Vol. 14, No. 1-4 293

NUCLEAR MAGNETIC RESONANCE PARTITIONING STUDIES OFSOLUTE ACTION IN LIPID MEMBRANESLan Ma, Theodore F. Taraschi, and Nathan Janes

Department of Pathology and Cell Biology,Thomas Jefferson University, 1020 Locust St., Philadelphia, PA 19107

INTRODUCTION: Lipid theories ofanesthesia implicate perturbation of mem-brane lipids as the locus for acute anestheticaction. [1] Chronic exposure to alcohols andanesthetics induces an adaptive response inmembrane phospholipids that confers resis-tance to many of the acute actions of alcoholsand anesthetics. [2]

We have proposed a colligative thermody-namic reformulation of the Meyer-Overtonhypothesis for anesthetic action. [3,4] Thisreformulation implicates configurationalentropy (Scf), the entropy imparted by a soluteupon a membrane structure in the partitioningprocess, as the driving force of solute actionon cooperative membrane equilibria. Solutepotency is determined by the competing con-tributions of configurational and thermalentropy (ASt). Equilibria most susceptible tosolute action (where dilute concentrations ofsolutes induce a perturbation equivalent to alarge change in temperature) involve largechanges in configurational entropy and smallchanges in thermal entropy according to thefollowing relation. [3]

AT/T m = AScf/ASt (1)

AT is the perturbation of the midpoint tem-perature, Tm, from its value in the absence ofsolute. The thermal entropy of an equilibriumis deduced from calorimetry and is approxi-mately constant for solute levels of biologicalrelevance. The remaining unknowns are theconfigurational entropy, which is determinedfrom the partitioning of the solute, and theperturbation of the equilibrium midpoint.

The colligative thermodynamic frameworkimplicates solute partitioning as the energeticforce that drives perturbations of cooperativemembrane equilibria by altering the relativefree energies of membrane states. Tests ofthe framework require simultaneous measuresof solute partitiomng and membrane structureover a range of solute concentrations andtemperatures.

Spin label* partitioning protocols have oftenbeen used in ESR studies of membrane struc-ture. [5] Such studies are designed so that thespin label partitioning probe is anonperturbing reporter of membrane struc-ture. To study solute action on membranes,however, a partitioning probe should serve themultifarious role of membrane perturbant,reporter of perturbations, and reporter ofsolute partitioning. Since NMR methods arenot limited to dilute solute levels, such flexi-bility is offered. Furthermore, complemen-tary structural information is available fromsimultaneous wideline XH [6], 2H [7], or 31P[8] studies.

PARTITIONING APPROACH TO SOLUTEACTION: In this abstract, we describe a HNMR partitioning approach based on theuncharged local anesthetic alcohol, benzylalcohol. Benzyl alcohol is a clinically usedtopical bacteriostatic agent. A variety ofcommercial pharmaceutical agents preparedfor injection contain benzyl alcohol for itspreservative properties and for pain relief.

The partitioning approach is based on (/) thesensitivity of the ring proton chemical shift tothe polarity of its environment and («) thesensitivity of the ring proton linewidth tomembrane binding. The chemical shift of thering resonances in a hydrophobic environmentare shielded and resolved from the ring reso-nances of the aqueous alcohol. The sensitivityof the ring proton resonance to its environ-ment provides a means of discriminating theaqueous alcohol resonance from the parti-tioned alcohol resonance. The dependence ofthe three chemically distinct ring protonchemical shifts on their environment is shownin Table 1 for benzyl alcohol (5 mole fraction%) in a variety of bulk solvents at 22°C. Theresonance exhibits a diamagnetic shift inhydrophobic solvents. A modest correlationbetween the chemical shift and Hildebrandt'ssolubility parameter ($*) for the solvent isevident.

294 Bulletin of Magnetic Resonance

TABLE 1

SOLVENT

WaterMethanol1-Propanol1-Butanol1-Octanol1-DecanolAcetoneMethylene ChlorideChloroform-djCarbon TetrachlorideHexane

5*

23.414.511.911.410.3

9.99.89.28.67.3

7.417.327.297.297.277.277.347.297.327.197.20

(ppm from TMS)

7.417.327.257.247.217.217.307.297.327.167.20

7.417.237.177.167.127.127.217.297.327.167.13

S (ppm) avg

7.417.297.247.237.207.207.287.297.327.177.18

I7.8

I7.6

I7.4 7.2

PPM7.0 6.8

Figure 1: The ring proton resonances of benzyl alcoholare shifted upfield and broadened upon binding tolecithin membranes.

