Simone I. S.
Hendrikse
*ab,
Rafael
Contreras-Montoya
c,
Amanda V.
Ellis
a,
Pall
Thordarson
b and
Jonathan W.
Steed
c
aDepartment of Chemical Engineering, The University of Melbourne, Melbourne, VIC 3010, Australia. E-mail: shendriksela@unimelb.edu.au
bSchool of Chemistry, University of New South Wales (UNSW), Sydney, NSW 2052, Australia
cDepartment of Chemistry, Durham University, Durham DH1 3LE, UK
First published on 30th November 2021
The building blocks of life – nucleotides, amino acids and saccharides – give rise to a large variety of components and make up the hierarchical structures found in Nature. Driven by chirality and non-covalent interactions, helical and highly organised structures are formed and the way in which they fold correlates with specific recognition and hence function. A great amount of effort is being put into mimicking these highly specialised biosystems as biomaterials for biomedical applications, ranging from drug discovery to regenerative medicine. However, as well as lacking the complexity found in Nature, their bio-activity is sometimes low and hierarchical ordering is missing or underdeveloped. Moreover, small differences in folding in natural biomolecules (e.g., caused by mutations) can have a catastrophic effect on the function they perform. In order to develop biomaterials that are more efficient in interacting with biomolecules, such as proteins, DNA and cells, we speculate that incorporating order and handedness into biomaterial design is necessary. In this review, we first focus on order and handedness found in Nature in peptides, nucleotides and saccharides, followed by selected examples of synthetic biomimetic systems based on these components that aim to capture some aspects of these ordered features. Computational simulations are very helpful in predicting atomic orientation and molecular organisation, and can provide invaluable information on how to further improve on biomaterial designs. In the last part of the review, a critical perspective is provided along with considerations that can be implemented in next-generation biomaterial designs.
Key learning points(1) Inter- and intra-molecular order in native biomolecules are essential to life on earth as molecular organisation, handedness and spatial distribution are crucial in biomolecule recognition and interaction.(2) Careful molecular design and preparation strategies are a foundation to creating ordered and helical bio-mimetic nanostructures. (3) Computational simulations provide invaluable insights into nanoscale molecular packing and can be used to screen molecular building blocks to predict desired physicochemical properties a priori, including intermolecular cohesion and secondary structure. (4) Including structural order into biomaterial design is demonstrated to improve cell interactions in both covalent and supramolecular systems and is anticipated to be even more important in achieving improved reciprocal interactions with living systems and hence better performance in biomedical applications. (5) In addition to molecular order, higher complexity and other design parameters, including mechanical properties, need to be matched according to the cell type and application in mind, paying attention to the dynamic nature and traction forces exerted by cells. |
Why is Nature so selective towards chiral molecules? The building blocks of life – amino acids, nucleotides and saccharides – are synthesised as homochiral molecules with mostly one handedness.3 The origin of this phenomenon is still under debate, and can have an evolutionary and/or physical origin.4 Nevertheless, homochirality in biomolecules is necessary to translate the chirality at the molecular level into three-dimensional functional folds, as a racemic mixture will give disordered and hence inactive structures. However, the opposite enantiomers can be present as well, although are typically correlated with diseases. Normally, peptides and proteins are synthesised with L-amino acids, however, upon aging, an inversion of configuration from L- to D-amino acids sometimes arises.5 Especially in long-living or permanent proteins, D-aspartic acid accumulates in many tissues, including dentine, bone, brain, heart and liver. This amino acid inversion can cause conformational changes, a loss of protein functioning, or trigger protein aggregation. Not surprisingly, protein isomerisation is enhanced in neurodegenerative diseases, such as Alzheimer's disease.6
From the amino acid sequence (i.e., primary structure) that encodes a particular protein, the formation of secondary structures in solution is triggered by non-covalent interactions, mainly driven by hydrophobic interactions and hydrogen bonds. Regions adopt a right-handed α-helical structure, displaying approximately 3.6 amino acids per turn, in which a single intermolecular hydrogen bond per amino acid pair connects a carbonyl oxygen with an amide hydrogen four amino acids downstream, or extended beta-sheets where two hydrogen bonds per pair run between adjacent strands’ carbonyl oxygens and amide hydrogens in a parallel or anti-parallel direction (Fig. 1).7 Although single amino acids have intrinsic propensities to favour α-helical or β-sheet structures, the sequence order, solvation and environment dictate the overall fold.8 According to a simple model neglecting hydrogen bonding, the fold is a result of maximising entropy, in which the α-helix has overlapping excluded volume from adjacent turns (with an optimal pitch to radius ratio c* = 2.5122), while an increasing helix radius locally unwinds the helix forming β-sheets.9 The next hierarchical layer is the formation of so-called tertiary and quaternary structures that include double- or triple-helices, and the formation of fibrils of laterally aligned helices, or alternatively, globular structures composed of multiple subunits, respectively. These layers of hierarchical ordering are a result of chiral amplification of the monomeric units through a three-dimensional (3D) structure fold by non-covalent interactions that is favoured by a reduction in free energy. Although partially or wrongly folded structures (i.e., kinetically trapped) can arise, typically the thermodynamically favoured structure is the one that is functional.
