From an engineering perspective, skeletal tissues are remarkable structures because they are lightweight, stiff and tough, yet produced at ambient conditions. The biomechanical success of skeletal tissues is largely attributable to the process of biomineralization — a tightly regulated, cell-driven formation of billions of inorganic nanocrystals formed from ions found abundantly in body fluids. In this Review, we discuss nature's strategies to produce and sustain appropriate biomechanical properties in mineralizing (by the promotion of mineralization) and non-mineralizing (by the inhibition of mineralization) tissues. We review how perturbations of biomineralization are controlled over a continuum that spans from the desirable (or defective in disease) mineralization of the skeleton to pathological cardiovascular mineralization, and to mineralization of bioengineered constructs. A materials science vision of mineralization is presented with an emphasis on the micro- and nanostructure of mineralized tissues recently revealed by state-of-the-art analytical methods, and on how biomineralization-inspired designs are influencing the field of synthetic materials.
Most processes and structures in nature follow the principle of “minimum inventory/maximum diversity” (Ref. 1). The diversity of natural forms is even more astonishing if we consider their multifunctionality as a design strategy2. Biomineralization is a fundamental example of how a narrow selection of biologically available elements is used for metabolic and physiological processes, and also provides elegant engineering solutions to biomechanical challenges. Indeed, calcium and phosphate ions are among the most important ions in our body.
Calcium is used in muscle contraction, nerve signal transmission and in blood clotting3, and phosphate serves as a basis for the metabolic pathway between adenosine di- and triphosphate conversions, as well as being a potent route by which proteins are modified specifically for cell signalling4 and for binding to mineral crystals in the body5,6. The concentrations of calcium and phosphate ions are tightly regulated in blood serum, in the extracellular tissue fluid and intracellularly. Bone is the essential storage depot for these ions in the form of crystallites of calcium phosphate mineral (a carbonated form of hydroxyapatite), which together with the organic matrix fulfils the mechanical functions of the skeleton by providing support, locomotion and protection. The mechanical tasks of bone require stability and strength combined with the capacity to grow and self-repair; however, the physiological roles of ionic calcium and phosphate require high reactivity and turnover, and fine regulation. Finally, both the biomechanical and metabolic roles rely on efficient transport, fast mobilization and appropriate sequestration of calcium and phosphate ions without inducing the skeleton to dissolve and without hardening soft tissues. This phenomenon raises two important connected questions, which are not only of medical relevance but of major interest to materials chemistry: namely, the way in which the aforementioned processes are coordinated and regulated appropriately in a living organism, and how their perturbation links to pathological conditions (Tables 1,2). This Review addresses these questions from a materials science perspective. We outline the analytical techniques used to study mineralized and non-mineralized tissues in healthy and diseased settings, and review selected bioengineering approaches that have been developed to replicate the properties of mineralized tissues and to initiate their repair.
Determinants of biomineralization
Mineral homeostasis. Homeostasis maintains key physiological variables, such as ion concentrations, temperature and pH, in an optimized range in vivo. However, the body is constantly interacting with the environment; as a result, homeostasis de facto relies on a dynamic equilibrium provided by multiple control mechanisms and feedback loops. Moreover, with respect to ion homeostasis, there is a specific set point for each physiological milieu. Although the calcium concentration in serum is maintained by the endocrine system in a narrow range of 2.2–2.7 mmol l−1 (Ref. 7), the intracellular calcium concentration is four orders of magnitude lower3. In bone, calcium ions are found as part of carbonated hydroxyapatite, which belongs to the category of sparingly soluble minerals8. The solubility of ionic compounds is described by the solubility product constant, Ksp — the point of equilibrium between ionic, dissociated compounds and the undissolved solid. For carbonated hydroxyapatite at physiological conditions, the logarithm of Ksp is reported to range from −58 to −59 (Refs 9,
In the body, an astonishing variety of tissues are based on a common building block — the collagen fibril. Collagen fibrils constitute a major component of the so-called connective tissues (which include the specialized skeletal and dental connective tissues that mineralize). The fibrils are formed from triple helices stabilized by intermolecular hydrogen bonds. The fine fibrillar nature of collagen allows the lateral self-assembly and the formation of a staggered arrangement to give extensive arrays of collagen fibrils14 that are further stabilized by covalent crosslinking15,16. Enzymatic crosslinking of collagen is initiated by lysyl oxidase — a copper-dependent enzyme — resulting in precisely positioned aromatic bonds, namely one per collagen chain17. The complexity and hierarchical organization of collagen is tissue specific18,19, as is the crosslinking (more specifically, the chemical nature and specificity of crosslinking and the combination of different types of collagen)17,20.
