Integral membrane proteins significantly decrease the molecular ...

Proc. Nati. Acad. Sci. USA Vol. 84, pp. 3704-3708, June 1987 Biophysics

Integral membrane proteins significantly decrease the molecular motion in lipid bilayers: A deuteron NMR relaxation study of membranes containing myelin proteolipid apoprotein (lipid-protein interaction/multipulse dynamic NMR/order parameters/rotational correlation times)

P. MEIER*, J.-H. SACHSEt, P. J. BROPHYt, D. MARSHt, AND G. KOTHE* *Institut fur Physikalische Chemie der Universitat Stuttgart, Pfaffenwaldring 55, D-7000 Stuttgart 80, Federal Republic of Germany; tMax-Planck-Institut fur biophysikalische Chemie, Am Fassberg, D-3400 Gottingen-Nikolausberg, Federal Republic of Germany; and *Department of Biological Science, Stirling

University, Stirling FK9 4LA, United Kingdom Communicated by Harden M. McConnell, February 9, 1987

ABSTRACT The influence of the myelin proteolipid apoprotein on lipid chain order and dynamics was studied by 2H NMR of membranes reconstituted with specifically deuterated dimyristoyl phosphatidylcholines. Quadrupolar echo and saturation recovery experiments were fitted by numerical solution of the stochastic Liouville equation, using a model that includes both inter- and intramolecular motions [Meier, P., Ohmes, E. & Kothe, G. (1986) J. Chem. Phys. 85, 3598-3614]. Combined simulations of both the relaxation times and the quadrupolar echo line shapes as a function of pulse spacing allowed unambiguous assignment of the various motional modes and a consistent interpretation of data from lipids labeled on the C-6, C-13, and C-14 positions of the sn-2 chain. In the fluid phase, the protein has little influence on either the chain order or the population of gauche rotational isomers but strongly retards the chain dynamics. For 1-myristoyl-2-[132H2]myristoyl-sn-glycero-3-phosphocholine at 35°C, the correlation time for chain fluctuation increases from 20 nsec to 650 nsec and for chain rotation from 10 nsec to 180 nsec, and the gauche isomer lifetime increases from 0.15 nsec to 1.75 nsec, on going from the lipid alone to a recombinant of protein/lipid ratio 0.073 mol/mol. The results are essentially consistent with spin-label ESR studies on the same system [Brophy, P. J., Horvath, L. I. & Marsh, D. (1984) Biochemistry 23, 860-865], when allowance is made for the different time scales of the two spectroscopies.

lipid-protein systems (see ref. 1 for review). The 2H NMR spectra of deuterated lipids, by contrast, have been found to consist of a single component in various lipid-protein systems, indicating that all lipid populations are in fast exchange, on the millisecond time scale. 2H NMR investigations have therefore concentrated on the ordering of the lipid chains. It is found that there is little or no change in chain ordering, although spectral line broadening does indicate a reduction in the rate of chain motion (e.g., see refs. 2 and 3). The complexity of the systems involved dictates that a detailed description of the effects of integral membrane proteins on lipid chain dynamics can best be achieved by a combination of multipulse NMR experiments with comprehensive theoretical simulations. Such an analysis is currently lacking. In the present work, we have investigated myelin proteolipid apoprotein reconstituted with specifically deuterated dimyristoyl phosphatidylcholine ([Myr2]PtdCho) as a model system for lipid-protein interactions in biological membranes. A motional model has been employed that includes both inter- and intramolecular motion (i.e., both long-axis motion and trans-gauche isomerization) and which is valid in both fast and slow motional regimes (4, 5). The simulation of quadrupole echo spectra as a function of pulse spacing, together with measurements of spin-lattice relaxation time (Tz), allows discrimination of the different motional modes, leading to an unambiguous description of the molecular dynamics. A consistent interpretation is obtained of data from reconstitutions with [Myr2]PtdCho labeled at the C-6, -13, and -14 atoms of the sn-2 chain. It is found that the protein has very little effect on either the degree of order of the lipid chain or the population of gauche rotational isomers but increases the rotational correlation times for chain fluctuation, chain rotation, and trans-gauche isomerism by a factor of 10 or more. These results are fully consistent with those obtained from ESR spectroscopy of spin-labeled lipids