Fig. 1 From molecular structure to hierarchical formation of functional structures composed of peptide, nucleotide and saccharide building blocks. Collagen, amyloid and DNA tertiary structures are adapted from ref. 16 with permission from the RCSB PDB, copyright 2015. |
As well as right-handed α-helical structures in proteins, DNA also typically adopts a right-handed, double helical structure driven by complementary Watson–Crick nucleobase pairing (adenine (A) binds to thymine (T), and cytosine (C) to guanine (G)) between adjacent single stranded DNA (ssDNA) strands as well as π–π interactions between lateral nucleobases (purines and pyrimidine) in the same strand (Fig. 1). Less commonly observed is Hoogsteen base-pairing (i.e., one nucleobase is in the syn rather than anti configuration). The most common form of DNA is the B-form, having about 10 nucleotides and a pitch of 3.4 nm per turn, whereas in the event of DNA–RNA binding and RNA–RNA duplex formation, a tighter double helical structure (i.e., A-form) is formed composed of about 11 nucleotides and a pitch of 2.8 nm per turn. When DNA is copied, and upon binding of RNA polymerase, DNA flips into a Z-DNA configuration having a left-handed twist instead of the usual right-handed rotation.10 This helix inversion was accidentally discovered in 1972 when DNA was placed in an aqueous solution containing a high salt concentration,11 although the molecular structure was only resolved in 1979.12 Since this time there has been considerable debate on how this helix inversion takes place. Several models have been proposed,13 and with the help of molecular dynamics simulations this includes local stretching and transient Watson–Crick hydrogen-bond disruption followed by an asynchronised base flipping.14 Although Z-DNA has been found in actively transcribed regions of the genome, and has important roles in, amongst others, gene expression, it also has been associated with human diseases, including cancers, immunogenic conditions and neurodegenerative disorders.15 The Z-form is more disordered and stretched compared to the A- and B-form, containing about 12 nucleotides and a pitch of 4.5 nm per turn.
