Microscopic observations

The first insight into elastin assembly came from histological observations. As elastin possesses few charged residues, its affinity for common dyes was low and specific dyes were developed to detect it in tissues (Dempsey and Lansing, Reference Dempsey and Lansing1954). Unfortunately, their use did not permit to assess the fine structures of the imaged elastic fibres. However, they underlined the regular and cylindrical shape of elastic fibres, and the possibility of fibrillar elements, at least in ligamentum nuchae. Further, histological analysis clearly demonstrated that the 3D organisation of the elastic network was tissue-dependent (Partridge, Reference Partridge1963).

The application of electron microscopy to the study of elastic fibres was not possible until thin section preparation was possible. In 1986, the group of Dino Volpin in Padova (Italy) demonstrated that, above 40 °C, tropoelastin assembled into thin 5 nm-thick filaments that could, in turn, merge to form an elastin fibre adopting a rope-like appearance (Bressan et al., Reference Bressan, Pasquali-Ronchetti, Fornieri, Mattioli, Castellani and Volpin1986). Elastin was conventionally described as unstructured and amorphous based on transmission electronic microscopy observations where the elastic fibres appeared devoid of any internal order/organisation. This discrepancy is explained by the fact that most reports used conventional sample preparation techniques. However, a possible internal organisation of elastin had been proposed earlier using quick freeze deep etch methods to observe elastic fibres ultrastructure (Pasquali Ronchetti et al., Reference Pasquali Ronchetti, Fornieri, Baccarani-Contri and Volpin1979). The fact that elastin sample preparation greatly influences its appearance is now accepted (Kozel and Mecham, Reference Kozel and Mecham2019), as is the rope-like organisation of elastin within the fibre.

Historical models

The random chain network (Hoeve and Flory, Reference Hoeve and Flory1958) was the first model proposed for elastin. It describes elastin as an ideal rubber and explains its mechanical behaviour with reference to the classical theory of rubber elasticity. In this model, elastin is described as a collection of long, unordered, cross-linked hydrophobic chains swollen in water.

In 1970, an alternative model was proposed to explain elastin structure and elasticity (Weis-Fogh and Andersen, Reference Weis-Fogh and Andersen1970). In the liquid drop model, tropoelastin molecules are assembled in a regular 3D arrangement creating intramolecular spaces filled by bulk water. According to this model, upon extension, the individual tropoelastin units change shape and expose their hydrophobic side chains to bulk water resulting in the formation of low-entropy water of hydrophobic hydration. This model therefore proposes that hydrophobic hydration would be the driving mechanism for entropic elasticity and hence that water is a critical actor in mediating elastin elasticity. This model was supported by the experimental finding that elastin contains ‘water spaces’ (Partridge, Reference Partridge1967). However, Hoeve and Flory (Reference Hoeve and Flory1974) contested this view by stating that ‘polymer backbone, not solvent, must bear the force’.

The oiled coil model proposed in 1973 (Gray et al., Reference Gray, Sandberg and Foster1973) was devised following extensive analysis of partial sequences derived from elastin. This model describes elastin as a network where the cross-linking domains are rigid α-helices while the hydrophobic domains are organised in a spring-like structure explaining elasticity. In this coil structure, glycyl residues are in external positions where they can interact with the solvent, while the more hydrophobic residues are buried. In this configuration, the backbone is fully hydrated. This model suggests that in a static state, elastin chains are organised in a particular way that would be lost upon extension but that this organisation could be fully recovered after stretching. In this context, both the chains and water seem to be important for elasticity.

After considerable work performed on the VPGVG sequence, which is tandemly repeated in bovine elastin, Urry proposed the β-spiral model in 1977 (Urry and Long, Reference Urry, Long, Sandberg, Gray and Franzblau1977). In this model, repeated motifs form consecutive β-turns that organise in a helix-like structure, which would be expected to manifest as a filamentous organisation at the supramolecular level. In the folded β-spiral, water surrounds β-turn chains that are free to rotate about their backbone bonds. These moieties are expected to rock or ‘librate’. According to Urry’s model, the entropic nature of elastin originates from these librational motions that decrease when the polymer is stretched.

When the full sequence of tropoelastin became available, it became clear that the sequences that gave rise to the oiled coil and β-spiral model only represented a small portion of the sequence and could not fully explain elastin 3D structure. While the oiled drop model was rapidly abandoned by the community, the β-spiral was still discussed because numerous experimental data suggested that the β-turn would be an important secondary structure in elastin.

After extensive conformational analysis of synthetic elastin-derived sequences, Tamburro’s sliding turn model was proposed (Tamburro, Reference Tamburro1990). As Urry’s, this model mainly focuses on the conformational behaviour of the hydrophobic domains. But while Urry proposed regular arrangement of turns, the sliding turn model states that these regions are characterised by the presence of labile and rapidly exchanging secondary structure features, including turns, distorted β-stands or poly-proline II elements (Tamburro, Reference Tamburro2009). This dynamical interchange is amplified by hydration and makes turns ‘slide’ along the chains in a perpetual movement (Tamburro et al., Reference Tamburro, Guantieri, Pandolfo and Scopa1990; Lelj et al., Reference Lelj, Tamburro, Villani, Grimaldi and Guantieri1992). For Tamburro, while the cross-linking domains are rather stable and rigid, the elastic domains of elastin are dominated by these sliding structures, which explains the high conformational entropy required for elasticity.

Based on small-angle X-ray and neutron scattering experiments, Weiss and colleagues (Baldock et al., Reference Baldock, Oberhauser, Ma, Lammie, Siegler, Mithieux, Tu, Chow, Suleman, Malfois, Rogers, Guo, Irving, Wess and Weiss2011) described the shape of hydrated tropoelastin. Further, they analysed the elastic behaviour of the isolated monomer using the worm-like chain model of polymer elasticity (Lapidus et al., Reference Lapidus, Steinbach, Eaton, Szabo and Hofrichter2002) and concluded that elastin elasticity is explained by the mechanical properties of its monomer. In this model, elastin is described as an ‘oiled coiled coil’ in which tropoelastin molecules are arranged head to tail to form intertwined microfibers. Different from other models, this model is not directly linked to a specific explanation for the elastin entropic behaviour as it does not provide data about the conformational behaviour of structural ensembles found within the tropoelastin shape. Because this model consists of a single conformation that resembles the shape derived from SAXS data, it is not readily reconcilable with the concept of structural disorder and statistical mechanics prevailing for elastin structure and function. Consequently, Weiss model for elastin should be better described as a model for tropoelastin assembly.

