Interaction of the P-type cardiotoxin with phospholipid membranes Peter V. Dubovskii, Dmitry M. Lesovoy, Maxim A. Dubinnyi, Yuri N. Utkin and Alexander S. Arseniev Shemyakin & Ovchinnikov Institute of Bioorganic Chemistry, Russian Academy of Sciences, Moscow, Russian Federation The cardiotoxin (cytotoxin II, or CTII) isolated from cobra snake (Naja oxiana) venom is a 60-residue basic membrane- active protein featuring three-finger beta sheet fold. To assess possible modes of CTII/membrane interaction 31 P- and 1 H-NMR spectroscopy was used to study binding of the toxin and its effect onto multilamellar vesicles (MLV) composed of either zwitterionic or anionic phospholipid, dipalmitoylglycerophosphocholine (Pam 2 Gro-PCho) or dipalmitoylglycerophosphoglycerol (Pam 2 Gro-PGro), res- pectively. The analysis of 1 H-NMR linewidths of the toxin and 31 P-NMR spectral lineshapes of the phospholipid as a function of temperature, lipid-to-protein ratios, and pH values showed that at least three distinct modes of CTII interaction with membranes exist: (a) nonpenetrating mode; in the gel state of the negatively charged MLV the toxin is bound to the surface electrostatically; the binding to Pam 2 Gro-PCho membranes was not observed; (b) penetrating mode; hydrophobic interactions develop due to penetration of the toxin into Pam 2 Gro-PGro membranes in the liquid-crystalline state; it is presumed that in this mode CTII is located at the membrane/water interface deepening the side-chains of hydrophobic residues at the tips of the loops 1–3 down to the boundary between the glycerol and acyl regions of the bilayer; (c) the penetrating mode gives way to isotropic phase, stoichiometrically well-defined CTII/ phospholipid complexes at CTII/lipid ratio exceeding a threshold value which was found to depend at physiological pH values upon ionization of the imidazole ring of His31. Biological implications of the observed modes of the toxin– membrane interactions are discussed. Keywords: cytotoxin II (cardiotoxin); membrane binding mode; multilamellar phospholipid vesicles; 31 P-NMR; isotropic phase. Understanding the physical principles underlying mem- brane protein structure and dynamics advanced rapidly during the past decade. A number of high resolution membrane protein structures available is growing. How- ever, all of them are either helical bundles or b-barrels [1]. An important question is whether new motifs would emerge. The positive answer has been obtained by consid- ering insertion of cytotoxins (CTs) into membranes. CTs (Fig. 1A) are single chain all b-sheet proteins with a common fold provided by four disulfide linkages forming a globular head from which three major loops emerge [2,3]. Due to their cytotoxic and hemolytic activities it was suggested [4,5] that CTs act on biological membranes. This has been demonstrated comprehensively with membrane models such as micelles, monolayers, liposomes [6–8]. Surface [9,10] and transbilayer [11–13] modes of the insertion of CTs into membranes were suggested. CTs were classified into P- and S-types [6]. The S-type CTs contain Ser28 and are thought to insert only loop I into membranes. The Pro30 residue is typical of the P-type CTs which interact with membranes by all three loops [14]. Lytic activity of CT molecules has been ascribed to a change in their orientation at the membrane surface [10] or in the positioning from a surface location to a transbilayer one [13]. A recent theoretical study of the binding of S- and P-type CTs to membranes suggested that P-type CTs are inserted into membranes via the tips of the three loops while S-type ones are inserted through the tip of the loop I only [15]. Experimental data supporting this hypothesis have been found for the interaction of the P-type CTII (Fig. 1B) with micelles [16] and phospholipid vesicles [17]. The effect of CTII on phospholipid membranes was not studied. A wide- line 31 P-NMR spectroscopy was used for this purpose in this work. Different phospholipid phases characterized by specific modes of molecular motion result in specific line shapes of chemical shift anisotropy (CSA)-dominated wideline 31 P-NMR spectra [7]. The induction by CTs of bilayer- to-isotropic phase transitions in membranes composed of anionic [18] or zwitterionic phospholipids [13] has been shown. An analysis of redistribution of the intensities within powder type 31 P-NMR spectra of MLV as a result of their deformation by the magnetic field of the spectrometer was performed and modulation of this effect by peptides [19], CTs [20] in particular, was studied. In the present work CTII/phospholipid interactions for MLV composed of either zwitterionic Pam 2 Gro-PCho or anionic Pam 2 Gro- PGro were analysed with this technique. These data were Correspondence to A. S. Arseniev, Shemyakin & Ovchinnikov Institute of Bioorganic Chemistry, 16/10 Miklukho-Maklaya str., V-437 Moscow, 117997 Russia. Fax: + 7 95 335 50 33, E-mail: aars@nmr.ru Abbreviations: CSA, chemical shift anisitropy; CT, cytotoxin; L/P, lipid to protein molar ratio; MLV, multilamellar vesicles; NaOAc, sodium acetate; PtdCho, phosphatidylcholine; Pam 2 Gro-PCho, dipalmitoylglycerophosphocholine; PtdGro, phosphatidylglycerol; Pam 2 Gro-PGro, dipalmitoylglycerophosphoglycerol. (Received 3 December 2002, revised 4 March 2003, accepted 18 March 2003) Eur. J. Biochem. 