Neutron diffraction analysis of the structure of rod photoreceptor membranes in intact retinas

J. Mol. Biol. (1980) 137, 315-348

Neutron Diffraction Analysis of the Structure of Rod Photoreceptor Membranes in Intact Retinas MARKYEAGER',BENNO SCHOENBORN,DONALDEN~ELMAN~, PETER MOORERAND LUBERT STRYER~ Department

of Biology, Brookhaven National Upton, N.Y. 11973, U.S.A.

Laboratory

IDepartment of Molecular Biophysics and Biochemistry Yale University, New Haven, CT 06520, U.S.A. =Department

of Structural

Fairchild Center, Stanford Stanford,

Biology

University School of Medicine

CA 94305, U.S.A.

(Received 30 August 1977, and in revised form 3 November 1979) Neutron diffraction data have been collected from samples containing ten darkadapted Rana catesbiana bullfrog retinas in 100, 80, 60, 40 and 30% DaOt Ringer’s solution using a step-scanning Soller slit diffractometer. Diffraction was also recorded from retinas equilibrated in De0 solutions with varying osmolarity. Rocking curve experiments demonstrated that the rods are disoriented in a cylindrically symmetrical fashion. Structure factor amplitudes were obtained using semi-automated curve-fitting procedures, and phases were obtained by interpreting the D,O-Ha0 and osmotic Patterson maps. In DzO Ringer’s solution the first four structure factors are -353+25, 246=19, 434513 and 383+19. Neutron scattering density profiles were calculated to 75 A resolution using these structure factors. These neutron diffraction data are consistent with the view that the lipid bilayer is a major structural motif of the rod outer segment disc membrane. Neutron Fourier syntheses in different mixtures of D,O and H,O indicate that the intradisc and extradisc spaces are predominantly aqueous, consistent with the increase in the intradisc and extradisc volumes as the Ringer’s solution is made more hypotonic. In isotonic Ringer’s solution, the thicknesses ofthe intradisc and extradisc spaces are about 36 A and 160 A, respectively, and the center-tocenter separation between the 50 A thick lipid bilayers is 88 A. Assuming that the intradisc space is occupied by pure Ringer’s solution, the maximum contrast match point for the hydrocarbon region of the disc membrane is 0.34 x lo-r4 cm/As, corresponding to 13% D,O. If the 30 A thick hydrocarbon region is occupied exclusively by anhydrous protein and hydrocarbon, then the estimated maximum fraction of rhodopsin in the hydrocarbon region is only ~20%. Neutron scattering density profiles in D,O Ringer’s solution are strikingly density on the extradisc side of the disc membrane asymmetric with a scattering

t Abbreviation

used: D, deuterium. 315

0022-283S/80/07031&34/$02.00/0

0 1980 Academic

Press Inc. (London)

Ltd.

3 1(i

Al. YEAAGER

Ir:l’

A I,.

lower tllan tliat, of t,lle intradisc spaw. Models t,hat) orient, rhodopsiu asymniotrically on tlte cytoplnsmic face of tile disc membra,rlc f3tclIditrp into tllcx cxtradisc aqueous space 8w suggested by thfx obsorvcd asymmetry. Howexvr, rhodopsin only accounts for roughly a tlrird of tlw observed asymmetry, suggesting tllat, a substantial amount of llydrogenated material resides in t’hc cstradisc rqioll. Tllis mat’erial may account for tile st,abilizat,ion of tile regular, parallel arrangclments of discs in t’he rod outer segment.

1. Introduction Visual excitation originates with the absorption of light by rhodopsin, the photoreceptor protein in the rod cells of t.he retina. Rhodopsin is an intrinsic membrane opsin, and an 11-cis retinal chromophore. The protein composed of an apoprotein, absorption of a single photon can excite a rod cell (Hecht et al., 1942). Wald (1968) showed that the photoisomerization of 1 l-&s retinal to the all-truns configuration is a critical early event in visual excitation. However, the sequence of events leading from photoisomerization t’o the hyperpolarization of the plasma membrane of the rod cell (Tomita, 1970; Hagins, 1972) is not known. Understanding the molecular basis of visual excitation requires a detailed knowledge of the architecture of rod photoreceptor membranes and of rhodopsin. More generally, structural studies on rod cell membranes may help elucidate some of the principles underlying the organization of other biological membranes. Electron and light microscopy (Sidman, 1957; Nilsson, 1965; Dowling, 1967; Bownds & Brodie, 1975) have shown that frog rod outer segments are cylindrical with a diameter of ~6 pm and a length of ~50 pm. Rod outer segments are aligned approximately parallel to one another in the retina and contain a periodic stack of about 1500 discs. Along the long axis of the rod, the 300 d repeating unit contains two densely staining disc membranes with a narrow intradisc space and a wider extradisc space. The periodic stacking of the disc membranes and the parallel alignment of the rod outer segments in the retina makes the system act like a onedimensional crystal, allowing investigation of rod structure by diffraction methods in intact retinas. However, because the discs are neither perfectly parallel nor stacked with perfect regularity, the limiting resolution of such a diffraction study is about 30 d. Although it is not feasible to determine the structure to atomic resolution, information can be obtained about the distribution of molecular components, protein, lipid and water, along the long axis of the rod outer segment. X-ray diffraction analysis has provided compelling evidence that a major portion of the lipids in the disc membrane are arranged in a bilayer (Blaurock & Wilkins, 1969,1972; Gras & Worthington, 1969; Corless, 1972; Chabre, 1975). However, the location of rhodopsin in the membrane and the structural changes that occur after light absorption have not been clearly defined. Although the diffraction patterns recorded at several laboratories are quite similar, the interpretations have been strikingly divergent : models have been proposed that localize rhodopsin on the intradisc side (Worthington, 1973,1974), on both sides (Blaurock, 1972; Wilkins, 1972 ; Vanderkooi & Sundaralingam, 1970), and partly on the extradisc side of the disc membrane (Corless, 1972 ; Chabre, 1975). A pure lipid bilayer model with no protein (Fig. 1) is quite similar to the experimental X-ray scattering density profile of

