|Home | About | Journals | Submit | Contact Us | Français|
The osteochondroma is a common, benign, primary tumor of bone. A mechanism for its pathogenesis has not been identified, but loss of function of EXT genes is implicated in sporadic and hereditary multiple osteochondromas. Recent advances in the understanding of other molecular signaling pathways in the physis cast doubt on the latest pathogenetic theories. These advances are reviewed and used as the basis for a revised theory for pathogenesis: A clone of proliferating chondrocytes without functional EXT1 (or EXT2) expression fails to produce heparan sulfate; lack of heparan sulfate at the cell surface disrupts fibroblast growth factor signaling and Indian hedgehog diffusion, leading to focal overproliferation and adjacent bone collar deficiency, respectively; together these effects are proposed to contribute to osteochondroma pathogenesis.
Osteochondromas, or exostoses, are cartilage-capped excrescences of bone that develop during physeal growth. Although their incidence may be underestimated given the fact that many sporadic osteochondromas cause no symptoms, they are nonetheless considered the most common of benign primary bone tumors 55.
Hereditary multiple exostoses (HME), alternatively called diaphyseal aclasis or osteochondromatosis, is a highly penetrant, autosomal dominant trait characterized by slightly stunted growth of long bones and multiple osteochondromas30. These osteochondromas are indistinguishable morphologically from the solitary cases (see Figure 1). HME has an incidence of about 1 in 50,000 live-births55. Many patients with HME require resection of an osteochondroma due to a mass effect or neurovascular impingement symptoms. Importantly, up to 3 percent of patients with HME will eventually develop a chondrosarcoma in the cartilaginous cap of the lesion21,55,69,72. Improved understanding of the pathogenesis of osteochondromas has implications not only for patients with HME, but for the more general understanding of mechanisms of neoplastic transformation and possible subsequent malignant degeneration. In this article, we will critically review the current theories of osteochondroma pathogenesis as well as how they came about; we will also propose an alternative theory, derived from the latest advances in physeal molecular signaling.
Through the twentieth century, varied etiological theories for osteochondromas have been derived from histological observation. Müller suspected osteochondromas to arise from erroneous differentiation of cells in the periosteum42. Others have hypothesized a migration of physeal chondroctyes into the metaphyseal periosteum49. Still others have supposed the perichondrial groove of Ranvier (see Figures 2 and and3)3) to be the source of osteochondromas46,58. The only animal model of osteochondroma formation involved irradiation of the groove of Ranvier in rabbits15. Nevertheless, the cells of origin and pathogenesis of osteochondromas remain unclear.
Over the last 2 decades, we have witnessed a dramatic advance in molecular biology and genetics. These have permitted a glimpse into the molecular players underlying these lesions. Genetic linkage analysis has located three etiological genes for HME-EXT1 (8q24.1)36, EXT2 (11p11-p12)31,74, and EXT3 (19p)29. Interestingly, mutations in any of these genes demonstrate very similar clinical manifestations. The human EXT1 and EXT2 genes have been cloned1,75, as have homologues in mice11,34,61, Caenorhabditis elegans11, and Drosophila melanogaster4,25. These EXT loci have defined a new class of putative tumor suppressor genes, to which have been recently added three related genes, EXTL173, EXTL275,76, and EXTL354,66, which have also been cloned24,26.
Because both sporadic osteochondromas and those associated with HME have been associated with loss of heterozygosity (or somatic loss of function of the wildtype allele) of one or more of the EXT loci, a neoplastic model of pathogenesis has been suggested48. The Knudson 'two-hit' theory of carcinogenesis27, derived from familial retinoblastoma and displayed elegantly by Vogelstein and his colleagues in the genetics of colorectal cancer67, has been applied to the osteochondroma. Both copies of the EXT1 gene have been observed to be microscopically deleted in osteochondromas of both sporadic and familial varieties7,41. The Knudson theory's application to osteochondroma pathogenesis has been strengthened by noted EXT gene losses and mutations in chondrosarcomas arising from osteochondromas7,19,20,50.
These data are consistent with a neoplastic model of pathogenesis for osteochondromas; they provide the basis for the presumed tumor suppressor function of the EXT family of genes. However, while the classification and localization of genes in the EXT family continues, the biochemistry of EXT1 and EXT2 function presents questions as to how these genes can function as tumor suppressors.
