Structural and functional characterization of the pore-forming domain of pinholin S²¹68

Significance

With growing antibiotic resistance in bacteria, understanding the lytic cycle of bacteriophages can help developing phage therapy. In the double-stranded DNA bacteriophage φ21, the integral membrane protein pinholin S²¹68 triggers lysis by forming small “pinholes.” It has been proposed that the second transmembrane helix (TMD2) is sufficient to form these pores, which we could confirm in a reconstituted model system, showing a pore diameter of about 2 nm. Using NMR, we determined the structure of TMD2 as it is embedded and aligned in the membrane, revealing a right-handed glycine zipper motif in a prime position for homo- and heteromeric helix–helix interactions. Molecular models are derived from these data to illustrate the assembly of the pinholes via inactive and active dimers.

Abstract

Pinholin S²¹68 triggers the lytic cycle of bacteriophage φ21 in infected Escherichia coli. Activated transmembrane dimers oligomerize into small holes and uncouple the proton gradient. Transmembrane domain 1 (TMD1) regulates this activity, while TMD2 is postulated to form the actual “pinholes.” Focusing on the TMD2 fragment, we used synchrotron radiation-based circular dichroism to confirm its α-helical conformation and transmembrane alignment. Solid-state ¹⁵N-NMR in oriented DMPC bilayers yielded a helix tilt angle of τ = 14°, a high order parameter (S_mol = 0.9), and revealed the azimuthal angle. The resulting rotational orientation places an extended glycine zipper motif (G₄₀xxxS₄₄xxxG₄₈) together with a patch of H-bonding residues (T₅₁, T₅₄, N₅₅) sideways along TMD2, available for helix–helix interactions. Using fluorescence vesicle leakage assays, we demonstrate that TMD2 forms stable holes with an estimated diameter of 2 nm, as long as the glycine zipper motif remains intact. Based on our experimental data, we suggest structural models for the oligomeric pinhole (right-handed heptameric TMD2 bundle), for the active dimer (right-handed Gly-zipped TMD2/TMD2 dimer), and for the full-length pinholin protein before being triggered (Gly-zipped TMD2/TMD1-TMD1/TMD2 dimer in a line).

Upon infecting a host, bacteriophages release their newly produced offspring into the environment by lysis of the host cell. The infection cycle of double-stranded DNA bacteriophages that infect gram-negative bacteria is regulated by small viral membrane proteins, so-called holins (1–3). These holins accumulate in an inactive form in the hosts’ cytoplasmic membrane during phage morphogenesis (3). At an allele-specific time, the holins start to form big membrane lesions. Through these holes the muralytic endolysins escape from the cytoplasm and start to degrade the peptidoglycane layer (4, 5). Host lysis is then completed by the Rz–Rz1 spanin complex that disintegrates the outer membrane (6). In addition to this canonical holin endolysin system, a second class of holins, referred to as pinholins, has been described (3, 7). This class is represented by pinholin S²¹68 of lambdoid phage φ21 (7, 8). In contrast to canonical holins, for example those encoded by phage λ, the holes formed by pinholins are much smaller (“pinholes”) and do not allow the passage of proteins (7, 9).

The pinholin S²¹68 protein is encoded by gene S²¹ of the lysis cassette, which possesses a dual start motif to allow expression of an antipinholin as well. This antipinholin protein has three additional N-terminal amino acids (M₁-K₂-S₃) and serves as a specific inhibitor of pinholin (Fig. 1A) (8, 10). The (anti)pinholin protein structure consists of two transmembrane domains (TMDs) and an unstructured C terminus (8). It is postulated that TMD1 possesses regulatory functions, while TMD2 is responsible for the actual pinhole formation (8, 11–13). During phage morphogenesis, (anti)pinholin initially accumulates as inactive heterodimers in the cytoplasmic membrane. For pinhole formation, TMD1 has to flip out of the membrane to allow TMD2–TMD2 interactions that build up the pinhole (Fig. 1B) (8, 14). The presence of antipinholin reduces the pinholin activity, because the additional charge on the N terminus delays the flipping of TMD1 (8, 11). Previous cross-linking and modeling studies by Pang et al. (9) have suggested that the pinholes are composed of seven pinholin S²¹68 monomers. These heptamers are supposed to form small holes with an inner diameter of around 1.5 nm, which lead to a collapse of the proton gradient across the membrane (9).

Fig. 1.

Pinholin S²¹68. (A) Primary structure of the (anti)pinholin. The dual start motif of gene S²¹ of phage φ21 allows the expression of both antipinholin S²¹71 and pinholin S²¹68. The postulated TMDs are indicated, and our synthetic TMD2 fragment (W₃₆-E₆₂) is underlined. (B) In its inactive state, both helices of pinholin S²¹68 are supposed to be aligned in a transmembrane state (Left). Above a certain threshold concentration, TMD1 is supposed to flip out of the membrane into a surface bound state (Right) (14). This triggering allows further self-assembly through TMD2–TMD2 interactions and leads to the formation of a small pinhole pore by a putative heptamer (9, 13).

The machinery also contains associated endolysins carrying an N-terminal secretory signal, called the signal-anchor release (SAR) domain. In their membrane-tethered state, the SAR-endolysins are enzymatically inactive, which prevents premature lysis (15, 16). This inactive state is highly dependent on the presence of an intact membrane potential (7, 17). After disruption of this potential by the formation of pinholes, the SAR-endolysins are released from the membrane, and become refolded and activated during this process (7, 16).

Drew et al. (18) demonstrated recently that the wild-type protein is predominately α-helical (∼83%) in DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine; di-14:0-PC) vesicles. Ahammad et al. (14) from the same group used electron paramagnetic resonance (EPR) experiments on synthetic reconstituted pinholin to show that TMD1 is partly externalized from the lipid, whereas TMD2 is stably membrane inserted. In the present study, we aimed to determine the detailed membrane orientation of the postulated pore-forming TMD2 fragment and examine its putative role as the basic homo-/hetero-/oligomeric assembly unit. We used synchrotron radiation-based circular dichroism spectroscopy (SRCD) for quantitative secondary structure determination, and oriented CD (SROCD) for qualitative orientational analysis. Accurate parameters on the helix alignment were obtained by solid-state NMR spectroscopy (ssNMR) on nonperturbing selective ¹⁵N-isotope labels, which yielded a detailed picture of the TMD2 as it is positioned in the membrane. Using a vesicle leakage assay based on ANTS (8-aminonaphthalene-1,3,6-trisulfonic acid disodium salt)/DPX (p-xylene-bis-pyridinium bromide), we proved that TMD2 is indeed the pore-forming domain and could estimate the size of the pinhole from the leakage of fluorescein-isothiocyanate-dextrans (FITC-dextrans, FDs). Based on these experimental data, we have constructed putative models for the pinholin in all three stages of its functional cycle.

