Dynamics of Amino Acid Side Chains in Membrane Proteins by High Field Solid State Deuterium Nuclear Magnetic Resonance Spectroscopy

We have obtained the first deuterium NMR spectra of individual types of aromatic amino acids in a defined membrane protein, bacteriorhodopsin, in the photosyn-thetic purple membrane of Halobacterium halobium R1. Isotopic labeling and high field (8.5 Tesla) operation permitted relatively rapid data acquisition at a variety of temperatures. At the temperature of growth (37 “C), we find that all 7 tryptophan residues are rigid on the time scale of the NMR experiment s), except for likely librational motions of =loo amplitude. By contrast, nearly all (9 2 2) of the 11 tyrosines and (13 & 2) 13 phenylalanines undergo rapid (>lob s-I) 2-fold ro- tational flips about CWr, causing formation of line shapes dominated by effectively axially asymmetric (asymmetry parameter q = 0.66) deuteron electric field gradient tensors. On cooling the phenylalanine- and tyrosine-labeled samples to “30 “C, all such motions are frozen out, ie. occur at rates <lo4 s-’, and axially symmetric (q = 0.05) line shapes are observed. At T > 91 “C, phenylalanine-, tyrosine-, and tryptophan-la- beled membrane spectra undergo dramatic narrowing to an isotropic line of -4-9 kHz width. This transition is a reflection of the loss of tertiary structure in the membrane protein with resultant fast unrestricted mo- tion of these aromatic side chains, and is only


Dynamics of Amino Acid Side Chains in Membrane Proteins by High
Field Solid State Deuterium Nuclear Magnetic Resonance Spectroscopy PHENYLALANINE, TYROSINE, AND TRYPTOPHAN* (Received for publication, March 16, 1981, and in revised form, April 30, 1981) Robert A. Kinsey We have obtained the first deuterium NMR spectra of individual types of aromatic amino acids in a defined membrane protein, bacteriorhodopsin, in the photosynthetic purple membrane of Halobacterium halobium R1. Isotopic labeling and high field (8. 5 Tesla) operation permitted relatively rapid data acquisition at a variety of temperatures. At the temperature of growth (37 "C), we find that all 7 tryptophan residues are rigid on the time scale of the NMR experiment s), except for likely librational motions of =loo amplitude. By contrast, nearly all (9 2 2) of the 11 tyrosines and (13 & 2) 13 phenylalanines undergo rapid (>lob s -I ) 2-fold rotational flips about CWr, causing formation of line shapes dominated by effectively axially asymmetric (asymmetry parameter q = 0.66) deuteron electric field gradient tensors. On cooling the phenylalanine-and tyrosine-labeled samples to "30 "C, all such motions are frozen out, ie. occur at rates <lo4 s-', and axially symmetric (q = 0.05) line shapes are observed. At T > 91 "C, phenylalanine-, tyrosine-, and tryptophan-labeled membrane spectra undergo dramatic narrowing to an isotropic line of -4-9 kHz width. This transition is a reflection of the loss of tertiary structure in the membrane protein with resultant fast unrestricted motion of these aromatic side chains, and is only partly reversible. These results, in conjunction with those obtained using [y-2H8]valine-labeled bacteriorhodopsin (Kinsey, R. A Over the past 10 years there has been considerable experimentation aimed at determining the nature of protein-lipid interactions in model and intact biological membranes. Some early studies (1,2) emphasized the idea that proteins acted in a manner similar to that of the rigid tetracyclic sterol choles-* This work was supported by the United States National Institutes of Health (Grant HL-19481). by equipment grants provided by the United States National Science Foundation (Grants PCM 78-23021 and PCM 79-23170) and the Alfred P. Sloan Foundation, and has benefitted by use of equipment made available by the University of Illinois National Science Foundation Regional Instrumentation Facility (Grant CHE 79-16100). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18  terol (3,4), causing an "ordering" of the lipid hydrocarbon chains, a view supported by several more recent theoretical studies (5)(6)(7)(8), and experimental work using fluorescence and ESR techniques (9)(10)(11)(12). More recently however, deuterium nuclear magnetic resonance spectroscopic techniques, in both model protein-lipid systems (13)(14)(15)(16) and in biological membranes themselves (17)(18)(19), have given no indication of lipid ordering by protein molecules, which has thus led to a revision by some workers of previous protein-lipid interaction models. A popular view now is that lipids may be "immobilized" o r perhaps trapped by proteins, but are not necessarily ordered (16,(20)(21)(22), but the crucial point about the protein-lipid interaction problem still remains unprobed, what does lipid "do" to protein? Since lipid phase behavior (23) and the presence of cholesterol (24) apparently have large effects on membrane enzyme activities, we have therefore begun a program aimed at elucidating the dynamic structures of proteins in membranes. As a "benchmark," we have already obtained NMR spectra of individual sites in crystals of some soluble proteins (25, 26) using a new magnetic ordering method which may eventually be applicable to membrane protein studies.
