Hexapeptide Steric Zippers Form Globular Structures on Route to Fiber Formation.

Steric zippers reside in proteins across the phylogenetic tree (19). As model hexapeptides, we chose 2 sequences from different genetic phyla, eukaryotic and bacterial, both of which have been identified by the Zipper database (ZipperDB) software (18, 22) to be of a high propensity to form fibers (relevant interaction parameters are in SI Appendix, Table S1). The hexapeptide, NNQQNY (henceforth referred to as NY6), is a classical amyloid eukaryotic hexapeptide of the yeast prion Sup35 that is commonly used as a model for amyloid fiber formation (23⇓⇓–26). NY6 forms cross–β-sheet crystals as confirmed by X-ray diffraction (15, 19, 27). By contrast and for comparison, we chose the hexapeptide, GVASAA (henceforth referred to as GA6), from the bacterial protein TasA, an extracellular matrix and fiber-forming protein (28⇓–30). The sequences of GA6 and NY6 are shown in Fig. 1 A and B, respectively. Unlike NY6, which is self-paired via hydrogen bonds between the side chains, the side chains in GA6 are mostly nonpolar and therefore, are potentially paired through van der Waals interactions.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Early and late aggregation of the hexapeptides GA6 and NY6. The hexapeptides’ molecular structures are shown in (A) GA6 and (B) NY6. AFM topography images of the hexapeptides showed globular structures at early timescales (∼30 min). C and D correspond to GA6 and NY6, respectively. Hexapeptides formed fibers following a 3-d incubation (1 mg/mL with 10% acetic acid). E and F correspond to GA6 and NY6, respectively. (Scale bars: 200 nm.)

In order to probe the nanomechanical properties of the hexapeptide layers prior to fiber formation, we adsorbed the hexapeptide layers on a polyethylene glycol (PEG) layer (molecular weight = 2,000 Da). The PEG layer was covalently attached to a silicon surface through a short silane molecule (Materials and Methods has details) and used in order to minimize surface effects on the formation of the hexapeptide layers. AFM topography images of the surfaces following the different surface modification steps (prior to the hexapeptide adsorption) are shown in SI Appendix, Fig. S1. AFM images of the hexapeptide-modified surfaces (Fig. 1 C and D) show that, following a 30-min incubation, the hexapeptides formed globular structures of (20 ± 7)-nm diameter and (4 ± 1)-nm height (GA6) and (40 ± 15)-nm diameter and (4 ± 1)-nm height (NY6). The size of the globular structures remained unchanged with a 10- to 20-times dilution of the hexapeptide layer (SI Appendix, Fig. S2). Hexapeptide globular structures also formed directly on a silane monolayer, showing that the globule formation is intrinsic to the hexapeptides (SI Appendix, Fig. S3). Considering the volume of a single hexapeptide to be 0.6 nm³ (GA6) and 1 nm³ (NY6) (31), the globular structures constituted ∼725 GA6 hexapeptides and ∼940 NY6 hexapeptides. [The number of hexapeptides per globule was calculated from the ratio between the volume of the globule and the volume of a single hexapeptide. For the globule volume, we used the equation V(globule) = π/6 × h × (3c² + h²), where h and c are the height and radius of the globule, respectively.] These globular structures remained unchanged for a few days following washing in a buffered solution. Repeating the adsorption protocol with a longer incubation of 1 mo yielded the formation of fibers (SI Appendix, Fig. S4), suggesting that indeed the hexapeptide globular structures were intermediate states on route to fiber formation. Fibers also formed following a 3-d incubation with acetic acid (Fig. 1 E and F). The fact that both hexapeptides formed globular structures (at the time point when we studied them), a geometry that minimizes their surface interaction with water, is suggestive of a self-preference of the hexapeptides over their interaction with water. This may be the initial driving force for the formation of fibers.

Pulling Hexapeptide Globular Structures Has a Characteristic Force Plateau Signature.

