Determination of the boundary lipids of sticholysins using tryptophan quenching

Sticholysins are α-pore-forming toxins produced by the sea-anemone Stichodactyla helianthus. These toxins exert their activity by forming pores on sphingomyelin-containing membranes. Recognition of sphingomyelin by sticholysins is required to start the process of pore formation. Sphingomyelin recognition is coupled with membrane binding and followed by membrane penetration and oligomerization. Many features of these processes are known. However, the extent of contact with each of the different kinds of lipids present in the membrane has received little attention. To delve into this question, we have used a phosphatidylcholine analogue labeled at one of its acyl chains with a doxyl moiety, a known quencher of tryptophan emission. Here we present evidence for the contact of sticholysins with phosphatidylcholine lipids in the sticholysin oligomer, and for how each sticholysin isotoxin is affected differently by the inclusion of cholesterol in the membrane. Furthermore, using phosphatidylcholine analogs that were labeled at different positions of their structure (acyl chains and headgroup) in combination with a variety of sticholysin mutants, we also investigated the depth of the tryptophan residues of sticholysins in the bilayer. Our results indicate that the position of the tryptophan residues relative to the membrane normal is deeper when cholesterol is absent from the membrane.

Vesicle preparation. Dry lipid films were prepared from the chosen organic lipid solutions combined in the desired molar proportion, followed by evaporation of the organic solvent under nitrogen flow at 40 °C. These films were then hydrated in buffer (10 mM Tris, 140 mM NaCl, pH 7.4) at 65 °C in a water bath for at least 30 min and suspended by gentle vortex. This method yields symmetric bilayers arranged in multilamellar vesicles, with all lipid species randomly distributed between the leaflets. Suspended lipid vesicles were then extruded at hydration temperature through polycarbonate filters to obtain large unilamellar vesicles (LUVs) of 200 nm average size. According to the manufacturer (Avanti Polar Lipids 57 ), this system is expected to yield samples with low polydispersity, if filters with pore size ≤ 200 nm are used. DLS measurements have confirmed this previously 58 and for the current samples (data not shown). Vesicle symmetry is expected to be maintained in the time Indicated are all five Trp residues of the protein, plus I6 and K67, which have been mutated to Cys to form a disulfide bridge that prevents the deployment of the N-terminal α-helix. Of all five Trp residues, only W110 and W114 (W111 and W115 in StnI) are inserted in the membrane 35,37 . W43 and W115 are buried in the protein core, and W146 is known to participate in protein-protein interactions upon oligomerization 10,37,39   [DMSO] sample was < 2.5% (v/v) to minimize possible effects on proteins or membranes. A lipid/protein (L/P) ratio of ~ 180 (1.6 mM and 9 µM, respectively) was used to hinder the potential crosslinking of non-interacting protomers. One can estimate the mean distance between monomers (or oligomers) based on their surface density 33,60 , as r av = √ σ 2 , where r av is the average distance between the particles, and σ is their surface density. This can be estimated from the L/P ratio, and the average lipid cross-sectional area. We used average cross-sectional areas values of 45 -60 Å 2 , depending on the membrane composition (see [61][62][63][64] for lipid coss-sectional areas). At this L/P ratio, the mean distance between free monomers on the membrane surface should be at least ~ 25 Å if the radius of the protein is also considered, being larger in less condensed membranes, with less or no Chol. Mean distance would increase to > 100 Å for octamers. Based on time scans performed using StnII WT with DOPC:PSM:Chol 1:1:1 as control (Fig. S3a), the reaction time with DSS was 10 min. The reaction was stopped using a 2 M solution of glycine, added to a final Gly/DSS ratio of 126, followed by 10 min incubation. Electrophoresis loading buffer with 0.5% (v/v) β-mercaptoethanol was added, and the samples were boiled for 20 min. Cross-linked products were analyzed using standard PAGE-SDS procedures 65 . Control experiments were also performed in absence of membranes to test if DSS produced intra-molecular adducts that affected the electrophoretic mobility of the toxins (Fig. S3b).