Exostosin-1 and exostosin-2, the protein products of EXT1 and EXT2 are widely expressed type II transmembrane glycoproteins of 746 and 718 amino acids, respectively. Both localize to the endoplasmic reticulum and Golgi complex38. There, they together perform the Nacetylglucosamine (GlcNAc) and D-glucuronic acid (GlcA) transferase activities of a heteroligomeric heparan sulfate polymerase56. They function as enzymes, attaching sugar moieties to the surface of proteins that will ultimately be secreted into the extracellular matrix or integrated into the cell membrane. Loss of functional EXT1 or EXT2 in a chondrocyte alters its ability to attach heparan sulfate to the proteins intended for its cell surface and its immediate extracellular milieu17,33,39,40. No other functions of EXT genes have yet been confirmed.
Possible mechanisms for tumorigenesis remain open for consideration, even if EXT genes exhibit cellular function limited to enzymatic catalysis of heparan sulfate proteoglycan (HSPG) synthesis alone. Insight was initially gained through study of the Drosophila homologue of EXT1, tout-velu. Tout-velu, like the EXT genes, is also a functional polymerase for HSPGs. It has been shown to be necessary for diffusion and long range signaling of hedgehog (Hh), a potent developmental patterning factor4. The hypothesis has thus been offered that EXT gene dysfunction results in failed long-distance signaling of human homologues of Hh. The known human homologues of Hh include Sonic hedgehog, desert hedgehog, and indian hedgehog (Ihh)47. Ihh specifically has been shown to be involved in signalling in the growing physis. Lending confidence to the extrapolation from the relationship between Hh and tout-velu-dependent HSPG synthesis, Ihh diffusion is indeed HSPG dependent 18. Further, EXT1 null mice fail to gastrulate properly, consistent with dysfunctional diffusion of Hh proteins known to be critical for spatial patterning32,33.
The explanatory theories most recently presented in the literature focus on disruption of Ihh signaling due to loss of EXT-dependent HSPG synthesis4,17. However, these theories have not expanded attention beyond the disruption of a single function of Ihh in the growth plate. The Ihh function in focus is a negative feedback loop whereby chondrocytes that have begun to hypertrophy express Ihh, which diffuses long-range to induce expression of parathyroid hormone related protein (PTHrP) in the reser ve zone and periarticular chondrocytes. PTHrP diffuses back to the yet proliferating chondrocytes to prevent their initiation of hypertrophy (See Figure 4). By this feedback loop, Ihh expression after cell cycle exit postpones the same cell cycle exit in the proliferating chondrocytes located just one layer closer to the epiphysis, allowing more rounds of proliferation.
The current theories of osteochondroma pathogenesis hypothesize that focal loss of this Ihh/PTHrP negative feedback loop for a clone of EXT null chondrocytes causes formation of an osteochondroma. However, the most recent knowledge regarding molecular signaling in the physis is incompatible with this hypothesis for two reasons.
First, osteochondromas represent an increased-although misdirected-focal linear growth of bone. Quite in contrast, a reduction in the PTHrP feedback loop results in focally decreased linear physeal growth. Without PTHrP signaling, chondrocyte differentiation, hypertrophy, and apoptosis occur after fewer rounds of proliferation, permitting the ossification front to focally advance into the physis9,10,28. The supposition that accelerated chondrocyte differentiation and early ossification could create a local excess of bone, simply fails to fit the data gathered from experiments in which PTHrP signaling to clones of cells was specifically disrupted 9,10. If any effect were achieved through disruption of this pathway, it would likely be the opposite of an osteochondroma.
Second, focal results of any kind would not be expected from interruption of the first of a two-step, longrange feedback loop. Poor diffusion of Ihh through an EXT-null clone of cells might result in poor return signaling of PTHrP, but the reduced long-range signal return would not be directed specifically toward the EXT-null cells. Instead the reduction in PTHrP signaling would more likely be distributed across the entire physis, EXT-null clone cells and neighboring wildtype cells alike. A change in the focal growth of an EXT-null clone according to this pathway would depend more upon the altered diffusion/signaling of PTHrP than that of Ihh itself. No disruption of PTHrP signaling in the absence of HSPGs has been confirmed or refuted. Focal loss of Ihh expression may mimic the focally disrupted Ihh diffusion through EXT-null chondrocytes. This has been tested in Ihh null/wildtype chimeric mice. Physis-wide stunted growth was observed in these mice10,28; they did not form osteochondromas. It is possible that such a mechanism contributes to the phenotypic shortening of long bones in individuals with HME, but that it would induce osteochondromas seems unlikely.