270, 2038–2046 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03580.x supplemented by 1 H-NMR data on the toxin binding to MLV. Finally, conclusions about the modes of CTII/ membrane interaction and their relation to biological activity were drawn. Materials and methods Purification of CTII CTII from Naja oxiana snake venom was purified as described previously [21]. Analytical RP-HPLC showed that purity of the CTII preparations was 98%. The phospholipase A 2 activity of CTII was found to be negli- gibly small. In the presence of CTII and EDTA, the effects of phospholipid degradation was not detected at all, even after long incubation (> a week). Sample preparation 31 P-NMR spectroscopy. The phospholipids used in the work, namely 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (Pam 2 Gro-PCho) and 1,2-dipalmitoyl-sn-glycero-3-[phos- pho-rac-(1-glycerol)] (Pam 2 Gro-PGro) were obtained com- mercially (Avanti Polar Lipids, Alabaster, AL, USA) and used without any further purification. 2 H 2 O(99.9%)was from IZOTOP (St. Petersburg, Russian Federation), sodium acetate (NaOAc), EDTA and KCl were from REACHIM (Moscow, Russian Federation) 1 . NaOAc buffer (0.2 M ) containing 10 m M EDTA, 150 m M KCl, in 2 H 2 O (direct pH meter reading of 5.5) was used for the hydration of phospholipids and dissolution of CTII. Samples containing CTII/phospholipid mixtures were prepared in the following way. NMR tubes (5 mm outer diameter 2 , Norell Inc, Landisville, NJ, USA) were loaded with 20 mg of the phospholipid powder. Fol- lowing this, buffer (the amount of buffer added provided at least 200 mol of water per mol of phos- pholipid) or CTII dissolved in the buffer [to provide a desired lipid to protein (L/P) molar ratio] was added. The dispersion obtained was cycled thermally in the range 20–50 °C and agitated mechanically to ensure its homogeneity. For experiments on pH titration, CTII/Pam 2 Gro-PGro mixture (20 mg of Pam 2 Gro-PGro, 2 H 2 O/Pam 2 Gro-PGro of 600 : 1 mol : mol) was prepared in 2 H 2 O containing 10 m M EDTA, 150 m M KCl without buffer. The pH was adjusted by adding small aliquots of concentrated NaOH or HCl solutions and pH values are given as direct pH meter readings. 1 H-NMR spectroscopy. One milligram of CTII was dissolved in 0.5 mL of 10 m M NaOAc (pH 5.5) buffer containing 95% H 2 O, 5% 2 H 2 O, 10 m M KCl and 1 m M EDTA. The 25 m M stock solution of Pam 2 Gro-PCho or Pam 2 Gro-PGro MLV were prepared in the same buffer. After addition of the required amount of lipid to the toxin solution, the mixture was cycled thermally and vortexed before measurement as described above. NMR data acquisition NMR spectra were obtained on an 11.4 T Bruker DRX- 500 spectrometer (Germany) with 1 H- and 31 P-resonance frequencies at x 0 /2p ¼ 500.13 and 202.5 MHz, respect- ively, using a standard broad-band 5 mm probehead. The magnetic field of the spectrometer was locked during acquisition through 2 H 2 O contained in the samples. The temperature of the samples was controlled by dry air and monitored by the VT-system (BVT 3000) to an accuracy of ±0.1 °C. Chemical shifts were referenced to external 85% H 3 PO 4 or internal HDO signal in the 31 P- and 1 H-NMR spectra, respectively. The WATERGATE scheme was used for water signal suppression in the 1 H-NMR spectra [22]. 31 P-NMR spectra were recorded with the spin ½ Hahn echo pulse sequence with full phase cycling of both transmitter and receiver [23] using 90° pulse of 8 lsand interpulse delay of 40 ls. The repetition time was 3–8 s, the longer time being used at higher temperatures. A total of 1000–3000 scans were accumulated for each spectrum of 60 606 Hz spectral width. During acquisition, broadband 31 P- 1 H decoupling was applied. The spectral processing was carried out with WINNMR software supplied by the spectrometer manufacturer (Bruker). The 2K time domain points were zero-filled to 4K data points multiplied by the Lorentzian (broadening factor of 1–5 Hz; this distorted minimally the line-shape of spectra) and Fourier transformed. Fig. 1. Schematic representation of three-fingered b-sheet structure of CTII (A) and its amino-acid sequence (B). In (A) the finger numbers, the N- and C-termini together with the hydrophilic residues situated at the membrane/water interface (dotted line) are marked. In (B) the disulfide bridges are shown in connecting lines, residues at the tips of the three loops interacting with dodecylPCho micelle at pH 5.5 [16] are shown in bold type. Ó FEBS 2003 Cardiotoxin/phospholipid interactions (Eur. J. Biochem. 