In this publication we report recent results obtained via 'H NMR spectroscopy at high field which indicate that the motions of individual types of amino acid residues in membrane proteins may now be studied in intact biological membranes. Our results show that protein dynamics in membranes may now be analyzed and compared with solution (27)(28)(29), micellar (30), and crystal data (31,32) to provide a basis for more meaningful studies of protein-lipid interactions in biological membranes.
As with studies of lipid dynamics in model or intact biological membranes (13,15,17) or in protein crystals (25,26), we have chosen to use the 'H nucleus as our structural probe. In this publication we concentrate on use of 'H NMR powder pattern line shapes to deduce information about the rates and types of amino acid side chain motion (33,34).
In this first detailed publication on amino acid dynamics in a "condensed phase" membrane protein, as opposed to the experimentally more accessible solubilized protein systems, we have chosen to investigate the "purple membrane" system of Halobacterium halobium R1. This system has the desirable NMR characteristics of only one protein, bacteriorhodopsin, in the purple membrane (35), its sequence is known (36-38), and its three-dimensional structure is becoming available (39,40). The system may also be enriched biosynthetically with a number of deuterated amino acids (40,41) without undue label "scrambling." Moreover, the system has been oriented using electric (42) or magnetic (43) fields, or by drying down onto glass or mica surfaces (40), and some preliminary results on formation of microcrystals have been obtained (44), opening up the possibility of obtaining oriented samples for NMR spectroscopy, which permits in favorable cases determination of residue orientations (25, 26). Finally, the enzyme has been shown to be susceptible to proteolytic cleavage and reassembly into an active proton-pumping system (37), which opens up the possibility of making specific NMR resonance assignments and studying the dynamics of individual residues.  (45). Finely divided platinum was prepared by the reductive procedure of Calf and Garnett (46), using NaBH4; 5 g of L-phenylalanine (Sigma) were added to 1   The mixture was then refluxed under Nz for 24 h, cooled, adjusted to pH 4.5 with 28% NH40H, and then kept at 4 "C overnight. The precipitate collected by filtration was washed with 3 volumes of cold water to remove salts, followed by 1 volume of cold 95% ethanol. After drying overnight a t 40 "C, 9 g of [lH']Tyr crystals were re- was then dissolved in this solution, and the flask was sealed and left to stand at room temperature, in the dark, for 3 days. The solvent was removed by rotary evaporation to give a light brown tar. The exchange was repeated three more times, and the percent deuteration was assayed after each exchange using proton NMR spectroscopy at 220 MHz. After the fourth exchange, the solution was adjusted to pH 6 with 28% NH40H, then kept overnight at 4 "C. The precipitate was filtered, redissolved in H20/ethanol (l:l, v/v), and decolorized using activated charcoal. The product was finally recrystallized from H20/ ethanol ( 3 2 , v/v), yielding 9 g of crystals. The product was 92 2 3% deuterated at positions SI, e,3, ( 2 , {,,, and q 2 as determined by 'H NMR spectroscopy at 220 MHz.

