Biochimie 71 (1989) 37-47 Socidtd de Chimie biologique/Elsevier, Paris 37

Polyene-sterol interaction and selective toxicity

Claude Michel GARY-BOBO

Service de Biophysique, D~partement de Biologie, CEN Saclay, 91191 Gif-Sur-Yvette. France

(Received 10-1-1988, accepted after revision 17-3-1988)

Summary - - From permeability experiments carried out with series of amphotericin B derivatives in both biological and model membranes, it was concluded that derivatives, whose carboxyl group at the C18 position is blocked by substitution, are much more efficient at inducing permeability in ergosterol- containing than in cholesterol-containing membranes, whereas derivatives whose carboxyl group is free and ionizable are equally efficient in both membranes types.

Binding measurements on erythrocyte membranes showed that all amphotericin B derivatives simply partition between membrane lipids and aqueous medium, according to their lipid solubility. There is no relationship between binding and efficiency in inducing permeability.

Permeability studies carded out on lipidic vesicles containing various sterels showed that: i) deriva- tives having their carboxyl free induced permeability of the 'channel' type, regardless of the sterol present, and no detectable permeability in sterol-free membranes; 2) derivatives whose carboxyl group is blocked induce channels only in membranes containing ergosterol or sterols having an alkyl side chain identical to that of ergosterol. In the presence of other sterols or in sterol-free membranes, their iono- phoric activity is poor and always of the 'mobile-carrier' type.

A model of polyene-sterol interaction is proposed, accounting for the data obtained with both biological and model membranes.

amnh•tprie in !~ doriv~tiv~_~ / ~te_rals / h i n d i n . / ian ic p e r m e a h i l i l y / e rythrocvtes / f luid ves ic les

I n t r o d u c t i o n

Polyene macrolides belong to the category of antibiotic ionophores. They interact with cell membranes, induce extensive permeability to ions and small non-electrolytes which eventually leads to cell death. The particular feature of polyenes is that their ionophoric action depends upon membrane sterol. The molecular mecha- nism of this action is not completely understood. However, on the basis of a large body of function- al evidence, mainly obtained with membrane models, such as black lipidic films and lipo- somes, the generally accepted hypothesis may be summarized as follows: polyenes penetrate into the lipid matrix of the membrane, where they

interact with sterol, forming complexes. These polyene-sterol complexes organize themselves into pores or channels spanning the membranes. Several molecular models of these multimolec- ular structures have been proposed ([1-3], for a review see [4]). Furthermore, it has been shown on both biological and model membranes that ergosterol-containing membranes are more sensitive to polyene than cholesterol containing ones. On this basis, polyenes are classified as anti-fungal antibiotics [4]. As a matter of fact, one of them, amphotericin B (AMB), is the main drug used in human therapy against syste- mic mycosis. However, the difference between the sensitivity of pathogenic microorganisms, whose membranes contain ergosterol, and that

Abbreviations: AMB: amphotericin B; AME: amphotericin B methyl ester; N-Fm-AMB: N-fructosyl amphotericin B; N-Ac-AMB. N-acetyl amphotericin B; N-Fm-AME: N-fructosyl amphotericin B methyl ester; N-Ac-AME: N-acetyi amphoteri- cin B methyl ester; FCCP:carbonyl cyanid-p-trifluoro-methoxyphenyl hydrazone.

38 C.M. Gary-Bobo

Amphotericin B , ~ O,,,,~CH 3

) - _ . . - . . . . - 1 . o. ' ; ' ,

i o. o. , . HzC OH

OH OH

oH + N-fruc'n'"' r . c o o - H6 oa~~ ' N - a c e t y l "~H O~Z~'~ " 'c°° -- OCH,

c", c . ,

OH N.HzC H z'll~ HO~ . ~ ' ~ OH N-fructosyl .JL H" r nn+. HI~ %~'"Y' V "~.H~ H NHCOCH~, methyl ester~ )" "~ . . . . . . . 3 - " . " N - a c e t y l .... o"~%..cooc"~

Am B ~ O H methylAmeSfer~."B ~ " ' OH"~ ~ ° H l 1

Fig. 1. Structural formulae of amphotericin B and of the four main derivatives studied.

of animal cells, whose membranes contain cholesterol, is rather small. As a result, the selective anti-fungal toxicity of AMB is poor, and its clinical use is severely limited by its very detrimental side effects, such as nephrotoxicity, anemia and electrolyte disturbances [5].

