Atomic Structure of the LL-37(17-29) Human Antimicrobial Peptide Reveals Functional Helical Fibril One Sentence Summary: The antibacterial activity of human and primate LL-37(17-29) is controlled by self-assembly into densely packed helical fibrils

Here we demonstrate, by crystal structures, the self-assembly of the human and primate antibacterial LL-37 active core (residues 17-29) into a densely packed hexameric fibrillar architecture of amphipathic helices. The fibril is composed of four-helix bundles with a hydrophobic core, while a network of polar interactions stabilizes contacts between bundles, overall forming a stable fibrillar configuration that is also thermostable in solution. Despite similarity in sequence and the formation of fibrils composed of amphipathic helices, the LL-37(17-29) fibril structure was significantly different from the cytotoxic bacterial PSMα3 peptide, which fibrillates into amyloid cross-α fibrils. LL-37(17-29) formed wide, ribbon-like fibrils, which co-localized with bacterial cells; structure-guided mutagenesis analysis indicated the importance of its helical self-assembly in antibacterial activity and interactions with bacteria. This work extends the previously reported link between antibacterial activity and the formation of ordered amyloid fibrils, to helical, stable, hexameric fibrils. This fibril-antibacterial link suggests a tunable mechanism of action and offers a prospective to design antimicrobial peptides with improved stability and bioavailability.