Cross-links formation

Since their first identification (Partridge et al., Reference Partridge, Elsden and Thomas1963), considerable effort has been made to understand the processes leading to the formation of mature elastin via cross-linking of tropoelastin. To date, this process is not fully understood but important advances have been made. The identification of cross-links joining domains 10, 19, and 25 was a key step (Brown-Augsburger et al., Reference Brown-Augsburger, Tisdale, Broekelmann, Sloan and Mecham1995). Indeed, this finding suggested that, at least in the initial steps of polymerisation, a regular organisation of tropoelastin molecules would be required to assemble elastin. However, more recently, extensive mass spectrometry analyses performed by Schmelzer’s group convincingly established that the formation of elastin cross-links was mainly a random process (Schräder et al., Reference Schräder, Heinz, Majovsky, Karaman Mayack, Brinckmann, Sippl and Schmelzer2018; Hedtke et al., Reference Hedtke, Schräder, Heinz, Hoehenwarter, Brinckmann, Groth and Schmelzer2019). Therefore, it seems reasonable to propose that, despite some level of ordered structures, cross-link formation is a random process that would enhance the overall entropic nature of the polymer.

Fractality

The supramolecular organisation of bovine ligamentum nuchae elastin and α-elastin either as coacervate or as a lyophilised powder has been studied by scanning and transmission microscopy (Tamburro et al., Reference Tamburro, De Stradis and D’Alessio1995). Importantly, these experiments evidenced filaments, fibrils, fibres, networks and dendritic, leaf-like forms, that could be rationalised in terms of fractal geometry.

Fractal geometry describes natural occurring objects with non-integer dimensions, including complex chemical systems (Borman, Reference Borman1991) and large aggregates (Family, Reference Family1993). The most important property of a fractal is its self-similarity. A piece of a fractal object looks like the whole, namely it has no characteristic length scale. When a fractal is observed at increasing resolution, it shows increasingly higher level of detail and the following power law (Sander et al., Reference Sander, Sander and Ziff1994) holds: $ N=c\;{s}^{-D} $ , where N is the smallest number of boxes of linear dimension s that can cover the object. D is the fractal dimension and c is a constant. For elastin, the fractal dimension D was determined by using the box-counting method and was in the range between 1.4 and 1.9 (Tamburro et al., Reference Tamburro, De Stradis and D’Alessio1995). Comparable fractal characteristics were calculated for some aggregates from recombinant elastin-inspired partially ordered polypeptides. A two-stage aggregation process was suggested and predicted by coarse-grained simulations for these peptides depending on temperature: when it is raised beyond the transition temperature (T_t) the mesoscopic nuclei rapidly link, forming fractal networks (Roberts et al., Reference Roberts, Harmon, Schaal, Miao, Li, Hunt, Wen, Oas, Collier, Pappu and Chilkoti2018). This type of coacervation is mirrored in tropoelastin which undergoes a multistage process (Yeo et al., Reference Yeo, Keeley and Weiss2011).

Reductionist approach

The fractal nature of assembled elastin and hence self-similarity means that even very short elastin sequences should show molecular and supramolecular features which are similar to those of the whole elastomer. As a consequence, past studies performed on short synthetic peptides corresponding to repeating sequences (Long et al., Reference Long, Rapaka, Volpin, Pasquali-Ronchetti and Urry1980; Tamburro et al., Reference Tamburro, Guantieri, Pandolfo and Scopa1990, Reference Tamburro, Guantieri and Gordini1992) or to recombinant polypeptides spanning regions encoded by exons 20 to 24 (Miao et al., Reference Miao, Cirulis, Lee and Keeley2005; Reichheld et al., Reference Reichheld, Muiznieks, Stahl, Simonetti, Sharpe and Keeley2014) may be highly informative regarding larger-scale elastin assembly. However, the choice of specific sequences within elastin chains may mean that such studies are not representative of all exons within the monomer.

To fill this gap, Tamburro and colleagues took advantage of the cassette-like organisation of the eln gene, where exons encoding hydrophobic sequences alternate regularly with lysine-rich domains containing of those rich in alanine (KA domains) and those containing proline (KP domains), and undertook the exon-by-exon synthesis of human tropoelastin domains (Tamburro et al., Reference Tamburro, Bochicchio and Pepe2003, Reference Tamburro, Pepe and Bochicchio2006; Pepe et al., Reference Pepe, Flamia, Guerra, Quaglino, Bochicchio, Pasquali-Ronchetti and Tamburro2008). The working hypothesis of their ‘reductionist approach’ (Salvi et al., Reference Salvi, Moscarelli, Bochicchio, Lanza and Castle2013) was that each domain would adopt a specific and independent conformation in the polymer, and possibly play a specific function.

CD is a useful and quick tool for assessing the secondary structure of polypeptides in solution, which, in the context of the reductionist approach, depends on temperature and solvent. In particular, temperature is a critical parameter for extended conformations such as poly-l-proline II (PPII) and folded structures like turns (Tamburro et al., Reference Tamburro, Pepe, Bochicchio, Quaglino and Ronchetti2005; Bochicchio and Pepe, Reference Bochicchio and Pepe2011). Therefore, studies as a function of temperature were carried out both in aqueous solution and 2,2,2-trifluoroethanol (TFE). TFE is a solvent which is less polar than water and which favours regular secondary structures such as helices and turns stabilised by intramolecular hydrogen bonds. The rationale behind the use of TFE is to assess the possible conformations adopted by the polypeptides in a more shielded environment as compared to the real local conditions they experience in hydrated elastin. These in situ conditions are unknown and more likely intermediate between the two solvent extremes, water and TFE. Of particular interest, is the finding that in an aqueous solution the dominant conformation for hydrophobic domain peptides is apparently the PPII often in equilibrium with the so-called ‘unordered’ state. The results are summarised in Table 1.

Table 1.

Summary of secondary structure elements presence in elastin domains as evidenced by CD and NMR following the reductionist approach

#d, domain number. Domain types: h, hydrophobic domain; KP, KP cross-linking domain; KA, KA cross-linking domain; CT, C-terminal. Secondary structures: α, α-helix; PPII, poly-proline II helix; β1, type I β-turn; β2, type II β-turn; T, turn; U, unordered. ^†Domain 26A is not a distinct domain but rather an extension of domain 26 generated by read-through beyond the normal donor splice site in the genomic sequence. The protein sequence is highly uncharacteristic of elastins, rarely expressed, and may be artefactual (Bisaccia et al., Reference Bisaccia, Castiglione-Morelli, Spisani, Ostuni, Serafini-Fracassini, Bavoso and Tamburro1998).