Further discrimination stems from themotional restrictions imparted by the mem-brane environment that is reflected in thespin-spin relaxation. The ring resonancescorresponding to the free and bound drug areshown in Figure 1 for a lecithin model mem-brane in the L_ state. The resonance of thebound agent is oroadened due to immobiliza-tion in the membrane. The T2 of the boundagent is approximately 6 msec, while the T2 ofthe free agent is more than three orders ofmagnitude greater (11 sec). The different

relaxation properties allow for spectral editingbased on T2 using spin echoes.

The partitioning of benzyl alcohol intomembranes is modest. Consequently, thelipid to water ratios of the sample must belarge to obtain accurate simulations of thebroad bound resonance, while a sample sizeand geometry consistent with high Zeemanfield homogeneity must be maintained. Inpractice, to reduce sample demands, an inter-nal acetate standard was used to determinethe aqueous alcohol concentration in a dilutemembrane suspension. Since the lipid con-centration is known, the intramembrane con-centration is obtained by difference to yieldthe partition coefficient. To ensure that theintegrated aqueous resonance is not contami-nated by the broad bound resonance, aCPMG sequence is used to delay acquisitionby 25 msec in order to filter the broad bound-drug resonance. This filtering method alsoremoves the dipolar broadened lipid reso-nance to improve baseline definition.

Since the colligative thermodynamic frame-work equates the action of solute and temper-ature through entropy, precise temperatureregulation is required. The aqueous benzylalcohol resonance exhibits a temperaturedependent chemical shift. In order to main-tain a consistently reproducible temperature,we have taken advantage of this temperaturedependence. The chemical shift differencesbetween the HOD and free benzyl alcoholresonances as a function of temperature isshown in Figure 2.

Vol.

Q.0 .

UIo

ccUIu.u.au.Xin

14, No. 1-4

2.9

2.8

2.7

2.6

2.5

2.4

2 ^c

Figure 2:

i 1-

-

-

1 , 1

) 10 20

1

1

30

TEMPERATURE (C)

The chemical shift difference

_

-

-

_

1

40 5

between t

295

benzyl alcohol ring proton resonance and the HOD res-onance shows a temperature dependence.

ANALYSIS OF PARTITIONING: Thedegree of anesthetic partitioning into a mem-brane system is sensitive to and characteristicof the state of the lipid assembly. The equi-librium constant, K , is deduced from thepartitioning changes characteristic of theinterchange between membrane states. Thetemperature dependence of the partitioningexhibits the following functional form for astate change between two membrane struc-tures. [3]

KP =expC(T-Tm)

(2)

AHC = vH „

RTTm

(3)

The partition coefficient for the membranestates a and ft are Kp« and Kp", respectively.These partition coefficients are not necessar-ily constant and may be altered to include atemperature dependence. The total partitioncoefficient is Kp. The midpoint temperatureis Tm. A fit of the experimental data to thisfunction yields partition coefficients for eachphase, the midpoint temperature, and thevan't Hoff enthalpy (AHvH), a measure of thecooperativity of the equilibrium.

TEMPERATURE (K)Figure 3: The molal partition coefficient of benzyl alco-hol into multilamellar DPPC membranes is shown as afunction of temperature for two concentrations of benzylalcohol. The fit corresponds to the theoretical multipa-rameter least-squares analysis described in the text. Thederivative of the fit to the data is shown offset below.The percent mole fraction intramembrane benzyl alco-hol concentrations at the Lp,p p equilibriummidpoint are as follows: Panel A: Lp , = 0.23%, Pp ' =1.1%; Panel B: 2.2%, 16.8%; The mole fraction benzylalcohol concentrations at the P^/ -* La equilibriummidpoint are as follows: Panel A: Pp, = 0.84%, La=2.1%; Panel B: 11.7%, 24.2%; For comparative pur-poses, general anesthetic intramembrane concentrationsare considered less than 5 mole fraction percent.

The analytical framework presented is notspecific to the partitioning analysis. It isbroadly applicable to any technique in whichthe observable is characteristic of each state.