Antimicrobial peptides (AMPs) are canonical components of the innate immune system of many organisms (1). AMP self-assembly bears functional relevance and can enhance antimicrobial activity (2). Certain AMPs assemble into well-ordered fibrils that resemble amyloids (3)(4)(5)(6)(7), which are proteins known to form cross- fibrils composed of tightly mated β-sheets, and have been associated with neurodegenerative and systemic diseases (8,9). Correspondingly, recent evidence of antimicrobial properties among some human amyloids suggests a potential physiological role of proteins otherwise known as pathological (10)(11)(12)(13)(14). Human LL-37 (hLL-37) AMP, a hCAP-18 protein cleavage product which plays an important role in the first line of defense against pathogens (15), is also known to self-assemble (16,17). Fibrillation of LL-37 was found critical for binding DNA and affecting receptors in the immune system (18).
The hLL-3717-29 fragment was suggested to serve as the active core of the AMP, showing a different spectrum of antibacterial activity as compared to the full-length protein and other fragments (33)(34)(35). Although not directly detected in-vivo, hLL-3717-29 can be cleaved from hCAP-18 or LL-37 by either proteinase K or staphylococcal peptidase I on its N-terminal side, and by trypsin on its C-terminal side. hLL-3717-29 is also the region within hLL- 37 showing the highest sequence similarity to PSMα3 (Fig. S1), and like the latter, forms a helical monomeric structure shown by NMR experiments (36). More specifically, hLL-3717-29 generates an amphipathic helix with a large hydrophobic moment compared to the entire hLL-37 and to PSMα3 (Table S2). hLL-3717-29 elicited dose-dependent inhibition of Gram-positive Micrococcus luteus growth (Fig. S2), with a minimal inhibitory concentration (MIC) (37) of 22 µM (Fig. 1). It was also active against the Staphylococcus hominis bacterium, with a MIC of 39 µM (Fig. S3). We found that hLL-3717- 29 formed long (several micrometers and longer), ribbon-like, fibrils, visualized using transmission electron microscopy (TEM) ( Fig. 2A). Cryogenic electron microscopy (CryoEM) showed that the wide (few hundred nanometers) fibrils are composed of lateral association of thinner fibrils (Fig.   S4). The wide fibrils also formed in the presence, and interacted with M. luteus cells (Fig. 2B).
Our determination of the crystal structure of hLL-3717-29 at 1.35 Å resolution, revealed selfassembly of amphipathic helices into a densely packed and elongated hexameric structure forming a central pore (Table S1, Fig. 3A-D and Movie S1). There were two helices in the asymmetric unit of the crystal, with 67% of their solvent accessible surface areas buried within the assembly, indicating compact packing. For comparison, in the PSMα3 cross-α structure, each helix is 62% buried in the fibril (19). In contrast, structures of full-length LL-37, co-crystallized alone or with different lipids, resulted in different levels of assembly, including monomeric, dimeric, tetramers and fiber-like structure of oligomers (38). When co-crystallized with dodecylphosphocholine (PDB ID 5NNT) (38), full-length LL-37 showed a repetitive architecture of juxtaposed 'head-totail' amphipathic helices, with interactions mediated by detergent molecules. This structure formed a much looser packing compared to the LL-3717-29 structure, with only 45% of the helix buried within the protein assembly.
The structure LL-3717-29 lacked amyloid continuous sheets with individual molecules stacked perpendicular to the fibril axis, and correspondingly did not bind the amyloid indicator dye Thioflavin T, in contrast to the cross-α amyloid fibrils of PSMα3 (19) (Fig. S5). Rather, the fibrillar assembly of LL-3717-29 was comprised of associating four-helix bundles, each stabilized by a closely packed hydrophobic core (Fig. 3E&F). Arginine residues are lined on the surface of the bundles, with side chains extending outwards, providing an overall highly positively charged molecular electrical potential surface (Fig. S6). The interfaces between bundles comprises a network of polar interactions, including potential salt bridges between Asp26 and Arg23/Arg29 from two adjacent helices, and between Lys25 and the C terminal of an adjacent helix (Fig. S7).
Due to the symmetry in the structure, each bundle of four helices could form sixteen inter-helical polar interactions with adjacent helices in the assembly (Fig. S7). In addition, Asp26 could potentially form a salt bridge with Arg29 on the same helix. Such intra-helical salt-bridges are associated with increased α-helical stability (39). Phe11, facing towards another Phe11 residue from an adjacent helix, contributes to hydrophobic packing between bundles. The overall stable helical assembly of LL-3717-29, which includes a network of polar interactions and hydrophobic packing, corresponded with the thermostability of the fibrils, as visualized by electron micrographs after heating to 60°C or 80°C (Fig. S8). Some disassembly at the edge of the wide fibrils, into thinner fibrils, could be observed after the 80°C heat shock, or after further incubation following the 60°C heat shock (Figs. S8B&C) demonstrating the lateral fibril association.
The fibrillar assembly in the crystal structure created alternating hydrophobic and polar (positively charged) zigzagged belts on its surface (Fig. 3D), suggesting interactions with and disruption of negatively charged lipid bilayers, such as bacterial membranes (40). Confocal microscopy images of fluorescein isothiocyanate (FITC)-labeled hLL-3717-29, which showed antibacterial activity similar to that of the unlabeled peptide (Fig. S9), confirmed aggregation and co-localization of hLL-3717-29 with bacterial cells (Fig. S10 and Movie S2).
The role of the self-assembly in the antibacterial activity of hLL-3717-29 was examined using single-point mutations designed based on the solved crystal structure. Alanine substitution of Ile24, the most buried residue within the core of the four-helix bundle ( Fig. S11 and Table S3), abolished antibacterial activity against M. luteus (Fig. 1) and against S. hominis (Fig. S3). This mutation failed to exhibit peptide assembly around bacterial cells (Fig. S12). Correspondingly, confocal microscopy images of the FITC-labeled I24A mutant, which, like the unlabeled peptide, was also ineffective against M. luteus (Fig. S9), indicated absence of peptide aggregation (Fig.   S10). Likewise, the I24S, I24K, I24Q and I24D mutations all abolished antibacterial activity against M. luteus (Fig. 1), and confocal microscopy images of the FITC-labeled I24S inactive mutant (Fig. S9) showed no detectable aggregation (Fig. S10). Similarly, Phe27 was significantly buried within the fibrillar assembly (Table S3), forming contacts with residues on the four-helix bundle and with other helices in the fibrillar assembly (Fig. S11). The F27A mutation abolished antibacterial activity against M. luteus (Fig. 1) and failed to aggregate when incubated with the bacteria (Fig. S12). In contrast to the buried Ile24 and Phe27, Gln22 was the least buried residue in the assembly, forming minimal contacts within adjacent helices ( Fig. S11 and Table S3).
Consequently, the Q22A mutation resulted in minimal change in activity against M. luteus (MIC=33 µM; Fig. 1) and was also active against S. hominis, with a MIC of 53 µM (Fig. S3).
Electron micrographs of the Q22A mutant showed large fibrous nano-structures contacting the bacterial cells (Fig. S12). Correspondingly, confocal microscopy images of the FITC-labeled Q22A, which was slightly less active than the unlabeled Q22A peptide (Fig. S9), formed aggregates which co-localized with the bacterial cells (Fig. S10). Overall, the mutagenesis analyses indicated the importance of hLL-3717-29 self-assembly in its antibacterial activity and supported a plausible active fibril arrangement.
Investigation of the N-terminal residues which are not buried within the assembly (Table   S3) but which face the central pore, showed that the F17A and K18A mutants display a similar ~3fold reduction in antibacterial activity against M. luteus, with a MIC of 60 µM (Fig. 1). Maintaining the positive charge via a K18R substitution, showed a very similar MIC to that of hLL-3717-29, while the K18H substitution showed slightly reduced activity (Fig. 1). In contrast, substitution to the polar but uncharged glutamine (K18Q) fully abolished activity (Fig. 1). The results suggest that the two residues facing the pore are important for activity, with the positive charge of Lys18 being the critical determinant. We therefore cannot conclude about the specific role of the central pore in activity, as the effect of substitutions might be related to the reduced positive charge (21,41) regardless of its structural location.
To conclude, the atomic structures of human and primate antibacterial LL-3717-29 showed a fibrillar assembly of densely packed amphipathic helices. It presents a type of self-assembly which, to the best of our knowledge, is distinct from other protein fibrils, with a role in direct killing of bacterial cells still to be fully determined. Mutagenesis analyses, designed based on the crystal structure, supported the role of self-assembly and of this particular structure in antibacterial activity and in direct interactions with bacterial cells. The sequence similarity between the human LL-37 and the bacterial PSMα3, and their shared ability to form helical functional fibrils, suggest a possible molecular or structural mimicry mechanism used by the bacteria to provide immune-evasive and survival strategies (42,43). Links between fibril formation and antimicrobial activity are accumulating (3)(4)(5)(6)(7)(10)(11)(12)(13)(14), and here, we provide atomic-level insight.
Further elucidation of the interplay between antimicrobial activity and fibril formation and morphology will aid the design of AMPs with enhanced potency, selectivity, stability, bioavailability and shelf-life. Moreover, under specific conditions, tunable self-assembly of AMPs might provide another layer of control over antimicrobial activity, in the form of antibacterial therapeutics or coating of medical devices, and will may target other roles of AMPs in immunomodulation and in killing cancerous cells (44).