An AFM tip was modified with a covalently attached hexapeptide as is explicitly shown in the schematic in Fig. 2A. We measured the forces between the modified tip and the hexapeptide-modified surface on retraction from contact. The typical F-D curves, observed with both GA6 and NY6 hexapeptides, are shown in Fig. 2B. Curves I to III show nonspecific binding events owing to the interaction between the AFM tip and the substrate (curves I and III) or no binding events (curve II). Curve IV shows a worm-like chain (WLC) behavior observed when a PEG molecule, anchored to the substrate through the hexapeptide tail, is pulled away from the substrate with a parabolic restoring force (32⇓⇓⇓–36). Fitting a representative F-D curve with a WLC model is shown in SI Appendix, Fig. S5. In contrast to the F-D curves shown in curves I to IV, which have been previously observed in different experimental setups (37), particularly interesting curves are shown in curves V and VI, where steps of a constant force were observed. Curve V represents a single step, and curve VI represents multiple steps. In curve V (referred to here for simplicity), the plateau in the force region (marked by ** in curve V in Fig. 2B) was characterized by a certain distance, ΔL, and a rupture force, ΔF, as shown by an enlarged view into curve V (Fig. 2C). While the different curves shown in Fig. 2B were observed with both the GA6 and NY6 systems, the frequency of these events significantly differed in these 2 systems as outlined in the table in Fig. 2D. Specifically, in the GA6 system, more than 70% of the curves showed a step-like plateau in the force curve (curves V and VI). In comparison, less than 10% of the force curves in the NY6 system showed the step-like behavior, and the rest were distributed between curves I and IV, with a nonspecific adhesion F-D (curve I) dominating.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Representative F-D curves measured between a hexapeptide on an AFM tip and globular structures of the same hexapeptide on a PEG layer. A schematic of the experimental system shows the hexapeptide covalently attached to the AFM tip through a PEG linker that is bound to the AFM tip with (3-Aminopropyl) trimethoxysilane (APTMS) (A). Representative F-D curves are shown in B. Curves I to III represent nonspecific interactions between the AFM tip and the surface. Curve IV shows a WLC behavior. Curves V (single step) and VI (multiple steps; positions of steps in a curve are denoted by numbers) show the characteristic nanomechanical fingerprint of the hexapeptide globular structures. The enlarged view at curve V (C) highlights the characteristic step rupture force (ΔF) and step length (ΔL). Measurements between a GA6-modified AFM tip and surface and between an NY6-modified tip and surface showed similar F-D curves, but the frequency of the different F-D curve types varied between these experiments. The table in D depicts the percentage of each F-D curve type (I to VI in B) of the total number of analyzed F-D curves. N (number of F-D curves analyzed) = 5,760 to 9,212. We suggest that our pulling experiments are similar to pulling a macromolecular chain from a reservoir as described schematically in E. The asterisks in E correspond to different pulling stages along the F-D trace shown in curve V in B as marked with starts accordingly. Initially, the modified AFM tip detaches from the surface (* in curve V in B); then, the AFM tip is pulled with a constant force as the hexapeptide chains are being pulled from the (shrinking) globular reservoir (** in curve V in B) into the solution. A final rupture event occurs when dF/dX > K (spring constant) as the AFM tip detaches from the hexapeptide chain (*** in curve V in B).

We performed several control experiments in order to test whether the force plateau F-D curves were specific to the interaction between the hexapeptide modified tip and the hexapeptide globular structures on the surface. All of the control experiments have been performed with GA6 due to the higher percentage of force plateau events in the GA6 relative to the NY6 system. SI Appendix, Table S2 lists the different configurations that we tested in the control experiments as well as the percentage of the plateau F-D curves in each experiment. Similar force curves as those shown in Fig. 2B were also observed in the control experiments, but the F-D curves with a plateau force only dominated in setups where both the AFM tip and the surface were modified with the hexapeptide. We concluded that, for a pulling force plateau F-D curve to occur, the AFM tip had to be modified with a complementary sequence to that on the surface.

Interestingly, similar force plateau F-D curves have been observed on surface detachment of a polypeptide chain (38⇓–40) as well as α-synuclein protofibrils (11). By analogy, since the globular structures were larger than the size of a single hexapeptide, we suggest that during the force plateau events, we pulled on hexapeptide chains. The hexapeptide chains either preformed in the globular structures or they were shear induced, as they were pulled away from the globular structures by an AFM tip modified with a complementary hexapeptide. The force plateau (per step) implies a triangle potential for the chain-unwinding process, where the energy varies linearly with the distance as a chain is pulled from a globule (41⇓–43). From a molecular perspective, we suggest that the globules act as a reservoir of hexapeptide chains and that, following a nonspecific unbinding event (* in curve V in Fig. 2B), a constant force (** in curve V in Fig. 2B) is necessary to unwind the chains, possibly due to the equality between the chain-unwinding force and the restoring force of the chain. Then, the AFM tip desorbs from the chain, and a rupture event occurs (*** in curve V in Fig. 2B). Fig. 2E shows a schematic representation of our explanation of the force plateau curves (curve V in Fig. 2B). A similar F-D behavior has been observed when pulling a hydrophobic polymer from a globule into an aqueous media (44) or when pulling lipid tubes or tethers from membranes (45⇓⇓–48), suggesting a common origin to the plateau F-D curves.