Results

Chemical Peptide Synthesis of the Pinholin TMD2 Fragment.

The sequence of TMD2 (W₃₆ to F₅₈) was chemically synthesized with an extension of four residues from the natural sequence (K₅₉-I₆₀-R₆₁-E₆₂) at the amidated C terminus. This extension increased the synthesis yield and facilitated peptide handling. Park et al. (8) showed previously that a ΔTMD1 mutant comprising residues D₂₉ to E₇₁ is still capable of pinhole formation in living cells. For ssNMR experiments, we incorporated a selective ¹⁵N-labeled leucine residue at either position L₄₂, L₄₅, L₄₇ or L₅₀. In addition, a fifth peptide comprising all four labeled residues (¹⁵N-L_{42, 45, 47, 50} TMD2) was synthesized (Table 1). For our functional studies using a fluorescent vesicle leakage assay, we synthesized two additional mutant peptides, namely S₄₄A TMD2 and G₄₀A S₄₄A G₄₈A TMD2. For the sake of simplicity, throughout this paper the TMD2 KIRE-amide peptide will be referred to as the TMD2 fragment.

Table 1.

Sequences of the chemically synthesized TMD2 fragments

Fragment	Sequence
TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
15N-L42 TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
15N-L45 TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
15N-L47 TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
15N-L50 TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
15N-L42, 45, 47, 50 TMD2	WAAIGVLGSLVLGFLTYLTNLYFKIRE-amide
S44A TMD2	WAAIGVLGALVLGFLTYLTNLYFKIRE-amide
G40A S44A G48A TMD2	WAAIAVLGALVLAFLTYLTNLYFKIRE-amide

¹⁵N-labeled leucine residues are underlined, and mutations are shown in bold.

Determination of the Secondary Structure of the Pinholin TMD2 Fragment.

Using SRCD, we investigated the secondary structure of the pinholin TMD2 fragment in small unilamellar lipid vesicles (SUVs). Compared to conventional CD on a benchtop instrument, SRCD enables measurements at lower wavelengths, gives a better signal-to-noise ratio, and is less affected by light scattering artifacts (19). The peptide was examined in SUVs reconstituted in phosphate buffer (10 mM; pH 7) with a peptide-to-lipid ratio (P:L) of 1:50 (mol:mol). In accordance with our SROCD and ssNMR experiments below, we examined the conformation of TMD2 in stable DMPC vesicles, and compared these with unsaturated POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine; 16:0–18:1-PC) vesicles that had to be used for further leakage experiments. In Fig. 2A the CD spectra of TMD2 in DMPC (solid line) and POPC (dotted line) are depicted. Both SRCD spectra show a typical α-helical line shape (20), with a positive band around 195 nm and two negative signals at 209 nm and 221 nm. Deconvolution of the spectra (as described in Methods using the CDSSTR and CONTIN-LL algorithm with different reference sets) resulted in an average α-helix content of 87% in DMPC and 90% in POPC (Table 2). Regarding the full sequence of the peptide with 27 amino acids, our findings suggest an effective α-helix length of 23 to 24 residues, which fits very well to the suggested TMD2 length (W₃₆-F₅₈). Notably, our data also show that the lipid systems used do not affect the overall fold of the peptide, as the two spectra are virtually identical.

Fig. 2.

Structure analysis of membrane bound TMD2. (A) SRCD analysis of S²¹68 TMD2 in SUVs composed of DMPC (solid line) or POPC (dotted line) with a P:L ratio of 1:50. Both measurements show a typical α-helical line shape, with a positive signal around 195 nm and two negative bands at 209 and 221 nm. The spectra are almost identical and show that the lipid systems used in this study do not affect the secondary structure. (B) SROCD spectrum of the pinholin TMD2 fragment in oriented DMPC bilayers. The spectrum shows an α-helical line shape with a positive fingerprint band at 208 nm (dotted line). This indicates that the TMD2 helix is oriented upright in DMPC bilayers as expected for a transmembrane state. (C) One-dimensional solid-state ¹⁵N-NMR spectra of labeled TMD2 fragments, measured in oriented membranes at the standard 0° sample angle (see Inset, Upper Right). Four peptides were synthesized with a single ¹⁵N-leucine label at position L₄₂, L₄₅, L₄₇, or L₅₀, plus a peptide containing all four labels. All peptides were reconstituted in macroscopically oriented DMPC bilayers with a P:L ratio of 1:50. The peptide ¹⁵N-L_{42, 45, 47, 50} TMD2 (black) shows four distinct signals between 180 and 215 ppm representing the four labels. The overlay clearly shows that each signal corresponds to one of the singly labeled peptides. Following the peptide sequence we assign ¹⁵N-L₄₂ TMD2 (red) at 196 ppm, ¹⁵N-L₄₅ TMD2 (blue) at 213 ppm, ¹⁵N-L₄₇ TMD2 (green) at 185 ppm, and ¹⁵N-L₅₀ TMD2 (purple) at 180 ppm. All labels give rise to signals between 180 and 215 ppm, showing that the TMD2 fragment is stably inserted essentially upright in a transmembrane state.

Table 2.

Secondary structure elements of the pinholin TMD2 fragment in DMPC and POPC SUVs (P:L = 1:50) evaluated from the CD spectra

Lipid	αR	αD	Sum α-helix	β-Strand	β-Turn	Unordered
DMPC	65 ± 4	22 ± 2	87 ± 4	3 ± 2	5 ± 1	6 ± 3
POPC	68 ± 4	22 ± 3	90 ± 4	3 ± 2	4 ± 1	3 ± 1

Secondary structure elements (in %; the sum is not always 100% due to rounding errors) of the pinholin TMD2 fragment in DMPC and POPC SUVs (P:L = 1:50) evaluated from the CD spectra. Average values and the SD for different algorithms are given (α_R: ideal helix; α_D: distorted helix).

Determination of the Membrane Alignment of the Pinholin TMD2 Fragment.

To investigate the 3D orientation of the pinholin TMD2 fragment in the lipid bilayer, we reconstituted the peptide in macroscopically aligned lipid bilayers made from DMPC with a molar P:L ratio of 1:50. We used SROCD and static solid-state ¹⁵N-NMR spectroscopy to characterize the membrane orientation of the TMD2 helix. The two methods are complementary and well suited for an entirely nonperturbing structure analysis of small helical membrane proteins. Both approaches make use of macroscopically oriented membrane samples, containing the reconstituted protein under full hydration. They not only provide the membrane alignment of helical proteins or peptides, but also reveal any possible change of orientation, and loss of conformation due to protein aggregation (21–27). The SROCD, as well as the ssNMR measurements, utilize the standard sample alignment of 0°, in which the bilayer normal is aligned parallel to the incident light beam or to the static magnetic field B₀, respectively.