Syntheses
Production of Labeled Membranes H. halobium strain R1 was the kind gift of Professor T. Ebrey, and was grown in a salt medium basically according to Onishi et al. (48) with the addition of 2% malate (49), except that either ['HsIPhe or ['Hz]Tyr were substituted for the normal nonlabeled amino acids. The growth medium thus typically contained the following amino acids (in g/10 liters): L-alanine, 2.15; L-arginine hydrochloride, 2.0; Lcysteine, 0.5; L-glutamic acid, 13.0; glycine, 0.6; L-isoleucine, 2.2; Lleucine, 8.0; L-lysine hydrochloride, 8.5; L-methionine, 1.85; L-phenylalanine, 1.3; L-proline, 0.5; L-serine, 3.05; L-threonine, 2.5; L-tyrosine, 2.0; and L-valine, 5.0. Note that the growth medium does not normally contain tryptophan, thus, for the Trp-labeled membrane system we incorporated ['Hs]Trp at a level of 5.0 g/10 liters into the medium. Purple membranes were isolated according to Becher and Cassim (50) and were then exchanged with 'H-depleted water (Aldrich) to remove some background HO'H. Samples were generally exchanged twice, then finally concentrated by ultracentrifugation for 12 h at 100,OOO x g, prior to NMR spectroscopy.

Radiotracer Experiments
To determine the level of deuterated amino acid breakdown and reincorporation into lipid and other amino acids second dimension). Individual amino acids were detected using ninhydrin spray, after which the "spots" were removed from the TLC plates and counted on a Tracor Analytic Model 6892 scintillation counter (Elk Grove, IL), using Aquasol-2 (New England Nuclear).
Our results indicate that 55% of ["CIPhe and -0% [14C]Tyr counts "scrambled into other amino acids, strongly suggesting a similar low level of incorporation into lipid. Similarly, with [I4C]Trp, even though there is amino acid breakdown during hydrolysis, comparison of the relative number of counts in the region of the Trp hydrolysis product with those on the rest of the chromatogram indicated little, if any, incorporation of 'H label into any other amino acid. Our results with Phe incorporation are consistent with those of Engleman and Zaccai (40) who, using ['HIPhe, determined that less than 1% "H was incorporated into other amino acids and whose neutron diffraction data indicated only small 'H label incorporation into lipid upon ['HI Phe supplementation.

Spectroscopic Aspects
Nuclear magnetic resonance spectra were obtained on a "homebuilt" Fourier transform NMR spectrometer which consists of an 8.5-Tesla, 3.5-inch bore high resolution superconducting solenoid (Oxford Instruments, Osney Mead, Oxford, United Kingdom), together with a variety of digital and radiofrequency electronics. We used a Nicolet   55.273 MHz. Deuterium NMR spectra were recorded on this instrument using an 800-pl sample volume and a quadrupole echo (51,52) pulse sequence. The 90" pulse width varied between 2.0 and 3.5 ps; 90" pulse widths and phase quadrature between the two radiofrequency pulses was established by viewing quadrature free induction decay signals of S-[methyl-*H3]methionine. The same settings were used for data acquisition on *H-labeled membranes. In essentially all cases, no phase corrections were necessary after Fourier transformation. Typically, data were collected using sampling rates of 500 &point. The zero frequencies of both instruments were established using a 1% D20 reference, the zero frequency for the protein samples investigated being offset -2 ppm downfield from this position, Samples were run as solid high speed pellets (100,000 x g for 12 h). Sample temperature was regulated either by means of a liquid nitrogen boil off system or by using a heated air flow. The temperatures reported were measured using a calibrated Doric Trendicator (San Diego, CA) with a copper-constantan thermocouple and are the gas flow temperatures measured next to the sample. Separate experiments indicate that this temperature is accurate to *l-2 "C over the entire sample volume. All samples had normal optical absorption spectra after NMR spectroscopy, except after heating to >90 "C when "loss" of chromophore 560 nm absorbance occurred.
