Membrane Partitioning and Translocation of Hydrophobic

By immobilized liposome chromatography (ILC), membrane partition coefficients (KLM) of the cations were measured. The KLM values were coincident with ...
0 downloads 0 Views 166KB Size
Download PDF
7528

J. Phys. Chem. B 2000, 104, 7528-7534

Membrane Partitioning and Translocation of Hydrophobic Phosphonium Homologues: Thermodynamic Analysis by Immobilized Liposome Chromatography Qing Yang,†,‡ Xue-Ying Liu,† Koujirou Umetani,§ Tatsuya Ikehara,§ Seiji Miyauchi,§ Naoki Kamo,*,§ Takashi Jin,| and Jun Miyake† National Institute for AdVanced Interdisciplinary Research, Agency of Industrial Science and Technology, 1-1-4 Higashi, Tsukuba, Ibaraki 305, Japan; Laboratory of Biophysical Chemistry, Graduate School of Pharmaceutical Sciences, Hokkaido UniVersity, Sapporo 060-0812, Japan; and Institute of Electrical Sciences, Hokkaido UniVersity, Sapporo 060-0812 ReceiVed: March 31, 2000; In Final Form: May 31, 2000

Partitioning of hydrophobic cations, tetraphenylphosphonium (TPP+) and triphenylphosphonium homologues ((Phe)3-P+-(CH2)nCH3, n ) 0-6), into lipid bilayers was studied by chromatography on immobilized small and large unilamellar liposomes (SUVs with mean diam of 30 nm and LUVs with 100 nm). The liposomes were immobilized stably in chromatographic gel beads by avidin-biotin (Yang et al. J. Chromatogr. B. 1998, 707, 131). By immobilized liposome chromatography (ILC), membrane partition coefficients (KLM) of the cations were measured. The KLM values were coincident with molar partition coefficient, Km, which was determined for some of cations by an ultrafiltration method. Temperature dependence of KLM gave the thermodynamic quantities of membrane partitioning, ∆H° (enthalpy change) and ∆S° (entropy change). The same quantities were also measured with isothermal titration calorimetry (ITC), and the values of both methods were similar to each other. This confirms the usefulness of ILC on the basis of its simplicity. Except for TPMP+ (n ) 0), both ∆H° and ∆S° were positive, meaning that their membrane partitioning is entropydriven. The values of both ∆H° and ∆S° did not depend on whether n was odd or even. Thus, the odd-even pattern of the activation energy for the planar lipid membrane transport of these phosphonium cations (Ono et al. Biochemistry 1994, 33, 4312-4318) originates from the odd-even pattern of the translocation process within the hydrophobic core of the bilayer membrane. The origin of the odd-even pattern was discussed.

Introduction Hydrophobic phosphonium cations have been used as probes to estimate the membrane potential of cells, vesicles, and liposomes because they are permeable to lipid membranes.1-5 The interactions of the lipophilic ions with membranes have been studied using model membranes such as liposomes and planar lipid membranes. Ketterer et al.6 successfully presented a theoretical model for the lipophilic ion interaction with lipid bilayers. Flewelling and Hubbell7 calculated the potential energy profile within bilayer membranes of hydrophobic ion, and the deduced potential profile was essentially the same as that of Ketterer et al.6 Their model showed a potential profile including a binding region near the membrane surface and an energy barrier in the middle of the membrane (Figure 1). Lipophilic ions must surmount this potential barrier for their translocation across membranes (also see Appendix). Ono et al.8 demonstrated the transport of triphenylphosphonium homologues ((Phe)3-P+-(CH2)nCH3 (n ) 0-5)) through a planar lipid membrane when a voltage was applied across the membrane. In accordance with the theory of Ketterer et al.,6 they obtained βki values of various phosphonium ions and * Fax: +81-11-706-4984. E-mail: [email protected]. † National Institute for Advanced Interdisciplinary Research. ‡ Present address: Georgetown University Medical Center, Lombardi Cancer Center, Research Building E420, 3970 Reservoir Rd. NW, Washington, DC 20007. § Graduate School of Pharmaceutical Sciences, Hokkaido University. | Institute of Electrical Sciences, Hokkaido University.

Figure 1. Potential profile for lipophilic ion interaction with membrane bilayer. The free energy of binding (∆G°) and the energy barrier (∆Gq) of the translocation are shown. δ represents thickness of the binding region located near the surface. β is a linear partition coefficient, a ratio of the surface density of adsorbed ions to their volume density in an aqueous solution. ki is the rate constant for the ith translocation through membrane. The translocation rate of TPP+ was reported to be 10-2-10-3 s-1 (see Flewelling and Hubbell7).

measured the temperature-dependence of βki values, where β is a ratio of the surface density of adsorbed ions to their volume density in the aqueous solution and ki is a rate constant of transmembrane ion transport.8 The thermodynamic treatment

10.1021/jp001237k CCC: $19.00 © 2000 American Chemical Society Published on Web 07/15/2000

