Osteogenesis Imperfecta, an Ever-Expanding Conundrum

Osteogenesis imperfecta (OI), or brittle bone disease, a rare heritable connective tissue disorder, was described more than a century ago.1 Its hallmark, bone fragility, may cause frequent fractures, spine and limb deformities, and chronic pain. Another characteristic of the OI syndrome is its wide spectrum of clinical severity, from the very mild to the lethal. Over the years, there have been several attempts at classifying the various forms of OI either by number or by degrees of severity (reviewed in Glorieux and Rowe2). All such efforts have only been partially successful because, except for the extremes of the spectrum, clinical types are overlapping. Furthermore, it has been oftentimes observed that within the same family, affected subjects show a wide variation in clinical presentations. This variability in phenotype severity is matched by an ever-expanding genotypic variability. OI is in most cases associated with mutations in COL1A1 and COL1A2, the genes encoding type I collagen, the main component of the bone matrix. Detailed analysis of the large number of identified mutations allowed for two main groups to be delineated. The first one corresponds to mutations leading to haploinsufficiency and overall to a mild phenotype. The second group includes mutations that cause structural abnormalities in the collagen molecule compromising the organization of the bone matrix. Associated phenotypes are usually moderate to severe or, rarely, lethal. All such mutations are transmitted as autosomal dominant traits. However, in a small number (less than 10%) of OI cases, no such association could be established, and in most of them there was evidence for autosomal recessive inheritance. In the first reported cases from a population isolate,3 mutations were identified in CRTAP, the gene encoding a cartilage associated protein part of the prolyl 3-hydroxylation complex involved in the processing of the type I collagen triple helix.4 Subsequently, mutations were found that affected the other two subunits of the complex, prolyl 3-hydroxylase 1 and cyclophylin B, as well as chaperones encoded by SERPINH1 and FKBP10 (reviewed in Forlino and colleagues5). Thus, up to that point all abnormalities could be linked to the synthesis, secretion, processing, and matrix integration of the type I collagen molecule. Recently, other forms, where no abnormalities related to collagen type I could be elicited, were identified by detailed analysis of undecalcified bone sections obtained from iliac crest biopsies. Pathophysiologic aspects of two such entities are discussed in this issue.6, 7 The first one, type V OI, was described 13 years ago.8 Unlike the other “new” forms of OI, it is transmitted as an autosomal dominant trait. Besides bone fragility, its main characteristics include short stature, ossification of the interosseous membrane of the forearm often leading to dislocation/subluxation of the proximal radial head, presence of radiodense bands under the growth plates, and the frequent development of hyperplastic calluses after fractures or corrective surgery. Its unique defining feature is, however, found on bone sections as a loss of the birefringent aspect of the matrix under polarized light illumination. It rather exhibits a meshlike appearance not observed in any other bone disease. Last year, two groups simultaneously reported the association of type V OI with a single identical mutation (c.-14C > T) in the untranslated region of IFITM5.9, 10 It introduces a new initiation codon that adds five residues at the N-terminus of the gene product, an osteoblast-specific transmembrane protein, BRIL.11 The findings were rapidly corroborated in a third publication,12 which added 42 patients to the 21 from the initial reports. Remarkably, all subjects carry the exact same mutation despite a wide variability in its phenotypic expression. Shapiro and colleagues further document this fascinating observation in an additional 17 patients.6 BRIL was discovered as part of a high-throughput screen for cDNAs encoding secreted and membrane proteins in osteoblastic cells.13 BRIL is part of an evolutionary conserved family of so-called interferon-inducible transmembrane (IFITM) proteins for which there are at least four closely related members in the human (IFITM1, 2, 3, and 10)14, 15 and two others in the mouse (IFITM6 and 7). The inclusion of BRIL as part of this family is based on genomic and protein structure considerations. BRIL, IFITM1, 2, and 3 are all clustered within 25 kb on chromosome 11; they all possess a similar gene architecture comprising two small coding exons separated by a short intron and potentially similar predicted protein topology having two transmembrane domains. Immunolocalization studies using tagged BRIL11 and IFITM316, 17 confirmed that they have both their N- and C-termini extruding out into the extracellular space, although this predicted model has been challenged recently at least for IFITM3.18, 19 Furthermore, other members like