We then performed a detailed statistical analysis of the characteristic parameters in force plateau curves, such as curves V and VI in Fig. 2B: the step rupture force, ΔF, and the length of the step, ΔL. Fig. 3 shows the distribution of these parameters (Fig. 3 A–D) calculated from all of the force plateau F-D curves with single and multiple steps. Gaussian fits to the distribution of the rupture force, ΔF, show similar values for both the GA6 and NY6 systems: ΔF = 90 ± 20 pN (GA6) and ΔF = 95 ± 20 pN (NY6) (Fig. 3 A and B). The single-peak distribution of the plateau rupture force, essentially lacking additional peaks of multiples of the rupture force, is suggestive of pulling a single chain (44, 49). Analysis of the distribution of ΔL and a Gaussian fit to the data yielded characteristic length for the GA6 system ΔL = (13 ± 2) nm and characteristic length for the NY6 system ΔL = (20 ± 10) nm (Fig. 3 C and D); both are of the order of the average diameter of the globular structures (∼20 nm for GA6, and ∼40 nm for NY6). Plotting the rupture force against the step length, ΔL (Fig. 3 E and F), shows that the rupture force is relatively constant over all of the step length ranges. The independence of ΔF on ΔL (as well as the lack of WLC events) is suggestive of a lack of an entropic stretching during the force plateau step event that is observed, for example, in pulling on a titin molecule (33, 34). This supports our suggestion that a chain was pulled from a reservoir rather than from a surface anchor.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Statistical analysis of F-D force plateau curves (curves V and VI in Fig. 2B). Distribution of the rupture force, ΔF, is shown for experiments with GA6 (A) and with NY6 (B). Distribution of step length, ΔL, is shown for experiments with GA6 (C) and with NY6 (D). A plot of the rupture force, ΔF, against the step length, ΔL, is shown in E for GA6, and in F for NY6. N = 250 to 1,400. Gaussian fits of the histogram distributions are shown in red together with the most probable ΔF and ΔL values.

We measured the F-D curves on retraction under different pulling speeds. We focused on the GA6 system because of its dominant force plateau behavior (as was done with the control experiments). The results, shown in SI Appendix, Fig. S6, show that the rupture force and step lengths are independent of the pulling speed, suggesting that, in the range of velocities tested, the system was in equilibrium with respect to molecular motion occurring during the pulling events. Similar behavior has been observed with a few other amyloid and polymer systems (13, 14, 50).

Pulling on surfaces covered with mature GA6 fibers, we observed multiple irregular F-D curves (a representative curve is shown in SI Appendix, Fig. S7). While the F-D curves showed steps occasionally, they were highly complex, and we could not easily classify them with the well-defined F-D curve types presented in Fig. 2B. The pulling events, sometimes extending to micrometers in length and nanonewtons in force (SI Appendix, Fig. S7), can be attributed to pulling on fibers with a more complex internal structure than the globular structures or to their surface detachment (11). In addition to the irregular F-D curves, we also observed nonbinding and nonspecific binding events. Since the fiber-coated layer itself was not as uniform and highly covered as that of the globules (Fig. 1E), these events can be attributed to the interaction of the modified tip with a bare surface. It can also be explained by the inaccessibility of the hexapeptide binding groups to the modified tip either due to them being buried inside the fibers or because they are more tightly packed (14).

The Characteristic F-D Curve Is Sequence Dependent.

