The amphibian antimicrobial peptide uperin 3.5 is a cross-α/cross-β chameleon functional amyloid

Significance

We determined the crystal structure of the full-length amphibian antimicrobial peptide (AMP) uperin 3.5 and showed fibrillation into helical “cross-α” amyloid fibril, correlated with its antibacterial activity. This provides a molecular basis for the link between AMPs which are largely helical in nature, and amyloid formation. Uperin 3.5 is a cross-α amyloid discovered in eukaryotes, following a previously reported cross-α amyloid fibril for Staphylococcus aureus PSMα3 cytotoxin, hence demonstrating the existence of the cross-α amyloid architecture across kingdoms of life, with potential functional roles in early evolution. Furthermore, the findings revealed a chameleon cross-α/cross-β secondary structure switch of uperin 3.5 fibrils, likely related to regulation of its activity.

Abstract

Antimicrobial activity is being increasingly linked to amyloid fibril formation, suggesting physiological roles for some human amyloids, which have historically been viewed as strictly pathological agents. This work reports on formation of functional cross-α amyloid fibrils of the amphibian antimicrobial peptide uperin 3.5 at atomic resolution, an architecture initially discovered in the bacterial PSMα3 cytotoxin. The fibrils of uperin 3.5 and PSMα3 comprised antiparallel and parallel helical sheets, respectively, recapitulating properties of β-sheets. Uperin 3.5 demonstrated chameleon properties of a secondary structure switch, forming mostly cross-β fibrils in the absence of lipids. Uperin 3.5 helical fibril formation was largely induced by, and formed on, bacterial cells or membrane mimetics, and led to membrane damage and cell death. These findings suggest a regulation mechanism, which includes storage of inactive peptides as well as environmentally induced activation of uperin 3.5, via chameleon cross-α/β amyloid fibrils.

Antimicrobial peptides (AMPs) are found in all kingdoms of life, serving roles in the host defense system by fighting microbial infections and killing cancerous cells (1–4). AMPs mainly target and disrupt membranes, leading to cell death (1), and, in some cases, their self-assembly into supramolecular structures enhances antimicrobial activity (5). Specifically, certain AMPs, such as dermaseptin S9, assemble into well-ordered fibrils that resemble amyloids, proteins which form a cross-β architecture comprising tightly mated β-sheets (6–17). Human amyloids have primarily been associated with neurodegenerative and systemic diseases (18, 19), and evidence of antimicrobial properties for some, including the Alzheimer’s associated amyloid-β and Parkinson’s associated α-synuclein, suggests a physiological role in fighting infections threatening the brain (6, 20–26).

Our earlier investigations of functional amyloids revealed a distinct structure−function correlation in the phenol-soluble modulins (PSMs) family of peptides secreted by the Staphylococcus aureus bacterium, which are involved in virulence activities (27, 28), and form amyloid fibrils with specific morphologies (29–31). Specifically, the biofilm-associated PSMα1 and PSMα4 adopt the amyloid ultrastable cross-β architecture (29), likely to serve as a scaffold rendering the biofilm a more resistant barrier. Exceptionally, PSMα3, which plays roles in cytotoxicity against human immune cells (27, 32), forms cross-α amyloid fibrils that are composed entirely of amphipathic α-helices. The helices stack perpendicular to the fibril axis into mated “sheets” (31), just as the β-strands assemble in amyloid cross-β fibrils. Furthermore, a short segment from PSMα3 forms atypical β-rich fibrils with antiparallel orientation and shows a mild antibacterial activity (29). Overall, PSMαs show diverse activities as well as different morphologies of amyloid and amyloid-like fibrils. We previously showed that the ability of PSMα3 to form cross-α fibrils is critical for cytotoxicity, likely mediated through a dynamic process of coaggregation with membrane lipids (31). Coaggregation of amyloids with membrane lipids was previously suggested, for example, for α-synuclein (33). Recently, various synthetic peptides, unnatural enantiomers and protein mimics have been shown to form an architecture resembling cross-α (34–38).