Fig. 2B shows the SROCD data of TMD2 in lipid bilayers made from DMPC. In the analysis of oriented CD spectra, the fingerprint band at 208 nm gives information about the orientation of an α-helical protein in the membrane. If a spectrum displays a pronounced negative band at 208 nm, the α-helix is aligned perpendicular to the incident light and to the bilayer normal, hence the α-helix is oriented parallel to the membrane surface. If the α-helix is tilted, the signal at 208 nm is reduced toward zero. When the band is completely absent, this indicates a helix oriented parallel to the incident light and bilayer normal, hence a transmembrane helix. This signature occurs together with a red-shift of the positive band maximum around 190 nm (23, 28). Our data show a typical α-helical line shape with a positive signal around 208 nm, and the positive and negative bands are slightly shifted to longer wavelengths (compared to the isotropic CD spectra), which proves that the TMD2 fragment is inserted upright in DMPC bilayers.

To refine this result, we used 1D ssNMR with selectively ¹⁵N-isotope–labeled peptides (Table 1). The positions of four leucine residues (L₄₂, L₄₅, L₄₇, L₅₀) were chosen to represent and cover the entire helix surface. If an α-helix is aligned parallel to the magnetic field (which corresponds in our set-up to an orientation parallel to the bilayer normal), the ¹⁵N-chemical shift signal will arise between 170 ppm and 220 ppm. In contrast, an α-helix oriented perpendicular to the membrane normal would give rise to signals between 80 ppm and 130 ppm. Fig. 2C shows the superimposed 1D ssNMR spectra of the¹⁵N-labeled samples. All signals lie between 180 and 215 ppm, indicating that all TMD2 fragments are stably inserted upright in the membrane, with the helix axis more or less parallel to the bilayer normal. Notably, the singly labeled peptides show distinct signals with ¹⁵N-L₄₂ at 196 ppm, ¹⁵N-L₄₅ at 213 ppm, ¹⁵N-L₄₇ at 185 ppm, and ¹⁵N-L₅₀ at 180 ppm. Even the spectrum of the fourfold-labeled ¹⁵N-L_{42, 45, 47, 50} TMD2 fragment is so well resolved that it shows distinct signals representing the four leucine residues, which allows easy assignment. The samples have a high quality of peptide alignment, because they contain only a negligible fraction of a powder component, which reflects peptide segments that are nonoriented and therefore spread over every possible orientation in the magnetic field.

To determine both, the detailed helix tilt of angle of TMD2 as well as its rotational alignment, we performed 2D SAMMY experiments (29, 30), which correlate the ¹⁵N-chemical shift with the ¹H–¹⁵N-dipolar coupling. The SAMMY spectra were collected directly after the 1D measurement on the same samples. To assess the integrity of the oriented bilayers before and after the ¹⁵N-ssNMR experiments, we measured also the corresponding ³¹P-ssNMR spectra (SI Appendix, Fig. S1). They show that all samples were well oriented at the beginning of the measurement. After the measurement, some of the ³¹P-ssNMR spectra indicated a slight loss of orientation, which is due to sample drying and lipid degradation at the end of these 3-day measurements. It is important to mention that our ¹⁵N-ssNMR analysis is not affected, however, because deterioration occurs only toward the end of the 2D measurements, and any emerging broad components do not change the sharp oriented signals analyzed here.

The 2D experiments provide PISA (polarity index slant angle) wheels, from which the tilt angle τ relative to the membrane normal can be determined (32–34). These wheels also give information on the mobility of a helix within the lipid bilayer, by yielding a so-called order parameter S_mol (35). Fig. 3A shows the 2D spectra of the fourfold-labeled ¹⁵N-L_{42, 45, 47, 50} TMD2 peptide (Fig. 3A, black), and the singly labeled peptides (Fig. 3A, L₄₂ red, L₄₅ blue, L₄₇ green, L₅₀ purple) in oriented DMPC bilayers (for individual spectra, see SI Appendix, Fig. S2). The 2D spectrum of ¹⁵N-L_{42, 45, 47, 50} TMD2 shows, just like the 1D measurement, ¹⁵N-chemical shifts between 180 and 213 ppm, and ¹H–¹⁵N-dipolar coupling frequencies between 5.8 and 7.8 kHz. The spectrum displays four distinct signals, representing the four labeled leucine residues, which can be easily assigned via the singly ¹⁵N-labeled samples (L₄₂ = 196 ppm, 7.8 kHz; L₄₅ = 213 ppm, 7.2 kHz; L₄₇ = 185 ppm, 5.8 kHz; L₅₀ = 180 ppm, 6.8 kHz). Only the dipolar coupling frequencies of the selectively labeled peptides ¹⁵N-L₄₂ TMD2 and ¹⁵N-L₄₅ TMD2 are seen to be slightly increased compared to the fourfold-labeled sample. First, we determined the tilt angle and the order parameter of the TMD2 helix with the help of the fourfold-labeled ¹⁵N-L_{42, 45, 47, 50} TMD2 peptide. Then we calculated the signal positions of the four labels using an ideal α-helix (assuming 100° pitch angle between consecutive residues) as an underlying helix model. This way, we found a best-fit for a helix tilt angle of about τ = 14°, an order parameter of S_mol = 0.91, and rotation angles for the individual labeled residues of ρ_L47 = 50°, ρ_L45 = 210°, ρ_L42 = 270°, and ρ_L50 = 350°.

Fig. 3.