Partitioning and Translocation of Hydrophobic Phosphonium TABLE 1: Chemical Structure of Lipophilic Phosphonium Cation Homologues Used

yielded values of (∆Hb + ∆Hq) and (∆Sb + ∆Sq), where ∆Hb and ∆Hq stand for the enthalpy changes of the binding of phosphonium ions and the transmembrane ion transport (the enthalpy change of the potential barrier), respectively, and ∆Sb and ∆Sq, the corresponding entropy changes. Ono and his colleagues8 observed that both (∆Hb + ∆Hq) and (∆Sb + ∆Sq) for a phosphonium with odd n were larger than those of even n. They called this the “even-odd pattern”. Unfortunately, they could not measure the thermodynamic values of the binding and translocation separately. Immobilized liposome chromatography (ILC) is a recently innovated method (see Lundahl and Beigi9) to measure the interaction of solutes with membrane precisely and conveniently; variations of retention volume depend on the extent of solutemembrane interaction.10-12 Very recently, we successfully applied the ILC method to determine the partition coefficients for the membrane binding of these phosphonium cations.12 The chromatographic data are likely to reflect binding of the cation to outer leaflets of the liposomal membranes but may not reflect the transport across the membranes (see Discussion and Appendix). In the present paper, temperature-dependence on binding of the cations to unilamellar liposomes was measured by ILC to obtain ∆G° () -RTlnK), ∆H° and ∆S° for the binding of phosphonium cations to liposome membranes. Here, K represents a distribution coefficient of a phosphonium cation between liposomes and the aqueous solution, and ∆G°, ∆H° and ∆S°, respectively, represent the standard Gibbs free energy, enthalpy and entropy changes of the binding. In addition, K or ∆H° was measured by direct methods of filtration and isothermal titration calorimetry (ITC), and they agreed well with value obtained by ILC. Using the data of Ono et al.,8 we were able to obtain the thermodynamic parameters for the translocation of the cations through the energy barrier shown in Figure 1. Our results reveal that the odd-even pattern of the activation energy observed previously8 is originated from the odd-even pattern of the translocation process. Materials and Methods Materials. Sephacryl S-1000 Superfine was purchased from Pharmacia Biotech (Uppsala, Sweden). Lipophilic phosphonium cations used are listed in Table 1. Triphenylmethylphosphonium (TPMP+), triphenylethylphosphonium (TPEP+), triphenylpro-

J. Phys. Chem. B, Vol. 104, No. 31, 2000 7529 pylphosphonium (TPPP+), triphenylbuthylphosphonium (TPBP+), triphenylamylphosphonium (TPAP+), triphenylhexylphosphonium (TPHP+), triphenylheptylphosphonium (TPHPP+), and tetraphenylphosphonium (TPP+) were from Tokyo Kasei (Tokyo, Japan). Egg-white avidin was from Calbiochem (La Jolla, CA). Egg yolk phosphatidylcholine (EPC > 99%), and 1,2-dioleoylphosphatidylethanolamine-N-(cap biotinyl) (biotin-cPE) were from Avanti Polar Lipids (Alabaster, AL). 2-[4-(2hydroxyethyl)-1-piperazinyl]-Ethanesulfonic acid (HEPES) was from Dojindo Laboratories (Kumamoto, Japan). 4-Nitrophenyl chloroformate and 4-(dimethylamino)pyridine were from Aldrich (Milwaukee, WI). Synthesis of Avidin Adsorbent. Sephacryl S-1000 gel was activated by 4-nitrophenyl chloroformate to the chloroformate density of 20-30 µmol/mL gel, and then avidin was coupled to the activated gel at about 3 mg/mL gel.11 Washing of the gels was done on a 10-µm filter (Millipore, Bedford, MA) fixed in a glass funnel. Alternatively, avidin was coupled to CNBractivated Sepharose 4B at 3.0-3.5 mg/mL of gel bed according to the manufacturer’s specifications. The avidin-gels were stored at 4 °C in buffer H (10 mM HEPES, 150 mM NaCl, pH 7.4) supplemented with 3 mM NaN3. Preparation and Immobilization of Biotinylated Liposomes. Small and large unilamellar liposomes, SUVs and LUVs, respectively, were prepared by probe sonication and extrusion as described previously.12 The SUVs or LUVs were composed of EPC supplemented with 2 mol % of biotin-cPE. The mean diameter was 30 ( 10 nm for the SUVs and 100 ( 20 nm for the LUVs as analyzed by dynamic light scattering.12 For immobilization, the biotinylated liposomes were mixed with avidin-Sepharose 4B (for SUVs) or Sephacryl S-1000 (for LUVs) for 2-3 h at 23 °C or overnight at 4 °C under nitrogen atmosphere. Nonbound liposomes were removed by washing with buffer H on a 10-µm filter. Phospholipids of the immobilized liposomes in the gel beads were determined essentially according to the method of Bartlett13 with modifications as described in detail by Yang et al.14 Immobilized Liposome Chromatography (ILC). The immobilized liposomes were packed in a 5-mm i.d. glass column (HR 5/5, Amersham Pharmacia Biotech) to form a 1 mL gel bed. The liposome column was placed in a column oven (CO8020-C, Tosoh) equipped with an injector, connected to a HPLC pump (CCPM-II, Tosoh) and an UV detector (UV-8010, Tosoh). The whole system was controlled by an IBM computer, and chromatograms recorded were analyzed by HPLC System 1 (Tosoh). Lipophilic phosphonium cations (1-2.5 mM, 10-15 µL) were applied to the liposome column, eluted with buffer H at a flow rate of 0.3 mL/min at 23 °C, and detected at a wavelength of 267 nm. Previously, we found that the elution volumes were independent of the flow rate.12 A coiled tubing ca. 200 cm in length was connected to the outlet of the pump to increase the column back-pressure for an accurate flow of the HPLC pump operating at low flow rate. For the temperature dependence of the solute retardation upon ILC, the temperature in the column oven was varied from 4 to 40 °C with an accuracy of (0.1 °C. The sample was loaded in a 2-mL loop in the oven to be equilibrated with the oven temperature for 10 min before injection. A buffer reservoir was placed in a water bath controlled at the same temperature as the run temperature in the oven. To attain the complete temperature equilibration, the experiments started after at least 30 min of reequilibration. The running buffer was passed through a mixer, a coiled tubing ca. 200 cm in length, and a 2-mL loop, which were all placed inside