In an attempt to improve the anti-fungal selec- tive toxicity of polyene macrolides and to obtain more efficient and less toxic drugs-, numerous natural and semi-synthetic derivatives of AMB have been studied. The selective toxicity of these derivatives has been tested, comparing the damage, such as potassium leakage and growth inhibition induced in yeasts and mycoplasma (representatives of pathogenic microorganisms) to those, such as potassium leakage and hemoly- sis induced in human red blood cells, (represen- tatives of host cells). Along this line of research, several derivatives of AMB and vacidin A (aromatic heptaene) have been studied [6] by comparing the results obtained on yeast and human red blood cells with those obtained in a model system: the large unilamellar lipid vesicle

(LUV). Typical examples of these derivatives are listed in Fig. 1. They differ mainly by the structure of what may be called their 'polar head', that is, by modification of the amino group on the glycosamine residue and the carboxyl group at the C~8 position of the macro- cycle. In the cases of AMB and vacidin A, both groups are free and ionized at physiological pH. The d6rivatives may be classified into three groups: 1) derivatives whose amino group is blocked by amidation as in N-acetyl amphoteri- cin B (N-Ac-AMB); 2) derivatives whose car- boxyl group is either absent (as in perimycin) or blocked by esterification as in amphotericin B methyl ester (AME); 3) derivatives whose amino and carboxyl group are blocked, as in N-acetyl-amphotericin B methyl ester (N-Ac-AME).

Typical results obtained with these various derivatives are listed in Table I. The conclusions which can be drawn from this study may be summarized as follows: 1) very good parallelism can be drawn between the results obtained, on

Polyene-sterol interaction and selective toxicity 39

Table !. Ionophoric activities of amphotericin B derivatives.

AMB Yr,50 a Er,50 a P50 b (/zg / ml) derivative

(/zg / ml) (/zg / ml) ergosterol cholesterol vesicles vesicles

AMB 0.58 0.48 0.10 0.20

N-Ac-AMB 8.10 9.10 0.4 0.8

AME 0.28 7.6 1.4 28.0

NoAc-AME 10.0 12.0 - -

aYe0, Era0: polyene concentration in the suspension necessary to obtain 50% potassium leak out of yeast cells and human erythrocytes, respectively. bPs0: polyene concentration in the lipidic vesicle suspensions necessary to obtain 50% sodium and proton exchange.

the one hand, on biological membranes and, on the other hand, on purely lipid membrane models. This indicates, in agreement with many observations previously made, that the permea- bility pathway induced by polyenes are primarily related to their interactions with membrane lipids. Furthermore, the greater sensitivity of yeast cells is clearly related to the presence of ergosterol in this biological membrane.

2) The efficiencies of the different derivatives, as measured by the concentrations correspond- ing, for instance, to a 50% K ÷ leakage from yeast (Yr,50), red blood cells (Er, s0) and the same data obtained on ergosterol and cholesterol- containing vesicles, vary extensively depending upon the type of modification. However, it appears that the selectivity, as measured by the difference between these parameters for choles- terol and ergosterol membranes is not directly related to overall efficiency. Rather, the selecti- vity is mainly exhibited by only one group of derivatives whose carboxyl at C~8 position on the macrocycle is either absent or blocked (as in perimycin or AME).

This important observation was more recently fully generalized by a systematic study of a large series of derivatives carried out on lipid model membranes [7] and biological membranes [8]. From these studies, the respective influence of the modification in the polar head groups of the polyene antibiotics has been more precisely assessed.

First, substitution of the amino group of the amino sugar always results in a drop of efficiency which may reach two orders of magnitude in concentration. This effect seems to be related

primarily to the proton giving ability of this amino group which is the greatest for the unsubs- tituted amine. However, these substitutions do not confer any selectivity to the polyene. High or low, the efficiency remains approximately the same for cholesterol-containing and ergosterol- containing membranes.