Acknowledgments
We

Figure S1. Sequence alignment of human LL-37 and bacterial PSMα3
Sequence alignment between human LL-37 (UniProt ID P49913) and S. aureus PSMα3 (UniProt ID H9BRQ7). Amino acids are color-coded by their physicochemical properties (45). Identity and similarity between the two sequences were 19% and 24%, respectively. The hLL-3717-29 segment within the full sequence of hLL-37 is underlined and constitutes the most conserved region between the two peptides, with 31% identity and 39% similarity to the equivalent segment in PSMα3.         (F) and I24S (G) inactive mutants (Fig. 1) do not undergo aggregation.

Figure S11. Structural location of mutated LL-3717-29 residues
A zoom-in view of the fibrillar assembly, focusing on one four-helix bundle, with each helix colored differently. In each panel, residues that were substituted in our assays are individually shown using space-filling model. (A) Gln22 faces outward from the bundle, forming very few contacts with adjacent helices, with only 13% of its solvent accessible surface area (SASA) buried in the assembly (Table S3). (B) Ile24 is completely buried (95% of its SASA) inside the four-helix bundle. (C) Phe27 faces away from the bundle yet contacts both other residues on the same bundle and adjacent helices, with 85% of its SASA buried in the assembly.   (Table S1)). The N-termini of the helices, with Phe17 in hLL-3717-29 or Ser17 in gLL-3717-29, and Lys18, lining the central pore. The pore of gLL-3717-29 is more occluded, due to the orientation of the lysine residues extending into the pore. In the hLL-3717-29 structure, the lysine residues are almost perpendicular to the crosssection of the pore.  (b) R-meas is a redundancy-independent R-factor defined in (47).
(d) Number of reflections corresponds to the working set.
(e) Rwork corresponds to working set. within the four-helix bundle. 2 The number indicated is the percentage of the SASA per residue on an isolated helix versus the SASA of this residue within the fibrillar assembly, averaged for the two chains. The higher the indicated percentage, the more buried the residue is on an isolated helix by surrounding helices within the fibrillar assembly. 3 The number indicated is the percentage of the SASA per residue on the four-helix bundle buried by surrounding helices in the fibrillar assembly, averaged for the four chains. The higher the indicated percentage, the more buried the residue is by surrounding helices within the fibrillar assembly but not by other helices on the same four-helix bundle. This indicates the contact of each residue with helices surrounding the four-helix bundle. 4 The percentage of SASA buried in the four-helix bundle versus the SASA buried in the fibrillar assembly. The higher the percentage difference per residue, the more contacts there are with surrounding residues on the four-helix bundle compared to residues on helices outside the bundle.

