Biological materials provide a unique source of inspiration for the design of analogous next-generation high-performance materials. In particular, damage-tolerant biological structures are comprised of hard and soft components hierarchically arranged into multiphase composites with characteristic dimensions ranging from nano- to macroscopic length scales.1–5 These complex multi-scale structures endow natural biocomposites with a combination of strength, stiffness, and toughness that is crucial for their physiological functions, ranging from structural support to defence to hunting.6–10
The peacock mantis shrimp (Odontodactylus scyllarus) offers such an example in its pair of hammer-like clubs that deliver repetitive high-energy strikes during close-range hunting or combat without undergoing catastrophic failure, despite their ability to break some of the strongest marine exoskeletons, i.e. highly mineralized mollusc shells (Figure 1).10,11 Detailed physico-chemical and micromechanical analyses of the club have identified the key microstructural factors bestowing it with its remarkable damage tolerance.11,12 The club is a multiphase composite that combines a stiff, hard and impact-resistant external region with a softer interior that provides additional dissipation of impact energy.13 The interior region is a 3-component structure consisting of: (i) fibrillar chitin organized into a highly expanded helicoidal pattern, (ii) a protein-rich matrix that mediates mineralization of the club; and (iii) amorphous calcium phosphate and calcium carbonate (Figure 1). In the exterior impact region, these three main components are also present but in different relative content, organization, and polymorphic forms. Chitin fibrils are organized in a sinusoidal pitch-graded chiral nematic structure, whereas the mineral phase is more abundant and consists of crystalline fluorapatite (FAP) nanorods.11 The protein matrix throughout the club is thought to bind the composite together and to regulate biomineralization, with the main protein (called “Club Mineralization Protein-1”, CMP-1) found to harbor both a chitin-binding domain as well as an acidic-rich domain that mediates apatite crystallization (Figure 1).14 The exterior and interior work synergistically to enhance the overall damage tolerance of the club. 13 While the impact region exhibits a quasi-plastic deformation associated with interfacial sliding of the nano-crystals, the interior region undergoes strain-hardening due to densification of the chitin-rich helicoidal structure.13 This intricate complex multi-layer architectural design is not unique to the dactyl club. Other structural materials also combine a stiff outer layer with a tough inner layer, most notably human teeth.4
Herein, we report a new class of a biologically-inspired multiphase nanocomposites with a graded structure that mimics key features of the mantis shrimp dactyl club (Figure 1). We successfully fabricated a biocomposite made of analogous building blocks – namely cellulose as a replacement to chitin, artificial proteins with both cellulose-binding domains (CBD) as well as the acidic domain of CMP-1 regulating biomineralization, and apatite – which is strong and damage-tolerant. Furthermore, we processed this composite into intricate shapes by manufacturing it into a dental implant crown with periodic patterns of micro-reinforcement orientation, and a bilayer architecture similar to human teeth (Figure 1). The composite consists of cellulose nanocrystals (CNCs) with long-range helicoidal organization mixed with two types of genetically engineered protein matrices. One CBD-containing protein was rationally designed to increase its interfacial strength with the chiral nematic oriented CNC network through non-covalent interactions and is spatially located in the interior region of the crown. The second protein was designed with the acidic-rich CMP-1 in order to harden, strengthen, and stiffen the surface of oriented CNCs by mediating nucleation and growth of apatite crystals.
