Molecular Architecture of PSM α 1 Functional Amyloids

16 Bioﬁlms are structurally and functionally complex networks of bacteria and nanoscale macromolecules that play an important role in a myriad of settings from personal health and agriculture to power productions and fuel storage. Amyloid nanoﬁbers are integral components of many bioﬁlms and serve various purposes ranging from virulent to structural. Nonetheless, the precise characterization of bacterial amyloid nanoﬁbers has been elusive, with incomplete and contradicting results. The present work focuses on the molecular details and characteristics of PSM α 1-derived functional amyloids present in Staphylococcus aureus bioﬁlms, using a combination of computational and experimental techniques. Results from molecular dynamics simulations, guided and supported by a variety of experiments, show that nanoscale nanoﬁbers present a helical structure formed by two-protoﬁlament PSM α 1 amyloid nanoﬁbers. PSM α 1 peptides assembles into cross- β -sheet structures with an average diameter of about 12nm , adopting a left-handed helical structure with a periodicity of approximately 72nm . Strikingly, the chirality of the self-assembled nanoﬁbers, an intrinsic geometric property of its constituent peptides, is central in determining the growth and shape of the ﬁbers. The presented ﬁndings provide structural insights into the properties of the functional amyloids, hypothesize the role of chirality on the formation of ﬁbers, and aid in strategies for the design of anti-amyloid compounds. experimental results from this work open the door to the possibility of designing anti-amyloid nanoparticles 44 that present speciﬁc supramolecular interactions with the nanoﬁbers. The 2 β PSM α 1 amyloid nanoﬁber model can be used to study nanoﬁber- antimicrobial interactions to elucidate a mechanism for bioﬁlm manipulation 44 using man-made biomimetic nanostructures.

serve as key virulence factors that stimulate inflammatory responses, alter the host cell cycle, lyse human cells, and contribute to 37 biofilm structuring [21][22][23] . PSMs, α-helical amphiphatic peptides, are classified depending on their length. The smallest peptides 38 (21 amino acids in length) are α-type, PSMα1-4 and δ -toxin. The longer peptides, which are 44 amino acids in length, are 39 PSMβ 1 and PSMβ 2. Despite their sequence similarity, not all PSMs form ordered amyloid structures and not all of them follow 40 the same structural motifs. As such, PSMα3, the most toxic member, forms cross-α nanofibrils, while PSMα1 and PSMα4 41 form canonical cross-β amyloid nanofibers. A truncated PSMα3 instead presents atypical β -rich nanofibril architectures, 42 highlighting the importance of structure as basis of functional diversity exhibited by S. aureus PSMαs 24 . 43 To date, various open questions remain about the formation and characteristics of PSM-derived amyloid nanofibers. Some 44 studies have identified the crystal structures of functional amyloids reporting a consistent nanofiber diameter of approximately 45 10-12 nm 25, 26 , but a specific reason for this mechanism of growth has not been completely deciphered. Similarly, the chirality 46 and helical structure of the nanofibers remains controversial 13,24 . From an experimental perspective, a critical issue is that large studies. There is, therefore, a clear need to gain additional insights on the structures of PSMα1 functional amyloid nanofibers. 58 In the present study, we report on the molecular structure of PSMα1 nanofibers and their characteristics, such as diameter, 59 chirality and periodicity, and advance hypotheses on the role of chirality on the mechanisms of nanofiber assembly. Leveraging a 60 combination of fully-atomistic molecular dynamics (MD) simulations and experimental data obtained via mass spectroscopy and 61 microscopy-based techniques, we probe the characteristics of several in silico candidate structures for the amyloid nanofibers. 62 We find compelling evidence that a cross-β -sheet two-protofilament (2beta) structure is the most plausible structural model 63 for PSMα1 nanofibers in solution, that matches the experimental values of chirality, diameter, and periodicity of mature PSMα1 64 nanofibers in solution.

