An ELIXIR for bone loss?^†

Roman Gaul—roughly modern-day Switzerland, Belgium, and France sans Provence—was divided into three parts.1 In a similar fashion, the 48 member superfamily of nuclear receptors (NRs) is classified into three groups.2

First are “classical” steroid and related receptors with high-affinity ligands such as estrogen (ER), progesterone (PR), glucocorticoids (GRs), mineralocorticoids (MRs), androgens (ARs), thyroid hormone (TR), retinoic acid (RAR), and 1,25 dihydroxyvitamin D3 (VDR).3 The first four bind DNA as homodimers and the others as heterodimers with the retinoid X receptor (RXR). Second, rapid advances in DNA cloning technology revealed that all mammalian genomes contain many proteins with domain structures similar to classical NRs.4 Because the endogenous ligands for the novel molecules were unknown, they were designated orphan receptors. Third, those orphans whose endogenous partners were identified became, appropriately, adopted receptors. Notably, while maintaining the canonical 11–13α helical loop structure, many of the orphan/adopted receptor (OAR) ligand domains bind multiple agonists.4 Good examples are the aryl hydrocarbon and constitutive androstane receptors (ARH and CAR respectively), which recognize a wide range of xenobiotics and stimulate hepatic detoxification.

The OARs are interesting in several novel ways. An initial surprise was the identity of endogenous ligands, which are largely lipophilic and often relatively common metabolites. Thus, liver X receptors α and β (LXRs) bind oxysterols,5, 6 making them sensors of cholesterol metabolism, whereas peroxisome proliferator-associated receptors (PPARs α, β, and γ) are activated by fatty acids and regulate multiple aspects of lipid metabolism and more.6 Farnesoid X receptors (FXRs) ligate bile acids and their amphipathic salts,7, 8 whereas the liver receptor homolog 1 (LRH-1) binding pocket can accommodate a range of phospholipids.9 Reflecting the problems with defining physiological agonists, a number of NRs are still orphans despite intensive research. For example, it was shown only recently that the important subgroup of retinoic acid–related orphan receptors (RORs), which regulate numerous molecular events,10 bind oxysterols and thus suppress transcription.11 Given that ligation of RORs leads to activation of many genes,10 it is reasonable to speculate that the endogenous ligand(s) for these remain to be identified. (Nuclear RORs should not be confused with the receptor tyrosine kinase-like family of cell membrane receptors [Rors 1 and 2], which play important roles in Wnt-mediated osteoblast and osteoclast function.12, 13)

The second novel aspect comes from comparing the mechanisms of action of OARs versus classical NRs. Similarities include the fact that DNA binding sites are almost invariably direct or inverse repeats (DRS or IRs) of a six-base pair sequence in target promoters, separated by a variable number of base pairs.14 Conversely, even though they usually act with RXR, PPARγ and RORα can bind as monomers, whereas LRH-1 does so invariably.4 Finally, whereas classical receptors reside in the cytoplasm and translocate only following hormone binding, the same is not true for OARs, which are always nuclear.

Third, the target organs/cells for many of the newly identified receptors are different from those associated with many classical NRs. For example LXRs are central to cholesterol homeostasis, via cells as diverse as hepatocytes, adipocytes, enterocytes, and macrophages. Furthermore, they synergize with different combinations of FXRs (intestine),5 PPARs (the immune system, liver, and macrophages),5 or LRH-1 (female reproduction and the hepatic acute phase response).9 In summary, LXRs are emerging as critical molecules in the regulation of a wide range of homeostatic and disease-related processes, including cancer, and thus represent potential drug targets. Their ability to execute this plethora of functions in the face of limited ligands is explained by differential expression of the two isoforms, LXRα and LXRβ, as well as by specificity for their nuclear targets by events at the level of DNA binding; thus, tissues with high levels of lipid metabolism, such as liver, intestine, and macrophages, express LXRα, whereas LXRβ is ubiquitous.