7530 J. Phys. Chem. B, Vol. 104, No. 31, 2000

Yang et al.

the oven. The temperature of the running buffer was thus adjusted to the desired value. Membrane Partition Coefficients and Thermodynamic Parameters Measured by ILC. The membrane partition coefficient (KLM) for the solute partitioning between the mobile and stationary liposome membrane phase can be calculated by eq 112 as below:

KLM ) (Vr - Vo)/0.755A

(1)

where Vr is the retention volume of a solute on the liposome column; Vo is liquid volume of the liposome column (mL). A is the amount of immobilized liposomes in the column (mmole of phospholipid). Derivation of this equation has been described previously11,12 in detail. Briefly, a distribution coefficient (K) for a solute between the mobile and stationary phases is generally calculated by the chromatographic equation,15

K ) (Vr - Vo)/Vs

(2)

where Vs is the stationary phase volume of a column. This equation can be applied to ILC to obtain the membrane partition coefficients since the solute retention on the liposome column is mainly dominated by the solute-membrane partitioning mechanism. The stationary phase volume of the immobilized unilamellar liposome membranes can be calculated from the volume per phospholipid molecule in the liposomal membrane (for EPC, 1253 Å3 in anhydrous membranes). On the basis of the linear correlation between log KLM and the reciprocal of temperature (1/T) (shown in Results), thermodynamic quantities for the cation-membrane partitioning were analyzed by the following equation:

Isothermal Titration Calorimetry (ITC). The heat produced through the liposome (LUVs) binding of phosphonium cations were measured directly with a VP-ITC microcalorimeter (MicroCal, Northampton, MA). Prior to measurements, solutions were degassed under vacuum. The sample cell had a volume of 1.73 mL (Vc) and contained LUVs suspended in buffer H at a lipid concentration of 10 mM. Solutions of 10 mM phosphonium cations were placed in a 250-µL syringe, with which LUVs were titrated; 10-µL aliquots were injected every 4 min under continuous stirring of the LUVs solution. After 20 injections, the amounts of heat generated were somewhat leveled off except for TPMP. This leveling may come from the saturation of the binding to the liposomes and was not complete, which gave some errors for estimation of values by iterative fitting described below. We did not increase the concentrations of titration phosphonium solution, because we wanted to examine the binding at as low concentration of phosphonium cation as possible. The heat, ∆Q, through the binding should be proportional to the amounts of phosphonium-LUVs complex, then ∆Q is expressed by

∆Q ) ∆HAb

where ∆H is the molar enthalpy change and Ab is the amount of phosphonium cations bound to LUVs. Since as described above, the heat generated showed the level-off at the later stage of the titration, ∆Q was analyzed by the following equation:

AoCLKcalCf V 1 + KcalCf c

∆Q ) ∆HAb ) ∆H

CtVc ) CfVc + Ab

log KLM ) -∆G°/2.3RT ) -(∆H°/2.3RT) + ∆S°/2.3R (3) where ∆H° and ∆S° are enthalpy and entropy changes associated with the solute binding to membranes, respectively, R is the gas constant, and T is the absolute temperature. Molar Partition Coefficients Determined by Ultrafiltration. One milliliter of a SUV suspension (4-16 µM of phospholipids) was mixed with a phosphonium cation (4-15 µM) and applied to the upper chamber of an ultrafiltration cassette (MPS micropartition kit, Amicon, Beverly, MA) with a YM10 membrane (Mw cutoff, 10 000) followed by centrifugation at 2000g for 15 min at 25 °C. The concentration of a phosphonium cation in the filtrate was determined by reversephase HPLC on Mightysil RP-250-4.6 column (bead diameter 5 µm, Kanto Kagaku, Tokyo) at a wavelength of 267 nm. The liposome-bound phosphonium cation was thus calculated by subtracting the free amount (in the filtrate) from that added. Concentrations of the liposomes were determined as phosphorus by a modified method of Bartlett.14 The molar partition coefficient, Km, for the cation-membrane binding was calculated as follows:

Km ) (Cint - Ceq)ω1/Ceqω2

(4)

where Cint and Ceq are the concentrations of the phosphonium cations added before and after the filtration, respectively. ω1 and ω2 are the weights of the water and lipid, respectively. Since the cations interacts only at the outer leaflet of the liposomal bilayer in the absence of membrane potential (see Results), and since the ratio of the phospholipid molecules in the outer and inner leaflets of the liposomes is 2.1:1 for SUVs,16 ω2 was calculated as (2.1/3.1) × (total weight of the lipids).

(5)

(6) (7)

where Ao represents the maximum amounts of the binding per mole of lipid; Kcal is the association constants determined by the calorimetry; CL is lipid concentration; Ct and Cf are the concentrations of total and free phosphonium cations, respectively. Elimination of Cf from eqs 6 and 7 yields the following equation:

(

)/

∆Q ) ∆HVc A - xA2 - 4AoCLCt 2

(8)

A ) 1/Kcal + AoCL + Ct

(9)

where

Eqs 8 and 9 describe that ∆Q is expressed by Ct, and Kcal was estimated by an iterative nonlinear least-squares fitting of ∆Q data. At this fitting, we used the relation that KLM ) Km ) AoKcal. The former relation was confirmed in the present work (see Results and Discussion). Limiting Molar Conductivities of the Phosphonium. Molar conductivity of the various phosphonium cations, Λ, in deionized water was measured at 1 kHz with an AC bridge (model 4265B, Yokogawa-Hewlett Packard, Tokyo). Deionized water with a specific resistance of 18.3 MΩcm was obtained from Milli-Q SP (Millipore, Bedford, MA). The samples were degassed by bubbling water-saturated nitrogen gas. Temperature was kept constant with an accuracy of (0.02 °C. Values of Λ were plotted against the root of the phosphonium concentrations to estimate the molar conductivity at infinite dilution, Λ°. The limiting molar conductivities of the phosphonium cations, λ+° were calculated by subtracting λBr° (limiting conductivities of

Partitioning and Translocation of Hydrophobic Phosphonium

J. Phys. Chem. B, Vol. 104, No. 31, 2000 7531

TABLE 2: log KLM Valuesa for Phosphonium Cationsb as Determined by ILCc immobilized liposomesd bilayer outer layer

SUVs LUVs SUVs LUVs

TPMP+ (n ) 0)

TPEP+ (n ) 1)

TPPP+ (n ) 2)

TPBP+ (n ) 3)

TPP+

TPAP+ (n ) 4)

TPHP+ (n ) 5)

TPHPP+ (n ) 6)

1.33 1.28 1.49 1.58

1.31 1.22 1.47 1.52

1.32 1.20 1.48 1.50

1.51 1.37 1.66 1.67

1.59 1.53 1.74 1.83

1.88 1.71 2.03 2.01

2.35 2.14 2.51 2.44

2.88 2.72 3.03 3.02

a The ILC zonal runs were done at 23 °C. b The number n indicates the number of methylene groups in the cations (see Table 1). c Values are calculated assuming either outer leaflets only or bilayers of the liposomes as the stationary phase (see eq 1). d SUVs and LUVs are small and large unilamellar vesicles, respectively.

bromide anion) from Λ°. The values17 of λBr° used were 63.20 (15 °C), 70.57 (20 °C), 78.17 (25 °C), and 86.00 (30 °C) Scm2mol-1.

TABLE 3: Molar Partition Coefficient (Km) for the Phosphonium-Liposome Interactions Determined by Ultrafiltrationa expt.