65
The presence of randomly coiled amorphous-like regions at the interface of nanofibers and water and between fibers further 66 support the importance of molecular simulations, as rigid crystal structures would fail to capture the entropic and imperfect 67 features of disordered regions of otherwise highly-ordered nanofibers 35  are detected as early as day 4, via β -sheets signal in circular dichroism (CD) spectroscopy ( Fig. S1 in SI), and visualized at day 80 9 via transmission electron microscopy (TEM; Fig. S2 in SI); however, the process is longer than previously reported 31 . Fibers 81 keep evolving for about two more weeks: CD spectra taken for samples older than 14 days are characterized by strong β -sheets 82 and disordered structures signals, while diameters measured from TEM images show a significant decrease in the standard error 83 of the mean diameters ( Fig. S3 and Tab. S1 in SI) for samples taken at day 21 and later. Leveraging this gradual stabilization, in 84 the following discussion, we will only analyze data of the mature fiber (i.e., after day 14).

85
The mature PSMα1 amyloid nanofibers are stable when subjected to thermal and mechanical (i.e., sonication) stress (see 86 Figure 1. Examples of a PSMα1 (a) 2β and (b) 2α nanofibers in water, with colors and text illustrating the terminology and labels used in this work. Individual PSMα1 peptides (black) form long β -strands that pair (e.g., light and dark orange) to form protofilaments (blue). When computing properties, it is convenient to define a layer (green), which is composed of PSMα1 from different strands and different protofilaments, approximately on a plane perpendicular to the nanofiber axis of elongation. The first character in each nanofiber name (1, 2, 3) represents the number of protofilaments in the structure; the second term (α or β ) describes the main structural motif of each peptide molecule within the nanofiber.
Based on these data and the information available in existing literature, we selected six classes of deformylated-PSMα1 91 aggregates as plausible candidates for the amyloid nanofibers (see Tab. 1). Namely, we simulated aggregates formed by 92 one, two, or three laterally-aggregated protofilaments ( Fig. 1 for clarifications regarding terminology) of PSMα1 in either 93 α-helical or β -sheet configuration, and estimated their characteristics by performing classical all-atom MD simulations. This 94 approach is more time consuming than starting from an experimentally-estimated structure, which is what is generally done for 95 non-bacterial amyloids, but it avoids introducing any bias due to measurement limitations (e.g., crystallization). The α-helical 96 secondary structure was chosen because single PSMα1 peptides in solution adopt this configuration, while β -sheet structure is

101
Simulated nanofibers assembled with α-and β -motifs share only limited similarities. Both types of aggregates are stabilized 102 by a combination of hydrogen bonding, hydrophobic, hydrophilic, and Coulombic interactions, but where these interactions 103 occur differentiates the two types of aggregates (Fig. 2). Within each protofilament, the PSMα1 peptides of β -sheet nanofibers 104 assemble into parallel β -sheet strands (Fig. 2a), with hydrogen bonds primarily occurring between residues seven through 105 thirteen. By contrast, the peptides in the α-nanofibers do not form hydrogen bonds with other peptides, but rather within 106 each peptide (Fig. 2b). The resulting α nanofibers are weakly stabilized. During the simulations, long aggregates break 107 into smaller clusters, indicating that α-helical peptides are unlikely to form stable assemblies in the absence of external 108 factors. This difference in stability is consistent with the characteristics of the strand interactions. In the α-nanofibers, the Hydrophobic regions in (c) β -sheet nanofibers occur inside each protofilament at a 4-to-6-residue steric zipper, and in (d) α-sheet nanofibers between protofilaments; water atoms in the nanofiber's proximity are shown in red. Salt bridges between GLU16 (pink bubbles) and LYS9, LYS12, or LYS21 (cyan bubbles) are found in (e) β and (f) α nanofibers. (g) TEM images of PSMα1 nanofibers in solution pre-and post-treatment with HFIP and TFA. (h) Fourier-transform infrared spectroscopy of pre-and post-TFA and HFIP treatments. The pre-treated secondary structures of mature nanofibers are primarily β -sheets (1600 cm −1 to 1625 cm −1 ) and β -turns (1700 cm −1 ), where shaded regions are associated with β -sheet (gray), disordered/random-coil (blue), α-helical (yellow), and β -turn (purple) secondary structures.
The structures of multi-protofilament nanofibers are also very different: α-protofilaments aggregate in assemblies with a 115 common hydrophobic core but different diameter, while β -protofilaments stretch side-by-side, forming locally planar structures.