Despite these problems with the recent thinking about osteochondroma etiology, the attention paid to disrupted Ihh function may not be misdirected. The theories simply need to be updated with the most recent information from the rapidly advancing knowledge of physeal physiology and signaling. In addition to the Ihh control of PTHrP secretion by periarticular cells68, Ihh has direct signaling to the proliferating chondrocytes35, where it is a powerful mitogen10,23,62, and to the perichondrial and primary spongiosal mesenchymal stem cells10,62, which it induces toward osteoblastic differentiation43,59 (see Figure 4).
The effect of lost EXT-dependent HSPG synthesis on Ihh mitogenic signaling to proliferating chondrocytes depends on whether diffusion alone is disrupted or ligand-receptor interaction is as well. If only Ihh diffusion is blunted around EXT deficient cells, then it could be reasoned that proliferating chondrocytes without functional EXT "see" a higher concentration of Ihh. As it cannot pass by, Ihh might build up near EXT-null cells, more powerfully signaling them to proliferate. This could contribute to focal overproliferation of cells and the formation of an osteochondroma. However, if signaling is less efficient or impossible without HSPGs, as has been suggested in EXT knockout mice33, then this disrupted Ihh function would result in reduced proliferation, detracting from, rather than contributing to osteochondroma pathogenesis.
The third known function of Ihh in the growth plate is its regulatory role in inducing the perichondrial and primary spongiosal mesenchymal stem cells to differentiate into osteoblasts62. Expression of Cbfa1, which is required for osteoblast differentiation, does not occur in the absence of Ihh signals diffusing from the prehypertrophic chondrocytes to these two populations of cells22. Mice that do not express functional Ihh form no endochondral bone62. Specifically, focal loss of Ihh expression at the periphery of the growth plate results in a focal defect in the bone collar, or the advancing lip of the cortical bone, that normally forms to surround the metaphyseal aspect of the physeal chondrocytes10.
It can be reasoned that a peripherally located, EXT-null clone of proliferating chondrocytes, unable to synthesize HSPGs, will prevent the diffusion of Ihh to the perichondrial cells in the abutting region of the groove of Ranvier (see Figure 5B). A focal defect will therefore develop in the bone collar peripheral to the EXT-null chondrocytic clone as those perichondrial cells fail to differentiate into osteoblasts in the absence of Ihh signaling. Perhaps, the loss of this rigid structural constraint of the bone collar is tantamount to losing some of the restraint to proliferate only longitudinally. When the bone collar is prevented from forming by reflection of the perichondrium overlying the growth plate, the chondrocytes beneath have been shown to proliferate in the peripheral direction14. As the EXT-null clone proliferates and "overflows" the defect in the bone collar, it maintains an outer lining of undifferentiated perichondrial cells. As the next generation of proliferating chondrocytes progresses toward hypertrophy, the bone collar is again formed around the next clone of normal chondrocytes, upstream to the outpouching osteochondroma (see Figure 5C).
The perichondrial cells lining the protruding EXT-null chondrocyte clone still express functional EXT. When Ihh signal diffuses to their location, their receptors will undoubtedly receive the signal, differentiate into osteoblasts, and appose a lip of cortical bone. However, because the outpouching cartilaginous cap of the forming osteochondroma does not permit Ihh diffusion, the only surrounding perichondrial cells that receive this Ihh signal are those immediately adjacent to the chondrocytes beginning hypertrophy, which express Ihh directly. In this way, the cartilaginous cap of an osteochondroma-although often compared to a normal physis directed peripherally-will never match the morphology of the normal physis. It will never form a surrounding, constraining bone collar. The surrounding lip of cortical bone that forms around the stalk of the osteochondroma will not catch up with the chondrocytic proliferation in the cap until cessation of growth at skeletal maturity (see Figure 5D). This effect may create the mushroom appearance of the typical, pedunculated osteochondroma.
While loss of Ihh function and the resultant effects on the bone collar may be critical to the formation of an osteochondroma, unless it is true that only Ihh diffusion and not signaling are disrupted by loss of HSPGs, the apparent overproliferation of the clone chondrocytes has not yet been explained by loss of EXT function. However, Ihh is not the sole regulator of proliferation in the physis. Some researchers have recently argued that the Ihh signaling pathways may indeed be downstream, at least in part, from signaling through the fibroblast growth factor receptor 3 (FGF-R3)8,53.
Many fibroblast growth factor (FGF) signaling pathways are highly mitogenic. However, FGF signaling to proliferating chondrocytes elicits an anti-proliferative, pro-differentiation response by inducing FGF-R3 signal transduction44,53. This effect is the opposite of the effects from Ihh mitogenic signaling and the PTHrP feedback loop. FGF-R3 signaling may be upstream of Ihh expression, as FGF-R3 may control the initiation of hypertrophy, which begin Ihh expression. Ihh expression by the pre-hypertrophic cells6 that have already exited the cell cycle may represent the beginning of negative feedback loop balanced against FGF-R3 initiation of hypertrophy 8,53.
