Biological membranes are constituted by a variable and complex mixture of lipids and proteins organized into a lipid bilayer structure (Sperelakis, 2012; Nickels et al., 2017). Besides membrane inserted proteins, in vivo systems can also contain significant amount of biomolecules adsorbed on their surface, also known as protein coronas (Nguyen and Lee, 2017). These protein layers may alter numerous membrane-related interactions and properties of natural systems, e.g., viral envelopes or extracellular vesicles (Monopoli et al., 2012; Bros et al., 2018; Buzás et al., 2018; Cai and Chen, 2019; Ezzat et al., 2019), but may also influence delivery or immunorecognition of synthetic nanoparticles (Oh et al., 2018). Since protein coronas form the first line in contact with other species, they are important both for the better mechanistic understanding of a diverse set of membrane active molecules, i.e., peptides, proteins, and also for understanding vesicular trafficking or viral fusion (Bros et al., 2018; Pattipeiluhu et al., 2020). However, compared to standard membrane models, our knowledge is very little on structural properties of biomembranes with an associated protein corona. Typical model membrane systems in structural biophysics are Langmuir monolayers, vesicles, liposomes, and solid supported lipid bilayers (Knobloch et al., 2015; Palivan et al., 2016), which in one hand provide systems that can be easily produced, but on the other hand may often limit the overall insight. Having an expanded set of complex membrane models is thus important to reach for advanced membrane biophysics (Stone and Deber, 2017; Rojalin et al., 2019). Accordingly, several directions are addressed, from preparation of more complex lipid bilayers by membrane protein reconstitution (Shen et al., 2013; Goers et al., 2018; Smirnova et al., 2018) to purified membranes from e.g., red blood cells (Li and Lykotrafitis, 2014; Rossi et al., 2019; Deák et al., 2020). Related, extracellular vesicles (EVs) may also offer new model systems to address problems described above. EVs have recently become an intensively studied area due to their extreme biomedical potential. These species also contain a phospholipid bilayer, with characteristic diameter ranging from 40 to 1000 nm and arising from diverse origin (Valadi et al., 2007; Vlassov et al., 2012; Zonneveld et al., 2014; Théry et al., 2018). Some of their subtypes may carry complex cargo, such as RNA (Keerthikumar et al., 2016; Kim et al., 2017; Driedonks and Nolte-'t Hoen, 2019), thus rendering them to play an important role in intercellular communication and potentially in diagnostics as biomarkers (Szabo and Momen-Heravi, 2017). Due to their in vivo origin, EVs also can contain adsorbed proteins and protein coronas (Rosa-Fernandes et al., 2017; Karasu et al., 2018; Xia et al., 2019). One of the frequently used EVs are red blood cell derived EVs (REVs), which offer a reproducible system with high particle concentration (Szigyártó et al., 2018; Kitka et al., 2019; Deák et al., 2020; Varga et al., 2020).

To progress in use of EVs as improved model membrane systems, here we employed red blood cell derived EVs and two selected membrane active peptides (MAPs) to characterize their interaction network utilizing several biophysical methods. Among membrane active compounds, antimicrobial peptides (AMPs) receive considerable interest as they have the potential to combat e.g., resistant bacterial species (Bradshaw, 2003). Natural AMPs, or host defense peptides, are components of the innate immune system, with a short sequence (~10–40 amino acids) and high content of cationic and hydrophobic residues. Although AMPs have multiple intracellular targets, the majority of their action mechanisms are based on the lysis of membranes (Brogden, 2005; Pasupuleti et al., 2012; Haney et al., 2019). Their in vivo structure-function relationship, however, is still not well-understood, which can be also due to the major differences between synthetic liposomes and natural biomembranes. For current studies the natural peptide melittin and its synthetic derivative, CM15, was selected as these are well-characterized with several of their membrane-related functions and hemolytic activities understood. Melittin is the major component of honey bee venom with potent lytic activity, as well as with multiple pharmacological actions e.g., antibacterial, antiviral, anticancer, and antiinflammatory effects (Dosler et al., 2016; Liu et al., 2016; Uddin et al., 2016). Its membrane binding mechanism can vary, either acting by a carpet mechanism, adsorbing on by bilayers in a partial helix conformation (Ohki et al., 1994) or forming toroidal pores (Smith et al., 1994; Sengupta et al., 2008; Lee et al., 2013; Hur et al., 2017) which probably requires self-assembly to tetrameric forms (Kurgan et al., 2019). Similarly to melittin, for the cecropin-melittin hybrid CM15, both carpet mechanism (Pistolesi et al., 2007) and pore formation were observed in a concentration dependent manner (Milani et al., 2009), but, in contrast to melittin, without significant hemolytic activity. To study the interactions between REV membranes and these MAPs, we performed complex biophysical characterization using linear dichroism (LD), circular dichroism (CD), attenuated total reflection Fourier-transform infrared spectroscopy (ATR-FTIR), dynamic light scattering (DLS), microfluidic resistive pulse sensing measurements (MRPS), fluorescent labeling coupled to size exclusion chromatography, and freeze-fracture transmission electron microscopy (FF-TEM). Current results indicate several novel functions of the investigated MAPs, including eradication of surface proteins from REVs upon their interaction.