The reductionist approach underlined that PPII is a significant contributor to the conformational ensemble in flexible, unordered polypeptides (Bochicchio and Pepe, Reference Bochicchio and Pepe2011). In fact, recently a new picture of unordered conformations emerged from the analysis of unfolded proteins and unordered peptides that points to more restricted conformational sampling in the unordered state with a limited range of accessible conformations.

Many spectroscopic studies and molecular dynamics simulations propose that short stretches of 2–5 residues in PPII are the most favoured conformations in solution for unordered and unfolded peptides (Bochicchio et al., Reference Bochicchio, Pepe, Crudele, Belloy, Baud and Dauchez2015). Furthermore, the unordered form is constituted not only by PPII-helical stretches with three or more residues but also by conformers in which individual residues are in PPII, α-, or β-regions of the Ramachandran map, rapidly interconverting. In TFE, in contrast, folded conformations, such as type I and type II β-turns are found together with α-helices in polyalanine sequences and with unordered conformations in hydrophobic and KP domains (Tamburro et al., Reference Tamburro, Pepe and Bochicchio2006).

In conclusion, because the polarity of the microenvironment might be intermediate between water and TFE, the reductionist approach led to the proposal of a conformational equilibrium for elastin, comprising PPII, β-turns, and unordered conformations, as illustrated in Figure 3. Each of the three terms at the equilibrium should comprise, on their own, additional internal fluctuations. Such is, for example, the case of PPII, where most probably the trans-conformational transitions involve the entire upper left region of the Ramachandran map, that is, the region of the (quasi) extended conformation, including also β-strands. Similarly, the term ‘β-turns’ refers to fluctuations around the energy minimum of isolated β-turns and also equilibria such as type I ⇄ type II, which have been proposed as a common structure in elastin, and are central to the ‘sliding β-turns’ model proposed by Tamburro as an explanation for elastin elasticity (Tamburro et al., Reference Tamburro, Guantieri, Pandolfo and Scopa1990; Lelj et al., Reference Lelj, Tamburro, Villani, Grimaldi and Guantieri1992).

Figure 3.

Conformational equilibrium characterising the elastic domains of elastin. (a) Elastic domains do not present stable conformations but rather rapidly and constantly interchanging turns, PPII and unordered states engaged in a conformational equilibrium sustained by the presence of surrounding water molecules. This process explains elastin exceptional conformational entropy. Boxes point at turns sliding along the sequence. (b) Illustration of a turn sliding along the sequence. The example sequence is GVGGL. The H-bond defining the turn shifts from positions G₁ → G₄ to V₂ → L₅.

As a native, relaxed state at high entropy is essential for conferring elasticity according to the classical theory of rubber elasticity, a pivotal role in this context is played by the PPII flexibility. In fact, the main peculiarity of PPII lies in its high flexibility undoubtedly due to the absence of intra-chain hydrogen bonds. Accordingly, the possibility that PPII is involved, as suggested above, in multi-conformational equilibria such as PPII/unordered/β-turns, should conceivably increase the entropy of the relaxed state, and therefore allow the expression of elasticity. In this sense, the reductionist approach strongly supports the general equilibrium extended ⇄ folded, proposed by Tamburro model as the source of elastin high entropy in the relaxed state.

Numerical simulations

Molecular dynamics (MD) simulation is a powerful computational technique for studying the properties of biological molecules at the atomic level. A protein is represented as a set of atoms that interact according to the laws of classical mechanics. By simulating the motion of these atoms, the conformational changes and dynamic properties of the protein can be sampled. The result is a trajectory that describes the positions and velocities of all the atoms in the molecule over time. Simulations have been used to study a wide range of biological processes, including protein folding/unfolding, ligand binding and enzyme catalysis, and have provided valuable insights into the molecular basis of these phenomena. Despite its computational cost, MD simulation has become an essential tool for studying globular proteins, filling the gap between experiment and theory and shedding new light on the fundamental principles governing protein structure and function (Bottaro and Lindorff-Larsen, Reference Bottaro and Lindorff-Larsen2018).

MD simulation is an essential tool for studying poorly structured or intrinsically disordered proteins (IDPs), that is proteins lacking a well-defined 3D structure under physiological conditions. When studying IDP by MD, simulation allows the exploration of the IDP conformational space and dynamics, generating an ensemble of conformations which can be compared to experimental observables. Although the conformational flexibility of IDPs makes their simulation particularly challenging, MD simulations constitute a valuable complement to experimental techniques, notably in the study of IDPs (Hernández et al., Reference Hernández, Crowet, Thiery, Kruglik, Belloy, Baud, Dauchez and Debelle2020; Salvi et al., Reference Salvi, Zapletal, Jaseňáková, Zachrdla, Padrta, Narasimhan, Marquardsen, Tyburn, Žídek, Blackledge, Ferrage and Kadeřávek2022).

MD relies on a force field, a mathematical model describing the interactions between atoms in a molecular system. It consists of a set of equations that computes the forces acting on each atom based on its position and the positions of neighbouring atoms. These forces, typically derived from experimental or theoretical data, determine the motion of the atoms and therefore the behaviour of the whole system. A force field typically includes terms for bond stretching, angle bending, torsional rotation, non-bonded interactions (such as van der Waals forces and electrostatic interactions), and other contributions such as solvent effects. Different force fields can have different levels of accuracy and computational efficiency, and their selection depends on the specific system being studied and the research question of interest. The choice of a force field can have a significant impact on the simulation results, and it is important to validate the model against experimental data whenever possible.

Several force fields have been developed for several decades, mainly for globular proteins. These force fields have shown their efficiency for simulating such systems, but have limitations for simulating IDPs because of their inability to reproduce the conformational ensembles explored by disordered systems (Rauscher et al., Reference Rauscher, Gapsys, Gajda, Zweckstetter, De Groot and Grubmüller2015). Consequently, new sets of parameters have been developed or adjusted from existing force fields for these systems. For instance, the CHARMM36m force field (Huang et al., Reference Huang, Rauscher, Nawrocki, Ran, Feig, De Groot, Grubmüller and MacKerell2017), developed from CHARMM36, allows to generate conformational ensembles for IDPs, the sampling method being based on the search for interactions between protein and water, to explore more specifically extended conformations. Currently, convenient simulation protocols are still under development.

In the field of elastin research, molecular modelling was used to investigate the structural organisation, elasticity, or phase transition of tropoelastin-inspired peptides. Pioneering computations were initially performed in vacuo (Chang et al., Reference Chang, Venkatachalam, Prasad and Urry1989; Lelj et al., Reference Lelj, Tamburro, Villani, Grimaldi and Guantieri1992; Villani and Tamburro, Reference Villani and Tamburro1995), then with one added shell of water molecules around (VPGVG)_n (Wasserman and Salemme, Reference Wasserman and Salemme1990), and later on, on a fully hydrated system with explicit water (Li et al., Reference Li, Alonso, Bennion and Daggett2001a).

