Introduction

Bones are highly hierarchical, mineralized tissues of which the main purpose in the human body is to provide mechanical stability, protection and, along with muscles, tendons and ligaments, to facilitate locomotion. To serve this purpose, bones have to be stiff and tough. Stiffness is the fundamental property that determines the deformability of the tissue and is attributed to the inorganic salts, mainly impure hydroxyapatite (HAp), of the bone tissue. Toughness on the other hand is a somewhat more complex concept because of the misdiscrimination between ‘strength’ and ‘resistance to fracture’ (ie, fracture toughness). Strength is the property that describes the “resistance to permanent (plastic) deformation”. As such, it is defined either at the point of maximum load of the elastic region (ie, yield strength), or at the maximum load of the plastic region (ie, ultimate strength). In brittle materials, where there is practically no or very little plastic deformation, the measured ‘strength’ coincides with the ‘strength to fracture.’ However, as the measured strength of a given sample is also a function of the pre-existing damage distribution, it cannot characterize the resistance to fracture, if the damage distribution is not known [1]. In the case of bone, the resistance to fracture—or simple fracture toughness—is also dependent upon a number of energy dissipation mechanisms, which manifest themselves on various hierarchical levels including the micro- and nanoscale [1–4]. When it comes to bone, toughness is either measured from conventional mechanical data (ie, un-notched specimens), or notched samples specifically prepared to measure fracture mechanics. In the first approach, toughness is often defined as the strain energy density accumulated until fracture [5, 6], although there are also other definitions, which pertain to repeated load cycles [7]. As mentioned such measures are dependent on the distribution of defects already present in the samples, such as microcracks or voids. Therefore, experiments on notched samples try to overcome this limitation be specifically inserting a stress concentrator, from which a crack is initiated and propagated through the sample until failure. This way, the amount of the consumed energy to failure (fracture toughness), and the amount of the consumed energy to propagate a crack (crack growth toughness) can be measured independently of the defect concentration [1].

Recently, characterization of bone fracture toughness has drawn the attention of bone research community because of its putative correlation to bone pathology, and in particular osteoporosis and aging [8]. Fracture toughness measurement techniques have been used in a number of studies to quantify bone fracture resistance in animal and human bone [9–17]. Most of these studies use standard material characterization techniques like compact tension (CT), or single-edge notched bending (SENB) specimen–based tests. Such techniques have been able to quantify changes in fracture toughness as a function of age, characterize the anisotropy of fracture toughness at different anatomical sites and reveal a number of toughening mechanisms that act at the tissue- and the microscale. Collagen orientation, and in this sense also osteonal orientation, play a key role in fracture behavior of bone, with fracture toughness being highest when cracks propagate perpendicularly to the long axis of the collagen fibers, ie, as a bone would break because of trauma, compared with that of parallel crack propagation [17, 18]. Interestingly, studies providing direct proof of a lower fracture toughness in osteoporotic compared with healthy human cortical bone seem to be missing. The closest evidence is perhaps a correlation between the indentation distance increase parameter from hand-held reference point indentation to crack growth toughness, reported by Diez-Perez and co-workers [19]. The prior actually exhibits a significant difference in patients who had experienced a fracture compared with ones that had not.

Nevertheless, it is generally thought that not only with age, but also due to disease, toughness as well as fracture tough of bone is reduced, which presumably is a contributing factor to the increased risk of fracture exhibited by aged and/or pathologic bones. However, the pertaining question is: why does this happen? Further, with fracture toughness being a benchtop rather than a bedside technique, there is almost no clinical data published on this topic. As a result of this, the exact relationship between fracture toughness or crack growth toughness and fracture risk has not been directly established. The assessment of fracture risk of an individual is currently based on clinical measurements such as dual-energy X-ray absorptiometry (DEXA) or the Fracture Risk Assessment Tool (FRAX). FRAX tool [with or without bone mineral density (BMD)]. While DEXA assesses the risk via epidemiological data on known risk factors with respect to the areal bone mineral density (BMD) [20–22], FRAX takes into account further clinical risk factors [23, 24]. BMD suffers from low specificity (high false negative rate) for a given sensitivity (rate of correct diagnosis) [25].