Although proteins are typically synthesised containing L-amino acids, saccharides are synthesised as D-analogues. Polysaccharides comprise of monosaccharide building blocks attached to one another using α- or β-glycosidic linkages and can be a linear or branched polymer. Due to a large array of hydroxyl groups (and sometimes together with negatively charged carboxylates and/or sulfates as found in proteoglycans), many non-covalent interactions between chains and surrounding water are formed, followed by the consequent formation of higher-order hierarchical structures. A well-known example is the formation of thick cellulose fibrils due to lateral alignment of the single cellulose fibres that are immobilised by a multivalent array of hydrogen bonding between the hydroxyl side groups (Fig. 1).16 Although far less investigated due to their flexible nature, many polysaccharides have been shown to form helical structures similar to their protein and nucleotide relatives. The microscopic fibrils that are formed in cellulose have an intrinsic right-handed twist,17 whereas amylose, containing the same glucose repeats, although with α(1–4) rather than β(1–4) glycosidic bonds, forms left-handed single-helical chains.18 Another example can be found in amylopectin, which is a large, highly branched molecule, that has chains intertwined and tightly packed in double-helical chains providing crystalline regions separated by amorphous regions (branching points) consequently giving a highly organised and hierarchical structure in starch granules.19 Interestingly, the macroscopically visible chirality in plant growth (i.e., over 90% of plant vines are right-handed) is dictated by alignment in cellulose in the plant cell walls guided by intracellular microtubules.20
The peculiar development of these well organised structures from the early Earth has resulted in the parallel evolution of proteins and enzymes that recognise and bind these biomolecules by their unique chiral signature. Natural enzymes, like proteinases, are intrinsically asymmetric and therefore stereoselectively bind L-amino acid containing proteins in contrast to their D-enantiomers.21 Although enzymes exist that bind both enantiomers, e.g., racemases, they operate with a chiral preference. As such, in drug discovery, it is crucial to develop enantiomerically pure compounds since their enantiomers can be non-active or even toxic. Learning from this field, in biomaterials science, the presence of a chiral preference in the cellular environment cannot be ignored as the predominant aim is to selectively interact with cellular components for a wide variety of biomedical applications, including for regenerative medicine, combatting antimicrobial resistance and tissue engineering. Therefore, upon the introduction of exogeneous biomaterials to cells, the organisation and hence spatial distribution of functional groups should be considered to allow better reciprocal interactions with living systems.
In this tutorial review we first highlight selected examples of the first steps in synthetic bio-inspired ordered structures that contain peptides, nucleotides and/or saccharides. Design considerations will be highlighted as well as external techniques to induce order. Experimental techniques can elucidate a great deal of information about morphology, order and configuration, although complementary computational simulations can resolve the nanoscopic packing in more detail. They can also be used in the rational design of biomaterials with desired secondary structure. The review concludes with a perspective and considerations that could be employed for the design of ordered biomaterials.
The RGD sequence is the most widely studied cellular adhesive peptide. It is present in fibronectin, as well as in other ECM proteins.30 The key interaction site for cell adhesion in fibronectin is a highly exposed RGD-containing loop that has a hairpin-like conformation. The glycine residue renders the key part of this loop flexible which is why scrambling this sequence (e.g., RDG or RAD) diminishes cellular adhesion. In synthetic fibronectin mimics it has been shown that displaying the RGD sequence in a loop31 or circular form,32 compared to the linear form, influences integrin binding selectivity. Moreover, the sequence Pro-His-Ser-Arg-Asn (PHSRN) present in the neighbouring (i.e., 9th) domain of full length fibronectin has been shown to play a synergistic effect, of which the distance between the two sequences is crucial, as competitive binding to integrin receptors inhibits cellular adhesion.33 This demonstrates the importance of amino acid orientation into space for specific cellular binding. This discovery represents a huge leap forward in biomaterials design, transforming inert materials into bioactive biomaterials by incorporating RGD sequences.34
Hydrogels are the most used materials to mimic the ECM as they are composed of a 3D network of polymers, chemically crosslinked and/or physically entangled, retaining typically up to 99% water in a similar way to living systems. The polymers generate a mechanically supportive matrix whereas their porous nature allows the diffusion of molecules, such as nutrients and oxygen. Hydrogels based on the non-covalent self-assembly of peptides are particularly interesting for mimicking the ECM as they are highly biocompatible and biodegradable.35 Their self-assembly is typically driven by hydrophobic effects and hydrogen bond formation (and e.g., β-sheet formation), similar as to in amyloidosis in vivo, which is the aggregation of denatured proteins, commonly seen in diseases like Alzheimer's.
