L. A. Morozova-Roche, Department of Medical Biochemistry and Biophysics, Umeå University, 901 87 Umeå, Sweden Fax: +46 90 786 9795 Tel: +46 90 786 5283 E-mail: email@example.com
Protein oligomeric complexes have emerged as a major target of current research because of their key role in aggregation processes in living systems and in vitro. Hydrophobic and charged surfaces may favour the self-assembly process by recruiting proteins and modifying their interactions. We found that equine lysozyme assembles into multimeric complexes with oleic acid (ELOA) at the solid–liquid interface within an ion-exchange chromatography column preconditioned with oleic acid. The properties of ELOA were characterized using NMR, spectroscopic methods and atomic force microscopy, and showed similarity with both amyloid oligomers and the complexes with oleic acid and its structural homologous protein α-lactalbumin, known as humanα-lactalbumin made lethal for tumour cells (HAMLET). As determined by NMR diffusion measurements, ELOA may consist of 4–30 lysozyme molecules. Each lysozyme molecule is able to bind 11–48 oleic acids in various preparations. Equine lysozyme acquired a partially unfolded conformation in ELOA, as evident from its ability to bind hydrophobic dye 8-anilinonaphthalene-1-sulfonate. CD and NMR spectra. Similar to amyloid oligomers, ELOA also interacts with thioflavin-T dye, shows a spherical morphology, assembles into ring-shaped structures, as monitored by atomic force microscopy, and exerts a toxic effect in cells. Studies of well-populated ELOA shed light on the nature of the amyloid oligomers and HAMLET complexes, suggesting that they constitute one large family of cytotoxic proteinaceous species. The hydrophobic surfaces can be used profitably to produce complexes with very distinct properties compared to their precursor proteins.
The process of protein self-assembly has become the focus of much current research as a broad manifestation and consequence of protein instability. The propensity of protein molecules to aggregate markedly increases if they are destabilized or partially unfolded [1–3]. The increased exposure of hydrophobic surfaces in partially unfolded states leads to spontaneous protein aggregation. Protein destabilization can be achieved using mild denaturing conditions, such as acidic or basic pH, heating, chemical denaturants and ligands, as well as at solid–liquid interfaces [4–7]. Among self-assembled protein complexes, oligomers have attracted special attention because of their involvement in amyloid formation and their distinct properties, which often differ from those of their precursor monomers. Specifically, during amyloid formation, oligomers may serve as nuclei for further aggregation [8–10]. It has also been suggested that they can fulfil the role of major cytotoxic agents compared with more inert amyloid fibrils [11–15]. Because of the transient nature of oligomeric species, which tend to associate into larger aggregates or split into monomers, it is difficult to produce their stable fractions [16–20]. A number of attempts have been made to stabilize the oligomers of amyloidogenic polypeptides using fatty acids and surfactants [21–25]. In our research, we have produced stable oligomeric complexes of equine lysozyme with oleic acid (ELOA), which we subsequently studied in detail with regard to their structural and cytotoxic properties.
Complexes of human α-lactalbumin with oleic acid were first described in the 1990s by Svanborg and coworkers, and named human α-lactalbumin made lethal to tumour cells (HAMLET) [26,27]. HAMLET was produced in vitro in an affinity column loaded with oleic acid, and it was also shown that HAMLET is present naturally in the casein fraction of human milk . Recently, HAMLET has been formed at higher temperatures of 50 and 60 °C, which facilitated the dispersal of oleic acid and structural changes in the protein . Complexes of bovine α-lactalbumin with oleic acid have also been produced using column chromatography and were designated as bovine α-lactalbumin made lethal for tumor cells (BAMLET) [29,30]. Because of their unique antitumor activity, the structure and function of HAMLET and BAMLET have been studied extensively, however, the nature of the conformational changes occurring in the proteins upon their complex formation and the mechanisms of cytotoxicity of the complexes are still debated [26,34]. It has been shown that in both complexes human and bovine α-lactalbumins are partially unfolded or misfolded even under physiological conditions, and this may be crucial for the cytotoxicity of their complexes [30,32]. A complex of bovine α-lactalbumin with polyamines has also been produced and denoted as LAMPA ; the partially unfolded state of α-lactalbumin within this complex was distinct from all other states of monomeric α-lactalbumin characterized to date. The same authors have shown that monomeric α-lactalbumin in the absence of fatty acids can bind to histone H3, which is the primary target of HAMLET , but free α-lactalbumin has not been found to have antitumor activity. Recently, it has been also shown that oleic acid can inhibit the amyloid fibril formation of bovine α-lactalbumin, acting at the initial stages of oligomerization and fibrillation .