Second, the absence or the substitution of the carboxyl group at C~8, either by esterification or amidation, always confers selectivity to the poiyene. These poiyenes, whose carboxyl group is blocked or absent, exhibit very low efficiency in inducing permeability in cholesterol-contain- ing membranes, whereas they remain very efficient in ergosterol-containing membranes. This difference in efficiencies between the two types of membranes may reach more than one order of magnitude in concentration.

In order to interpret these data and to try to understand the mechanism by which ionophoric activity and selectivity can be controlled by the molecular properties of the different polyenes, it is useful to consider, according to the currently proposed hypothesis, the successive steps of their mode of action on membranes. The p01yene, introd-'ced into the external aqueous medium of the ~.,ll, in which it is in general poorly soluble, interacts with the membrane. This primary step results in the penetration into the lipid matrix, probably in oligomeric form (at least dimeric). The arr~unt of polyene thus incorporated, for a given external concentra- tion, may vary to a large extent, depending upon both the nature of the membrane lipid and the physicochemical properties of the polyene considered. Therefore, the first hypothesis

40 C.M. Gary-Bobo

which could be ad,, anced is that the efficiency of this polyene mostly depends upon its actual amount present in a given membrane; the great- er this amount, the greater its efficiency.

The second step presumed (tightly correlated with the first) is that, when incorporated, the polyene forms complexes with the sterol present. This ability to form these complexes depends upon the structures of both partners. This process could also be considered the most important parameter. Finally, based on these polyene-sterol complexes, permeability path- ways are formed and the extent of the membrane permeability induced depends, in turn, not only upon their number but also upon their intrinsic permeability properties. The latter could be very important in governing both efficiency and selectivity.

Physicochemical properties in solution

The juxtaposition in the macrocycle ring of hydrophilic and hydrophobic moieties confers to polyene molecules particular physicochemical properties which are not very well known. In general, they are poorly soluble in both water and organic solvents, other than dimethyl sulfoxide or dimethyl formamide. This poor solubility is related to their strong self associa- tion.

As a matter of fact, the question arises if polyenes exist in monomeric form in aqueous solution even at concentration well below 10 -8 M [9]. It appears that in the range 10 -8-10 -6 M they are mostly in oligomeric forms (at least di- or trimeric). Their aggregation numbers increase with their concentration and beyond 10 -6 M (or 10 -5 M depending upon the deriva- tive considered) aqueous solutions become

cloudy and polyene aggregates tend to sediment. It must be kept in mind that, within the range of concentrations of biological interest, 10-8-10-5 M, a polyene solution represents, in fact, a mixture of different aggregation states of the molecule. Furthermore, polyene molecules have a strong tendency to surface adsorption.. The extent of this phenomenon depends upon the nature of the substitution on the polar head and the elec- tric net charge of the molecule, as well as the nature of the surface [10]. This makes it difficult to compare, on a quantitative basis, results obtained under different experimental condi- tions, since they depend upon the nature of the laboratory vessels, the surface/volume ratios and various other parameters.

Octanol-water partition coefficients of 4 radiolabeled AMB derivatives, representative of the 4 classes of substitution on the polar head groups have been measured (see Table II).

The partition coefficients were found to be independent of concentration between 10 -8 and 10 -6 M. They depend upon ionic strength and pH. The pH dependency corresponds to the titration curves of the carboxyl group, for N-Ac-AMB. In the case of zwitterionic N-fruc- tosyl amphotericin B, the pH dependency is very similar, beside the pH shift due to the protona- tion of the amino group. The inverse pH depen- dant-v af AMI7 ;e r~lat~rl t a th,~ nr,r~t,r~n-al-~an r~[ the amino group alone. Of course, the partition coefficient of the non-electrolyte N-Ac-AME is pH independent. However, it appears that the carboxyl group has a predominant influence on the solubility and, in fact, the partition ratio values fall into two groups regardless of the pH: the methyl esterification of the carboxyl group in AME and N-Ac-AME results in partition coef- ficients about 5 times higher than those of the two other derivatives, N-Fru- and N-Ac-AMB.