Previous studies by Bowie, Carver, Martin, and coworkers showed that the AMP uperin 3.5, secreted on the skin of Uperoleia mjobergii (Australian toadlet) (39), forms amyloid fibrils, and suggested an interaction with bacterial membrane lipids that stabilize its α-helical conformation (40–43). Moreover, by using uperin 3.5 mutants, they showed that a high α-helical content and lower net charge contribute to a higher aggregation rate of the peptide (41, 42). Here, we demonstrate at atomic resolution that uperin 3.5 formed cross-α fibrils, which we suggest are essential for its toxic activity against the Gram-positive bacterium Micrococcus luteus. Moreover, the presence of bacterial membrane lipids induced a structural transition of uperin 3.5 into helical species in solution and in the fibrils, whereas their absence revealed a chameleon behavior of a secondary structure switch into cross-β fibrils, which correlated with reduced antibacterial activity.

Results

Uperin 3.5 Is a Functional Cross-α Amyloid.

The AMP uperin 3.5 self-assembled to form elongated amyloid fibrils, as visualized using transmission electron microscopy (TEM), which bound the amyloid indicator dye thioflavin-T (ThT) (SI Appendix, Fig. S1), similar to previous reports (40). The crystal structure of the full-length, 17-residue AMP, uperin 3.5, determined at 1.45-Å resolution (Table 1) (Protein Data Bank [PDB] ID code 6GS3), revealed a cross-α fibril architecture (Fig. 1). The cross-α fibrils of uperin 3.5 formed from stacks of amphipathic α-helices, positioned perpendicular to the fibril axis, creating a unique arrangement of mated “helical sheets” extending along the fibril axis, similar to the β-sheets which form cross-β amyloid fibril architecture (6–17). Overall, the cross-α fibrils resemble cross-β fibrils in their quaternary structures and in the ability to induce ThT fluorescence.

Table 1.

Data collection and refinement statistics for uperin 3.5 cross-α fibril

Characteristic	Value
PDB ID code	6GS3
Beamline	ESRF MASSIF-3
Date	April 19, 2017
Data collection
Space group	P 1 21 1
Cell dimensions
a, b, c (Å)	19.70, 28.44, 20.32
α,β,γ (°)	90.00, 106.95, 90.0
Wavelength (Å)	0.968
Resolution (Å)	19.47–1.45 (1.49–1.45)
R-factor observed (%)	8.5 (60.1)
Rmeas* (%)	8.9 (64.8)
I/σI	15.7 (2.7)
Completeness (%)	97.0 (94.7)
Redundancy	10.6 (7.6)
CC1/2† (%)	99.9 (90.8)
Refinement
Resolution (Å)	19.44–1.45 (1.49–1.45)
Completeness (%)	97.2 (94.7)
No. reflections‡	3398 (238)
Rwork§ (%)	18.5 (24.4)
Rfree (%)	22.9 (28.0)
No. atoms	275
Protein	254
Ligand/ion	5
Water	16
B-factors
Protein	15.1 (Chain A)
	12.7 (Chain B)
Ligand/ion	41.4 (K+)
	19.3 (SCN−)
Water	21.6
rms deviations
Bond lengths (Å)	0.016
Bond angles (°)	1.783
Clash score (85)	0
Molprobity score (85)	0.5
Molprobity percentile (85)	100th percentile
Number of crystals used for scaling	1

Values in parentheses are for highest-resolution shell.

^*R_meas is a redundancy-independent R-factor defined in ref. 86.

^†CC_1/2 is percentage of correlation between intensities from random half-datasets (87).

^‡Number of reflections corresponds to the working set.

^§R_work corresponds to working set.

Fig. 1.

The cross-α amyloid fibril architectures of uperin 3.5 versus PSMα3. The crystal structures of uperin 3.5 (PDB ID code 6GS3) (green and gray ribbons) and PSMα3 (PDB ID code 5I55) (30) (blue and gray ribbons) cross-α fibrils are shown in two orientations: down, and along the fibril axis. Two mated helical sheets are displayed in each panel, with six layers of laterally stacked α-helices depicted in each sheet (fibrils are likely composed of thousands of layers). The α-helical sheets interact via their hydrophobic face to create a tight interface. In contrast to the parallel orientation of the helices along the sheets of PSMα3, the helices in uperin 3.5, colored in different shades of green for clarity, take on an antiparallel orientation. The distances between helices along the sheets and between sheets are shown in each panel. In both panels, side chains are shown as sticks with heteroatoms colored by atom type (nitrogen in blue, oxygen in red, and sulfur in yellow).