Spectral simulations were carried out on the University of Illinois Digital Computer Laboratory's Control Data Corporation Cyber-175 system, which is interfaced to a Tektronix 4006 graphics terminal and interactive digital plotter (Tektronix, Beaverton, OR) in our laboratory, as described previously (15).

RESULTS AND DISCUSSION
We shall discuss first the basic background theory to the study of the 'H NMR of amino acid dynamics in membrane proteins, concentrating on a residue that we intuitively expect to be one of the most rigid, or irrotationally bound tryptophan. We show in Fig. 2  correspond to +1 c-* 0 and 0 c, -1 and give rise to a "quadrupole splitting" of the absorption line with separation Avgl between peak maxima of where 8 and \k define the orientation of the principal axis of the electric field gradient tensor (usually the C-D bond vector) with respect to the laboratory coordinates, eq is the field gradient V,,, and eQ is the deuteron quadrupole moment.
Assuming axial symmetry (asymmetry parameter q = 0) to eliminate \k dependence, as is certainly appropriate for aliphatic C-2H bonds (56), then Equation 1 may be rewritten in a simpler form For rigid polycrystalline solids all values of 8 are possible and one obtains a so-called "powder pattern," Fig. 2 (55) show, however, that the best fit to the experimental spectrum (Fig. 2B) is obtained using e2qQ/h = 183 k 3 kHz and an asymmetry parameter q = 0.05 f 0.02 ( Fig. 2A and Table I). The observed quadrupole coupling constant for ['H5]Trp (Fig. 2B), is therefore considerably in excess of the -168 kHz found in aliphatic C-'H systems using NMR methods (56). This result is nevertheless consistent with the increased e2qQ/h values found in a variety of other aromatic compounds (34,56), the observed trends for the electric field gradient values for C-D bonds being sp > sp2 > sp3 (60). The average value for naphthalene and anthracene, perhaps the most reasonable published models for E2H5] Trp, is -184 kHz (61-63). Also, in addition to having substantially larger coupling constants, it is well known that C-D bonds in aromatic systems may have non-zero asymmetry parameters, q (64). In those aromatic systems where asymmetry parameters have been investigated, q values = 0.053 k 0.015 have been determined (56).
Our results therefore suggest that there is essentially no large amplitude motion of the Trp molecule on the time scale of the 'H NMR experiment s) at 25 "C, since the quadrupole splitting is that expected for a rigid, crystalline, aromatic species.
What motions then, if any, do the Trp residues of bacteri-    (33). We believe that the first model (53) is the more plausible. Tryptophan residues in the purple membrane are therefore properly throught of as being rigid at the temperature of growth (37 "C).
Upon heating the ['H5]Trp-labeled purple membranes above -90 "C, there is a dramatic change in the 'H N M R spectrum, as shown in Fig. 3. The large quadrupole splitting collapses and a relatively narrow "isotropic" line, of -4-kHz width, is obtained. Assuming isotropic rotational motion, a quadrupole coupling constant e2qQ/h = 183 kHz and TJ = 0.05 ( Fig. 2), this line width corresponds to a rotational correlation time TR of -100 ns (65) and is presumably due to protein denaturation. Upon cooling to 37 "C, there is a reappearance of some -140-kHz component, although this does not necessarily imply any prot.ein renaturation.
The results obtained with tryptophan were not unexpected and are supported by the observation of essentially irrotationally bound Trp residues in a variety of proteins in solution (27)(28)(29)(65)(66)(67). With this body of solution N M R data in mind, it is therefore worth asking if the types of motions observed for other aromatic residues, such as phenylalanine and tyrosine (27,29), carry over into the solid state. In particular, we would like to know if these residues may "flip," as has been shown in a number of 'H solution N M R studies (27,29,67,68).
