Molecular-level insights into the surface-induced assembly of functional bacterial amyloid

Abstract

Protein coating material is important in many technological fields. The interaction between carbon nanomaterial and protein is especially interesting since it makes the development of novel hybrid materials possible. Functional bacterial amyloid (FuBA) is promising as a coating material because of its desirable features, such as well-defined molecular structure, robustness against harsh conditions, and easily engineerable functionality. Here, we report the systematic assembly of the functional amyloid protein, CsgA, from Escherichia coli (E. coli) on graphite. We characterize the assemblies using scanning tunneling microscopy (STM) and show that CsgA forms assemblies according to systematic patterns, dictated by the graphite lattice. In addition, we show that graphite flakes induce the fibrillization of CsgA, in vitro, suggesting a surface-induced conformational change of CsgA facilitated by the graphite lattice. Using coarse-grained molecular dynamics simulations, we model the adhesion and lamellar formation of a CsgA-derived peptide and conclude that peptides are adsorbed both as monomers and smaller aggregates leading initially to unordered graphite-bound aggregates, which are followed by rearrangement into lamellar structures. Finally, we show that CsgA-derived peptides can be immobilized in very systematic assemblies and their molecular orientation can be tuned using a small chaperone-like molecule. Our findings have implications for the development of FuBA-based biosensors, catalysts, and other technologies requiring well-defined protein assemblies on graphite.

Graphical abstract

Significance

Amyloids are protein aggregates with a well-defined cross-β structure and are often found to form as a result of misfolding, leading to human disease. In contrast, functional bacterial amyloids are evolutionarily evolved to self-assemble into fibrils that play functional roles for the bacteria by strengthening biofilm and increasing bacterial virulence. CsgA may be a promising bio-nanomaterial due to its high tolerance for protein engineering and modifications; e.g., by the addition of fusion proteins. Here we demonstrate the potential use of CsgA to form a hybrid bio-nanomaterial with graphite where CsgA forms highly systematic assemblies, dictated by the graphite lattice. We believe our findings will influence the development of amyloid-based biosensors, catalysts, and other technologies involving systematic protein assemblies on graphite.

Introduction

Protein-based surface modification is important in technological fields such as biomedicine, bioengineering, catalysis, and industrial equipment (1,2). Protein coating of carbon nanomaterials (CNs) such as graphite, graphene, and carbon nanotubes, is an important field of study due to its relevance in many fields, including nanotechnology and medical sciences (1, 2, 3). CNs have multiple desirable properties, including mechanical uniformity, strength, and electrical conductivity (4, 5, 6). Given the versatility of proteins’ binding and coating properties, many new applications for hybrid CN-protein materials can be expected. CN can be solvated by proteins, and proteins can be immobilized by CN, producing a combined product that cannot be achieved with the two materials alone (4, 5, 6). Further, CN has been shown to induce a conformational change of peptides from ⍺-helical and random-coil secondary structure to β sheet secondary structure (4,7, 8, 9), indicating that CN can alter protein properties. Covering CNs with proteins capable of self-assembling into systematic structures is especially interesting as it allows the construction of consistent and well-organized structures (9,10).

In recent years, we have characterized a group of proteins called functional bacterial amyloids (FuBAs). These proteins are remarkably efficient at self-assembling into amyloid fibers with β sheet secondary structure (so called cross-β structure) (11). Functional amyloids are especially common in bacteria and often form structural components of biofilms, enhancing strength, stiffness, and adhesive properties (12, 13, 14, 15). The most well-characterized FuBA is curli from Escherichia coli (12). CsgA is the main structural component of curli, but curli production requires six more assisting proteins transcribed from two operons (csgBAC and csgaDEFG) (16). These include a transcription regulator (CsgD), a nucleator protein (CsgB) (17,18), chaperones (CsgC, CsgE, and CsgF) (19, 20, 21), and an outer membrane pore protein (CsgG) (22).

CsgA can self-assemble under a wide range of conditions (23), including high concentrations of denaturant (24) and are remarkably stable once formed (25). Given the extreme robustness and strong assembling capability of CsgA, the protein holds potential use as a nanomaterial. In addition, CsgA effectively binds to a wide variety of different surfaces (26,27) to provide new functionalities, whether used as a purified component or produced in biofilm (27, 28, 29, 30, 31, 32). Purified CsgA shows very robust surface interaction and the resulting coats withstand harsh conditions (27). Further, CsgA is readily engineered for functionalization, opening up for, e.g., SpyCatcher-SpyTag linker approaches (27).

CsgA contains five imperfect repeat sequences, each of which is predicted to form a β hairpin (Fig. 1) (33, 34, 35, 36). Each repeat sequence consists of about 22 residues and is connected by β turns of four or five residues (Fig. 1 b) (35,36). The repeat sequences are believed to stack on top of each other to form a β helix (Fig. 1 a) (33,36). This structure of CsgA has been predicted from homology modeling (33,37) combined with molecular dynamics (MD) (36) and is supported by experimental NMR data (38). Monomeric CsgA is an intrinsically disordered protein (IDP) in vitro and only folds upon fibril formation (39, 40, 41).

Figure 1.