Comprehensive ssNMR analysis of TMD2 helix tilt and rotation angles. (A) Two-dimensional SAMMY spectra of the fourfold-labeled ¹⁵N-L_{42, 45, 47, 50} TMD2 fragment, and of the four corresponding singly labeled peptides. ¹⁵N-L_{42, 45, 47, 50} TMD2 (black) shows four distinct signals with ¹⁵N-chemical shifts between 180 and 213 ppm, and with ¹H-¹⁵N-dipolar coupling frequencies between 5.8 and 7.8 kHz. Each singly labeled peptide gives rise to a distinct signal, which can be assigned as ¹⁵N-L₄₂ TMD2 (red) at 196 ppm and 7.8 kHz, ¹⁵N-L₄₅ TMD2 (blue) at 213 ppm and 7.2 kHz, ¹⁵N-L₄₇ TMD2 (green) at 185 ppm and 5.8 kHz, and ¹⁵N-L₅₀ TMD2 (purple) at 180 ppm and 6.8 kHz. As an overlay over the spectra, we show our fitted signals (black dots), which reveal the underlying structural parameters of the helix as τ = 14°, S_mol = 0.91, ρ_L47 = 50°, ρ_L45 = 210°, ρ_L42 = 270°, ρ_L50 = 350° (see definitions in the text). (B) Helical wheel illustration of TMD2, showing the local rotational angles determined for each label. Our calculation shows that the peptide is tilted by 14° toward G₄₃ and L₅₀ (black arrow), which positions the glycine zipper sideways, together with a line of polar residues (G₄₀, S₄₄, G₄₈, T₅₁, T₅₄, N₅₅). In the hydrophobic lipid environment, this orientation makes the polar strip on TMD2 available for right-handed helix-helix interactions. The helical wheel plot was prepared using the Protein ORIGAMI web application (http://www.ibg.kit.edu/protein_origami/) (31), with the color code: yellow, hydrophobic; light blue, polar H-bonding; green, Gly.

From this self-consistent and unique solution, we can thus conclude that: 1) The conformation of the pinholin TMD2 fragment is fully consistent with an essentially ideal straight α-helix, 2) this helix is tilted by about 14° away from the membrane normal, and 3) it does not wobble strongly in the membrane. The latter statement is based on the best-fit order parameter of S_mol = 0.91, given that a value of 1.0 would indicate complete uniaxial order, while 0.0 would correspond to isotropic averaging. Accordingly, TMD2 undergoes only slight wobble, which may indicate some stabilization via helix–helix interactions. Another important structural parameter can be obtained from this comprehensive ssNMR analysis, namely 4) the overall azimuthal rotation angle of the TMD2 helix, which describes the 3D alignment of its hydrophobic/hydrophilic faces relative to the membrane. The rotational analysis of the TMD2 fragment shows that a glycine zipper motif together with several polar residues (G₄₀, S₄₄, G₄₈, T₅₁, T₅₄, N₅₅) are positioned sideways along the tilted helix. This result suggests that 5) the polar strip is freely available for lateral contact with other proteins, whose importance had already been pointed out in mutational studies (9, 11, 13). From the determined alignment, we can conclude that 6) any helix–helix interaction involving this motif would have to occur via a right-handed assembly of transmembrane segments.

Determination of the Function of the Pinholin TMD2 Fragment Using a Fluorescent Vesicle Leakage Assay.

Having confirmed the stable transmembrane alignment of TMD2, we modified a standard ANTS/DPX-based vesicle leakage assay in order to examine the formation of holes (36, 37). The peptide was reconstituted in large unilamellar vesicles (LUVs) made from POPC, at different P:L ratios (mol:mol) ranging from 1:400 to 1:50. The lipid POPC was chosen, because its low phase transition temperature kept the vesicles fluid and stable while handling them. The fluorophore ANTS and the quencher DPX were entrapped inside the LUVs, and external material was removed by gel filtration. We reasoned that, if the peptide is able to form holes per se, then the entrapped fluorophore and quencher can escape from the LUVs. Due to the resulting massive dilution (the volume of the vesicles being less than 0.1% of the total volume), the fluorescence of ANTS would be dequenched and hence an increase in fluorescence registered. The resulting fluorescence was measured over a fixed period of time, for 900 s at 22 °C. At this point, the nonionic detergent Triton X-100 was added to completely solubilize the vesicles, resulting in a signal corresponding to 100% leakage (Fig. 4A). Measurements were performed at least three times for each condition, and a control with pure POPC LUVs gave essentially no leakage within 900 s. We determined the extent of TMD2-induced leakage from POPC LUVs at several different P:L ratios (Fig. 4B). Even the lowest peptide concentration (P:L = 1:400) shows around 30% leakage, and at P:L ratio of 1:200 it is increased to 77%. At 1:150 and higher, leakage reaches essentially 100%. These data prove that the TMD2 peptide induces strong leakage, which is in perfect agreement with its postulated function as the hole-forming segment in S²¹68 pinholin.

Fig. 4.

ANTS/DPX-based vesicle leakage assay. (A) The timing of the experiment starts when the proteoliposomes are loaded onto a gel-filtration spin column to remove external ANTS and DPX, assuming that the reconstituted hole-forming TMD2 is already fully assembled into active pinholes. Release of the small entrapped fluorophores from the liposomes leads to a dequenching of ANTS, hence an increase in fluorescence is detected, and can be extrapolated back to 0% at t = 0 s. After 900 s, Triton X-100 is added to solubilize the vesicles to define 100% leakage. (B) Extent of leakage from POPC LUVs induced by the TMD2 fragment and by TMD2 glycine zipper mutants at different P:L ratios after 900 s at 22 °C (average values with SD from at least three measurements). The wild-type TMD2 fragment produces massive leakage (dark gray bars), giving already 30% at the lowest P:L ratio of 1:400. Mutating the central Ser₄₄ in the GxxxSxxxG motif to Ala (S₄₄A TMD2) impairs leakage at low concentration, which only partially recovers at higher P:L ratio (gray bars). Destruction of the entire glycine zipper motif (G₄₀A S₄₄A G₄₈A TMD2) essentially abolishes leakage (white bars), giving a maximum of 9% at the highest P:L = 1:50. These data prove 1) the ability of the TMD2 fragment to self-assemble into small holes, for which 2) the extended glycine zipper is of critical importance.

Our rotational analysis of the TMD2 segment revealed that an extended glycine zipper motif (G₄₀xxxS₄₄xxxG₄₈) is freely available for lateral helix–helix interactions, and earlier mutational studies had also highlighted the importance of this motif (9, 11, 13). Therefore, we examined in our vesicle leakage assay the role of the three putative contact-forming residues on TMD2 (Fig. 4B). At first, we mutated the central polar Ser₄₄ to Ala, to inhibit possible polar and H-bonding interactions in the hydrophobic bilayer interior. Indeed, at low peptide concentrations (P:L = 1:400 to 1:150) the S₄₄A TMD2 mutant no longer induces any significant leakage. However, at higher concentrations (P:L = 1:100 to 1:50) an increasing degree of leakage is observed, but never reaching the dramatic effect of the wild-type. In a second step, we additionally made Ala-mutations of both Gly residues flanking Ser₄₄, giving the triple-mutant G₄₀A S₄₄A G₄₈A TMD2, lacking the entire glycine zipper motif. Alanine was chosen instead of even bulkier residues in order to minimize the influence on the overall hydrophobicity of the peptide. The leakage data clearly show that this subtle mutation to alanine nearly completely abolishes TMD2-induced leakage at all concentrations, revealing the functional importance of the extended glycine zipper motif G₄₀xxxS₄₄xxxG₄₈. To rule out any impact of the mutations on the structure and membrane orientation of the peptides, we performed CD measurements, and no significant changes were visible (SI Appendix, Fig. S3).