To understand the structural effect of membrane active peptides on REVs polarized light spectroscopy measurements were performed, using circular dichroism (CD) and flow-linear dichroism (LD) (Kogan et al., 2011, 2014; Rocha et al., 2016). Flow-LD spectroscopy is regularly used to study shear-aligned vesicles, where shear is created in a Couette flow cell, which measurements can provide both qualitative and quantitative details about membrane orientation and membrane associated molecules (Rodger et al., 2002; Nordén et al., 2010; Svensson et al., 2011). Furthermore, as reported earlier (Szigyártó et al., 2018) can be employed also to study molecular details of REV systems. The latter study has demonstrated that REVs contain membrane associated hemoglobins as part of their protein corona, which provides additional contributions to the LD spectra, owing to their heme groups (Caesar et al., 2009; Szigyártó et al., 2018). Considering LD measurements, the spectrum of control REV samples shows a major, positive band at ~220 nm, arising most probably from a mixture of the n-π^* and π-π^* amide transitions of helices present in REV proteins, where helices are predominantly aligned parallel to the macroscopic orientation (Figure 1). Note that due to the complex sample composition at this region significant light scattering occurs as well (Nordén et al., 2010). The spectrum also displays a positive peak at ~420 nm, which corresponds to the Soret band of heme containing proteins (Caesar et al., 2009; Szigyártó et al., 2018). In the case of REVs, only hemoglobin can be considered, thus the particular peak belongs to membrane associated hemoglobin molecules. The gradual increase in peptide concentrations up to 50 μM, shows the decrease of both positive peaks (Figure 1). In case of melittin, the peak at 420 nm shows a decrease even at 1 μM (Figure 1C), whereas for CM15 the decrease is slower (Figure 1D). Moreover, upon further addition of melittin and CM15, a small red and blue shift can be detected, respectively and shoulders emerge at higher wavelengths. The deconvoluted LD band (Szigyártó et al., 2018) shows components close to 417 and 430 nm, corresponding likely to a mixture of oxygen-bound (oxyHb) and oxygen-free (deoxyHb) hemoglobin (van Kampen and Zijlstra, 1983).

When comparing signal changes of the LD peak at ~220 nm, arising from amide regions of all oriented proteins, a reduced signal is observed upon increasing peptide concentration (Figures 1, 2). This indicates that MAPs affect all membrane associated proteins involving both membrane embedded or surface associated molecules during the titration process. However, the same trend is found for the Soret band of hemoglobin suggesting its effective removal from the REV surface initiated by these peptides. Further on, the relative loss of peak intensities at 220 and 420 nm during titration follow a similar trend for each particular peptide (Figure 2 and Table S1). Note that while for melittin the signal is almost completely lost at 20 μM peptide concentration, for CM15 the peaks are still observable even at 40 μM suggesting a slower effect for the latter membrane active peptide.

As previously reported (Szigyártó et al., 2018) the CD spectrum of REVs arising from all proteins accompanied with the vesicles exhibits a strong, broad negative band ~226 nm originated from n–π^* transition and a much weaker unresolved π−π^* transition component at 210 nm. The interactions of MAPs with REVs were investigated using the same titrations as described in the LD section. Addition of melittin and CM15 rendered the decrease of the main negative band at 226 nm (Figures 3A,B) accompanied with small blue shifts. Concomitantly, at higher peptide concentrations, the band at 208 nm displayed as a shoulder was blue shifted and appeared as a separate peak with increased intensity. All of these changes are most prominent for the use of melittin. The spectral variations resemble to those observed previously for REVs upon storage (Szigyártó et al., 2018), where conformational changes of membrane-associated proteins, especially, hemoglobin, have been detected.