Biomineralization has evolved to diversify the mechanical properties of connective tissues. For example, the type I collagen fibrils of the dermis are not mineralized to keep the skin compliant, whereas a similar collagen-based matrix must be mineralized in bones and teeth to render them rigid. Importantly, in bone, the extent of mineralization is maintained within a narrow range and is normally far from being uniform18. Interestingly, in some anatomical locations, such as at tendon attachment sites to bone or in the supporting tissues of a tooth, the interface between hard and soft tissues must be precisely graded for successful physiological performance. Thus, at hard/soft tissue interfaces, it is not only the extent of mineralization that has to be precisely moderated, but also its sub-micrometre spatial gradients.
Promotion and inhibition. In a simplified way, the regulation of biomineralization is a hierarchical cascade of events in which abundant calcium and phosphate ions of the extracellular milieu are initially prevented from precipitation by systemic biomineralization inhibitors; however, in skeletal tissues, the inhibitory mechanisms are annulled by the action of local biomineralization promoters (typically, enzymes). In addition, local biomineralization inhibitors are active in soft tissues that are at a higher risk of pathological mineralization (for example, cardiac valves, arteries and articular cartilage) and at hard/soft tissue interfaces (for example, tendon/ligament-bone attachments, cranial sutures and the periodontal ligament).
Pyrophosphate (PPi) and fetuin A are examples of systemic inhibitors of calcium phosphate precipitation from body fluids. PPi is a ubiquitous functional antagonist of inorganic phosphate (Pi), and an increase in the PPi/Pi ratio prevents spontaneous precipitation of mineral21. Interestingly, PPi concentrations maintained by osteoblasts in forming bone constitute an extracellular depot of compartmentalized phosphate until the PPi is actively enzymatically cleaved. Extracellular polyphosphate (polyPi; an extended analogue of PPi) also serves in the transport and compartmentalization of phosphates22. In the same way that glucose concentration is controlled by the formation and destruction of glycogen, Pi concentration is regulated by the formation and destruction of polyPi — a substrate for enzymatic cleavage in the skeleton23. PolyPi-packed granules chelate calcium ions, forming neutrally charged and amorphous complexes23; thus, it seems that calcium-binding PPi and polyPi are a bioreservoir of mineral ions and effectively preclude collagen-based soft tissues from undesired mineralization.
The systemic mineralization inhibitor fetuin A is a circulating protein that prevents the growth of nascent crystal nuclei in the blood and facilitates mineral particle recycling by macrophages. A single molecule of fetuin A can sequester up to 90–120 calcium atoms and 54–72 phosphate ions24. Fetuin A forms colloidal complexes with calcium and phosphate ions (calciprotein particles) that are about 30–150 nm in size25. In addition, fetuin A has a strong affinity for bone tissue and comprises up to 25% of the non-collagenous proteins of bone25.
Mineralization is central to skeletal function; therefore, to abolish the effect of biomineralization inhibitors in bone, osteoblasts express high levels of the PPi-degrading enzyme tissue-nonspecific alkaline phosphatase (TNAP)26. The enzymatic activity of TNAP on PPi not only removes a potent inhibitor of mineralization but has the added advantage of simultaneously generating phosphate ions, which in turn promote mineralization. In soft tissues that are not destined for mineralization, the ratio of PPi/Pi remains high to inhibit mineralization. It has been suggested that polyPi granules contain alkaline phosphatase and other proteins; the notion here being that if alkaline phosphatase is activated, hydroxyapatite nucleates within the granule and displaces the protein component to the granule surface to result in the formation of a crystalline core coated with an amorphous shell23.
The biomineralization promoter bone sialoprotein (BSP) is a calcium-binding small integrin-binding ligand N-linked glycoprotein (SIBLING), the function of which is mineral nucleation. It is present in the extracellular matrices (ECMs) of bone and teeth but not in non-mineralized tissues27. Dentin matrix acidic phosphoprotein 1 (DMP1) is also important in promoting ECM mineralization, as shown by the mouse models and patients with deletions or mutations in the gene encoding DMP1 who have autosomal recessive hypophosphatemic rickets characterized by osteomalacia. The promotion of mineralization through the formation of mineral–protein complexes probably involves the stabilization of disordered mineral precursors by negatively charged proteins that localize to collagen fibril microdomains where the potential energy is lowest28. At these sites, the mineral component of the complex electrostatically interacts with collagen to influence fibril mineralization. At low concentrations of collagen reconstituted in vitro, the presence of negatively charged non-collagenous proteins can lead to in vivo-like intrafibrillar mineralization, whereas their absence results in the precipitation of extrafibrillar mineral globules28. However, at higher tissue-like densities of collagen fibrils, in which dense monodispersed fibrils display cholesteric 3D alignment, the formation of intrafibrillar mineral is possible without the involvement of non-collagenous proteins, and the morphology of the forming mineral reflects the spatial constraints of the dense collagen matrix29. As a consequence, the importance of mineral nucleators probably depends on the tissue suprafibrillar hierarchical architecture.