Lipid-protein interactions are important determinants of biological membrane structure and function and, for this reason, have been the subject of intensive study by physicochemical methods. Magnetic resonance spectroscopy has made major contributions in this area because of its unique sensitivity to anisotropic molecular motion. In consequence of the different time scales of the two spectroscopies, nuclear magnetic resonance (NMR) and electron spin resonance (ESR) are sensitive to different aspects of the lipid dynamics, and this has given rise to rather divergent views of the lipid-protein interaction. The ESR spectra of spin-labeled lipids interacting with integral membrane proteins have been interpreted in terms of two components in slow exchange, on the nanosecond time scale. One component is characteristic of the lipid mobility in fluid, protein-free bilayers, and the other corresponds to lipids whose mobility is restricted by direct interaction with the intramembranous hydrophobic surface of the protein. The two-component nature of the spectra has allowed direct study of the stoichiometry and specificity of the interaction in a wide variety of different

(6, 7). THEORY Analysis of the various 2H NMR relaxation experiments was performed using a comprehensive motional model, described in refs. 4 and 5. The basis of the model is the density matrix formalism. The action of the different radiofrequency (if) pulses on the density matrix p(t) is represented by unitary transformations, involving Wigner rotation matrices. BeAbbreviations: PtdCho, phosphatidylcholine; [Myr2]PtdCho,

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.

dimyristoyl phosphatidylcholine; [Myr2]PtdCho-n-d2, 1-myristoyl-2[n-2H2]myristoyl-sn-glycero-3-phosphocholine; [Myr2]PtdCho-14d3, 1-myristoyl-2-[14-2H3]myristoyl-sn-glycero-3-phosphocholine. 3704

Biophysics: Meier et al.

Proc. Natl. Acad. Sci. USA 84 (1987)

tween pulses the density matrix is assumed to obey the

stochastic Liouville equation (8, 25)

-P(flt) at

second evolution period and so on. Finally, after the nth pulse, the observable NMR signal is given by

L(t, T1, T2, . . Tn-1) = Tr[p(t, r1, r2, . * Tn-l)-I+],

=

-

-(i/h ) HX(fl).p(,t)

-

r'flp(Qft) - Peq(fl)]b

[1]

which is solved using a finite grid method (9, 26). HX(f) is the Hamiltonian superoperator of the spin system, which depends on the orientation and conformation of the molecule, specified by the Euler angles fQ. 1F is the time-independent Markov operator for the various motional processes, where the equilibrium distribution Peq(Q) is given by

FflPeq (4) = 0

[2]

and Peq(Q) is the equilibrium density matrix. The Markov operator includes both intermolecular and intramolecular motions. The intermolecular motion is the anisotropic rotational diffusion of the lipid molecule as a whole, within an orienting potential. The intramolecular chain motion consists of trans-gauche isomerization, which is represented by ajump process. The dynamics of the system are characterized by three correlation times: TRI, and TRR for rotation about the diffusion tensor axis and rotation of this axis, respectively, and Tj for trans-gauche isomerization. The equilibrium distribution Peq(Q) is described in terms of internal and external coordinates. The internal part accounts for different conformations, and the external part for different orientations. Generally, there are four different conformational states corresponding to the different allowed orientations of a particular chain segment relative to the chain axis. The Euler angles OK,K (K = 1, 2, 3, 4) characterizing these orientations are listed elsewhere (4, 5). The conformational populations may be used to set up a segmental order matrix (4), which on diagonalization yields the segmental order parameters Sz z and Sxx - Sy r . They express the ordering of the most-ordered segmental axis Z' and the anisotropy of the order, respectively. The orientational distribution of the phospholipid molecules is described in the mean-field approximation, using an orienting potential from the molecular theories of liquid crystals (10). For axially symmetric ordering the normalized distribution function

f(f) = Njexp(A cos23)