To better understand the effect of the membrane-peptide interaction on the global CD spectra, reference measurements using the same titration were also performed with DOPC (dioleoyl-phosphatidylcholine) and DOPC/DOPG (dioleoyl-phosphatidylcholine/dioleyl-phosphatidylglycerol, 80/20 n/n%) liposomes, which can be considered as the simplest model system to mimic the outer leaflet of red blood cell membranes (Viswanad et al., 2014) (Figures S1, S2). When comparing the arithmetic sums of the individual CD curves of the free REV plus peptides (Figure S3) as well as MAPs bound to liposomes (Figure S1) it is obvious that the CD profile found for the REV plus MAP mixtures is distinct from such a simple combination of the two separate components. Nonetheless qualitative similarities can be observed, especially in the increasing helicity upon peptide additions. Nevertheless, the more than 80 proteins that have been identified in REVs (Prudent et al., 2018), are highly likely to undergo some conformational changes especially if they are detached from the surface upon peptides additions, thus the observed CD signal variations cannot be deduced with confidence.

To support changes in secondary structure observed by CD, FTIR spectra were collected focusing on amide I (~1660 cm⁻¹) and amide II (~1542 cm⁻¹) spectral contributions of proteins and peptides (Barth, 2007). Analysis of the two main band components of amide I, centered at 1632 and 1655 cm⁻¹ are characteristic of antiparallel β-sheets and α-helices, respectively (Barth, 2007; Mihály et al., 2017) (Figure 3C). A definite change in the ratio of the two amide I components was observed for both MAPs with increasing concentration, suggesting the dominance of α-helical conformation. This effect can be related to the interaction of the peptides with the REV membrane proteins and with the lipid bilayer. Note, that sample drying required for the IR measurements may influence the structural properties of peptides. However, for both studied systems the results are in close agreement with those of the solution phase CD data, which strongly suggests that such an effect was not significant here.

Qualitative changes on REV morphology was further investigated by transmission electron microscopy. The original REV surface contained granules representing native assemblies of membrane associated proteins (Figure 4A) (Varga et al., 2014, 2020; Deák et al., 2015; Szigyártó et al., 2018). The average size of these vesicles was also verified using dynamic light scattering (DLS) measurements. These measurements indicated that original REVs have spherical morphology with a mean diameter of ~200 nm, which size remains mostly similar after addition of peptides (Table S2 and Figure S4), whereas remarkable changes were observed in REV membrane surface morphology after addition of CM15 and melittin (Figures 4B,C).

Protein aggregation on the vesicle surface and partial removal of surface proteins can be observed already at 5 μM concentration of the MAPs (Figures S5, S6). Further increase in the peptide concentrations up to 50 μM shows significant decrease in the native protein particles by removing them from the vesicle membrane surface (Figures 4B,C). The FF-TEM measurements reveal that both peptides cause detachment of proteins from membrane surface, whereas the membrane remained mainly intact. Proteins detached from the surface can be observed close to the remaining clear intact REV vesicles (Figure S6).

Note, that melittin is also known to be hemolytic (Sato and Feix, 2006). This could affect the above results as for instance loss of LD signal intensities can be attributed also to membrane disruption. Consequently, to address and quantify how prominent is the membrane disrupting effect of melittin on REVs, microfluidic resistive pulse sensing (MRPS) measurements were employed by determining the size distribution and concentration of vesicles. The absolute size distribution of REVs was determined with and without melittin (Figure 5). The overall particle concentration of the control REV sample was 1.57·10¹⁰ ± 8·10⁷ particles/mL, which is reduced to 9.58·10⁹ ± 5·10⁷ particles/mL in the presence of 50 μM melittin. This indicates that by the final concentration of our titrations melittin disrupts ~37% of the vesicles. This is in line with the previously described effect of this peptide on red blood cells (Watala and Gwozdzinski, 1992), and is reasonably a low disrupting effect for REVs if one considers that in the samples the used final peptide concentration, 50 μM, corresponds to ~2·10⁶ melittin molecules per one vesicle.

To support MRPS results, and to better compare and understand at a near quantitative level the effect of melittin on both protein corona members and membrane inserted proteins, size exclusion chromatography with fluorescence detection (Flu-SEC) was used. For investigating membrane protein-melittin interactions glycophorin A (CD235a), a characteristic REV membrane protein, was chosen. Control and melittin-treated REVs were labeled with anti-CD235a conjugated with phycoerythrin (PE-CD235a) and the absorbance of hemoglobin as well as the fluorescence of the antibody was detected (Figure 6). The absorbance chromatogram of control REVs shows a single peak at 3.1 min retention time, which corresponds to the void volume of the column, and indicates intact EVs (Figure 6A). In contrast, for REVs treated with 50 μM melittin the peak was remarkably reduced in intensity with only 3% area (AUC) compared to the control sample, and a new peak appears at 7.6 min retention time, which corresponds to hemoglobin molecules detached from REVs.