Locally in the ECM, mineralization-inhibiting proteins such as osteopontin (which is abundant in bone) and matrix Gla protein (MGP; which is abundant in blood vessels and cartilage) either regulate crystal growth in the skeleton and dentition to fine-tune the overall favourable mineralization or inhibit crystal growth completely. In the ECM of bone, nascent mineralization foci are always associated with osteopontin, which is subsequently found throughout mineralized bone matrix30. Osteopontin also has an important role in bone physiology as an interfacial protein, bridging new bone to old bone (for example, at the cement lines located between osteons of different generations) and new bone to implant surfaces31. Osteopontin is present at interfaces where mineralization has to be abruptly quenched: for example, at tendon insertions to bone (entheses) or the periodontal ligament32. Osteopontin-knockout mice do not exhibit macroscopic skeletal abnormalities (redundancy for regulating mineralization appears to be provided by other SIBLINGs), but the size of the mineral crystals and their level of perfection are higher, indicating a lower crystal growth restriction33,34, as would be expected given the loss of inhibitory osteopontin.
Although mineralization is systemically inhibited at most soft-tissue locations, articular cartilage and arteries appear to be at high risk for ectopic mineralization35,36. The consequence of this is especially detrimental for their physiological function, and thus an additional line of defence against mineralization is required. Both chondrocytes and vascular smooth muscle cells express MGP, and this is incorporated into their ECM. Defects in the gene encoding MGP in humans (causing Singleton–Merten syndrome and Keutel syndrome) lead to catastrophic mineralization that causes fatal rupture of the aorta and/or premature growth plate fusion in long bones, and joint ankylosis35. For other mineralization inhibitors, such as matrix extracellular phosphoglycoprotein (MEPE) and the peptide acidic serine and aspartate-rich motif (ASARM) found in SIBLINGs, readers are referred to excellent recent reviews27,37,
Compartmentalization of calcium and phosphate ions. Reports of the presence of vesicles in the ECM and of amorphous calcium phosphate mineral in developing bone started to emerge nearly 50 years ago41,42. Although the plasma calcium level is high, the intracellular calcium level is kept low through sequestration by cytosolic organelles and proteins; however, calcium needs to be labile for its signalling role3. Protective mechanisms are therefore in place, whereby cells can urgently decrease dangerously high levels of cytosolic calcium3 by the budding off of calcium-rich vesicles into the ECM. As a result of calcium sequestration by proteins, PPi and polyPi, and following Ostwald's rule of phase transformation (that is, the first forming phase is likely to be metastable in an environment of high saturation), extracellular vesicles contain disordered calcium phosphate43. Importantly, biomineralization proceeds via an amorphous pathway, which is well recognized in the realm of invertebrates that produce their skeletons from calcium carbonate44.
More recently, high-resolution imaging of native frozen hydrated specimens of developing bone tissue by cryo-electron microscopy with elemental analysis has confirmed the localization of amorphous calcium phosphate in intracellular and extracellular vesicles45. Despite the high elemental density of the vesicles loaded with disordered mineral, their calcium/Pi ratio was lower than that found in mature bone mineral and in stoichiometric hydroxyapatite. A feasible explanation is that the phosphate-rich phase (possibly polyPi) forms first and sequesters intracellular calcium that is otherwise toxic to cells43. An in vitro study of an osteoblastic culture showed that analogous intracellular granules containing amorphous calcium phosphate are associated with mitochondria46. In an in vivo study of developing bone in a zebrafish model using fluorescence confocal microscopy and Raman imaging, a nucleotide-like compound was also identified in the amorphous calcium phosphate-rich extracellular vesicles47. This is in agreement with observations that ATP is an effective stabilizer of disordered calcium phosphate mineral41. Moreover, ATP is a substrate for TNAP26, which is highly expressed by bone cells and not only cleaves phosphate to provide ions for mineralization but also destabilizes the amorphous phase and promotes the formation of nanocrystals in the collagenous template of bone. There is a possibility that amorphous mineral transport in vesicles is not only limited to bone cells and the ECM but also takes place in the peripheral bloodstream47,48.
