Collagen and Bone Strength

Bone strength is a complex property which we understand in only a limited and incomplete fashion. Bone density and bone strength are highly but imperfectly correlated.1 The relative amounts and properties of the mineral (apatite) and the organic matrix (mainly type I collagen) in bone, as well as the organization of the bone at both the microscopic and macroscopic scales (microarchitecture and anatomy, respectively), determine its mechanical strength.1-7 That there is a strong genetic component to bone density and, more generally, to bone strength is well established,8-20 but we still know relatively little about the contributions of specific genes to bone mineral content and bone strength. As we discuss below, the evidence for the importance of various candidate genes is uneven in its quality but is compelling in the case of the type I collagen genes. Readers of this journal know that bone is a metabolically active tissue, with bone resorption and deposition proceeding as ongoing processes. Bone modeling in response to load, as described by Wolff,21 enables bones to adapt their material structures to optimize both their mechanical properties and their mass. Osteoporosis in many but not all settings, including those associated with estrogen insufficiency, steroid use, nutritional abnormalities, or excessive exercise, develops when these opposing processes are unbalanced.22 Whether the cause is osteoblastic insufficiency or osteoclastic overactivity, the result at the tissue level is always a loss of bone mass/density and deleterious changes in bone architecture. At the inorganic molecular level, the smaller mineral crystals disappear, and the larger more brittle crystals persist.23 There is some suggestion that collagen structure and properties may also be altered at the organic level.24-27 In an article by Puustjärvi et al.28 on page 321 of this Journal issue, the authors probe the mechanism of bone loss associated with heavy exercise. They report that long-term running significantly reduced bone density in beagle dog vertebrae without changing the bone's mechanical (indentation) properties. The authors suggest that changes in the orientation of the collagen network, without changes in collagen maturation (cross-linking) and content (hydroxyproline per tissue), contributed to the maintenance of strength in the face of decreased mineral density. The purpose of this commentary is to review additional documentation for that view while indicating the limitations of the present study. First, we briefly review genotype-phenotype associations of collagen. Next, the impact of collagen mutations on the microscopic structure of the connective tissue are considered. Then, we summarize what is known relating these microarchitectural changes to biomechanical tests. Finally, we consider how Puustjärvi et al.'s28 paper fits into this background. The existence of a mechanistic link between type I collagen and bone strength has been amply demonstrated by an extensive literature on osteogenesis imperfecta (OI) and other clinical syndromes attributable to mutations of type I collagen genes (COLIA1) and COLIA2 (summarized on-line29, 30). Most of these data depend on family studies in which inheritance of putative pathogenic alleles could be followed. Such investigations have been able to relate the clinical severity of OI to the nature of the specific mutation present in each kindred. The range of syndromes is broad, extending from in utero death to extremely subtle manifestations, with most patients suffering disability due to multiple fractures and their consequences. This variability led to the hypothesis that some type I collagen mutations might predispose to fragility fractures or osteoporosis but might be otherwise clinically silent (e.g., Refs. 27 and 31–33).31-33 Conversely, while most OI patients demonstrate subnormal bone mineral density (BMD), some have diminished bone strength but normal BMD.34 Several mouse models of OI arising from mutations of the type I collagen genes Cola1 and Cola2 have been used for similar analyses. Naturally occurring mutants include the oim/oim mouse,35 which makes a type I collagen α1 chain trimer due to a nonsense mutation of Cola2, and the retroviral insertion Mov-13, which disrupts Cola1. While oim/oim mice model moderate to severe OI, oim/+ heterozygous mice are an excellent model of mild OI.36 Mice harboring transgenes encoding genetically engineered glycine substitutions of the type I collagen α1 model the perinatal lethal form of OI.37 Ultrastructural and biomechanical investigations of bones from these animal models6, 36, 38-41 have linked the brittleness of OI bones to abnormalities in the collagen structure. However, these biomechanical studies are limited in important ways, as discussed below. Our understanding of OI encompasses detailed information regarding the pathogenic mechanisms by which bone fragility arises in affected individuals or organisms.42-44 Among the characteristics of the bone in OI are an increased number of abnormally thin collagen fibrils,45 the presence of smaller than normal mineral crystals46 with altered mineral characteristics39, 47 and the presence of mineral outside the collagen fibrils.47 One of the reasons that OI collagen is thought to form thinner fibrils is that it is excessively post-translationally modified. This excessive post-translational modification, in turn, is due to the alterations in primary structure that retard and often prevent the assembly of individual collagen molecules.43 These abnormalities are amplified by accelerated matrix deposition and remodeling as the cells try to eliminate the abnormal collagen molecules. Collagen in acquired skeletal fragility states shows similar biochemical abnormalities. Osteoporotic bone has been reported to be overhydroxylated,26 overglycosylated,24 and to differ in its cross-links25 from age-matched nonosteoporotic bones. While in some cases the reports of altered post-translational modification of collagen are