Determination of the Pinhole Size Using an FD Leakage Assay.

The observed rapid leakage of small fluorophores is consistent with the postulated heptameric assembly of TMD2 in the membrane (9). To estimate the actual diameter of the putative hole, we used yet another modified leakage assay (Fig. 5A), based on FDs of different sizes that are quenched by antifluorescein antibodies (38, 39). Experiments were performed at 22 °C at a constant P:L ratio of 1:100, which is sufficient to induce strong leakage. We reconstituted TMD2 in POPC LUVs in the presence of 6-carboxyfluorescein (CF) (molecular mass = 0.376 kDa), FD4 (molecular mass ∼4 kDa), FD20 (20 kDa), FD40 (40 kDa), or FD70 (70 kDa). After removal of external fluorescent material by gel filtration, an antifluorescein antibody was added to the vesicles. If the pinhole formed by TMD2 is big enough for a certain dextran-coupled fluorophore to pass through, it will leak out from the LUV and become quenched by the antifluorescein antibody. By comparing the leakage of various FDs with different molecular masses, the pinhole diameter can be estimated from the known hydrodynamic radii of the FDs.

Fig. 5.

Dextran leakage assay for estimating the pinhole size. (A) As in Fig. 4, the timing starts when the reconstituted proteoliposomes are loaded onto a gel-filtration spin column to remove external FD. When any FD leaks out from the LUVs, the fluorescence signal will decrease, because an antifluorescein antibody has been added outside to quench the escaped fluorophore. The POPC LUVs are solubilized by the addition of Triton X-100 after 2,500 s. (B) The change in fluorescence was monitored at 525 nm for various FD of different molecular mass at P:L = 1:100 (showing averages of the normalized curves from two independent measurements for each analog). Each signal can be extrapolated back to 100% over the first 200 s (fitted curves: dashed lines), while the sample was on the gel-filtration column to remove nonencapsulated FD. CF (blue) without any dextran leaves the LUVs very rapidly because of its small size, showing comparable leakage as in the ANTS/DPX assay with similarly sized dyes. FD4 (red) with a molecular mass of 4 kDa shows a decrease in fluorescence to 50% within about 200 s, ending up at a 30% level. Larger FDs (FD20 green, FD40 yellow, FD70 purple) only show a minor decrease of less than 10% fluorescence after 2500 s, as they are clearly not able to leave the LUVs through the small pinholes.

The TMD2 fragment was cosolubilized with the lipids, so the leakage started as soon as the nonencapsulated FD was removed by gel filtration. The dashed lines in Fig. 5 depict the first 200 s of the measurement, which were fitted as described in Methods. Fig. 5B shows the decrease of the fluorescence at 525 nm over 2,500 s for the different fluorescein derivatives. CF (Fig. 5B, blue) is the smallest molecule and shows a rapid decrease of fluorescence within 200 s almost down to zero. FD4 (Fig. 5B, red) with a mass of roughly 4 kDa shows an exponential decline in fluorescence, reaching 50% after 200 s and leveling out around 30% at t = 2,500 s. FD4 leaves the LUVs much slower than CF, but still a significant proportion of the FD4 molecules leak out. The larger FDs (FD20, FD40, FD70), on the other hand, only show a marginal decrease of less than 10% fluorescence after 2,500 s. Obviously, these large FDs cannot leave the LUVs effectively. The slight decrease in fluorescence is attributed to the fact that dextrans are not spherical molecules, but ellipsoids (40), which may occasionally slip through the pinhole. Our data thus show that the hole formed by TMD2 has to be smaller than the hydrodynamic radius of FD20, but big enough for FD4 to leave the LUVs. Given the hydrodynamic radii of these fluorescein derivatives, the diameter should be less than 2.8 nm, based on the Stokes radii for FDs provided by the supplier [∼1.4 nm for FD4, 3.3 nm for FD20, 4.5 nm for FD40, and 6.0 nm for FD70 (41); and a Stokes radius for CF of 0.63 nm (42)].

Discussion

During phage morphogenesis, pinholin S²¹68 forms small holes in the inner membrane of Escherichia coli inducing cell lysis. Pinholins consist of two TMDs, of which TMD2 is supposed to be involved in hole formation, while TMD1 has a regulatory function (8, 11–13). So far, there is no experimental structure available for the proposed pinhole. Hence, this study focuses on the hole-forming fragment TMD2, prepared by Fmoc solid-phase peptide synthesis. We investigated its secondary structure, characterized its membrane alignment, and examined its function using several biophysical methods.

SRCD confirmed the helical structure of the TMD2 fragment (Fig. 2). Line shape analysis revealed about 90% α-helix content of our 27-amino acid TMD2 construct (W₃₆ to E₆₂) when reconstituted in DMPC or POPC bilayers. This corresponds to 23 to 24 residues making up the helix, which is in full agreement with the length of the predicted TMD2 segment (W₃₆ to F₅₈), anchored in the membrane by the two aromatic flanking residues. These findings are in accordance with the expected secondary structure (9), given that Drew et al. (18) recently found 83% helix content for the full-length protein. Using SROCD, we could demonstrate that the TMD2 fragment is indeed oriented in a transmembrane alignment in DMPC membranes (Fig. 2B). This result was confirmed by 1D ¹⁵N-ssNMR experiments on four specifically ¹⁵N-labeled peptides containing a single label each, and on one peptide containing all four labels at the same time (Table 1). All of these peptides show a ¹⁵N-chemical shift between 180 ppm and 215 ppm (Fig. 2C), indicating a transmembrane alignment with the helix axis oriented more or less parallel to the bilayer normal.