[3]

depends on a single parameter A, characterizing the orientation of the molecules with respect to the local director. The orientational order parameter, Szz, is related to the coefficient A by a mean value integral: r

Szz = - N1 j [3 (cos2p8) - 1]exp(A cos2/3)sin P d, [4] 2 0 The molecular order of the phospholipid molecules is thus specified by the orientational order parameter Szz and the segmental order parameters Sz'z and Sxyx - Syry. The time evolution of the density matrix for a general pulse sequence is set up using the appropriate Wigner rotation matrices and the equation of motion. Before an rf pulse is applied, the spin system is at thermal equilibrium, Peq, with a Boltzmann population of the spin states. Application of the first pulse creates a defined nonequilibrium state p(O). After the pulse, the density matrix evolves according to the stochastic Liouville equation (Eq. 1). A second pulse is then applied, preparing a new initial condition, followed by a

3705

[5]

where I+ is the nuclear spin-raising operator and ri, r2 . . . Tnri are the various pulse-separation times. Fourier transformation of L(t, rT, T2, . . Tn-1) starting from time t = r1 + 72 . . * + Tn-i + Ti corresponding to a spin echo, yields single quantum spectra, which depend sensitively on the actual pulse sequence. From the decay of the echo amplitude as a function of i, various relaxation times can be evaluated. Two different pulse sequences are mainly employed in this study: the quadrupole echo sequence, (X/2)x - T - (f/2)y, and the saturation recovery sequence, (X/!2) - T1 - (X/2)x T2- (iT/2)y (4, 5). Since the echo spin-spin relaxation time, T2E, is dominated by motions with correlation times of the order of the inverse of the 2H quadrupole coupling constant, TR- (e2qQ/h )-1, the quadrupole echo sequence is sensitive to motions in the range 10-8 sec < TR < 10-4 sec. Saturation recovery is determined by the spin-lattice relaxation times, Tlz, which are particularly sensitive to motions with correlation times at the inverse of the Larmor frequency, TR 1. For a field strength of B = 7.0 T, this corresponds to overall sensitivity in the range 10-1" sec < TR < 10-7 sec. Thus, by combining quadrupole echo and saturation recovery studies, it is possible to follow dynamic processes over 7 orders of magnitude of correlation times.

MATERIALS AND METHODS Materials. Myelin was isolated from bovine spinal cord (11). The proteolipid was extracted and delipidated by chromatography on Sephadex LH-20 in CHC13/CH30H/0.1 M HC (50:50:1, vol/vol) as described (12). Chromatography was repeated in order to ensure complete delipidation, as judged by thin-layer chromatography and phosphate analysis. [6,6-2H2]- and [13,13-2lH2myristic acid were synthesized essentially as described (13). Acetic acid hexyl ester (or nonanoic acid ethyl ester) was reductively deuterated with LiAl2H4, and the deuterated alcohol was then converted to the bromide, as starting material for a Grignard reaction. The appropriate w-bromo fatty acid was converted to its chloromagnesium salt by using CH3MgCl. This was then coupled to the deuterated Grignard reagent, using a LiCuCl4 catalyst in tetrahydrofuran at -130C. [13,13-2H2]myristic acid was used to acylate sn-1-myristoyl lysophosphatidylcholine according to ref. 14, to yield [Myr2]PtdCho deuterated on the C-13 atom of the sn-2 chain ([Myr2]PtdCho-13-d2). The lysophosphatidylcholine was prepared from [Myr2]PtdCho by phospholipase A2 digestion, under conditions for the activity assay given by the suppliers (Boehringer Mannheim). [Myr2]PtdCho deuterated on the C-6 and C-14 atoms of the sn-2 chain ([Myr2]PtdCho-6-d2 and [Myr2]PtdCho-14-d3) were synthesized as described in ref. 4. Sample Preparation. The proteolipid protein was reconstituted with the specifically deuterated [Myr2]PtdCho by dialysis from 2-chloroethanol as described (12). The buffer used throughout was 0.1 M NaCl/1 mM EDTA/2 mM Hepes, pH 7.5. Samples in 2-chloroethanol (volume, 100-300 ml) were dialyzed against five changes of 5 or 10 liters (minimum 30-fold excess) of buffer at 4°C. Dialyzed samples were analyzed by sucrose density gradient centrifugation [10-55% (wt/vol) sucrose in buffer; Beckman SW40 rotor, 40,000 rpm, 3 hr] and recovered essentially as a single band. Lipid phosphorus (15) and protein content (16) of the reconstituted complexes were determined as described. For NMR measurements, the reconstituted complexes were pelleted (61,700 x g, 45 min), resuspended and washed