116
This behavior can be linked to the distribution of charged groups and salt bridges (a bond between the oxygen atoms of an 117 acidic residue and the nitrogen atoms of basic residue). Salt bridges form between residue 16 (glutamic acid) and one of three 118 lysine residues (9, 12, or 21) in both α and β aggregates, as shown in Fig. 2e & f, with differences, however, in both orientation 119 and most likely participating lysine. In the β -sheet nanofibers, the most common salt bridges form with LYS9 (and partially 120 with LYS12), which is located between the residues making up the steric zipper (ILE8 and VAL10), easily accessible to GLU16. 121 By contrast, the salt bridges in α-sheet nanofibers are predominantly formed with LYS12 and LYS21, as LYS9 is part of the 122 α-helix backbone and, therefore, not as readily available. 123 Our experiments indicate that the simulated β -structures are in better agreement with the characteristics of the β -nanofibers, 124 as infrared spectroscopy shows a strong signal associated with β -sheets and β -turns ( Fig. 2g & h). Additionally, treating 125 4/13 the mature nanofiber with HFIP, an aprotic surfactant, does not result in a complete nanofiber dissolution and only a partial 126 disappearance of the β -sheets signal, which is compatible with the observed inter-peptide hydrogen bonds, protofilament 127 hydrophobic interactions and location of β -turns. Finally, sample treatment with TFA, which affects the hydrogen bonding, 128 results in nanofiber dissolution and loss of the β secondary structure, with the appearance of a weak α-helix signal, a 129 phenomenon observed also in the simulations for unstable β -nanofibers (Fig. 5). These results speak to the fact that β -130 nanofibers are compatible with experimental observations; however, additional analysis is required to determine the number of 131 protofilaments that compose the PSMα1 nanofiber. (TEM) and 12 nm (AFM). This discrepancy is likely due to the difference in sample solvation during the two measures. In 137 order to take into account the experimental difference, we also simulated β -structures in vacuum. Of note, as mentioned before, 138 distribution from TEM images remains largely unchanged for the mature nanofiber, with a marginal reduction of the average 139 diameter occurring with the nanofiber aging.

140
By comparing corresponding experimental and simulated distributions (Fig. 3c), that is TEM with nanofiber in vacuum and 141 AFM with solvated nanofiber, we can exclude the 1β system, as it peaks at shorter distances and does not show any value of 142 the diameter above 10 nm, which are present in all the experimental distributions as well as literature data. When it comes 143 to selecting between the 2-protofilament and 3-protofilament structures, the comparison is not as discerning; while the 3β 144 simulated structure tends to have more frequent peaks at short range (thanks to the almost planar structure of the aggregate),