Time-resolved fluorescence measurements. Time-resolved fluorescence measurements were performed using a FluoTime100 spectrofluorimeter, equipped with a PicoHarp300E time-correlated single photon-counting module (PicoQuant GmbH, Berlin, Germany). A 297 ± 15 nm pulsed diode laser was used for excitation. Emission was collected through a long-pass filter (> 345 nm) to avoid stray light. When required, neutral density filters were used to attenuate excitation intensity. The instrument response function (IRF) was acquired using light scattered by a LUDOX solution in absence of colored filters up to ~ 10,000 counts in the peak channel. Sample decays were recorded up to ~ 12,000 counts in the peak channel. Inner filter effects were avoided by using fluorophore concentrations such that the optical density at the excitation wavelength was OD l/2 < 0.05. Constant stirring was maintained during all experiments. A Peltier element was used to ensure steady temperature. Details on the analysis are provided in the supporting information (S.I.) file.
Steady-state fluorescence spectroscopy. Steady-state fluorescence measurements were performed on a PTI Quanta-Master spectrofluorimeter (Photon Technology International, Lawrenceville, NJ, USA). Sample excitation was 295 nm to specifically excite Trp residues. Emission was recorded between 305 and 450 nm, at 1 nm intervals. Polarizers were used to minimize light scattering in the signal and to avoid distortions in the acquired spectra. The configuration was Em pol = 0°, Ex pol = 90°, as recommended these ends in 66 . Again, inner filter effects were avoided by using fluorophore concentrations such that OD l/2 λex < 0.05. Experiments were performed in PBS (10 mM phosphate, 140 mM NaCl, pH 7.4). Constant stirring was used. The temperature was controlled by a Peltier element. When the temperature effect was evaluated, the gradient was 5 °C/min. The fluorescent signal was corrected for inner and outer filter effects produced by vesicle-induced light scattering, as indicated in 66 , using the same amounts of toxin and lipids as in the original experiments, but with LUVs composed solely of POPC, to which no toxin binding could be detected. Details on the analysis of the shape of the emission spectra of Trp are provided in the S.I. file.
Toxin titration with quencher-free vesicles. To properly evaluate Trp quenching, complete membrane binding by the toxins was required. All toxin variants (100 nM) were titrated with quencher-free vesicles of the desired compositions to saturation. The results were used to estimate the fraction of bound protein as a function of lipid and protein content in the sample. Analysis was performed according to as described in 67,68 . In this equation, θ is the fraction of bound protein, which is a function of the total lipid and protein concentrations, [L T ] and [P T ] respectively. The values of [L T ] and [P T ] were corrected for dilution effects caused by LUV addition to the sample. K a is the association constant, and n is the stoichiometry of the interaction, understood as one monomer of protein binding to LUVs every n lipid molecules (without any implication on the possible direct interactions).
In all cases, goodness of fit was evaluated, at least, according to the residuals trace, the root mean squared deviation (RMSD), and the value of χ 2 red .
Analysis of tryptophan exposure to lipids. The exposure of a membrane-bound fluorophore to surrounding lipids can be studied using fluorescence quenching 40,[69][70][71][72][73] . Using increasing amounts of quencher-labeled lipids in the membrane, the number of boundary lipids that a membrane fluorophore is in contact with can be estimated from ) is the fluorescence signal measured at a given mol% of quencher in the membrane, F 0 is the emission of the sample in absence of quencher, F min is the intensity observed when the protein is in pure quencher lipid, and L n is the number of lipids that can establish direct contact with the studied fluorophore. F([Q]) was normalized as F/F 0 . Detail of the analysis is provided in the S.I. file.