Equine lysozyme was selected as the subject of our studies because it is the closest structural homologue of α-lactalbumin. Equine lysozyme has been used extensively as a model in protein folding and amyloid studies over the last two decades [4–6,16,38–44], and this has enabled us to reveal a wealth of information on the mechanisms underlying these processes. By contrast to conventional non-calcium-binding c-type lysozymes and similar to α-lactalbumins, equine lysozyme is a calcium-binding protein ; however, it still displays an enzymatic activity that is characteristic of lysozymes. As a consequence, it possesses a combination of the structural and folding properties of both superfamilies of structurally homologous proteins – lysozymes and α-lactalbumins. Equine lysozyme is characterized by significantly lower stability and cooperativity than non-calcium-binding lysozymes [4,5,39,40,44]. It forms a range of partially folded states under equilibrium destabilizing conditions similar to α-lactalbumins [4,5], and also populates kinetic folding intermediates during the refolding reaction similar to c-type lysozymes [45,46]. Its equilibrium and kinetic intermediates as characterized by similar structural properties. Specifically, equine lysozyme possesses a very stable core, which retains its native-like conformation even in the molten globule state [5,6] and which is rapidly folded and persists in kinetic intermediates [45,46]. Equine lysozyme also forms oligomeric and fibrillar amyloid assemblies under acidic conditions, where its partially folded state is populated [42,47]. Its amyloid oligomers, ranging from tetramers to ecosinomers, display an amyloid gain-on function, such as apoptotic activity [42,48]. These oligomers are populated in very small quantities of only a few percent and tend to convert rapidly to amyloid protofilaments. Therefore, the study of stable and well-populated oligomers of equine lysozyme with oleic acid, which share HAMLET-like and amyloid properties, may shed light on both phenomena. The application of a solid–liquid interface, facilitating protein self-assembly and protein–oleic acid interactions, proved to be an efficient approach to produce such complexes and to model their interactions, which may occur at the hydrophobic and charged surfaces in both biological systems and in vitro during the storage of proteinaceous materials.
ELOA complex formation
ELOA complexes were formed using an anion-exchange column preconditioned with oleic acid, as described in Materials and methods. ELOA was eluted as a strong peak at ∼ 1 m NaCl using a NaCl gradient of 0–1.5 m (Fig. 1). In the absence of oleic acid, equine lysozyme was eluted as a narrow peak at a lower NaCl concentration of 0.67 m. At the front of the ELOA elution profile there is a small peak, possibly corresponding to equine lysozyme according to its position in the salt gradient; this was not analysed further.
CD spectroscopy of ELOA
The far- and near-UV CD spectra of ELOA and equine lysozyme in 10 mm Tris buffer (pH 9.0) are presented in Fig. 2. The near-UV CD spectrum of ELOA (Fig. 2A) at 25 °C is much less structured than that of the native state equine lysozyme, i.e. the minima at 305 and 291 nm, and the maximum at 294 nm are no longer present, and the magnitude of the ellipticity is diminished (Fig. 2A). Thermal unfolding of equine lysozyme at pH 9.0 (Fig. 2C) closely resembles the protein unfolding transition observed previously at pH 4.5, leading to formation of the partially folded state of a molten globule type at 57 °C . The near-UV CD spectrum of the equine lysozyme molten globule at 57 °C is characterized by pronounced peaks at the same wavelengths as in the native state (Fig. 2A), in accord with results described previously [4,6,39]. By contrast, the overall amplitude of the near-UV CD spectrum of ELOA at 57 °C is significantly reduced compared with signals recorded at 25 °C, and resembles the spectrum of thermally unfolded equine lysozyme at 91 °C (Fig. 2A).
The thermal unfolding transition of ELOA was monitored by changes in ellipticity at 222 nm in the far-UV CD region (Fig. 2C). It was manifested in an overall decrease of the CD signal and occurred over a very board range of temperatures starting at ∼ 35 °C and proceeding up to 91 °C. In equine lysozyme alone, two unfolding transitions were observed over the same thermal range, with the first transition taking place between ∼ 35 and 57 °C, leading to an increase in the amplitude of the CD signal, and the second occurring between 57 and 91 °C, resulting in an overall decrease in CD ellipticity.
ELOA spectra in the far-UV CD region recorded at both 25 and 57 °C do not display the minimum at 230 nm typical of the native state equine lysozyme spectrum at 25 °C, but exhibit the same shape as the spectrum for the equine lysozyme molten globule at 57 °C (Fig. 2B). At 91 °C, both ELOA and equine lysozyme are characterized by the same residual ellipticity typical of the thermally unfolded state (Fig. 2B).
The near- and far-UV CD spectra of ELOA incubated at 37 °C for 24 h did not exhibit any changes, indicating that ELOA remained stable and did not undergo any structural changes under these conditions (data not shown). The CD spectra of equine lysozyme did not reveal any changes when the protein was coincubated with a 50 fold excess of oleic acid in solution for 2 h at 20 °C (Fig. 2D). This indicates the importance of the column environment for ELOA formation.