While the antibacterial amphibian uperin 3.5 and the cytotoxic bacterial PSMα3 showed low sequence similarity (SI Appendix, Fig. S2A), both contained amphipathic helices and assembled into helical sheets that mate via a hydrophobic core (SI Appendix, Table S1, and visualized for uperin 3.5 in SI Appendix, Fig. S2). However, while the PSMα3 helices are orientated in a parallel fashion (30), uperin 3.5 helices were arranged in an antiparallel manner (Fig. 1). Thus, the cross-α amyloid fibril architecture can be encompassed by either parallel or antiparallel helical sheets. Uperin 3.5, in comparison to PSMα3, showed shorter intersheet and interhelix distances, likely due to less bulky side chains (Fig. 1). The high-order crystal packing of uperin 3.5 showed alternating hydrophobic and hydrophilic interfaces between rows of sheets (SI Appendix, Fig. S3), as seen in PSMα3 (30). The solvent-accessible surface area buried within the mated sheets of the cross-α uperin 3.5 fibril resembled that of PSMα3 (SI Appendix, Table S1), despite the shorter sequence of uperin 3.5. Nevertheless, the shape complementarity between the uperin 3.5 mated sheets was smaller than that of mated PSMα3 sheets (SI Appendix, Table S1), which might be related to the staggered orientation of the sheets (Fig. 1). The staggered orientation might be compensated by higher-order packing in the fibril, with more than one pair of sheets per row (SI Appendix, Fig. S3). Overall, the eukaryotic uperin 3.5 and the bacterial PSMα3 share a cross-α architecture but display extensive polymorphism in helix orientation and stacking.

In addition to the extensive hydrophobic core between sheets along the fibril, uperin 3.5 was also stabilized via an array of interhelical polar bonds that ran along each sheet. These included interhelical electrostatic interactions between Asp4 and Lys14, and an interhelical hydrogen bond between the side chain of Arg7 and the amide carbonyl of Arg7 from an adjacent helix (SI Appendix, Fig. S2D). Overall, each helix is potentially involved in four polar interactions along the sheet of stacked helices forming the cross-α fibril. Previous studies have indicated the importance of Arg7 in lipid interactions (41, 42), which aligns with our hypothesis, discussed below, that membrane interactions and the formation of cross-α fibrils are correlated processes.

Uperin 3.5 Shows a Chameleon Propensity of a Secondary Structure Switch in the Presence of Bacterial Membrane Lipids.

In solution, uperin 3.5 formed a random coil structure with typical minima at 197 nm, as measured using circular dichroism (CD) spectroscopy. This was in contrast to PSMα3, which is helical in solution (31). Nevertheless, the addition of small unilamellar vesicles (SUVs) composed of 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE) and 1,2-dioleoyl-sn-glycero-3-phospho-(1′-rac-glycerol) (DOPG) at a 1:1 molar ratio, mimicking a Gram-positive bacterial membrane (44, 45), induced an immediate secondary structure transition of uperin 3.5 toward an α-helical conformation, with typical minima near 208 and 222 nm (Fig. 2A). A similar phenomenon has been reported for other AMPs, which showed a random coil structure in solution, and a structural transition in the presence of membrane lipids (46).

Fig. 2.

Bacterial lipids induce uperin 3.5 secondary structure transition into α-helical species in soluble and fibrillar states. (A) Solution CD spectra of uperin 3.5, indicating a random coil conformation (dashed curve). Upon addition of DOPE:DOPG SUVs, the uperin 3.5 structure immediately transitioned and stabilized in a dominant α-helical conformation (solid curve). (B) The ssCD spectra of uperin 3.5 fibrils, indicating the formation of β-rich fibrils (dashed curve). When incubated with DOPE:DOPG SUVs, the fibrils took on a α-helical conformation (solid curve). The ssCD spectra of the cross-α fibrils of PSMα3 are shown in SI Appendix, Fig. S4 for comparison. (C) The ssCD spectra of an uperin 3.5 thin film, indicating the formation of β-rich fibrils (dashed curve). The addition of DOPE:DOPG SUVs solution on to the thin film of preformed β-rich fibrils resulted in a secondary structure transition toward the α-helical conformation (solid curve). (D) ATR-FTIR spectra of the amide I′ region of uperin 3.5 fibrils (dashed curve) showed a major peak at 1,616 cm⁻¹, indicative of cross-β amyloids (57–59). Uperin 3.5 fibrils formed in the presence of DOPE:DOPG SUVs (solid curve) demonstrated a major peak at 1,652 cm⁻¹, indicative of α-helices (60, 61), and a minor peak at 1,616 cm⁻¹, indicative of residual β-rich fibrils. The dotted gray lines indicate wavenumbers of 1,616, 1,634, and 1,652 cm⁻¹. In all experiments (A–D), the signal of buffer only or DOPE:DOPG SUVs solution was negligible and was subtracted from each measured sample.