Depth-dependent fluorescence quenching of tryptophan. The position of a membrane-embedded fluorophore along the membrane normal can be estimated using quencher-labeled lipids in which the quencher group is located at varying positions along the acyl chains of the lipids, or at the headgroup [43][44][45][46][47][74][75][76][77] . The fluorescence intensity of the sample, or the lifetime, can be measured in the absence and presence of each of the quenchers. The depth-dependent quenching profile, QP(h), defined as QP(h) = F 0 /F-1 for intensity measurements 77 , can be obtained from these data. The QP(h) can be approximated, according to the distribution analysis, with two symmetrical Gaussian functions, as where h m is the center of the profile measured from the bilayer center (h = 0), indicating the average depth of the fluorophore in the membrane; σ is the distribution width, and corresponds to the fluctuations in h; and S is the area, related to fluorophore accessibility. The bilayer center is the symmetry axis of the mirrored Gaussian functions, which are required to account for trans leaflet quenching on deep fluorophores 43 . Values of membrane quencher depths from the bilayer center were 18.2 (TEMPO-PC), 12.1 (5-SLPC), 11.5 (7-SLPC), 10.1 (10-SLPC), 6.4 (12-SLPC), and 2.9 Å (14-SLPC), from 78 , where these were obtained in pure POPC bilayers, being the best available approximation.

Results
Toxin titration with quencher-free vesicles. All toxins were titrated with quencher-free vesicles of each of the selected compositions. This allowed us to estimate the stoichiometry of each toxin-membrane interaction ( Table 1 and Fig. S4). In all cases, based on fluorescence emission of each toxin variant, saturation is reached at L/P molar ratio < 80. Thus, for subsequent experiments, a L/P molar ratio of 120 was judged as enough to ensure complete binding in all cases. At the L/P ratio, the average distance between free monomers would be ~ 25 Å, increasing to ~ 100 Å if octameric pores are formed (not considering the surface corresponding to the lumen of the pores). It is interesting to compare the stoichiometry and affinity of the non-competent mutants, StnI-I6C/K67C ox and StnII-I7C/K68C ox , with that of the WT variants. In presence of Chol, when the l d and l o phases are present, the parameters obtained are very similar for all variants of each toxin isoform, with the stoichiometry of the mutants being slightly higher. Only StnII-I6C/K67C ox displays a higher affinity than its WT counterpart. Unexpectedly, when Chol is absent and the membrane is in the l d phase, the affinity of these mutants is much larger (~ 6x) than that of the corresponding WT variant, despite their inability to deploy the N-terminal α-helix.
Regarding the Trp residues, W114 seems to play a much more relevant role than W110 in maintaining binding affinity, despite W110 being more solvent exposed in the soluble structure. StnII-W43/110F behaves much like StnII-WT, just with a somewhat reduced membrane affinity toward Chol-containing membranes. However, the substitution of W114 in mutant StnII-W43/114F results in a toxin that can bind membranes with Chol but whose Trp emission is not increased when titrated with Chol-lacking membranes.
In all cases, the increase in fluorescence emission, calculated as F bound /F solution for each variant (Table 1), is much larger when Chol is present in the membrane.  (Fig. 2). This suggests that the conformation of the N-terminal α-helix is not relevant regarding adduct formation during the crosslinking reaction. The differences between isoforms probably reflect their distinct reactivity with DSS. The split bands observed at the expected position of the monomers (Fig. 2) together with the occurrence of the same bands in the controls (Fig. S3b) indicate that StnI is more prone to form intramolecular adducts and, overall, more reactive. Hence, exposure of oligomers to DSS would result in faster crosslinking and, consequently, higher abundance of low mobility adducts, as revealed by the different profiles observed for StnI and StnII and their corresponding mutants shown in (Fig. 2).