(a) CsgA consist of five repeat sequences (R1–R5), which are each predicted to form β hairpins that stack on top of each other, which results in an overall β-helical structure. (b) The sequence of CsgA contains an N22 signal sequence, which is involved in transport across the outer membrane in E. coli. From residues 23–131, five repeat sequences (R1–R5) are found. The seven vertical bars mark ladders formed from conserved residues stacked on top of each other. (c) The top view of the CsgA β-helical structure shows how conserved residues in R1–R5 stack on top of each other. Reprint from (37).

Although surface adsorption of CsgA has been investigated before, little is known about the interactions between CsgA and the underlying substrate, nor about the architecture of the coating layer, except when they have been mechanically applied to surfaces in specific patterns (32). Due to the systematic and efficient self-assembly of CsgA in solution, we hypothesize that even more systematic assemblies can be made from CsgA by combining the protein with a highly systematic surface material. Highly oriented pyrolytic graphite (HOPG) is a well-defined and atomically flat material with a surface that is easily renewable by cleavage and removal of the top graphite layers (42). These features make HOPG particularly suitable for scanning probe microscopy (42). Graphite (including HOPG) consists of multiple layers of graphene with sp² bonded carbon atoms arranged in a planar hexagonal lattice. Thanks to the hexagonal structure of graphite, three equivalent directions exist in the plane of the graphite surface (43). Graphite-adsorbed molecules can be oriented in distinct directions (1,44), indicating that the underlying graphite lattice influences the orientation of molecules. Indeed, peptides have been shown to self-assemble in a register that perfectly matches the underlying graphite lattice (45).

Scanning probe microscopy (SPM), represented by scanning tunneling microscopy (STM) and atomic force microscopy (AFM), has proved to be a powerful tool for studying organic and biological species at molecular or submolecular levels due to its high structural resolution and adaptability to various environments including ultra-high-vacuum (UHV) conditions and liquid-solid interfaces. Remarkable advances of STM and AFM have demonstrated atomic resolutions in UHV conditions (46, 47, 48, 49). Particularly, STM has remained the preferential experimental approach in the studies at liquid-solid interfaces prepared under liquid and ambient conditions for pursuing molecular or sub-molecular-level resolutions (10,50, 51, 52). These advances demonstrate the feasibility of using STM to study amyloidal peptides with high structural resolution and provide complementary insights to structural analysis by other methods.

Here we study the self-association behavior of CsgA on highly oriented pyrolytic graphite using STM, and show that CsgA assembles in highly systematic patterns, whose general orientation likely reflects the organization of the graphite lattice. In addition, we show that graphite flakes (GFs) accelerate aggregation of CsgA in vitro. Further, we use coarse-grained MD to simulate the dynamic process of adhesion and lamellar formation of a CsgA-derived peptide. Finally, we show that CsgA-derived peptide assemblies reach a higher level of order in the presence of the small chaperone-like molecule, 4,4′-bipyridyl (4BPY), which aligns the C termini of different CsgA molecules against each other. These results pave the way for FuBA-based engineered protein-surface biosensors, catalysts, or similar technologies requiring well-defined protein assemblies. To our knowledge, this is the first time that the self-assembly of a full-length protein (and a FuBA protein) has been investigated using STM; previously only peptides have been investigated. This has given us unique insight into how an amyloidogenic protein self-assembles on a graphite surface, with the particularly high resolution that STM allows.

Materials and methods

Materials

Recombinant CsgA was expressed and purified as described previously (24). R3, R4, and R5 peptides were purchased from Genscript (Piscataway, NJ) with a purity of 98%. The peptide sequences are SSIDLTQRGFGNSATLDQWNGKN (R3), SEMTVKQFGGGN-GAAVDQTASN (R4), and SSVNVTQVGFGNNATAHQY (R5). 4BPY and GFs (graphene nanoplatelets) were purchased from Sigma-Aldrich (St. Louis, MO).

Sample preparation for STM

Full-length CsgA (155 μM) in denaturing storage buffer (8 M GdmCl, 50 mM Tris-HCl, 500 mM imidazole, pH 8) was mixed 1:1000 with 4BPY (6.4 mM in Milli Q water) to a final concentration of 0.155 μM CsgA, 8 mM GdmCl, and 6.4 mM 4BPY. Peptides (0.5 mM) were mixed with 4BPY (6.4 mM in Milli Q water) to a final concentration of 0.5 μM peptide and 6.4 mM 4BPY. A volume of 20 μL of each solution was deposited on each of their freshly cleaved HOPG surfaces, and the solutions were allowed to evaporate at room temperature (r.t.) prior to STM analysis.

STM experiments

A Nanoscope IIIA scanning probe microscope system (Bruker, Billerica, MA) was used to perform STM experiments in constant-current mode under ambient conditions at tunneling conditions I = 400 pA, V = 600 mV. Newly mechanically cut Pt/Ir wires (80/20) were used as tips. The operation of STM and the Nanoscope system has been explained elsewhere (53,54). All samples were studied in triplicate.

Statistical analysis of STM images

STM images were analyzed using Gwyddion version 2.59 to determine the peptide strand lengths and distance between strands. To calculate the number of residues in each strand, a step size of 0.325 nm was assumed for each residue. The average distance between β strands was determined by measuring the distance across 10 strands at 110 different positions. A Gaussian distribution was used to fit the statistical histogram of the length distribution.