+

1 2d 3e

Results and Discussion Table 2 summarizes the relationship of log KLM and n at 23 °C. The liposomes are SUVs and LUVs. The values indicated by the “bilayer” were calculated under the assumption that the phosphonium ions are distributed in both the outer and inner leaflets of liposomes, and those indicated by the “outer monolayer” were calculated assuming that phosphonium distributes only in the outer leaflet. It is probable that the phosphonium penetrates little through the bilayer membranes. The plausible reasons for this difficulty are described in Appendix. In addition, the preliminary results using 31P NMR suggested no permeability of TPP+, TPMP+, or TPHP+ through bilayer membranes in the absence of electrical potential (Jin et al., unpublished), as is the same condition as the present ILC. As shown in a previous paper,12 the elution peak was single: If two populations (i.e., in the outer and inner leaflets or the inner space of liposome) exist, the elution pattern may be composed of two components. Even if some phosphonium might transport to reach the inner leaflet through the high potential barrier, the amounts of these ions might be minimal and they may be largely diluted to a concentration too low to be experimentally detected. This is supported by the fact that the retention time was independent of the flow rate as shown previously.12 The values indicated by the “outer monolayer”, therefore, may be the true values. If the present our assumption that the phosphonium cations interact only with the outer leaflet under the present experimental condition might not be right; i.e. if cations interact with both leaflets, the ∆G° values reported later are changed maximally by 0.22 kcal/mol for SUVs and by 0.41 kcal/mol for LUVs (see Table 2). It is very interesting that among TPMP+, TPEP+, and TPPP+, the log KLM value of TPMP+ is the largest; the interaction between TPMP+ and the membrane is strongest although the chain length of TPMP+ is the shortest, which is contrary to expectation. For n g 3, log KLM was linear when plotted against n, which accords with the expectation that more hydrophobic ions interact more strongly with liposome membrane. The anomaly of TPMP+ also appeared in the limiting conductance (see Figure 4). The molar partition coefficients (eq 4) were determined by ultrafiltration (Km) for TPAP+, TPHP+, and TPHPP+, and are listed in Table 3. The values for other phosphoniums were smaller and were difficult to determine precisely. Table 3 reveals fairly good agreement with values of the “outer monolayer” listed in Table 2. As shown later (Table 4), the values of ∆G°, ∆H°, and ∆S° were determined from the temperature dependence of KLM. Values for the binding of TPP+ to LUVs at 25 °C determined by ILC were similar to those reported by Flewelling and Hubbell.7 Their values of ∆G°, ∆H°, and ∆S°

phosphonium cation

c

TPAP TPHP+ TPHPP+

n

log Km (n ) 8)b

4 5 6

2.06 ( 0.05 (SD) 2.66 ( 0.04 (SD) 3.04 ( 0.03 (SD)

a

Values calculated using eq 4. Initial concentrations employed are given in footnotes c-e. b The number of experiments done. c Cint values were 5.28, 11.2, 14.7, 20.3, 24.2, 29.5, 33.3, and 40.6 µM, and the corresponding Ceq values were 3.11, 6.28, 7.22, 9.82, 13.0, 15.5, 17.1, and 19.5 µM; corresponding values of ω1/ω2 were in a range of 123124. d Cint values were 5.15, 10.2, 15.0, 19.5, 25.1, 30.0, 35.0, and 40.2 µM, and the corresponding Ceq values were 2.42, 4.63, 6.42, 9.03, 12.3, 15.2, 16.1, and 18.6 µM; corresponding values of ω1/ω2 were in a range of 395-401. e Cint values were 16.9, 22.7, 24.9, 34.2, 35.0, 42.7, 44.2, and 54.4 µM, and the corresponding Ceq values were 4.95, 5.80, 7.02, 8.74, 9.66, 10.9, 12.2, and 14.3 µM; corresponding value of ω1/ω2 was 401.

TABLE 4: Thermodynamic Quantities for the Binding of Phosphonium Cations to the Outer Leaflet of Unilamellar Liposomes ∆G° (kcal/mol)

∆H° (kcal/mol)

∆S° (cal/mol K)

cation

SUVs

LUVs

SUVs

LUVs

SUVs

LUVs

+

-2.12 -2.06 -2.06 -2.33 -2.86 -3.50 -4.22 -2.46

-2.19 -2.11 -2.09 -2.29 -2.79 -3.42 -4.13 -2.54

0.12 1.54 3.09 3.66 3.39 3.09 3.01 2.13

-0.21 2.46 4.53 6.27 5.79 5.69 5.24 3.41

7.52 12.1 17.3 20.1 21.0 22.1 24.3 15.4

6.63 15.4 22.2 28.7 28.8 30.6 31.5 20.

TPMP TPEP+ TPPP+ TPBP+ TPAP+ TPHP+ TPHPP+ TPP+

for the binding determined by equilibrium dialysis and spin label method were -2.8 ( 0.5 kcal/mol, 3.5 ( 0.3 kcal/mol and 21 ( 1 cal/mol K, respectively. The slight difference can be ascribed to the difference in physical property of the LUVs. Another reason for the difference might be the method used. This good agreement between ILC and the conventional methods proves that the novel ILC method can be useful for accurate measurements of membrane partition coefficients and thermodynamic parameters for the solute-membrane interactions. The ILC method has some advantages over conventional methods in that the experiments can be conveniently and rapidly performed using only one immobilized liposome column. In addition, even for weak interactions, the partition (interaction) can be detected precisely: for conventional methods such as equilibrium dialysis, ultrafiltration, or detection of a concentration change by an electrode method,18-19 if the interaction is weak, the result may contain errors because of small observed changes. Figure 2 shows van’t Hoff plots of log KLM of various phosphonium ions of varying n. This figure reveals that the larger value of TPMP+ is evident at lower temperature (4 or 10 °C), and less pronounced at the temperature range from 10 to 40 °C. The linearity of this plot is fairly good (r ) 0.96-

7532 J. Phys. Chem. B, Vol. 104, No. 31, 2000

Yang et al.

