Bone and Mineral Metabolism: Where Are We, Where Are We Going, and How Will We Get There?

Abstract

Context:

Advances in diagnosing and treating metabolic bone diseases will require ways to assess cellular signaling within human bones, ideally noninvasively. Only then will we be able to fully harness the increased molecular understanding of bone that derives from human genetics and model organisms, primarily rodents. New hormones regulating mineral ion homeostasis surely remain to be discovered, probably through advances in the study of human genetic disease.

Bones are pretty amazing. Evolution designed them to last a few decades and now they are expected to last for as long as a century while leveraging our muscles, protecting our organs, regulating our calcium and phosphorus homeostasis, and supporting hematopoiesis. It is not surprising that, despite their impressive design, bones fail to protect us as we age: two million osteoporotic fractures occur in the United States each year. We have found ways to identify weak bones and strengthen them enough to prevent a fraction of those fractures, but our understanding, diagnostic tools, and available therapies remain rudimentary. In the first part of this essay, I will briefly summarize current approaches to dealing with osteoporosis, their limitations, and how we might progress. Then I will discuss diseases of mineral metabolism.

Osteoporosis

We have fairly good ways to predict who might fracture by combining measurements of bone mass with assessment of epidemiological risk factors, most importantly age (1), but we need new approaches to choose people for therapy more appropriately.

A major predictor of fractures is bone mass. We measure bone mass using dual-energy x-ray absorptiometry (DXA), a tool with a precision unrivaled in clinical endocrinology. Nevertheless, most people who fracture have bone density measured by DXA that is outside of the so-called “osteoporosis” range (T score < −2.5). Part of the explanation for this paradox is that most hip and forearm fractures occur after falls, and some people fall a lot more than others. But the limitations of DXA measurements probably explain much of why some people fracture at higher bone densities than others. What we really want to measure is bone strength, not bone mass, so that we can compare that strength to plausible stresses caused by falls. Bone strength reflects the amount of bone (bone mass) but also architectural parameters (for example, how the struts of trabeculae connect together) and material properties of bone (for example, how strongly collagen is cross-linked). We are getting better at measuring architectural parameters now with the few high-resolution peripheral quantitative computed tomography (HR-pQCT) machines that can identify trabeculae and cortical pores. But these machines too have limitations: the resolution is not yet great enough to identify individual trabeculae satisfactorily and to measure degrees of cortical porosity accurately. Second-generation scanners may provide more useful information. Equally importantly, HR-pQCT today can only image peripheral bones: the tibia and radius, not the hip or spine in which most fractures occur. And HR-pQCT, as an x-ray technique, cannot begin to identify important material properties of bone. Minimally invasive techniques, such as microindentation of bone (2), are beginning to develop indices of bone's material properties, but much needs to be done to develop noninvasive ways to assess the molecular parameters crucial to bone strength. It would be wonderful to be able to distinguish the mineral phase from the bone matrix and thereby diagnose osteomalacia noninvasively, for example. That is currently impossible, although investigations using magnetic resonance imaging suggest that this may become feasible (3). Finite element analysis technology, originally invented to predict the strength of bridges and airplanes, has been applied to the analysis of bone strength with some success, but this has been limited by the available structural data used for the calculations (4). One can hope that over the next several years, continued progress in imaging will allow bioengineers to come up with better predictive estimates of bone strength in the clinical context.

As impressive as the new imaging methods have been, they look at bone matrix and ignore the complicated cell biology responsible for laying down and destroying that matrix. At the moment, the only way to get at those cells is with bone biopsy, a technique sufficiently invasive that patients do not flock to studies that involve biopsies. Bone biopsies can teach us a lot about the cells buried in bone (osteocytes) and on the bone surface (osteoblasts, lining cells, and osteoclasts), but they also reveal our profound ignorance of the mesenchymal cells even a few microns away from the bone surface. Where do osteoblasts, the cells that lay down the bone matrix on the bone surface, in adult humans come from? At the moment, that's anybody's guess; no one knows for sure. With marrow aspiration, we can isolate cells that, when grown on plastic or injected sc in mice, can become osteoblasts, chondrocytes, and adipocytes. These cells can self-renew, so they are called mesenchymal stem cells by some. They may prove useful as reagents in regenerative medicine, but their roles, if any, in actually forming bone in normal skeletal homeostasis has not been established. As yet, no one can identify precursors of osteoblasts unambiguously on bone biopsies, and this means that no one has yet been able to study, even in vitro and in animal models, the hormonal and paracrine signals that regulate the proliferation, differentiation, and lifespans of osteoblast precursors. A number of groups have used lineage-tracing technology and cell surface markers to identify precursors of osteoblasts in vivo in mice, but these are all preliminary studies (5–9). The lineage-tracing experiments identify groups of cells that include the precursors of osteoblasts, but all of the marked populations are genetically heterogeneous, probably with similarly heterogeneous fates. Still, the iterative process of using multiple approaches to identify precursors of osteoblasts at differing stages of differentiation is likely, in the next few years, to take us to where hematologists, who identified hematopoietic stem cells and their progeny years ago, have been for some time. An important goal would be to mark key osteoblast precursors in vivo in humans in a way that would allow the tracking of the fates of these cells and the effects of hormonal/paracrine manipulation and the effects of specific drugs and genetic polymorphisms on those responses. Right now, therapies for osteoporosis, although they target varying molecules and pathways, do not distinguish patients on the basis of the cellular basis of their disease. Only when we begin to understand the variation in the cellular basis of disease pathogenesis will we be able to devise and tailor therapies to specific kinds of pathophysiology.