In seminal work by the Gazit36 and Ulijn37 groups, 9-fluorenylmethyloxycarbonyl (Fmoc)-functionalised dipeptides (Fmoc-Phe-Phe) were developed that form fibres by π–π interactions between lateral phenylalanine residues, and the formation of antiparallel β-sheets held together by hydrophobic interactions and hydrogen bonding.38 Due to its crystalline character, fibres bundle together to form thick fibrils. The obtained pristine fibrous structures can be enriched with RGD sequences by attaching the RGD sequence to the end of the self-assembled peptide, and by subsequent co-assembly with Fmoc-Phe-Phe, to display RGD sequences on the fibre periphery.39 Hydrogels formed from the self-assembled fibrils have been shown to support anchorage dependent fibroblast growth and spreading in 3D. This work resulted in the foundation of the company BiogelX™ which commercialises hydrogels based on Fmoc derivatives of small peptide sequences found in the ECM as scaffolds for cell culture. Various bio-functionalised peptide amphiphiles with short ECM mimetic sequences have been developed by the group of Stupp.40 Amongst others, peptide amphiphiles containing the neurite-promoting laminin sequence IKVAV were developed and shown to support the differentiation of 3D encapsulated murine neural progenitor cells into neurons. Whereas in covalent polymers the RGD spacing should be tightly regulated to allow strong integrin binding, in supramolecular polymers ECM-mimicking monomers can migrate within the polymer backbone to allow optimal binding. As such, the epitope concentration can be significantly reduced to only 5 mol%. Since the short ECM mimetic peptides, that are installed at the end of supramolecular monomers, might be too close to the polymer backbone, and hence hinder epitope availability and consequently binding affinity, it has been shown that increasing the linker length with 5 glycines improved cell spreading.41 Although binding availability is improved upon increasing linker length, the gained flexibility can negatively influence binding affinity as well.
Not only has ‘bio-functionalisation’ been shown to influence cellular fate of peptide-based assemblies with ECM mimics, the helical direction of the fibres has also been shown to be important.42 Supramolecular polymers based on C2-symmetric L-phenylalanine derivatives self-assemble into left- or right-handed helices depending on the number of methylene units incorporated, also known as the ‘odd–even’ effect (Fig. 2A).43 Investigating the effect of chirality on 2D cellular adhesion and proliferation of mouse embryonic fibroblasts (NIH 3T3) and endothelial cells (Hy926) revealed that both cellular adhesion and proliferation is significantly higher on left-handed films than on right-handed surfaces. Moreover, left-handed helices obtained from L-phenylalanine derivatives displayed improved 3D cell culture and osteogenesis in mesenchymal stem cells (MSC) differentiation, whereas D-phenylalanine derivatives, that form right handed helices, favour adipogenesis in MSC differentiation, higher cytotoxicity towards cancer cells and higher protein absorption.44 Collectively, these results demonstrate (i) the high stereospecificity of cell–ECM interactions which goes beyond discerning molecular asymmetry, and (ii) the relevance of supramolecular chirality in the design of ECM mimicking biomaterials.
Due to the inherent stereoselectivity present in Nature, cells typically favour binding to L-amino acids over their D-isomers. However, exogenous administrated peptides and proteins composed of solely L-amino acids are rapidly cleared in vivo by proteases, which limits their use in a wide variety of biomedical applications. In order to increase their life-time, D-amino acids have been investigated and have already found great potential as they are resistant to degradation, have low immunogenicity and can improve therapeutic potency of peptide-based drugs by five orders of magnitude.45,46 However, care must be taken when considering amino acid isomerisation as it can completely alter side chain orientation and therefore affect secondary structure formation and hence protein binding.47,48 Even given this, the potential of using heterochiral monomers, containing both L- and D-amino acids, has still been explored.