The question that arises in light of the inability of areal BMD to accurately predict fracture risk for the individual is: what are other contributing factors complementary to bone mass and clinical risk factors that can aid in establishing an improved diagnosis? There are a number of approaches toward this aim, one of these being the quest for the origin of bone toughness (ie, the identification of toughening mechanisms and their relation to structure and composition). Because of the complex hierarchical structure of bone, such toughening mechanisms can manifest at a variety of length scales. Alterations in bone matrix production because of aging and osteoporosis disease likely initiate at the molecular level, hence, it is here where the absence or presence of toughening mechanisms would perhaps have a profound effect on bone fracture toughness. The aim of this review is to summarize general findings on bone fracture toughness and to then address in particular toughening mechanisms that act at the nanoscale.

Fracture Toughness Measurement of Bone and Identification of Toughening Mechanisms

As mentioned earlier, the current knowledge of bone toughness is primarily based on ‘conventional’ fracture toughness techniques that include assessment of the linear-elastic fracture mechanics (stress intensity factor, K c ), nonlinear-elastic fracture mechanics fracture toughness (J-integral, J c ) and crack-growth resistance (R-curves) [1]. In the majority of these studies, eg, [9, 10, 26–28], toughness was assessed by means of linear-elastic fracture mechanics. The reported single-value fracture toughness (K c ) for human cortical bones ranged from 2.0 to 6.5 MPa√m depending on the anatomic location and the orientation of the tested specimens with respect to anatomic axes [29]. Linear-elastic fracture mechanics, however, assume that the material is nominally elastic and inelastic phenomena occur only in a small region around the crack-tip. In the case of bone, this assumption has allowed for good estimates of bone fracture toughness and valuable comparative studies, but is not entirely true. Bone’s complex hierarchical structure and composition results in a number of toughening mechanisms that act during loading, damage initiation, and propagation, the effect of which cause deviation from the linear-elastic behavior. Such mechanisms span across the various hierarchical levels with the ones acting at the microscale having been studied in more detail.

Among the various microscale toughening mechanisms that have been identified, of particular significance are: the formation of diffuse damage and microcracks [30, 31], as well as the crack deflection – and crack bridging mechanisms [18, 32]. These are inherent defense mechanisms that act either in front or behind the main crack tip dissipating energy and, accordingly, increasing the resistance of the tissue to crack growth.

In their paper, Koester et al. combined nonlinear-elastic fracture mechanics analysis with microscopy techniques such as environmental scanning electron microscopy (ESEM) and synchrotron X-ray computed tomography (SμCT) to quantify the toughness behavior of cortical bone and generate full crack-growth resistance curves (R-curve) in the transverse orientation. This allowed capturing in great detail the aforementioned microscale toughening mechanisms and show that the contribution of these mechanisms to the ultimate measured toughness depends on the orientation of the crack propagation to the axial direction of the osteons. In the transverse orientation (normal to osteonal direction), crack deflection appears to contribute most towards fracture toughness and is thought to be responsible for a 5-fold increase in toughness compared to the longitudinal orientation with the values of fracture toughness ranging between 20–25 MPa√m and 2–5 MPa√m, respectively [18].

Recently, another toughening mechanism has been reported that also acts at the osteonal level of cortical bone and allows for load- and movement transfer between the subsequent osteonal lamellae in a manner analogous to the engineered “elastomeric bearing pads” used in large engineering structures. The phenomenon occurs during the loading of the tissue in the transverse C-R orientation [transverse C-R orientation refers to samples cut with their long axis transverse to the long osteonal axis and parallel to circumferential lamellae] (cf. [29]), when high strains accumulate within lamellar interfaces (interlamellar areas) leading to selective stiffening of these areas [33]. It is interesting that both studies [18, 33] show the importance of interfaces, yet our knowledge of the exact composition of these features and whether their properties change with age and disease is rather limited. Similar to cement lines, which have been identified to be different to lamellar bone, compositional distinction between lamellae and lamellar interfaces has also been proposed [33].