Selection of building-block components
To mimic the club’s energy-absorbent multilayered helicoidal organization of chitin nanofibrils, we selected plant-derived and abundantly available colloidal CNC rods (Figure 2a-b) as the bulk component of the composite given their outstanding physico-chemical properties.15–17 Most importantly for our purpose, their high inter-rod affinity above a critical concentration (5-6% w/v)18,19 results in phase separation and self-assembly into a left-handed chiral nematic (cholesteric) liquid crystal, which can be preserved upon drying (Figure 2c). Their use, however, is limited by their low fracture dissipation attributed mainly to their low aspect ratios. This limitation can be circumvented by exploiting the structural compatibility of CNCs with carbohydrate binding module (CBM)-containing proteins. First, we set out to explore the use of reinforcing structural proteins to produce a tough fracture energy-dissipating matrix (Figure 2d and Figure S1).20–23 We hypothesized that such proteins would enable bridging interactions between the stiff long-range chiral nematic helicoidal CNC scaffold, thus acting as a molecular binder. Second, our earlier reports demonstrate that to use these proteins as a binding matrix for biocomposites, it is crucial to initially induce their phase separation and condensation into highly concentrated assemblies, often described as coacervates (Figure 2e-f and S2-S6).24–27 Key advantages of concentrated coacervates include increased intermolecular interactions, low viscosity and interfacial tension, all of which greatly facilitate wetting and infiltration through the CNC network to result in strong binding to the reinforcing phase and, in turn, to high adhesive properties that could not be achieved in the absence of coacervation (Figure 2h).20–23
Furthermore, we have recently found that CMP-1 in the club of the peacock mantis shrimp has a 3-block primary structure (Figure 2d)14 that may result in all three of the characteristics required to act as a matrix of a tough biocomposite. First, it has a conserved mid-block carbohydrate-binding module that is likely used as an anchor to chitin fibrils in the club. While such modules have historically been almost exclusively described as components of carbohydrate -degrading enzymes, there is growing evidence illustrating that they also play a crucial role in the biological fabrication of high-performance structural biomaterials such as the squid beak27 and mollusk shells28 by forming noncovalent bridging interfaces between carbohydrates, proteins, and inorganic minerals. Second, its N-terminal contains long stretches of aspartic (Asp) acid-rich regions that are ideally suited to bind to Ca2+ ions and regulate apatite mineralization. Indeed, in vitro biomineralization assays demonstrated that CMP-1 could induce the formation of apatite minerals. And third, the C-terminal is enriched in Glycine (Gly), Proline (Pro), and Alanine (Ala) residues and was predicted to be intrinsically disordered, a feature common to many proteins exhibiting coacervation. Accordingly, incubation of CMP-1 with Ca2+ ions was found to induce the formation of coacervate-like microdroplets, which may be a way to sequester and concentrate Ca2+ ions to accelerate the biomineralization process.
Consequently, we rationally designed a set of five proteins with a 3-block architecture combining these three characteristics, namely: (i) a carbohydrate-binding motif; (ii) a self-coacervation domain; and (iii) a biomineralization domain to trigger in situ apatite crystallization (Figure 2d and S1). Four of which are called reinforcing proteins (RF1-4) and the last one mineralizing protein (MP1). Each design included a central domain consisting of a resilient structural protein with intrinsically disordered tendencies known to self-coacervate, as well as end domains conferring either binding to CNCs (i.e. CBMs) or mineralizing properties. In RP1, we included six repeats of motif 1 of the Eumeta variegate silk fibroin (BGW6).28 The central domain of RP2 is comprised of 80 repeats of the elastin-like polypeptide (ELP)_(VPGVG)80.29 In RP3, we used 4 repeats of the tandem repeat ((A)25K(VPGVG)15GD)4,29 which was called eSilEl. In RP4, we engineered a 4-motif repeating sequence ((YKYKYKY(VPGVG)5)3GD(A)9 K(A)12 KSVVYV)4 (named eBGWSilEl(+)4) by combining parts of BGW6 and eSilE, and introducing a lysine- (Lys) and Tyrosine-rich (Tyr) stretch (YKYKYKY) to further enhance binding interactions with negatively charged CNCs through electrostatic interactions. Finally, MP1, the mineralizing protein, was designed to mimic the highly mineralized hard exterior impact region of the club. It is a modified version of RP4 whereby the C-terminus CBM was replaced with the Asp-rich cluster residues 131-293 from CMP-114 (called eCMP1Asp), and all the poly-Lys stretches of eBGWSilEl(+)4 were mutated to glutamic acid (Glu).