NEUTRON

DIFFRACTION

OF

INTaCT

Llpid bllayer

RETINAS

317

rocarbon cham LIpId headgroup

Aqueous

Neutrons I D20)

Neutrons (Ii201 I

I

-150

-100

I

I

I

I

I

-50

0

50

loo

150

Distance C8) FIG. 1. A lipid bilayer model of the disc membrane is shown schematically, followed by the electron density image and the neutron scattering density profiles of this model in Da0 and H,O. The images of the membrane with X-rays (. . .) and neutrons are quite different. The neutron scattering density profile in D,O (--------) is quite different from that in Hz0 (-----). Thus, several images of the membrane can be obtained by changing the D,O to H,O ratio in the Ringer’s solution and using neutron irradiation. Neutron scattering densities of the membrane components were calculated according to Yeager (1975a) using atomic scattering lengths and molecular volumes. The neutron scattering density of water, pw (in cm/As), is given by:

PW T

~0.661

x lo-i4

-f 6.91 x 10-14/?,

where p is the volume fraction of D,O. The neutron scattering and of rhodopsin, pP, will depend on D,O concentration, since bonded to carbon are potentially exchangeable: pp = 1.9 x lo-l4

$

density of the those hydrogens

(1) lipid not

headgroup covalently

1.27 x IO-‘*/+,

where y is the fraction of potentially exchangeable hydrogens. The structure factors X-ray and neutron scattering density profiles were calculated by Fourier transformation respective centrosymmetric step function models: F(h)

:= 2

(2) for the of their

d/2 0

where s(h) is the structure factor for order h; p(z) is the scattering density at a real space distance z; and d is the unit-cell repeat spacing (James, 1965; Levine, 1973). The profiles represent the images of the membrane at 30 A resolution calculated by including the first 11 struoture factors (computed from the model using eqn (3)) in the Fourier transformation : (4) 13*

318

M. YEAGFR i

ET

AL

Blaurock & Wilkins (1972). There is no landmark in t’he profile that can be confidently attributed to protein. Neutron diffraction is a powerful tool for investigating biological membranes (see Schoenborn, 1976), and its utility has been demonstrated in investigations of myelin structure (Kirschner & Caspar, 1972; Kirschner, 1974; Kirschner et al., 1975) and of model membrane systems (Zaccai et al., 1975; Worcester & Franks, 1976; Schoenborn, 1976; Worcester, 1975). One of the advantages of neutrons over X-rays is illustrated by Figure 1. In X-ray diffraction of lipid bilayers, the greatest contrast arises between the phosphate headgroups and the hydrocarbon regions of the bilayer. Since the electron density of rhodopsin is almost the same as that of the headgroups, it cannot be detected in a lowresolution X-ray map if it happens to reside in that region. The contrast between protein and headgroups is much greater for neutrons and hence t’here is hope of detecting the protein. A second advantage of neutrons lies in the ease with which the scattering properties of the aqueous regions of a specimen can be varied over a wide range simply by including D,Ot in the solution. This kind of manipulation allows experiments analogous to isomorphous replacement in crystallography, allowing the determination of reflection phases. As far as is known, D,O-Hz0 substitutions are innocuous. We report here the first use of neutron scattering to investigate the structure of rod photoreceptor membranes in intact retinas. Saibil et al. (1976) have carried out similar experiments on isolated rod outer segments oriented in a magnetic field. Our neutron diffraction pattern in D,O Ringer’s solution extends to 33 A resolution (Yeager, 1975a). We have determined the phases for the first four reflections by D,O-Ha0 exchange and by osmotic shrinking and swelling. It is important to stress that our analysis of the neutron diffraction data is entirely independent of the earlier X-ray studies. Although the resolution of the neutron Fourier syntheses is only 75 A, their high contrast allows us to draw conclusions that complement and extend the X-ray results. Models having a substantial amount of rhodopsin on the cytoplasmic face of the disc membrane protruding into the extradisc aqueous space are most compatible with our neutron Fourier syntheses. Our neutron results also suggest that the extradisc aqueous space contains an appreciable amount of hydrogenated material. The existence of such material may account for the stabilization of the regular, parallel arrangements of discs in the outer segment. Summaries and an abstract of this work have been published (Yeager et al., 1974; Yeager, 1975a,6).