Supramolecular self-assembly is tightly dependent on a well-balanced hydrophobic-to-hydrophilic balance along with minimised steric hindrances, and thus optimal packing, hence order, in the laterally stacked monomers. As a result, atomic spatial arrangement and backbone flexibility are key parameters. It has been shown by Marchesan et al.49 that the handedness of heterochiral peptide-based assemblies is dictated by the nanoscopic packing of the monomers and can be controlled by tuning the molecular design of monomers. Whereas homochiral tripeptides did not form hydrogels, the inclusion of a D-amino acid in the middle of two phenylalanine units was shown to form rigid hydrogels by hierarchical bundling. The presence of linear amino acid side chains was shown to result in efficient packing of the units into hierarchically formed thick fibrils, whereas branched side chains formed thinner, although denser and more interconnected networks rather than bundles. Extending this research to Phe-Phe dipeptides showed that by changing the chirality of just one stereocentre (i.e., from L-Phe–L-Phe to D-Phe–L-Phe), hierarchical bundling was suppressed and transparent, self-supportive and stable hydrogels where cells could be cultured were generated as compared to cell-toxic and unstable hydrogels formed by L-Phe–L-Phe (Fig. 2B).50 The effect of amino acid sequence on nanoscale packing was also highlighted by the group of Stupp on peptide amphiphile self-assembly.40 For example, upon substituting alanine by more hydrophobic valines (Ala6 to Val4Ala2), first helix unwinding was observed followed by a higher degree of twisting than the homoalanine sequence (i.e., shorter pitch).51 Moreover, the same group revealed that a higher lipid membrane interaction was obtained for L-peptide amphiphiles compared to their D-enantiomers, together with a contradictory reduce in cell viability.52 This unexpected reduction in cell viability might be attributed to slow membrane destabilisation and/or peptide amphiphile engulfment within cells due to cell's traction forces able to disrupt the weak cohesion between monomers, both facilitated by electrostatic interactions between negatively charged cellular membranes and positively charged end-functionalised peptide amphiphiles. Consequently, the hydrophobicity, charge, steric hindrance and internal cohesion (and hence internal dynamics) between monomers in supramolecular polymers can affect cell viability and question the biocompatibility of the synthetic system. Altogether, this demonstrates the importance in molecular design of the monomeric building blocks and the subsequent nanoscale packing in the supramolecular polymers they form upon self-assembly.
The significance of introducing internal order in biomaterials has not only been demonstrated for supramolecular polymers, it has also been shown of importance in covalent polymeric systems, in which the backbone is fixated rather than transient. In a study by Hammond and coworkers,53 cell-adhesive peptides, that have both the RGD and PHSRN sequence (i.e. PHSRN-K-RGD), were grafted to 8-arm poly(ethylene glycol) and two polypeptides with different secondary structures crosslinked in a hydrogel network. Poly(γ-propargyl-L-glutamate), that forms a rigid α-helical conformation, was shown to display superior attachment and spreading of 2D cultured induced-pluripotent stem cell-derived endothelial cells as compared to poly(γ-propargyl-D,L-glutamate), containing 50% D- and 50% L-glutamate that forms a random-coil secondary structure, and the 8-arm poly(ethylene glycol) control. Strong attachment and spreading in α-helical containing hydrogels were also observed in the absence of adhesive peptides, also confirming that an ordered secondary structure, with potentially the influence of local nanoscale stiffness, can enhance cell responses.
Taken together, including an ECM mimetic peptide in biomaterial design is a good start to enable cell adhesion, however, having control over conformation, helicity and order can push cell interactions to the next level and can improve favourable reciprocal biomolecule interactions.
Next to the famous double helix, different secondary and tertiary structures can also form in Nature, that include hairpins, triple helices, i-motifs and G-quadruplexes. Not surprisingly, aptamers, which are short oligonucleotide sequences that can fold into small 3D structures, have shown great affinity and selectivity towards a wide variety of targets, including proteins, peptides, carbohydrates and small molecules, and comprise the ability to distinguish between enantiomers in a similar way to antibodies.57 Despite their great promise, extremely small aptamers that could potentially target small molecules and toxins, are more challenging to obtain, due to a reduced ability to fold in a unique 3-dimensional fold, and hence a decrease in specificity and stability.