Nanocomposite assembly of impact resistant interior region of the crown
Previously we described how a diverse set of structural proteins can undergo ion-induced liquid-liquid phase separation (LLPS) in vitro, resulting in the spontaneous formation of two immiscible liquid phases from the dispersed protein solution by loss of solvation. 21,25,30,31 In the present study, LLPS of the reinforcing proteins was initiated by mixing a solution of 1 M potassium phosphate at pH 7.4 with 100 μM protein, resulting in a clear dilute phase on top and a dense coacervate phase on the bottom of the vial with droplet sizes in the range of 0.05 – 1 μm (Figure 2g-f and S2-6). To assemble nanocomposite films, we intermixed CNC (3 % w/v) with the RPs in their coacervated form (10 % with respect to the dry mass of CNC). Under controlled conditions, the mixture was then concentrated to achieve the critical CNC concentration (6% w/v), after which it was allowed to fully dry into films. To evaluate the effect of individual protein variants on the mechanical properties, we performed tensile measurements (Figure 2i-j and S7). Compared to pure CNC, mixtures containing RP1, RP3 and RP4 displayed considerable improvement in stiffness, strength and toughness with only a minor reduction in the strain at failure (ca. 0.5-1 %). In contrast, the mechanical response of RP2 was fundamentally different, forming a brittle material. Such dramatic differences in properties can be attributed to the lack of poly-Ala stretches in ELP’s core sequence. Earlier observations have shown that Ala-rich sequences undergo conformational conversion from random coil/α-helical to a β-sheet dominated content,20,21 which increases protein-protein interactions and results in a more cohesive matrix with more efficient load transfer. Such a molecular network creates interlocking regions that dissipate the mechanical stress between the protein chains and have been found to significantly toughen natural silk (Figure S6).28,32 This is consistent with the Circular Dichroism (CD) measurements showing that RP1, RP3 and RP4 all underwent conformational transition, while RP2 retained its conformation after coacervation (Figure S3). Among the RP/CNC biocomposites, RP4 showed the highest stiffness (33 ± 3 GPa), strength (460 ± 35 MPa ) and toughness (11 ± 0.9MJ/m3) (Figure 2i-j and S7). We attribute the superior mechanical properties of this biocomposite to its stronger protein/CNC binding interactions brought by the additional Lys-rich stretch that can form ionic interactions with CNC, which further enhances the CNC/CBMs affinity. Subsequent experiments were all carried out on RP4 since this fusion variant displayed the most promising mechanical performance.
To gain more insights into the role of molecular interactions and relationships between molecular structure and mechanical properties, we systematically studied the effect of varying the coacervated RP/CNC mixing ratio. Upon increasing the weight ratio (at a constant CNC content of 6 % w/w) we noted a clear decrease in modulus, strength and toughness (Figure 3a and S8). High-resolution SEM on the fracture surfaces of the corresponding specimens suggested qualitative differences in the orientation of CNCs in the dried films (Figure 3b). Pure CNC, as well as 10% and 20% weight ratios, showed chiral nematic ordering, while higher mixing ratios were entirely isotropic. Synchrotron wide-angle x-ray scattering (WAXS) measurements on pristine CNC 6% w/v, RP4 10%/CNC 6% w/v and RP4 20%/CNC 6% w/v showed a four-quadrant symmetry signal with meridian and equatorial arcs for the (004) and (200) reflections. We calculated the Hermans orientation parameter (HOP) for these films from the azimuthal intensity profile of the (200) equatorial arcs, and found that HOP ranged from 0.67-0.69, indicative of a high degree of orientation. In contrast, scattering signals for all the other samples containing more than 20% RP4 had azimuthal Debye–Scherrer ring patterns (Figure 3b, S9) with an HOP of 0.0, illustrating no preferential orientation. In situ WAXS/SAXS structural evolution from liquid crystalline solutions to solid materials during the self-assembly and drying process followed a similar trend (Figure S10-S12). Furthermore, to demonstrate that the cholesteric CNC network is preserved after the incorporation of optimum weight ratio of RP4 (10%), we studied the hydrated samples using CD (Figure 3c). CD spectra consistently showed a clear single positive peak at 200-250 nm, illustrating left-handed chirality independent of CNC concentration (Figure S13-S15). This indicates that RP4 indeed can effectively adsorb to the CNC surfaces and follows the CNCs’ left-handed twist. Therefore, we conclude that there exists an optimum relative content of RP4, above which excessive addition of RP4 unfavourably results in stronger oppositely-charged interfacial complexation. This can kinetically arrest CNCs from self-assembling into the chiral-nematic superstructure. Together, these data provide the foundation to formulate composites with higher CNC concentration in order to construct denser 3D bulk structures instead of 2D films without altering the protein content. Another key advantage of increasing the CNC content was that the resulting bulk material displayed a highly expanded helicoidal organization with an exceptional higher orientation parameter (Figure 3d-e and S16-S17). For RP4 10%/CNC 10% w/v, HOP was found to be 0.96, ultimately leading to higher flexural strength and modulus which can be attributed to the denser arrangement of CNC rods and higher orientation degree. The ultimate flexural strength for RP4 10%/CNC 10% w/v bulk reached a value as high as 150 MPa and a flexural modulus of about 12 GPa (Figure 3f-g).