3706


four to five times in buffer prepared with deuterium-depleted water, and finally packed into 1-ml (8-mm-diameter) polyethylene sample tubes. NMR Methods. 2H NMR experiments were performed at 46.1 MHz with a Bruker CXP 300 spectrometer, using quadrupole echo and saturation recovery sequences. The ff/2 pulse width was 4 tusec, with a home-built probe (10-mm coil). All experiments were recorded using quadrature detection with a digitizing rate of up to 2 MHz and appropriate phase-cycling schemes (4, 5). The number of scans varied between 3600 and 36,000 (3-Hz repetition rate), with samples containing about 75 mg of deuterated lipid. Simulations of multipulse dynamic NMR experiments for I = 1 spin systems undergoing inter- and intramolecular motion in an anisotropic medium were performed as described (4, 5). The spin Hamiltonian for the Zeeman and quadrupole interactions includes nonsecular contributions. Within the Redfield limit (17), analytical expressions were used in the analysis. The constant parameters in the calculations were obtained from fast-rotational and rigid-limit quadrupole echo spectra of the pure phospholipids. As found previously (4), the quadrupole coupling constant of the aliphatic deuterons is e2qQ/h = 169 kHz. Residual quadrupolar echo relaxation times of TOE = 1 msec ([Myr2]PtdCho6-d2) and TOE = 2.5 msec ([Myr2]PtdCho-13-d2, [Myr2]PtdCho-14-d3) were used to account for the effect of dipolar interactions, which are omitted from the spin Hamiltonian. The adjustable parameters Szz, Szz,, (Sx'x' - Srr), TRii, TRI, and Ti (see Theory) were determined by simulation of the various experiments. In general, they need not be varied independently. In the fast-motional region, Szz, Sz'z,, and (Sxx, - Syry) can be obtained from the splittings of the quadrupole echo spectra (4). The dynamic parameters TRII, TR±, and Tj are then reliably evaluated by analyzing the different pulse experiments.

RESULTS Representative quadrupolar echo 2H NMR spectra of [Myr2]PtdCho-13-d2 in recombinants of different protein/lipid ratios, in the fluid phase, are given in Fig. 1. Clearly, both the spectral line shapes and the angular dependence of the quadrupolar echo relaxation time, T2E (as indicated by the change in line shape as a function of the pulse spacing, rl, in the quadrupolar echo sequence), show a very marked dependence on the protein/lipid ratio. Quadrupolar echo spectra of recombinants with the same protein/lipid ratio, but with different positions of the deuterium label on the sn-2 chain of the [Myr2]PtdCho, in the fluid phase, are given in Fig. 2. These spectra show clear sensitivity to the differential segmental motion at the various chain positions, in the presehce of the protein. In all cases the spectra consist of a single component, with no evidence for a second, broader, underlying component. The dependence of the quadrupolar echo relaxation time, T2E (defined by the pulse spacing, ri, at which the total integrated intensity of the quadrupolar echo spectrum is reduced to l/e of its initial value), and the spin-lattice relaxation time, Tlz, on the protein/lipid ratio for recombinants with [Myr2]PtdCho-13-d2 is given in Fig. 3. The relaxation curves were single-exponential decays, corresponding to just one spectral component, in all cases. Both the quadrupolar echo and spin-lattice relaxation times decrease strongly with increasing protein content in the recombinant, indicating a very pronounced effect of the protein on the lipid chain motion. Combination of the dependence of the echo line shapes on pulse spacing in Fig. 1, together with the measured relaxation times in Fig. 3, allows separation of the contributions from the different intra- and intermolecular chain motions, as

