Interfacial assembly of the protein at the water‒air interface
With respect to the model, the surface assembly properties of the proteins were studied. Although the concentration is known to play a role in the interfacial properties of proteins, given the high complexity of the system and for simplicity, this work used only a single concentration, ca. 6 mg/mL, and studied the effect of pH on protein assembly at the water‒air interface. Therefore, the first experiment was to determine the adsorption behaviour of the proteins at the water‒air interface and the effect of pH. To do this, interfacial surface tension (IFT) analysis was performed using a Wilhelmy plate. In this experiment, the force that a platinum plate is subjected to when wetted with a liquid is related to surface tension Eq. 1:
\({\gamma }=\frac{F}{L\text{cos}\theta }\) | Equation 1 |
where \({\gamma }\)is the interfacial surface tension, L is the length of the plate, and \(\theta\) is the wetting angle. The advantage of this system is that it allows for the measurement of adsorption kinetics on long timescales without suffering as much from evaporative effects. Inspired by the natural pH gradient within the insect and our own observations, only pH values of 8, 7 and 6 were used. Figure S3 shows the results from these experiments. In all the cases, very rapid adsorption was observed, with all the experiments showing a reduction in IFT even at the beginning of the experiment; the IFT of pure water was 72 mN/m. Despite the similar behaviors of the samples at pH 6 and 7, the protein absorbs to the interface much more quickly at high pH, with an apparent lower IFT value. Notably, the observed values agree with the RSF values observed at similar concentrations.13
To some extent, these experiments verify the observations of DLS,4 where the proteins had smaller diameters; hence, faster diffusion at higher pH values occurred. This indicates that the protein at pH 8 has a faster adsorption to the interface due to faster diffusion speeds under these conditions. Nevertheless, the difference in the IFT itself is more nuanced. It is possible to argue that there might be changes in the total accessible surface area of the protein, whereby at pH 8, the protein exists as a monomeric unit, with a greater likelihood of forming a different type of oligomeric unit as the pH decreases. Oligomerization or aggregation would then reduce the total accessible surface area of the protein. Because the concentration was constant, such a reduction in the IFT can be described by Eq. 2 in the SI.17 Further discussion and observations can be found in ST 2, describing more in-depth observations of a solid-to-solid transformation driven by stress; see Video 2 and Figures S4 and S5.
To characterize the morphology of the protein film at the water‒air interface, we used confocal laser fluorescence microscopy (CLFM). Here, intrinsic fluorescence was used with no labeling to avoid changes in assembly behaviour by incorporating fluorescent labels, which commonly react with free primary amines or carboxylate groups (i.e., Lys or Asp/Glu residues), which are present only at terminal domains or linker units. This finding is particularly relevant given the important role of the NTD in driving assembly. In these experiments, approximately 10 µL of solution equilibrated at either pH 8 or 6 was placed in individual wells and left to age for approximately 5 h before observing the air/water interface using a long working distance objective (x10). The results are shown in Fig. 2A and B. Despite the films being of similar thickness (10–20 µm, as shown in Fig. 2C), the relative fluorescence intensity of the protein at pH 6 was greater than that at other pH values, which could indicate an aggregation-induced emission (AIE) process.18 Freely rotating Tyr (mainly) might be locked in a single rotameric state, enhancing the fluorescence of these multimers/aggregates against the protein in solution. These observations support our hypothesis of a higher degree of order/oligomerization at lower pH. However, a high degree of order is also imparted by the geometrical restriction of the interface itself and the elongated nature of the protein, which also produces a significant signal at higher pH.
To further characterize the mechanics of the films, they were subjected to dynamic microindentation measurements. A small indenter was gently placed in contact with the film and later oscillated, with small oscillation amplitudes, at a range of frequencies (0.1–10 Hz). Using this method, we determined the viscoelastic response in terms of the separate elastic and viscous components of the same material. The results are shown in Fig. 2D-F, with the sample pH shown in the right lower corner. In these experiments, the measurements are reported as stiffness (K) against frequency, with stiffness calculated as the measured force divided by the indentation depth, or half the oscillation amplitude. K’ and K” refer to the elastic and viscous components, respectively. As expected, at pH 6, the material exhibited a relatively higher K’ (160 ± 1 mN/m) than did its counterparts at pH 7 or 8 (60 ± 2 and 60 ± 3 mN/m, respectively); these two pH values exhibited little difference. Once again, we propose that the switch-like behavior occurs at pH values just below 7.4 At lower pH, the nature of the protein in solution is likely oligomeric, driven by NTD interactions; hence, the effective MW of the system increases, increasing the total cohesiveness, which translates to a higher modulus.