These results are in accordance with an EPR study by Ahammad et al. (14), which qualitatively proved that TMD2 is inserted in the membrane but did not provide a detailed orientation. Using a 2D ssNMR experiment correlating the ¹⁵N-chemical shift with the ¹H–¹⁵N dipolar coupling (33, 34) (Fig. 3A), we could determine the tilt angle τ of the helix, together with its order parameter S_mol and the azimuthal rotation angle ρ. With τ = 14°, TMD2 is found to be tilted only slightly in DMPC bilayers. The membrane-embedded portion of TMD2 comprises 23 amino acids, corresponding to a length of 34.5 Å (1.5 Å per residue along the helix axis). With a hydrophobic thickness of only 25.4 Å [calculated according to Marsh (43); for experimental data see Kučerka et al. (44)], DMPC bilayers are too thin to accommodate the TMD2 in a completely upright manner, as this would induce a positive hydrophobic mismatch. As a result, a tilting of the helix would be generally expected to avoid exposure of hydrophobic residues at the bilayer surface. Notably, the observed tilt of τ = 14° still seems too small to accommodate the designated TMD2 portion within the hydrophobic region of the DMPC bilayer. However, a look at the putative membrane-embedded sequence (W₃₆AAIGVLGSLVLGFLTYLTNLYF₅₈) shows that at least the C-terminal region carries several polar residues that could reach into the amphiphilic headgroup layer. In addition Ahammad et al. (14) had proposed, based on their EPR studies, that the C-terminal residues N₅₅ and L₅₆ are located in the interface region, and they furthermore proposed a local increase in hydrophobic thickness based on their ²H-ssNMR study (45).

Probably the most informative outcome of our ¹⁵N-ssNMR structure analysis is the finding that an extended glycine zipper G₄₀xxxS₄₄xxxG₄₈ is positioned on the tilted helix such that it faces toward one side. This well-known helix–helix interaction motif (46, 47), has already been highlighted in several mutational studies on pinholin (9, 11, 13). Together with further H-bonding residues (T₅₁, T₅₄, N₅₅), our 3D structure places this strip of polar residues perpendicular to the direction of tilt, making it ideally available for helix–helix interactions (Fig. 3B). In the following paragraph, we will discuss the ensuing implications not only on the structure of a pinholin pore, but also on the assembly of a TMD2 homodimer and the heteromeric TMD1–TMD2 interaction within the folded full-length protein.

An earlier computational analysis of Pang et al. (9) had suggested two possible interaction surfaces in a heptameric pinhole model. They described a hydrophilic interaction surface “A” containing W₃₆A₃₇xxG₄₀V₄₁xG₄₃S₄₄xxL₄₇G₄₈xL₅₀T₅₁xxT₅₄, which resembles quite closely those residues that we propose to be available for interaction. Another more hydrophobic interaction surface “B” from their analysis, on the other hand, contains A₃₈xxV₄₁L₄₂xxL₄₅xxxF₄₉xxY₅₂L₅₃xN₅₅L₅₆xF₅₈. These residues are largely oriented along the tilt direction according to our experimental findings, which would make them less accessible to lateral helix–helix contacts. It has been proposed (9, 11, 13) that pinhole assembly starts from inactive dimers, in which surface “A” is shielded by TMD1. An activated dimer is formed via TMD2, once TMD1 has flipped out the membrane to release surface “A” for homotopic interactions. Subsequently, several dimers with homotopic contacts via “A” are supposed to switch to heterotopic interactions involving surface “B” in a heptameric pinhole.

Our experimental results on the TMD2 membrane alignment (Fig. 6A) support the importance of surface “A” for helix–helix interactions. The laterally accessible glycine zipper provides a continuous right-handed strip (Fig. 6 A and B), which could readily assemble with a second TMD2 helix to form a right-handed dimer (i.e., the proposed active dimer) (9, 11, 13) (Fig. 6E). The modest tilt angle of 14° allows shielding of T₅₁ and N₅₅ along with T₅₄ forming polar interactions with Y₅₂ (Fig. 6E), and probably allows for further stabilization given that V₄₁ fits snuggly in the pocket formed by G₄₃. We cannot provide any experimental information on the actual oligomeric state of the TMD2 fragment in our NMR samples. The order parameter of S_mol = 0.91 suggests that the TMD2 fragment possesses a high uniaxial order. This value of S_mol would be surprisingly large for a single membrane-embedded transmembrane helix, but it would make perfect sense if the peptide forms dimers or assembles into even larger oligomeric structures (48). As we observe massive leakage in our fluorescence assay at a P:L ratio of 1:50 in POPC (Fig. 4), these data suggest that pinholes should also be formed at the same peptide concentration in DMPC, even though the sample hydration and membrane curvature may differ somewhat in the oriented bilayer stacks used for ssNMR.

Fig. 6.

Structural model of pinhole formation based on the ssNMR analysis. Color code of residues: Yellow, hydrophobic; light blue, polar H-bonding; green, Gly. (A) The TMD2 helix is found to be tilted by τ = 14°. Its azimuthal rotation angle positions the hydrophilic strip with the glycine zipper (G₄₀, S₄₄, G₄₈, T₅₁, T₅₄, N₅₅) perpendicular to the direction of tilt, making it available for helix–helix interactions. (B) Helical mesh projection of TMD2 showing the polar strip with the glycine zipper along an otherwise hydrophobic helix. (C) Helical mesh projection of TMD1, which interestingly also contains an extended glycine zipper on one face, and a polar H-bonding strip on the opposite face (seen as light blue residues flanking the mesh on both sides). The helical mesh projections were prepared using the Protein ORIGAMI web application (http://www.ibg.kit.edu/protein_origami/) (31). (D) View of a putative heptameric pinhole, assembled from our observed TMD2 orientation (only five of the seven helices are depicted). In this right-handed architecture, the polar strip containing the glycine zipper motif makes up the hydrophilic lining of the pinhole. Stabilizing knob-in-hole contacts are conceivable between V₄₁-G₄₃, G₄₈-L₅₀, and H-bonding between Y₅₂-T₅₄. (E) Structural model of the active dimer (after externalization of TMD1) based on our ssNMR data, allowing for homotypic TMD2–TMD2 interactions. The glycine zippers G₄₀xxxS₄₄xxxG₄₈ on both helices align with each other giving a right-handed dimer, which could be further stabilized by interactions between G₄₃-V₄₁, T₅₁-T₅₁, and Y₅₂-T₅₄ (F) Predicted model of the inactive full-length pinholin, based on inspection of the TMD1 sequence, where G₄₀xxxS₄₄xxxG₄₈ on TMD2 aligns with G₁₀xxxG₁₄xxxG₁₈ on TMD1. This leaves the opposite polar side of TMD1 free for interactions to form inactive dimers. (G) Process of pinhole formation based on our structural models. Top views (from left to right) of the pore with the blue dot indicating S₄₄ (D), of the active TMD2/TMD2 homodimer (E), and of the inactive full-length dimer (based on F). In the latter, intramolecular TMD1–TMD2 interactions are facilitated by the glycine zippers (green dots), while the polar strips (blue) on TMD1 are available for intermolecular interactions. The arrows indicate the timeline of pinhole formation.