What, therefore, are the manifestations of such large amplitude residue motions in the 'H N M R spectra of amino acids in membrane proteins in the solid state? We show in Fig. 4 a spectrum of ['H5]Phe in the crystalline solid state, together with simulated 'H N M R spectra, for a variety of possible amino acid motions. For the case of no motion we obtain a rigid lattice powder spectrum (Fig. 4A), which as expected is in good agreement with the experimental result obtained with ['H5]Phe (Fig. 4B). The quadrupole coupling constant For the case I!?' = go", a quadrupole splitting Av -17.0 kHz is predicted, having an intensity corresponding to four deuterons. In addition, a 20% intensity contribution from the {-"H is predicted, having the same breadth as the rigid lattice spectrum, since the C(-"H vector is along the axis of motional averaging, i.e. p = 0. The composite spectrum (corrected for the nonuniform labeling of our sample) is shown in Fig. 4C. It is quite dissimilar to that observed for [*H5]Phe in any native system investigated.
By contrast, a 2-fold "jump" model for phenylalanine motion, whereby the aromatic ring executes 180" reorientational flips about Cy-Cl, and which has been detected previously in solution NMR studies of proteins by means of chemical shift data, predicts a very different result (Fig. 4 0 ) .
Assuming that motion is "fast" compared to the breadth of the rigid bond coupling (>>2 X lo5 s"), a motionally averaged  (Fig. 40). This dominant feature is easily detected in some intact membrane spectra, as discussed below. Note that we have not yet considered the {-'H. This lies on the axis of motional averaging and its spectrum will therefore be essentially unaffected by tne motion (p = 0", Fig.   5 ) , except for a second order effect due to an initial non-zero q. The composite spectrum of Fig. 4 0 therefore contains two components: 75% intensity from 2H61.S2~f1~f2 having 17 = 0.66, A v g l = 30 kHz, and a total breadth of -181 kHz, and 25% from having 17 -0.05 and a total breadth of -270 kHz (Fig. 5 ) .
It is perhaps worth noting that in general for a 2H powder spectrum, the splittings of the singularity, step, and edge as a function of breadth (vg) and asymmetry parameter (7 or qeff) are as follows:  such that UQ and 7 are fully determined by measurement of any two of the frequency parameters, AuQ,. In practice, however, especially in systems as complex as cell membranes, where different residues of the same chemical type may undergo different motions at any given temperature, and where adventitious 'H label incorporation may have occurred due to label scrambling, either in synthesis or biosynthesis, such analyses of spectral line shapes will require considerably higher spectral signal-to-noise ratios than those obtained so far.
The above results have concentrated exclusively on the deuterium NMR spectra of flipping phenylalanine rings: such line shape changes as discussed above will also, of course, be manifest in I 3 C spectra, and similar calculations have been presented by Spiess (71) for the chemical shift tensor, u.