ALCOHOL ACTION IN MODELMEMBRANES: The lecithin membrane,DPPC (l,2-dipalmitoyl-.stt-glycero-3-phospho-choline), adopts three well-studied structuresor phases, a gel-structure (L«,), a ripple-structure (Pp,), and a fluid bilayer-structure(La). [9] Since the interchange between thesethree membrane structures is driven by

296 Bulletin of Magnetic Resonance

entropy, changes in solute, temperature andpressure alter the energetic balance to favor agiven structure. The Lp, •* P^/ equilibrium(pretransition) exhibits an equilibrium mid-point temperature determined by calorimetryas 34.8°C. [9] This change in state is accom-panied by a small change in thermal entropy(12.5 J mol"1 K-1). The P^, -* La (main transi-tion) exhibits an equilibrium midpoint tem-perature determined by calorimetry as 41.0°C.This change in state is accompanied by a rela-tively large change in thermal entropy (85.6 Jmol'1 K*i). [9] The large difference betweenthe thermal entropy changes associated withthese two equilibria provides a simple systemin which to test the predictions of the colliga-tive thermodynamic framework, that soluteaction occurs through entropy and that equi-libria characterized by a small thermalentropy change should be most susceptible toperturbation.

The temperature dependence of benzylalcohol partitioning at two substantially dif-ferent alcohol concentrations is shown inFigure 3. Figure 3A corresponds to benzylalcohol concentrations below that required forgeneral anesthesia; whereas, the concentra-tion in Figure 3B is near that required forlocal anesthesia. The partition coefficientsobtained by the NMR method are in excellentagreement with direct radiolabel measures.[10] Two discontinuities correspond to thetwo membrane equilibria. It is these changesin partitioning that provide the configura-tional entropy by which solutes perturb equi-libria. The low entropy Le, -+ Ppr equilib-rium exhibits greater sensitivity to the alcoholthan the high entropy P^, -*• La, as qualita-tively predicted by the thermodynamic model.

The quantitative test for the model isshown in Figure 4. The partitioning methodprovides intramembrane solute concentra-tions, which, in turn, provide the magnitude ofthe configurational entropy imparted to eachmembrane structure. This contributionlowers the free energy of each state accordingto the magnitude of the partitioning, andthereby alters the difference in free energyand shifts the equilibrium. The experimentalpoints are in good agreement with the theo-retical predictions (represented by the lines)at dilute alcohol concentrations for which thethermodynamic treatment is derived andwhich corresponds to pharmacological levels

1 10

INTRAMEMBRANE HOLE FRACTION U ) DIFFERENCE

Figure 4: The dependence of the equilibrium midpointtemperature of DPPC on the presence of benzyl alcohol.The benzyl alcohol intramembrane concentration differ-ence between the initial and final states at the equilib-rium midpoint is shown. Data are presented for the lowentropy Lp, -+ Fg, (pretransition; filled circles) andthe high entropy Fp, -*• La (main transition; open cir-cles) equilibria. The coUigative thermodynamic predic-tions (eq. 1) are represented by the lines. The solidportions of the lines designate the averageintramembrane concentrations at the midpoint whichcorrespond to the range of pharmacological relevance

for general anesthesia.Particularly striking is the dramatic contrast

in benzyl alcohol sensitivity exhibited by thesetwo equilibria. At average intramembraneconcentrations of 5 m.f.%, the low entropyequilibrmm is perturbed by approximately12°C, whereas the high entropy equilibrium isperturbed by approximately 1°C. Not onlydoes this observation support the predictionsof the thermodynamic model, but it demon-strates that remarkably low intramembraneconcentrations of nonspecific solutes can pre-cipitate quite substantial effects upon mem-brane structure.

ALCOHOL ACTION IN LIPOSOMESMADE FROM RATS CHRONICALLYEXPOSED TO ANESTHETICS: Rat livermicrosomes obtained from rats exposed tonitrous oxide or fed ethanol were isolated andliposcmes formed from the extracted phos-pholipids. Shown in Figure 5 are representa-tive benzyl alcohol partitioning traces for theliposomes prepared from the ethanol-fed andcontrol animals. The partition coefficient into

Vol. 14, No. 1-4 297

UJ

FFIC

]

UJot i

TION

(IR

TI

«*.Q.