2. Materials and Methods (a) Dissection

and mounting

of retinas

for neutron

diffraction

Rana catesbiana bullfrogs were obtained from the Connecticut Valley Biological Supply Co. (Southampton, MA) or from the Mogul-Ed Biological Supply Co. (Oshkosh, WI). Tetracycline (250 mg daily) was administered orally to frogs that exhibited “red-leg” disease (Gibbs, 1963; Nate et al., 1974). Frogs were kept at room temperature in cages that provided both an aqueous and a dry environment. Bullfrogs were used because the dissected retinas were quite large, ~1.5 cm in diameter. t Abbreviation

used: D, deuterium.

NEUTRON

DIFFRACTION

OF INTACT

RETINAS

319

The Ringer’s solution contained 115 mM-NaCl, 2-O rnM-KCl, 2.0 mM-CaCl,, 20 mMdextrose, 10 mM-Trizma base, 1.0% (w/v) streptomycin sulfate and 1.0% (w/v) pencillin-G. The pH or pD of the Ringer’s solution was titrated to 7.51tO.5 at 5°C using HCl or DC1 (Thompson-Packard). For experiments in different D,O/H,O mixtures, a solution with the desired volume fraction of DzO was prepared by mixing the appropriate volumes of H,O and D20 Ringer’s. Hypertonic Ringer’s solution was prepared by adding sucrose to isotonic D,O Ringer’s (w/v), and hypotonic Ringer’s solutions were prepared by mixing isotonic D,O Ringer’s with pure D,O (v/v). All operations with DzO were carried out as rapidly as possible to minimize exchange with atmospheric HzO. A dozen eyes were dissected for each neutron scattering experiment. Frogs were darkadapted for at least 6 h, and the dissections were performed under dim red light with wavelengths > 640 nm using Corning C&2-64 filters. Frogs were sacrificed by decapitation. The pigment epithelium was removed from the retinas before experimental use. Diffraction patterns from specimens with and without the pigment epithelium were indistinguishable. Decapitation, enucleation, and removal of the retinas and pigment epithelium required about 75 min for 12 eyes. Retinas were allowed to equilibrate with buffer for about 30 min on ice before use. To examine the effect of light, ret)inas were bleached in white light during equilibration. For the experiments in D,O Ringer’s solution of varying osmolarity, the retinas were draped over aluminum slats 4.13 cm x 0.79 cm x 0.064 cm, which were then mounted on 2 parallel rods in an aluminum specimen cell (6.4 cm x 5.7 cm x 1.27 cm) (see Yeager, 19753). When the aluminum slats were clamped toget’her, the 1.2 mm space between adjacent slats allowed the photoreceptor layers of apposed retinas to just touch one anot>her. Sample cells with a path-length of 3.2 mm were used for the experiments in different mixtures of DzO and Hz0 because of the increased beam attenuation from incoherent scattering of hydrogen. The mounting of the retinas in the specimen cell required about 30 min. The two principal obstacles in neutron diffraction work with biological tissue are the finite stability of the samples and the low flux of the neutron beam. Specimen holders accommodated a parallel array of 10 dissected retinas to provide a sufficient number of rods in the neutron beam for diffraction to be observed. To maintain the stability of the diffraction, the dissections were carried out as rapidly as possible without compromising sample orientation ; the retinas were maintained at 5°C; oxygenated Ringer’s solution containing antibiotics and glucose was continuously cycled through the sample cell by peristaltic pumping at a flow-rate of about l-5 ml/h. Rod outer segment birefringence after the neutron diffraction experiments was routinely strong (Yeager, 1975b). Experiments demonstrating the osmotic integrity of the rod outer segments (Fig. 3) are described in Result)s, se&ion (a). X-ray diffraction patterns were collected for specimens in D,O and Hz0 Ringer’s solution. These were identical, showing that D,O did not affect retinal structure adversely (Yeager, 1978). No evidence of specimen deterioraton due to the use of aluminum was detected. Aluminum was used because it is opaque to light and transparent to neutrons. Also, neutrons do not bleach rhodopsin (Yeager, 1975a).

(b) Neutron diffraction

methods

Neutron diffraction data were collected on a paired Soller slit, step-scanning diffractometer at, the High Flux Beam reactor at Brookhaven National Laboratory (Nunes, 1973; Moore et al., 1974). The dimensions of the spaces between Soller slits were 1.9 cm x 0.24 cm. The Soiler slit vanes were 71 cm long and the aperture measured 1.9 cm x 1.9 cm. A pyrolytic graphite monochromator was used to select a wavelength band at 4-19 A (AX/h = 0.025) from the scattered neutrons, which were Bragg reflected into a 3He detector. The horizontal beam divergence was 9’, and the vertical beam divergence (-0.4” at the sample) was defined by the beam pipe rather than by the apertures and heights of the Soiler slit collimators. The wavelength was calibrated by 0 :28 scans of the 002 reflection from a pyrolytic graphite crystal placed in the sample position.

320

31. YEAGER

The 20 axis and w axis of the diffractometer

ET

AL.

were parallel

to one another

dicular to the beam. The x axis WBBparadlel t.n the beam and perpendicular

and perptrl-

to the w axis.