Table 11. Bound / free ratios computed as classical equilibrium partition coefficients between the membrane lipid phase and the aqueous solution, PRBC.

AMB derivatives PRBCa × 102 Poor b PRBC / Poor x 102

~Fm-AMB 2.4±0.2

~Ac-AMB 1.8±0.1

~Fru-AME 6.7±0.2

~Ac-AME 8.0±0.2

0.20±0.01 12.0

0.19±0.01 9.5

0.99±0.05 6.8

0.93±0.05 8.9

aPRBc: erythrocyte membrane lipid/buffer partition coefficient. bPocr: OCta~oI/buffer partition coefficient (pH 7.4).

Polyene- sterol interaction and selective toxicity 41

Polyene binding to red blood cell membranes

The binding of the 4 radiolabeled derivatives to human erythrocytes has been studied in relation to their efficiency in inducing membrane permeability [10]. The binding to erythrocyte suspensions containing 2x109 cel ls /ml was measured in an isotonic phosphate saline buffer containing sucrose to prevent swelling (140 mM NaCl, 5 mM sodium phosphate, 30 mM sucrose, pH 7.4). For the derivatives studied, the bound/ f ree ratios, measured between 10 -8 and 10 -4 M, were found to be constant, independent of the concentration. No saturability, and no cooperativity were detected in this range, so the binding appears to be a simple dissolution, without any detectable sites of fixation.

Binding measurements on resealed ghosts indicate clearly that hemoglobin does not parti- cipate significantly in the binding. Moreover, the binding to erythrocytes is fully reversible and the kinetics of adsoption and desorption are relati- vely fast (equilibrium is reached within minutes) and monoexponential.

This behavior of polyene binding is typical of a simple partitioning between phases, mem- brane matrix and aqueous medium. Therefore, the bound/ f ree ratios can be computed as classical equilibrium partition coefficients between the membrane lipid phase and the aqueous solution, PRBC. The values obtained are listed in Table II. They compare well with the octanol /water partition coefficients Pocr. The ratio PRBC/Pocr is roughly the same for the 4 derivatives. It may be considered that the 'affinity' of a given po!yene for the erythrocyte membrane depends primarily upon its lipid solu- bility. Such a conclusion will be discussed later. The efficiency in inducing permeability has been tested under the same experimental conditions. Binding and K + efflux from the cell were measur- ed on the same sample. The results are given in Fig. 2 in which K ÷ leak, expressed as percent of the total K ÷ content in erythrocytes, was plotted versus the total polyene concentration in the suspension. The two derivatives with a free carboxyl group, N-fructosyl- and N-acetyl- amphotericin B, whose PRBC are 3 - 5 times lower than those of the two methyl esterified derivatives, exhibit a much higher ionophoric activity. In Table III, the concentrations induc- ing a 10% K ÷ leak are listed; in the first column of this table the total polyene concentration, as in Fig. 2, are given. In the second column, the actual concentration present in the membrane,

"o 0)

_e

+

60

40

20

| j

I

7

,4"7

/ / ~,

/ . ] . ..o..o. .toc~oSYlo.~o~O . ,0 ' , ' - " . . . 10 "s 10 -s Ctot. (M)

Fig. 2. Potassium release by human erythrocytes after 90 min incubation periods in isotonic saline phosphate buffer with increasing concentrations of amphotericin B derivatives•

computed from PRBC, expressed as the number N of polyene molecules/cell ar, d in the third column the molar ratio R of membrane l ipid/bound molecule. The comparison of this set of values shows than the difference in activity in inducing permeability between esterified and non-esterified polyene derivatives is larger when estimated from the amount of polyene molecules actually present in the membrane than when estimated from the overall concentration in the suspension, as in Fig. 2. These values clearly demonstrate that the ionophoric activity in the cell membrane does not depend upon 'affinity', since the polyene whose affinity is the lowest exhibits the highest activity. The finding that the binding of ABM derivatives behaves as a simple equilibrium partition ratio apparently does not support the basic hypothesis according to which their ionophmic action depends upon their ability to form complexes with membrane sterol. However, the binding data simply imply that if specific binding sites to cholesterol exist, they are either already fully saturated below 10 -8 M or far from saturation even at 10 -4 M. The former possibility cannot be excluded. However, fixation to such sites of high affinity cannot be correlated to permeability because permeability is induced at polyene concentrations greatcr than 10 -6 M.