Figure 2. van’t Hoff plots of KLM. 3, TPMP+ (n ) 0); b, TPEP+ (n ) 1); 0, TPPP+ (n ) 2); O, TPBP+ (n ) 3); 2, TPP+; 4, TPAP+ (n ) 4); 1, TPHP+ (n ) 5); ], TPHPP+ (n ) 6).

0.99). According to eq 3, thermodynamic quantities of ∆G°, ∆H° and ∆S° which were associated with the binding were calculated and are listed in Table 4 and plotted in Figure 3. The enthalpy changes were measured directly by isothermal titration calorimetry (ITC), and values estimated are shown by data points marked by ]. The value of ∆G was calculated by -RTlnKcal. Although there are differences in values of these thermodynamic quantities, the tendency against n (described below) is the same for both. As described in Materials and Methods, the values estimated from ITC might have some errors. It is obvious that the binding is entropy-driven. There is no significant difference in ∆G° between SUVs and LUVs, while values of ∆H° and ∆S° differ between them (discussed below). For n e 3, ∆G° stays relatively constant (although the curve is somewhat convex) with both ∆H° and ∆S° being increased with an increase in n. In contrast, for n > 3, the increase in ∆H° ceases followed by a slight decrease, and the increase rate in ∆S° becomes smaller. Thus, the behaviors of the phosphonium binding are different below and above n ) 2 or 3. This is probably caused by the fact that the binding distance in membranes for lipophilic ions (δ in Figure 1) is restricted (approximately 3-6 Å). The distance between P+ and the end of phenyl is 4.5 Å, and that between P+ and the methyl end of TPBP+ is 5.2 Å, suggesting that this site might be able to accommodate the (Phe)3-P+-(CH2)4 and the longer alkyl chain moiety might reside in the hydrocarbon chain in the phospholipids. Seelig and Ganz,20 and Beschiaschvili and Seelig21 measured the thermodynamic quantities of the transfer of lipophilic molecules from aqueous phase to liposomes. They used tetraphenylborate (TPB-) as one of the lipophilic ions which is a structural analogue of tetraphenyphosphonium (TPP+) used in the present work. Interestingly, the large negative enthalpy change was observed for the TPB- binding to SUVs, implying that the binding is enthalpy-driven contrary to the common concept of entropy-driven.22 On the other hand, the present work shows that the TPP+ binding is entropy-driven. As described above, the direct enthalpy measurement gave a very small positive value. Why are the enthalpy changes of these two lipophilic ions so different although they are very similar structural molecules aside from the charge? This awaits for an interesting further investigation. Seeling and his colleagues20,21 also reported the following: (1) the enthalpy change of the binding of a cyclic peptide or 2-(p-toluidinyl)naphthalene-6-sulfonate is strongly dependent on the vesicle size, but (2) the Gibbs free energy is relatively

Figure 3. Thermodynamic quantities of ∆G°, ∆H°, and ∆S° of the liposome binding of various phosphonium cations. The abscissa, n, represents number of methylene in the alkyl chain (see Table 1). The closed circles (b) represent data of LUVs and open circles (O), SUVs. The data (]) and broken lines are those obtained from ITC experiments.

independent; in addition, (3) for the larger vesicle, the enthalpy change increased (due to a decrease in the absolute values of negative enthalpy). Their second observation is the so-called “entropy-enthalpy compensation rule”. The results in Table 4 and Figure 3 accord with their observations in 2 and 3 above. As discussed above, the retention of the lipophilic cations on the liposome column is believed to be caused by the interaction between the cation and the outer leaflet of bilayer membranes. Therefore, the values of ∆G°, ∆H° and ∆S° determined by ILC are thought to correspond to the binding parameters of ∆Gb, ∆Hb and ∆Sb defined by Ono et al.8 Since the sum of the free energy (∆G° + ∆Gq), enthalpy (∆H° + ∆Hq), and entropy (∆S° + ∆Sq) for the binding and transport inside membranes (see Figure 1) has been determined, we were able to calculate ∆Gq, ∆Hq and ∆Sq. Results are listed in Table 5.The values of the energy barrier (∆Gq) are positive, as is expected from Figure 1. With an increase in the alkyl chain length, ∆Gq gradually decreased, but the values essentially stayed constant, suggesting that the part contributing the most

Partitioning and Translocation of Hydrophobic Phosphonium

J. Phys. Chem. B, Vol. 104, No. 31, 2000 7533

TABLE 5: Estimation of Thermodynamic Quantities for the Translocation of Phosphonium Cations Across the Hydrophobic Core of Planar Lipid Bilayersa

can be estimated from the limiting molar conductivity λ+0 by the following equation:

phosphonium cation

∆Gq (kcal/mol)

∆Hq (kcal/mol)

∆Sq (cal/mol K)

TPMP+ TPEP+ TPPP+ TPBP+ TPAP+ TPHP+ TPP+

19.4 18.4 17.7 17.6 17.1 17.0 16.5

30.5 34.6 20.1 35.1 18.7 47.6 25.7

37.5 54.5 8 59 1.7 102.7 30.8

a The ∆Gq + ∆G , ∆Hq + ∆H , and ∆Sq + ∆S obtained by Ono et b b b al.8 were subtracted by ∆G°, ∆H°, and ∆S° which are assumed to equal to ∆Gb, ∆Hb, and ∆Sb. Here, ∆G°, ∆H°, and ∆S° are those in LUVs of Table 4.