Quenching of tryptophan emission by spin-labeled lipids: Effects on membrane-binding and
Trp emission intensity, shape, and lifetime: effect of temperature. The emission at 330 nm under 295 nm excitation was recorded to test the effect of spin-labeled lipids on membrane binding. In all cases, a time-dependent increase in the emission as a response to lipid addition to the sample was observed (Fig. S5). The kinetic parameters of the interactions of each toxin variant with vesicles of each composition (Fig. S6) show that all Stn variants were faster in Chol-containing membranes. In these cases, the dominant component was the fastest one (α 1 > ~ 0.8; τ 1 < 5 s, typically). A second component (with τ 2 ~ 50-70 s) was also detected in all cases. Increasing the mol% of 7-SLPC in membranes with Chol did not significantly perturb the kinetics. Only those of StnII-W43/114F appear to be affected by the quencher (Fig. S6).
The binding process to membranes without Chol was slowed down by increasing 7-SLPC. The fastest component was observed to become less dominant. Interestingly, this did not occur for the inactive mutants, for which only the secondary process seemed to be affected.
To ensure that the same degree of binding was achieved regardless of the presence of 7-SLPC, the shapes of the emission spectra (analyzed using eq. S2.) were compared. These spectra were recorded after the time-dependent signal had plateaued. Only a small quencher-dependent broadening (< 2 nm) and red-shift (0-2 nm) of the spectra was observed for the WT, the inactive mutants, and StnII-W43/110F in Chol-containing membranes (Fig. S7).
In the other instances, the emission of the Trp mutants was broadened and red-shifted (~ 10 nm and ~ 6 nm, respectively). The combination of spectral and kinetic data suggests that, while increased 7-SLPC content in the membrane can affect membrane binding, it only slows down the process, without preventing it. The spectral differences observed are likely a consequence of the lost contribution of the quenched Trp residues to the total spectra of the toxin, which maintain the emission of the shielded Trp.
Quenching of Trp emission by spin-labels has been shown to be essentially collisional rather than a static process 46,76,79 . However, given that the diffusion coefficients of lipids in membranes are 10 -7 -10 -8 cm 2 s -180 , molecules with an excited state lifetime < 50 ns would experience less than one collision during the excited state unless already in contact with the quencher 40 . In fact, integral membrane proteins appear to diffuse together with annular lipids 81 , reducing the effective diffusion coefficient of those lipids. Hence, the observed quenching would appear as static quenching, even if the intrinsic nature of the process is collisional. This is the basis of the model with which the number of lipids that surround a fluorophore can be estimated through fluorescence quenching 40,70,71 .
To test quenching of Stn's Trp by spin-labels, the emission of the sample was measured using steady-state and time-resolved spectroscopy. Three lifetime components of about 8, 3, and 1.5 ns were required to obtain satisfactory fits to the fluorescence intensity decays. A plot of τ 0 /τ (intensity-weighted average lifetimes) obtained without (τ 0 ) and with increasing amount of quencher (τ), revealed, in most cases, a dynamic component of quenching (Fig. 3). However, the F 0 /F values obtained using steady-state data are much larger, indicating that quenching www.nature.com/scientificreports/ occurs, mainly, by a process equivalent to static quenching. In nearly all cases, results displayed a downward curvature, away from the y axis, with increasing mol% of 7-SLPC. This deviation from the typical Stern-Vollmer behavior is expected when only part of the total fluorophore population is accessible to the quencher, as in Stns (see Fig. 1) 69,82,83 . Lifetime data indicates that collisional quenching is minimal or does not occur in the situations studied. However, given that all Stn variants used have three or more Trp residues, and that even single-Trp proteins often display multiple exponential decays 69,84,85 , it was possible that the shielded Trp residues could mask the reduced lifetime of the quencher-exposed Trp residues. Hence, given that collisional quenching increases with temperature, the steady-state emission was also recorded for the samples containing 30 and 70 mol% of 7-SLPC while increasing the temperature of the sample (Figs. S8 and S9). In most cases for Chol-containing membranes, quenching increased with temperature, though not steadily. This suggests that, while collisional quenching may occur, it depends on membrane organization around the protein complexes, with the main mechanism being static. In absence of Chol, quenching does not vary, and, in some cases, it even diminishes with increasing temperature. Rearrangements in the vicinity of the toxin oligomers may diminish Trp-quencher contacts.