Whereas all-atom simulations can be performed on characteristic peptide motifs of tropoelastin, more recent approaches like coarse-grained modelling do not consider each atom individually, but rather one particle as a group of several atoms. This approach is difficult to use with molecules such as tropoelastin and its motifs (Depenveiller et al., Reference Depenveiller, Wong, Crowet, Debelle, Baud, Dauchez and Belloy2023). Coarse-grained simulation approaches have been applied to ELP, but they require a systematic re-parameterisation of the potentials of the dihedral angles to prevent the secondary structures from ‘freezing’ during the simulation.

The dynamical behaviour in vacuo of an elastin-derived peptide, GLGG, was modelled (Villani et al., Reference Villani, D’Alessio and Tamburro1997) and small amplitude motions named solitons were observed. In this picture, the molecular motion is essentially quasiperiodic and of rather lower entropy. In dynamical system theory it is well established that, increasing the amplitude fluctuations of solitons, a transition to chaos of high entropy is expected (Walker and Ford, Reference Walker and Ford1969). Then, peptide chaotic motion is suggested to take place in solution or at higher temperature where large motions should occur. Moreover, considering that the common characteristic of the proposed elastin elasticity mechanism is a higher molecular freedom inside either the peptide chains or solvent molecules for the relaxed state with respect to the extended one, a transition to chaos has been proposed as a source of entropy differences and the driving force for the refolding of stretched elastin when the external force is removed. This was the starting point for a theoretical study based on non-linear systems theory (Villani et al., Reference Villani, D’Alessio and Tamburro1997; Villani and Tamburro, Reference Villani and Tamburro1999). One of the global variables considered was the trajectory of the end-to-end vector R_ee defined as the difference of position components of the Cartesian coordinate of the end carbon atoms of acetyl and N-methyl-amide groups. The perfectly irregular trajectory R_ee(t) found is considered typical of fractional Brownian motion (chaos). A typical Brownian motion is that resulting from the random walk assumed in the classical theory of elasticity (Flory, Reference Flory1953) and according to the fractal nature suggested for elastin. The large conformational freedom in solution of the free chains, either folded or extended, is apparent from the structural parameter trajectories. The motions are, in all cases, larger than the corresponding ones in vacuo. Moreover, this picture is in agreement with proposed plasticiser role of water (Megret et al., Reference Megret, Lamure, Pieraggi, Lacabanne, Guantieri and Tamburro1993). In fact, the flexibility of the peptide backbone is increased by the interaction with the solvent molecules and by efficient solvation of peptide units that reduces the intramolecular hydrogen-bonding interactions.

With increases in computational power, elaborate elastin-related systems have been simulated in explicit water. The analysis of the corresponding trajectories evidenced essential features of hydrated elastin chains dynamics. Based on the reductionist approach, numerical studies have been performed on peptides bearing the characteristic sequence patterns of tropoelastin, namely XPGXG, XGGZG, XGXXPG, XGXPGXGXG (Figure 2).

Domain 18 of bovine elastin contains 11 VPGVG tandem repeats (Raju and Anwar, Reference Raju and Anwar1987). Similar VPGXG motifs are used to design intrinsically disordered biomaterials (Roberts et al., Reference Roberts, Dzuricky and Chilkoti2015). The XPGXG motifs originating from tropoelastin sequence have been extensively studied, both experimentally by different methodologies, and in simulations, VPGVG being the consensus motif classically described (Muiznieks and Keeley, Reference Muiznieks and Keeley2017). The VPGEG and VPGKG forms are also found and widely used in biomedical applications and biomaterials (Rodríguez-Cabello et al., Reference Rodríguez-Cabello, Arias, Rodrigo and Girotti2016; Wang et al., Reference Wang, Dzuricky and Chilkoti2017; Rodriguez-Cabello et al., Reference Rodriguez-Cabello, Gonzalez De Torre, González-Pérez, González-Pérez and Montequi2021).

Most simulations performed on these peptides have been run on structural models of elastin. From such data, Urry proposed a model where the repeating (VPGVG)_n sequence would form a β-spiral (Venkatachalam and Urry, Reference Venkatachalam and Urry1981). However, such an organisation implies restricted conformations, which is incompatible with the high degree of conformational flexibility of elastin (Rauscher and Pomès, Reference Rauscher, Pomès, Fuxreiter and Tompa2012). Further, this β-spiral structure has been shown to be unstable during simulations (Li and Daggett, Reference Li and Daggett2002).

The XGGZG type motifs, with X and Z being mainly permutations of L or V residues, are also important motifs of tropoelastin. They are found in domains 7, 28, and 30 (Figure 2). These sequences play a major role in aggregation and elastic fibre formation (Bochicchio et al., Reference Bochicchio, Aït-Ali, Tamburro and Alix2004). A variant of a natural sequence, VGGVG, is often used in biomaterial studies. Numerical simulations performed on these patterns (Bochicchio et al., Reference Bochicchio, Pepe, Crudele, Belloy, Baud and Dauchez2015) show that, in solution, these peptides mostly adopt a type I β-turn rapidly exchanging with type II β-turns and irregular β-stands.

The VGVAPG is the archetypical XGXXPG-type motif. It is found six times in domain 24 of human tropoelastin (Indik et al., Reference Indik, Yeh, Ornstein-Goldstein, Sheppard, Anderson, Rosenbloom, Peltonen and Rosenbloom1987) where it is tandemly repeated three times, followed by VGLAPG and then again tandemly repeated three times. The VGVAPG motif has deserved considerable attention because it presents numerous biological activities (Le Page et al., Reference Le Page, Khalil, Vermette, Frost, Larbi, Witkowski and Fulop2019). As such, VGVAPG belongs to the group of matrikines and more specifically to the group of elastokines, bioactive peptides derived from elastin sequences (Maurice et al., Reference Maurice, Blaise, Gayral, Debelle, Laffargue, Hornebeck and Duca2013). Numerical simulations performed on XGXXPG peptides show that this pattern adopts a mixture of extended PPII and folded turn-like conformations in equilibrium (Floquet et al., Reference Floquet, Héry-Huynh, Dauchez, Derreumaux, Tamburro and Alix2004; Rauscher et al., Reference Rauscher, Baud, Miao, Keeley and Pomès2006). This description supports the sliding turns model (Tamburro et al., Reference Tamburro, Bochicchio and Pepe2003, Reference Tamburro, Bochicchio and Pepe2005). The canonical β-turn classification defines 8 types of β-turns (Wilmot and Thornton, Reference Wilmot and Thornton1988). Strikingly, within the VGVAPG peptide, the type VIII β-turn is very frequently encountered. This propensity of VGVAPG (and XGXXPG peptides) to adopt a type VIII β-turn conformation correlates with their biological activity (Floquet et al., Reference Floquet, Héry-Huynh, Dauchez, Derreumaux, Tamburro and Alix2004; Moroy et al., Reference Moroy, Alix and Héry-Huynh2005; Fuchs et al., Reference Fuchs, Bonvin, Bochicchio, Pepe, Alix and Tamburro2006; Blanchevoye et al., Reference Blanchevoye, Floquet, Scandolera, Baud, Maurice, Bocquet, Blaise, Ghoneim, Cantarelli, Delacoux, Dauchez, Efremov, Martiny, Duca and Debelle2013).