In Fig. 6 we present the fist 2H NMR results on ['HsIPhelabeled bacteriorhodopsin in the purple membranes of H. halobium, together with comparison spectra of ['HHb]Phe in the solid state (Fig. 6A). The deuteron spectra of all "native" membranes ( Fig. 6, B--F) are extremely broad and have characteristic line shapes (Fig. 4, A, B , and D). At low temperatures (-85 "C, -30 "C, Fig. 6, B and C ) , the spectra of ['HslPhe-labeled purple membranes are virtually superimposable on that of the solid amino acid at room temperature (Fig.  6A). The quadrupole splittings of the membrane spectra at low temperatures are, as with the ['H~slTrp-labeled membranes (Fig. 2), only very slightly temperature-dependent between -0 and -85 "C, but in contrast to the Trp sample, upon warming above "30 "C there is a continuous change in spectral line shape up to -55 "C, characterized by a loss of the -130-kHz component and a growth of a -30-kHz component (Fig. 6, E-G). As will by now be clear, this narrow spectral component (AuQ, -31 kHz, qeff = 0.66) most likely originates from ['Hs]Phe rings undergoing 2-fold flips. At the temperature of growth (Fig. 6F), all (13 f 2) of the 13 Phe residues appear to be undergoing rapid 2-fold jumps, and as expected at >50 "C, all Phe residues again appear to be undergoing such motions, although our spectral simulation at 55 "C is less satisfying than those obtained at lower tempera-tures, since adequate agreement between experiment and computer simulation is only achieved using q,ff = 0.60 (rather than qeff = 0.66), suggesting that additional motions to those discussed above have begun to occur. Interestingly, the spectra of Fig. 6, D-F, show little evidence of broad line shapes due to flipping at "intermediate exchange" frequencies. One likely explanation for this result is that the *H N M R spectra of such intermediate exchange residues will be characterized by rather rapid decays of the quadrupole echo intensity in the 9O0-~-9Oow~ experiment (72) and we have indeed observed rather rapid and anisotropic quadrupole echo decay rates at 3" (in addition to large differential spin-lattice relaxation behavior) in Fig. 6D. Accurate determinations of flipping to nonflipping ratios at intermediate temperatures will thus require considerably higher signal-to-noise ratio spectra than we have obtained so far, in addition to 7-dependence studies, the use of slow motional models (73,74) to accurately simulate the spectra of residues undergoing intermediate rate motions (-104-10' s"), computations of echo decay rates in such intermediate exchange situations, together with development of instrumentation having dead times shorter than those currently available. Nevertheless, at -3 "C our results are quite well simulated using a simple superposition of two individual states and suggest that -45% of the spectrum of  (Fig. 6F) to a relatively sharp "isotropic" line shape having a width of -9 kHz (Fig. 6G). There seem to be two main possible explanations for this result. First, the protein could simply be denatured, resulting in an isotropic line shape or series of overlapping Lorentzian lines, due to motion of the Phe ring with a correlation time -200 ns (65). The second possibility is that the Phe rings begin to undergo fast continuous rotational diffusion about CD-CY a t >90 "C, resulting in a line having AVQ -17 kHz (Fig. 4C), the splitting being in addition further reduced by some additional off-axis motions, together with a contribution from HO'H helping to obscure the quadrupole splitting. However, it seems unlikely that the HO'H component's total integrated intensity is sufficiently large to effect this obfuscation. The spectrum of Fig. 6G may be quite well simulated using a -10% flipping component (AvQ, -31 kHz) and a -90% isotropic component ( W -9 kHz). We thus favor the notion that the bacteriorhodopsin has become at least partially unfolded by heating above -90 "C. Upon cooling the sample of Fig. 6G to 37 "C, a broad splitting is again obtained, superimposed on an isotropic line. Such a spectrum is not observed in the initial heating runs, suggesting that there is a partial refolding of the protein after cooling. Although this spectrum (Fig. 6G) is rather difficult to simulate, we estimate that -40% of the Phe residues remain in the mobile, denatured state. Optical absorption spectra indicate loss of the 560 nm chromophore absorption and presumably a non-native protein structure is obtained after such high temperature excursions. It may be worth noting that the collapse of the native membrane type spectrum occurs over a rather narrow range of temperature, between 85 and 90 "C.
The results of Fig. 6 indicate that phenylalanine side chains in bacteriorhodopsin of the purple membrane of H. halobium undergo only a limited variety of motions a t their growth temperature. In particular, it seems that the vast majority undergo 2-fold flipping motions a t 37 "C about Cp-CY, and only between 0 and 2 are rigid. If therefore seems reasonable to ask whether such motions are also seen with the aromatic residue tyrosine, since 2-fold flips have previously been noted in 'H NMR spectra of several proteins in solution (29,68).