questionable because the tissue analyzed was obtained from fractured femoral heads, these reports still point to a link between collagen structure and bone strength. Animal models with diet-induced defects of collagen cross-linking, such as lathyritic rats and chickens or rats and chicks raised on vitamin B6-deficient diets, develop long bones with decreased torsional stiffness and torsional strength.48 Copper-deficient animals, which have an impaired ability to form the reactive aldehydes needed for cross-linking, also have bones with decreased torsional strength.49 Each of these alterations could contribute to the ability of collagen to align in oriented fibrils. Thus, both inherited alterations of collagen synthesis and nutritional factors can alter bone density and bone mechanical properties. Recently, Uitterlinden and colleagues performed a cross-sectional study in which they reported that a restriction site polymorphism within an Sp1 recognition motif in the first intron of COLIA1 was associated with both BMD and fracture risk in an elderly Dutch population sample.50 These authors reported that the homozygous presence of a G residue (S allele) at the polymorphic site favored high BMD compared with heterozygosity for G and T residues or homozygosity for T (s allele). Moreover, they found that s alleles were overrepresented among subjects with incident nonvertebral fractures, conferring an approximate 50% increase in fragility fracture risk. These findings represent tantalizing extensions from the established role of type I collagen in determining bone strength in two more speculative ways. First, Uitterlinden's investigation follows a recent trend of correlating allele identity by state and phenotype in cross-sectional populations as an alternative to correlating allele identity by descent and phenotype in families. This study design will be familiar to those who have followed the extensive literature on vitamin D receptor polymorphisms and bone density (reviewed partially by Cooper and Umbach51 and Econs and Speer52). Second, this report deals with an allelic variant of uncertain functional significance, albeit one for which a plausible biological role might be hypothesized. While each of these issues warrants further scrutiny, only the second is considered here and only as it relates to Puustjärvi and coworkers' report28 in this issue. Puustjärvi's paper is significant in that it provides novel, interesting information that may help us to close the gap between allelic variation and the mechanisms by which phenotype is established. Their observation of collagen deposition parallel to lines of mechanical stress suggests that the pattern of collagen deposition as well as collagen's primary sequence may contribute to the ultimate biomechanical performance of bone. Should subsequent experiments reveal that Sp1 activity is responsive to directional biomechanical stress, this will help us both to understand the mechanistic basis of Wolff's law and provide evidence in support of S/s allelism in determining bone strength. There have been numerous studies (e.g., 2-5) associating mechanical strength with bone mineral content (ash density/ash weight) or its radiographic correlate, BMD, but most ignore the contribution of collagen and the underlying architecture it defines. It is therefore worthwhile to consider some findings regarding the structural organization of bone at the micoscopic level. In both cortical and trabecular bone there is a nonuniform distribution of mineral content and mineral crystal properties. Originally the variation in mineral content was demonstrated by back-scattered electron images.53 More recently, Fourier transform infrared microspectroscopy (FTIRM) and infrared imaging (FTIRI) have demonstrated variations in both mineral content and mineral crystal size, with the most mature mineral having the largest, more perfect crystals.54-57 Since mineral content contributes to the material component of bone's biomechanical performance,2 it is likely that crystal properties also make a contribution. FTIRM and FTIRI also reveal differences in collagen maturity throughout normal trabeculae that are not detected in osteoporotic trabeculae. A comparable variation is noted across the normal osteon, but is absent in osteons in osteoporotic biopsies.58 Another issue to consider regarding the relationship between ultrastructure and mechanical performance is that bone is anisotropic. This is not surprising given the fibrillar organization of collagen in connective tissues. In a study of the equine radius,59 collagen fiber orientation was highly correlated with mechanical properties. Additionally, Martin et al. reported that in the equine radius, longitudinally oriented collagen was correlated with greater modulus and monotonic strength, and remodeling was associated with more transverse collagen.60 In contrast, a 1995 paper designed to test the hypothesis that the direction of collagen fibers in bone is a structural response to the type of mechanical load to which the bone is subjected, found no correspondence between the strain type and the direction of collagen fibers.61 These findings from cortical bone samples, together with Wolff's original postulate,21 are closely related to Puustjärvi and colleagues' proposal that collagen organization contributes to trabecular bone strength in the exercised dogs.28 One of the challenges in evaluating trabecular bone strength is separating the contribution to strength from the bone tissue itself from the overall strength displayed by the open, porous trabecular architecture. The most common approach for measuring mechanical properties of trabecular bone has been compression tests performed on standardized samples of trabecular bone with geometries large enough for the architecture to be considered as an average continuum of bone material and open space.62 The properties defined in such a test depend both on the material properties of the bone and the distribution of bone material throughout the structure (as