Fibrillization experiments in vitro

CsgA in denaturing storage buffer (8 M GdmCl, 50 mM Tris-HCl, 500 mM imidazole, pH 8) was buffer exchanged into 50 mM Tris-HCl, pH 7.4 using PD-10 desalting columns (GE Healthcare Life Sciences, Brøndby, Denmark) according to the manufacturer’s protocol. The protein concentration was estimated by measuring the absorbance at 280 nm (ε = 11,460 M⁻¹cm⁻¹ or 0.822 mg/ml⁻¹ cm⁻¹, MW = 13.9) using a NanoDrop 1000 (ND-1000 Spectrophotometer, Scientific, Waltham, MA). GFs were resuspended in MilliQ water to a stock concentration of 0.5 mg/mL, which was mixed with the CsgA solution to obtain a final concentration of 0–0.2 mg/mL and 0.3 mg/mL of GFs and CsgA, respectively. A concentration of 1.2 mM Thioflavin T (ThT) in 50 mM Tris-HCl, pH 7.4, was added to obtain a final concentration of 40 μM. Samples were added to a Corning 96-well half-area low-bind microplate (Corning, NY). Fibrillization was monitored at 37°C under quiescent conditions in a FLUOstar Omega plate reader (BMG Labtech, Ortenberg, Germany) with excitation at 448 nm and emission at 485 nm at a gain of 900. Measurements were made every 5 min. Half-times were estimated using KaleidaGraph (Synergy Software, Paramus, NJ).

Fourier transform infrared spectroscopy

Fourier transform infrared spectroscopy (FTIR) analysis was performed on R5 (1.5 mg/mL) with and without 4BPY (1 mg/mL) after 1 h of incubation at r.t. in MilliQ water. The instrument used was a Tensor 27 (Bruker, Billerica, MA) with a reflection diamond attenuated total reflectance (ATR) cell. A sample with a volume of 2 μL was deposited on the ATR cell and dried using N₂ gas. Spectra were collected at 4000–1000 cm⁻¹ with 64 accumulations. Baseline correction and normalization were performed using the associated OPUS 5.5 system.

Circular dichroism

R5 peptide (0.3 mg/mL) with and without GFs (1 mg/mL) and/or 4BPY (0.1 mg/mL) was incubated in 10 mM Tris-HCl, pH 7.4, for 1 h at r.t. Circular dichroism (CD) spectra were then recorded from 280 to 190 nm on a Chirascan circular dichroism spectrometer (Applied Photophysics, Leatherhead, UK) in steps of 1 nm and 1-nm bandwidth using a 1-mm path length. The temperature was kept constant at 25°C during acquisition. Three scans were recorded and averaged for each sample.

Coarse-grained MD simulations and analysis

Simulations of CsgA peptide adsorption on graphite were performed using the Gromacs 2016 package (55,56) and a modified version of the Martini CG force field (57, 58, 59). The bonded and nonbonded parameters of graphite were taken from previous studies by Gobbo et al. (57) and Piskorz et al. (58). The initial state of peptides was created by coarse-graining of atomistic peptides build in Pymol using the martinize.py script (60). Β strand secondary structure was assigned to all residues of the LNIYQY peptide during coarse-graining. The GF with a SASA of 80 nm² was placed in a cubic box together with 25 randomly positioned peptides with a minimum distance of 4 Å, reaching a final peptide concentration of 12 mM. The box was solvated using ∼23,000 Martini water molecules and ∼4000 Martini antifreeze water molecules to avoid freezing of water at the graphite surface (59,61,62). The simulation details with respect to temperature and pressure coupling were similar, as described by Piskorz et al. (58) with the exception that simulations were run at 323 K. The simulation runtime was 1.5 μs and simulations were run in six replicas. The burial of graphite surface area was determined for each simulation frame by rolling a spherical probe of 1.4 Å around the GF using the SASA.tcl script of VMD (63). The angle of each β strands relative to the graphite lattice direction was determined in the plane of the GF using the following procedure: a reference direction was chosen from one of the three directions of the graphite lattice, and the angle of all strands was measured relative to this reference. The angle of each strand was calculated from the principal axis of inertia of the backbone beads in each strand, which was determined using the measure inertia command of VMD (63). The radial pair distribution function plugin of VMD (63) was used to analyze the intermolecular distances of Ile 3–Ile 3 backbone beads of the LNIYQY peptide. Simulations were visualized using Pymol version 1.9 (Schrödinger, New York, NY). Bonds were drawn between backbone beads using Pymol and sidechain and water beads were hidden in simulations snapshots for clarity.