The meridional Bragg diffraction was recorded at w = 0” and x : 0” with t,lle diffracting planes of the rods perpendicular to the 20 axis of the detector, and parallel to the beam direction. The equatorial diffraction was obtained by rot’atinp the sample to x z 90”, SO that the planes were parallel to the 28 axis. Rocking curves were recorded with the detector set at a fixed 20 value, and the angle of t.he rods wit11 respect t’o tile incident, beam was changed by rotating the sample axis in W. Refer to Yrager (1978) for further det,ails of the diffraction apparatus and specimen geometry. The meridional and erluatorial diflrnction patterns were recorded from 28 = 0.24” to 7.20” in increments of 0.08”. The computer that controlled the diffractometer was programmed to collect a specified number of monitor counts at each scattering angle (10s cts/-110 s). In 100°/c and 80% DZO Ringer’s solution the region from 0.24” to 2.80” was scanned for -35 min (~1 min/angle), and the region from 2.80" to 7.20” was scanned for ~2.5 h (-2.5 min/angle). In SOY/, and 40% D,O Ringer’s the diffraction pattern was scanned for -3.5 h (-2.5 min/angle) and -5 11 in 30% D,O. Iu 20% and 0% D,O no A total of 82 days of signal could be detected when scarming for -7 11 (-5 min/angle). beam time was used in 9 poriods from May 1973 to February 1975.

3. Results (a) Neutron

&@-action patterns

Neutron diffraction patterns from retinas equilibrated in different D,O/H,O mixtures (Fig. 2) exhibit meridional Bragg reflections with a period of 29515 A. The smooth curves through the Bragg peaks are the sums of the diffuse equatorial scattering (dotted curves) and Gaussian peaks fitted to the meridional scattering after subtraction of the equatorial scattering. The difference between the Bragg peaks and the equatorial scattering is directly related to the coherent intensity of the reflections. The decrease in the coherent intensity of the reflections as the D,O content of the Ringer’s solution decreases suggests that the contrast within the structure diminishes as the H,O concentration increases. Neutron diffraction experiments conducted in D,O Ringer’s of varying osmolarity (Fig. 3) were recorded from dark-adapted retinas equilibrated in 2% (w/v) sucrose, isotonic, O-8 diluted, and 0.6 diluted D,O Ringer’s solutions. Diffraction patterns recorded from bleached retinas equilibrated in D,O Ringer’s solution were quite similar to the dark patterns. The osmotic sensitivity of the rod outer segments is shown in Figure 4, in which the reflections shift to lower scattering angles and the repeat spacing between discs increases as the hypotonicity of the Ringer’s increases. This behavior is consistent with electron microscopic (Dowling, 1967; DeRobertis & Lasansky, 1961; Brierley et al., 1968; Clark & Branton, 1968; Cohen, 1971; Heller et al., 1971; Korenbrot et al., 1973) and X-ray diffraction experiments (Blaurock and Wilkins, 1972 ; Blaurock, 1972; Corless, 1972 ; Chabre & Cavaggioni, 1975). Although the reflections recorded in 0.4 diluted Ringer’s solution were broadened, they could be indexed to a unique period of 375 8. However, in 4% sucrose Ringer’s solution, two repeat spacings of 285 .& and 230 A were detectable. Two repeat spacings in hypertonic sucrose solutions have also been observed by X-ray diffraction (Chabre & Cavaggioni, 1975).

I

40%

0

0.01

0.02

0.03

0.02

D20

o-03

I04

103

102

IO 0

0.01 ts,-’

FIG. 2. Neutron diffraction patterns of intact retinas in lOO%, SO%, SO%, 40% and 30% D,O Ringer’s solution. Continuous curves through the data bars are the sums of polynomial fits to the background scattering (. . . .) and Gaussian peaks fitted to the background-subtracted reflections. The height of the data bars here and in succeeding Figures represents f 1 standard deviation, assuming Poisson counting statistics.

M. YEAGER

:122

ET

‘:“T/

AL.

/r

(b)

I

I

I

I 0.01

I

I

0.02

o-03

(d)

0

0.01

0.02

0.03

0

-I CR CR FIG. 3. Neutron diffraction patterns of intact retinas equilibrated in D,O Ringer’s solution of varying osmolarity. (a) 2% sucrose Ringer’s solution, d = 294 A; (b) isotonic Ringer’s solution, d = 298 A; (c) 0.8 diluted Ringer’s solution, d = 308 a; (d) 0.6 diluted Ringer’s solution, d = 328 ip. Continuous curves through the data bars are the sums of polynomial fits to the background scattering (. . . .) and Gaussian peaks fitted to the background-subtracted reflections.