Tryptophan exposure to PC lipids. Given that most of the quenching observed for all Stn variants used could be interpreted as static quenching, we proceeded to analyze the steady-state data presented in Fig. 3 according to Eq. 2. These analyses (Figs. S10 and S11) allowed us to estimate the fraction of the fluorophore population that is inaccessible to quencher and, most importantly, the number of PC lipids in contact with accessible Trp residues (Table 2).
Results indicate that, for most Stn variants, the inaccessible fluorophore population would be responsible for 60-75% of the Trp emission. StnII-W43/110F is an exception, showing that W114 is mostly protected from 7-SLPC.
Trp accessibility results are most interesting. StnI appears to be exposed to 2.5 PC lipids in both membrane systems studied. However, the presence of Chol greatly affects the accessibility of StnII's Trp. When Chol is absent, StnII has contact with > 4 PCs. However, in Chol-containing membranes, only ~ 1 PC molecule can reach the Trp residues of StnII. The inactive mutants show the same behavior as the corresponding WT protein. The values obtained for StnII-W43/110F are not very reliable, possibly due to the low resolution allowed by the small

Position of the Trp residues of Stns relative to the membrane normal. Quenching experiments
using spin-labeled lipids that differ from each other in the position of the quencher group were performed to estimate the membrane depth of the Trp residues of Stns. For this, 30 mol% quencher lipid was included in the corresponding membrane system. The lipid arrangement in the pores formed by actinoporins is unclear. Toroidal and non-toroidal architectures have been proposed 10,32,86,87 . Hence, since the depth of the quenchers has been measured for undistorted membranes 46,47,77 , the success of this approach depended on the existence of a membrane distortion induced by a toroidal architecture and the extension of this effect.
Results shown in Fig. 4 suggest that the lipid acyl-chains still are more or less parallel to the membrane normal, at least in the vicinity of the Trp residues of the competent proteins. It appears that the average position of the Trp residues of the WT proteins is slightly deeper (~ 2 Å) when Chol is absent from the membrane, according to fitting results (Table 3). Though not as clearly, this also occurs to StnII-W43/110F. Results from the Trp mutants indicate, as expected, that W110 is placed deeper in the membrane than W114. Not only that, but W110 also seems to be much less protected by either the protein structure or by other lipids.
Data from the inactive mutants is not so clear, but still seems to support that acyl-chain orientation is not affected near the Trp side chains. In agreement with results from previous sections, the Trp residues of the double-Cys mutants seem to be more exposed to the quenchers. Again, this seems reasonable given that, for these proteins, the volume that corresponds to the pore-lumen in the rest of the toxins is not devoid of lipids nor occupied by the N-terminal α-helices of these proteins.

Discussion
In this study, the lipid environment of the membrane-embedded Trp residues of Stns has been examined in membranes with and without Chol. We began by studying the interaction of six different Stn variants, two WT, and four mutants, with membranes of two different compositions, DOPC:PSM 80:20 and DOPC:PSM:Chol 70:15:15.
As observed for other lipid compositions 37 , the substitution of W114 by Phe results in a mutant (StnII-W43/114F) that is essentially unable to bind to Chol-lacking membranes. Given the different phase state of the two chosen compositions (see Fig. S1), it could be hypothesized that W114 plays a role in sensing the phase state of the membrane. The fact that the absence of W114 prevents membrane binding when Chol is absent too suggests that either Chol or W114 are required for the recognition of the SM-headgroup. In fact, W116 of EqtII has been predicted to participate in a cation-π interaction with the SM-headgroup 25 . This interaction should be present whenever SM is present in the membrane. The relevance of this interaction is highlighted by StnII-W43/110F, and not StnII-W43/114F, being capable of binding DOPC:PSM 80:20 membranes.