The XGXPGXGXG sequence motif found in domain 26 of human tropoelastin defines a class of elastokines distinct from the XGXXPG (Long et al., Reference Long, King, Prasad and Urry1988; Maeda et al., Reference Maeda, Mizoiri, Briones and Okamoto2007). Numerical and experimental analyses have been used to address their conformational behaviour (Hernández et al., Reference Hernández, Crowet, Thiery, Kruglik, Belloy, Baud, Dauchez and Debelle2020). As with other elastin-related sequences, these peptides are extremely dynamic when hydrated. Nevertheless, it appears that their intrinsic conformational changes favour type II β-turn conformation suggesting a possible role in their biological activities.

Overall, independently of the considered sequence motif, numerical simulations strongly support a conformational equilibrium in hydrated elastin peptides (Figure 3) as suggested by the reductionist approach. In agreement with spectroscopic studies, the structure of the hydrophobic domains of elastin therefore consists of sparse, transient, and local secondary structure elements which can be described in a statistical way.

Following the description of the molecular shape of tropoelastin derived from the SAXS/SANS envelope of a recombinant tropoelastin and some of its fragments, namely domains 2–18 and 2–25 (Baldock et al., Reference Baldock, Oberhauser, Ma, Lammie, Siegler, Mithieux, Tu, Chow, Suleman, Malfois, Rogers, Guo, Irving, Wess and Weiss2011), the first molecular model of a 698-residue long tropoelastin was obtained by folding an extended monomer combining implicit and explicit replica exchange molecular dynamics simulations (Tarakanova et al., Reference Tarakanova, Yeo, Baldock, Weiss and Buehler2018). Although these results are remarkable, the proposed model only represents one possible conformation. Further, given the demonstrated highly flexible and mobile nature of elastin chains, one can wonder about the generation of a unique model, lacking publicly available coordinates.

The tropoelastin model presents several convincing features, but it is incompatible with the established concept of structural disorder and the notion of conformational equilibrium that rule elastin function. Therefore, we feel that this model should be cautiously considered as it is misleading to represent a highly disordered protein by a single conformation.

Monolayer, multilayer, and bulk water

Water equilibrium sorption–desorption isotherms of initially dry elastin revealed critical water contents for the gradual formation and completion of the protein hydration (Panagopoulou et al., Reference Panagopoulou, Kyritsis, Vodina and Pissis2013). The S-shaped sorption isotherm satisfactory fits the well-known Guggenheim-Anderson-de Boer equation (Timmermann, Reference Timmermann1989), allowing to define monolayer, multilayer, and bulk water. A first critical value x_w = 6.5% (h = 0.07) is found for the filling of the first sorption layer, corresponding to 0.3 molecules of water by residue of elastin. A second critical value x_w = 11.5% (h = 0.13) is found as the onset of water clustering (filling of multilayers); a third critical value of x_w = 18% (h = 0.15) must be reached for the organisation of water in the protein hydration shell, which roughly corresponds to 0.7 molecules of water per residue and could be related to the beginning of swelling. Finally, a maximum water uptake of 23% (h = 0.3) is determined from vapour sorption experiments leading to significant bulk water, which roughly corresponds to 1.4 molecules of water per residue. It must be noted that usually about 0.3–0.7 g of water remains associated per gram of ‘dry’ protein (Nandi et al., Reference Nandi, Bhattacharyya and Bagchi2000). In the case of elastin, the low value of 0.3 ranks elastin in the category of natural hydrophobic proteins, but in comparison with synthetic polymers it also allows to classify elastin as a hydrophilic polymer.

Freezable and unfreezable water

Differential scanning calorimetry (DSC), that measures the variation of the heat capacity during a temperature program, is a well-suited technique to determine phase transitions in any kind of sample. In calorimetric experiments of polymers, water molecules are categorised into two types according to their behaviour during freezing and thawing. Freezable water gives crystallisation endotherms during cooling and melting endotherms during heating. In contrast, unfreezable water cannot freeze and does not generate transitions.

The term ‘freezable water’ should be preferred to the term ‘free’ water, reserved to vibrational or relaxational techniques (Rault et al., Reference Rault, Ping and Nguyen1995; Tang et al., Reference Tang, Samouillan, Dandurand, Lacabanne, Lacoste-Ferre, Bogdanowicz, Bianchi, Villaret and Nadal-Wollbold2017). In the same way, care should be taken when associating unfreezable water with ‘bound’ water, since hydrogen bonding is not the ‘exclusive’ factor influencing water crystallisation. Indeed, nanocavities in the polymer can be also an important reason for the formation of unfrozen water (Liu and Yao, Reference Liu and Yao2001).

The first DSC experiments performed on native, purified and soluble elastin (Ceccorulli et al., Reference Ceccorulli, Scandola and Pezzin1977) evidenced a maximal amount of unfreezable water at x_w = 25% (h = 0.33) in the elastin-water system, leading to a molar ratio of unfreezable water per residue close to 1.5. This value was consistent with the hypothesis of a direct solvation of the main chain and interpreted according to a model previously proposed for polyamides, with three water molecules for two residues: the first strongly bound to two carbonyl groups of elastin, the others weakly bound, and inserted between pre-existing hydrogen bonds between carbonyl and amine groups.

As shown in Figure 4, calorimetric experiments on hydrated elastin evidenced the appearance of freezable water for water amounts above x_w = 22%, that is, h = 0.3 (Samouillan et al., Reference Samouillan, André, Dandurand and Lacabanne2004, Reference Samouillan, Tintar and Lacabanne2011; Panagopoulou et al., Reference Panagopoulou, Kyritsis, Vodina and Pissis2013), in accordance with equilibrium sorption–desorption isotherm experiments. Cross-links and conformational changes can explain the slight decrease of the maximal molar ratio of unfreezable water per elastin residue (1.4) when compared to soluble elastin (1.5).