42 C.M. Gary-Bobo

Table 111. Comparison of polyene concentrations: total polyene concentration, actual concentration present in the membrane, and molar ratio R of membrane / lipid bound molecule.

AMB Ctotal a N b R c derivatives (gM)

N-Fru-AMB 2 1.1 x 105 4364

N-Ac-AMB 4 1.7 x 105 2824

N-Ac-AME 10 1.2 × 106 399

N-Fru-AME 100 1.1 x 107 44

aCtotal-- amphotericin B derivative concentrations corresponding to 10% K ÷ release by human erythrocyte (see Fig. 2). bN: number of molecules bound / cell. oR: phospholipid/bound AMB derivative molar ratios.

The second possibility that saturation is far from being reached at 10 -4 M is more probable, since at this concentration, the l ipids/bound polyene molecular ratios range between 40 and 200, depending upon the PRaC measured. These values are far from 1/1 or 1 / 2 stoichiometry assumed for AMB-sterol complex [1, 2]. Furthermore, it must be kept in mind that in the erythrocyte membrane, cholesterol molecules interact strongly with phospholipids and that AMB derivatives have to compete with them to form complexes.

A ~ ,~t concentrmions . . . . . . . very . . . . . . . lnuucmg a slgnlncant K+ leak, which represent less than 1 mole- c u l e / 4 x l 0 -3 lipid molecules in the case of N- Fru-AMB, the amount of highly permeable, channels formed by these cholesterol-polyene complexes is too small to be detected by the radioactive method of binding measurement, although it can be detected by the very sensitive spectroscopic method of circular dichroism [ 11].

To conclude this binding study, it seems that the first hypothesis made in the introduction, i.e., that polyene efficiency in inducing permea- bility might be related to its affinity for the membrane, should be discarded. The possibility that polyene efficiency and selectivity depend upon its ability to form highly permeable channels with sterol remains to be investigated. This has been done using the large unilamellar lipid vesicular system.

The role of the carboxyl group of polyenes in their interaction with sterols

The permeability induced by a series of AMB derivatives on LUV containing various sterols

has been studied, using the proton-cat ion exchange method [12]. In this method, a trans- membrane pH of about 2 units (5.5 inside and 7.5 outside) is established in the vesicle suspen- sion, driving protons outward, while a corre- sponding cation gradient drives them inward. Since anions (sulfate and phosphate) cannot cross the membrane, any cation movement can take place only by a one to one electro-neutral exchange with proton. Under this condition, the cationic permeability induced by an ionophore is adequately determined by measuring proton

I l i a [ movements , provlmng these movements are not rate limiting. This is assured by the addition of a specific protonophore, FCCP (carbonyl cyamid- p-trifluoro methoxyphenyl hydrazone). Proton movements themselves were measured by moni- toring the change in time of the position and intensity of the intravesicular phosphate NMR signal [13]. This alp NMR method has the deci- sive advantage of monitoring simultaneously both the intravesicular proton concentration and the pH clistribution. It enables subpopulations in the vesicle suspension to be distinguished on the basis of their internal pH.

These findings make it possible not only to compare the activities of different ionophores, but also to distinguish the ionophoric mechanism involved, i.e., mobile carrier or channel [12]. This is shown in Fig. 3 which compares the evo- lution with time of the NMR signal of the inter- nal phosphate in vesicles subjected to the action of either valinomycin, a typical mobile carrier ionophore, or gramicidin D, a typical channel forming ionophore.