Biomechanical implications of inhibition and promotion. Most mineralization inhibitors, that may have a dual role as mineral nucleators under certain circumstances27, are abundant in physiologically mineralized tissues. Redundant inhibitors are incorporated into bone for a strategic reason — mineralization has to be confined to a narrow range of about 65–70 wt%49. As bone stiffness is inversely proportional to toughness and directly correlates with mineral content, uncontrolled mineralization can reduce work-to-fracture in bone50,
Heteronucleation — the simultaneous formation of numerous crystallites in a chemically impure and biologically crowded milieu — is the most likely scenario in collagenous tissue biomineralization41, in which multiple nucleation sites are present in the ECM. However, matrix vesicles, the membrane of which differs slightly from the cellular membrane in its composition, provide a potential for significantly more abundant nucleation sites or promoters in the form of proteins and lipids, especially acidic phospholipids53, in combination with their metastable ionic cargo and TNAP. As a result, the bone ECM is a polycrystalline composite, in which billions of nanocrystals have astonishingly reproducible habit, size and purity. Carbonated hydroxyapatite crystals in bone are reported to be plate-shaped, as opposed to being hexagonal or needle-shaped as they are in geogenic apatites54, and crystal size in healthy bone does not exceed about 50 nm in the longest dimension55,56. Finally, from an engineering point of view, bone hydroxyapatite is of ‘low quality’, owing mainly to its many ionic and organic inclusions, substitutions and other imperfections8.
Some studies on bone mineral structure have indicated that although the core of particles is crystalline, the periphery is highly substituted and practically amorphous13,23,57. It has been demonstrated that structuring water is trapped as a rigid layer in the disordered outer shell of mineral crystallites and facilitates their co-oriented stacking at an additional hierarchical level57. Although amorphous calcium phosphate particles are transient in nature, stable amorphous calcium phosphate hydrophilic coatings on carbonated hydroxyapatite are present even in mature bone57. Thus, the structuring role of water is as important in the inorganic phase of skeletal tissues as it is in the architecture of organic moieties: structural water imparts unique properties to bone mineral, such as the nanoscale size of the crystallites and their poor crystallinity57. From a physicochemical aspect, the nanoscale size of the crystallites and their low purity, to a limited extent, compensate for the sparingly soluble nature of carbonated hydroxyapatite. From a biological perspective, the implications of the small, impure crystals are high reactivity (high surface-to-bulk ratio; excessive ions can be sequestered or lacking ions can be mobilized); compatibility with the organic matrix (they can fit within and between crosslinked collagen fibrils without disrupting the matrix integrity); and multiple interfaces established at the nanoscale that hinder crack propagation. Another benefit of having small (rather than large) crystalline particles is that smaller stiff and brittle elements are less sensitive to defects and stress concentrators58.
The small size of hydroxyapatite crystals lodged in a continuous collagenous framework allows for the implementation of one of nature's most elegant biomechanical solutions — pre-stress59,60. The hydration shell of collagen is provided by small leucine-rich proteoglycans (SLRPs) that decorate the surface of collagen fibrils61. The precisely located, enzymatic covalent crosslinks in the collagenous scaffold confine bound water to the interfibrillar space62. As a result of the crosslinking, the collagenous matrix is effectively a continuous framework with limited extensibility that cannot swell to accommodate bound water, and the osmotic pressure provided by SLRPs builds up to the kilopascal range63,
The non-collagenous proteins intercalated with the stress-bearing collagenous framework serve as a basis for an important biomechanical phenomenon — sacrificial bonds. Sacrificial bonds bridge dissimilar constituents in a composite68; they can be severed at critical loads but then re-established at rest50,69. Sacrificial bonds provide the first tier in bearing high stress. Further stress is absorbed by the collagen chains themselves, with only the highest stress being imparted to the mineral crystallites69. Indeed, as found by simultaneous monitoring of strain in bone by a strain gauge, small-angle X-ray scattering and wide-angle X-ray diffraction, the strains experienced by the macrospecimen, its collagen framework and the crystals are not equal. Of the net strain experienced by the specimen, only 80% is effectively transduced to collagen fibrils, meaning that before they are loaded there is a limited amount of shear between adjacent fibrils. In turn, of the net strain experienced by collagen fibrils, only about 90% is transduced onto the hydroxyapatite crystallites, with the remainder being absorbed by the interfacial proteins69,70.