270) 2039 Line shape simulation Theoretical 31 P-NMR spectra were calculated and fitted to the experimental ones within program MATHEMATICA (ver- sion 4.0, Wolfram Research). Theoretical spectra were represented as the convolution of the Lorentzian, Gaussian (or their linear combination) and function of angular distribution for ellipsoids [19]. This procedure takes into account that MLV of phospholipids in the magnetic field of the spectrometer adopt ellipsoidal shapes [19,24]. In case where the experimental spectra were combinations of isotropic and broad lines, the adjustable parameters in the simulation protocol were: components of the tensor of chemical shift anisotropy, integral intensity, the semiaxis ratio of ellipsoidal MLV, halfwidths of Lorentzian or Gaussian lineshapes and their proportion. At the attained level of signal-to-noise ratio this procedure ensured the error of the spectral decomposition into isotropic and broad components to be within  2–4%. Molecular graphics Fig. 1 was drawn with the MOLMOL program [25]. Results and discussion In the present study 31 P-NMR spectroscopy was used to evaluate the effects of CTII on MLV and thus, in combination with the structural data on the protein [16,21], to investigate its membrane binding modes. The extent of the toxin binding to MLV was determined by 1 H-NMR spectroscopy as the first step. Binding of CTII to Pam 2 Gro- P Gro vs. Pam 2 Gro- P Cho MLV 1 H-NMR spectra. The binding of peptides to lipid vesicles is seen in the 1 H-NMR spectra of peptide/lipid mixtures as broadening and/or chemical shift changes of the peptide proton signals because the binding affects overall rotational correlation time and environment of a peptide [26]. The signals in the 1 H-NMR spectra of CTII in aqueous solution at 30 °C are sharp (Fig. 2). Their spectral intensities were attenuated by the addition of Pam 2 Gro-PGro MLV while line widths and chemical shifts remained unchanged (Fig. 2, from top to bottom). Evidently, this corresponds to the case when the exchange rate between free and lipid bound toxin falls into the slow time scale (with an upper boundary of the order of  s )1 ). In this case only the ÔfreeÕ peptide is seen spectroscopically while the 1 H-NMR signals of the mem- brane-bound CTII are extremely broad due to low overall rotational correlation time of the complex. When the L/P ratio becomes >  10 : 1, CTII signals in the spectra are not observed (Fig. 2, bottom), i.e. all CTII is bound to Pam 2 Gro-PGro. It is of note that the 31 P-NMR spectra of the samples corresponding to L/P ¼ 7, 10, 12 (Fig. 2) were similar to spectrum taken at L/P ¼ 14 (see below) 3 . Thus the bilayer structure of Pam 2 Gro-PGro membranes in the gel phase is intact at these conditions. Raising the temperature above the gel to liquid crystalline phase transition of Pam 2 Gro-PGro ( 41 °C) at L/P ¼ 14 : 1 results in the domination of the isotropic signal in the 31 P-NMR spectra 4 . However no signal of free CTII was observed in 1 H-NMR spectrum under these conditions. This means that the transition of Pam 2 Gro-PGro to an isotropic phase is not accompanied by a redistribution of CTII between aqueous and lipid phases. The titration of CTII with MLV of zwitterionic Pam 2 Gro-PCho at 30 and 50 °C, i.e. at the temperatures of the gel and liquid crystalline phase, respectively, to an L/P ratio of 50 : 1 did not influence high-resolution 1 H-NMR spectra of the toxin (spectra are not shown). This indicates that binding of CTII to MLV of Pam 2 Gro-PCho is negligibly small and, thus, electrostatic attraction of the positively charged CTII molecule to a negatively charged membrane is required for the binding of the toxin. 31 P-NMR spectra. The partitioning of the P-type CTs into zwitterionic membranes was shown to depend upon surface pressure in the outer membrane leaflet [27]. CTs partitioned into lyso-PtdCho micelles [6], sonicated vesicles of sphingo- myelin [28] or egg PtdCho [17] but not 100 nm unilamellar extruded vesicles of dimyristoyl glycero phosphocholine 5 (P. V. Dubovskii & M. A. Dubinnyi, unpublished obser- vations). 1 H-NMR data of the present study showed that CTII remains in aqueous phase of Pam 2 Gro-PCho MLV. Indeed, the 31 P-NMR spectra of Pam 2 Gro-PCho/CTII dispersions (spectra are not shown) were found to be identical to those of pure Pam 2 Gro-PCho dispersions in the temperature range studied (30–55 °C). Thus, both 1 H-NMR and 31 P-NMR data argue for the absence of CTII binding to Pam 2 Gro-PCho. The influence of CTII binding to MLV of Pam 2 Gro- PGro on the lineshapes of the 31 P-NMR spectra of the phospholipid was studied at different temperatures and L/Ps. These observations were made in the temperature interval (30–55 °C) encompassing the gel-to-liquid crystal transition of Pam 2 Gro-PGro bilayers ( 41 °C), while the L/P ratio varied from 7 to 200. In Fig. 3, the 31 P-NMR spectra of the MLV of pure Pam 2 Gro-PGro and in the presence of CTII (L/P ¼ 129 : 1) arecompared.Thespectrainthegelstate(30 °C) exhibit no difference while in the spectra corresponding to the liquid Fig. 2. 