Biomimetic silk-like fibre fabrication and characterization
Taken together, these observations indicate that an elastic yet dynamic film forms very rapidly at the water/air interface, and upon extension of the strain/stress applied by pulling, it is possible to assemble insoluble silk-like fibres. For simplicity and because the film was more resilient at pH 6, this condition was used and exploited to produce multimeter long single fibres by continuously pulling the formed fibre using the method depicted in Video 3 and simplified in Fig. 3A. Simply by depositing small droplets onto standard plastic petri dishes, it was possible to pull fibres at different reeling speeds. The achieved reeling speeds ranged from 1.8 mm/s to approximately 53 mm/s. However, the fibre formation was less stable at higher speeds. We hypothesize that there is competition between film formation and protein depletion through the formation of fibres. However, the coverage range contains the natural silkworm spinning speeds estimated to be between 10 and 30 mm/s.19
Using this method, from droplets between 50 and 100 µL (Fig. 3B), it was possible to collect several meters of single fibres (Fig. 3C). All the fibres produced presented a highly hierarchical morphology, with the main fibre formed by bundles of submicron fibres composed of smaller nanofibrils, as shown in Figure S2. The results obtained by SEM corroborate these observations and are summarized in Fig. 4A-G. The fibrillar network extends far from the location where the mature fibre emerges, as outlined in Fig. 4A, with the experimental observations shown in Fig. 4B. We were able to observe individual nanofibrils without much evidence of branching. In other words, the smaller observable fibrils seem to grow in one dimension and only interact by relatively weaker lateral interactions to form larger bundles. These observations are not only in line with the current understanding of natural silk fibre,20 but also predicted from our proposed fibre assembly model; Tyr residues limit the lateral docking of strands, much like the lateral docking of β-solenoids, facilitating the nanofibrillar interface but also facilitating other types of interactions, such as π-π or methyl-π interactions. It has been observed that native silk fibroin fibres can be exfoliated into increasingly thinner fibrils, ranging from 20–100 nm bundles down to 3.1 ± 0.8 nm and ultimately into an extended chain with a diameter of approximately 3.7 ± 0.9 Å.20
In the interest of understanding the biological relevance of our method in the context of in vivo fibre production, we measured the forces that were required to pull these fibres. At steady-state. We measured a force of approximately 0.45 ± 0.04 mN at a reeling speed of 15.5 mm/s, well within the forces that the insect can exert.21,22 No significant differences in the forces were observed at the different speeds (Figure S6 A), indicating that by varying the reeling speeds, we were more likely to change the strain rate rather than the overall stress. Further chemical analysis of the fibres, beyond the nanoscopic similarity with natural silk fibres, showed a remarkable resemblance of the molecular structure, with the amide-I peak showing similar nominal β-sheet, β-turn and statistical coil compositions (Fig. 4D). Thus, soluble silk fibroin (silk-I) was transformed to the well-known insoluble conformation (silk-II) by just using pH control and mechanical stimulation. Notably, no significant differences were observed in the FTIR spectra when the reeling speed was varied.
To further understand the effect of reeling speed at the molecular level, we conducted fibre X-ray diffraction (fXRD) experiments on the different fibres immediately after they were produced at different reeling speeds. In silk fibres, the protein adopts an extended chain conformation, wherein the hydrogen bond network is perpendicular to the long axis of the fibre. However, this conformation is drastically different from that of our Silk-I model, which has a β-solenoid structure. In the proposed fibre formation process, the solenoids would be aligned parallel to the water/air interface plane, and upon drawing, these would align with the long axis of the created filament. The solenoids are denatured/stretched upon application of a critical strain rate and stress, and the known Silk-II configuration emerges. Hence, we would expect to observe a transition from a dominating Silk-I structure to an increasingly more dominating and aligned Silk-II structure upon increasing the reeling speed. Indeed, this was observed going from a reeling speed of approximately 1.5 mm/s to approximately the maximum possible speed of approximately 52.7 mm/s. As shown in Fig. 5A, there is a reflection corresponding to approximately 17 Å at the lowest reeling speeds, which disappears as the reeling speed increases. We interpret this reflection as coming from the hexagonal packing of hydrated solenoid units, which upon reaching a critical strain rate are disrupted, prompting extension of the backbone and collapse of the chains in β-sheets, consistent with the prediction in our previous work.4