Binding of fluorescent dyes to ELOA
ELOA binds hydrophobic dye 8-anilinonaphthalene-1-sulfonate (ANS), which leads to an ∼ 10-fold increase in dye fluorescence compared with the free dye in solution (Fig. 3A). A shorter wavelength shift of the spectrum maximum from 515 to 495 nm was also observed, indicating that ANS is present in the bound form in a more hydrophobic environment. These results suggest that the ELOA complex is characterized by exposed hydrophobic surfaces.
ELOA also binds thioflavin-T (ThT) dye, which is known for its ability to bind specifically to amyloid species. In the presence of ELOA, the fluorescence of ThT increases by approximately sixfold compared with the free dye in solution (Fig. 3B). This indicates that ELOA possesses the tinctorial property of amyloids.
Atomic force microscopy
ELOA was analysed using atomic force microscopy (AFM) and the images are presented in Fig. 4. ELOA is characterized by a spherical morphology reflected in spherical-cup specs of 10–30 Å height as measured in AFM cross-sections (Fig. 4A). In samples deposited on mica preincubated with 10 mm NaCl, we observed ring-shaped assemblies of spherical species, with a height of ∼ 10 Å measured along the circumference (Fig. 4B–D) and a diameter of ∼ 30 nm between the highest points of the circumference. Because NaCl balances negative charges on the mica surface and facilitates the adhesion of ELOA, which is also negatively charged at pH 9.0, this may stabilize the ring assemblies of the ELOA oligomers.
1D and NOESY 1H NMR spectra
The 1D 1H NMR spectrum of ELOA at pH 9.0 exhibits very broad aromatic resonances at 8–6 p.p.m. (Fig. 5C,D) and a complete absence of resolved methyl peaks in the low-field region of 2.5–0.5 p.p.m. (data not shown). By contrast, the 1D 1H NMR spectra of equine lysozyme, either eluted from the column without oleic acid preconditioning (Fig. 5A) or freshly dissolved in D2O, are characterized by well-dispersed resonances in both the aromatic and aliphatic regions, closely resembling the spectra reported previously and assigned to the native equine lysozyme at pH 4.5 .
The positions of resolved resonances of oleic acid in ELOA were compared with those of free oleic acid in solution (Fig. 5D). All are consistently shifted up-field: the peak for free oleic acid at 5.4 p.p.m. is positioned at 5.24 p.p.m. in the ELOA complex, the 2.1 p.p.m. peak is at 1.9 p.p.m., the 1.3 p.p.m. peak is at 1.2 p.p.m. and the 0.9 p.p.m. peak is at 0.8 p.p.m., respectively. This indicates that oleic acid exists in a different environment within the ELOA complex than its free form. The spectrum of ELOA was also recorded in 10 mm NaCl/Pi at pH 7.4 (data not shown), and closely resembled the spectrum shown in Fig. 5C. The amount of bound oleic acid in ELOA was determined by comparing the peak area of oleic acid in its bound form at 5.24 p.p.m. (2 olefinic protons) with the peak area corresponding to aromatic proton resonances of lysozyme in the 1D 1H NMR spectrum of ELOA. In approximately 20 consecutive preparations, the ratio of oleic acid to equine lysozyme in ELOA varied from 11 to 48 depending on the specific chromatographic conditions during the complex formation. In general, repetitive saturation of the column with oleic acid resulted in the formation of ELOA with a higher oleic acid content.
The 2D 1H NOESY spectrum of ELOA at pH 9.0 is shown in Fig. 6 and similar results were obtained at pH 7.0 (data not shown). The spectrum arising from the proteinaceous part is characterized by very broad resonances and we present it at a high contour level to demonstrate the resonances from oleic acid molecules integrated into the complex structure. Indeed, the positive NOE cross-peaks between oleic acid signals at 5.2, 1.8 and 1.1 p.p.m. (Fig. 6A) indicate that oleic acid is not present in its free form, but within a large molecular complex. Positive NOE cross-peaks were also observed between oleic acid proton resonances and the aromatic residue resonances of equine lysozyme in the region of 6.5–7.5 p.p.m., as shown in Fig. 6B. This indicates intermolecular binding between lysozyme and oleic acid and that aromatic residues of equine lysozyme are involved in oleic acid binding.
Pulsed field gradient diffusion measurements
The diffusion coefficients of ELOA, native monomeric equine lysozyme and molten globular equine lysozyme at pH 2.0 were determined using pulse field gradient (PFG) diffusion measurements (Fig. 7). Diffusion coefficients were calculated by analysing diffusion decays (a representative example is shown in Fig. 7A) according to Eqn (1). Because equine lysozyme is present in a molten globule state within ELOA, the diffusion coefficient of the molten globule was used as a reference when calculating the molecular volumes and masses of ELOA complexes, according to Eqn (2). The diffusion coefficient of the native state of equine lysozyme was 1.18 times larger than the corresponding value for the molten globule, indicating an ∼ 18% larger hydrodynamic radius and an ∼ 60% larger molecular volume for the molten globule state.