IFITM3 also seem to localize preferentially into the endosome compartment,18, 19 whereas BRIL localizes mostly to plasma membranes.11 Unlike other IFITM members, BRIL transcription is not responsive to interferons.20 Rather, its gene transcription is regulated by Sp1/Sp3 and GLI2 transcription factors.21 More important, the expression of BRIL is mostly restricted to osteoblasts, whereas other IFITMs are ubiquitous. A functional link between an aberrant BRIL and the specific characteristics of type V OI is not readily apparent. For instance, the so-called calcification of the interosseous membrane is in fact deregulated periosteal proliferation and new bone formation,12 and the hyperplastic callus formation may represent its extreme expression. At the same time, trabecular bone formation is impaired and bone fragility evident. Marked phenotypic variability despite the presence of an identical mutation in all patients also has to be entered into the equation. Solution to this conundrum will probably come from the study of various mouse models that are undoubtedly already on the treadmill. New therapeutic avenues may then be developed. The second article, published by Bogan and colleagues,7 is precisely about this next logical step, which is the generation and analysis of animal models mimicking OI subtypes. This report describes for the first time on the skeletal manifestation in a global PEDF-knockout mouse model. PEDF (pigment epithelium-derived factor) is a ubiquitously expressed secreted protein that accumulates in body fluids. PEDF presents sequence homology to the serpin protease inhibitor family (hence the name SERPINF1 for its encoding gene) but does not possess protease inhibitory activity. It has been ascribed many different functions such as antiangiogenic, antitumorigenic, and neurothrophic ones.22, 23 PEDF levels are also clinically associated with fat metabolism, the metabolic syndrome, insulin resistance, and diabetes type 2. Surprisingly, mutations in SERPINF1 were recently found to be the cause of OI type VI.24-26 Unlike the classic forms of OI, it is transmitted as an autosomal recessive trait and is characterized by moderate to severe bone fragility, causing fractures and a mineralization defect. Histomorphometric observations are akin to those found in osteomalacia. Under polarized light, bone matrix completely lacks its normal birefringent pattern, suggesting severe disorganization. In this model, first described 10 years ago,27 the prominent phenotype was one with increased vascularization defects of the stromal compartment of the pancreas and prostate and epithelial cell hyperplasia. Another PEDF-null mouse, generated independently,28 was shown to be fertile with no overt developmental defects and presented with mild and transient increases in eye vasculature.29 In light of the purported pleiotropic and whole body metabolic regulatory roles of PEDF, it would not be surprising that the full range of phenotypic manifestations is still incomplete and could expand in future studies. In the current article, Bogan and colleagues focus specifically on bone and present evidence that this mouse model recapitulates features observed in the skeleton of patients with OI type VI, although the severity appears to be less in the mouse than in human subjects. Consistent with previous reports on its immunolocalization and expression in bone cells,30-32 PEDF was found to be associated mostly with osteoblasts and osteocytes. Although PEDF is a secreted product, it did not seem to be readily detected over the extracellular matrix. Another interesting finding was the reported presence of PEDF in growth plate chondrocytes, which could in principle affect longitudinal growth, but this has not been studied. Cartilage and growth plate being essentially avascular, it is tempting to speculate that this could be attributable in part to the presence and anti-angiogenic function of PEDF. However, this is surprising at the same time, given the inverse relationship between PEDF expression and hypoxia. Coherent with the PEDF anti-angiogenic properties, immunostaining for a marker of endothelial cells (PECAM1) was found to be increased in the PEDF-knockout. Just as in OI patients, the mouse bone volume was decreased at 12 weeks of age. Histomorphometric parameters showed increases in osteoid thickness, osteoid maturation time, and mineral:matrix ratio. Mechanical testing revealed that the bones were structurally more brittle. Calvarial osteoblast cultures established from the PEDF knockout also formed more mineralized nodules, with an increased mineral:matrix ratio. Although a general systemic “priming” of osteoblasts in vivo is still possible to explain the in vitro data, the results would tend to suggest that the defect is cell autonomous. Gene expression monitoring in osteoblast cultures indicated that Enpp1, Dmp1, Phex, and Bsp were downregulated in the cells lacking PEDF. Given that these target genes are essential for bone formation and mineralization, they represent probable direct