The larger increment in the fluorescence emission when the membranes used contain Chol, and this being even larger for the inactive mutants, is also revealing. This result is probably a consequence of the acyl-chain order near the Trp residues. Lipids in the l d phase allow more water molecules across the membrane [88][89][90] , which can result in fluorescence quenching by water despite the toxins being membrane-bound. Given that the Cholcontaining membranes used in this study are in a mixture of l d and l o phases, this result also suggests that pores in these membranes find themselves in an environment that is more like the l o phase than the l d phase in which the Chol-lacking membranes are. Following the water-quenching argument, the further increased emission of the oxidized double Cys mutants could result from the absence of a pore lumen, regardless of the oligomerization state of these toxin variants. The Trp residues of these mutants would be less accessible to water since the membrane integrity is less affected by their binding than by that of a fully functional Stn that penetrates the membrane with its N-terminal α-helix. Table 2. PC-lipid contacts. Number of PC lipids (L n ) in contact with the membrane inserted Trp residues of the indicated Stn variants in membranes with and without Chol. The value of the non-quenchable fluorescence (F min ) is also indicated. Values are result ± SD, obtained by fitting procedures. www.nature.com/scientificreports/ Regarding the observed binding kinetics reported by Trp emission, the fastest resolved component could be attributed to the water-membrane transition, whereas the slower one could be related to oligomerization and the rearrangements leading to the final pore structure. The main effect of increasing 7-SLPC in membranes without Chol on the binding kinetics was reducing the amplitude of the fast component. This could be due to a diminished initial raise in fluorescence emission, which occurs in absence of quencher by the solvent-to-membrane transition of the toxins. This effect would be smaller in Chol-containing membranes, as the exposure to PC lipids is also diminished in that case.
Though minimal in most cases, quencher-induced changes in the shape of the emission spectra were significant for StnII-W43/114F and StnII-W43/110F. Since a plot of FWHM vs λ max of these data (Fig. S7) follows the  Table 3. Fitting results of the depth-distribution analysis. Values were obtained from fitting Eq. 6 to the quenching results shown in Fig. 4. Parameters are average depth (h), width of the distribution (σ), and area (S), related to fluorophore accessibility. Indicated data for StnII-W43/110F in presence of Chol were obtained using data from Chol-lacking membranes as guidance since the initial fit was unsuccessful. Data shown for StnII-W43/114F are from the dashed line fit in Fig. 4 www.nature.com/scientificreports/ same trend as the data obtained by titration with quencher-free vesicles (data not shown), we observe that the spectra are essentially the same if W110 and/or W114 are quenched, regardless of whether it is by water or by 7-SLPC. Not only that, but also that W110 and W114 are the only residues affected by the presence of 7-SLPC. Despite Trp quenching by spin-labels being of dynamic nature 46,76,79 , for Stns, this shows itself as static quenching. Despite the intrinsic nature of the mechanism being dynamic, observed quenching can still appear as static if the quencher participates in a molecular complex in which the residence time of the components is long enough so that replacement of these molecules is unlikely to happen during the excited state of the fluorophores involved. The fact that in some cases quenching diminishes with increasing temperature suggests looser complexes that can dissociate with the effect of temperature. In fact, the traces that show a reduction in quenching are precisely those of StnI-I7C/K68C ox and StnII-I6C/K67C ox , precisely the mutants whose interaction with lipids might not be as tight given their inability for membrane penetration. This would allow for an easier rearrangement, with a possible preference towards SM, given the higher T m of this lipid.