FGF-R3 causes cell cycle exit by activating the transcription factor STAT1, which induces expression of p2163. After binding with cyclin E and cyclin-dependent kinase 2, p21 increases the concentration of Rb (retinoblastoma) relative to its phosphorylated counterpart, forcing cell cycle arrest at the G1-S checkpoint2. This effect is best illustrated by the autosomal dominant, gain of function mutations in FGF-R3 which generate the dwarfing chondrodysplasias: achondroplasia, hypochondroplasia, and thanatophoric dysplasia5,45,51,52,57,64,65,70,71. Alternatively, the opposite effect is seen in FGFR-3 null mice, which exhibit skeletal overgrowth13,16.
Crucial to its relation to EXT deficiency, FGF-R3 signaling has been demonstrated to be exquisitely dependent on HSPGs3,12. Therefore, a clone of EXT-null chondrocytes is equivalent to a clone of FGF-R3 null chondrocytes with regard to FGF signaling; the chondrocytes will be freed from a powerful negative control on proliferation. This lost signaling increases the proliferation and postpones the differentiation of the EXT-null clone of chondrocytes, which, associated with the bone collar defect caused by blocked Ihh diffusion to the perichondrial groove of Ranvier, may begin the outpouching that ultimately yields an osteochondroma (see Figures 5A and 5B).
The plausibility of a theory for pathogenesis of an osteochondroma depends on its adherence to all available data regarding the implicated pathways involved and its ability to explain the character of the resulting lesion. Both Ihh diffusion and FGF-R3 signal transduction are demonstrably dependent on HSPGs present at the cell surface. Loss of EXT function results in HSPG synthesis deficiency. The proliferative results of lost FGF-R3 signaling and the focal absence of the bone collar from disrupted Ihh diffusion to the perichondrium have each been shown independently, as discussed above. While other signaling pathways are undoubtedly affected by loss of HSPGs in a clone of chondrocytes, the predictable effects on these two pathways form a theory for osteochondroma pathogenesis which fits all the available evidence.
In addition, this pathogenetic theory explains some of the characteristics of an osteochondroma. First, means for the creation of the resultant structural morphology of an osteochondroma are provided by the theory. The radiographic diagnosis of an osteochondroma depends upon demonstration of cortical and medullary continuity. In this theory, the osteochondroma cap forms from chondrocytes proliferating in the peripheral direction, but from the otherwise normal proliferative zone of the physis (see Figure 5D). Therefore the fronts of chondrocyte differentiation, hypertrophy, and apoptosis, and subsequent primary spongiosal ossification are in continuity with those in the rest of the physis. In addition, the focal distruption of the bone collar, which is then restored epiphyseally once the next wildtype clone of cells moves through the proliferative zone, provides a mechanism for the diagnostically important cortical continuity.
A final feature of the osteochondroma which any pathogenetic theory must address, is that of malignant degeneration into a surface chondrosarcoma. This was once thought to happen in up to 25 percent of patients with HME37. While that figure has more recently been reduced to 0.5 to 3%21,55,69,72, the resulting surface chondrosarcomas do make up approximately one sixth of all the chondrosarcomas in humans 60. While this theory for osteochondroma pathogenesis does not entirely explain malignant degeneration, it explains a scenario which is predisposed toward accumulation of genetic mutations; this would, at a certain frequency, lead to malignant degeneration.
In a classic parallel, familial adenomatous polyposis leads to colorectal cancer because a given population of cells accumulates somatic mutations while undergoing more than the usual number of cell cycles prior to terminal differentiation67. This theory of osteochondroma pathogenesis includes an extension of chondrocyte generations from a single clone due to disrupted FGF-R3 signaling. This necessarily increases risk for accumulating important malignant somatic mutations by genomic replication error alone. However, in contrast to the case of familial adenomatous polyposis, the osteochondroma cells are not intrinsically immortalized by the EXT mutations in this theory. Instead, they are freed from a potent extracellular restraint to cell cycling. Once the extracellular restraint of FGF-R3 signaling is removed, extended cell cycle iteration is permitted and with it, the increased stochastic likelihood of other proneoplastic mutations. Thus, the putative tumor suppressor function of EXT genes is not only indirect, but only enables oncogenesis in a specific milieu of extracellular signals.
The authors would like to thank Jeff Stevens, Ph.D., for provision of the photomicrograph of the physis, and D. Lee Bennett, M.D., for provision of the knee radiograph for illustrations.