Osteogenesis imperfecta (OI) is a heterogeneous group of heritable disorders affecting bone and connective tissue. It is most notably characterized by bone fragility, decreased bone mass, and recurrent fractures that can lead to skeletal deformities. The clinical severity can vary widely depending on the genetic mutation involved.1–4 The majority of individuals with OI have dominantly inherited mutations in either COL1A1 or COL1A2, which encode the α chains of type I collagen. In a minority of individuals with OI, however, the disease exhibits an autosomal recessive inheritance pattern. In most cases in which the affected genes have been identified, they have been found to encode proteins whose functions can be plausibly linked to the maturation and processing of collagen. For example, three of these genes encode components of the prolyl-3-hydroxylase complex, which can modify Pro986 in the α1 chains of type I collagen (CRTAP, LEPRE1, and PPIB), and two of these genes encode chaperones (SERPINH1 and FKBP10), which may be important for proper folding of procollagen (MIM #605497, #610339, #123841, #600943, #607063).5 Hence, the genetic data have thus far been consistent with the hypothesis that abnormalities in collagen biosynthesis and maturation are primarily responsible for the etiology of OI. Two recent papers by Becker et al.6 and Homan et al. (in this issue),7 however, have reported mutations in another gene, SERPINF1, which encodes pigment epithelium-derived factor (PEDF),8, 9 whose link to collagen is less apparent, thereby raising the possibility of an alternative disease mechanism for OI.
In the March issue of the American Journal of Human Genetics, Becker et al. reported the identification of three different truncating mutations in SERPINF1 in four individuals with autosomal recessive OI.6 Becker et al. initially focused on one patient whose parents were second cousins from the United Arab Emirates. Based on the consanguinity, the authors reasoned that the patient likely would be homozygous for the same mutation, and they carried out exome sequencing in order to identify the gene responsible. By applying a variety of statistical analyses, the authors homed in on one likely candidate gene, namely SERPINF1, in which they had identified a single nucleotide change predicted to cause a premature stop. This patient was homozygous for the point mutation, as was the patient's affected brother (both parents and two unaffected sisters were heterozygous). The authors then went on to examine the SERPINF1 gene in other unrelated patients with similar clinical features and identified two additional truncating mutations in patients of Turkish descent with parental consanguinity.
In the current issue of Journal of Bone and Mineral Research, Homan et al. independently report two additional SERPINF1 mutations, each predicted to cause premature stops.7 This group identified one of these homozygous mutations in members of a consanguineous French Canadian family and the other in an Italian patient, all of whom were classified as having OI type VI, a unique form of OI first described by Glorieux et al. in 2002.10 The investigators performed extensive homozygosity mapping and next-generation sequencing of the candidate gene region in order to identify SERPINF1 as the causative gene. OI type VI is clinically different from the other recessive forms in that the children do not fracture until approximately 6 months of age, unlike those with other recessive OI disorders who fracture at birth. Bone biopsy reveals a distinctive pattern; specifically, there are large amounts of osteoid devoid of mineralization and a “fish-scale pattern” of bone lamellation.10, 11 Homan et al. provided histologic and bone biopsy data for their three patients; all revealed the characteristic excess of unmineralized osteoid typical of OI type VI, and one had the additional unique fish-scale appearance with abnormal lamellation. Biopsy data, however, were not provided by Becker et al. Of note is that patients with OI type VI are also unique in that their responses to cyclical pamidronate therapy result in less improvement in mobility and a lower reduction in fracture incidence than in patients with OI types I, III, and IV.11
The types of OI have expanded as the number of responsible identified genetic mutations has increased, thereby resulting in a system of classification that has become increasingly complex. This problem was addressed in the “Nosology and Classification of Genetic Skeletal Disorders: 2010 Revision,”4 with the goal of basing the classification on clinical severity as per the original classification by Sillence et al.12–14 Before these two studies implicating SERPINF1, it had been suggested that mutations in FKBP10, a chaperone protein, could potentially be responsible for OI type VI15 (MIM #607063). Homan et al., however, found no FKBP10 mutations in their reported cohort and propose, instead, that SERPINF1 mutations are the cause of OI type VI. For purposes of discussion of these two reports, SERPINF1 mutations are classified as causing OI type VI (MIM #613982), and the patients discussed in the Becker et al. article have clinical findings consistent with those in the current Homan et al. report.
Both Homan et al. and Becker et al. provide convincing genetic data showing that homozygous mutations in SERPINF1 could indeed cause OI type VI, and that it is the loss of function of PEDF, the protein product of the SERPINF1 gene, that leads to the bone abnormalities. All five mutations identified by the two groups are predicted to lead to protein truncations, and three of the five mutations occur relatively early in the coding sequence. Moreover, two of these mutations were shown to occur early enough in the coding sequence to cause reduced levels of the mutant transcript, consistent with nonsense mediated decay. Finally, the two 3'-most mutations were identified in the last exon, and studies by another group have shown that the C-terminus of PEDF is essential for its secretion from transfected cells. Hence, all of the genetic data suggest that most, if not all, of these mutations result in null alleles.