With increasing age, both fracture toughness and crack-growth toughness of bone decrease [29, 31, 34] and in some individuals can reach a critical level below which patients suffer “enhanced bone fragility and a consequent increase in fracture risk” [35]. The latter, according to the National Institute of Health, signifies the definition of osteoporosis. It is, thus, becoming clear that to assess the occurrence of osteoporosis, one should be able to accurately assess the fracture risk, which in turn requires knowledge of factors beyond BMD and FRAX, ie, including some measure of bone toughness.

Changes that affect toughness because of age and disease could occur at various length scales. At the microstructural level for example, the number and the morphology of osteons affect bone toughness with toughness being lower in bones with high density of smaller osteons [36]. Similarly, increased cortical porosity with age and disease leads to a decreased fracture toughness [37]. Toughening mechanisms that act at lower structural levels have also been identified, but so far received less attention. Nevertheless these mechanisms may play an important, albeit unknown at present, role in overall fracture risk.

Toughening Mechanisms at the Nanoscale

Obviously failure must initiate and propagate at the nanoscale level. Toward this end, several studies have shown that bone failure ultimately occurs through two different mechanisms. These are the disruption of mineralized collagen fibrils and delamination of such fibrils in between their extra-fibrillar mineral compartments [38–42]. Therefore, it is reasonable to assume that there are nanoscale effects at work that will determine the required energy to initiate and propagate a crack. In fact, chemical treatment of freshly fractured surfaces as well as electron microscopy have revealed that the particles observed when investigating bone fracture surfaces at the nanoscale are indeed extrafibrillar mineral particles that cover the collagen fibrils (as opposed to the intrafibrillar mineral) [40]. Further investigations point to an unstructured matrix in between the extrafibrillar mineral particles [43, 44] that was identified to consist in large part of phosphorylated serine present in noncollagenous proteins (NCPs) but not in collagen [43] as well as of osteopontin (OPN) and bone sialoprotein (BSP). While NCPs have known properties of influencing crystal growth and shape, mediating collagen crystal attachment and cell attachment, they have also been proposed to be of mechanical relevance. In fact, Fantner et al. have shown that under the right conditions, networks of purified bone proteins provide molecular self-healing properties by forming weakly cross-linked networks, via unspecific ionic bonds similar to ones formed in hydrogels that repeatedly dissipate considerable amount of energy [45]. While these bonds are weak, a large number of them would still provide substantial strength [44]. Importantly, NCP networks, eg, of OPN, contain a smart way of repeatedly dissipating energy, through the sacrificial-bond and hidden length-mechanism [44], which might be an important mechanism for fatigue resistance of bone. Further, it was shown that freshly cleaved surfaces of bone once remarried offer similar self-healing behavior as NCP networks [44]. In addition to this effect, NCP matrices might even be covalently crosslinked in bone as they are substrates for transglutaminase 2 [46]. One would perhaps consider the covalent crosslinking to increase strength but on the other hand this could also impair the self-healing properties of NCP networks.

In addition to structural investigations, there are animal models to consider based on the depletion of the proteins osteopontin (OPN) and osteocalcin (OC). Experiments on bones from mice deficient of OPN showed a 30 % lower fracture toughness compared with ones from wild-type mice [47••]. While a thorough investigation of further structural and mechanical properties was conducted, the only other significant change found was a decrease in elastic modulus of OPN-deficient bones, but this was too weak to account for the decrease of fracture toughness. The absence of OPN appears to be influencing the mechanics at the nanoscale and ultimately manifesting a change at the tissue level. Another animal study involving germline deletion of osteocalcin (OC) and a double OPN and OC deletion, confirmed the lower fracture toughness of OPN depleted mouse bone and showed a comparable decrease on the absence of OC [48••]. The double knockouts exhibited even lower fracture toughness while microscopy investigation revealed that all the knockouts had significantly lower amounts of diffuse damage. A nanoscale toughening mechanism, the formation of so-called dilatational bands, was proposed that involved both OC and OPN and adds additional toughness to individual mineralized collagen fibrils [48••].