Results

CsgA forms β strand-rich assemblies on graphite

To investigate how CsgA self-organized on a well-defined molecularly flat surface, CsgA (expressed recombinantly in E. coli and subsequently stored in the monomeric state in high concentrations of denaturant) was diluted 1:1000 into water and applied in solution to a surface of HOPG. CsgA was allowed to self-assemble while it slowly dried, and was then analyzed by STM. This revealed large (>10,000 nm²) film-like CsgA assemblies (Fig. 2 a). The CsgA assemblies are visible as parallel lamellar stripes ∼4 nm in width, separated by narrower darker stripes. The lighter lamellar stripes can be attributed to the β sheet secondary structure of packed CsgA assemblies, while the darker areas are likely random-coil structure, which is expected to be less conductive because of less contact with the HOPG surface. Individual β strands are clearly visible in higher-resolution images (marked with white bars in Fig. 2 b). We analyzed the topography of CsgA assemblies as indicated by the white bar (profile 1) in Fig. 3 a along the cross-β axis and plotted the resulting profile (Fig. 3 b). In the profile, we see distinct peaks separated by ∼4.8 Å. We attribute these peaks to the backbone region of β sheets. We measured the distance between β strands in multiple regions of the assemblies and the average distance between β strands was determined to 4.8 Å (±0.1 SD) (Fig. 2 c), which is in good agreement with the canonical inter-strand distance seen from, e.g., fiber diffraction (64). The average β strand length was 3.8 nm (±0.7 SD), with a length distribution from 2.275 to 5.2 nm (Fig. 2 d). This corresponds to seven to 16 residues (given 0.325 nm/residue in the extended conformation (65)) with 11–12 as the most prevalent number of residues per β strand.

Figure 2.

(a) Large-scale STM image of CsgA FL on HOPG. The outline of a light lamellar stripe is marked with white bars. Scale bar, 10 nm. (b) Higher-resolution STM image of CsgA FL on HOPG. The five lower-left white bars and the upper-right bar mark five adjacent β strands and the distance measured over 10 strands (∼4.8 Å), respectively. Scale bar, 5 nm. (c) Histogram of distances between β strands fitted with a Gaussian curve (n = 110). (d) Histogram of the length distribution of CsgA FL β strands fitted with a Gaussian curve (n = 350). (e) Autocorrelation plot of the length of adjacent CsgA FL β strands. The red line is the exponential decay fit.

Figure 3.

(a) Analysis of topographical features of CsgA FL assemblies. The topography was analyzed along the white bar in the lower-left corner of the main image (profile 2). Inset: zoomed view of the marked area. The topography was analyzed along the white bar in the inset (profile 1). Scale bar, 5 nm, inset scale bar, 1 nm. (b) Cross-sectional profile of the topography along the white line in the upper-right part of (a). The black bar indicates a distance of 4.8 Å. (c) Cross-sectional profile of the topography along the white line in the lower-left part of (a). (d) Proposed model of the CsgA FL assemblies on HOPG.

In solution, CsgA is initially an IDP which only gains structure upon self-association (23). We queried whether the assemblies formed on graphite share structural similarities with the β-helical structure predicted to form in CsgA fibrils in vitro (36). As a first step in our structural analysis, we examined whether there were any internal correlations in strand lengths. We first measured the length of a large number of sequential strands (Fig. S1 a). For each strand, we then analyzed how its length correlated with neighboring strands at increasing distance to it. We determined the Pearson correlation coefficient (Fig. S1 b and c) and plotted the results in Fig. 2 e. To determine over how many strands the length correlation exists, we fitted a single exponential decay to our autocorrelation curve. Our analysis showed that the decay length is 1/0.21 or 4.8 ± 0.4 strands and that the length of the strands is, therefore, correlated over about five sequential strands. This analysis showed that there are systematic clusters of similar strand lengths in the CsgA assemblies. Next, we analyzed the topography across three cross-β backbones in the orthogonal direction relative to the β strands (profile 2 in Fig. 3 a). There is a distinct height profile, originating from the cross-sectional profile of three β strands giving rise to three ∼4-nm broad peaks (Fig. 3 c). However, these peaks are not horizontal but show a very modest slope-like topography where one end of each β strand is at a slightly higher level than the other end (Figs. 3 c and S2). We suggest the structural model of CsgA assemblies on graphite shown in Fig. 3 d, where fibrils are positioned side by side in a systematic pattern to form a film-like amyloid surface.

The graphite surface controls the direction of CsgA β strands

Our next step was to investigate whether the graphite lattice influenced the orientations of the CsgA assemblies. This was done by searching for CsgA assemblies with different directions on the HOPG surface (Fig. 4 a). The perpendicular angle of the cross-β axis was measured at many positions and plotted. We found that three different directions of the β strands existed within the observed area which we denoted 60°, 0°, and −60° (Fig. 4 a and b). This observation correlated well with the three expected axes in the plane of the hexagonal HOPG lattice (i.e., the three a axes of the four-axis Miller-Bravais system of hexagonal crystals (66)). This indicates that the graphite lattice guides the orientation of the CsgA assemblies, possibly by stabilizing peptide-graphite interactions at preferred angles. Our images contained a somewhat higher number of strands with an angle at −60° compared with strands with an angle at 60°, 0°. However, it is not expected that this angle is preferred over the others, since the graphite lattice contains a three-axial symmetry. We suggest two different types of β strand-graphite lattice interactions that can lead to β strand assemblies formed in three different directions with strand angles at 60°, 0°, and −60° (Fig. S3). In the proposed assembly type A (Fig. S3 a), the β strands lie orthogonal to the $\left[01\overline{1}0\right]$, $\left[\overline{1}010\right]$, and $\left[\overline{1}100\right]$ directions (i.e., β strands are parallel to the sides of the hexagon). This would lead to a theoretical β strand distance of 4.9 Å, which is close to our experimentally determined strand β distance of 4.8 Å. In assembly type B, the β strands lie orthogonal to the $\left[11\overline{2}0\right]$, $\left[\overline{2}110\right]$, and $\left[1\overline{2}10\right]$ directions, leading to a theoretical β strand distance of 4.3 Å. Assembly type B has been experimentally identified in HOPG-adsorbed assemblies formed from a peptide segment (hIAPP_20–29) of human islet amyloid peptide (hIAPP) (45). We believe that assembly type A is the most likely for CsgA assemblies because of the similar values of the theoretical and experimentally determined β distances. However, to unambiguously determine the orientation of the CsgA assemblies with respect to the graphite surface, it would be necessary to obtain STM images where both the protein assemblies and the underlying graphite surface were imaged with atomic resolution in the same area. Despite obtaining many images of the lamellas, we were unable to obtain clear images of the border area between the lamellas and the underlying graphite substrate. In our two suggested models, we have hypothesized commensurability between the β strands of CsgA and the graphite lattice. Although commensurability between amyloidogenic β strands of hIAPP and graphite lattice has been demonstrated previously, it should be noted that this is not necessarily the case for CsgA deposited on graphite (45).