(b) Systematic

errors and correction of the experimental

data

In addition to statistical errors, which limit the precision of the observed diffraction, several sources of systematic error must be considered in evaluating the integrated intensity of a reflection from experimental data. These include extinction, disorder, chromatic and geometrical aberrations, absorption, background scattering, and the Lorentz and disorientation corrections. The reader is referred to Arndt & Willis (1966) for an excellent discussion of technical considerations in X-ray and neutron diffraction studies. The linearity of the structure factor of the strong third order versus percentage D,O (Fig. 7) indicated that there was little or no primary or secondary extinction in this system (Caspar & Phillips, 1975). Consistent with this view was the observation that

NEUTRON

DIFFRACTION

OF

INTACT

RETINAS

323

FIG. 4. The reflections in 2% sucrose Ringer’s solution (. . . . .) with a spacing of 294 A (primed) are shifted to higher reciprocal spacings compared with the reflections in 0.8 diluted Ringer’s solution ( ------) with a repeat spacing of 308 A.

diffraction patterns recorded from samples in which the path of the beam through the retinas was 7-9 mm (Fig. 3(b)) and 3.2 mm (Fig. 2) were indistinguishable. Disorder can be due to variation of the unit cell repeat spacing (lattice disorder) or variation of the structure within the unit cell (substitution disorder). Schwartz et al. (1975) have carried out a detailed analysis of disorder from X-ray studies of retinal rods. In the analysis of our neutron scattering data, lattice disorder was allowed for by fitting Gaussian peaks of variable width to the Bragg reflections. Chromatic and geometrical aberrations that result from the use of a diffractometer with finite apertures and a beam of finite bandwidth have been discussed by Schmatz et al. (1974) and by Moore (1975). If the Soller slits caused substantial smearing, the resolution of the minima between the Bragg peaks would have improved as the height of the Soller slits was reduced. Since the resolution of the reflections did not change as the height of the Soller slits was reduced (Yeager, 19753), it was concluded that smearing effects must be quite small. Chromatic aberrations were also negligible, since the bandwidth of the neutron beam was small (AA/h = 0.025). The instrumental angular resolution can be estimated from (Nunes, 1973): (5) where A,, is the width of the direct beam and the scattering angle 20 is less than -30”. For the highest scattering angle measured, +i’.2”, a direct beam width of 0.24” (FWHMt) and AX/X = 0.025, the angular smearing, A(28), is O-3”. This angular smearing in 28 t Abbreviation

used : FWHM,

full

width

at half

maximum.

324

ItI. YEAGER

ET

rl I,

was corrected for by fitting Gaussian peaks of variable width to the Bragg reflections, as done for lattice disorder. Absorption effects arise because the pat,h-length of the beam through t,he sa,mple is dependent on the scattering angle. For a sample of thickness t, the emergent intensit.y I is related to the incident intensity I,, by (Amdt & Willis, 1966) : I - = e-/d, IO

(6)

where p is the linear absorption coefficient. The specimen holders used in the osmotic variation experiments (Fig. 3) had a thickness of 7.9 mm with 0.64 mm aluminum slats, and 1.2 mm for two apposed retinas. If we assume that the rod outer segments of two apposed retinas formed a 0.1 mm thick diffracting layer centered in the 1.2 mm tissue space, then the diffracted neutrons were scattered through retinal tissue up to 20 = 4@0”. At higher angles the diffracted neutrons would pass through both retina and aluminium. The structural analysis including four reflections (see Structural Analysis) included data collected up to a scattering angle of 3.8”. Over the range in 20 from 0 to 3.8”, there was virtually no change in path-length through the sample (t/cos20). Also, the neutrons only passed through retinal tissue. (In the sample holders with a thickness of 3.2 mm used to collect the D,O-H,O contrast variation data shown in Figure 2, scattered neutrons up to 28 = 9.75” passed exclusively through retinal tissue.) If we assume that the linear absorption coefficient is constant throughout the retina, then I/I, did not vary with 20. Therefore, no absorption correction had to be applied within a given data set. In the 7.9 mm sample cell the structure factors for orders 5 to 8 in D,O Ringer’s were overestimated by less than 7% due to scattered neutrons passing through aluminum. When comparing data sets collected under different conditions, attenuation was corrected for by normalizing the scattering profiles to the intensity of the direct beam transmitted through the samples. The normalization took into account the presence of the aluminum slats in the sample holder, which had the effect of increasing the direct beam intensity compared to what it would have been with only tissue in the beam. The observed I/I, value was given by the sum of two exponential terms for the absorption due to aluminum and tissue, weighted by the volume fractions of aluminum and tissue in the beam. From this equation, values for the linear absorption coefficient of the bulk tissue, pr, were obtained for each experiment. Knowing t+, the normalization factor for the scattering profile followed from equation (6). Average values for pLT of the retina in lOOO/& SO%, 60% and 40% D,O were 1.0, l-6, 2-2 and 2-8 cm-l, respectively. Corresponding values for the linear absorption coefficient of loo%, SO%, From these values 60% and 40% D,O were 0.55, 1.4, 2.2 and 3-O cm-l, respectively. the hydration of the retina is about 75% by volume, and the linear absorption coefficient of the dry retina is -2.4 cm-l. Special care was taken to obtain the best estimate of the background scattering, since the entire retina was exposed to the beam. The background in some previous low-angle studies was determined by fitting a line or smooth curve between the minima around a Bragg reflection (Kirschner, 1971; Worthington & McIntosh; 1974). This approach is valid if the minima between reflections are well-defined plateaux, which indicates that adjacent reflections are resolved. The regions between reflections