Proc. Natl. Acad. Sci. USA 84 (1987) 10 kHz

b

a

c

I'

c

b

J

c

FIG. 1. Experimental (-) and simulated (---) 2H NMR spectra of myelin proteolipid apoprotein/[Myr2]PtdCho-13-d2 recombinants at different protein/lipid ratios, at 350C. All spectra refer to quadrupole echo sequences. (Top) Pure lipid (protein/lipid ratio = 0) with Tr = 60 Asec (spectra a), 270 usec (spectra b), or 600 tusec (spectra c). (Middle) Protein/lipid ratio = 0.033 mol/mol, with T1 = 60 usec (spectra a), 120 ,usec (spectra b), or 270 psec (spectra c). (Bottom) Protein/lipid ratio = 0.07 mol/mol with T = 60 pisec (spectra a), 90 Asec (spectra b), or 120 Asec (spectra c). The simulations were obtained with the parameters of Fig. 4. Spectra were normalized to reflect the actual intensity in a quadrupole echo sequence.

indicated in ref. 4. The simulated spectra (dashed lines), obtained by using a single set of motional parameters for each protein/lipid ratio, are compared with the experimental spectra for the [Myr2]PtdCho-13-d2 lipid in Fig. 1. There is a generally good agreement between experiment and simulation, in terms of both line shapes and intensities, throughout the quadrupolar echo sequences. The calculated dependence of the relaxation times on protein/lipid ratio is given by the solid lines in Fig. 3. The spin-lattice relaxation times were fitted independently, using Redfield theory, and serve to determine the values for the fast correlation time, Tj. Only when Tj - 1 nsec, does this parameter influence the quadrupolar echo line shapes, and in this region the two methods give values that are in good agreement. The quadrupolar echo relaxation times are determined completely by the parameters from the line shape and intensity simulations in Fig. 1. The good agreement between experimental and calculated relaxation times in Fig. 3 therefore further demonstrates the consistency of the description of the line shapes and relaxation behavior. Simulation of the quadrupolar echo spectra for the different label positions is given in Fig. 2 (dashed lines). In this case, exactly the same simulation parameters were used for the three different label positions, with the exception of the segmental order parameters for the [Myr2]PtdCho-6-d2 lipid. (The values of the latter are very close to those for the lipid alone at 350C: Sz'z = 0.75, Sxx - Sryy = 0.12). This again confirms the consistency of the motional model, since all other motions should be independent of label position, and the motions at the 14 position are conformationally predictable from those of the 13 position. The protein/lipid ratio dependence of the motional correlation times and segmental order parameters, deduced from the integrated description of the spectral parameters for the [Myr2]PtdCho-13-d2 lipid, are given in Fig. 4. It is clear that although the protein has virtually no effect on the chain order, the rates of all the different modes of chain motion are