Figure 4.

(a) Large-scale STM image showing the three directions of CsgA FL assemblies. The three different directions of β strands are indicated as 60°, −60°, and 0° (marked 1, 2, and 3). The angles were measured as illustrated in the inset in the upper-right corner. All angles were measured relative to the 0° axis, which is shown as a solid black arrow in the inset. Scale bar, 20 nm. (b) Histogram of estimated β strand angles showing three different angles of CsgA FL assemblies on HOPG (n = 30). (c) CsgA FL fibrillization curves in vitro with and without 0.2 mg/mL graphite flakes (GFs). Error bares are standard deviation (n = 3). (d) Half-times of CsgA fibrillization as a function of [GF]. Error bares are standard deviation (n = 3).

GFs induce the fibrillization of CsgA in solution

STM provides end-point images of the aggregation-absorption process but does not allow us to follow real-time kinetics of aggregation. However, the impact of a graphite surface on aggregation kinetics can be followed kinetically using GFs. We, therefore, decided to investigate whether GF could promote the fibrillization of CsgA in vitro. The GF are only a few nanometers thick, consist of relatively few stacked graphene layers, and have a diameter of up to 2 μm. As a result, the GF, like HOPG, have a large surface area consisting of the planar hexagonal graphite lattice. CsgA was mixed with increasing concentrations of GF, and the rate of the fibrillization was monitored by measuring the fluorescence of the amyloid-specific fluorophore Thioflavin T (ThT) (Fig. 4 c). It was assessed how GFs influenced the rate of the self-assembly of CsgA by determining the half-time of fibrillization (t_½^{fibrillization}). This parameter specifies the time at which the self-assembly has reached half of its maximum value and is a common method for evaluating the rate of the self-assembly process (67). At concentrations above 0.025 mg/mL, GFs significantly increased the fibrillization rate of CsgA, leading to a decrease in the t_½^{fibrillization} to less than 50% of the GF-free CsgA at 0.125 mg/mL GFs, after which t_½^{fibrillization} stayed constant. These results indicated that GFs promote CsgA self-assembly.

Graphite induces the rate of structure formation of a CsgA backbone segment by facilitating longer-range structure and constriction of β strand angles

We hypothesize that GF immobilizes CsgA monomers and that the graphite lattice assists conformational change of CsgA from the original IDP state to an elongation-ready structure, influenced by the order-inducing orientation of the GF surface. However, our STM analysis does not allow us to monitor the kinetics of self-assembly but only provides an end-point view; conversely, ThT fluorescence provides kinetic information but no structural insight. To gain insight into the dynamics of the adsorption and self-assembly process in high time resolution, we turned to MD using the coarse-grained (CG) Martini model. Since simulations involving self-assembly of multiple full-length (121-residue) CsgA molecules are very computationally demanding, we focused on a CsgA segment LNIYQY from the R1 repeat. The structure of fibrils formed from LNIYQY peptides was recently solved and indicated that the sequence is an important segment of the CsgA fibril spine (68). We simulated the adsorption and self-assembly of LNIYQY peptides on a GF with a solvent-accessible surface area (SASA) of 80 nm². (Fig. 5 a–e). The peptides adsorbed rapidly on the GF within ∼250 ns, leading to a drastic reduction in peptide diffusion as well as coverage of the graphite surface (Fig. 5 f). Peptides were adsorbed both as monomers and as smaller aggregates (Fig. 5 b), leading initially to an unordered surface coverage of peptides (Fig. 5 c), which, after 600–1200 ns, partially matured into well-ordered lamellar structures (Fig. 5 d and e). In parallel, there was a further modest increase in graphite surface burial as the peptides self-organized (Fig. 5 f). During the simulation, we observed a rapid formation of smaller peptide aggregates in solution (Fig. S4 a), which is in agreement with previous observation of the peptides' ability to aggregate in solution (68). Peptide aggregation in solution can be observed in our data from the distinct population of intermolecular ∼5-Å distances between Ile 3-Ile 3 backbone beads at 0–130 ns before any significant peptide adsorption has taken place (Figs. 5 g and S4 a). Interestingly, the population of ∼5-Å backbone distances transiently disappears during the early part of the adsorption plateau phase at 300–450 ns, while a broad peak ∼8 Å simultaneously appears (Fig. 5 g). This indicates that small aggregates formed in solution partially lose structure and become more loosely associated during the early phase after adsorption. The loose peptide association is illustrated in Fig. S4 b, where the majority of intermolecular Ile 3-Ile 3 backbone distances are ∼8 Å or longer. At 450–600 ns, the ∼5-Å peak reappears and two additional peaks at ∼8 and ∼11 Å gradually build up and become well-defined at 750–900 ns, indicating that the unordered adsorbed peptides undergo a rearrangement, leading to ordered assemblies with longer-range systematic distances at ∼5, ∼8, and ∼11 Å. Closer examination of the intermolecular clusters showed that backbone distances of 8 Å (±1 Å) and ∼11 Å (±1 Å) are abundant in lamellar structures (see simulation snapshot Fig. S4 c). In contrast, when peptides were simulated in the absence of the GF, less molecular order was obtained within the duration of the simulation, since the radial distribution plot lacks any well-defined peaks above ∼5-Å distance (Fig. S5).