candidates to explain the osteomalacia phenotype. For instance, ENPP1 (nucleotide pyrophosphatase phosphodiesterase-1) acts as a physiological inhibitor of calcification by generating PPi, an inhibitor of mineral growth and also a substrate for alkaline phosphatase.33 Although speculative at this point, changes in the delicate balance in the local production of organic and inorganic regulators of mineralization could have come into play in the more global scheme of effects led by absence of PEDF. Alternatively, gene expression changes could have been a response of the osteoblast sensing an abnormal extracellular matrix or by a direct change in transcription because intracellular PEDF can also translocate to the nucleus.34 Distinction among those possibilities will require further investigations. Another point raised by Bogan and colleagues is whether the gene expression changes are a direct effect of PEDF acting on osteoblast cell surface–mediated signaling events. Binding studies have identified that PEDF is a ligand for at least three different receptors: patatin-like phospholipase domain containing protein 2 (PNPLA2),35 F1 ATP synthase β-subunit,36 and a laminin receptor.37 Although PNPLA2 was reported to be expressed in osteoblasts by Bogan and colleagues, it is still unclear whether any of these receptors are present on bone cells and whether they are mediating effects of extracellular PEDF on osteoblast signaling and contributing to downstream gene expression. Lessons learned from individual knockouts of those receptors do not permit at present the establishment of any correlation to the bone phenotype. For example, the PNPLA2 knockout mice die prematurely from cardiomyopathy associated with excessive lipid and glycogen accumulation.38 In addition to those aforementioned receptors, PEDF was also found to be an antagonist of the canonical WNT pathway, binding to LRP6 and LRP5 at least in retinal pigment epithelial cells.39 Thus, as another mechanistic possibility, the lack of PEDF in OI type VI could perturb WNT signaling and deregulate osteoblastogenesis. Highlighting the crucial role of WNT signaling in mesenchymal stem cells' commitment and differentiation to osteoblasts,40 we and others have recently reported mutations in WNT1 as a cause of moderate/severe OI.41-43 The most important question remains what is the role of PEDF in bone and what is the molecular mechanism of its action(s), PEDF being expressed almost ubiquitously and secreted systemically. Another question is whether bone requires local production of PEDF to act in close vicinity of or even inside the osteoblast. Previous reports have shown that PEDF interacts with a number of extracellular molecules including collagen itself, hyaluronan, and proteoglycans.22 Through such interactions, PEDF could potentially regulate initiation or progression of extracellular matrix mineralization. However, could the effects of PEDF be partly mediated intracellularly, within the secretory pathway of collagen modification and/or assembly? Is PEDF produced in other organs and delivered through blood essential for bone homeostasis and mineralization? Unfortunately, the current global mouse knockout model does not permit us to address those issues. Tissue-specific knockout mice, with PEDF ablated only in osteoblasts or other bone cells, will be essential to further understand the role of PEDF produced in bone versus systemically. Conversely, overexpression of PEDF in all tissues in mice under the control of the chicken beta-actin promoter did not lead to any overt defects, except for lower body weight and decreased growth rate and kidney size.39 Although the phenotype description of that mouse was scant, this might suggest that too much PEDF is of no obvious benefit and that circulating levels need to be kept within a “physiological” window. Beyond answering such basic questions, the model described nonetheless provides a proof of concept for the potential treatment of OI type VI. The great interest for PEDF in cancer therapy has already yielded a vast number of data on potential active forms of PEDF, and these will hopefully expedite delineating a treatment for OI type VI. The formulation and methodology for systemic delivery of PEDF as replacement therapy will certainly bring its own challenges, especially if it needs to selectively gain access to the bone matrix. Perhaps bone-targeting technologies44, 45 developed for treatment of other skeletal disorders such as hypophosphatasia could be applied. Clearly, the story is far from over. In the past two months, three groups have reported on recessive WNT1 mutations associated with moderate/severe OI.41-43 Already in 2001, mutations in LRP5, another component of the canonical Wnt signaling machinery, had been shown to cause the autosomal recessive osteoporosis-pseudoglioma syndrome, initially called osteogenesis imperfecta with blindness.46 Thus this other major regulator of bone homeostasis has to be considered in terms of pathophysiology and potential future therapeutic avenues.