The microheterogeneity of bone is an important toughening mechanism, which probably also defines the difference between the properties of young and senescent bone. The majority of mature skeleton is composed of lamellar bone, which can be thought of as a plywood-like layering of collagen arrays oriented in different directions. In fact, in lamellar bone 10–15% of collagen fibrils form disordered arrays71,72. These disordered arrays are positioned at the boundaries of 2–3-μm-thick ordered lamellae with a distinct orientation of collagen fibrils. The disordered layers at the interlamellar boundaries are less mineralized, in addition to the loose packing of collagen and higher hydration73. This contributes to microscale fluctuations of Young's modulus; the numerous interfaces of varying stiffness, together with alternating collagen alignment, explain the observation that a crack in healthy bone never follows a straight line (that is, an energetically low-cost path). Deviation of the crack leads to energy dissipation and is visible as a ‘zig-zag’ pattern at the fracture surface50.
The cardiovascular system is the second most mechanically challenged system after the musculoskeletal system (Box 1). Loading of the heart and arteries is automatic and repetitive (Fig. 2), driven by the pace-making apparatus of the beating heart and moderated by autonomous nervous signals and adrenocortical humoral signals. The life-long provision of steady orthograde blood flow, together with the attenuation of the pulsatile waves in the peripheral tissues, puts a special set of mechanical requirements on the components of the cardiovascular system: namely, compliance, resilience and fatigue resistance.
With ageing, the cardiac valves and the arterial walls become more rigid, and as a result the ability to ensure one-way blood flow and to dampen pulsatile waves deteriorates. The higher rigidity of aged valves and arteries can be attributed to reduced collagen compliance and impaired elastin recoil. With gradual cardiovascular stiffening often comes mineralization — a qualitative transition whereby a compliant organic structure accrues an abnormal and debilitating inorganic constituent. There are four principal types of cardiovascular mineralization: atherosclerotic calcification, medial artery calcification, cardiac valve calcification and vascular calciphylaxis. These types of calcification can coexist, escalating the severity of a clinical condition74. Although calcification in atherosclerosis can be viewed as a compensatory response to a chronic inflammation (as also observed in tuberculosis lesions, carcinomas and parasite invasion sites), and vascular calciphylaxis is an extreme example of metabolic disorder with a mortality rate of up to 80%74, in this Review, we focus mainly on medial and valvular calcification (Table 2).
To understand the pathophysiology of cardiovascular calcification, the balance between the local anti- and pro-mineralization factors acting in a dynamic and mechanically challenged milieu must be considered75,
Cardiovascular mineralization occurs systemically if the pro- and antimineralization factors are not balanced, as in the example of the genetic loss of MGP function (Singleton–Merten80 and Keutel81 syndromes), or as in the case of an abnormal concentration of circulating phosphate in uraemia82. An elevated plasma phosphate level readily leads to the formation of ectopic calcification foci in various soft tissues (such as the skin, kidneys and tendons) but is most significant in arteries, specifically the tunica media. Medial arterial calcification can not only occur in uraemia but may also be present in type 2 diabetes, and it aggravates the prognosis for both83.
Intriguingly, we have reported that highly-crystalline spherical mineral particles can be found in both the aortic and valvular tissue that do not manifest any macroscopic calcification, and also within calcific lesions in pathological vascular tissue and cardiac valves84. In some of the macroscopic calcification sites, it was possible to identify these particles as either associated with mineralized fibres or embedded within compact calcified areas84. The origin of the spherical mineral particles remains of great interest: in particular, with regards to their high crystallinity, spherical shape and relative lack of structural association with ECM fibres. Unveiling the full mechanism of pathological cardiovascular calcification is thus an active area of research, and many questions remain on the origin of the mineral and its nucleation and growth. Importantly, the cardiovascular calcification comprises a different variety of calcium phosphate to bone (that is, whitlockite instead of hydroxyapatite85), and cardiovascular mineralization is thus quite different from bone proper. Nonetheless, a postulated driving force for cardiovascular calcification at the local level is provided by cells similar to osteoblasts, which can transdifferentiate from local pericytes and smooth muscle cells, or can differentiate from circulating pluripotent cells74. Vascular smooth muscle cells display signs of transdifferentiation towards chondroprogenitors and/or osteoprogenitors, and their transcription profile ultimately overlaps with that of osteoblasts (although it does not fully match). However, it is less clear whether mineralization precedes this cellular change86 or follows it87. The chondrogenic and/or osteogenic shift of the transcription profile is associated with pro-inflammatory factors, such as tumour necrosis factor (TNF), and vascular calcification in general is accompanied by neoangiogenesis74. It is conceivable that the inflammatory component is more characteristic of local cardiovascular calcification primarily associated with cell transdifferentiation87, whereas systemic vascular calcification is induced independently of inflammation by metabolic imbalance and alteration of the ionic equilibrium83.