Figure 5.

Snapshots from the simulation of the CsgA spine segment LNIYQY during the adsorption phase (a, b), rearrangement phase (c) and formation of lamellar structures (d, e). Lamellar structures are colored red and orange. For clarity, only backbone beads are shown and water molecules are not shown. (f) Plot showing the peptide burial of graphite surface area and diffusion coefficient as a function of time. (g) Radial distribution function of backbone beads of Ile 3 of LNIYQY computed during six 150-ns intervals (0–900 ns) in presence of graphite. (h) Histograms of β strand angles relative to the graphite lattice showing three distinct populations of angles at −60°, 0°, and 60°. The histograms of two representative peptides are shown over the duration of the simulation trajectory.

In our STM analysis, we observed confined angles of CsgA β strand at 60°, −60°, and 0°, which suggested that final CsgA assemblies were affected by the direction of the graphite lattice. In light of this observation, we next investigated whether the graphite lattice influenced the dynamics of the self-assembly process of our simulation. First, we analyzed the angles of β strands in the X-Y plane relative to the graphite lattice during the simulation. We observed three distinct populations of β strand angles at −60°, 0°, and 60° (histograms of two representative peptides are shown in Fig. 5 h). During a major part of the simulation, the −60° angle is the most dominant angle taken up by the peptides (Fig. S6), indicating a homogeneous buildup of peptide backbone orientations. This is supported by our STM analysis where clusters of β strands with one particular direction often exist over large areas, containing hundreds of β strands (Figs. 1 and 3). Interestingly, a distinct population of peptides at −60° was observed very early in the simulation at 150–300 ns, which corresponds to the period immediately after peptide adsorption to the graphite surface and before any β sheet-like assemblies were observed (Fig. S6). This indicates that the graphite may act as a template for β sheet formation by forcing β strands to orient and align in certain directions.

The molecular orientation of CsgA peptide assemblies on graphite can be controlled by a co-assembly with a chaperone molecule

The CsgA protein is built from five modular repeating segments that are predicted to stack on top of each other (R1–R5), resulting in a β helix (Fig. 1). As a next step, we investigated the possibility to modulate the CsgA assemblies using a small molecule, 4BPY, which can be detected by STM and has previously been shown to co-assemble with compounds as diverse as stearic acid (69) and peptides (70) on HOPG. 4BPY can hydrogen bond with the carboxyl group of the stearic acid and the C terminus of peptides (69,70). First, we analyzed the potential co-assembly of CsgA FL and 4BPY on HOPG using STM. However, only homogeneous assemblies of CsgA FL without any visible 4BPY were observed (data not shown). We speculate that the failure of 4BPY to tether the C terminus of CsgA FL may be caused by the large size of the CsgA FL compared with 4BPY and that CsgA is expected to form a β helix formed from five stacked β hairpins (Fig. 1). In view of this, we hypothesized that co-assemblies of peptide and 4BPY would be more likely to form if the size of the protein could be reduced while maintaining the sequences predicted to be responsible for the aggregation propensity of CsgA. To achieve this, we focused on the individual repeat sequences of CsgA that are predicted to be the core regions of full-length CsgA (Fig. 1). The terminal repeats R1 and R5 are particularly aggregation prone; R3 shows slightly reduced aggregation propensity, whereas R2 and R4 do not aggregate under experimental time scales in vitro (39). We chose to investigate the assembly of R3, R4, and R5 to have a representative selection of peptides. As anticipated, R3 and R5 formed co-assemblies with 4BPY (Fig. 6 a and c), whereas no assemblies of R4 could be observed. This allowed us to control the orientation of peptides by coupling the C terminus using 4BPY. In the STM images, the bright parallel stripes are 4BPY molecules, whereas the lamellar stripes between the parallel 4BPY stripes are peptides. The brighter appearance of the 4BPY molecules is caused by the conjugated π-electron system, which leads to higher electron conductance. The average β strand length was 3.5 nm (±0.2 SD) and 3.8 nm (±0.2 SD) for R3 and R5 peptides, respectively (Fig. 6 b and d). The distribution of strand lengths for the 4BPY-bound peptides was narrower than the full-length CsgA proteins, indicating a more regular and fine-tuned assembly. The angle (⍺) between the axis formed by the 4BPY stripes and the peptides was 33° (±2 SD) for R3 and 30° (±2 SD) for R5. This angle is similar to peptide-4BPY assemblies described previously (8,44,70, 71, 72, 73, 74). Based on our STM analysis of the 4BPY-bound R3 and R5 peptides, we propose the schematic model shown in Fig. 6 e. To investigate whether graphite and 4BPY influenced the peptide secondary structure, we analyzed R5 in the presence and absence of GFs and 4BPY using circular dichroism (CD) (Fig. 6 f) after 1 h of incubation. All samples showed a significant peak ∼220 nm, which is characteristic of β sheet secondary structure, supporting our observation that R5 forms β sheet secondary structure both in the presence and absence of graphite. The presence of amyloid β sheet secondary structure of R5 with and without 4BPY was supported by Fourier transform infrared spectroscopy (FTIR), which showed the amyloid characteristic peak at 1620 cm⁻¹ of both samples (Fig. 6 g).