The well-characterized mechanisms of NR-controlled transcriptional activation or repression involve binding to a cognate DNA repeat, followed by recruitment of coactivators or co-repressors, respectively.14 ER and GR also act as trans-repressors wherein they do not bind directly to DNA, but rather interact with other proteins, impeding the gene-regulatory capacity of the latter. The issue of posttranslational modification (PTM) of classical steroid receptors15 is well studied but less is known about the analogous processes in OARs. The data suggest that most if not all members of the NR family undergo PTMs, including serine phosphorylation, lysine acetylation, and ubiquitinylation and, perhaps less commonly, SUMOylation. Although studies are limited, several OARs mediate a novel form of trans-repression, involving SUMOylation. Thus, addition of the small ubiquitin-related modifier (SUMO) moiety to PPARγ or LXRs suppressed release of inflammatory cytokines by macrophages following exposure to Toll-like receptor (TLR) ligands.16 A similar observation was made recently in the context of hepatocyte-mediated acute phase response.17 In both instances the SUMOylated OAR was recruited to a preexisting repressor complex, suppressing its disassembly, with the net result being decreased transcription of a subset of target genes.

A number of NRs impact bone mass, often by regulating the number and function of osteoclasts, but a detailed description of their cellular targets and roles is beyond the scope of this commentary. In brief, estrogen regulates the osteoclast in multiple ways, suppressing production of inflammatory cytokines by macrophages and T cells,18 and separately by decreasing production of reactive oxygen species, which stimulate osteoclast activity.19 Activation of the VDR in osteoblasts increases expression of receptor activator of NF-κB ligand (RANKL), the key osteoclastogenic cytokine, by binding to multiple vitamin D response elements (VDREs) in the regulatory region of the gene.20 In contrast, activation of GRs results in decreased expression of osteoprotegerin,21 the endogenous RANKL decoy receptor; in either instance the RANKL/osteoprotegerin (OPG) ratio is increased, resulting in augmented osteoclast generation and hence bone loss. Additionally, glucocorticoids suppress osteoclast formation directly.22 Finally, diabetic patients given rosiglitazone lose bone because the drug suppresses osteoblastogenesis while simultaneously increasing formation and activity of osteoclasts.23 In the latter instance the PPARγ effector coactivator 1α (PGC-1α) acts as a coactivator of osteoclast differentiation. PPARγ also induces yet another NR, estrogen-related receptor α (ERR-α), which increases mitochondrial number, thereby enhancing production of the protons required for dissolution of the inorganic component of bone.

The above background establishes a framework in which to dissect the work of Kleyer and colleagues24 in this volume on the role of LXRs in bone biology. Three previous studies examined this question, using different approaches. Long-term administration of two LXR agonists to wild-type mice had minimal impact on bone mass, structure, or turnover.25 When the in vivo phenotype of LXR-deleted mice was analyzed,26 bone mass of single- or double-knockouts was altered minimally, although mice lacking LXRα had more, but less active, cortical osteoclasts, whereas LXRβ-null animals had decreased numbers of hyperactive trabecular osteoclasts. Interestingly, the double-knockout had unchanged bone mass, perhaps as a result of the counterbalancing effects of the two receptors. Of importance, levels of OPG were unchanged in the single-knockouts but significantly higher in the doubly-null cohort. The same authors went on to show that an LXR agonist suppressed in vitro osteoclast formation from bone marrow macrophages (BMMs) exposed to the key osteoclastogenic cytokines macrophage colony-stimulating factor (M-CSF) and RANKL via a mechanism involving blockade of RANKL-induced signaling.27

The current authors also had access to mice carrying all three mutant LXR genotypes, but chose to use the animals mainly as a source of osteoblasts and BMMs for in vitro studies. The two cell types were co-cultured in the presence of an LXR agonist or parathyroid hormone, which stimulates osteoclastogenesis in this setting by enhancing production of M-CSF and RANKL by the mesenchymal cells. The data show that the osteoblast is the major target of treatment with LXR, which decreases secretion of RANKL but leaves that of OPG unchanged. Consequently, osteoclast formation and function are diminished. Incongruously, when BMMs monocultures were performed, osteoclast number was unaltered but their size and function was decreased. The sole in vivo experiments involved oophorectomy of wild-type mice given an LXR agonist, which completely suppressed trabecular bone loss, while increasing serum OPG and decreasing that of RANKL.