1 H-NMR spectra of CTII in the presence of increasing amounts of Pam 2 Gro-PGro MLV (from top to bottom) at 30 °C. The L/P ratio is indicated on the right. The impurity peaks are marked with asterisks. 2040 P. V. Dubovskii et al. (Eur. J. Biochem. 270) Ó FEBS 2003 crystalline state (50 °C) the redistribution of the intensities of the high-field peak and the low-field shoulder are clearly seen. It is known that the negative diamagnetic susceptibility anisotropy of the fatty acyl chains of the phospholipid molecules causes a preferential orientation of the long molecular axis perpendicular to the external magnetic field [24,29,30]. As a result, MLV of phospholipids adopt shape of a prolate ellipsoid instead of a sphere. Spectral simula- tions of the lineshape of the 31 P-NMR spectra allow to extract the ratio of ellipsoid long/short axis (c/a) [31,32]. This ratio is dependent upon hydration of Pam 2 Gro-PGro which changes with the amount of CTII added. To elucidate theeffectofCTIIonthedeformationofMLVat50°C, the c/a values calculated for MLV of Pam 2 Gro-PGro solvated by either buffer alone or buffered toxin solution are plotted in Fig. 4A against L/P. The addition of the buffer or CTII aliquots resulted in the same phospholipid hydration in both samples. The decrease of L/P below of 20 : 1 induces nearly spherical shape of Pam 2 Gro-PGro MLV (c/a  1), i.e. CTII inhibits the deformation of MLV by the magnetic field (Fig. 4A). Addition of CTII to MLV of Pam 2 Gro-PGro in the liquid crystalline phase results in the appearance of the isotropic signal in the 31 P-NMR spectra. Dependence of the amount of isotropic signal in the 31 P-NMR spectra of Pam 2 Gro-PGro upon amount of CTII added was deter- mined at 50 °C (Fig. 4B). The main feature of this dependence is an existence of a threshold of the L/P ratio below which the transition of bilayer to isotropic phase is initiated. At L/P > 20 the amount of the isotropic signal is zerowithinanexperimentalerror( 3%). At lower L/P ratios the amount of isotropic signal increases nearly proportionally with amount of CTII added. The complete transformation of bilayer into isotropic phase at 50 °Cis Fig. 3. 31 P-NMR spectra of Pam 2 Gro-PGro MLV (water/Pam 2 Gro-PGro ¼ 230 : 1, mol/mol) in the presence (left) of CTII (L/P ¼ 129 : 1) and its absence (right) at 30 °C (bottom), 50 °C (top). The spectra are scaled to the intensity of the high-field peak. The com- puter-simulated shapes of MLV at 50 °Care shown in the top. Fig. 4. Change in the ellipticity c/a of Pam 2 Gro-PGro MLV with an amount of CTII (L/P and L/ 2 H 2 O) or buffer (L/ 2 H 2 O) added (A) and the amount of isotropic signal in the 31 P-NMR spectra of Pam 2 Gro- PGro/CTII dispersions (B). The experiments were carried out at 50 °C. Ó FEBS 2003 Cardiotoxin/phospholipid interactions (Eur. J. Biochem. 270) 2041 observed at L/P ratio of  10:1(Fig.4B).Itisinteresting to note that under these conditions (pH 5.5) CTII bears a net positive charge of  10 and thus CTII/Pam 2 Gro-PGro complexes forming isotropic phase are electrically neutral. The temperature variation of 31 P-NMR spectra of Pam 2 Gro-PGro in the presence of CTII at L/P ratio of 14:1isshowninFig.5. The decomposition of the spectra (see example in Fig. 5) gives the amount of the isotropic component. Its tempera- ture dependence is featured by a sigmoidal shape (Fig. 6A). The temperature of the half-transition of bilayer to isotropic phase (Fig. 6A) coincides that of the gel-to-liquid crystal transition of Pam 2 Gro-PGro MLV. The CSA values of Pam 2 Gro-PGro bilayers in the presence of CTII across the whole temperature range studied (30–55 °C) were similar to their respective values in the absence of the toxin. Thus the average conformation of the head group in Pam 2 Gro-PGro bilayers is insensitive to the presence of CTII under the chosen experimental conditions. This does not contradict a suggestion of an insertion of CTII between the negatively charged polar PtdGro groups. Similar observations were made in studies of positively charged peptides such as polymyxin B [33], melittin [34], gramicidin S [35]. 