Synthetic biomaterials have been designed that include either only nucleobases, nucleosides (also (deoxy)ribose) or full nucleotides, either naturally occurring or modified. Due to the highly negative charge present on the nucleotide backbone, nucleotide conjugates, including block copolymers and inorganic conjugates, typically form spherical micelles as a result of electrostatic repulsion.58,59 Though, 1D fibrous structures can be obtained when the hydrophilic-to-hydrophobic balance is restored or by the help of cosolvents.60,61 In another approach to circumvent the charged backbone, nucleobases (i.e., without (deoxy)riboses and negatively charged phosphates) have been attached to polymers, including peptide backbones. Peptide nucleic acids (PNAs) have been intensely investigated and have shown great promise in a wide range of applications, due to their increased stability and ability to hybridise to natural DNA and RNA by strand invasion.62 This invasion is very effective as hybridisation of PNA–DNA is stronger than DNA–DNA and PNA–PNA hybridisation partly due to its neutral character, reducing repulsive interactions that are present during DNA–DNA duplex formation. PNAs have shown to be able to bind to hairpin, quadruplex and RNA targets as well, of which the binding is typically stabilised by overhanging nucleotides. However, PNAs in their unbound state have the tendency to collapse into globular structures, probably due to a hydrophobic collapse of the nucleobases. Therefore, backbone modifications have been included to pre-organise the polymer by using α- or γ-modifications with respect to the amine conjugated nucleobase unit (Fig. 3A). A right-handed helical twist is obtained upon γ-modifications with L-amino acids, whereas left-handed helices can be obtained using D-amino acids. Upon increasing the number of L-serine γ-modifications resulted in an increase in DNA and RNA affinity, observed as an increase in melting temperature (Fig. 3A).63
Other peptide–nucleotide hybrid materials have been developed that have shown remarkable properties. Shape shifting nanofibers were obtained from DNA block copolymers by Park and coworkers60 when either the polarity of the co-solvent was changed, or upon protease treatment. Nanofibres were initially obtained due to π–π interactions between phenylalanine side chains and disrupted hydrogen bonding, whereas strengthening these hydrogen bonds upon increasing solvent polarity or backbone cleavage resulted in a transition to nanosheets due to lateral hydrogen bonding alignment.
Short ssDNA and RNA have been used as templates to fabricate chiral supramolecular polymers of small organic molecules, including naphthalene, anthracene and porphyrins amongst others, into both left- and right-handed nanofibers driven by non-covalent interactions.64 Control, and the ability to switch between helicities, can be achieved by altering ionic strength, pH, and cooling rate. Likewise, the binding of adenosine–triphosphate (ATP) facilitated by cationic ammonium or guanidinium, and metal coordinated organic groups, such as dipicolyl ethylene diamine, have also been shown to induce supramolecular chirality. Remarkably, the helicity of such a system could be tuned by ATP dephosphorylation, yielding right-handed helices upon ATP binding, and left-handed helices upon adenosine diphosphate (ADP) binding (Fig. 3B). Interestingly, such materials can be augmented with out-of-equilibrium and autonomous behaviour, and hence life-like properties similar to actin filaments present in Nature, upon the addition of enzymes.65 ATP binding initiates supramolecular self-assembly, whereas enzymes (e.g., ATPases) trigger depolymerisation due to ATP hydrolysis. By recovering ATP using e.g., creatine phosphokinase (CPK), the system reassembles resulting in a living and transient supramolecular system.
Compared to the non-covalent interactions that drive peptide and amphiphile assembly – mainly hydrophobic effects and hydrogen bonding – the intrinsically weak intermolecular carbohydrate–carbohydrate interactions are typically overlooked, despite that these are of utmost importance in key cellular processes, including in cell and pathogen interactions.67 These weak interactions allow carbohydrates to screen (cell) surfaces, and upon recognition, cooperativity is initiated to tighten the binding by the multivalent effect. Multiple saccharides line up and bind to their host to reach a synergistic binding that is significantly stronger than the sum of the individual components, collectively giving the multivalent carbohydrate its biological function. Next to inter- (and also intra-) molecular interactions, water plays an important aspect in carbohydrate ordering as well, yielding remarkable gelling properties that can effectively resist tensile forces (e.g., as found in cartilage). Although investigating chiral amplification and hierarchical order in synthetic saccharide-based systems in aqueous solutions is still in its infancy, seminal contributions have been reached in this field.