Witnessing the differences between the GA6 and NY6 systems, we asked whether the F-D curves that we measured were sequence dependent. We studied the nanomechanical properties of shuffled hexapeptide sequences. The shuffled NY6, NNQYQN sequence (henceforth referred to as NY6S), and the shuffled GA6, VAAASG (henceforth referred to as GA6S), were used, because they both receive the lowest ZipperDB energy and composite score of all possible NY6 and GA6 shuffles (SI Appendix, Table S1). The shuffled hexapeptide NY6S (molecular structure is in Fig. 4A) also formed globular structures at the surface of a PEG layer (Fig. 4B) with an average diameter (30 ± 10 nm), which is between the diameters of GA6 (20 ± 10 nm) and NY6 (40 ± 15 nm). The F-D curves measured with NY6S exhibited a similar set of forces as that presented in Fig. 2B. However, shuffling the NY6 sequence resulted in ∼60% force plateau events of the total number of measured F-D curves relative to less than 10% measured with the NY6 system. Surprisingly, despite its low-ZipperDB score, the shuffled GA6 hexapeptide (molecular structure is in Fig. 4E) assembled into a mixture of globules and fibrils (Fig. 4F) under the conditions where GA6 formed only globules (Fig. 1C). Unlike the NY6S case where shuffling the sequence increased the frequency of force plateau events, with GA6S, only 3% of the F-D curves (relative to 70% with GA6) showed a force plateau behavior. This difference in step frequency suggests that NY6 and GA6S better pack in a globule, and thereby, they become less accessible to interaction with the hexapeptide-modified AFM tip. Analysis of the force plateau F-D curves measured with NY6S showed a step rupture force ΔF = 80 ± 20 pN (Fig. 4C) and a step length ΔL = 30 ± 20 nm (Fig. 4D). The shuffled GA6 structures that could be unzipped (3% of all of the F-D curves measured) showed a wide step length distribution (Fig. 4H), representing the large variety of the GA6S structures (the topography AFM image is in Fig. 4F). The distribution of step lengths of the GA6S system could not be fitted with a conventional statistical distribution function, but it included 3 dominant regions, peaking around 25 nm (corresponding to globular diameter) (SI Appendix, Fig. S8), ∼100 nm, and ∼170 nm (corresponding to fibrils’ lengths ranging between 50 and 250 nm). The step rupture force (70 ± 15 pN) (Fig. 4G) was single peaked, implying that we could not discriminate between globular structures and (young) fibril pulling in our experimental setup. Similarly to the unshuffled systems, the step rupture force was independent of the step length in the GA6S and NY6S systems (SI Appendix, Fig. S9).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

The effect of sequence on the F-D measurements. The molecular structure of NY6S and the shuffled sequence of the hexapeptide NY6 are shown in A, and the molecular structure of GA6S and the shuffled sequence of the hexapeptide GA6 are shown in E. NY6S forms globular structures at the surface of a PEG layer as shown by the AFM topography image in B, and GA6S forms a mixture of globules and fibrils (F). Analysis of the force plateau F-D curves measured between a hexapeptide-modified AFM tip and the hexapeptide layer yields the distribution of the step rupture force (ΔF) and the step length (ΔL) as shown for NY6S in C and D, respectively. The distributions of ΔF and ΔL are shown for GA6S in G and H, respectively. Gaussian fits of the histogram distributions are shown in red together with the most probable ΔF and ΔL values. The data in H could not be fitted with a standard statistical distribution function. (Scale bars: 200 nm.)

The Force Plateaus and Lengths Are Quantized.