The secondary structure of dry uperin 3.5 fibrils was evaluated using solid-state CD (ssCD) spectroscopy (47–51), previously shown to be highly applicable for characterization of aggregating proteins and peptides (48, 52–54). The ssCD spectrum measured for uperin 3.5 incubated in the absence of bacterial lipids indicated a β-rich conformation, with a typical minimum at 218 nm (Fig. 2B). In contrast, uperin 3.5 fibrils incubated in the presence of DOPE:DOPG SUVs had a seemingly α-helical conformation (Fig. 2B). While the fibril sample in the ssCD spectrum showed a deeper 222-nm minimum as compared to the 208-nm minimum, the sample in solution showed the opposite trend, with a deeper 208-nm minimum (Fig. 2). The same trend was observed when analyzing a thin film of preformed PSMα3 fibrils (SI Appendix, Fig. S4). The switch in the depth of the minima of fibrils versus soluble uperin 3.5 could be related to different secondary structure subpopulations, association between α-helices (55, 56), or other physicochemical properties. Adding DOPE:DOPG SUVs to the preformed β-rich uperin 3.5 fibrils led to a transition toward an α-helical conformation (Fig. 2C), providing further evidence of lipid-induced helicity, even after the fibrils were already formed.

The secondary structure of dry uperin 3.5 fibrils was further studied using attenuated total internal reflection Fourier transform infrared (ATR-FTIR) spectroscopy, which was shown to be useful for characterization of fibrillar architectures in amyloid proteins and peptides (57–61). The major peak in the amide-I′ region of the IR absorption spectrum of uperin 3.5 fibrils was at 1,616 cm⁻¹, indicative of the cross-β amyloid architecture (57–59), while the minor peak at 1,652 cm⁻¹ was indicative of α-helices (60, 61) (Fig. 2D), suggesting a mixed population with a predominant cross-β structure. Forming the fibrils in the presence of DOPE:DOPG SUVs resulted in a shifted IR spectrum, with a strong major peak at 1,652 cm⁻¹, indicative of a majority of α-helices (Fig. 2D). Additional minor peaks at 1,616 and 1,634 cm⁻¹, indicative of cross-β amyloids and regular β-sheets, respectively, suggested some remaining β-rich structures. The X-ray fiber diffraction of uperin 3.5 incubated in the absence of bacterial lipids showed a cross-β signature (62), with orthogonal reflection arcs at 4.7-Å spacing (sharp) and at 7 Å to 11 Å (diffused) (SI Appendix, Fig. S5). A fiber X-ray diffraction of uperin 3.5 fibrils grown in the presence of bacterial lipids showed strong reflections of the lipids, and the pattern of the protein was not adequately coherent for interpretations. Overall, these findings suggest a chameleon behavior of uperin 3.5, with transition from a random coil conformation in its soluble state to either a cross-β fibril conformation, formed in lipid-free solution, or a cross-α fibril structure, induced by the presence of bacterial membrane lipids.

Thermostability of Uperin 3.5 Fibrils Might Stem from Their Chameleon Propensity.

Uperin 3.5 fibrils were found to be thermostable, as observed in electron micrographs after a heat shock treatment of incubating the fibrils at 60 °C for 10 min (Fig. 3B). The ssCD spectrum of preformed uperin 3.5 fibrils indicated largely on β-rich species (Figs. 2B and 3A), while, after the heat shock treatment, a deeper minimum at 218 nm was observed, as compared with non−heat-shock-treated fibrils of the same sample, indicating a larger population of β-rich species (Fig. 3A). In comparison, S. aureus PSMα1 fibrils, which bear a cross-β configuration (29), were also thermostable, while the PSMα3 cross-α fibrils dissolved after the heat shock treatment (SI Appendix, Fig. S6), indicating on a fibril secondary structure-dependent thermostability.

Fig. 3.