Upon the addition of valinomycin, at low concentration, to a vesicular suspension (in the

Polyene- sterol interaction and selective toxicity 43

R = 3 , S x 1 0 - 6

I A

I 1 ! ! I t l ~ p p m ) L, 2 0 - 2

- 5 R = 6 0~.1 0

6(pp n'O 4 2 0 -2

VALINOMYCIN GRAMICIDIN

Fig. 3. Evolution in time of the internal phosphate NMR signal of large unilamellar lipidic vesicles under the influence of valinomycin and gramicidin D. Under the recording conditions used, neither the external phosphate nor the phospholipid phosphate signals are visible in the NMR spectra (for explanation see text). R: antibiotic / lipid molecular ratios in the vesicular suspension.

44 C.M. Gary-Bobo

presence of FCCP) the phosphate signal, initially centered at 5 - + 0.25 ppm (correspond- ing to pH 5.5), progressively shifts until it reaches 8 = + 2.2 ppm, which corresponds to the equilibrium pH of 7.5.

During this shift, the signal broadened significantly. This reflects a pH distribution among the vesicles, due to a distribution of ion flux rates. This distribution is related in part to the relative heterogeneity in size of the vesicles, but mostly to the statistical distribution of vali- nomycin molecules in the vesicule population. Under the experimental conditions chosen, the number of valinomycin molecules added to the suspension corresponds only to 0.5 to 1 /vesicle. In spite of the fact that, according to the statisti- cal distribution law, the majority of vesicles has no valinomycin in their membrane, all of them are involved in the permeability process. The progressive discharge of the proton gradient of the whole population indicates that valinomycin molecules are exchanged between vesicles relati- vely fast as compared to the ion flux rate they promote. Therefore, what is measured in the

case of mobile carrier ionophore is effectively the ion transporting efficiency which is the limit- ing factor.

The situation is completely different in the case of gramicidin D. After the addition of this channel-forming ionophore, the internal phosphate signal rapidly splits into two parts: one remains centered at 8 = 0.25 ppm, the initial pH value, while the second is centered at ~ = 2.2 ppm which is the equilibrium value, at pH 7.5. With time, an increase of this latter signal is observed, at the expense of the signal at 0.25 ppm, which finally disappears. No inter- mediate signal is observed.

The splitting of the signal into two parts reflects, as in the valinomycin case, the statistical distribution of gramicidin molecules among vesi- cles. Only a part of them has enough ionophore in the membrane for channels to be formed. But channels have a very high intrinsic transporting capacity and the vesicles forming channels attain equilibrium too quickly to allow intermediate steps to be observed. With time, and relatively slowly, gramicidin molecules are exchanged

ERGOSTEROL R-7 ;..10 -5

A M B

I 1 1 I I I I

i ( ppm) i, 2 0 -2

CHOL ESTEROL 0- L AMB R:2 ,,1

1

Omm

I J 1 t I 1 1

4 2 0 -2

CHOtJESTEROL 3 AHE R.I~,I O-

I

6 h

1 1 l 1 l S ppm) 4 2 0 -2

E11~GOSTEROL _ R,, 6,,I 0

I l L 1 I S(pO m) /, 2 0 -2

Fig. 4. Evolution in time of the internal phosphate NMR signal of cholesterol- and ergosterol-containing vesicles under the influence of amphotericin B (AMB) and amphotericin B methyl ester (AME). Conditions were the same as those for Fig. 3.

Polyene-sterol interaction and selective toxicity 45

between vesicles and progressively all of them go to equilibrium. It can be concluded that, in the case of channel-forming ionophores, the rate of proton efflux measured reflects the rate of channel formation and not the transp,~rting capacity of the channel once formed.

On the basis of these data, the behavior of AMB derivatives was studied in vesicles contain- ing either cholesterol or ergosterol. Fig. 4 shows the evolution of the internal phosphate NMR signal of the two types of vesicles upon addition of either AMB or AME. The behavior of this signal for AMB, in both cholesterol- and ergo- sterol-containing vesicles, is identical to that of gramicidin D and it can be concluded, in agree- ment with many previous observations, that AMB is a channel-forming ionophore. In the case of AME, the behavior of the signal differs in ergosterol and cholesterol vesicles. In ergoste- rol-containing vesicles, this behavior is typical of channel-forming ionophores, whereas in choles- terol-containing vesicles, the internal phosphate signal evolution is similar to what is observed with the mobile carrier valinomycin. This dual behavior has been observed with all the derivati- ves whose carboxyl group is blocked, either Oy esterification or amidation. On the other hand, all the derivatives whose carboxyl group is free and ionizable behave like channel-forming iono- phores, whether the sterol is cholesterol or ~ f.lm, tlo I-/~ ,..i~ I %,1 ~ Ik /~ Lli~,l I..,lll.