3707

.In 20 kHz I

K

b

C

0

AlI a

10 kHz

i

b

c

I

-

L-

t

To

,4 2 kHz

c

b

a

FIG. 2. Experimental (-) and simulated (---) 2H NMR spectra of [Myr2]PtdCho-6-d2, [Myr2]PtdCho-13-d2, and [Myr2]PtdCho-14-d3 in recombinants of similar protein/lipid ratios, at 35TC. All spectra refer to quadrupole echo sequences. (Top) [Myr2]PtdCho-6-d2 (protein/ lipid ratio = 0.039 mol/mol) with T1 = 30 psec (spectra a), 60 ,Asec (spectra b), or 90 pisec (spectra c). (Middle) [Myr2]PtdCho-13-d2 (protein/lipid ratio = 0.04 mol/mol) with Tr = 60 ,usec (spectra a), 90 Asec (spectra b), or 120 jusec (spectra c). (Bottom) [Myrj]PtdCho14-d3 (proteih/lipid ratio = 0.039 mol/mol) with T1 = 30 psec (spectra a), 90 Asec (spectra b), or 180 Asec (spectra c). The simulations were obtained with the parameters of Fig. 4, except for Szz = 0.73 and Sxx - Sr = 0.11 in the case of [Myr2]PtdCho-6-d2. Spectra were normalized to reflect the actual intensity in a quadrupole echo sequence.

strongly influenced by the lipid-protein interaction. The results from the [Myr2]PtdCho-6-d2 and tMyr2]PtdCho-14-d3 lipids are also consistent with this conclusion (Fig. 2), Since little change is found in the various order parameters, but the

0

E S.

0.

Protein/lipid ratio (mol/mol) x 10-2 FIG. 4. (Upper) Rotational correlational times characterizing the inter- and intramolecular dynamics of [Myr2]PtdCho-13-d2 in myelin proteolipid recombinants, as a function of protein/lipid ratio. Circles refer to chain fluctuation (TR,), triangles denote chain rotation (TRII), and squares refer to trans-gauche isomerization (Tj). Temperature, 350C. (Lower) Order parameters characterizing orientational and conformational order of [Myr2]PtdCho-13-d2 in myelin proteolipid recombinants, as a function of protein/lipid ratio. Circles denote the orientational order parameter Szz, squares refer to the segmental order parameter Szz, and diamonds denote the anisotropy Sx, Srr of the segmental order. Temperature, 350C.

At

E

12.5 2

4

Protein/lipid ratio (mol/mol)

6 x

10-2

FIG. 3. 2H quadrupolar echo relaxation times T2E (O, left ordinate) and spin-lattice relaxation times Tjz (i, right ordinate) of [Myr2]PtdCho-13-d2 as a function of protein/lipid ratio in the recombinant. Symbols denote experimental values taken from the integrated absorption powder pattern. Solid lines represent calculations of the relaxation times, using the parameters of Fig. 4. Temperature, 350C.

rotational correlation times are much increased. [For the pure lipid, very little difference is found in the motional parameters for the different labels in the fluid phase, with the exception of the segmental order parameters, Szz,, Sr y, and Sxyxy (see ref. 4), Therefore, the parameters Szz, TRfl, TRL, and Tj at a protein/lipid ratio = 0 in Fig. 4 can be taken as representative for all three positional isomers.]

DISCUSSION In the fluid phase, the 2H NMR spectra of the deuteriumlabeled lipids all consist of a single component and display single-component relaxation times, indicating either a homogeneous lipid environment or rapid exchange via lateral diffusion (at a rate > 106 sec-1) between environments of differing lipid mobility. ESR experiments on the same samples, using a [Myr2]PtdCho probe spin-labeled on the 13