6 In all those cases an increase of peptide concentration in PtdGro bilayers did not affect 7 the 31 P CSA values of bilayers but lead to the occurence of an isotropic peak in the 31 P NMR spectra. Under conditions chosen (L/P ¼ 14 : 1) the magnetic field induced deformation of MLV of Pam 2 Gro-PGro was determined in the temperature range of 30–55 °Candthe values obtained were compared to those for pure Pam 2 Gro- PGro (Fig. 6B). It is seen that CTII extinguishes the temperature dependence of the deformation of Pam 2 Gro- PGro MLV observed in the absence of the toxin. These data support the above conclusion on hampering of the defor- mation of Pam 2 Gro-PGro MLV in the magnetic field by CTII. pH-dependence of the CTII/Pam 2 Gro-PGro interaction The only group of CTII which changes its ionogenic state in the physiological pH range is the imidazole ring of His31 residue (Fig. 1A). This group is at the membrane-binding site of CTII [16] and thus might influence deepening of CTII into membrane. Indeed, the amount of the isotropic signal in the 31 P-NMR spectra of Pam 2 Gro-PGro/CTII mixture (Fig. 7A) showed a characteristic pH-dependence. The pK a value of 6.1 of this process (Fig. 7B), is close to the value of 5.8 found for His31 of CTII in dodecylPCho micelle [16]. Thus, the neutralization of His31 residue results in the increase of the deteriorating effect of the toxin onto Pam 2 Gro-PGro bilayers. It was shown recently [36] that CTII molecule changes its disposition in dodecylPCho micelle depending on the ionization state of His31 residue. By the change of the ionogenic state of the imidazole ring from a protonated state to a deprotonated one, the molecule of CTII inserts loop 3 more deeply into the micelle. The inserted volume of CTII molecule within the micelle and within the membrane increases when the imidazole group of His31 becomes neutral. This is reflected in the pH-dependence of the ability of CTII to induce an isotropic phase in the membranes of Pam 2 Gro-PGro. Modes of CTII interaction with the phospholipid membranes CTII was characterized in detail in aqueous solution by NMR spectroscopy [21]. This toxin has two slowly inter- converting forms called ÔmajorÕ and ÔminorÕ [21]. They have a Val7-Pro8 peptide bond in trans-orcis-configuration, respectively. Both ÔmajorÕ and ÔminorÕ formsofCTIIarenot Fig. 5. Temperature dependence of the 31 P-NMR spectra of Pam 2 Gro- PGro (molar ratio of water/Pam 2 Gro-PGro is 200 : 1) in the presence of CTII at L/P ¼ 14 : 1. Theoretical spectrum corresponding to the experimental one taken at 38 °C is shown on the right with the decomposition bands drawn in broken lines. Fig. 6. Temperature dependence of the amount of isotropic signal (A) and of the MLV deformation (B). In (A) L/P ¼ 14 : 1. In (B) data for Pam 2 Gro-PGro (diamonds) and Pam 2 Gro-PGro/CTII mixture (L/P ¼ 14 : 1) (triangles) are shown. The molar ratio of water/Pam 2- Gro-PGro in the samples was 200 : 1. The ellipsoidal shape of MLV with semiaxes a and c was assumed in the computer simulations of the data in (B). The vertical bars correspond to the error in the parameter estimation. 2042 P. V. Dubovskii et al. (Eur. J. Biochem. 270) Ó FEBS 2003 bound to zwitterionic membranes of Pam 2 Gro-PCho at the membrane hydration levels used in the present study ( 200 water molecules per lipid molecule). The observations of the present study agree with a previous suggestion that CTs interact with membranes by a combination of electrostatic and hydrophobic forces [5]. In this respect the interaction of CTII with negatively charged membranes is similar to that of other basic peptide toxins: thionin, purothionins, crambin, viscotoxin and delta-haemo- lysin [37,38]. The data from this suggest that the modes of CTII interaction with the membranes can be described as follows (Table 1). Mode 1: Attracted to the membrane surface (Table 1). It was suggested [39] that a positively charged peptide can be attracted by a negatively charged membrane–water interface due to electrostatic interaction and then partitioned into membrane hydrophobically. It is likely that CTII is immobilized by Pam 2 Gro-PGro in the gel state without hydrophobic partitioning. 1 H-NMR study of CTII interaction with MLV of Pam 2 Gro-PGro at 30 °C showed (Fig. 2) that at L/P > 10 : 1 all CTII is bound to lipid. At the same time 31 P-NMR data indicated that bilayer structure of Pam 2 Gro-PGro liposomes is not disturbed at 30 °C (Fig. 5). This suggests that all lipid surface is covered by CTII molecules at L/P  10 : 1 as a further decrease of L/P results in the appearance of unbound CTII (see Fig. 2, L/P ratios below 10 : 1). Assuming that Pam 2 Gro-PGro molecules occupy the same surface area in the gel state of thebilayerasPam 2 Gro-PCho, i.e. 0.48 nm 2 [40] the CTII molecule would have  4.8 nm 2 accessible surface area on the bilayer at L/P ¼ 10 : 1. The maximal cross-section of CTII molecule perpendicular to its long axis is  5nm 2 and thus no more than one toxin molecule per 10 lipids can be accommodated on the surface of Pam 2 Gro-PGro MLV in the gel phase. Thus at L/P ¼ 10 : 1 the Pam 2 Gro-PGro liposomes in the gel phase are covered with CTII in a carpet- like fashion suggested for cecropins [41] and thionins [42]. It is of note that the membrane charge is fully compensated by the charge of CTII at these conditions. Mode 2: Partitioned (inserted) state (Table 1). In the liquid crystalline phase of Pam 2 Gro-PGro membrane the hydrophobic partitioning of CTII is favoured. In the case of the P-type CTs the insertion is accomplished via the tips of three hydrophobic loops (fingers) of the molecule [14–16,43]. The tightly bound water molecule in the second loop of these CTs provide W-shape to this loop [16,21,43,44] and the tips of the loops 1–3 are joined into a continuous hydrophobic column. The modelling of CTII binding with negatively charged membranes suggested that the only ÔmajorÕ form of CTII is energetically favoured to insert into membranes [15]. The mode of the insertion is similar to one described for CTII bound to zwitterionic dodecylPCho micelles [16]. The depth of the penetration is determined by the width of the hydrophobic motif of CTII molecule ( 1 nm) and disposition of positively charged side-chains which (Fig. 1, lysines 4,5,12,23,50, arginine 36) are bound to the negatively charged groups of the lipid molecules at the lipid bilayer/water interface. The liquid crystalline bilayer of Pam 2 Gro-PGro can accomodate a definite number of CTII molecules (L/P  20 : 1, Fig. 4B) without induction of an isotropic phase. The gel to liquid crystal transition is accompanied by an  0.16 nm 2 increase of the surface area per lipid molecule in Pam 2 Gro-PCho bilayer [40] and we assume the same value for bilayers of Pam 2 Gro-PGro. Thus, the surface area occupied by 20 molecules of Pam 2 Gro-PGro upon transi- tion from gel to liquid crystalline phase increases by 0.16 · 20 ¼ 3.2 nm 2 . This value is smaller than the molecular area of 5 nm 2 occupied by a single molecule of CTII on the surface of a dodecylPCho micelle [16]. This suggests that insertion of CTII into Pam 2 Gro-PGro membrane is likely to cause expansion of the membrane. Mode 3: Isotropic phase (Table 1). When concentration of CTII in the liquid crystalline membranes of Pam 2 Gro- PGro exceeds a threshold value (L/P  20 : 1), the liposomes are not able to accommodate CTII further and Fig. 7. pH-dependence of the 31 P-NMR spectra of Pam 2 Gro-PGro/ CTII (14 : 1) dispersions at 50 °C (A), the anisotropic component in the spectra is shown with broken line, the isotropic signal is truncated. (B) The amount of the isotropic signal plotted vs. pH and approximated by the best nonlinear fit for the determination of pK a value. Ó FEBS 2003 Cardiotoxin/phospholipid interactions (Eur. J. Biochem. 270) 2043 are transformed into an isotropic phase with L/P ratio of 10. The induction of this phase by CTs in anionic membranes was suggested to be due to formation of stoichiometrically well defined electrically neutral complexes of CT/ phospholipid having inverted micellar structure [18,45]. This observation seems to be valid for CTII/Pam 2 Gro- PGro mixtures too. Biological implications It is well known that CTs preferentially target and disrupt bilayers that are rich in acidic phospholipids on the extracellular side of the plasma membrane [46]. This effect was related to ability of CTs to induce formation of an isotropic lipid phase [19,45]. At a lower concentration, at which cytotoxic effect is less pronounced, CTs modulate Na + and Ca 2+ ion fluxes acting on specific membrane channels [47]. This effect can be attributed to CTs bound to the lipid bilayer with the tips of three fingers. We conclude from this work that CTs influence the bending elastic modulus of the membrane and, hence, the transbilayer pressure profile originating from the stiffening of lipid molecules in the polar but not in the acyl region of the membrane. (In this respect the influence of CTs on the membrane stiffness is opposite to the action of the cholesterol [48] which is ubiquitous in the membranes of animal cells.) The membrane stiffness is important for functioning of membrane receptors and ion channels [49]. Thus a deleterious effect of low CT doses on living cells might be due to insertion of CTs into membranes (see mode 2, Table 1) resulted in improper functioning of the ion channels and receptors. Acknowledgements This work was supported partially by the Ministry of Science and Technology of Russian Federation and by the Russian Foundation of Basic Research (RFBR) grants 00-04-55024, 00-15-97877, 01-04-48548. References 1. White, S.H., Ladokhin, A.S., Jayasinghe, S. & Hristova, K. (2001) How membranes shape protein structure. J. Biol. Chem. 276, 32395–32398. 2. Dufton, M.J. & Hider, R.C. (1988) Structure and pharmacology of elapid cytotoxins. Pharmacol. Ther. 36, 1–40. 3. Kumar, T.K., Jayaraman, G., Lee, C.S., Arunkumar, A.I., Sivaraman,T.,Samuel,D.&Yu,C.(1997)Snakevenomcardio- toxins-structure, dynamics, function and folding. J. Biomol. Struct. Dyn. 15, 431–463. 4. Dufourcq, J. & Faucon, J.F. (1978) Specific binding of a cardio- toxin from Naja mossambica mossambica to charged phospholi- pids detected by intrinsic fluorescence. Biochemistry 17, 1170–1176. 