In order to understand better what differences in structure between the two sterols could play such an important role in the case of esterified derivatives, experiments were carried out with AMB and AME in vesicles containing: 1) no sterol or sterol whose hydroxyl group in the 3/3 position was absent or blocked; 2) sterol contain- ing one or two double bonds in the steroid nucleus; 3) sterols whose aliphatic tail contains double-bonds in different places and different substitutions. As a typical example Of this study, Fig. 5 shows a comparison of the evolution with time of the internal phosphate NMR signal induc- ed by AME in vesicles containing 7-dehydro- cholesterol and brassicasterol. In 7-dehydro- cholesterol-containing vesicles, the phosphate NMR signal evolution is identical to that observ- ed in cholesterol-containing vesicles, in spite of the presence of two double bonds in the nucleus as in ergosterol. On the contrary, with brassica- sterol, whose nucleus is identical to cholesterol but whose aliphatic tail is identical to ergosterol, the behavior of the internal phosphate signal is identical to that observed with ergosterol. The

general conclusion which can be drawn from this study may be summarized as follows: derivatives having their carboxyl free and ionizable, such as AMB, exhibit a channel-like behavior, regard- less of the sterol present, provided that it has a free hydroxyl group in the 3/3 position; in sterol-free membranes or in membranes contain- ing sterol without a free hydroxyl group~ these derivatives have no detectable ionophoric activity.

Derivatives whose carboxyl is blocked, such as AME, exhibit a channel-like activity only in

membranes containing either ergosterol or sterols having an alkyl side chain identical to that of ergosterol, even though the nucleus is diffe- rent. In the presence of other sterol or in sterol- free membranes, the ionophoric activity is much lower and always of the mobile cartier type.

On the basis of these results, a coherent and relatively precise hypothesis can be formulated concerning the role of the polyene-sterol inter- actions on the mode of action of these antibio- tics, accounting for the specific anti-fungal activ- ity of this class of derivatives characterized by the absence of free ionizable carboxyl group.

The stability of the polyene-sterol complex is based upon the strength of two types of interac- tions: hydrophilic, that is the hydrogen bonding between the polar head groups of the polyene and the 3/3-OH group of the sterol; hydropho- bic, that is Van der Waais interactions between the rigid polyene moiety of the macrolide ring and the body of the sterol. Depending upon the polyene polar head structure, the relative impor- tance of the two types of interactions in complex stability is different.

Fundamentally, the H-bond is established between the protonable amino group of the amino sugar moiety of the polyene, as the proton donor, and the oxygen of the 3/3-OH group of sterol, as the acceptor. However, the conditions under which tiffs H-bond is established strongly depends upon the polyene carboxyl group. If this group is free and ionized, first a strong elec- trostatic :teraction between this group and the amino group maintains the amino-sugar moiety in a fixed conformation in space; second comple- mentary binding can be established between the carboxylate and the sterol. Such a complete H-bond system is very strong and can ensure, by itself, the stability of the polyene-sterol complex, the hydrophobic interaction being of secondary importance. If the carboxyl group is blocked, the amino-sugar is free to rotate. H-bond between the amino group and sterol is

46 C.M. Gary-Bobo

,qo "

f H, C~,

7- DEHYDROCHOLESr EROL -5 t~t,~ R=Txl0

7-DEHYDROCHOLESTEROL AME R= 1 0 -3

,I il 0 m

t t i 1 1 I I ~(ppm)/, 2 0 -2 ~(ppm) 4 2 0 -2

BRASSICASTE ROL - 5

A NB R= %10

I I G(ppm) & 2 0 ~-2

BRASSICASTE ROL 1 AME R: I 0 -3

,3m° A

~ppm) 1- 2 0 -2

Fig. $. Evolution in time of the internal phosphate NMR signal of 7-dehydrocholesterol- and brassicasterol-containing vesicles raider the influence of amphotericin B (AMB) and amphotericin B methyl ester (AME). Conditions were the same as those for Fig. 3.