Figure 6.

STM image of R3 (a) and R5 (c) co-assembled with 4BPY. ⍺ denotes the angle between the axis formed by the 4BPY stripes and the peptide β strands. Scale bars, 6 nm. (b and d) Histograms of length distribution of β strands in R3 (n = 170) and R5 (n = 110) assemblies, respectively. (e) Cartoon illustrating the co-assemblies of peptide and 4BPY. (f) CD spectrum of R5, R5-4BPY, and R5 in presence of GFs (R5 graphite) and R5 in presence of both GFs and 4BPY (R5 4BPY graphite). (g) FTIR spectrum of R5, R5-4BPY, and 4BPY.

Discussion

FuBA and graphite are well-organized partners

To the best of our knowledge, this is the first reported example of the construction of large (>10,000 nm²) uninterrupted FuBA-based two-dimensional (2D) amyloid films with a structural order not previously seen for surface-adsorbed FuBA. While HOPG-adsorbed assemblies have been reported for shorter pathological peptides like amyloid β (10) (42 residues) and hIAPP (9) (37 residues), this is the first reported example of amyloid films from a full-length amyloidogenic protein with a predicted β-helical structure. Taking the unique features of FuBA into account, such as extreme robustness, engineerability, and effective fibrillization, we expect that these results have many useful ramifications. The CsgA assemblies formed in our study showed highly regular and repetitive structures with restricted β strand directions at 60°, 0°, and −60°, suggesting that the graphite lattice templates the directions of the graphite-adsorbed fibrils. In our MD simulations, we observed that a significant population of peptides was already positioned at one predominant angle before any β sheet-like assemblies. The well-organized structure of FuBA and the regular structure of the graphite surface combine to make the two materials excellent partners for the production of hybrid bionanomaterials.

Structural insights of CsgA assemblies may be important for the design of fibrillization inhibitors

Understanding the structure of CsgA is important, as curli may play a role in pathogenesis and invasion of host cells (35), and the structure-based design of curli inhibitors naturally depends on a well-characterized CsgA structure. In our present work, we gained insight into structural details about CsgA assemblies, including the average β strand length, and found that the length of strands is correlated over at least five sequential β strands. Care must naturally be taken about generalizing protein structure identified on graphite to also exist in solution as exemplified by graphite’s ability to induce conformational change from ⍺-helical or random-coil structure to β sheet secondary structure (4,7, 8, 9). However, this is probably of less concern for CsgA since a conformational change from random-coil structure to β sheet secondary is expected for CsgA in solution, and the resulting β sheet-dominated secondary structure lacks any fibril polymorphism (75) and is remarkably unaffected by highly varying environmental conditions (23). Furthermore, the validity of graphite-adsorbed amyloid structures is also supported by previous examples of pathological amyloids that showed structural similarity between graphite-adsorbed amyloids and their in vitro-formed counterparts (9,10).

Amyloid surface adsorption is important in both bio-nanomaterial and disease

Absorption of peptides onto a surface strongly reduces the peptides’ exploration space from a three-dimensional (3D) volume to a 2D surface. Specifically, in our MD simulations, the peptides explore a volume of 3375 nm³ (i.e. a 15-nm cube) in solution but only a solvent-accessible surface area of ∼80 nm² after adsorption to the GF, leading to a higher local concentration of peptides and localization of peptides into a single plane. The peptides were still able to diffuse in the X-Y plane of the graphite surface after adsorption, indicating that peptides can diffuse along the graphite surface and attach to fibril ends. Considering that the kinetics of amyloid formation are highly sensitive to protein concentration (76), it is not surprising that graphite induces the self-assembly of CsgA. These observations are in agreement with previous studies, indicating that surfaces can induce fibrillization if the protein-surface interactions are favorable (77,78). In graphite, the relatively generic fibrillization-inducing effect of surfaces is probably significantly enhanced by the elongation-inducing effect of the graphite lattice (4). Thus, the elongation-inducing effect of the graphite lattice works in perfect synergy with graphite’s tendency to attract proteins to bind on its surface.