Although atherosclerosis-associated vascular calcification merits a separate review, it is intriguing from the biomechanical perspective that a higher heterogeneity of a calcifying plaque, and the presence of numerous interfaces between its organic and inorganic constituents, leads to plaque instability and the higher risk of disintegration, followed by thrombosis88. By contrast, macrocalcification of the plaque sustains plaque integrity and lowers the risk of thrombosis88. In comparison with physiological mineralization, the presence of abundant interfaces with abruptly changing moduli is essential for adequate toughness in bone50.
Cells are generally responsive to the stiffness of the ECM to such an extent that the fate of stem cells is strongly influenced by the Young's modulus of the substrate89. It seems reasonable that abnormal stiffening of the substrate might contribute to the acceleration of osteogenic transdifferentiation and mineralization86. The effect of the substrate stiffness is not the only factor governing cell fate but is probably enhanced synergistically by the differentiation potential (namely, the stemness) of the cells in question. Indeed, the injection of mesenchymal stem cells into relatively stiff infarcted heart tissue in mice led to the eventual formation of calcific lesions, possibly occurring through differentiation of the cells towards the osteogenic lineage90. By contrast, injected fibroblasts or haematopoietic progenitors in the same mouse model were not associated with cardiac tissue calcification. Nevertheless, tissue stiffness may also be increased by the accumulation of nonspecific crosslinks in the collagen framework induced by advanced glycation end products (AGEs). This non-enzymatic reaction between sugar metabolites results predominantly in protein crosslinking (but also lipid and nucleic acid crosslinking) and increases tissue rigidity (Table 3). Although primarily attributed to uncompensated diabetes, this phenomenon is in fact associated with ageing in general and largely affects tissues with a slow turnover rate. Because lipids are also a substrate for AGEs, non-enzymatic crosslinking aggravates atherosclerotic plaque instability. Thus, the nonspecific AGE crosslinking of collagen, elastin and lipids could explain the higher incidence of both atherosclerotic (intimal) and non-atherosclerotic (medial) vascular calcification in the elderly owing to their generally slower metabolism. In fact, the devastating effect of AGE-initiated non-enzymatic crosslinking can be found throughout all connective tissues, including bone, cartilage, dermis, the ocular lens capsule, blood vessels and cardiac valves, and is inevitably associated with age, and amplified by glycaemia (either diabetic or dietary) and impaired renal clearance.
Cardiovascular prostheses composed of processed cadaveric or xenogeneic tissue are commonly used to replace defective cardiac valves or arteries. These bioprostheses possess excellent hydrodynamic behaviour and low thrombogenicity. Glutaraldehyde crosslinking is commonly used for preservation, fixation and sterilization of the bioprosthetic tissue, which would otherwise evoke an inflammatory foreign body reaction, followed by resorption of the graft by the host91. However, aldehyde-treated tissues possess inferior resilience under cyclic loading92, and eventually deteriorate and rupture following pathological calcification. Interestingly, glutaraldehyde fixation considerably stiffens the tissue and reduces its stress relaxation rate. As a nonspecific tissue stiffener, the glutaraldehyde effect resembles that of AGEs (Table 3).
A promising research direction for cardiovascular grafts is the evaluation of alternative fixation methods that would allow the maximal conservation of the biomechanical properties of a tissue and defer the calcification events. For example, the use of a carbodiimide-based compound has been reported as a crosslinking agent that results in better biocompatibility and minimal calcification of a transplant in the short term93. Genipin, an alternative crosslinking agent that is applied for collagen hydrogels94, shows encouraging results in cardiovascular transplant processing in animal models95.
Engineering control of biomineralization
The strategies of biomineralization control observed in nature have inspired the field of artificially engineered materials: for example, the use of biomolecules for the inhibition or the nucleation of mineral. It is noteworthy that phosphate and calcium behave in radically different ways when binding to biomolecules. Phosphate ions can be polymerized or covalently attached to polymers, whereas calcium ions bind electrostatically to negatively charged polymers. Hence, phosphate metabolism requires the action of enzymes, such as TNAP, and calcium metabolism is controlled by pH and the salinity of the solution. Considering that calcium and phosphate are both needed for the formation of mineral, there is great flexibility in controlling the mineralization process.