NEUTRON

DIFFRACTION

OF

INTACT

RETINAS

325

in the neutron diffraction patberns from retinas were not plateaux, which indicates that adjacent reflections were not completely resolved. Therefore, the background was ascertained experimentally. The equatorial scattering, obtained by rotating the retinas 90” in angle x from the Bragg reflecting position, was used as an estimate of background. The diffuse intensity centered at l/55 A-l in equatorial X-ray diffraction patterns has been interpreted as diffraction between rhodopsin molecules in the plane of the disc membrane (Blasie, 1972). The equatorial neutron diffraction patterns in Figure 5, however, are quite featureless compared w&h the Bragg diffraction (Fig. 2) and do not display peaks of intensity centered at l/55 A-l. The patterns in Figure 5 do change as the D,O concentration is varied. As the D,O concentrat’ion decreases, the scattering at reciprocal space distances greater than O-016 d-l increases because of incoherent scattering from hydrogen. The scattering at distances less than @016 .J-’ is much greater than that observed with only buffer in the sample cell, and this diffuse scattering probably arises from those constituents of the retina not periodically organized to give Bragg diffraction. The scattering at distances less than 0.016 Am1 decreases as the D,O concentration decreases, which suggests that structures causing this diffuse scattering (e.g. protein, lipid and carbohydrate) are being contrast-matched as the D,O concentration decreases. This diffuse scattering probably arises from non-periodic constituents of the retina, which do not appear to contain substantial oriented features as seen in the electron microscope (SjGstrand, 1959). Assuming the diffuse scattering to be isotropic, it can be used as an estimate of the background scattering underlying the meridional Bragg reflections. This approach for obtaining the background would be inapplicable in situations where there is oriented equatorial scattering from the structure. In such situations the scattering between the meridional and equatorial reflections, for instance at 45”, could be used to obtain the background. (The scattering at 45” could not be used in our experiments, however, because of the large mosaic spread of the samples (see below).) Polynomial functions fitted to the equatorial scattering were subtracted from the observed meridional Bragg diffraction. The background-subtracted reflections were fitted by Gaussian peaks. The sum of these Gaussians and the polynomials corresponding to the equatorial scattering closely matched the observed meridional diffraction patterns (Fig. 2). The observed reflection intensities I(h) were given by the product of the height and full width at e-t height of the Gaussian peaks, and the repeat spacing was calculated from the positions of centers of the Gaussian peaks. The details and evaluation of the data analysis procedure have been published elsewhere (Yeager, 19753). The integrated intensity lint of reflection h is given by Arndt & Willis (1966) :

(7) where K is a proportionality constant and the polarization factor P(h) equals unity at low angles and in neutron diffraction. The Lorentz factor L is a geometric correction that takes into account the fact that different reflections spend different amounts of time intersecting the Ewald sphere to satisfy the Bragg condition. In the rotating crystal method the rate of rotation of a zero-level equatorial reciprocal lattice point is inversely proportional to the reciprocal space distance between the origin in reciprocal

M. YEAGER

ET

AL.

105 I05

104 104

103

102

IO

0

0.02

0.01

0.03

(xl-1 FIG. 5. Intact the equatorial regression fits.

retinas were rotated 90” in angle x from the Bragg reflecting position to obtain scattering. The smooth cwves drawn through the data bars are polynomial

NEUTRON

DIFFRACTION

OF INTACT

space and the point of intersection of the reflection with reciprocal of the Lorentz factor, L- l, in such an experiment L-l(h)

= sin28 g h

RETINAS

the Ewald sphere. is given by:

327

The

(8)

and so L-l is approximated by h at low angles. Therefore, the observed intensities must be multiplied by sin20 to make them proportional to the squares of the structure factors in equation (7). Rotating crystal geometry is equivalent to that of fiber diffraction in which a parallel bundle of fibers is held stationary and the fibers have random rotational orientations about their long axes (James, 1965). The reciprocal lattice points are spread out into circles with radii proportional to the distance between the fiber axis and the intersection of the reflection with the Ewald sphere. Since a greater fraction of the intensity is recorded for reflections at a smaller radius than for those at a greater radius, the Lorentz factor given by equation (8) will apply to the equatorial reflections on the zero layer-line. A further disorientation correction D(h) is necessary if the fibers are not exactly parallel to one another but exhibit disorientation about the fiber axis (Franklin & Gosling, 1953; Arnott, 1965). Elegant treatments of the Lorentz and disorientation corrections in fiber diffraction have been’published (Cella et al., 1970; Dears, 1952; Holmes & Barrington Leigh, 1974; Stubbs, 1974). Retinal rod outer segments have random rotational orientations about their long axes, and the discs are cylindrically symmetric. The reciprocal Lorentz factor given by equation (8) will therefore apply to the meridional reflections from rods, and the structure factor amplitude l>(h)\ is given by: l&h) 1 = @D(h)