In their previous study, Pang et al. (9) suggested a heptameric assembly according to negative-stain electron microscopy on delipidated purified protein. They checked the stability of different oligomeric states by simulated annealing of TMD2 helices, followed by molecular dynamics simulations. In Fig. 6 D and G we also depict such a heptameric pinhole, which is consistent with the pore size range determined from our leakage data. However, our ssNMR data are clearly compatible only with a right-handed oligomeric structure, and not with the left-handed bundle depicted by Pang et al. (9), which also has a much higher helix tilt angle. That is because according to Fig. 6D the polar stretch containing the glycine zipper motif has to be oriented toward the luminal face of the hole and not toward the hydrophobic outside. Our oligomeric model offers a likely mode of stabilization based on the observation that V₄₁ and L₅₀ fit snuggly into the pocket formed by G₄₃ and G₄₈, respectively, and a further polar interaction can occur between Y₅₂ and T₅₄. Mutations on G₄₃, G₄₈, and T₅₄ did indeed lead to lysis-defective pinholins (11).

Our TMD2 construct lacks the regulating TMD1 segment; however, we may nonetheless speculate on the structure of the inactive pinholin protein (Fig. 6F). It has been shown by mutational analysis (11) that A₃₇, A₃₈, V₄₁, L₄₇, and G₄₈ must be part of the TMD1/TMD2 interface. Except for A₃₈, all of them are positioned around the glycine zipper motif on TMD2, which is freely available for interactions based on our ssNMR structure. Furthermore, the central S₄₄ of the glycine zipper on TMD2 was also shown to interact with TMD1 (13). Interestingly, TMD1 also possesses a glycine zipper motif on one face of the helix, plus a polar strip on the opposite face (Fig. 6C). Based on mutational analysis (11) it has been previously proposed that the side of the membrane-embedded TMD1 helix containing most of the glycine zipper (G₁₀, G₁₄, A₁₇, G₂₁, L₂₈) is facing toward the lipid acyl chains. The interpretation was based on the fact that mutations to leucine delayed or abolished triggering, whereas mutations to glutamine accelerated triggering. Seeing that increased hydrophobicity or polarity delayed or accelerated flipping of TMD1, respectively, it was concluded that these residues should be oriented toward the lipids rather than being engaged in helix–helix interactions. However, the same mutational data could also be simply explained by the fact that it is energetically more unfavorable to expose hydrophobic residues to the hydrophilic environment in the flipped state, irrespective of their orientation in the membrane. Therefore, we propose that the right-handed glycine zipper (G₁₀xxxG₁₄xxxG₁₈) on TMD1 can readily interact intramolecularly with the freely available G₄₀xxxS₄₄xxxG₄₈ motif on its own TMD2 (Fig. 6F). This interaction makes the opposite, polar surface of TMD1 freely available for intermolecular polar interactions with another inactive pinholin protein, forming the proposed inactive dimer (Fig. 6 G, Right). Interestingly, A₁₂ lies in the middle of this polar surface, and it has been shown by mutational analysis (11) that this residue compromised the antipinholin character of irsS²¹68_A12L when it is present together with the wild-type protein (irsS²¹68 is a mutant with a strong antipinholin character, preventing the flip of TMD1 in a heterodimer with pinholin), which proves that it must be engaged in an intermolecular contact. Notably, our heterodimeric model places all helices in a straight line, unlike the helix bundle illustrated by Pang et al. (11).

The structural model of the inactive pinholin protein in Fig. 6F offers another interesting detail. Based on the proposed interaction, the TMD1 is not able to span the lipid bilayer completely, leading to the “problem” that its charged N terminus is pulled toward the bilayer interior, which is energetically unfavorable and should thereby destabilize the transmembrane alignment of TMD1. This tension would nicely explain the role of the elongated antipinholin sequence. So far it had been postulated that the additional charge on the N terminus of antipinholin delays the flipping (8) (the N-terminal residues preceding TMD1 are MDK in pinholin, and MKSMDK in antipinholin). However, it is questionable whether a single additional charge should evoke such tremendously different behavior (even if D and K are engaged in an intramolecular salt bridge). It is more likely that the increased length allows the charged NH₃⁺ terminus to reach out of the membrane, thereby stabilizing the transmembrane alignment of TMD1. It seems that introduction of the polar Ser and charged Lys residues, which can snorkel into the headgroup region, is a good choice by nature, because it does not increase the hydrophobicity and thereby still sustains a delayed flipping behavior.

Using the ANTS/DPX vesicle leakage assay, we examined the proposed function of TMD2 as the hole-forming segment. The fact that the peptide causes massive leakage of the vesicles at high concentration (P:L = 1:50) would be compatible with many possible mechanisms of action, from micellization to well-ordered hole formation. However, we found that TMD2 induces considerable leakage in POPC LUVs already at a very low P:L ratio of 1:400 (30% within 900 s) (Fig. 4B). At such a low concentration, there is not enough peptide for micellization, hence it seems more likely that the peptides actually self-assemble into ordered structures, like pinholes. Pang et al. (11) have already shown that a ΔTMD1 mutant is biologically active. Here we present experimental evidence that the TMD2 fragment per se is able to form holes in lipid vesicles using biophysical approaches. These results strengthen the hypothesis that TMD1 plays only a regulatory role: That is, they support the model that TMD1 has to flip out of the membrane in order to allow TMD2 to form the pinholes (9, 11, 17).

Furthermore, we have addressed the importance of the glycine-zipper motif for TMD2-induced vesicle leakage. As already suggested by mutational studies (9, 11, 13), we could prove that the glycine zipper is indeed crucial for pinhole formation. The TMD2 mutant G₄₀A S₄₄A G₄₈A, lacking the entire glycine zipper motif, nearly completely abolished leakage at all tested peptide concentrations, while the single mutant S₄₄A still induces a considerable degree of leakage at high peptide concentration (Fig. 4). The latter activity might be attributed to the fact that the central polar Ser₄₄ seems to be, based on our structural model, not directly engaged in building up the pinhole. Ser₄₄ points centrally into the luminal face of the pinhole, as seen in Fig. 6G, whereas G₄₈ is proposed to stabilize the wall of the pinhole. Of course, Ser₄₄ plays a central role in building up the active dimer. Because of crowding at high peptide concentration, the specific formation of an active dimer beforehand would probably not be as critical as it must be at low peptide concentration. Therefore, the leakage induced by the S₄₄A TMD2 mutant at high concentration might be simply explained by lateral crowding of the reconstituted peptides in the lipid bilayer. This interpretation would support the critical role of protein concentration and active dimer formation in the proposed timeline of pinhole assembly (13) (Fig. 6G). It is also in full accordance with an earlier fluorescence study visualizing pinholes in vivo, where a S₄₄C mutant showed a uniform distribution within the membrane, whereas the wild-type protein accumulated into rafts after triggering (17).