Thermal stability and the effect of heat on the secondary structure of uperin 3.5 fibrils. (A) The ssCD spectra showing the effect of heating on the secondary structure of preformed thin films of uperin 3.5 fibrils formed alone or in the presence of DOPE:DOPG SUVs. (Left): Uperin 3.5 incubated alone showed a predominant β-rich fibril conformation before and after a 10-min, 60 °C heat shock treatment (blue and red curves, respectively). (Right): Uperin 3.5 fibrils formed in the presence of DOPE:DOPG SUVs showed predominant α-helical fibril conformation (blue curve), while, after a 10-min, 60 °C heat-shock treatment, they showed a secondary structure transition toward a β-rich conformation (red curve). The dotted gray lines indicate wavelengths of 208, 218, and 222 nm. The same sample was tested following temperature change in both assays. (B) TEM micrographs of 1 mM uperin 3.5 incubated for 2 d at room temperature alone, or in the presence of DOPE:DOPG SUVs, showing thermostable fibrils. (Scale bars, 300 nm.) The same sample was tested following temperature change in both assays. TEM micrographs of DOPE:DOPG SUVs incubated alone as control are shown in SI Appendix, Fig. S7.

Uperin 3.5 incubated in the presence of DOPE:DOPG SUVs showed massive fibril formation and thick fibrils around and on the SUVs (Fig. 3B; SUVs incubated alone as control are shown in SI Appendix, Fig. S7). The ssCD spectra indicated on a mixed population of species with a predominant α-helical conformation (Fig. 3A). When subjected to a heat shock treatment, the uperin 3.5 fibrils formed in the presence of the SUVs remained stable, whereas the SUVs seemed to partially denature, as observed in the electron micrographs (Fig. 3B), and the ssCD spectra indicated on a transition to a predominant β-rich architecture (Fig. 3A). Overall, these findings demonstrate a heat-induced transition of uperin 3.5 fibrils to a predominantly β-rich conformation, which probably led to thermostability. This corresponds to the higher stability of the cross-β PSMα1 as compared to cross-α PSMα3 (SI Appendix, Fig. S6), which are homologous sequences with fibrils of different secondary structures.

Uperin 3.5 Fibrillation Is Involved in Antibacterial Activity.

The antibacterial activity of uperin 3.5, examined using the disk (agar) diffusion test against four Gram-positive pathogens, demonstrated its more potent activity against M. luteus as compared to Staphylococcus hominis, Staphylococcus epidermidis, and S. aureus (SI Appendix, Fig. S8), as previously reported (39). Using the standard broth dilution assay for quantification, a minimum inhibitory concentration (MIC) of freshly dissolved uperin 3.5 against M. luteus, defined as the lowest concentration that prevented bacterial growth for 24 h, was 2 µM (Table 2). Membrane surface topography analysis of M. luteus bacterial cells, visualized using scanning electron microscopy (SEM), indeed showed that overnight incubation with 2 µM uperin 3.5 induced some membrane damage, while 6 µM uperin 3.5 induced severe morphological damage (Fig. 4A). Higher-resolution TEM micrographs of 4 µM uperin 3.5 incubated for 24 h with M. luteus showed massive fibril formation around the bacterial cells, which led to cell death (Fig. 4B). Uperin 3.5 incubated under the same conditions without the bacteria showed only scarce fibril formation (Fig. 4B), suggesting that the presence of bacterial cells induced aggregation. A similar effect was observed in samples of uperin 3.5 with DOPE:DOPG SUVs, in which massive fibril formation was observed around the SUVs (Fig. 3B).

Table 2.

Uperin 3.5 antibacterial activity against M. luteus

Uperin 3.5	MIC (μM)	MIC (μM) after 60 °C treatment
Freshly dissolved	2	2
5-d incubated	4	16
Freshly dissolved FITC labeled	8	—

Fig. 4.

Electron micrographs of uperin 3.5 fibrils formed on bacterial cells and of uperin 3.5-induced membrane damage. (A) SEM images showing M. luteus cells incubated for 24 h in the (Left) absence or presence of uperin 3.5 at (Middle) 2 μM and (Right) 6 μM, and resultant membrane dents. (Scale bars, 500 nm.) (B) Transmission electron micrographs of M. luteus cells incubated in the (Left) absence or (Middle) presence of 4 μM uperin 3.5 for 24 h, showing massive fibril formation around bacterial cells (Middle). (Right) Scarce fibril formation is shown upon incubation without cells. (Scale bars, 1 μm for M. luteus [Left], and 200 nm for M. luteus incubated with uperin 3.5 [Middle] and for uperin 3.5 incubated alone [Right].)