Figure 4.

Freezable and unfreezable water in elastin. DSC heating thermograms (10 °C/min) of elastin-water systems evidencing the detection of endothermic melting of ice (i.e. detection of freezable water) for h = 0.28 (x_w = 22%) and intensification of the endotherm for higher hydration. Reprinted (adapted) with permission from Samouillan et al. (Reference Samouillan, Tintar and Lacabanne2011). © 2011 Elsevier.

Interestingly, solid-state ¹³C NMR spectra on differentially hydrated elastins exhibit a clear evolution above h = 0.3, supporting the existence of a localised water phase under this peculiar hydration (Perry et al., Reference Perry, Stypa, Tenn and Kumashiro2002). As water is added, one plausible model might include an extensive hydrogen-bonding network from peptide backbone to water molecules in the first, second and subsequent hydration layers.

The endothermic peak corresponding to the melting of freezable water is widely used to quantify freezable water in proteins and tissues, by dividing the area of the peak by 0.334 J.kg^‑1, corresponding to the melting enthalpy of pure ice at 0 °C (Heys et al., Reference Heys, Friedrich and Truscott2008; Panagopoulou et al., Reference Panagopoulou, Kyritsis, Vodina and Pissis2013). Total hydration can be reached by thermogravimetric analysis and the amount of unfreezable water by a simple difference (Tang et al., Reference Tang, Samouillan, Dandurand, Lacabanne, Lacoste-Ferre, Bogdanowicz, Bianchi, Villaret and Nadal-Wollbold2017). The broadness of the endotherm or presence of shoulders can be due to the subdivision of freezable water into different categories as reported for various biopolymers/water systems: (1) bulk water in excess, which can form its classical four H-bonds with neighbouring water molecules and freezes/melts at the usual freezing point of water, (2) intermediate water (Liu and Yao, Reference Liu and Yao2001), which freezes/melts at a lower temperature and can be associated with water in mesopores (Hay and Laity, Reference Hay and Laity2000; Tang et al., Reference Tang, Samouillan, Dandurand, Lacabanne, Lacoste-Ferre, Bogdanowicz, Bianchi, Villaret and Nadal-Wollbold2017), (3) freezable ‘bound’ water, less closely bound and immobilised (Hatakeyama et al., Reference Hatakeyama, Nakamura and Hatakeyama1988). Another kind of water, often named ‘hydrophobic hydration’, corresponds to clathrate-like structures of water around hydrophobic solutes (Teeter, Reference Teeter1984). In water/hydrophobic compounds mixtures, clathrate-like structures of water generally dissociate in the 273 K region (Nguyen et al., Reference Nguyen, Diger Berger, Wagner, Thiel, Rgen Butt and Graf2021) or at slightly higher temperature (López-Bueno et al., Reference López-Bueno, Herreros-Lucas, Suárez-Rodríguez, Bittermann, Amigo, Woutersen, Giménez-López and Rivadulla2021) due to the specific arrangement of water around the hydrophobic solute, arising from a steric effect (Rezus and Bakker, Reference Rezus and Bakker2007).

Even if hydrophobic hydration is hardly detectable by calorimetric measurements in hydrated cross-linked elastin because it overlaps with frozen water melting, this type of hydration is crucial to explain the thermal behaviour of aqueous solutions of elastin polypeptide analogues (Luan et al., Reference Luan, Parker, Gowda and Urry1992; Urry, Reference Urry2004; Ribeiro et al., Reference Ribeiro, Arias, Reguera, Alonso and Rodríguez-Cabello2009). It was shown to be intricately linked with the hydrophobic effect, that plays a major role in the coacervation phenomenon leading to fibrillogenesis of elastin. The hydrophobic effect can be described as the tendency of apolar groups to associate in an aqueous solution, minimising the total hydrophobic surface exposed to water (Urry, Reference Urry2004; Rezus and Bakker, Reference Rezus and Bakker2007) – and thereby the hydrophobic hydration – when the temperature is raised, leading to a phase separation. During this phase separation, the decrease in order of structured water surrounding hydrophobic groups that becomes bulk water exceeds the ordering of the protein or polypeptide, implying that the entropy of the whole system (water and protein) increases. This transition referred as an inverse temperature transition (ITT) in proteins or lower critical solution temperature (LCST) in polymers, has been evidenced by classic calorimetric measurements to be dominated by the endothermic process of the clathrate-like water melting occurring in the 20–40 °C range (Urry et al., Reference Urry, Luan, Parker, Gowda, Prasad, Reid and Safavy1991; Rodríguez-Cabello et al., Reference Rodríguez-Cabello, Reguera, Alonso, Parker, McPherson and Urry2004; Ribeiro et al., Reference Ribeiro, Arias, Reguera, Alonso and Rodríguez-Cabello2009).

Modulated differential scanning calorimetry (M-DSC) is required to detect the minor exothermic part due to Van der Waals cohesive interactions of the polypeptide chains accompanying the association (Rodriguez-Cabello et al., Reference Rodriguez-Cabello, Alonso, Diez, Caballero and Herguedas1999; Urry, Reference Urry2004). Calorimetric experiments on specific domains from elastin have shown that the temperature for the formation of clathrate-like water, which depends on sequence, was closely connected to the coacervation temperature (Dandurand et al., Reference Dandurand, Samouillan, Lacabanne, Pepe and Bochicchio2015) and could be used to tune the coacervation of polypeptides. The dependence of ITT/LCST on the amino acid sequence points out that the water structures in elastin present a heterogeneous distribution (Ribeiro et al., Reference Ribeiro, Arias, Reguera, Alonso and Rodríguez-Cabello2009).

Free, bound, and hydrophobic water

Interfacial water is considered as a part of the structures of macromolecules. The mobility of these water molecules is very different from that of pure bulk or ‘free’ water. Dynamically oriented and exhibiting restricted motions, these ‘bound’ water molecules are identified by a slowing down of their dynamics as shown by NMR spectroscopy, quasi-elastic neutron-scattering spectroscopy and dielectric relaxation techniques (Gniadecka and Jemec, Reference Gniadecka and Jemec1998; Kaatze et al., Reference Kaatze, Behrends and Pottel2002; Sun and Boutis, Reference Sun and Boutis2010; Cametti et al., Reference Cametti, Marchetti, Gambi and Onori2011; Reuhl and Vogel, Reference Reuhl and Vogel2022).