We have a few serological markers that reflect roughly total rates of bone formation (for example, procollagen I-amino-terminal peptide) or of bone resorption (for example, immunogenic fragments of collagen 1 containing characteristic cross-link moieties [NTX and CTX]). These assays that all measure portions of collagen I, the major protein of bone and its precursor, procollagen, reflect the action of osteoblasts to deposit bone matrix and osteoclasts to resorb the matrix, but they cannot get at the mechanisms regulating the numbers and activities of the bone-forming and -resorbing cells. Investigators are beginning to assess blood levels of local bone regulators, such as the wnt antagonist, sclerostin (10), but the meaning of the levels in the blood of such proteins, which act only a few cells away from the cells that make them, is not known.

To understand the enormous variation in bone strength between individuals and within individuals over time, I suspect we will need new ways of assessing, in humans in vivo, the activities of pathways known to regulate bone cells, such as the wnt, notch, fibroblast growth factor (FGF), bone morphogenetic protein, IGF, and TNF pathways and their molecular components, on cells in living bone. We are not close to attaining that goal, but advances in marking cells with antibodies and detecting metabolites, combined with new imaging modalities, may make that goal attainable.

When we can identify relevant bone cells and their precursors in vivo, as well as the pathways controlling their fates, one can imagine then being able to apply the endocrinologist's tools, stimulation and suppression tests, to bring out abnormalities in these cells and pathways. In the case of bone, of course, the stimuli that are relevant include not just chemical activators or inhibitors of pathways, but also mechanical forces that can be applied to elicit important responses. When we find ways to assess the cellular responses within bone, I am guessing that we may then be able to exploit more fully the already substantial data from genome-wide association studies regarding bone mass and fractures. The goal would be to develop ways of assessing the biology behind an individual's bone mass/strength and perhaps develop and identify individual therapies adapted toward the specific biology of each individual. Right now, the idea of personalized medicine in the osteoporosis world has no real meaning. Therapies are designed for broad epidemiologically defined groups and result in variable effectiveness. Our most powerful therapies, the anabolic agents, PTH (1–34) and antibodies to sclerostin, work well but lose their effectiveness rather quickly for, at the moment, completely unknown reasons. Some answers will come from studies of genetically manipulated mice, but we will need equivalently revealing studies in humans to understand how these agents work and how to develop better agents for strengthening bones.

Mineral Metabolism

The most exciting advance in the understanding of mineral metabolism in the last decade was the discovery of the phosphate-regulating hormone, FGF23. FGF23 acts on the renal proximal tubule to decrease the resorption of phosphate and to suppress the activity of the 25-hydroxyvitamin D 1-α hydroxylase. New hormones do not come along all that often any more, so it is worth reflecting upon how the actions of FGF23 were discovered and why they were not discovered sooner. The latter question is easy to answer; there is no FGF23 gland, so that the classic approaches of endocrinology, gland ablation and hormone purification from glands, was not possible. FGF23 is made primarily by osteocytes, the major cells in bone, surrounded by bone matrix (containing most of the body's phosphate, of course), so that traditional gland ablation experiments could not be contemplated. In fact, the actions of FGF23 were discovered serendipitously when examining the causes of inherited forms of rickets. Most revealingly, in autosomal dominant hypophosphatemic rickets, the mineralization disorder was caused by point mutations in the coding region of FGF23. This result was surprising because, before the discovery of FGF23, there were no obvious physiological mysteries that demanded the presence of a new important hormone. Nevertheless, we now know, largely through the studies of FGF23 knockout mice, that normal levels of FGF23 are required for normal phosphate and vitamin D homeostasis. It took human genetic studies and studies of a cancer syndrome, tumor-associated osteomalacia, to demonstrate potential roles for FGF23 in humans.