3708


position of the sn-2 chain (data not shown), have all yielded two-component spectra, with the proportion of the more Notionally restricted component increasing with increasing protein content of the recombinant. These latter results are in full agreement with previous ESR studies on the proteolipid apoprotein/[Myr2]PtdCho system, obtained using PtdCho spin-labeled on the 14 position of the sn-2 chain (6). The ESR results therefore indicate an inhomogeneous lipid population, but with exchange between the component populations that must be fast enough to give averaging of the NMR spectra. Indeed, recent ESR estimations of the exchange rate of the lipids on and off the surface of the myelin proteolipid apoprotein have yielded values 107 sec' for PtdCho, which would ensure fast exchange on the NMR time scale (7). The NMR results have been successfully interpreted in terms of a single-component motional model in which the lipid chain dynamics are progressively perturbed by increasing protein content. An unambiguous discrimination of the various motional modes has been achieved by a comprehensive analysis of the effects on the spin dynamics in the different pulse sequences, as described previously for pure lipid systems (4). Strikingly, a consistent description can be given for the line shapes of all three different positional isomers, even though those for [Myr2]PtdCho-13-d2 and [Myr2]PtdCho-14-d3 in no way approximate to a conventional Pake doublet, as has been assumed in most previous (motional-narrowing) analyses (2, 3). Both the segmental and orientational lipid chain order parameters are essentially unaffected by the lipid-protein interaction (Fig. 4 Lower), indicating that the ensemble average chain orientation and gauche rotational isomer population at the protein surface are not appreciably different from those in fluid lipid bilayers. Similar qualitative conclusions have been reached from 2H NMR studies on myelin proteolipid apoprotein (18) and a variety of other lipidprotein systems (2, 3), although these previous analyses have not explicitly distinguished chain order from chain dynamics. ESR experiments have suggested that the spin-labeled lipid chains associated with the protein do not possess a high degree of orientational order (19, 20), in qualitative agreement with the present results. In contrast, fluorescent-probe measurements on the other systems have been interpreted to suggest that an increase in order of the lipid chains is induced by the protein (21, 22), contrary to the present direct measurements. The boundary conditions imposed on the chain order parameters in the fluid phase by the present measurements will clearly have important implications for the interpretation of theoretical models of lipid-protein interactions (23). The protein has a very strong effect on the lipid chain dynamics; all correlation times are increased by a factor of 10 or more at the high protein contents (Fig. 4 Upper). Clearly, both the chain-axis motion and the trans-gauche isomerism are slowed down at the protein interface, but with preservation of the differential rates between the inter- and intramolecular motions. The degree of restriction of chain motion may be greater than that for other integral proteins, at least for the faster motions, since the effects on Tjz are greater. Typically, integral proteins have been reported to decrease Tjz by 20-30% (2, 3), whereas at high concentrations the proteolipid protein decreases Tjz by up to a factor of 5 (Fig. 3). ESR measurements on the proteolipid apoprotein/ [Myr2]PtdCho system are in qualitative agreement with the NMR results, in that the effective rotational correlation time of the motionally restricted lipids is a factor of =10 longer than that of the fluid lipids (6), although in the ESR case a detailed analysis of the chain dynamics was not possible. The protein/lipid-ratio dependence of the NMR-derived correlation times plateaus at a protein/lipid ratio -0.073