5. Vincent, J.P., Balerna, M. & Lazdunski, M. (1978) Properties of association of cardiotoxin with lipid vesicles and natural mem- branes. A fluorescence study. FEBS Lett. 85, 103–108. 6. Chien, K.Y., Chiang, C.M., Hseu, Y.C., Vyas, A.A., Rule, G.S. & Wu, W.G. (1994) Two distinct types of cardiotoxin as revealed by the structure and activity relationship of their interaction with zwitterionic phospholipid dispersions. J. Biol. Chem. 269, 14473–14483. Table 1. Modes of CTII/Pam 2 Gro-PGro interaction. No Mode a Lipid phase Conditions 1 Gel T < 41 °C L/P P 10 : 1 2 Liquid crystalline T > 41 °C L/P P 20 : 1 3 Isotropic T P 41 °C, L/P < 20 : 1 a The number of lipid molecules in the membrane (only upper leaflet is shown) does represent 8 the stoichiometry of the CTII/lipid interaction. The coordinates of CTII molecule correspond to those calculated in dodecylPCho micelle (code of 1FFJ). Only backbone atoms are shown. 2044 P. V. Dubovskii et al. (Eur. J. Biochem. 270) Ó FEBS 2003 7. Watts, A. (1998) Solid-state NMR approaches for studying the interaction of peptides and proteins with membranes. Biochim. Biophys. Acta 1376, 297–318. 8. Maget-Dana, R. (1999) The monolayer technique: a potent tool for studying the interfacial properties of antimicrobial and mem- brane-lytic peptides and their interactions with lipid membranes. Biochim. Biophys. Acta 1462, 109–140. 9. Ksenzhek, O.S., Gevod, V.S., Omel’chenko, A.M., Semenov, S.N., Sotnichenko, A.I. & Miroshnikov, A.I. (1978) Interaction of the cardiotoxin from the venom of the cobra Naja naja oxiana with phospholipid membrane model system. Molekulyarnaya Biologiya 12, 1057–1065. 10. Bougis, P., Rochat, H., Pieroni, G. & Verger, R. (1981) Penetra- tion of phospholipid monolayers by cardiotoxins. Biochemistry 20, 4915–4920. 11. Lauterwein, J. & Wu ¨ thrich, K. (1978) A possible structural basis for the different modes of action of neurotoxins and cardiotoxins from snake venoms. FEBS Lett. 93, 181–184. 12. Bilwes, A., Rees, B., Moras, D., Menez, R. & Me ´ nez, A. (1994) X-ray structure at 1.55 A ˚ of toxin gamma, a cardiotoxin from Naja nigricollis venom. Crystal packing reveals a model for insertion into membranes. J. Mol. Biol. 239, 122–136. 13. Sue, S.C., Rajan, P.K., Chen, T.S., Hsieh, C.H. & Wu, W. (1997) Action of Taiwan cobra cardiotoxin on membranes: binding modes of a beta-sheet polypeptide with phosphatidylcholine bilayers. Biochemistry 36, 9826–9836. 14. Dauplais, M., Neumann, J.M., Pinkasfeld, S., Menez, A. & Roumestand, C. (1995) An NMR study of the interaction of cardiotoxin gamma from Naja nigricollis with perdeuterated dodecylphosphocholine micelles. Eur. J. Biochem. 230, 213–220. 15. Efremov, R.G., Volynsky, P.E., Nolde, D.E., Dubovskii, P.V. & Arseniev, A.S. (2002) Interaction of cardiotoxins with membranes: a molecular modeling study. Biophys. J. 83, 144–153. 16. Dubovskii, P.V., Dementieva, D.V., Bocharov, E.V., Utkin, Y.N. & Arseniev, A.S. (2001) Membrane binding motif of the P-type cardiotoxin. J. Mol. Biol. 305, 137–149. 17. Dubinnyi, M.A., Dubovskii, P.V., Utkin, Y.N., Simonova, T.N., Barsukov, L.I. & Arseniev, A.S. (2001) ESR study of the inter- action of cytotoxin II with model membranes. Russian J. Bioorg. Chem. 27, 102–113. 18. Picard, F., Pezolet, M., Bougis, P.E. & Auger, M. (1996) Model of interaction between a cardiotoxin and dimyristoylphosphatidic acid bilayers determined by solid-state 31 P NMR spectroscopy. Biophys. J. 70, 1737–1744. 19. Picard, F., Paquet, M.J., Levesque, J., Belanger, A. & Auger, M. (1999) 31 P NMR first spectral moment study of the partial mag- netic orientation of phospholipid membranes. Biophys. J. 77,888– 902. 20. Picard, F., Pezolet, M., Bougis, P.E. & Auger, M. (2000) Hydro- phobic and electrostatic cardiotoxin–phospholipid interactions as seen by solid-state 31 P NMR spectroscopy. Can. J. Anal. Sci. Spectr. 45, 72–83. 21. Dementieva, D.V., Bocharov, E.V. & Arseniev, A.S. (1999) Two forms of cytotoxin II (cardiotoxin) from Naja naja oxiana in aqueous solution: spatial structures with tightly bound water molecules. Eur. J. Biochem. 263, 152–162. 22. Piotto,M.,Saudek,V.&Sklena ´ r ˇ , V. (1992) Gradient tailored excitation for single-quantum NMR spectroscopy of aqueous solutions. J. Biomol. NMR 2, 661–665. 23. Rance, M. & Byrd, R.A. (1983) Obtaining high-fidelity spin ½- powder spectra in anisotropic media: phase cycled Hahn echo spectroscopy. J. Magn. Reson. 52, 221–240. 24. Qiu, X., Mirau, P.A. & Pidgeon, C. (1993) Magnetically induced orientation of phosphatidylcholine membranes. Biochim. Biophys. Acta 1147, 59–72. 25. Koradi, R., Billiter, M. & Wu ¨ trich, K. (1996) MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graphics 14, 51–55. 26. Jelicks, L.A., Broido, M.S., Becker, J.M. & Naider, F.R. (1989) Interaction of the Saccharomyces cerevisiae alpha-factor with phospholipid vesicles as revealed by proton and phosphorus NMR. Biochemistry 28, 4233–4240. 