We show therefore in Fig. 7 'H NMR spectra of ['H'lTyr in the crystalline solid state (Fig. 7A), together with spectra of H81.8s.rl.c2,t , this could correspond t.o 9 (k2) of the 13 Phe rings "' A. Kintanar and E. Oldfield, unpublished results. the membrane, as a function of temperature. At room temperature the amino acid spectrum may best be simulated (Table I, Fig. 7A) using e'qQ/h = 181.3 k 3 kHz and TJ = 0.05 0.02. An essentially identical spectrum is obtained for the ['H2]Tyr-labeled purple membrane a t "90 "C (Fig. 7B), where spectral simulation reveals an observed splitting of -124 kHz and asymmetry parameter q = 0.05, corresponding to a coupling constant e2qQ/h = 180 kHz. As with the case of the [2H5]Phe-labeled membranes, this slight reduction in coupling in the membrane may be due to peptide bond formation, or to lattice effects, e.g. hydrogen-bonding differences between amino acid crystal and lipoprotein membrane, or more likely due to experimental error. As with the ['H5]Phe results (Fig.  6 ) , increasing temperature again results in a decrease of the broad spectral component and an increase in the narrower component having A V Q~ -27-29 kHz (Fig. 7). At the temperature of growth (37 "C), the spectrum of the ['H2]Tyr-labeled purple membrane (Fig. 7) is qualitatively quite similar to that of the ['Hs]Phe membrane at the same temperature (Fig. 6D). We find that good agreement with the experimental result of Upon heating the sample of Fig. 7 to -85 "C, there is relatively little change in the 'H NMR spectrum, but between -86 and 92 "C the protein again apparently unfolds and a narrow line spectrum is obtained at 95 "C, as seen previously with ['HJPhe. Upon cooling to 37 "C, the majority of the narrow component remains, indicating little return to the native protein structure.
The results we have presented in this publication represent the first attempt at detailing the motions of aromatic amino acids in a functional biological membrane protein, bacteriorhodopsin in the purple membrane of H. halobium R1. Our results indicate that tryptophan, phenylalanine, and tyrosine residues are rigid at low temperatures (<-30 "C), but phenylalanine and tyrosine residues are both highly mobile at the temperature of growth of the H. halobium purple membrane (37 "C), undergoing fast (>105-106 s-') 2-fold jumps about CD-CY. Tryptophan residues do not undergo this type of motion even at 85 "C, immediately prior to protein denaturation. Upon denaturation at -90 "C "narrow line" spectra (having line widths -5-10 kHz) are obtained for all three aromatic amino acids, suggesting fast large amplitude motions.
The above results are to be compared with those reported previously (41) for the aliphatic system [y-2Hs]valine-labeled bacteriorhodopsin. In all instances, there is no evidence for fast motion about C"-C8 at any temperature investigated. In the case of valine-labeled purple membranes, motion about C' -CY is fast (>lo6 s-') at all temperatures investigated (down to 120 OK). The increased bulk of the benzenoid rings in Tyr and Phe greatly impede motion of these side chains. When they do begin to move (at about the growth temperature of the organism), rotation is not diffusive but occurs by a 2-fold flipping process, as has been detected previously in solution NMR studies of soluble proteins (27,29,67,68). The additional bulk of the Trp-indole ring prevents even this motion and only small angle librations are allowed. These results are supported by spin-lattice relaxation data to be reported elsewhere,' which show that all systems except for the [y2H6] valine membranes have spin-lattice relaxation times which decrease with increasing temperature since correlation times ( T~) are all >>wo".
Basically similar results with 'H-labeled aromatic amino acids in this and other systems have recently been reported in abstracts by three other independent groups (75)(76)(77), indicating that the types of motions we have observed in the solid state may be of general occurrence.
We believe that the way is now open to examining a whole new area in membrane molecular biology by focusing on the active species, the membrane enzymes, rather than solely observing the membrane lipids, which for technical reasons have been the most attractive species to study during the last 10 years. Clearly, spectral sensitivity is now sufficient to permit extremely detailed investigation of the rates and types of motion of amino acid residues in membrane proteins, including the effects of, e.g. cholesterol and membrane lipid "fluidity" on protein structure.*