I(h),

(9)

where L-l(h) = h and I(h) is the background-subtracte intensity of reflection h integrated over angle 213. In order to calculate a disorientation correction for each reflection, the intensity distribution across the disorientation arcs was mapped by rocking curve analysis. The detector was fixed at a scattering angle 28 corresponding to the peak maximum for a particular reelection, and the sample was rocked in angle W. The rocking curves shown in Figure 6(a) were recorded with the sample oriented to observe both the meridional and equatorial diffraction. Since the equatorial scattering with the sample at x = 90” would be expected to be spatially isotropic, the rocking curves should be horizontal lines. The equatorial rocking curves in Figure 6(a) are fitted quite well by lines and the profiles are almost horizontal. To correct for the decreased amount of material in the beam and the lower beam flux as w increased, the scattering observed at each value of w was normalized by using direct beam measurements through the sample with the detector at 20 = 0”. The slight upward tailing of the equatorial profiles for reflections 1 and 2 in Figure 6(a) may indicate an incomplete correction for this effect. The equatorial rocking curves were considered as the background for the meridional rocking curves in Figure 6(a). The background-subtracted rocking curves shown in Figure 6(b) are fitted quite well by Gaussians. It is clear from these profiles that the mosaic spreads for orders 1 to 4 are almost the same and are quite large. The mosaic spreads are defined by the full width at e-* height of the Gaussians (fl

M. YEAGER

ET

AL.

lb)

I’ -40

-20

0

20 40

jl -40

lilI,.d -20

0

20

40

-40

-20

/-L-L& 0

20

40

-40

-20

J 0

20

40

wCO) FIG. 6. Rocking curves for the first 4 reflections from neutron diffraction of retinas in D,O Ringer’s solution with scattering angles of 0.8”, 1.6’, 2.4’ and 3.2”, and scale factors of 6.0 x 10V4, 2.6 x 10-s, 2.86 x 10-s and 8.33 x 10m3, respectively. Lines were fitted to rocking curves with the sample aligned to observe the equatorial scattering at x = 90”. (a) The curves drawn through the data recorded with the sample aligned to observe the Bragg diffraction at x = 0” are the sums of the linear equatorial rocking curves and Gaussian peaks fitted to the background-subtracted meridional rocking curves. (b) Background-subtracted rocking curves for the first 4 reflections from neutron diffraction of retinas in D,O Ringer’s solution with scale factors of 667 x 10m4, 5.0 x 1O-3, 3.33 x 10m3 and 1.25 x 10T3, respectively. The data bars display the rocking curves after subtraction of the equatorial rocking curves in (a). The full width at e-* height of the Gaussian profiles fitted to the data defines the mosaic spread 7. The mosaic spreads for reflections 1 to 4 are 558~0~6”, 5&O& 1.8”, 58.611.2” and 55.2&1.4”, respectively.

standard deviation). These widths for orders 1 to 4 are 55*8&0.6”, 550&1*8”, 58.6f l-2” and 55*2* l-4’, respectively. The Gaussian rocking curves in Figure 6(b) prove that the rods are disoriented in the plane that intersects the Ewald plane. Photomicrographs of the birefringence in the plane of the retina (Yeager, 19753) indicate that disorientation of the rods about their long axes is indeed present, and that the disorientation does not occur along a particular axis of the plane of the retina. Thus, the rods are disoriented in a roughly cylindrically symmetrical fashion so that the reflections in reciprocal space should be approximately circular in a cross-section perpendicular to the Ewald plane. Having characterized the meridional reflections, we can now calculate the disorientation correction D(h) for each reflection (Saxena & Schoenborn, 1977). (No allowance needs to be made for the analyzer crystal, since its size and mosaic spread were sufficient to detect all neutrons in the measured wavelength range that passed the Soller slits.) D(h) is the reciprocal of the fraction of the intensity of reflection h

NEUTRON

DIFFRACTION

OF INTACT

RETINA8

329

accepted by the Soller slit aperture closest to the graphite monochromator. D(h) will of the higher increase with increasing h, since a smaller fraction of the intensity orders will be recorded compared with the lower orders. To calculate D(h) the sample was considered to be made up of a one-dimensional vertical array of scattering elements over the sample height. The diffracted rays from a single element are smeared vertically due to the mosaic spread 7. (Since the vertical beam divergence was -lo and 7 -BOO, the additional smearing due to beam divergence is negligible.) For a single element the vertical height of the smeared reflection in the plane of the Soller slit aperture is 20 . 7 . L, where L is the horizontal distance from the sample to the aperture. The total intensity in the vertical direction of reflection h was obtained by numerical integration over all elements assuming the slit to be of infinite height. D(h) was then given by the ratio of this total imensity to the integrated intensity with finite limits given by the slit height. Using the values L = 86-8 cm, sample height. = 19 mm, slit height = 19 mm, X = 4.19 A and d = 300 A, the values of D(h) for orders 1 to 8 were 1.2, 15, 2.0, 2.7, 3.4, 4.1, 4.7 and 54, respectively. Thus, the percentage of the vertical intensity that was recorded for orders 1 to 8 was 83,66, 49, 37, 30, 25, 21 and 18, respectively. A preliminary analysis of diffraction patterns recorded using a two-dimensional, position-sensitive detector (Yeager, 1975a), in which the total vertical intensity of the reflections is recorded, confirmed the above treatment of the Soller slit diffraction patterns. The above treatment of the Lorentz and disorientation corrections was justified, because the mosaic spread was the dominant influence on the recorded intensity; r] was substantially larger than the beam divergence, the chromatic bandwidth of the neutron beam and the mosaic spread of the graphite monochromator.