With the emergence of 2D NMR spectroscopies, magnetic relaxation methods have played a leading role in the study of protein hydration dynamics (Halle, Reference Halle2004). The dynamics properties of water are probed by the correlation time between the proton spin–lattice (longitudinal) relaxation time T1 and the proton spin–spin (transverse) relaxation time T2. Complementing the earlier work of Ellis and Packer (Reference Ellis and Packer1976), four reservoirs of water could be discriminated by their mobilities in the elastin-water system (for an hydration h = 1) at 25 °C using deuterium 2D T1-T2 NMR (Sun et al., Reference Sun, Mitchell, Huang and Boutis2011; Sun and Boutis, Reference Sun and Boutis2011; Silverstein et al., Reference Silverstein, Bilici, Morgan, Wang, Zhang and Boutis2015). Correlation times ranged from 5 ps, for the free water outside the fibres, to 112 ns for the more restrained water, closer to the protein (Figure 5). Interestingly, with increasing strain on the polymer, experimental ¹³C NMR data evidenced an increase of the frequency of the protein backbone motion associated with an entropy decrease, as commonly described for the rubber elasticity of elastomers.

Figure 5.

Four types of water in hydrated elastin are identified by their distinct dynamics. ²H T1–T2 NMR relaxation times of water in hydrated elastin from bovine ligament (h = 1) at 25 °C. Component α ₁ corresponds to free water external to elastin fibres, component α ₂ to free water between the elastin fibres, component β to mobile water within the elastin fibres moving along tortuous channels and component γ to more restricted water. Reprinted (adapted) with permission from Sun et al. (Reference Sun, Mitchell, Huang and Boutis2011). © 2011 American Chemical Society.

Surprisingly, an increase of the random tumbling of localised water is also observed in this water-elastin system. This fact can be interpreted by an increase of the orientational entropy of water molecules. This could appear in contradiction with the short-time dynamics simulations performed on (VPGVG)₁₈ (Li and Daggett, Reference Li and Daggett2002) suggesting a decrease of water orientational entropy near hydrophobic groups during the elongation process. Nevertheless, it agrees with the simulations of (VPGVG)₃ which show that the lifetime of peptide-water H-bonds decreases with increasing strain, while the peptide entropy decreases (1.63–1.46 kJ/mol.K).

Dielectric relaxational spectroscopy (DRS), which consists of measuring the complex dielectric permittivity of a sample as a function of frequency and temperature, has been widely used to investigate the arrangement of water in hydrated biopolymers or aqueous solutions of proteins (Bone and Pethig, Reference Bone and Pethig1985; Nandi et al., Reference Nandi, Bhattacharyya and Bagchi2000). Indeed this technique provides information on the orientational dynamics of molecular dipoles and covers all kinds of polarisation fluctuations in the ms to ps time scales (Cametti et al., Reference Cametti, Marchetti, Gambi and Onori2011). To extract reliable information about protein hydration dynamics from DRS, which gives average information, relatively high protein concentrations are required since only water molecules in the vicinity of protein surface are perturbed (Halle, Reference Halle2004).

At temperatures above 0 °C, several relaxation processes are detected in the dielectric spectra of hydrated biomacromolecules or aqueous protein solutions. They are assigned to molecular mechanisms involving motion of the protein backbone, of side groups in the protein conformation, and of various layers of strongly and/or weakly bound water and free water. In order of increasing relaxation frequency, the three main relaxations are generally associated with protein tumbling (frequency f < 10 MHz, relaxation time τ > 15 ns), to hydration water in the interfacial region surrounding the biomolecule (τ in the 20 ps–10 ns range) and to the orientational polarisation of the free water molecules (f > 10 GHz, τ < 15 ps), respectively) (Kaatze, Reference Kaatze1997; Cametti et al., Reference Cametti, Marchetti, Gambi and Onori2011). Dielectric studies on concentrated solutions of polypentapeptide analogues of elastin (Buchet et al., Reference Buchet, Luan, Prasad, Harris and Urry1988; Urry et al., Reference Urry, Peng and McPherson1997), an elastin coacervate (Urry et al., Reference Urry, Henze, Redington, Long and Prasad1985), or fibrous elastin (Bone and Pethig, Reference Bone and Pethig1985; Urry et al., Reference Urry, Hugel, Seitz, Gaub, Sheiba, Dea, Xu and Parker2002) highlighted both the low-frequency dielectric dispersion due to a librational mode of the protein/polypeptide and the high-frequency dispersion due to water. This latter dispersion was found to be a composite dispersion (Urry et al., Reference Urry, Peng and McPherson1997) due to at least three kinds of water: free water, water hydrogen bonded to the polar groups of the polypeptide backbone and clathrate-like water associated with the hydrophobic side chains. In the case of elastin-like polypeptides, the magnitude of water relaxation, depending on the hydrophobicity, decreased with increasing temperature between 0 and 60 °C, while the magnitude of the librational mode increased showing an ITT to a more ordered structure. These results corroborate calorimetric studies and demonstrate the temperature dependence of the water-polypeptide system, highlighting the role of hydrophobicity in the structure formation of these bioelastomers.

Complementary dielectric measurements on hydrated elastin were also performed at low temperature, in a zone where the relaxation time of the hydration water orientation drastically slows down and can be examined in detail. These experiments evidenced three relaxation modes for the complex formed by elastin and its hydration shell (Figure 6). This first one is a rapid process, known as the universal relaxation in deeply super-cooled confined water due to the reorientation of water molecules in the external layer of the hydration shell of elastin (secondary hydration shell), with an associated energy corresponding to the breaking of two hydrogen bonds (Gainaru et al., Reference Gainaru, Fillmer and Böhmer2009; Sudo and Yagihara, Reference Sudo and Yagihara2009). This process is also detected by ²H NMR spin–lattice relaxation (Lusceac et al., Reference Lusceac, Vogel and Herbers2010). The second is an intermediate process associated with the dynamics of the internal layer of the hydration shell directly linked to the protein (primary hydration shell). This third is a slower, highly cooperative process, involving large motions and mainly due to protein dynamics.

Figure 6.

Dynamics of elastin and its hydration shell. A. Dielectric relaxation map of hydrated elastin (h = 0.25) as a function of frequency and temperature, showing the fastest (β₁), intermediate (β₂) and slowest (p1) relaxation modes. Reprinted (adapted) with permission from Samouillan et al. (Reference Samouillan, Tintar and Lacabanne2011). © 2011 Elsevier. B. Relation between the relaxation time of the two fastest β of hydrated elastin indicating that the motions of the protein/water complex are controlled by the fast motions of water in the secondary hydration shell.