We then assessed the fracture response of the 3D bulk composite using both linear-elastic fracture mechanics (LEFM) protocol to obtain the fracture toughness at crack initiation, as well as crack resistance curves in order to evaluate the stable crack propagation behaviour. The initiation fracture toughness KIc was ca. 2 MPa m1/2; however, the composites did not fail in a brittle fashion. Instead, they exhibited significant R-curve behavior (stable crack propagation) with an equivalent fracture toughness (KJeq) reaching values as high as 5 MPa m1/2 prior to catastrophic failure (Figure 3h-i). To obtain more insights into the fracture mechanisms, we looked at the fracture surfaces of bulk RP4 10%/CNC 10% w/v after three-point flexural tests. Cracks exhibited large in-plane periodic break patterns through crack twisting and interfacial sliding of helicoidally-arranged CNC rods, very similar to fracture surface features identified in the dactyl club and suggested to be associated with energy dissipating and toughening mechanisms (Figure 3j).33 Moreover, we noted less delamination of CNC layers in the bulk RP4 10%/CNC 10% w/v, indicating stronger adhesion between individual CNC layers due to binding with RP4, which can delay catastrophic failure (Figure 3k and S18). In contrast, the pristine CNC fracture surface showed pronounced delamination and pull-outs.
Nanocomposite assembly of highly mineralized stiff and hard exterior region of the crown
The complex step of mineral formation in most biological systems is induced synergistically by highly ordered 3D fibrillar frameworks (e.g. collagen in bone or chitin in seashells) and intrinsically disordered proteins (IDPs) (e.g. Phov in bone or Pif80 in nacre). While the 3D framework guides the spatial deposition of minerals, it is believed that IDPs control the nucleation and growth of biominerals. We also observed this trend while investigating apatite mineralization in native dactyl clubs, in particular finding that CMP-1 can mediate nucleation and growth of apatite crystals on the surface of the chiral nematic chitin 3D framework.14 CMP-1 undergoes LLPS induced by Ca2+ ions and we suggested that this preassembly could be a key intermediate step in the biofabrication of the club by increasing the local concentration of Ca2+ ions in order to accelerate the biomineralization process.14
To this end, we explored whether MP1, (Figure 2d) could induce biomimetic apatite mineralization on the surface of the helicoidally-organized RP4/CNC 3D framework and emulate the highly mineralized external impact region of the dactyl club. We incubated MP1 for 24 days with modified simulated body fluid (SBF) (pH ~7.4) 34 and probed apatite biomineralization using polarized light microscopy (PLM) (Figure 4a). We noted considerable differences in the birefringence intensity as a function of time, suggesting biomineralization was occurring. We then collected Raman spectra of bulk mineralized specimens formed for periods of 2 to 16 days (Figure 4b). The spectra showed vibration peaks in the 400 to 1200 cm-1 wavenumber range assigned to (ν1-ν4) of PO43- of HAP that increased in intensity with time, indicative of increasing crystallinity. Consequently, we recorded 1H→31P heteronuclear correlation (HETCOR) two dimensional (2D) magic-angle spin nuclear magnetic resonance (MAS-NMR), with the expectation that this would enable selective enhancement of protonated species in close proximity with phosphate ions of crystalline apatite mineral (Figure 4c and Figure S19). The 16-day sample showed a relatively sharp 31P signal at 1.57 ppm correlated with OH-1 at δ (1H) at around 0.0 ppm corresponding to ordered PO43- of the apatite crystallite as previously described.35,36 In all other samples, the corresponding 31P/1H cross-peak was broader, indicating lower crystallinity (Figure S20). Using scanning transmission electron microscopy (STEM) we then visualized specimens mineralized over 16 days (Figure 4d). STEM images showed a relatively large amount of electron-dense granule build-up with a size distribution in the range of 5 to 200 nm. Selected area electron diffraction (SAED) analysis on these granules showed characteristic diffraction patterns matching crystalline HAP as well as an amorphous halo ring, indicating the co-existence of both amorphous and crystalline HAP. Energy-dispersive x-ray spectroscopy (EDS) mapping confirmed the presence of P, Ca, and O expected for a calcium phosphate mineral compound (Figure 4e).37