The diffusion coefficients for the ELOA complexes were determined by following separately the strong signals of the aromatic residues of equine lysozyme and the oleic acid protons at 1.15 p.p.m. The diffusion coefficients determined by following the proton resonances of aromatic residues of lysozyme molecules were slightly smaller than those derived from monitoring the signals of the oleic acid protons, indicating that the ELOA preparations contain a small amount of free oleic acid, estimated to be < 10%. The diffusion coefficients for the ELOA complexes were 0.28–0.56 times that of the molten globule state of equine lysozyme (Fig. 7B). Using these values and taking into account the amount of oleic acid bound to each protein molecule, the number of equine lysozyme molecules in the ELOA complexes was estimated to be 4–9 in most cases and 30 molecules in one particular preparation.
Trypan blue cell-viability assay
The effect of ELOA, oleic acid, equine lysozyme and the mixture of equine lysozyme with oleic acid on cell viability was examined using a Trypan blue staining assay. A mouse embryonic liver cell culture (Fig. 8A) and mouse embryonic fibroblasts (Doc. S1) were used for this purpose. ELOA was added at a concentration of 1.8–12.4 μm. The concentrations of equine lysozyme and oleic acid used were equivalent to their content in the ELOA complex. The cells were incubated with the corresponding compounds for 1.5, 5 and 24 h. The viability of mouse embryonic liver cells decreased significantly within 1.5 h of incubation at all ELOA concentrations used; in the presence of 1.8–8.9 μm ELOA it decreased by ∼ 20%, at a higher ELOA content of 12.4 μm it decreased by ∼ 40%. Cell viability decreased by ∼ 70% upon the addition of 8.9 μm ELOA after 5 h and by ∼ 80% after 24 h of incubation. The survival of cells treated with 12.4 μm ELOA did not exceed ∼ 20% after either 5 or 24 h of incubation. Even at its highest concentration, equine lysozyme alone did not affect the viability of mouse embryonic liver cells (data not shown). The reduction in cell viability induced by 85–596 μm oleic acid was within ∼ 10% (Fig. 8B); the same effect was observed when cells were added to a mixture of oleic acid within the same concentration range and equine lysozyme at its highest concentration (data not shown).
Mouse embryonic fibroblast culture was also treated with ELOA and the results of the cell viability assessed by Trypan blue staining assay are presented in Fig. S1. Cell viability decreased by ∼ 90% in the presence of 8.9 μm ELOA after 1.5–24 h of incubation, whereas 85–596 μm oleic acid reduced cell viability by ∼ 10% after 1.5 h and by ∼ 30% after 24 h of incubation. Equine lysozyme alone did not induce cellular toxicity and a mixture of equine lysozyme and oleic acid at their highest concentrations within the range examined here produced the same effect as oleic acid alone (data not shown).
The ELOA complexes with different protein to oleic acid ratios were used in the cytotoxicity experiments, including ratios of 1 : 20, 1 : 40 and 1 : 48. Their cytotoxicity depended on the concentration of the proteinaceous component, determined by measuring absorbance spectra. This indicates that the proteinaceous component, but not oleic acid, is a critical factor in defining the cytotoxicity of ELOA complexes. Further studies are needed to provide more detail on the structure–function relationship of ELOA complexes.
Acridine orange/ethidium bromide staining
Mouse embryonic liver cells treated with ELOA (12.4 μm), equine lysozyme (12.4 μm) and oleic acid (596 μm) for 1.5, 5 and 24 h were subjected to acridine orange and ethidium bromide staining. Representative images of the stained cells after 5 h of treatment are given in Fig. 8C. Acridine orange permeates all cells leading to green fluorescence. In the presence of equine lysozyme and oleic acid, live cells appeared green in ∼ 90% of cases. Ethidium bromide is taken up by cells if their cytoplasmic membrane integrity is lost. Ethidium bromide interacts with DNA in apoptotic cells, giving an orange fluorescence; ethidium bromide fluorescence usually predominates over acridine orange uptake. Orange/green staining was seen in ∼ 80% of all cells treated with ELOA (Fig. 8C), indicating apoptotic type cell death .