reflected, for example, in the apparent density and in morphometric factors such as connectivity and trabecular length and diameter). Given that cancellous bone is anisotropic and its structure quite variable, compression tests results can be affected by a number of factors, including loading orientation, specimen geometry and size, storage conditions, temperature, and the frictional characteristics between the bone sample and the apparatus used to apply load to the sample.63-65 Mechanical tests of individual trabeculae have been reported.66-68 Such tests have the advantage of eliminating the architectural considerations in favor of measuring the true tissue properties of the bone. Working with such small samples, however, is tedious, difficult, and fraught with many of the same limiting factors (e.g., orientation and frictional effects) as in large sample tests. Because of these difficulties, many investigators have relied on indirect measures, such as X-ray tomography,69-71 nuclear magnetic resonance scanning,72 or ultrasound.72, 73 Finite-element models based on high-resolution images have also been used to predict trabecular bone mechanical properties.74 Such studies have confirmed the earlier observation that the compressive strength of trabecular bone is related to bone density62 and have shown the importance of specimen geometry.75 The indentation technique used in the paper by Puustjärvi et al.28 is yet another method for measuring mechanical properties. The technique has been used recently to study the compressive properties of cartilage (e.g., Refs. 76 and 77)76, 77 and is one of the few mechanical measures readily available for analyses of trabecular bone in situ. In such a test, strength is determined as the maximum force (the force required to fail the material) required to push a cylindrical, flat-bottomed indentor into the material divided by the area (presumably of the indentor) over which the force is applied. In indentation tests on cartilage, the indentor can be applied directly to the intact cartilage surface. The size of the indentor (1.96 mm in diameter) is quite large relative to the size of the structural components of the cartilage, so the continuum assumption is valid. In the case of trabecular bone, however, the trabecular surface must be exposed by cutting into the bone, so that while loading is done on an “intact” surface, the load is applied to trabeculum exposed by cutting through the bone. Therefore, the trabeculae being loaded are no longer connected to the rest of the structure that was removed, changing the support provided by adjoining trabeculae.63 Furthermore, the indentor covers but a few trabeculae, so the test approach is somewhere between that of compressing a single trabeculae and that of testing a sufficient area of trabecular bone to constitute an appropriate structural continuum. Other considerations in interpreting results from indentation testing include possible variations caused by placing and aligning the indentor over the cut surface. Placement is crucial because of variations in trabecular shape and orientation, as well as variations in mineral distribution and properties throughout individual trabeculae.55, 56 These limitations were appropriately recognized by Puustjärvi et al.28 who realized that differences in mechanical properties of the trabecular bone between their exercised and control animals could have occurred that were not resolved by the indentation technique. Despite such limitations, however, their attempt to measure mechanical properties is important because it represents one of only a few studies (e.g., Ref. 78)78 that have focused on the impact of matrix contributions to the compressive strength of trabecular bone. There are also a limited number of studies of the relationship between trabecular bone strength and collagen content. Recently, in a study of the age-dependent properties of human tibial trabecular bone, Ding et al.79 showed that the amount of collagen in the whole bone (collagen density) as opposed to collagen per bone weight varied with failure energy, while collagen content (collagen per dry weight) predicted ultimate strain (strain at which the bone failed). This suggests that the arrangement of the collagen, as indicated in the Puustjärvi et al. paper,28 is a critical property. This leads us finally to why the exercised dogs in the Puustjärvi study28 maintained their bone strength in face of a loss of bone density but no changes in collagen content or cross-linking. One possibility, as noted above, is that the dogs might have had bone with decreased mechanical strength, but the sensitivity of the assay did not allow its determination. A second possibility is that the bulk determination of collagen content and collagen cross-links masked the local changes in collagen density in the trabeculae. Perhaps if collagen contents had been expressed per bone, rather than per dry weight of bone, differences might have been detected. However, Wolff's law predicts and studies of loaded bones demonstrate that radiographic trabecular architecture can change in response to load. Since the collagen provides the oriented template for mineral deposition, finding of altered collagen arrangement in response to exercise is an exciting molecular insight into Wolff's law. Future studies will also have to consider the matrix proteins associated with the collagen80 and how alterations in bone strength are related to variations in these noncollagenous matrix proteins. There is but a single study demonstrating alterations in extractable bone noncollagenous proteins when normal and osteoporotic biopsies are compared.81 Because these proteins are involved in regulation of the mineralization process,80 they too may be altered when bone modeling and remodeling are accelerated. Furthermore, because many of these proteins are involved in regulation of collagen organization,80 it will not be surprising to find an association between their expression and the type of alteration in collagen organization.