Subsequently, we evaluated whether the coacervated state of MP1 affected apatite formation on the surface of the 3D framework and probed both structural and mechanical properties of the resulting composite. To this end, pre-assembled RP4/CNC films were rehydrated and infiltrated with MP1, followed by a biomineralization step for 16 days. Raman spectra collected from the non-mineralized specimens exhibited only bands associated with the organic phase (protein and CNC, Figure 4f), In contrast, after biomineralization peaks corresponding to the characteristic ν4 vibrations of crystalline HAP emerged. The other vibration peaks (ν1 and ν3) could not be observed due to overlapping signals from CNC, with the exception of the infiltrated sample with the MP1 followed by the mineralization step. High magnification SEM images illustrated the formation of a significant mineral phase on the surface of CNC rods in the RP4/CNC/MP1 specimens (Figure 4g). We noted the formation of granular-shape clusters covering the entire surfaces of aligned CNC rods organized into planar sheets, with a size range of 5 to 150 nm. These clusters completely connected in-plane CNCs within a single layer and were also found between adjacent layers. Interestingly, the formation of apatite on the surface of CNC rods did not disrupt the long-range helicoidal arrangement and resulted in a mechanical reinforcing effect. While the flexural modulus increased noticeably, the flexural strength did not change significantly and most importantly, this came at the expense of only a slight decline in specific fracture toughness (Figure 4h-i). The composite prepared with MP1 in the coacervate state (RP4/CNC/MP1) was the most prominent among all the samples in term of mineral assembly. Although there were slight shifts in the peak positions, the overall Raman spectrum was similar to the mineral phase formed in solution. High-resolution SEM images also confirmed the formation of larger crystals at the interface of CNCs and this composite exhibited the most significant enhancement in mechanical properties with a modulus of 34 GPa and a flexural strength of 164 MPa (Figure 4h-i). These properties are very similar to naturally-occurring impact-resistant structural materials, including natural nacre and cortical bone (see Figures 4h, i)
These findings imply that sequestration of MP1 on the surface of the helicoidally organized RP4/CNC 3D scaffold could induce biomimetic apatite mineralization and emulate the highly mineralized stiff and hard exterior impact region of the dactyl club. Thus, CNCs bearing negatively-charged carboxyl groups synergistically work in concert with MP1 to enhance sequestration of inorganic ions and apatite crystallization, therefore improving surface hardening.
Design and fabrication of a bioinspired dental implant crown
As a proof of concept, we applied our two-component phase separating system to produce a functionally-graded material replicating the complex shape of human teeth, with material properties suitable for use as a dental implant crown (Figure 5). We began with a CNC hydrogel (80% dry w/v) to which RP4 was added (10%). The mixture was injected into custom-made crown moulds and dried under pressure at 60 ͦ C (Figure S21) to form a canine, a first pre-molar, and a second pre-molar (details of the manufacturing steps included in the supplementary information). These resulting crowns exhibited a closely-matched shape resemblance to their natural counterparts, including the finer morphological details (Figure 5a-b and Video S1). To verify that a gradient of structural and mechanical properties from the exterior to the interior of the material was achieved, a brick-shaped sample was cut from the crown (Figure 5b) and imaged by SEM (Figure 5c) and phase mapping. Simultaneous WAXS diffraction patterns and x-ray fluorescence spectra from the exterior region showed an exceptionally high degree of crystallinity, with HAP as the dominant mineral phase as evidenced from the high intensity of the major reflections (002), (211), and (112) (Figure 5d) and the Ca Kα/Kβ signals (Figure 5e). Moving from the exterior to the interior region of the crown, a gradual decrease in the mineral phase was observed, until only the diffraction pattern and fluorescence signal corresponding to the protein/CNC mixture was detected, which remained constant throughout the rest of the interior region. This resulted in distinct and clearly-delineated microstructural regions as observed by SEM (Figure 5c). The microstructural gradient was then correlated to the mechanical properties by conducting nanoindentation mapping of the entire crown (Figure 5f). Accordingly, the outermost region was the stiffest and hardest, with the elastic modulus ranging from ca. 25 to 36 GPa and the hardness from 1.9 to 3.2 GPa, respectively. In contrast, the interior region was more compliant (elastic modulus 18 to 22 GPa) and softer (hardness 1.5 to 1.7 GPa.). Comparing the obtained values to the bilayered structure of the natural human tooth, we found a similar trend, with elastic and hardness value of 70-80 GPa and 4-4.5 GPa for the enamel (exterior layer of the natural tooth) and 15-20 GPa and 0.5-0.8 GPa for dentine (interior region).38 We finally performed in vitro cytocompatibility assessment on the surface of the crown by culturing adult human dermal fibroblasts (AHDF) on the surface of a 2D film with the same composition as the outer surface of the bioinspired crown (Figure 5g-h). The assays included cell viability and proliferation as well as cytotoxicity using Alamar Blue™ and lactate dehydrogenase. The surface material of the bioinspired crown could support the adhesion, growth and proliferation of AHDF without any apparent cytotoxicity, with no significant differences compared to the control tissue culture plate.