Imaging of ELOA interactions with live cells
In order to observe interactions between ELOA and live cells, the complex was fluorescently labelled with the amine-reactive dye Alexa Fluor 488 and live PC12 cells were subsequently incubated with fluorescently labelled ELOA. A concentration of fluorescently labelled ELOA of 850 nm was determined in bulk medium, using quantitative imaging by confocal laser scanning microscopy (CLSM)  and fluorescence correlation spectroscopy (FCS), techniques that enable nondestructive observation of molecular interactions in live cells with single-molecule sensitivity. The time course of ELOA interactions with live cells was studied using time-lapsed CLSM (Fig. 9). We observed that ELOA accumulated continuously in the vicinity of the cell membrane over a period of 58 min, reaching a 10-fold higher local concentration than the bulk concentration in solution. During this time, cells were able to ‘resist’ ELOA and significant uptake of the complex was not detected. At a pivotal time point of coincubation (59 min), cell membranes ruptured in a cooperative manner and ELOA streamed into the cells, filling the whole cellular interior almost instantaneously (60 min). Such effect was not observed for equine lysozyme alone (data not shown), which did not disrupt cellular membranes and did not cause cell damage over 6 h of observation.
We demonstrated that the self-assembly of equine lysozyme into stable oligomers can be induced in an anion-exchange chromatography column preconditioned with oleic acid, as outlined in Fig. 10. It is important to note that coincubation of a 50 fold excess of oleic acid with equine lysozyme in solution did not lead to ELOA formation, as evident from the near-UV CD measurements (Fig. 2C). Oleic acid molecules bound to the ion-exchange matrix constitute an extended surface, facilitating both charged and hydrophobic interactions with equine lysozyme molecules. Such a surface may effectively model the cell lipid membranes able to induce protein–ligand interactions, which would not otherwise occur in solution. Indeed, in solution, oleic acid, like many other small aliphatic molecules, would be present as a micelle. Concomitantly, the solid–liquid interface may induce partial unfolding of equine lysozyme and exposure of the hydrophobic surfaces buried in the native state; this may also be critical for ELOA complex formation. It is important to note that extensive studies have recently been conducted to characterize the conformational changes occurring at the solid hydrophobic interfaces in hen egg white lysozyme, which is a structural homologue to equine lysozyme . A suggested model of conformational change included conversion of the initial α-helical structures into random coil/turn and subsequently into β sheet [51–53]. Such structural changes are a key event in oligomeric and fibrillar amyloid assembly. Equine lysozyme is significantly less cooperative than hen egg white lysozyme [5,6,46] and is more prone to structural rearrangement and aggregation. Therefore, under our experimental conditions, it readily assembled into well-defined ELOA complexes, preserved as a stable fraction in solution for up to a week. It is worth noting that complexes of hen egg white lysozyme with oleic acid were also produced under the same conditions, but they were significantly less populated and easily lost oleic acid (data not shown). Complexes of human α-lactalbumin with oleic acid, HAMLET, were also produced using column chromatography . Remarkably, a multimeric active complex of α-lactalbumin with oleic acid was isolated and purified from the casein fraction of human milk [26,27] and denoted as multimeric α-lactalbumin (MAL), which indicates that the solid–liquid interfaces of the chromatography column may mimic in vivo conditions.
Equine lysozyme within the ELOA complex is present in a partially unfolded state, as evident from the near- and far-UV CD spectra (Fig. 2A,D), ANS binding (Fig. 3A) and the decreased dispersion seen in the 1D 1H NMR spectrum (Fig. 5C). The near-UV CD spectrum of ELOA exhibits lower ellipticity values and largely overlapping peaks compared with the native and even molten globule states of equine lysozyme (Fig. 2A) [6,39]. This indicates that the protein tertiary structure within ELOA may be even more disordered than in its molten globule state. Examination of the 1D 1H NMR spectrum of ELOA clearly shows up-field shifts of the resonance of oleic acid incorporated within the complex compared with the resonances of free oleic acid, demonstrating that oleic acid molecules are an integral part of ELOA. They interact directly with the aromatic residues of lysozyme, as demonstrated by the presence of cross-peaks between the protons of aromatic residues and oleic acid observed in the 1H NOESY spectrum of ELOA (Fig. 6B).
The number of protein and oleic acid molecules varies within ELOA complexes produced in different preparations. We have shown that 11–48 oleic acids can bind to each equine lysozyme molecule, depending on the specific chromatographic conditions during complex formation. The number of equine lysozyme molecules in ELOA can also vary from 4 to 30, as determined by PFG diffusion measurements. Previously, we observed the formation of oligomers of equine lysozyme under amyloid-inducing conditions at acidic pH, which also ranged from tetramers to ecosinomers and larger , however, they never constituted more than a few percent of the total amount of monomeric equine lysozyme in solution. This is in contrast to ELOA, which constitutes the majority of molecular species in the samples. In this respect, ELOA resembles the HAMLET-type complex of α-lactalbumin with oleic acid extracted from the casein fraction of human milk, which is also oligomeric in nature [26,31].