PEDF is a secreted glycoprotein in the serpin superfamily. Interestingly, in mice globally lacking PEDF secondary to an engineered deletion mutation of the Serpinf1 gene, there have been no bone abnormalities reported.17 It is certainly possible that Serpinf1−/−mice do have bone abnormalities that may become apparent only after a more detailed analysis. Becker et al. also point out that the deletion mutation engineered in the mice eliminated the middle portion of the molecule, leaving the formal possibility that the remaining coding sequence might suffice to preserve whatever activity is required for proper bone development and homeostasis. It is also possible that PEDF might simply play different roles in mice versus humans. Given that PEDF is part of a large family of proteins, it is not inconceivable that these proteins might exhibit functional redundancy and that the relative importance of a given family member for a particular biological process might vary from species to species.
PEDF was originally identified as a protein secreted by retinal pigment epithelial cells and shown to be capable of inducing a neuronal phenotype in cultured retinoblastoma cells.8, 9 Subsequent studies showed that PEDF is a potent inhibitor of angiogenesis that seems to block the effects of vascular endothelial growth factor (VEGF).18, 19 Indeed, the role that PEDF plays in regulating the vasculature in vivo has been clearly shown in Serpinf1−/−mice, which exhibit increased growth of stromal vasculature and epithelial tissues in the prostate and pancreas.17 As a result, there has been considerable interest in developing PEDF as an anti-angiogenic drug, and clinical trials for PEDF are underway to treat patients with age-related macular degeneration.19, 20
Additional work has identified other potential functions for PEDF, including studies showing that PEDF is expressed by osteoblasts and, to a lesser extent, osteoclasts in culture21 and by both chondrocytes and osteoblasts during endochondral bone formation.22 Moreover, PEDF has been shown to be capable of inducing expression of osteoblast differentiation markers in a pre-osteoblastic cell line as well as increasing mineralized nodule formation by osteosarcoma cells.23 PEDF has also been shown to be capable of inhibiting osteoclast differentiation, RANKL-mediated osteoclast survival, and osteoclast bone resorption activity.24 Hence, loss of function of PEDF in patients with inactivating mutations in SERPINF1 might be predicted to result in decreased numbers of osteoblasts and/or enhanced numbers or activity of osteoclasts, which would then lead to decreased bone formation and/or enhanced bone resorption.
An alternative model is that the effect of mutations in SERPINF1 is indirect, namely, that the primary defect is in vascular development, which then leads secondarily to the OI phenotype. In support of this model, the balance between PEDF and VEGF expression is believed to regulate neovascularization in bone,21, 22, 25 and there is evidence that VEGF-mediated capillary invasion during bone formation is important for proper cartilage remodeling and ossification.25 If this model is correct, the question is whether one can intervene in patients with disease by regulating the extent of vascularization in bone remodeling. If so, the potential therapeutic strategies would encompass not just those related to restoring PEDF function but also to any interventions designed to block angiogenesis, keeping in mind that VEGF may be acting upstream of PEDF to regulate this process. Homan et al. point out that this may also have implications for the use of systemic anti-angiogenic agents utilized for the treatment of other patients (for example, those with cancer) as such agents may have unexpected effects on bone.
Of course, it is also possible that PEDF may play a direct role in regulating collagen biosynthesis or processing that has not been previously recognized, and that the pathogenesis of OI in patients with mutations in SERPINF1 may not be fundamentally different from that in other patients with OI. Interestingly, PEDF is known to bind avidly to collagen type I,25, 26 although this might simply represent a mechanism by which PEDF is sequestered in the extracellular matrix. In this respect, Becker et al. performed biochemical analyses to examine effects on collagen using dermal fibroblasts isolated from one of the patients with a SERPINF1 mutation and found no alteration in post-translational modifications or collagen secretion. The authors state that they cannot rule out subtle defects, such as delayed secretion, as has been observed in fibroblasts isolated from patients with other mutations.15, 27–29 Nevertheless, the identification of SERPINF1 mutations in patients with OI raises the possibility that the OI phenotype may result from defects in pathways and processes that may be collagen-independent.
With respect to OI, Homan et al. and Becker et al. raise both novel and crucial questions regarding the pathways and molecules that may play a role in the etiology of this disorder. These findings could have important implications for the diagnosis of OI type VI, including screening for circulating PEDF levels. These studies also have potential implications for the development of new strategies for therapeutic intervention, possibly involving PEDF replacement. The broader question raised by these two important studies is whether PEDF plays a role in regulating homeostasis in both developing and mature bone. In this respect, could altered levels of PEDF activity lead to abnormal bone formation and/or loss of bone mineral density in other disease states, including osteoporosis? Similarly, can manipulating levels of PEDF be a therapeutic strategy for enhancing bone structure and function in patients, possibly titrating the effect according to the degree of bone loss? Given the enormous number of patients who suffer from low bone mass, it will be important to elucidate precisely what role PEDF plays in regulating bone development and homeostasis. Critical to understanding this process will be a detailed understanding at the molecular and cellular level of the regulation and function of PEDF in bone, as well as a more detailed characterization of the consequences of PEDF loss in experimental animals, most notably in Serpinf1−/−mice.