To regulate the phosphate contribution to biomineralization, the use of TNAP-enzyme control has proved to be a useful lever. In the case of both healthy skeletal development and deterioration of an aldehyde-treated bioprosthetic cardiac valve, phosphate ions are released by TNAP and precipitate together with calcium on the collagenous framework; however, in one case the effect is beneficial96 and in the other case it is detrimental91,97 (Tables 1,2). Thus, by intervening and moderating TNAP activity — for example, by the addition of AlCl3 or FeCl3 salts to transplanted valves to block TNAP activity — the onset of valve calcification can be delayed97. Conversely, the delivery of active TNAP as an enzyme-replacement therapy has recently brought to clinical practice a treatment for hypophosphatasia to promote bone mineralization98. Indeed, hypophosphatasia results from loss-of-function mutations in the gene encoding TNAP such that the mineralization-inhibiting substrate (that is, PPi) for this enzyme accumulates in skeletal and dental tissues, causing osteomalacia, odontomalacia, hypercalcaemic seizures and, commonly, death in infancy from respiratory insufficiency (due to a soft hypomineralized rib cage). In hypophosphatasia, the ECM collagenous framework of bone and the inorganic ions are available, but biomineralization is hindered because of excessive local inhibitory PPi. In enzyme-replacement therapy, recombinant TNAP is targeted to residual skeletal and tooth mineral in patients via a mineral-binding deca-aspartate peptide sequence — a delivery method that contributes to the rescue of the skeletal and dental mineralization defects98,99. Precisely targeting TNAP to residual bone and tooth mineral because of the presence of the mineral-binding peptide prevents possible adverse effects, such as unwanted mineralization of soft tissues at the site of injection (the skin) or more distally after its uptake in the circulation99. However, the relative contributions to the success of this patient treatment of the mineral-targeted enzyme versus the circulating enzyme remain to be determined.
Immobilizing TNAP can also induce mineralization on the surface of biomaterials meant to replace defective bone. For example, the coupling of TNAP to a chitosan matrix favours cell adhesion and results in increased scaffold toughness attributable to localized mineralization, as validated by X-ray diffraction analysis100.
Owing to the potential electrostatic interactions of calcium, either a pro- or antimineralization milieu can be artificially created for use in biomaterials development by the judicious selection of appropriately charged substrates (for example, polymers, peptides and recombinant proteins). Furthermore, collagenous matrices can be pre-mineralized by applying concentrated solutions of calcium and phosphate ions (for example, using simulated body fluid), and then used as osteoinductive scaffolds that promote osteoblast binding101,102. However, achieving the fine intrafibrillar mineralization found in bone, as previously discussed, requires important regulatory determinants in addition to the calcium phosphate product and the available collagenous framework. Intrafibrillar collagen mineralization has been achieved in vitro by adding polyaspartate peptide to a neutral solution of calcium and phosphate in which a highly hydrated amorphous-phase mineral is stabilized as a polymer-induced liquid precursor103. The stabilized amorphous mineral then infiltrates collagen fibrils and crystallizes, thus forming a hierarchical structure. This process, when applied to either demineralized bone or reconstituted collagen fibrils, produces aligned intrafibrillar nanocrystalline hydroxyapatite104.
Even in the absence of molecular or ionic biomimetic cues, biomaterial-induced mineralization can be tailored through changes in biomaterial properties such as topography and stiffness. The stiffness of the substrate, as discussed above, has been shown to influence whether mesenchymal stem cells take a neurogenic or osteogenic differentiation route, with the latter being favoured in the presence of a higher Young's modulus of 25–40 kPa versus 1 kPa (Ref. 89). Furthermore, even small changes in surface roughness105,106, or the order/disorder of nanoscale pits on a substrate, can strongly influence mineralization107. A strong osteogenic commitment was reported for a culture grown on a substrate with a surface roughness of about 2–3 μm and the distance between micro-asperities of 50–70 μm (measured by alkaline phosphatase activity and type I collagen gene expression)108. However, in other cell evaluation experiments, the effect of surface roughness levels off as a culture develops and matures109, perhaps owing to deposition of the ECM by the cell culture and masking of the original surface roughness. Currently, scaffolds composed of functionalized materials can be custom built from an array of techniques to suit the size and type of defect, as well as the regenerative capacity of the patient. Ideally, these scaffolds should be a combination of the macroscopic geometry and stiffness required to support loads, and of meso- and cell-scale structures, stiffness values and materials that take into consideration pro- or antimineralization cues required for particular applications.