CsgA has previously been shown to adhere to very diverse types of surfaces, including poly(tetrafluoroethylene) (PTFE) and glass (27,30). However, in these previous cases, there has not been any evidence for the formation of film-like CsgA assemblies, nor that the assemblies were restricted to specific orientations with respect to the underlying surface. This indicates that CsgA-graphite assemblies are unique in their formation of systematic film-like structures with specific orientations, suggesting that the graphite lattice guides the orientation of the surface-adsorbed CsgA fibrils.

While this combined effect is ideal for the production of hybrid nanomaterial from graphite and CsgA, it may be a serious problem for the biocompatibility of graphite in the presence of aggregation-prone pathological proteins as discussed elsewhere (4). We speculate that the effective graphite coverage of CsgA is a result of an optimal combination of the protein’s amino acid residue composition and the flexible conformation of CsgA in the IDP state. The importance of flexible peptide conformation in favor of efficient surface coverage has been demonstrated in previous MD simulations (79). Peptide residue composition plays an important role for the kinetics and thermodynamics of surface adsorption, but, remarkably, homopeptides with identical residue compositions but different lengths may obtain different conformations in solution, which eventually leads to differences in the kinetics and thermodynamics of surface adsorption (80). In our in vitro experiments, we observed an induced rate in CsgA fibril formation in the presence of GFs. A possible scenario for this observation could be that CsgA adsorbs and assembles on the graphite surface and induces subsequent folding of CsgA and formation of fibrils as revealed by ThT fluorescence results. Importantly, Minton (78) demonstrated that the extent of both fibrillization and surface adsorption can be dramatically altered by even small changes in the interaction strength between the fibrillizing protein and the surface, leading to a sharp transition between surface-adsorbing and non-adsorbing protein. This phenomenon may open the possibility of engineering CsgA with different graphite-interaction strengths and thereby tailoring CsgA to either form fibrils in solution or form surface-adsorbed assemblies. This may be a real possibility since CsgA can be extensively re-engineered (29,81) and mutated (82) without losing its ability to form fibrils. Interestingly, CsgA can be satisfied by a broad range of surface interactions with both hydrophobic and hydrophilic surfaces (26,27), although simulations showed a slightly more favorable adhesion to graphite compared with mica (26). Nevertheless, because of CsgA’s favorable interaction with a broad range of surfaces, it may be challenging to manipulate the graphite-CsgA interaction through protein engineering.

Chaperone-mediated fine-tuning of CsgA peptide assemblies

In our STM analysis, we observed the formation of co-assemblies between 4BPY and aggregation-prone CsgA peptides. However, no co-assemblies between full-length CsgA and 4BPY were observed. We speculate that the failure of 4BPY to tether the C terminus of full-length CsgA may be caused by the large size of the CsgA compared with 4BPY and that CsgA overrules the preferred interaction of 4BPY with graphite and makes it more favorable to form CsgA-graphite interactions without co-assembling with 4BPY. In the peptide-4BPY assemblies observed involving R3 and R5, the peptide:4BPY stoichiometry is 1:1, which is similar to other peptide-4BPY assemblies, described previously (8,44,70, 71, 72, 73, 74). In these assemblies, peptides are assembled head to head, which also allows 4BPY molecules to be closely positioned, leading to the formation of 4BPY stripes. However, for a β-helical structure like full-length CsgA, where only one 4BPY molecule would bind per five stacks of β strands, the distance between 4BPY molecules would likely become too large to form consecutive stripes. Peptides and 4BPY each have a preferred way to interact with the graphite substrate. The peptide-graphite interaction is dominated by van der Waals interactions, whereas the 4BPY-graphite interactions are π-π dominated (45). Previous studies have shown that, due to the different nature of graphite interaction with peptide and 4BPY, respectively, the peptides and 4BPY molecules in co-assemblies of peptide-4BPY compromise their individual preferred interactions with graphite to allow favorable interaction of the overall peptide-4BPY co-assembly with graphite (45). This causes a twist of the peptide-4BPY assemblies with respect to the graphite surface (45), which may be acceptable for peptides up to a certain size but unfavorable for larger molecules like full-length CsgA protein. In our resulting peptide-4BPY assemblies, the angle between the axis formed from the 4BPY stripes and the β strands of the R3 and R5 peptides is ∼30°, which is similar to other peptide-4BPY assemblies, described previously (8,44,70, 71, 72, 73, 74). Because of this angle, the overall assembly direction is not orthogonal to the β strands as in the full-length CsgA assemblies, and, as a result, the peptides are slightly out of register. Such a slightly out-of-register-type assembly is acceptable for peptides that can assemble head to head since the majority of their β strands will still be able to form hydrogen bonds. However, a similar out-of-register-type assembly would likely be strongly unfavorable for full-length CsgA with a predicted β-helical native structure, where the repeat sequences R1–R5 are predicted to stack directly on top of each other (Fig. 1).

Author contributions

T.V.S., conceptualization, formal analysis, investigation, methodology, writing – original draft; Y.Z., investigation; P.W., investigation; C.W., formal analysis, methodology, supervision; D.E.O., conceptualization, formal analysis, funding acquisition, methodology, project administration, supervision, writing – review & editing.

Editor: Michael T. Woodside.