27. Bougis, P., Tessier, M., Van Rietschoten, J., Rochat, H., Faucon, J.F. & Dufourcq, J. (1983) Are interactions with phospholipids responsible for pharmacological activities of cardiotoxins? Mol. Cell Biochem. 55, 49–64. 28. Chien, K.Y., Huang, W.N., Jean, J.H. & Wu, W.G. (1991) Fusion of sphingomyelin vesicles induced by proteins from Taiwan cobra (Najanajaatra) venom. Interactions of zwitterionic phospholipids with cardiotoxin analogues. J. Biol. Chem. 266, 3252–3259. 29. Seelig, J., Borle, F. & Cross, T.A. (1985) Magnetic ordering of phospholipid membranes. Biochim. Biophys. Acta. 814, 195–198. 30. Speyer, J.B., Sripada, P.K., Das Gupta, S.K., Shipley, G. & Griffin, R.G. (1987) Magnetic orientation of sphingomyelin–leci- thin bilayers. Biophys. J. 51, 687–691. 31. Brumm, T., Mo ¨ ps, A., Dolainsky, C. & Bayerl, T.M. (1992) Macroscopic orientation effects in broadline NMR-spectra of model membranes at high magnetic field strength. A method preventing such effects. Biophys. J. 61, 1018–1024. 32. Pott, T. & Dufourc, E.J. (1995) Action of melittin on the DPPCcholesterol liquid-ordered phase: a solid-state 2 H- and 31 P-NMR study. Biophys. J. 68, 965–977. 33. Sixl, F. & Watts, A. (1985) Deuterium and phosphorus nuclear magnetic resonance studies on the binding of polymyxin B to lipid bilayer–water interfaces. Biochemistry 24, 7906–7910. 34. Monette, M. & Lafleur, M. (1995) Modulation of melittin-induced lysis by surface charge density of membranes. Biophys. J. 68, 187–195. 35. Prenner, E.J., Lewis, R.N., Neuman, K.C., Gruner, S.M., Kondejewski, L.H., Hodges, R.S. & McElhaney, R.N. (1997) Biochemistry 36, 7906–7916. 36. Arseniev, A.S. (2002) NMR, ESR and molecular modeling of protein–membrane interactions. In Xx th International Confer- ence on Magnetic Resonace in Biological Systems Toronto, Canada. 37. Dufourc, E.J., Dufourcq, J., Birkbeck, T.H. & Freer, J.H. (1990) Delta-haemolysin from Staphylococcus aureus and model membranes. A solid-state 2 H-NMR and 31 P-NMR study. Eur. J. Biochem. 187, 581–587. 38. Huang, W., Vernon, L.P., Hansen, L.D. & Bell, J.D. (1997) Interactions of thionin from Pyrularia pubera with dipalmitoyl- phosphatidylglycerol large unilamellar vesicles. Biochemistry 36, 2860–2866. 39. Beschiaschvili, G. & Seelig, J. (1990) Melittin binding to mixed phosphatidylglycerol/phosphatidylcholine membranes. Biochem- istry 29, 52–58. 40. Nagle, J.F. & Tristram-Nagle, S. (2000) Structure of lipid bilayers. Biochim. Biophys. Acta 1469, 159–195. 41. Gazit, E., Boman, A., Boman, H.G. & Shai, Y. (1995) Interaction of the mammalian antibacterial peptide cecropin P1 with phos- pholipid vesicles. Biochemistry 34, 11479–11488. 42. Caaveiro, J.M., Molina, A., Rodriguez-Palenzuela, P., Goni, F.M. & Gonzalez-Manas, J.M. (1998) Interaction of wheat alpha-thionin with large unilamellar vesicles. Protein Sci. 7, 2567–2577. 43. Sun, Y.J., Wu, W.G., Chiang, C.M., Hsin, A.Y. & Hsiao, C.D. (1997) Crystal structure of cardiotoxin V from Taiwan cobra venom: pH-dependent conformational change and a novel mem- brane-binding motif identified in the three-finger loops of P-type cardiotoxin. Biochemistry 36, 2403–2413. Ó FEBS 2003 Cardiotoxin/phospholipid interactions (Eur. J. Biochem. 270) 2045 44.Sue,S.C.,Jarrell,H.C.,Brisson,J.R.&Wu,W.G.(2001) Dynamic characterization of the water binding loop in the P-type cardiotoxin: implication for the role of the bound water molecule. Biochemistry 40, 12782–12794. 45. Batenburg, A.M., Bougis, P.E., Rochat, H., Verkleij, A.J. & de Kruijff, B. (1985) Penetration of a cardiotoxin into cardiolipin model membranes and its implications on lipid organization. Biochemistry 24, 7101–7110. 46. Gasanov, S.E., Alsarraj, M.A., Gasanov, N.E. & Rael, E.D. (1997) Cobra venom cytotoxin free of phospholipase A and its effect on model membranes and T leukemia cells. J. Membr. Biol. 155, 133–142. 47. Fletcher, J.E., Hubert, M., Wieland, S.J., Gong, Q.H. & Jiang, M.S. (1996) Similarities and differences in mechanisms of cardio- toxins, melittin and other myotoxins. Toxicon 34, 1301–1311. 48. Reinl, H., Brumm, T. & Bayerl, T.M. (1992) Changes of the physical properties of the liquid-ordered phase with tempera- ture in binary mixtures of DPPC with cholesterol. A 2 H-NMR, FT-IR, DSC, and neutron scattering study. Biophys. J. 61, 1025– 1035. 49. Perozo, E., Kloda, A., Cortes, D.M. & Martinac, B. (2002) Phy- sical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. 2046 P. V. Dubovskii et al. (Eur. J. Biochem. 270) Ó FEBS 2003 . through the tip of the loop I only [15]. Experimental data supporting this hypothesis have been found for the interaction of the P-type CTII (Fig. 1B) with micelles. analysis of redistribution of the intensities within powder type 31 P-NMR spectra of MLV as a result of their deformation by the magnetic field of the spectrometer