These studies provide in summary, evidence that the presence or absence of NCPs can have a significant impact on bone fracture mechanics even if mineral density measured via DEXA is comparable [47••, 48••]. Of course, the toughening mechanism of crack bridging is not exclusively facilitated through NCPs alone. Mineralized collagen fibrils are also important for crack bridging, however, as shown by Poundarik et al., NCPs also partly mediate this process. There are surely other factors as well such as increased crosslinking of the collagen due to glycation that will influence bone mechanics and with this the fracture toughness at the tissue level [49].

With respect to NCPs, it is important to note that both mechanisms proposed, ie, the sacrificial bonds—hidden length mechanism as well as the dilatational band formation, have a molecular self-healing nature. This also seems obvious by recent results indicating that diffuse damage itself does not invoke a remodeling response [50•], and heals through a mechanism not involving bone resorption [51•]. This implies that diffuse damage might be a sign of healthy bone matrix and its absence might suggest matrix alteration and increased brittleness.

The Relevance of Nanoscale Toughening Mechanisms Mediated by NCPs for Osteoporosis

The relevance of the nanoscale toughening mechanisms for osteoporosis is not clear. At best, we have animal studies showing lower fracture toughness in NCP-deficient mice when bone mineral density measured via DEXA (PIXImus, GE Lunar II, Faxitron Corp., Wheeling, IL) remains unchanged. To a degree this is because such studies are just a single approach among a number of approaches that can ultimately improve fracture risk diagnosis. Other strategies may include morphologic analysis of the trabecular bone compartment or the use of computer models simulating strength of whole bones. While all approaches are justified, researching bone mechanics at the nano- and microscale level, ie, attempting to link mechanics with structure, and composition, is rather complex because of the number of the involved length scales. As a result, such studies are not readily accessible to many groups in the biomechanics research community. Further, clinical cross-evaluations of bone composition in osteoporotic and healthy individuals and correlation of such findings to mechanical properties are nonexistent. It is somewhat puzzling, though, that there are significant changes in the NCP content of aged and especially osteoporotic bone as laid out by the seminal work of Grynpas et al. in 1992 [52]. Unfortunately, we are not much closer to establishing whether the mechanical properties of human bones are indeed linked to the overall presence and amount of NCPs. It may well be that properties like toughness are affected by the presence and amount of NCPs, such as OC and OPN. Whether the results obtained from the animal studies mentioned earlier are translatable to human pathology remains unclear. Yet, novel analytical approaches indicate that “older” human bone contains less OPN compared to “younger” human bone obtained from the same donor [53]. Therefore, the quest for the importance of NCPs for human bone toughness is probably one of the important open questions to be answered in the future [54]; the deficiency of specific NCPs in osteoporosis might not only deliver novel insights into the workings of this disease, but also offers potential to identify related signaling pathways, eg, mediated by TGF-β or WNT, behind the expression of relevant NCPs and means to influence the system one way or another.

Last, it should be also noted that NCPs and NCP-mediated processes are not the only possible candidates influencing bone toughness at the nanoscale level. In this context we have already mentioned the role of glycation of collagen, which leads to increased tissue brittleness [49]. Further, there have been insights elucidating the importance of closely bound water in bone [55, 56], which was found to correlate with toughness. Finally, Davies and co-workers have very recently reported novel insights on the inorganic phase in bone by [57•]. Their results suggest that citrate, present in the calcium phosphate crystals in bone, might play an important role in establishing crystal shape and size, which may in turn be important for bone mechanics.

Whatever the most important influencing factors for bone toughness and fracture toughness may be, the imperative task is to truly elucidate structure-function relationships for successful identification of toughening mechanisms or pathologic structural and compositional changes relevant for osteoporosis.

Clearly this poses a true challenge given the complex multi-scale organization of bone and the fact that at the lowest level it is a nanocomposite. Yet, it is also a great opportunity; with more research efforts dedicated towards structure-function relationships including nanoscale toughening mechanisms, doors might open for complementary diagnostic measures of bone fracture resistance as well as novel complementary pharmaceutical treatment strategies, targeting not only bone mass but also bone toughness and fracture toughness.