mol/mol, which corresponds quite closely to the number of Notionally restricted lipids per protein found previously by ESR (6). This latter value also correlates well with the estimated number of lipids that can be accommodated around the protein hexamer that is found in detergent-solubilized proteolipid apoprotein (24). The detailed protein/lipid-ratio dependence of the correlation times in Fig. 4 is not expected to conform to a strict two-component model for fast exchange, since it will also contain contributions from perturbations of the mobility of lipids beyond the first shell surrounding the protein. Nevertheless, the motional parameters measured at the higher protein/lipid ratios are expected to approximate reasonably closely those of the lipids at the protein interface and provide a rather complete description of the way in which the lipid chain mobility is restricted by the protein. The rates of all chain motions are reduced by a factor of 10 or more. It remains for future work to determine to what extent these results for myelin proteolipid apoprotein may be generalized to other integral membrane proteins. We thank Frau S. Schreiner and Mrs. S. Chattejee for their expert technical assistance in the sample preparation. This work was supported by a grant from the Multiple Sclerosis Society of Great Britain and Northern Ireland to P.J.B. and from the Deutsche Forschungsgemeinschaft to G.K. 1. Marsh, D. (1985) in Progress in Protein-Lipid Interactions, eds. Watts, A. & de Pont, J. J. H. H. M. (Elsevier, Amsterdam), Vol. 1, pp. 143-172. 2. Bloom, M. & Smith, I. C. P. (1985) in Progress in Protein-Lipid Interactions, eds. Watts, A. & de Pont, J. J. H. H. M. (Elsevier, Amsterdam), Vol. 1, pp. 61-88. 3. Seelig, J., Seelig, A. & Tamm, L. (1982) in Lipid-Protein Interactions, eds. Jost, P. C. & Griffith, 0. H. (Wiley-Interscience, New York), Vol. 2, pp. 127-148. 4. Meier, P., Ohmes, E. & Kothe, G. (1986) J. Chem. Phys. 85, 3598-3614. 5. Muller, K., Meier, P. & Kothe, G. (1985) Prog. Nucl. Magn. Reson. Spectrosc. 17, 211-239. 6. Brophy, P. J., Horvath, L. I. & Marsh, D. (1984) Biochemistry 23, 860-865. 7. Horvath, L. I., Brophy, P. J. & Marsh, D. (1987) Biochemistry, in press. 8. Kubo, R. (1969) in Stochastic Processes in Chemical Physics, Advances in Chemical Physics, ed. Shuler, K. E. (Wiley, New York), pp. 101-127. 9. Kothe, G. (1977) Mol. Phys. 33, 147-158. 10. Cotter, M. A. (1977) J. Chem. Phys. 66, 1098-1106. 11. Benjamins, J. A., Miller, S. L. & Morell, P. (1976) J. Neurochem. 27, 565-570. 12. Brophy, P. J. (1977) FEBS Lett. 84, 92-95. 13. Das Gupta, S. K., Rice, D. M. & Griffin, R. G. (1982) J. Lipid Res. 23, 197-200. 14. Mason, J. T., Broccoli, A. V. & Huang, C.-H. (1981) Anal. Biochem. 113, 96-101. 15. Eibl, H. & Lands, W. E. M. (1969) Anal. Biochem. 30, 51-57. 16. Lowry, 0. H., Rosebrough, N. J., Farr, L. & Randall, R. J. (1951) J. Biol. Chem. 193, 265-275. 17. Redfield, A. G. (1965) Adv. Magn. Reson. 1, 1-32. 18. Rice, D. M., Meadows, M. D., Scheinman, A. O., Goni, F. M., Gomez-Fernandez, J. C., Moscarello, M. A., Chapman, D. & Oldfield, E. (1979) Biochemistry 18, 5893-5903. 19. Jost, P. C., Griffith, 0. H., Capaldi, R. A. & Vanderkooi, G. A. (1973) Biochim. Biophys. Acta 311, 141-152. 20. Pates, R. D. & Marsh, D. (1987) Biochemistry 26, 29-39. 21. JUhnig, F., Vogel, H. & Best, L. (1982) Biochemistry 21, 6970-6978. 22. Rehorek, M., Dencher, N. A. & Heyn, M. P. (1985) Biochemistry 24, 5980-5988. 23. Abney, J. R. & Owicki, J. C. (1985) in Progress in Protein-Lipid Interactions, eds. Watts, A. & de Pont, J. J. H. H. M. (Elsevier, Amsterdam), Vol. 1, pp. 1-60. 24. Smith, R., Cook, J. & Dickens, P. A. (1984) J. Neurochem. 42, 306-313. 25. Freed, J. H., Bruno, G. V. & Polnaszek, C. F. (1971) J. Phys. Chem. 75, 3385-3399. 26. Norris, J. R. & Weissman, S. I. (1969) J. Phys. Chem. 73, 3119-3124.