ELOA complexes display properties similar to those of equine lysozyme amyloid oligomers, for example, ThT binding and their morphological appearance as shown by AFM. In a similar way to equine lysozyme oligomers, ELOA also forms ring-shaped assemblies (Fig. 4). By contrast to equine lysozyme and α-lactalbumin amyloid oligomers, which are populated on-pathway to amyloid fibrils [47,54], the ELOA complex did not produce polymeric structures upon prolonged incubation in our experiments. This suggests that oleic acid stabilizes the oligomeric complex, preventing its further conversion and assembly into larger polymers. Some other surfactants and compounds such as SDS and fatty acids were also applied to Aβ peptide, α-synuclein and other amyloidogenic proteins to stabilize their oligomers as opposed to fibrils . Although prefibrillar proteinaceous structures encompass a wide variety of species, studies of kinetically trapped ELOA complexes can shed light on the structural and functional properties of pre-fibrillar species and their role in ‘on-’ and ‘off’-pathway’ amyloid assembly.
It is interesting to note that the thermal unfolding transition of ELOA occurs over a very wide temperature range and broadly coincides with two unfolding transitions of equine lysozyme alone under the same conditions. However, two transitions were not noticed in ELOA and we did not observed an increase in ellipticity signals during ELOA unfolding, which is a distinguishing feature of the first transition in equine lysozyme [39,46]. This indicates that the conformational changes in ELOA and equine lysozyme alone may have different structural origina. Similarly, HAMLET was slightly less stable than human α-lactalbumin in the presence of calcium towards thermal denaturation and exhibited the same stability as human α-lactalbumin towards urea denaturation . This indicates that oleic acid has a similar effect on the structural stability of both complexes.
We have shown that ELOA is cytotoxic towards different cell types, including mouse embryonic liver cell culture, mouse embryonic fibroblast culture, a neuroblastoma cell line (SH-SY5Y) and a rat phreochromcytoma (PC12) cell line. Combined staining with acridine orange and ethidium bromide indicated that ELOA induces apoptotic-type cell death. In order to gain further insight into the mechanisms underlying cellular toxicity, we studied the interactions of ELOA with live cells by using single molecular techniques such as CLSM and FCS (Fig. 9). Our results showed that ELOA initially accumulated actively in the vicinity of the cell membrane, implying that the cell membrane is a primary target for ELOA toxic activity. We presume that interactions of ELOA with the cell membrane trigger apoptotic stimuli, proceeding from the plasma membrane to the cell interior without ELOA internalization per se and consequently trigger cell death. ELOA internalization occurred after the cell membrane rupture.
It is important to note that equine lysozyme oligomers are also cytotoxic, inducing apoptosis in similar cell types . HAMLET complexes have been shown to cause cell death in cancer and immature cells, but not in healthy differentiated cells [30,57]. Thus, a range of various protein oligomeric complexes can induce cytotoxicity, even though their structural properties differ from each other, and this requires further detailed investigation [7,11,58]. In all these complexes, including ELOA, cytotoxicity is a newly gained property, acquired as a result of their self-assembly and, in the case of ELOA, also because of the interaction with oleic acid. Oleic acid itself can induce some cytotoxic effects [59–62], but its cytotoxicity is significantly lower than that of proteinaceous complexes (Figs 8 and S1). These results emphasize the role of protein self-assembly in producing the cytotoxic effect. To date, extensive information has been gathered on the mechanisms behind the cytotoxicity of HAMLET and amyloid oligomers, however, there is no clear consensus. Because equine lysozyme can form both ELOA complexes and amyloid oligomers, in-depth studies of their molecular properties and induced cytotoxicity would provide a clearer insight into both these phenomena and any link between them.
In conclusion, using hydrophobic surfaces in column chromatography, we produced highly populated ELOA complexes, composed of partially unfolded protein molecules and oleic acid. These complexes have some common structural and cytotoxic features with amyloid oligomers of equine lysozyme and with HAMLET. These complexes are stable and therefore amenable to structural characterization at atomic resolution, whereas the amyloid oligomers are often transient in nature and not populated in significant proportions. By producing ELOA, we have shown that other proteins besides human and bovine α-lactalbumins can form such structures, which widens the scope of the HAMLET-type phenomenon. Proteins provide an unlimited source of varying properties and functions, among them protein complexes, which if well-characterized, can be used profitably in various therapeutic and biotechnological applications with the potential to target specifically undesirable cells.
Materials and methods
Equine lysozyme was purified from horse milk, as described previously . Oleic acid and all chemicals were purchased from Sigma (Stockholm, Sweden), unless stated otherwise. The protein concentration was determined by absorbance measurements on a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) at 280 nm using an extinction coefficient of E1% = 23.5.