Monitoring the fluorescence chromatogram of control REV (Figure 6B) two peaks at 3.2 and 6.5 min retention times corresponding to REV-bound and free antibodies were observed. The intensity of the REV peak at 3.2 min is reduced to 55.9% in the presence of melittin compared to the control sample. This reduction is in good agreement with the loss of vesicle number determined by MRPS measurements, i.e., the reduced REV-peak on the fluorescence chromatogram can be attributed to the decreased REV concentration. These observations drive to the conclusion that melittin is capable to remove surface bound proteins but does not affect significantly the membrane-embedded molecules like glycophorin A. The loss of signal intensity lies close to that observed for MRPS measurements indicating a fraction of the vesicles being disrupted.

The combined results confirm that both melittin and CM15 efficiently remove proteins adsorbed to the external surface of REVs, which is a new functionality that has not been seen for these relatively well-studied MAPs. The TEM images clearly display that the REV surfaces will become much smoother. As hemoglobins is one of the most abundant proteins found in human protein corona (Solorio-Rodríguez et al., 2017) in line with this, the abrupt decrease of the Soret band at 420 nm indicates the removal of membrane-bound, LD active hemoglobin molecules. Furthermore, since melittin can form membrane pores (Smith et al., 1994; Hur et al., 2017) it is likely that besides removing hemoglobins from the external surface, it also causes leakage of internal hemoglobins from the vesicles. This assumption is supported by Flu-SEC results of REVs treated with melittin showing almost complete loss of peak intensity of REV-associated hemoglobin (Figure 6A). The TEM images show that the proteins removed from the surface aggregate most likely together with the peptides, and their assemblies can be found quite near to smooth vesicles from which they are likely to directly originate. Based on the MRPS and fluorescence-SEC measurements membrane inserted proteins most likely are only little affected. This is also supported by LD spectra collected at higher melittin concentrations, where the peak ~220 nm gradually disappears, and a smaller negative band appears at 208 nm. While the initial peak will be a contribution from all the proteins present on the oriented vesicles, and thus challenging to interpret, the latter negative peak will likely arise from a narrower distribution of structural elements. Its absolute value lies close to helices typically inserted into the bilayer preferentially parallel to the membrane normal, as this band arises from the π−π^* amide transitions that are polarized along the helix axis (Figure 1A) (Nordén et al., 2010).

In summary, current results demonstrate that melittin and CM15 drastically affects protein corona adsorbed to the REV lipid bilayer. The detachment of these surface proteins by these MAPs can clearly be reached and tracked, potentially providing improved applications for surface engineering and production of EV-based therapeutics enhancing particle-based delivery strategies (Bhunia et al., 2017; Nguyen and Lee, 2017). Increased understanding on these interactions and how this can be controlled by other biomolecules is an interesting direction, which we aim to explore in the future. The combination of several biophysical techniques also allowed the qualitative separation and the nearly quantitative identification of apparent lysis and surface protein-removing activities of melittin, which offers improved understanding for different processes occurring simultaneously during peptide-biomembrane interactions. The observed new functionality of the studied peptides suggests that they may have further alternative activities in vivo beyond those observed previously on synthetic liposomes. Results also imply that extracellular vesicles provide an excellent opportunity to better follow and understand interactions of membrane active compounds with complex biomembranes. Considering that membrane active peptides are also interesting for loading various drug compounds into EVs and for aiding their trans-bilayer penetration, we expect that peptide-EV interactions will receive increasing focus in EV-engineering.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Scientific and Research Ethics Committee of the Hungarian Medical Research Council (ETT TUKEB 6449-2/2015). The patients/participants provided their written informed consent to participate in this study.

TB-S, IS, and ZV conceived the study. PS and IS performed CD, LD, ATR-FTIR experiments, and data analysis. MR and JM contributed to ATR-IR data evaluation and discussion. FZ and TJ contributed to data interpretation and discussion. SB performed peptide synthesis. ÉB and JK performed FF-TEM measurements. DK and ZV performed MRPS and Flu-SEC experiments and data evaluation. IS and TB-S wrote the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.