As highlighted above, a well-balanced hydrophobic-to-hydrophilic ratio is key in driving the supramolecular assembly of biomaterials. For carbohydrate-functionalised materials, in particular, this is quite challenging due to the increased disorder in the system as a consequence of increased free rotation of glycosidic bonds, the different conformations that carbohydrates can adapt, their non-planar character, and the ability for only weak non-covalent donor and acceptor interactions. Similar to bulk microphase separation,68 which is dependent on the volume fraction of the blocks, linker length (i.e., degree of polymerisation) and degree of block incompatibility (i.e., Flory–Huggins parameter), micelles are formed when the carbohydrate headgroup is too hydrophilic and bulky, whereas vesicles can be formed when the linker length increases, and fibrous structures when all three parameters are in balance, as was illustrated by pioneering work by Lee and coworkers.69
Carbohydrate-functionalised birefringent hydrogels with micron scale fibre alignment were developed by Hudalla et al.70 The incorporation of the monosaccharide N-acetylglucosamine at the end of a self-assembling peptide building block drives the anisotropic fibre alignment by carbohydrate–carbohydrate interactions and solvent depletion in a molecular crowded environment (Fig. 4A). Carefully balanced carbohydrate interactions have been shown to be crucial as diglycosylated disaccharides and unglycosylated hydroxyl analogues were unable to form aligned nanofibers. The densely packed aligned glycosylated peptide nanofibers were shown to be resistant against non-specific proteins, mammalian cells and bacteria binding, yet endowing specific interactions to lectins. This highlights the importance of molecular packing and functional group orientation in supramolecular assembly in facilitating biomolecule interactions. Chiral amplification of monosaccharides was also obtained upon the incorporation of glucose and mannose to achiral supramolecular amphiphiles, a result of a combination of hydrophobic effects of the core and intermolecular carbohydrate–carbohydrate interactions.71 Fibrous structures with opposite helicity were obtained due to α- versus β-glycosylation (Fig. 4B). More hydrophilic and bulky disaccharide-functionalised amphiphiles were unable to form ordered fibres, however, upon the co-assembly with glucose-functionalised monomers, chiral amplification and hence carbohydrate–carbohydrate interactions were restored.
In molecular dynamics (MD) simulations, the polymeric system is built in a confined simulation box using empirical force fields, and the molecular motion of the included molecules is followed over time in either a bottom-up or top-down approach.72 In the bottom-up approach, the formation of a self-assembled polymer and hence the formation of non-covalent bonds starting from the monomeric building blocks is probed, whereas in a top-down approach, an already assembled polymer, with an idealised pre-determined structure, is simulated over time to probe different structural configurations. Although multiple packing motifs underpinning these alternative configurations might arise, these simulations can provide the potential energy or order of the aggregate, which determine the overall stability of the aggregate. By simulating the formation of the non-covalent interactions over time, key drivers of the growth process can be determined (e.g., hydrophobic or π–π interactions), and self-assembled pathways formed, including kinetic intermediates. Importantly, a fundamental understanding of the secondary structure can be obtained.
Information concerning the nanoscale molecular packing is usually obtained using classic all-atom MD simulations. This technique is, however, limited by the use of 106 particles and 100 ns simulation time, which does not adequately simulate the formation of bigger and more complex aggregates, and the formation of long-range order. To overcome this limitation, coarse-grain simulations, that simplifies the molecules into adhesive beads (i.e., 4 heavy atoms including their hydrogen atoms are represented as 1 bead), and enhanced sampling techniques, to sample the kinetic pathways involved, including intermediates, in the formation of the final assembled polymer, can be employed. Also, these techniques have their disadvantages, although collectively provide invaluable information into self-assembly and secondary structure formation. By comparing experimental findings with the performed computational simulations a complete fundamental understanding spanning over several length scales can be obtained. In a study by Marrink et al., MD simulations and semiempirical quantum chemical calculations to predict UV-vis, circular dichroism and infrared absorption spectra were performed, next to experimentally acquired electron microscopy and spectroscopy, to elucidate the self-replication mechanism of peptide macrocycles.73 This complementary investigation showed good agreement between calculated and experimentally acquired spectra, and revealed a cooperative self-assembly process driven by alignment between aromatic headgroups and β-sheet forming peptides giving favourable hydrophobic interactions and dipole–dipole interactions, with the inner core being collapsed to expel unfavourable water interactions (Fig. 5A).