The direct power law found between the relaxation time of the two fastest processes demonstrates that the fast dynamics of water in the secondary hydration shell drive the slower dynamics of the complex protein-strongly confined/bound water. As shown for hydrated globular proteins (Fenimore et al., Reference Fenimore, Frauenfelder, McMahon and Young2004; Frauenfelder et al., Reference Frauenfelder, Chen, Berendzen, Fenimore, Jansson, McMahon, Stroe, Swenson and Young2009), internal proteins motions are conditioned by the fluctuations of their hydration shells.

It must be pointed out that even for a low hydration level (h = 0.1), the dynamics of the complex protein/primary hydration shell are detectable, and like those observed for higher hydration levels. As formerly demonstrated for hydrated elastin (Bone and Pethig, Reference Bone and Pethig1985), in contrast to what is observed for globular proteins, bound water exhibits some freedom of motion, even at very low hydration (Samouillan et al., Reference Samouillan, Tintar and Lacabanne2011), correlated to the onset of flexibility of the protein (Nandi et al., Reference Nandi, Bhattacharyya and Bagchi2000).

It was recently shown by DRS/NMR/DSC that the dynamics of hydrated elastin are exactly the same as those of highly asymmetric mixtures of glass formers (Reuhl and Vogel, Reference Reuhl and Vogel2022). This finding strongly supports the fact that water is a crucial and indissociable partner of elastin. Water does play an essential role in the structure and dynamics of hydrated elastin, such as in an extensive hydrogen-bonding network, where water molecules cannot be considered just as a discrete phase (Perry et al., Reference Perry, Stypa, Tenn and Kumashiro2002).

Biomaterials

One of the first and direct use of ELPs was their incorporation in the design of biomaterials, because the cassette-like organisation of the eln gene facilitated the design and production of biomaterials incorporating elastin sequences (Girotti et al., Reference Girotti, Reguera, Rodriguez-Cabello, Arias, Alonso and Testera2004). These biomaterials are now commonly considered in tissue engineering strategies as they present valuable properties (Chilkoti et al., Reference Chilkoti, Christensen and Mackay2006). First, they can be genetically encoded allowing important production. Second, they are responsive to stimuli (notably temperature) and, finally, they are biocompatible. Therefore, ELPs are now commonly considered in tissue engineering strategies as they constitute a new paradigm in biomaterial design (Rodríguez-Cabello et al., Reference Rodríguez-Cabello, Prieto, Arias, Reguera and Ribeiro2006). Further, ELP can be incorporated and used as hydrogels (Pepe et al., Reference Pepe, Maio, Bracalello, Quintanilla-Sierra, Arias, Girotti and Bochicchio2021; Sharma et al., Reference Sharma, Sharma and Roy2021) or electrospun materials of various shapes (Boland et al., Reference Boland, Matthews, Pawlowski, Simpson, Wnek and Bowlin2004; Laezza et al., Reference Laezza, Pepe and Bochicchio2022). The biomechanical properties of ELP-based hydrogels can be designed (Muiznieks and Keeley, Reference Muiznieks and Keeley2017) and they are sensitive to the medium, notably to pH (Hollingshead and Liu, Reference Hollingshead and Liu2020). ELP-based hydrogels have been successfully used for wound care (Wen et al., Reference Wen, Mithieux and Weiss2020) and to repair cartilage (Chen et al., Reference Chen, Zhang, Li, Wei, Zhao and Chen2021). Nevertheless, their usage is more noticeable in cardiac (Gonzalez de Torre I et al., Reference de Torre, Alonso and Rodriguez-Cabello2020; Hume et al., Reference Hume, Kanagalingam, Deshmukh, Chen, Mithieux, Rashid, Roohani, Lu, Doan, Graham, Clayton, Slaughter, Kizana, Stempien-Otero, Brown, Thomas, Weiss and Chong2023) and vascular (Mahara et al., Reference Mahara, Kiick and Yamaoka2017; Natsume et al., Reference Natsume, Nakamura, Sato, Ohtsuki and Sugawara-Narutaki2022) regenerative medicine. Recently, a bioink containing modified and functionalised ELPs was successfully used to print scaffolds presenting excellent biocompatibility and mechanical properties (Dai et al., Reference Dai, Belaïdi, Fleury, Garanger, Rielland, Schultze and Lecommandoux2021), offering new perspectives in tissue engineering.

Coating

As it is possible to tune and control the behaviour of ELP in solution (Hassouneh et al., Reference Hassouneh, Zhulina, Chilkoti and Rubinstein2015), their controlled deposition as surface coating has been explored and tested, notably in the vascular context where the use of uncoated material can result in adverse effects such as restenosis. ELP coatings have been shown to improve the blood compatibility of cardiovascular devices as their presence limits platelets activation (Woodhouse et al., Reference Woodhouse, Klement, Chen, Gorbet, Keeley, Stahl, Fromstein and Bellingham2004; González-Pérez et al., Reference González-Pérez, Acosta, Rütten, Emonts, Kopp, Henke, Bruners, Gries, Rodríguez-Cabello, Jockenhoevel and Fernández-Colino2022). Likewise, the deposition of genetically designed ELPs on the surface of CoCr stents enhanced endothelial cell adhesion and spreading on the metal surface (Castellanos et al., Reference Castellanos, Zenses, Grau, Rodríguez-Cabello, Gil, Manero and Pegueroles2015).

Targeting and delivery

ELP-tag have been shown to be very efficient for delivery and targeting (Jenkins et al., Reference Jenkins, Milligan and Chilkoti2021). In particular, Meyer et al. (Reference Meyer, Kong, Dewhirst, Zalutsky and Chilkoti2001) designed an ELP drug carrier that would coacervate at the temperature found in solid tumours, thereby targeting the drug preferably to the tumour site. Likewise, when ELP are fused with a cell-penetrating peptide, they improve the targeting and delivery of anti-tumour drugs (Walker et al., Reference Walker, Perkins, Kratz and Raucher2012). In this context, the nanoparticle behaviour of ELPs and their versatility have found numerous applications. For instance, they have been used for intra-vitreal delivery to prevent age-related macular degeneration (Sreekumar et al., Reference Sreekumar, Li, Wang, Spee, Hinton, Kannan and MacKay2018) and as vectors for the efficient delivery of cytokines in various pathological contexts (Gong et al., Reference Gong, Yang, Zhang and Gao2022). Recently, a machine learning approach has been proposed (Cobb et al., Reference Cobb, Engel, Seale and Janorkar2021) to control the size of ELP aggregates and their stability to optimise delivery.