In accordance with the above findings, light microscopy images showed that fluorescein isothiocyanate (FITC)-labeled uperin 3.5, which retained fibril-forming and antibacterial abilities (Table 2 and SI Appendix, Fig. S9), accumulated on or inside rhodamine-labeled DOPE:DOPG SUVs (SI Appendix, Fig. S10A) or M. luteus cells (SI Appendix, Fig. S10B). The latter led to membrane permeation, followed by propidium iodide uptake, indicating cell death. Taken together, these findings demonstrate the bidirectional effects of uperin 3.5 and bacterial cells, namely, the role of bacterial lipids in inducing cross-α fibrillation, and the role of cross-α fibrillation in antibacterial activity.

The heat shock-treated uperin 3.5 soluble peptide retained its antibacterial function, with a similar MIC against M. luteus cells, before as compared to after heat treatment (Table 2). In contrast, preformed uperin 3.5 fibrils (preincubated for 5 d) exhibited a heat-induced reduction in their antibacterial activity, with an MIC of 16 µM versus 4 µM following, as compared to before heat treatment, respectively (Table 2). Of note, it is difficult to compare the activity of fresh and incubated samples due to the rapid aggregation rate and change in effective concentration. The heat-induced effect was compared using the same sample immediately preheating and postheating, yet potential changes in fibril dynamics must also be considered when analyzing the factors that affect activity. Moreover, since heat and the presence of lipids had opposite effects on the secondary structure of preformed uperin 3.5 fibrils (Figs. 2 and 3A), it must be assumed that addition of uperin 3.5 to bacterial cells may partially reverse some of the heat-induced effects. Thus, the overall fourfold reduction in antibacterial activity of preformed fibrils after heating is likely the result of a combined effects of heat-induced secondary structure transition toward a β-rich conformation, altered dynamics between fibrils and monomers, changes in biophysical properties and effective concentration, and additional factors. Overall, the polymorphism within the same peptide sequence, and the mixture of populations, including different secondary structures and in/soluble states, present great challenges to structure−function−fibrillation relationships of functional fibrils in general. The presence of other components such as lipids and cells further affect the dynamic within species, increasing complexity and interpretations.

Discussion

Here, we revealed and characterized the functional cross-α fibril architecture formed by the eukaryotic uperin 3.5 AMP, and its chameleon secondary structure switch, putatively regulating activity. The helical fibril formation of uperin 3.5 was largely induced by, and formed on, bacterial cells, where it consequently led to severe membrane damage and cell death (Fig. 4 and SI Appendix, Fig. S10). Of note, crystal structures of short helical AMPs are rather rare (63); nevertheless, quasiracemate crystal structures of frog AMP magainin 2 derivatives showed laterally stacked helices (64), similarly to the cross-α architecture. While correlations between amyloid formation and antibacterial activity have long been drawn (6–17, 20–26), this work provided an atomic-level description for this link, involving a cross-α amyloid configuration. The departure of the secondary structure of the cross-α amyloid uperin 3.5 from the typical cross-β configuration clarifies the connection between AMPs, which are largely helical in nature, and amyloid formation. Interestingly, amyloids secreted from microbes, either during a pathogenic infection or even by the gut microflora, have been implicated in the development of neurodegenerative and amyloid systemic disorders (65, 66). The overall connection between microbes, AMPs, the immune system, and neurodegenerative and systemic diseases requires more extensive studies.

In line with the reported polymorphism of cross-β amyloids (67), we observed polymorphism within the cross-α fibrils of uperin 3.5 and PSMα3. Both showed helical mated sheets of stacked amphipathic α-helices, which lay perpendicular to the fibril axis. PSMα3 contained parallel α-helices, while uperin 3.5 contained antiparallel helices within the sheets (Fig. 1), further recapitulating the parallel and antiparallel β-sheet configurations. Similarly, an antiparallel cross-α crystal structure was shown in a designed synthetic peptide (35). Furthermore, the interhelix and intersheet distances of uperin 3.5 were smaller than those in PSMα3, likely due to less bulky side chains, and due to the staggered sheets of uperin 3.5 (Fig. 1). As amyloid polymorphism has been correlated with different toxicity levels and prion disease strains (68, 69), the polymorphisms within cross-α structures may dictate toxicity levels, mechanisms of action, cell specificity, and fibril stability. While PSMα3 was stable as an α-helix in solution and in fibrils, uperin 3.5 helicity was dependent on the presence of lipids or other components. Moreover, while PSMα3 cross-α fibrils dissolved when subjected to heat shock, uperin 3.5 was thermostable (Fig. 3B and SI Appendix, Fig. S6), which is likely attributed to the heat-induced transition into cross-β fibrils (Fig. 3A). In PSMα3, secondary structure changes were shown inducible via mutations (31) or truncations (29) which formed mixed α/β populations or exclusively β-rich structures (29, 31). Of note, since we cannot determine the exact composition of the uperin 3.5 samples, namely, soluble vs. insoluble species and mixture of different secondary structures, in real time of bacteria interaction, we can only suggest a trend, based on our combined results, of a reduced activity of the β-rich polymorph.