40LA

L

40

35

30

25

• • ! 1

• •

- i — ' 1 ' r~

\mr •^ T

1 , 1 , 1

—i 1 1 r—

-

-

i . \

20 30 40 50 60 70

TEMPERATURE

Figure 5: Benzyl alcohol partitioning traces are shownfor liposomes made from rat-liver microsomal-phospho-lipids. The samples represented in the lower trace(filled triangles) were prepared from chronicallyethanol-fed rats. The samples represented in the uppertrace (filled circles) were prepared from their pair-fedlittermates.

the control samples is larger than the treatedsamples. This difference in partitioning ischaracteristic of 'membrane tolerance'. [2] Astructural equilibrium is apparent in the con-trol samples near 37°C that is lacking in thesamples obtained from treated animals. Sim-ilar results are obtained from the chronicnitrous oxide paradigm. These results evi-dence an adaptive response to the chronicpresence of anesthetic agents that results inaltered domain structure in the reconstitutedsystem. Similarly, structural lipid domains arepredicted at the anesthetic locus in our ther-modynamic reformulation of the Meyer-Overton hypothesis.

CONCLUSIONS: Alcohols and anestheticsact through the entropy imparted by parti-tioning to modify membrane architecture.Analysis of anesthetic action requires simul-taneous measures of solute partitioning andmembrane structure over a wide range of

solute concentrations. NMR partitioningmethods offer unique advantages in suchinquiry since the solute can serve the multi-faceted role of perturbant, reporter of mem-brane perturbations, and reporter of solutepartitioning.

METHODS: Spectra were obtained on a Bruker 8.5TAM spectrometer operating at 360 MHz with deuteriumlock. Lipids were dried under N2, evacuated (<5mTorr) for a minimum of 3 hours (natural lipids) orovernight (synthetic lipids) and hydrated in a cut-off 5mm tube with a tris-KCM^O buffer containing benzylalcohol and the acetate internal standard. The sampletube containing the multilamellar vesicles was placedcoaxially in a 10 mm tube and centered within the limitsof the decoupling coil for quantitative detection. Spectrawere obtained from a Bloch decay or from a spin echoCPMG sequence to delay acquisition 25 msec in orderto filter broad resonances. Typically the CPMGsequence was used for fluid phase lipids, and wheneverthe baseline was ill-defined due to the bound agent. TheBloch decay was typically used for gel-state lipids. 31pspectra were obtained periodically to ensure the absenceof small vesicular structures.

The chronic animal models and the membrane prepa-rations have been described elsewhere. [11]

The 'molal' partition coefficients reported here aremixed unit values ([moles alcohol / kg lipid] / [molesalcohol / liter deuterated buffer]) in order to referencethe heavy water buffer partition coefficient to literaturevalues for normal buffer. The molal partition coeffi-cients are converted to mole fraction partition coeffi-cients for all thermodynamic calculations.

ACKNOWLEDGEMENTS: Supported by PHSAA07215, AA07463, AA07186, AA00088.

REFERENCES:1. Seeman, P. Pharmacol. Rev. 24, 583-655

(1972).2. Taraschi, T.F. and Rubin, E. Lab. Invest.

52,120-131(1985).3 Janes, N., Hsu, J.W., Rubin, E. and

Taraschi, T.F. Biochemistry in press(1992).

4. Wang, D.-C, Taraschi, T.F., Rubin, E.,and Janes, N., unpublished.

5. McConnell, H.M., Wright, K.L. andMcFarland, B.G. Biochem. Biophys. Res.Comm. 47,273-281 (1972).

6. Janes, N., Rubin, E., and Taraschi, T.F.Biochemistry 29, 8385-8388 (1990).

Bulletin of Magnetic Resonance

7. Smith, R.L. and Oldfield, E. Science 225,280-288 (1984).

8. Taraschi, T.F., Lee, Y.-C, Janes, N., andRubin, E. Annals New York Acad. Sci.625,698-706(1991).

9. Chen, S.C. and Sturtevant, J.M.Biochemistry 20. 713-718 (1981V

10. Colley, CM. and Metcalfe, J.C. FEBSLetters 24,241-246 (1972).

11. Ellingson, J.S., Janes, R, Taraschi, T.F.,and Rubin, E. Biochim. Biophys. Acta1062,199-205 (1991).