Following these experiments, we conducted single-fibre tensile testing, and the results are presented in Fig. 5B-F. Figure 5B shows the average stress‒strain curves obtained for the fibres produced at different reeling speeds; these curves already show noticeable differences that are more detailed in Fig. 5C-F. As one would expect, we observed optimal mechanical properties at a reeling speed that corresponded to the maximum natural spinning speeds (ca. 30 mm/s), with the mechanical properties decreasing thereafter. Briefly, the elastic modulus showed no significant differences from the lowest reeling speed of 1.8 mm/s up to 15.9 mm/s, with values ranging between 4 and 5 GPa; however, from 21 to 32.2 mm/s, we observed an increased modulus, with three increasing between 8 and 10 GPa, decreasing back to 6 ± 1 GPa for the highest reeling speed (52.7 mm/s); see Fig. 5C. The values of tensile strength followed a similar trend, with strength increasing for reeling speeds between 21 and 26.3 mm/s to almost 200 MPa for these speeds and decreasing monotonically as the reeling speed increased thereafter; see Fig. 5D. Interestingly, a slightly different trend was followed by the maximum strain and toughness values, where these values showed an obvious second maxima at the minimal reeling speed (1.8 mm/s). The maximum extensibility was approximately 6% for both 1.8 and 5.9 mm/s, decreasing to approximately 3% for reeling speeds from 10.7 to 21 mm/s, after which the extensibility increased again to 6 ± 4% for the fibres produced at 26 mm/s, decreasing to approximately 3% for any of the faster reeling speeds (see Fig. 5E). Similarly, the same trend was observed for toughness (see Fig. 5F). Interestingly, the observed trends are different from the fibre diameter trends, which only showed a monotonic decrease from 6 ± 1 µm to 3.7 ± 0.2 µm for fibres produced at 1.8 and 52.7 mm/s (see Figure S6 B). Notably, the best performance of our fibres replicates the properties shown by degummed cocoon fibres to a great extent, as reported in the literature.23 Moreover, the overall results are readily explained by our fXRD data and indeed support our model. At the lowest reeling speeds, the relatively higher content of the Silk-I conformation would enable greater extensibility (and toughness) without necessarily enhancing other properties, such as the elastic modulus or tensile strength, as lower strain rates would also imply a lower orientation of the formed β-sheet crystallites. On the other hand, the observation of the overall properties showing a maximum at the peak of natural spinning speeds might indicate the formation of an optimal protein network architecture, where the balance of crystallite size, distribution and orientation maximizes the properties of the material, particularly its toughness. Although our system might not fully recapitulate the in vivo system, mainly due to concentration differences, it is notable that we observed an optimum at natural spinning speeds.
Our proposed fibre formation mechanism involves two separate steps: first, the oligomerization of NTD driven by pH reduction; second, the reconfiguration of the network and further denaturation of the protein fold driven by stress. In this sense, the oligomerization of NTD effectively reduces the degrees of freedom of the protein while also breaking the symmetry of the system. This first step promotes network formation and primes the protein for stress-driven assembly. The system we propose here recapitulates the two-step system, first by breaking the symmetry of the system by exposing it to an interface, becoming inherently asymmetric, and further driving the assembly by stress. Here, in purely aqueous solutions without any precipitants (salts or solvents) and by using high-quality protein feedstock, we were able to replicate the Silk-I to Silk-II transformation using relative speeds and forces that are accessible to the animal.
In addition to the fundamental implications of our observations, the presented method represents a facile, biomimetic process that allows for the easy and efficient fabrication of silk-like fibres. Its simplicity is amenable to the fabrication of several composite materials. For example, multiple fibres can be pulled simultaneously (Fig. 6A) while moving the platform along the collector/mandrel to fabricate nonwoven mats (see Video 4 and Figure S7 for further explanation of the principle). Using standard ideas for the fabrication of orthotopically modified materials, multiple fibrous mats approximately 2 × 4 cm in length were manufactured using approximately 500 µL of the protein solution (Fig. 6B).
Moreover, the process allows for the facile bottom-up incorporation of functionality into the produced fibres. In the past, functional silk fibres were formed by either modifying the surface chemistry of natural fibres24 or directly feeding insects functional particles.25 Here, by simply introducing functional particles into a low-viscosity protein solution, we were able to fabricate magnetic silk-like fibres with incorporated magnetite nanoparticles (Fig. 6C and Figure S8), as well as living composites with Escherichia coli (E. coli) transformed with green fluorescence protein (GFP) for use as fluorescent stimuli-responsive fibres. Despite the relatively large size of E. coli, these bacteria were readily homogeneously incorporated (Fig. 6D and E), with strong evidence of preferential alignment of the elongated bacteria in the same direction as the fibrils (Video 5), as shown by the directional analysis of the images in Figure S9. We believe that this orientation is induced mainly by the confinement offered by the hierarchical nanofibrillar morphology of the fibres and not necessarily by the flow, as the forces used here are low. We are now attempting to further optimize and fabricate devices.