Fig. 5 Using computational simulations to (A) elucidate the nanoscale packing in molecular self-assembly of peptide macrocycles and (B) the in silico prediction of the gelation of a virtual library of mono- and di-peptides utilising a neural network. (A) Modified and reproduced with permission from ref. 73. Copyright 2017 American Chemical Society. |
Computational simulations not only complement experimental findings to elucidate the full molecular picture of self-assembled polymers, they can also be used to predict supramolecular self-assembly propensity, gel formation and secondary structure formation, and hence allows the in silico selection and development of bio-inspired polymers with desired morphology and order. Several machine learning methods to generate quantitative structure–property relations were explored by Berry and coworkers to predict the gelation of an extensive library of mono- and di-peptides (Fig. 5B).74 They revealed that among the machine learning methods utilised, neural networks performed the best in successfully predicting gelation propensity, followed by support vector machines and random forests, which were experimentally confirmed with test compounds. The models also revealed that the five physicochemical properties; (i) number of rings, (ii) predicted molecular aqueous solubility, (iii) polar surface area, (iv) AlogP (i.e., octanol–water partition coefficient) and (v) number of rotatable bonds, dictated the prediction of gel formation. Predicting the secondary structure a priori is more challenging as initial configuration is usually derived from experimentally derived databases, and simulations usually only display small rearrangements. Initial steps in this space have been made, including in revealing unexpecting α-helical intermediates in β-sheet forming peptides, and predicting melting curve trends of short DNA–polymer conjugates and collagen-like peptides, however, needs to be further explored for more complex synthetic biopolymers and non-natural (e.g., amino acid and nucleobase analogues) building blocks.72,75
Given that cells induce traction forces to remodel the surrounding matrix raises a challenge for supramolecular hydrogels build from solely small molecular building blocks; since these building blocks are held together by weak non-covalent interactions, these monomers can be released from their scaffold and internalised (e.g., by endocytosis) or can even actively penetrate in the cell membrane (typically when the amphiphile contains a positive net charge as highlighted before). As such it is important to investigate and control the nanoscale packing, and hence dynamic property, of these monomeric building blocks within their fibrous support to match the cell's remodelling and traction forces. The influence of the internal dynamics of synthetic supramolecular biomaterials was investigated by Dankers et al.80 while employing ureidopyrimidinone functionalised supramolecular hydrogels. In excessively dynamic hydrogels, presentation of RGD ligands to cells was shown to be ineffective, whereas cell spreading and migration was significantly enhanced upon tuning the dynamic profile and RGD availability of the supramolecular system. As such, cell compatibility and the ability to control cell fate is dependent on several parameters that need to be carefully optimised to match the desired cell type specific response for a given application.
Realising such great control in self-assembly, and hence the helicity in synthetic materials, is already achieved by (i) molecular design in which the stereochemistry and hydrophilic-to-hydrophobic balance is crucial; (ii) including steric constraints to favour helicity, such as lactam bridges and employing β-peptides; or (iii) utilising external forces and techniques to induce molecular alignment, e.g., 3D printing and layer-by-layer assembly. Next steps towards improved, and hence next-generation, biomaterial designs should consider (i) incorporating several bio-active components, e.g., different ECM-derived peptides that can work synergistically; (ii) increasing complexity using multicomponent single- and multiple polymeric systems; (iii) inducing hierarchical order to achieve reciprocal interactions that span across several length scales (i.e., from nanoscopic interactions at the atom scale to synergistic whole organ level at the micron scale, Fig. 6). In addition to these considerations that aim to build more complex and sophisticated biomaterials with internal order, the amphiphilic nature of the supramolecular polymers, including molecular packing and internal positive charges, on one hand, and the traction forces exerted by cells due to mechanotransduction on the other hand, also need to be regarded as these can induce cell apoptosis, due to e.g. internalisation, membrane distortion or loss of support.
Fig. 6 Towards hierarchical materials that can match cell interactions across several length scales. Part of this figure is modified and reproduced from ref. 17. Copyright 2019 Springer Nature. |
Since there is no single material that suits all cell types and hence multiple (biomedical) applications, every type of biomaterial typically needs to be optimised according to the desired application by undergoing several iterations to match the desired biological responses. This screening of several parameters is unfortunately still stuck in a laborious trial-and-error process. However, rapid progress in computational simulations and artificial intelligence, potentially in conjunction with implementing statistical DoE approaches, bears great potential to reduce this laborious hurdle in a cost-effective manner and in a multi-disciplinary setting. As such, combining knowledge across several disciplines, from drug discovery to computer science, is key to create more sophisticated biomaterials, and realising improved biomedical applications for the near future.
This journal is © The Royal Society of Chemistry 2022 |