The serendipitous path to discovery of the roles of FGF23 in mammals has an important implication regarding the completeness of our list of important regulators of calcium and phosphate homeostasis. We have to anticipate that the list is incomplete. Further discoveries of such hormones are likely to come from further delineation of genetic causes of mineral disorders. It seems likely that the use of exome and whole genome deep DNA sequencing will reveal such new hormones in the next few years, as all the Mendelian diseases yield their mysteries.

That FGF23 was discovered only a few years ago also explains why we really know very little yet about FGF23 regulation and action. We know that very low levels of blood and total body phosphate are associated with low levels of FGF23 and that high levels of phosphate are associated with high FGF23 levels. But the changes in FGF23 levels in response to acute changes in blood phosphate occur slowly, and it seems clear that the regulation of FGF23 by phosphate is not a simple relationship analogous to the way that calcium regulates PTH secretion. It is not yet clear, for example, whether osteocytes directly sense changes in blood levels of phosphate and change the secretion of FGF23 accordingly. And, if phosphate does regulate FGF23 production and secretion in osteocytes directly, the intracellular mechanisms that phosphate regulates are not at all clear. Cells use sodium-phosphate transporters to move phosphate into cells, and some cells respond to the increase of intracellular phosphate by changing activation of kinases such as ERK1 and -2. But whether these kinase pathways or others in osteocytes regulate FGF23 synthesis and secretion in response to changes in extracellular phosphate remains to be determined.

We are only beginning to understand how FGF23 acts. FGF23 is a member of the large FGF family of ligands, most of which work by activating FGF receptors 1–4. The activation of this receptor family by FGF23 is weak and requires the presence of an FGF23-binding coreceptor, klotho. The genetic evidence that klotho mediates the specificity and actions of FGF23 is substantial, but many mysteries about klotho remain. Soluble forms of klotho circulate, and their roles in mediating actions of FGF23 and/or independent actions should be clarified in the next few years. FGF23 not only regulates phosphate levels and 1,25-dihydroxyvitamin D levels through actions on the proximal tubule, but also increases calcium and sodium reabsorption by activating the TRPV5 channel and by increasing the abundance of the sodium–chloride cotransporter, respectively, in the distal tubule. These actions of FGF23, and doubtless others, do not fit easily into a tidy role for the hormone primarily in phosphate metabolism. We need to be prepared for a more encompassing paradigm to explain the physiological roles and regulation of FGF23.

Not all of the actions of FGF23 require the presence of klotho. When klotho is removed from parathyroid cells in vivo, FGF23 can still decrease PTH secretion (11). Strikingly, high levels of FGF23, like those seen in renal failure, can increase cardiac hypertrophy in an apparently klotho-independent fashion (12). Because FGF23 is a better predictor of mortality in renal failure than measures of phosphate burden, this direct klotho-independent role of FGF23 in the heart will continue to receive much deserved attention.

The excitement about FGF23 should not cause us to forget how little we understand about PTH. Perhaps the largest mystery about PTH secretion is still our ignorance of the mechanisms whereby high extracellular calcium suppresses PTH secretion. The efforts of Brown (13) and others have shown that the calcium-sensing receptor somehow responds to changes in extracellular calcium levels by activating a number of intracellular signaling pathways that mediate the suppression of PTH secretion. But how these changes in intracellular signaling lead to suppression of PTH secretion remains a complete mystery. Undoubtedly, the continued absence of a cell line that mimics the properties of parathyroid chief cells has held back progress. Perhaps the use of three-dimensional scaffolds or other approaches successful with culture of other epithelial cell types will provide a suitable model to allow advances in solving this remaining mystery.

The last part of the title of this essay is “how we will get there.” Readers may have noticed that I've had an easier time talking about what we need to learn than how we will learn it. The good news, of course, is that the methods for discovery in human physiology and disease have never been stronger. Powerful animal models, undoubtedly using CRISPR-Cas technology not just to change genes in mice and cells but also in nonmurine animal models, will allow the generation of relevant animal models. Whole genome sequencing should lead to the identification of the causes of all diseases inherited in simple Mendelian fashion and greater understanding of polygenic diseases. But, undoubtedly, new technologies, undreamt of yet, will allow further progress in the next decade. Mechanisms for the design of sophisticated clinical trials, along with increasingly sophisticated ways of extracting important clues from the “big data” sources made possible through the introduction of computerized databases for managing patient care will provide ways to apply the insights of discovery science to the cure and treatment of disease. I'm not worried about a scarcity of good ideas if we can convince our government and other funding sources to support this effort.

Acknowledgments

This work was supported by National Institutes of Health Grant DK011794.

Disclosure Summary: H.M.K. consulted for Novartis and performed research sponsored by Amgen.

Abbreviations

 
• DXA

  dual-energy x-ray absorptiometry

 
• FGF

  fibroblast growth factor

 
• HR-pQCT

  high-resolution peripheral quantitative computed tomography.

References