still established, but in a much less favorable condition. Furthermore, there is no complement- ary bonding with the carboxyl. Therefore, the H-bond system is relatively weak and the hydro- phobic interaction between polyene and sterol body becomes of primary importance to ensure the stability of the complex. Since the magnitude of dispersion forces is highly dependent upon the distance of closest approach and matching of hydrophobic surfaces, the specificity of sterol structure is revealed.

The ionophoric activity of polyenes being the result of their ability to form transmembrane channel structure by complexation with sterol,

the stability of this complex is the predominant factor conditioning both activity and selectivity for membranes containing specific sterol.

Conclusions

The cytotoxicity of polyene antibiotics can be ascribed to their ability to form channels with sterol in biological plasma membranes. Cell death is the result of the massive ionic permeabi- lity induced in this way and with which cells are not able to cope.

It appears that a given class of polyene macro- lide, which lacks the ionizable carboxyl group

Polyene-sterol interaction and selective toxicity 47

at C18 in their macrolide ring, can form such channels only with ergosterol. This restricted possibility makes them active and toxic for pathogenic microorganisms whose membranes contain ergosterol, such as fungi, and poorly active on animal cells whose membranes contain cholesterol. The very good correlation between biological data and results on model systems corroborates this conclusion. However, it must be kept in mind that polyenes may also interact with membrane proteins [10] and there is evidence that they can induce many effects [14, 15]. Are these effects correlated to their ionophoric properties and the large variety of consequences on membranes of an increased permeability? Are these effects the results of more specific interactions with various functional proteins? Answers to these questions can only be determin- ed by experimenting on specific cellular systems at sublethal concentrations, aiming t o observe modifications of membrane function more subtle than massive permeabilization.

Acknowledgment

This work was supported in part by INSERM, grant no. 041003.

References

1 Finkelstein A. & Holtz R. (1973) in: Membranes

(Eiseman G., ed.), Marcel Dekker Inc., NY, Vol. 2, pp. 377-401

2 De Kruijff B. & Demel R.A. (1974) Biochim. Biophys. Acta 339, 57-70

3 Van Hoogevest P. & De Kruijff B. (1978) Bio- chim. Biophys. Acta 511,397-407

4 Hammond S. (1977) Prog. Med. Chem. 14, 104-179

5 Schaffner C.P. (1984) In: Macrolide Antibiotic Chemistry, Biology and Practice (Omura S., ed.), Academic Press, New York, pp. 457-507

6 Cybulska B., Ziminski T., Borowski E. & Gary- Bobo C.M. (1983) Mol. Pharmacol. 24,270-276

7 Cybulska B., Herve M., Borowski E. & Gary- Bobo C.M. (1986) Mol. Pharmacol. 24,293-298

8 Cheron M., Cybulska B., Mazerski J., Grzy- bowska J., Czerwinski A. & Borowski E. (1988) Biochem. Pharmacol. 37,827-836

9 Mazerski J., Bolard J. & Borowski E. (1982) Bio- chim. Biophys. Acta 719, 11 - 17

10 Szponarski W., Wietzerbin J. & Gary-Bobo C.M. (1988) Biochim. Biophys. Acta 938, 97-106

11 Szpornarski W. & Bolard J. (1987) Biochim. Bio- phys. Acta 897, 229-237

12 Herve M., Cybulska B. & Gary-Bobo C.M. (1985) Eur. Biophys. J. 12,121-128

13 Tran Dinh S., Prigent Y., Lacapere J.J. & Gary- Bobo C.M. (1981) Biochem. Biophys. Res. Com- mun. 99,429-435

14 Omura S. (1984) in: Macrolide Antibiotics, Che- mistry, Biology and Practice (Omura S., ed.), Academic Press, New York, pp. 510-546

15 Bolard J. (1986) Biochim. Biophys. Acta 864, 157 "~" - - , D U D