Movie S1. Crystal structure of hLL-3717-29 and the assembly of the peptide into fibril
The movie displays the hexameric crystal structure of hLL-3717-29 from different orientations and emphases. The helices are shown in grey ribbon presentation, with one representative four-helix bundle colored pink. The movie starts with a view into the fibril axis and zooms in to the representative bundle, showing its side chains. The orientation is then rotated 90˚ for a view along the fibril axis, and then further rotated along the fibril axis to display the overall helical assembly.
The view is again zoomed into the bundle, showing side chains. The orientation is then rotated back to a view into the fibril axis.

Movie S2. Confocal microscopy imaging of live M. luteus incubated with FITC-hLL-3717-29
A series of confocal microscopy images was taken for about two hours, at intervals of 10 minutes, and then merged into a movie. The images reveal a rapid accumulation of FITC-hLL-3717-29 on the bacteria cells and the formation of peptide foci, indicating rapid aggregation in the presence of M.
luteus. Depletion of the Hoechst blue signal was observed for bacterial cells localized with the peptide (green), indicating cell rupture and the release of DNA (49,50). Bacterial cells with no accumulation of the peptide maintained a strong Hoechst blue signal.

Materials and Methods
Peptides  controlled chamber at 100% relative humidity, 3 l of each sample were applied to a perforated carbon film supported by an electron microscope grid, which were pre-discharged, as described above. After 3 sec, the drop was blotted by Whatman filter paper and liquids were vitrified through rapid plunging of the grids into liquid ethane at its freezing point (55). Specimens were examined under a FEI Talos 200C high-resolution electron microscope, at an accelerating voltage of 200 kV, using a Gatan 626 cryo-holder. To minimize electron beam radiation damage, the low-dose imaging mode was used. Images were collected digitally by a FEI Falcon III directimaging camera and the TIA software, with the help of the "phase plates" (FEI), to enhance image contrast (56,57).

Crystallization conditions
Lyophilized human and gorilla LL3717-29 peptides were dissolved to 10 mM ( Crystals were flash-frozen in liquid nitrogen before X-ray data collection. Structure determination and refinement X-ray diffraction of gLL3717-29 was collected at the ID23-EH2 micro-focus beamline at the European Synchrotron Radiation Facility (ESRF), Grenoble, France. The wavelength of data collection was 0.873Å. X-ray diffraction of hLL3717-29 was collected at the EMBL micro-focused beam P14, at the high brilliance 3rd Generation Synchrotron Radiation Source at DESY: PETRA III, Hamburg, Germany. The wavelength of data collection was 0.976Å. Data indexing, integrating and scaling were performed using XDS and XSCALE (58). Phases were obtained by molecular replacement using Phaser (59). For the molecular replacement of gLL3717-29, a 13-residue polyalanine idealized helix was used as the search model. For the molecular replacement of hLL3717-29, the structure of gLL3717-29 was used as the search model. Crystallographic refinements were performed with Refmac5 (60). Further model building was performed using Coot (61) and illustrated with Chimera (including constructing Movie S1) (62). The structures of human and gorilla LL3717-29 were determined at 1.35 Å and 1.1 Å resolution, respectively. In both structures, there were two peptide chains in the asymmetric unit and water molecules. There were no residues detected in the disallowed region at the Ramachandran plot. Crystallographic statistics are presented in Table S1.

Sequence alignments and proteolytic digestion prediction
Sequence alignment between LL3717-29 and PSMα3 was performed using the MAFFT server (63).
Amino acids are colored by their physicochemical properties (45). Proteolytic digestion sites in the human LL37 were predicted using the Expasy's PeptideCutter Tool (64).

Calculations of structural properties
The electrostatic potential map, hydrophobicity and B factor scales presented in the figures were created using Chimera (62). The values of the hydrophobicity scale were according to Kyte and Doolittle (65). The electrostatic potential was calculated using APBS-PDB2PQR (66). Helix amphipathicity and the hydrophobic moment (Table S2) were calculated with HeliQuest (67).

Solvent-accessible surface area calculations
Solvent-accessible surface areas (SASAs) were calculated using AREAIMOL, with a probe radius of 1.4Å (68,69), via the CCP4 package (60). The solvent-accessible buried surface area of each chain in the asymmetric unit was calculated as the area difference between the isolated chain and the chain within the fibril assembly, and is presented as the percentage of the total SASA of the chain. The SASA per residue within different isolated helical assemblies are presented in Table   S3.