Production of ELOA by anion-exchange chromatography
ELOA was produced using 1 or 5 mL DEAE FF Sepharose columns (Amersham Biosciences, Piscataway, NJ, USA) connected to a Bio-Rad chromatographic system (BioLogic Wokstation, Bio-Rad, Hercules, CA, USA) and conditioned with oleic acid. Fifty microlitres of 99.5% oleic acid were dissolved in 50 μL of 99.5% ethanol and sonicated in a Transsonic 310 sonicator (Elma, Singen, Germany) for 15 min. Then, 700 μL of 10 mm Tris/HCl buffer, pH 9.0 were added and the final solution was sonicated again for 15 min. The resulting mixture was loaded onto the column and dispersed through the DEAE Sepharose matrix, using a linear NaCl gradient of 0–1.5 m in 10 mm Tris/HCl buffer, pH 9.0. The column was washed with a 10-bed volume of 10 mm Tris/HCl buffer, pH 9.0. Equine lysozyme in 10 mm Tris/HCl, pH 9.0, was loaded onto the column and the ELOA complex was eluted by a linear NaCl gradient of 0–1.5 m. ELOA was dialysed against a 3 × 200-fold volume excess of 10 mm ammonium acetate, containing 60 μm fatty acid free bovine serum albumin, pH 9.0, for a minimum of 2 h each time (Slide-A-Lyzer, membrane cut-off 3 kDa; Pierce, Rockford, IL, USA) and lyophilized.
CD spectra were recorded in a Jasco J-810 spectropolarimeter (Jasco, Tokyo, Japan) equipped with a Jasco CDF-426L thermostat, using 0.1- and 0.5-cm path length cuvettes. At least three scans were averaged for each spectrum.
Fluorescence measurements were performed on a Jasco spectrofluorometer FP-6500 (Jasco). The ThT-binding amyloid assay was carried out using a modification of the method described by LeVine . ThT fluorescence was recorded using excitation at 440 nm, emission between 450 and 550 nm and setting the excitation and emission slits at 5 nm. The fluorescence of the hydrophobic dye ANS was recorded using excitation at 350 nm and emission between 410 and 600 nm, with the excitation and emission slits set at 3 nm.
AFM measurements were performed on a Pico Plus microscope (Agilent, Santa Clara, CA, USA) in tapping mode, using a 100 nm scanner with acoustically driven cantilevers. TESP model cantilevers with etched silicon probes of diameter ≤ 10 nm (Veeco, Plainview, NY, USA) operated at frequencies of 170–190 or 320–370 kHz. The scanning resolution was 256 × 256 pixels. Scanning was performed in trace and retrace to avoid scan artefacts. Images were flattened and plane adjusted. Samples were diluted in Milli-Q water to a final concentration of 20–100 μg·mL−1, placed on mica, left for up to 5 min, rinsed three times with Milli-Q water and air-dried at room temperature overnight. Freshly cleaved mica (GoodFellow, Devon, PA, USA) or mica preincubated with 10 mm NaCl for 10 min was used. The dimensions of ELOA species were measured in cross-section in AFM height images using pico plus software (Agilent).
1D 1H NMR spectra were recorded using a Bruker 600 MHz spectrometer equipped with a 1H, 13C, 15N cryo probe. NMR samples were prepared by dissolving ∼ 0.5 mg lyophilized ELOA in 500 μL D2O, 10 mm Tris or 10 mm NaCl/Pi pH at 9.0 or 7.2 to yield a protein concentration of ∼ 50 μm. The molar ratio between oleic acid and lysozyme in the complex was determined by comparing the peak areas of oleic acid olefinic proton resonances with the lysozyme aromatic signals. 2D 1H NOESY spectra were recorded at 25 °C, using a mixing time of 150 ms, 8 scans and 272 increments (experimental time ∼ 1.5 h).
PFG diffusion measurements were performed using the bipolar pulse-pair diffusion experiment  and analysed as described previously . In the diffusion experiments, the signal intensity was attenuated as a function of gradient strength and, if other factors were constant, signal intensity (I) relative to that in the absence of gradients (I0) was given by Eqn (1):
where G, δ, Δ and τ correspond to amplitude, duration, time between PFGs, and recovery time after PFGs, respectively; γh is 1H gyromagnetic ratio and Dt is translational diffusion coefficient, respectively. Assuming that the proteinaceous particles are spherical and neglecting density changes, the mass of the complexes was determined using Eqn (2):
where MELOA and MEL are molecular masses, and DELOA and DEL diffusion coefficients of the ELOA complex and monomeric equine lysozyme in the molten globule state used as a reference. The number of equine lysozyme molecules in ELOA was calculated by dividing the molecular mass of the complex (MELOA) by the molecular mass of equine lysozyme coordinated with oleic acids, according to the stochiometry determined from 1D 1H NMR spectrum.