Figure 4. Limiting molar conductivities (λ+°) of various phosphonium cations at 25 °C. The abscissa represents n whose meaning is given in Figure 3 and Table 1.

is (Phe)3P+. Flewelling and Hubbell7 estimated ∆Gq of TPP+ to be 20 kcal/mol by experiment and 17.8 kcal/mol by theoretical calculation. Our value of 16.5 kcal/mol shows good agreement. As seen in Table 5, both ∆Hq and ∆Sq are positive, meaning that the positive value of the energy barrier (∆Gq) is originated from the positive ∆Hq, while ∆Sq is favorable for the translocation. Phosphonium cations of odd n have large values of both ∆Hq and ∆Sq, and the ∆Gq value remains relatively constant for all cations; enthalpy and entropy compensation seems to hold. What are the reasons for this even-odd pattern on a molecular basis? Although not clear, there have been several observations of the even-odd pattern: Ono et al.8 reported that bromide salts of phosphonium cations with odd n have less solubility in water than even n, and that the chemical shift of 31P NMR also shows the zigzag pattern. A similar zigzag pattern was observed in partial molar volume.23 In addition, the even-odd pattern was also observed for the electric conductivities in water;8 Figure 4 shows our data on the limiting molar conductivities of phosphonium cations, λ+0, which are essentially the same as those of Ono et al.,8 although our deflections of the zigzag seem smaller. The conductivity of TPMP+ (n ) 0) is smaller than the value extrapolated from the values of n ) 1-6, and it is noteworthy that the binding behavior of TPMP+ is different from others (see Figure 2 and Table 2). Samoilov et al.24 developed a theory concerning the ionic conductivity or hydration of ions: In aqueous solutions, one water molecule vibrates within a cell (compartment) at the most stable position. There exists a potential barrier (an activation energy) between these cells (denoted by Ei), over which water molecules must climb to diffuse. The presence of ions alters Ei, and the magnitudes of the change are denoted by ∆Ei which

1/λ+0(∂λ+0/∂T) + 1/T - (1/Dw)(∂Dw/∂T) ) ∆Ei/RT2 where Dw is the self-diffusion coefficient of water molecules in pure water. The positive value of ∆Ei means that the water molecule just adjacent to the ion is restricted by the ion. Results are 0.68 (n ) 0), 1.23 (n ) 1), 0.9 (n ) 2), 0.94 (n ) 3), 0.78 (n ) 4), 1.18 (n ) 5), and 0.62 kcal/mol. The values of phosphonium cations are larger than the ordinary inorganic ions reported: for example, 0.39, 0.17, -0.20, -0.30, -0.34, -0.10, -0.14, and -0.15 kcal/mol were reported for Li+, Na+, K+, Rb+, Cs+, Cl-, Br-, and I-, respectively. It is interesting that ∆Ei of the odd phosphonium is larger than the others. The limiting molar conductivities of odd phosphonium cations are larger (see Figure 4), and the partial molar volumes in aqueous solutions (at infinity dilution) are also smaller (data not shown; see Umetani23). These mean that the size of odd phosphonium ions including the hydrated water layer is smaller, but the interaction between the hydrated water and the cation is stronger. It is reasonable to consider that these “odd-even” properties concerned with the interaction of water molecules result in the odd-even pattern of ∆Hq and ∆Sq. One plausible interpretation is that in the binding site located near the membrane surface (see Figure 1), the bound phosphonium ions are still hydrated (especially for odd phosphonium cations which have hydrated water with strong affinity); the translocation process toward hydrophobic hydrocarbon chain moiety requires the dehydration which may give rise to the odd-even pattern of ∆Hq and ∆Sq. This is, however, only one possible interpretation, and an exact molecular mechanism must be investigated in future. In conclusion, the present paper showed that (1) the ILC method is a useful method for estimating the binding or partitioning of solute to liposomal membrane even though the interaction is weak, (2) this method is also pertinent for estimating the thermodynamic quantities, and (3) the odd-even pattern of the translocation through membranes is originated from the translocation of the phosphonium cations across the membrane hydrophobic core. It is necessary to elucidate the molecular mechanism of this pattern. Appendix This appendix treats the trans-membrane translocation of the phosphonium cations, because we, in this article, described that these cations are permeable under a certain condition and may not be under the other condition. The most important restriction on this respect is the electro-neutrality condition stating that the positive electric current caused by the ion movement should be compensated by the negative one when the electric field is not imposed externally. (1) The translocation of phosphonium cations through biological membranes or liposomes which contain an electrogenic pump or molecular machinery. Phosphonium cations are mostly used to estimate the membrane potential created by this pump.2-5 The electric current due to the translocation of the cation is electrically compensated by that of an ion-translocation made by the pump as well as the translocation of other mobile ions. Therefore, the lipophilic cations are allowed to permeate. (2) The translocation of phosphonium cations through bilayer phospholipid membranes when an external electric voltage is applied. In this case, the electro-neutrality condition does not necessarily hold, and the lipophilic phosphonium ions can