Chameleon proteins that can adopt diverse secondary structures in response to external chemical or physical stimuli have been previously described (70–72), some induced by lipids or heat (73–75), similar to what was observed here for uperin 3.5 fibrils (Figs. 2 and 3). Martin, Carver, and coworkers previously pointed on a potential chameleon propensity of uperin 3.5 (42), and discussed the role of helices as intermediates to aggregation (34, 42, 76, 77) along with the potential contribution of intermolecular helix–helix interactions in the formation of a fibrillation nuclei (42). A thermal-induced cross-α/cross-β transition was previously reported in a synthetic self-assembling heptapeptide, suggested to occur via an intermediate turn motif (34). In addition, a dipeptide protein mimic, containing triazole linkages as peptide bond surrogate, formed β-sheet−rich structures in crystals, while heat-induced polymerization into a pseudoprotein revealed transition into helical sheets resembling the cross-α configuration (38). The seemingly relatively unstructured C-terminal region of uperin 3.5 (Fig. 1) might be essential for the mechanism of cross-α/cross-β transition. An alternative suggestion for the transition mechanism of uperin 3.5 could stem from environmental effects on the equilibrium of mixed α/β species (Fig. 2).

Taken together, we hypothesize that, in the amphibian uperin 3.5, secreted on the skin of the toadlet (39), the chameleon secondary structure switch could be related to regulation of activity. Namely, uperin 3.5 monomers secreted on the amphibian skin have disordered structures, while the presence of bacterial cells induces helicity and cross-α formation, which putatively facilitates the antibacterial activity and eliminates the threat of infection. In the absence of bacteria, uperin 3.5 assembles into inactive cross-β fibrils. This is somewhat similar to human amyloid toxicity which is debatably attributed to prefibrillar oligomeric conformations, which can contain α-helices (76), while the mature β-rich fibrils are considered less toxic (78). The β-rich uperin 3.5 can also serve as a reservoir (34, 79), presenting a means of storing a high local concentration of inactive uperin 3.5 on the amphibian skin, which can be switched back to helical conformation with the appearance of bacterial cells (as tested on mimetics; Fig. 2C). We expect that additional studies will shed additional light on the underlying regulation.

The mechanism by which uperin 3.5 disrupts membranes is yet to be fully determined. Recently, we suggested a peptide−membrane coaggregation model for PSMα3 toxicity against human cells (31). In this model, toxicity is not caused by a particular entity, such as monomers, oligomers, or fibrils. Rather, toxicity entails a dynamic process of peptide aggregation that is induced by, and involves, membrane lipids (31). Thus, both the presence of soluble species and the ability to fibrillate are critical determinants of toxicity. We suggest similar aggregation dynamics for uperin 3.5, in which cross-α fibrillation is both directly related to the presence of, and incorporates, membrane lipids (SI Appendix). However, we expect differences in the mechanisms of action of PSMα3 and uperin 3.5, due to the different properties of their fibrils and different sequences.

Altogether, the presented findings emphasize the structural complexity and the polymorphic nature of the amyloid fold, as shown in cross-α fibrils formed by prokaryotic and eukaryotic toxic peptides, namely, the bacterial PSMα3 and the amphibian uperin 3.5. This extends the definition of amyloid structures, and of functional fibrils. The findings could lay the foundations for the design of novel synthetic AMPs with enhanced potency, stability, and bioavailability, and with controllable storage and activities, to fight the growing threat of aggressive and resistant microbial infections.

Materials and Methods

A full methodological description is provided in SI Appendix.

Peptides.