In our analysis of the force plateau curves, we thus far referred to the characteristic parameters ΔF and ΔL and their total distribution over all of the force curves disrespectfully of the number of the force plateau events per F-D curve (a representation of multiple steps in curve VI is in Fig. 2B). However, the number of steps observed per F-D curve is an additional characteristic of the force plateau curves, and it poses an important clue to the mechanism behind the force plateau steps. We propose that unwinding single-hexapeptide chains from globular structures during the force plateau events is similar to unwinding a yarn ball, where loops of equal length are being pulled from a reservoir with equal force (51). Using this analogy, we suggest that, in our system, multiple steps occur when unwinding hexapeptides loops from globular structures (schematic model is in Fig. 5I). Better packing of the chains within a globule will lead to a smaller number of steps (loops) and vice versa. Fig. 5 and SI Appendix, Fig. S10 show that pulling on different hexapeptide layers yielded a different number of steps in each F-D force plateau curve. Specifically, the number of steps in GA6S (2 steps) < NY6 (3 steps) < GA6 (4 steps) < NY6S (6 steps). The smaller number of steps of GA6S and NY6 together with the lower step frequency of GA6S (3%) and NY6 (8%) indicate that these hexapeptides pack better in a globule relative to the GA6 and NY6S globular structures.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Comparison between the step rupture force (ΔF) and step length (ΔL) measured in different step positions. Step position 1 is the first rupture event as specified in curve VI in Fig. 2B. The step rupture force was measured with GA6 (A), GA6S (B), NY6 (C), and NY6S (D), and the step length, ΔL, was measured with GA6 (E), GA6S (F), NY6 (G), and NY6S (H). N = 232 to 812. Data are plotted on the same scale for clarity. We explain the multiple steps of equal ΔF and ΔL by analogy with unwinding a yarn ball as shown schematically in I (not drawn to scale). The globular structures are composed of hexapeptide chains (either preformed or shear induced) that are unwound loop by loop. The loops’ length matches the size of a step length, ΔL, and it is of the order of the globule’s diameter (both marked with an arrow). Unwinding the equally sized loops gives rise to multiple force plateau steps of equal force and length for each hexapeptide. The final loop is smaller than its preceding, resulting in smaller steps in the final rupture events. The rupture force measured in the final step was used to calculate the hexapeptide–hexapeptide interaction energy.

We compared the unwinding forces and step lengths between the different step positions in a force curve (curve VI in Fig. 2B shows the definition of the step position). Fig. 5 A–H shows that, in all of the hexapeptide layers, the unwinding lengths and forces are quantized (i.e., each unwinding event required the same amount of energy in agreement with the yarn ball model) (51). Interestingly, the characteristic step length of each hexapeptide globule, deduced from the F-D curves in Figs. 3 and 4—ΔL = 13 ± 2 (GA6), 20 ± 10 (NY6), 30 ± 20 (NY6S) nm—is of the order of the average globule diameter, deduced from the AFM images—20 ± 10 (GA6), 40 ± 15 (NY6), 30 ± 10 (NY6S) nm. We did not refer to GA6S in this comparison because it formed both globules and fibrils; however, here as well, the size of the globules measured by the AFM (20 ± 7 nm) was similar to ΔL ∼ 25 nm, as shown in SI Appendix, Fig. S8. Furthermore, the distribution of the step lengths correlates well with the distribution of the diameters of the globular structures (SI Appendix, Fig. S8). The correspondence between the measured step lengths and the size of the globules supports the suggestion that the force plateau steps represent unwinding of hexapeptide loops that are of the order of the diameter of the globular structure as illustrated in Fig. 5I. In the final step, before the AFM tip desorbs from the chain, the AFM probe is pulling the remaining part of the chain, and therefore, it is not necessarily similar in length to the “ΔL quanta.”

Last, we were intrigued to compare the Eisenberg hexapeptide–hexapeptide interaction energy with that calculated from the force needed to detach the hexapeptide-modified AFM tip from the hexapeptide chain. We used the equation ΔE = ΔF length_hexapeptide to calculated the energy required to detach the hexapeptide-modified tip from the hexapeptide chain. Estimating the length_hexapeptide to be the length of a single hexapeptide on which the detachment force acts (∼2 nm) and plugging in the rupture force values at the final step (ΔF [GA6] = 90 ±10 pN, ΔF [NY6] = 120 ±10 pN, ΔF [GA6S] = 70 ± 15 pN, ΔF [NY6S] = 80 ± 20 pN), we found that the hexapeptide–hexapeptide interaction was ΔE (NY6)= 35 ± 3 kcal/mol, ΔE (GA6) = 26 ± 3 kcal/mol, ΔE (NY6S)= 23 ± 2 kcal/mol, and ΔE (GA6S) = 20 ± 4 kcal/mol. Remarkably, all of these values stand well within the steric zipper unzipping energy calculated by Eisenberg and coworkers (18). However, while the calculation predicts that only the nonshuffled sequences form fibers (SI Appendix, Table S1), our results show that the differences in hexapeptide interaction energy are subtle, and indeed, eventually all of the hexapeptides formed fibers after a long time (Fig. 1 E and F and SI Appendix, Figs. S4 and S11). Considering all of the nanomechanical properties of hexapeptide structures that we measured and discussed, we conclude that it is the number of steps and their relative frequency in the F-D measurements and not their measured interaction energy that better probe the differences in the packing of these hexapeptide structures.