7534 J. Phys. Chem. B, Vol. 104, No. 31, 2000 permeate through membranes whose permeate rate is measured by the electric current through the membrane.1,6,8 But, the translocation rate of phosphonium cations is generally much smaller than that of lipophilic anions,6 the rate of which is determined by the height of the barrier shown in Figure 1. (3) The translocation of phosphonium cations when lipophilic cations are added to tightly sealed vesicles suspended in a medium containing hydrophilic ions such as Na+, Cl-, H+, and OH-. The present experiment is of this category. Due to the electro-neutrality condition, the translocation rate of phosphonium cations is determined by the rate of these hydrophilic ions whose membrane permeation is very slow. In addition, it is noted that, as described above, the translocation rate within membranes is not high,7 ex. 10-2-10-3 /s for TPP+. Therefore, in the present experimental condition, the phosphonium cations seem not to translocate through the liposome membranes, if the liposome is tightly sealed: The tight seal is well expected from a calcein experiment described elsewhere.12 Acknowledgment. Q.Y is very grateful to Stanley Elec. Co. Ltd., Japan, for its financial support. This work was supported by the Protein Molecular Assembly Project, The Agency of Industrial Science and Technology, The Ministry of International Trade and Industry (Japan), and by Grants-in-Aid for Scientific Research from the Japanese Ministry of Education, Science, Sports and Culture. References and Notes (1) Bakeeva, L. E.; Grinius, L. L.; Jasaitis, A. A.; Kadziauskas, Yu. P.; Liberman, E. A.; Skulachev, V. P.; Topali, V. P.; Tsofina, L. M.; Vlaoimirova, M. A. Biochim. Biophys. Acta 1970, 216, 1-12.

Yang et al. (2) Rottenberg, H. Methods Enzymol. 1979, 55, 547-569. (3) Harold, F. M.; Papineau, D. J. Membr. Biol. 1972, 8, 27-62. (4) Schuldiner, S.; Kaback, H. R. Biochemistry 1975, 14, 5451-5461. (5) Kamo, N.; Racanelli, T.; Packer, L. Methods Enzymol. 1982, 88, 356-360. (6) Ketterer, B.; Neumacke, B.; Lo¨a¨uger P. J. Membr. Biol. 1971, 5, 225-245. (7) Flewelling, R. F.; Hubbell, W. L. Biophys. J. 1986, 49, 541-552. (8) Ono, A.; Miyauchi, S.; Demura, M.; Asakura, T.; Kamo, N. Biochemistry 1994, 33, 4312-4318. (9) Lundahl, P.; Beigi, F. AdV. Drug DeliV. ReV. 1997, 23, 221-227. (10) Beigi, F.; Gottschalk, I.; Lagerquist Ha¨a¨gglund, C.; Haneskog, L.; Brekkan, E.; Zhang, Y.; O ¨ sterberg, T.; Lundahl, P. Int. J. Pharm. 1998, 164, 129-137. (11) Yang, Q.; Liu, X.-Y.; Ajiki, S.-I.; Hara, M.; Lundahl, P.; Miyake, J. J. Chromatogr. B. 1998, 707, 131-141. (12) Yang, Q.; Liu, X.-Y.; Umetani, K.; Kamo, N.; Miyake, J. Biochim. Biophys. Acta 1999, 1417, 122-130. (13) Bartlett, G. R. J. Biol. Chem. 1959, 234, 466-468. (14) Yang, W.; Liu, X.-Y.; Yoshimoto, M.; Kuboi, R.-I.; Miyake, J. Anal. Biochem. 1999, 268, 354-362. (15) Freeman, D. H. Anal. Chem. 1972, 44, 117-120. (16) Huang, C.; Mason, J. T. Proc. Natl. Acad. Sci., U.S.A. 1978, 75, 308-310. (17) Robinson R. A.; Stokes, R. H. In Electrolyte Solutions; Butterworths: London, 1959. (18) Demura, M.; Kamo, N.; Kobatake, Y. Biochim. Biophys. Acta 1987, 903, 303-308. (19) Demura, M.; Kamo, N.; Kobatake, Y. Biochim. Biophys. Acta 1987, 894, 355-364. (20) Seelig, J.; Ganz, P. Biochemistry 1991, 30, 9354-9359. (21) Beschiaschvili, G.; Seelig, J. Biochemistry 1992, 31, 10044-10053. (22) Tanford, C. H. The Hydrophobic Effect: Formation of Micelles and Biological Membranes; Wiley: New York, 1980. (23) Umetani, K. Master’s Thesis, Graduate School of Pharmaceutical Sciences, Hokkaido University, 1997 (in Japanese). (24) Samoilov, O. Ya. Hydration of Ions: Structures of Electrolyte Solutions; (translated by Uedaira H. Chijin-Shyoten: Tokyo, 1976).