Finally, we also investigated the size of the resulting holes, using a modified leakage assay with FITC-FDs of different molecular masses. So far, electron microscopy and molecular dynamics simulations have suggested that holes are formed with a diameter of around 1.5 nm (9). If this assumption is correct, FDs larger than this should not be readily able to leave the LUVs. Indeed, we found that CF with a diameter of 1 nm (i.e., a similar size as ANTS and DPX, which were used in the other leakage assay) can pass freely through the holes. Our data show that up to 70% of labeled FD4 can slowly leave the LUVs over a period of 2,500 s (Fig. 5B). Any larger FDs showed only negligible leakage. Based on the estimated hydrodynamic radii of the used dextrans, we thus found that the diameter of the hole should not be larger than 2.8 nm in diameter (given the 1.4-nm hydrodynamic radius of FD4), which is in accordance with the predicted diameter of 1.5 nm (9, 11).

The reason why the determined hole size based on the estimated hydrodynamic radii of the FDs gives a slightly larger nominal value can be attributed to the fact that the actual shape of the dextrans is more elliptical than spherical (40), so the radii provided by the manufacturer are not perfectly well defined. An ellipsoid has two radii, one along the major axis and a second one along the minor axis. For FD4 the radius of the long axis has been determined as 2.37 nm, and 1.08 nm for the short axis (49). Therefore, the hole seems to be just about big enough to let FD4 pass through, head-on with an effective diameter of ∼2 nm around its long axis. Furthermore, the product specifications of the purchased FD4 reveal that a mixture of FD3, FD4, and FD5 is provided as “FD4,” so most likely only FD3 leaves the pinhole to a significant amount. The pinhole should thus be large enough to allow the passage of small molecules (e.g., ions), but too small for proteins with a typical size of 3 to 6 nm (50).

Taking these data together, we have presented here the detailed membrane alignment of the pinholin TMD2 fragment. We confirmed its helical conformation and elucidated its 3D alignment in DMPC membranes, which positions the functionally important glycine zipper sideways. We provide structural models for a right-handed oligomeric pinhole, for a right-handed active homodimer, and for the inactive full-length pinholin, based on our experimental ssNMR data. Furthermore, we have provided first-hand evidence for the hole forming properties of TMD2 per se in POPC vesicles and estimated the diameter of the formed pinholes to be smaller than 2 nm.

Methods

A detailed description of all experimental procedures and data processing protocols used in this study are available in SI Appendix, Supplementary Material and Methods.

Chemical Peptide Synthesis.

All peptides were synthesized on an automated parallel peptide synthesizer Syro II (MultiSynTech) using standard solid-phase Fmoc protocols. For ssNMR analysis, specific leucine residues were replaced nonperturbingly with ¹⁵N-leucine. Table 1 provides an overview of all synthesized peptides.

SRCD and SROCD Spectroscopy.

For SRCD experiments, the pinholin TMD2 peptide was reconstituted in SUVs and an aliquot was filled into a “Birkbeck-type” demountable calcium fluoride cell (51). For the SROCD sample, the TMD2 peptide was reconstituted in macroscopically aligned DMPC bilayers (P:L = 1:50) on quartz glass plates. Afterward, the sample was positioned perpendicular to the light beam in the SROCD measurement cell.

All SR(O)CD measurements were performed on the UV-CD12 beamline at the synchrotron radiation source Karlsruhe Research Accelerator at the Karlsruhe Institute of Technology, as previously described (23, 24), with experimental set-ups as described by Clarke and Jones (52) and Bürck et al. (19). To remove the background CD signal of the lipids, the averaged spectrum of a pure lipid sample was subtracted from the peptide spectra.

The secondary structure was estimated using the DICHROWEB online server (53, 54).

Solid-State NMR Spectroscopy.

The peptide was reconstituted in macroscopically aligned DMPC lipid bilayers. All samples were prepared with a P:L ratio of 1:50 (mol:mol).

NMR spectra were recorded on a Bruker Ultrashield 600 MHz spectrometer (Bruker Biospin) using a modified Bruker HFP flatcoil probe for ³¹P measurements, and a home-built HX lowE flat coil probe for ¹⁵N.

SAMMY experiments (30) employed the improved SAMPI4 pulse sequences (29) for ¹⁵N-chemical shift/¹H–¹⁵N-dipolar coupling correlation spectra. For displaying and processing of the data, TOPSPIN (Bruker) was used. By comparing our experimental data to simulated PISEMA spectra, we extracted the underlying tilt angle τ, the azimuthal rotation angle ρ, and the order parameter S_mol. The simulated spectra were calculated on the basis of an ideal α-helix (with torsion angles Φ = 60.7° and Ψ = 44.7°).

ANTS/DPX-Based Vesicle Leakage Assay.

For standard fluorescence leakage experiments, LUVs entrapping the fluorophore ANTS and the quencher DPX were prepared (37).

All fluorescence measurements were performed on a Fluorolog spectrofluorimeter (HORIBA Jobin Yvon). The ANTS emission was set to 515 nm (slit 4 nm) and its excitation to 355 nm (slit 4 nm). A long-pass filter (GG395, 3 mm; Schott) was used to make sure the excitation light is blocked out during detection. A pure lipid sample was recorded and used as a reference for the baseline. To obtain 100% leakage, Triton X-100 (0.27% of total sample volume) was added and the percentage of leakage was calculated as the fluorescence after 900 s divided by the fluorescence signal after the addition of Triton.

FD Leakage Assay.

For dextran leakage assays, fluorescein-coupled dextran derivatives of different sizes were entrapped in POPC LUVs. Antifluorescein antibodies (antifluorescein/Oregon green, rabbit IgG fraction; Invitrogen, Thermo Fisher Scientific) outside the vesicles quench the fluorescence of the fluorescein (38, 39). For the assay we used CF and the fluorescein isothiocyanate-dextrans FD4 (∼4 kDa), FD20 (20 kDa), FD40 (40 kDa), FD70 (70 kDa) (Sigma-Aldrich). Again, the Fluorolog spectrofluorimeter was used for the measurements. The emission of fluorescein was monitored at 525 nm with a 14 nm slit and a long-pass filter (GG495, 3 mm; Schott) blocking the excitation light (excitation 490 nm; slit 1.25 nm). For this assay, the fluorescence signal decreases when dye leaks out and is quenched by the antibodies. The measurement was terminated by addition of Triton X-100 (0.27% [vol/vol]). A sample without peptide was recorded and used to determine the background signal. To evaluate the amount of leakage, the initial signal is taken as 0% leakage, and the value after addition of Triton X-100 is taken as 100% leakage.

Data Availability.

All study data are included in the article and SI Appendix.