Mouse embryonic liver cell culture, mouse embryonic fibroblasts and human neuroblastoma SH-SY5Y cell line were cultured in a Dulbecco’s modified Eagle’s medium supplemented with 10% (v/v) fetal bovine serum and antibiotics in a 5% CO2 humidified atmosphere at 37 °C. Cells were plated at a density of 104 cells·well in 96-well plates, cell viability was assayed after 1.5, 5 and 24 h of coincubation with ELOA and respective controls. ELOA was diluted in serum-free culture medium to the required concentrations and then added to the cells. Oleic acid was diluted in ethanol prior to the addition to culture media. The effect on cells of equivalent concentrations of ethanol was examined and shown to not affect the cell viability.
PC12 cells, pheochromocytoma cells derived from rat adrenal medulla, were obtained from the American Type Culture Collection (ATCC). The cells were cultured in collagen-coated flasks in RPMI 1640 medium supplemented with 5% fetal bovine serum, 10% heat-inactivated horse serum, 100 U·mL−1 penicillin and 100 μg·mL−1 streptomycin (all from Invitrogen, Stockholm, Sweden), and maintained in a 5% CO2 humidified atmosphere at 37 °C. The medium was replaced every 2–3 days. For CLSM experiments, cells were plated in eight-well chambered coverslips (Nalge Nunc International, Rochester, NY, USA) and grown in phenol-red-free RPMI medium supplemented with 10% horse serum, 5% fetal bovine serum, penicillin (100 units·mL−1) and streptomycin (100 μg·mL−1). Average cell density at plating was ∼ 1 × 105 cells·cm−2 in 300 μL medium. The cells were observed for 2–3 days after plating.
Trypan blue assay
The viability of mouse embryonic liver cells and mouse embryonic fibroblasts (Supporting information) was measured by using a Trypan blue exclusion assay. Cells were harvested from the plates after 5 min treatment with 0.25% trypsin and 0.02% EDTA solution containing phenol red (Biological Industries, Kibbutz beit Haemek, Israel). Cells were maintained in culture media containing 10% fetal bovine serum (Biological Industries), then sedimented by centrifugation for 10 min at 400 g and resuspended in 100 μL NaCl/Pi solution (without Ca and Mg). Cells were then stained by adding 100 μL 0.4% Trypan blue solution for 5 min and counted under a light microscope using a Neubauer counting chamber. The viable cells were unstained, whereas the dead cells displayed a blue colour; the counts were evaluated using standard statistical analysis techniques.
Acridine orange/ethidium bromide staining
The dyes acridine orange and ethidium bromide were used to discriminate between live and dead cells on the basis of their membrane integrity. Acridine orange and ethidium bromide (100 μg·mL−1) were mixed in a ratio of 1 : 1. After harvesting, cells were centrifuged for 10 min at 400 g, and the cell pellet resuspended in 25 μL NaCl/Pi. Then 1 μL of the dye mixture was added and cells were examined immediately under a Leica fluorescent microscope (Leica, Wetzlar, Germany) equipped with a green excitation filter block G-2E/C. Cells were counted in four randomly selected areas each containing ∼ 100 cells to quantify cell viability.
Fluorescent labelling of ELOA
ELOA was labelled with Alexa Fluor 488 dye, using the protein labelling protocol provided by the producer (Invitrogen). Excess free dye was removed using a protein desalting column PD10 (GE Healthcare, USA).
FCS/CLSM measurements were performed on a uniquely modified LSM 510 instrument (Carl Zeiss, Jena, Germany), equipped with an inverted microscope for transmitted light and epifluorescence (Axiovert 200 m); a VIS-laser module comprising the Ar/ArKr (458, 477, 488 and 514 nm), HeNe 543 nm and HeNe 633 nm lasers and the scanning module LSM 510 mETA. The instrument was modified to enable avalanche photodiode imaging using silicon avalanche photodiodes (SPCM-AQR-1X; Perkin–Elmer, Fremont, CA, USA). Images were recorded without averaging, using a scanning speed of θ = 25.6 μs·pixel−1 and 512 × 512 pixel resolution. The C-Apochromat 40×/1.2 W UV-VIS-IR objective was used in all measurements. Alexa Fluor was excited using the 488 nm line of the Ar/ArKr laser. Quantitative measurements were achieved by quantitative APD imaging  performed on an integrated FCS/CSLM instrument.
All cell viability experiments were performed in triplicate. The experimental results were analysed by Student’s paired t-test and are shown as mean ± SEM. The level of statistical significance was set at P < 0.05 for the ELOA-treated cells versus the cells treated with oleic acid. *, P < 0.05; **, P < 0.01.
Images of equine lysozyme  were produced using molmol graphic program.
We thank Catharina Svanborg for drawing our attention to the very promising HAMLET field, Christopher Aisenbrey and Sohyun Kim for valuable research assistance and Nils Elfving for supplying equine milk from his horse farm in Arnäsvall, Sweden. This research was supported by grants from the Swedish Research Council, the Wallenberg and the Kempe foundations, The Swedish Brain Foundation, the Biotechnology program, and Insamlingstiftelsen, Umeå, Sweden.