The present in vitro and in vivo studies are worthy of discussion on several accounts. First, while seldom used, the use of embryonic stems cells on a C56/BL6 background for gene targeting is attractive, because it eliminates the need for back-breeding to eliminate the mixed genotype associated with the ubiquitous 129/sv-based approach. A concern, however, is whether the albino embryonic stem (ES) cells (the basis for establishing germline transmission) are indeed of the same genotype as the commercially available C57/BL6 mice used in the current work. This point is relevant because there is a marked difference between the current and previous data with respect to in vitro osteoclastogenesis of BMMs. While Kleyer and colleagues24 report that osteoclast number is not decreased when BMMs are the sole cellular target for LXR activation in culture, Remen and colleagues27 found that the same agonist inhibits polykaryon formation via an LXRβ-dependent mechanism. Second, the authors may be underestimating the decrease in RANKL expression by osteoblasts treated with the LXR agonist. Most osteoblast-associated RANKL is membrane-bound and so an ELISA measures only that which has been cleaved. Third, a possible additional explanation for decreased osteoclast function relates to the impact of LXR agonists on membrane cholesterol and the cytoskeleton, respectively. LXRβ, the major isoform in BMMs, mediates reverse cholesterol transport, which results in depletion of macrophage cholesterol levels.28 Again, this is germane because pharmacological removal of the steroid from osteoclastic culture decreases generation of the ruffled border,29 the functional organelle of the osteoclast. Similarly, dendritic cells treated with GX3965 exhibit altered adhesion and migration arising from decreased expression of fascin,30 an actin-bundling protein that controls formation of podosomes, a critical prerequisite for osteoclast function. Fourth, the claim that osteoblast-BMM co-cultures replicate the in vivo environment is probably an oversimplification. The work of Pacifici (see Ref. 31 and references therein) shows that at least mice T cells, which are responsive to LXR agonists, are important players in bone regulation in vivo both basally and specifically after oophorectomy, when they act as a source of tumor necrosis factor α (TNFα), a potent stimulator of bone marrow stromal cell–derived M-CSF and RANKL.

Finally, and most importantly, the authors will have the opportunity to perform a series of important studies both in vitro and in vivo that will likely reveal important mechanistic and potential translational insights. Further investigation of the mechanism by which LXRs suppress RANKL expression could include examination of the regulatory region of the gene for direct repeat 4 (DR4) elements and confirmation of this fact by a combination of electrophoretic mobility gel shift assay (EMSA), reporter assays, and, most convincingly, LXR chromatin immunoprecipitation sequencing (ChIP-seq). Indeed, whole-genome LXR ChIP seq has been reported for mouse liver,32 and mouse33 and human macrophages.34 Oophorectomy of their three genotype-altered mice should identify which isoform is critical for the bone loss associated with postmenopausal osteoporosis. Separately, T cell depletion in vivo by administration of anti CD4 and CD8 antibodies will uncover the role of these cells both basally and in response to estrogen loss. An exciting idea will be to determine if LXR agonists can ameliorate or induce rheumatoid arthritis. The reason these studies may be important can be found in the original report of Prawitt and colleagues,25 in which long-term administration of an LXR agonist had no impact on bone mass. If indeed, as suggested by the current work, modulation of LXR signaling suppresses bone loss only in the context of pathophysiological bone resorption, the translational implications may be enormous.

Returning to Gaul, where Caesar sealed the Roman conquest in 52 BCE by defeating a chieftain named Vercingetorix, the general is still alive in the minds of the huge cohort that read the hilarious “comic book” series starring Asterix and colleagues. Among these a central figure is Getafix (related to Vercin Getorix?), the purveyor of potions and the like. Perhaps a future volume will include his cry to the legions suffering from or at risk to osteoporosis: “Come get my eLiXiR for bone.” Hopefully Cacofonix will not drown out his message.

Disclosures

The author states that he has no conflicts of interest.