The measured exposure of the Trp residues of Stns is interesting. Our results in absence of Chol would support recent molecular dynamics simulations showing that the lipid environment of FraC octamers would not display a preference for SM over PC 36 . The situation when Chol is present might be different. In principle, if Stns did not display any selectivity, results should be about the same regardless of membrane composition. However, if specific lipid-protein interactions that prevent potential quencher-Trp contacts are favored, results should reflect the presence of the preferred lipid in the membrane. In this regard, the behavior of StnI and StnII seems to be strikingly different. While StnI-quencher contacts are apparently the same in both membrane systems used, a very different situation is found for StnII, which is much more accessible to PC lipids when Chol is absent in the membrane. This is most likely an outcome of Chol standing in the way of the quencher, perhaps in an interaction also involving SM. This would agree with previous observations 35 , supporting past results indicating that the activity of StnII is much more favored by Chol than StnI's 24,91,92 . Interestingly, this behavior, though helpful, in principle, for pore formation, appears to be independent of helix deployment since the inactive mutants appear to establish the same interactions as the WT proteins. A combination of these with the results obtained for the Trp mutants seems to indicate that W114 is the residue that would be most shielded by Chol. Not only that but its presence would also be required to establish the interactions with SM and Chol since its removal indicates that W110 is much more exposed to PC lipids. Since Trp residues have been shown to be unable to specifically interact with Chol by themselves 64 , Chol recognition would require W114 together with other residues.
This preferential interaction with Chol is not observed, in terms of quencher-shielding, for StnI. Since StnI also has an identical Trp residue (W115), the difference with StnII must lie somewhere else. Of the twelve residues that vary between StnI and StnII, only two are on the loops that directly interact with the membrane: D76 and T77 of StnII, and S77 and S78 of StnI. These residues are on the opposite loop to W110 and W114 (see Fig. 1). Depending on the exact nature of the interaction, oligomerization might be required if the mentioned residues all take part in the selectivity for Chol. This possibility cannot be ruled out, given that, according to our crosslinking analysis, StnII-I6C/K67C ox also forms oligomers. The acidic character of the Asp residue in contrast with the polar character of Ser might tune these interactions, modulating what the lipid population is in the first shell of annular lipids of these pores.
Depth-dependent quenching did not report significant differences, though all proteins seem to have their Trp residues closer to the membrane midplane in membranes without Chol. It could be argued that this might be an effect of Chol on the depth of the quencher groups. However, Trp depth is measured relative to these quenchers. Hence, while the measurement may not be precise in terms of absolute distance to the bilayer center (in Å) because the actual depths of the quenchers might not the same as those measured in POPC 78 , these results do show that Trp residues of Stns are at a deeper position in the membrane relative to the acyl chains of the modified PC molecules. As expected, our results for the Trp mutants of StnII indicate that the average position of W110 is deeper than that of W114.
The good fits obtained in most cases also indicate that, if Stn pores are toroidal, lipid acyl-chains are still mostly parallel to the membrane normal in the region surrounding the Trp residues (Fig. 5). One of the goals of this analysis was to detect membrane curvature induced by pore formation if this in fact occurred. The results obtained do not give many hints in this regard. Nevertheless, together with past results on the angle of the N-terminal α-helix 29,93 , these new data do suggest that the outer leaflet of the vesicles, the one on which binding occurs, is unlikely to be curved (Fig. 5).

Conclusions
Taken together, our results show that, regardless of their specific recognition of SM, Stn pores are also in contact with PC lipids. However, for StnII, but not for StnI, these contacts are reduced when Chol is present in the membrane. We propose that this preference, in which W114 may intervene, could perhaps be related to the difference between these toxins that occur in one of the loops that they insert in the membrane. Finally, it appears that acyl-chain orientation is not significantly altered by Stns, at least in the zone surrounding the membraneembedded Trp residues of these proteins.

Data availability
The data supporting the findings of this study are available from the corresponding author (juan.palaciosb1a@ gmail.com) upon reasonable request. The angle with the α-helix with the membrane normal is ~ 32º, according to 29,86 . The dashed line indicates the membrane shape in an ideal torus. Solid line is the suggestion of what the shape of the membrane limit (not necessarily the polar surface) could be in these pores. It seems unlikely that lipid headgroups could reside on the part of the line that is behind the α-helix. To see this figure in color, go online.