153
As both 2β and 3β are possible candidates for the nanofiber structures in dispersion, we leveraged the differences in structure 154 between these two classes of assembly, to determine if any (or both) structures are likely present in solution. Despite starting 155 6/13 from crystal-like topology, which does not resemble the structure in solvent or in vacuo, all the simulated β assemblies spontaneously evolve to adopt a helical configuration (Fig. 4a) over a short period of time. Moreover, all the β -nanofibers 157 display a left-handed chirality (Fig. 4b), which matches the handedness obtained from the CD spectra of the PSMα1 nanofibers 158 in solution (see Fig. S3 in the SI). The quantitative comparison for the half-periodicity length (peak-to-peak intensity in the 159 AFM image) shows that the 2-protofilament β -structure better matches the experimental results. Using the interquartile range 160 (IQR) as an estimate of the variability (similar to the standard deviation), we found that the IQR of the AFM data (from 23 nm 161 to 51 nm) overlaps with the IQR of the 2β structure (from 37 nm to 68 nm), which also has a median periodicity of 49 nm. 162 Conversely, the overlap of the experimental observation with the data for the 3β structure is minimal. 163 Interestingly, the 3β nanofibers tend to have a small layer-to-layer angle or an almost flat conformation (hence the long 164 period), and very short fibers (10 layers, ∼4 nm) prefer a right-handed chirality (this effect disappears for longer strands). This 165 behavior, together with the low stability of the flat conformation, resulted in instability of certain lengths of 3β nanofibers 166 during the simulations. This observation is in contrast with the other types of aggregates that rapidly assume their helical 167 structure when starting from flat conformation at any length. 168 These results suggest that the nanofiber can become locally unstable when more than two protofilaments are associated, 169 limiting the ability of the fiber to grow laterally, and that one of the potential roles of the associated extracellular DNA observed 3 α 9.88 ± 0.14 Ⓢ Ⓢ ⊗ ⊗    TFA is an excellent solvent for most peptides, and it is commonly used in both solid-and solution-phase peptide synthesis.

330
TFA is strong proton donor capable of breaking up intra-and inter-molecular hydrogen bonding, as well as removing many 331 protecting groups and the crude synthesized peptide from resins, but it is not strong enough to cleave peptide bonds.

332
As a solvent HFIP is polar and exhibits strong hydrogen bonding properties enabling it to dissolve substances that serve The nanofiber diameter distribution from TEM data was obtained with in-house code. The images were processed using kernel 344 filtering to determine the direction of the nanofiber in the frame of the image; we used 180 100×100 kernels for an angular 345 resolution of 1°after filtering the images to maximize the contrast between the edges and center of the nanofibers. Then, we 346 searched each filtered image in the direction perpendicular to the applied filter angle and estimated the diameter from the 347 distance between two consecutive high-contrast peaks.  Aliquots of 20 µL were deposited on freshly cleaved, buffer-washed muscovite mica (Grade V-4 mica from SPI, PA). The 351 samples were incubated for 1 min and dried under a gentle stream of nitrogen for 5 min before scanning. The images were 352 processed and analyzed using NanoScope software 6.13.

353
AFM image processing 354 The nanofiber periodicity and diameter distribution from ADF data was obtained with in-house code. The periodicity of the 355 PSMα1 structure was determined from AFM images by computing the distance between highest-intensity nanofiber peaks.

356
The AFM images were initially split in multiple images, each containing a single nanofiber, to obtain relatively consistent 357 intensity ranges. The highest point in the nanofibers where then identified using the Otsu thresholding method and the distance 358 11/13 between the center of closest peaks was collected. As the distance is dependent upon the Otsu threshold parameter, which in 359 turn depends on the intensity of the pixel, the threshold parameter for each image was determined to be the value, between 1.1 360 and 3.1, that minimizes the standard deviation for the values of the distances between the regions of highest intensity. Diameters 361 were obtained by first segmenting images in order to separate the fiber from the background. 3000 to 5000 points in the fiber 362 were then randomly chosen and, for each of them, the diameter was set equal to the shortest distance to opposite edges, which 363 was found by sampling all the directions with a 1°resolution.

394
The periodicity of the PSMα1 helical structure was determined from the average layer-to-layer angle and average layer-to-395 layer distance projected along the nanofiber axis, determined by linear regression of all atomic coordinates. For each layer 396 (see Fig. 1), first the center of mass and the principal axes of inertia were computed, followed by the layer-to-layer distance 397 (i.e., distance between the center of mass of consecutive layers) and by the layer-to-layer angle (dihedral angle from the largest 398 principal axis of each layer and the center of mass of two consecutive layers).

399
In the manuscript, we report the results obtained for the longest simulated fiber in each class, while data of shorter fibers 400 were used to test the convergence of the results, as they generally show a somewhat monotonic trend with aggregate length.

401
This effect becomes small when at least 20 layers are present even for the quantities that have the strongest dependence from 402 the nanofiber length (see Figs. S9 and S10).