Conclusions and future perspectives
Nature has created a redundancy of control mechanisms that together make the biomineralization process predictable and finely regulated, both temporarily and spatially. This elegant machinery involves the simultaneous action of inhibitors and promoters, and enzyme-enabled amplification loops. However, this versatility in the complex regulatory control of biomineralization comes at a price — a failure in just one nexus in this chain of regulation may have life-threatening consequences, as seen in individual genetic disorders such as osteogenesis imperfecta and vitamin D-resistant rickets. Although these genetic conditions are rare, the intricate biomineralization control mechanisms that have evolved over hundreds of millions of years for hardening bones and teeth, and for preventing the debilitating calcification of soft tissues in the cardiovascular system, are seemingly unable to keep pace with recent changes in human lifestyle, particularly modern diet and increasing longevity110.
The first known prosthetic replacement of a body part (a right big toe) took place around four thousand years ago in ancient Egypt111. The carved wooden toe had wear patterns indicating that it was functional for years. The desire to recreate missing or damaged tissues is even more pressing today, with a shift from using inert prosthetics that simply replace the lost organ, to the use of ‘cell-instructive biomaterials’, which are capable of a far more intimate integration into the host by means of maximizing the regenerative potential of the cells. The ultimate goal of cell-instructive biomaterials could be seen in the complete regeneration of tissues and organs achieved by the appropriate tuning of the host cellular responses using ionic dopants, signalling molecules or peptides, in combination with physical cues (for example, substrate roughness, stiffness and curvature).
As the field of materials analysis advances, a greater insight into the intricate structure of natural materials can be gained. For example, high-resolution electron microscopy imaging used today in the field of structural biology is routinely combined with advanced cryo-preservation methods to keep molecules and cells in their pristine, hydrated native state, as opposed to the chemical fixation that is associated with dehydration and distortion of the sample43,112,113. Biomineralization is a multistep process involving mineral–protein complexes and metastable compounds (such as amorphous calcium phosphate), and therefore, arguably the most informative way to interrogate this process in detail might be dynamic liquid-cell or atmospheric electron microscopy, to provide analysis under more native aqueous conditions113,114. As another example, Raman spectroscopy — originally developed for materials characterization — has been successfully adapted for in situ monitoring of mineralization in living cells and tissues in vivo and in vitro115. In this way, advances in state-of-the-art technology can unravel the mysteries of nature and provide tools for biomimetic design. As the resolution of modern imaging methods constantly improves, it is important to keep findings contextualized and to not lose track of the bigger picture. This is especially important for the imaging of hierarchically organized biological materials and in accounting for normal biological variation, in which correlative imaging is gaining momentum. By combining two, three or even more imaging methods that allow co-localization of a feature of interest, the resolution and the volume of imaging can span seven to eight orders of magnitude115,
The principles governing the mineralization of biological tissues, or those preventing them from mineralizing, have evolved over millions of years into robust mechanisms that function under various environmental challenges. A detailed understanding of such principles is naturally inspiring new approaches in engineering. This includes self-organization processes and mimicking the multiscale hierarchical structure of mineralized tissues. However, many processes are poorly understood, and, in particular, the dynamics of mineral (ion) transport, intermediate storage and on-time delivery remain a mystery in most mineralizing tissues. Moreover, mechanosensitivity of cells integrates mechanical stimuli into the growth and adaptation of skeletal tissues. This raises several important research questions, some examples of which are summarized in Fig. 4. These questions need to be addressed if we want to push our knowledge from a simple description of structures and their material properties to a more complete understanding of how these structures are formed, repaired and adapted to changing needs. Such understanding will also improve our abilities to conceive new types of active, adaptive and self-repairing materials. Finally, there is a counterintuitive character of many of nature's technological solutions — namely, that imperfections (in their common connotation) can be biologically advantageous in several different ways130. For example, skeletal fragility is reduced by the inhomogeneity of bone, while the biological activity of mineral crystals results from their imperfection, and mechanosensation of bone cells is facilitated by local disorder in bone collagen organization. The intentional productive use of imperfection is a hallmark of materials engineering, making biomineralization particularly inspiring for discoveries in materials science.
The authors thank M. McKee (McGill University, Canada), S. Bertazzo (UCL, London), J-P St-Pierre and T. Whittaker (Imperial College London, London) for the critical reading of this manuscript and insightful comments.