.

Finally, other events may be occurring within bone marrow that could influence adiposity and osteogenesis [23]. For example, changes in marrow cell composition and inflammation could result in local cues that lead to the altered bone phenotype. T1 diabetes is known to be associated with increased systemic inflammation, especially during disease onset when the expression of adipose genes increase and osteoblast genes decrease. More recently, increased inflammation within T1 diabetic bones has been identified, indicating yet another potential marrow function for bone: immune response tissue [57]. Local secretion of cytokines would have the ability to significantly increase their concentration in bone and cause inflammatory responses. This may be a way to decrease anabolic functions during times of disease or stress. Inflammatory factors, such as TNF-α and granulocyte colony-stimulating factor, are known to effectively suppress osteoblast maturation and bone formation [58–62]. Could adipocytes be involved in secreting these factors? It is more likely that they are the result of inflammatory cells within the marrow, since the inflammation is evident prior to adipocyte maturation. This suggests that anti-inflammatory agents, already in use to treat T1 diabetes, may prove beneficial in suppressing T1 diabetic bone pathology.

In summary, bone marrow adiposity is location dependent, varies with gender, functions similarly to and independently from other adipose stores and does not always correlate with bone loss. In T1 diabetic mice, leptin treatment and PPARγ antagonism prevent marrow adiposity but not bone loss, and vertebral bone loss occurs without increased marrow adiposity. These findings suggest that the increase in bone marrow adiposity is not directly causing bone loss. Nevertheless, based on the nature of physiology, marrow adiposity must have some role – either through cytokine or adipokine secretion, structural support, metabolic control or something else. With regard to treatment options, at this point, marrow adiposity seems an unlikely target on its own for the treatment of T1 diabetic bone loss. Anabolic treatments, perhaps in combination with marrow adiposity suppressing agents, seem a more likely effective treatment option.

Acknowledgements

Thanks to Regina Irwin, Katie Motyl and Lindsay Martin for their hard work and dedication, and for their critical review of this manuscript.

Financial & competing interests disclosure

Funding for the studies in Laura R McCabe’s laboratory was provided by the American Diabetes Association and NIH. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.