Uperin 3.5 peptide (Uniprot accession code P82042) of the sequence GVGDLIRKAVSVIKNIV-NH2, PSMα1 (Uniprot accession code H9BRQ5) of the sequence MGIIAGIIKVIKSLIEQFTGK, and PSMα3 (Uniprot accession code H9BRQ7) of the sequence MEFVAKLFKFFKDLLGKFLVNN were all custom synthesized at >98% purity. Peptide pretreatment, and preparation of the SUV, are described in SI Appendix.

Fibrillation assays of uperin 3.5, with and without lipids, and/or bacterial cells, were conducted using ThT fluorescence fibrillation kinetics assay and TEM. Thermostability assays were performed using TEM. All are described in SI Appendix.

Crystallization Conditions, Structure Determination, and Refinement.

Nonpretreated uperin 3.5 peptide synthesized with free (unmodified) termini was dissolved in ultrapure water, to 10 mM, followed by a 10-min sonication in a bath sonicator, at room temperature. All crystals were grown, at 20 °C, via hanging-drop vapor diffusion. The drop was a mixture of peptide in reservoir solution (0.1 M KSCN, 0.1 M MES pH 6.03 and 20%vol/vol Jeff 600). Crystals were flash frozen with cryoprotection of 20% ethylene glycol and stored in liquid nitrogen prior to data collection. X-ray diffraction data were collected at 100 K, at the microfocus beamline Massif-III (ID30a-III) of the European Synchrotron Radiation Facility (ESRF) in Grenoble, France; wavelength of data collection was 0.9677 Å. Data indexation, integration, and scaling were performed using XDS/XSCALE (80). Phases were obtained using ARCIMBOLDO_LITE (81), Model building was done using Coot (82). Crystallographic refinements were performed with Refmac5 (83), and illustrated with Chimera (84). Crystallographic parameters are listed in Table 1. Calculations of structural properties are described in SI Appendix.

Solution and Solid State CD Spectroscopy.

Pretreated uperin 3.5 was dissolved to 10 mM in ultrapure water and diluted to 1 mM in 20 mM potassium phosphate buffer and 100 mM NaCl, pH 7.3. Uperin 3.5 was diluted in the CD cuvette to a final concentration of 0.2 mM, in the presence or absence of 0.6 mM SUVs. For ssCD spectroscopy, uperin 3.5 (5 mg/mL) dissolved in ultrapure water and uperin 3.5 (1.8 mg/mL) mixed with DOPE:DOPG SUVs solution at a 1:3 peptide:lipid molar ratio were incubated at 37 °C, with 300 rpm shaking, for 2 d. PSMα3 (5 mg/mL) dissolved in ultrapure water was incubated at 37 °C, with 300 rpm shaking, for 2 d. Fibrillated samples were centrifuged at 10,000 g for 5 min and resuspended in ultrapure water. For the thermostability assays, uperin 3.5 pretreated samples were dissolved in ultrapure water to 1 mM (1.8 mg/mL), mixed with DOPE:DOPG SUVs at a 1:3 peptide:lipid ratio, and incubated for 5 d at room temperature. The ssCD spectra of each sample was measured at room temperature, and again after a 10-min, 60 °C heat shock, while still being placed as a thin film on the disk. Additional details are described in SI Appendix.

ATR-FTIR Spectroscopy.

Uperin 3.5 was dissolved to 1 mg/mL in 5 mM hydrochloric acid and sonicated in a bath sonicator for 5 min, at room temperature. Immediately prior to measurements, the dry peptide was dissolved in D₂O, either with or without SUVs, to a final concentration of 20 mg/mL, with a 1:1 peptide:lipid molar ratio. Additional details are described in SI Appendix.

Antibacterial Assays.

M. luteus (an environmental isolate) was a kind gift from Charles Greenblatt from the Hebrew University in Jerusalem, Israel. S. hominis (ATTC 27844), S. aureus (ATCC 29213), and S. epidermidis (ATCC 12228) were purchased. M. luteus was cultured in Luria−Bertani medium at 30 °C, with 250 rpm shaking, overnight. S. aureus, S. hominis, and S. epidermidis were cultured in brain−heart infusion medium, at 37 °C, with 250 rpm shaking, overnight. Disk (agar) diffusion test, determination of the MIC, SEM, and fluorescence microscopy are described in SI Appendix.

Data Availability.

Structure coordinates data have been deposited in PDB (ID code 6GS3).