(c) Structure factor amplitudes Ideally, the D,O-H,O contrast variation data should be collected on a single sample. This was not possible in these experiments because of the finite stability of the tissue. Data collected on different specimens in different D,O/H,O mixtures were normalized to constant beam flux and monitor counts, thereby placing the structure factors derived from them on the same relative scale. The amplitudes from different experiments a.t the same D,O concentration varied more than predicted on the basis of counting statistics (Fig. 7), reflecting differences in the number of rods contributing to diffraction from preparation to preparation. However, since variation of this kind was several times less than the changes produced by varying DzO concentration over the range considered, comparison of amplitudes between experiments is reasonable. Plots of P(h) versus percentage D,O (Fig. 7) suggest that P(h) depends heady on percentage D,O, & expected for hydrated centrosymmetric structures (Bragg & Perutz, 1952; Zaccai et al., 1975; Worcester & Franks, 1976). The least-squares regression lines shown in Figure 7 had linear correlation coefficients of 0.97, O-92, 0.98 and 0.95 for orders 1 through 4, respectively, consistent with this hypothesis. The data were pooled, and “best” estimates for structure factor amplitudes (Table 1) were obtained using the regression lines shown in Figure 7. The structure factor amplitudes for the experiments in D,O Ringer’s solution of varying osmolarity (Fig. 3) are listed in Table 2.

M.

YEAGER

ICY’

A

r,.

h=2

I

240

120

60

0

i

0

20

40

60

00

100

% D,O

FIG. 7. Structure factor amplitudes F(h) for reflections 1, 2, 3 and 4 increase linearly as the II,0 concentration is increased. The error bars represent +. 1 standard deviation obtained by propagation of statistical errors through background subtraction and fitting of Gaussian reflections to the Bragg peaks. Each point plotted represents an independent experiment carried out on new fresh retina preparations. No attempt was made to normalize measured intensities for differences in the amounts of retinal material in different specimens.

TABLE

1

Structure factor amplitudes from neutron diffraction of intact retinas in different mixtures of D,O and H,O

h

100

80

y” l&O 60

40

1 2 3 4

199; 16 136118 242117 214&20

151*14 106115 18Si 14 167+17

103+12 76&13 134113 121115

55*11 45112 79&11 751 13

The structure factors were calculated from the [east-squares The errors are 1 standard deviation obtained from the standard ordinate intercept of the regression lines.

regression deviation

30 31110 30111 52+10 52&12 lines shown in Fig. 7. values of the slope and

NEUTRON

DIFFRACTION

OF TABLE

INTACT

RETINAS

2

Structure factor amplitudes from neutron diffraction in D,O Ringer’s solutions

h 1 2 3 4

d(A)

2% (w/v)

sucrose

331

Ringer’s solution Isotonic

of intact retinas

0.8 diluted

0.6 diluted

31912 23142 38214 32813

33312 25412 441%:~ 378f3

315&2 210&3 397&4 31613

221&3 16545 284&3 237&5

294

298

308

328

The structure factors were calculated from eqn (9) using the data shown in Fig. 3. The errors are 1 standard deviation obtained by propagation of statistical errors through background subtraction and fitting of Gaussian reflections to the Bragg peaks.

(d) Effect of light Data were collected on dark-adapted and bleached retinas and structure factors were obtained for both kinds of specimens. To correct for variation in the number of diffracting rods in different samples, data sets were scaled to one another by normalizing to the sum of the intensities for the first eight reflections. Within error, the structure factors from both kinds of specimens were the same. Only small structural changes can be taking place in the disc membrane upon bleaching.

4. Structural Analysis The intensities, phases and disorientation corrections are considerably more reliable for the first four orders than for the higher orders. The structural analysis described below will therefore consider only the first four strong reflections. (a) Patterson maps One-dimensional Patterson the corrected intensities :

maps were calculated

P(x) = ;,&

h D(h) I(h) cos

by Fourier

transformation

of

,

where P(x) is the value of the Patterson function at a real space distance x and n = 4. Patterson maps calculated on the same relative scale for experiments in different D,O/H,O mixtures (Fig. 8) displayed broad peaks centered at about 84 A with magnitudes of about 60% of the origin peaks. The movement of the 84 A correlation to larger distances as the DsO concentration decreases was not observed in all experiments and may not be significant. As the D,O concentration of the Ringer’s solution decreased, the reproducible characteristics of the Patterson maps were a reduction in the magnitude of the origin peak and the 84 A correlation. This behavior is consistent with a reduction in the contrast within the unit cell as the D,O concentration

M.

332

25

0

YEACER

50

ET

75

A I,

100

150

125

Distance (A) FIG. 8. Patterson the first 4 reflections (---), soy0 (.

maps calculated on the same relative for neutron diffraction experiments . .). 60% (-----), and 4094 (-.-.-.)

scale from the corrected intensities of on intact retinas equilibrated in 100% I>,0 Ringer’s solution.

decreases and is also suggested by the reduced coherent intensity of the reflections as the Ha0 content of the Ringer’s solution increases (Fig. 2). The general modulations of the Patterson maps at all D,O concentrations are qualitatively the same, which suggests that at 75 11 resolution the principal effect of changing the D,O concentration from lOOo/o to 30% is to alter the scattering density of the aqueous regions without changing the internal contrast of the disc membrane. Patterson maps calculated from the intensities from experiments in D,O Ringer’s solution with varying osmolarity (Fig. 9) also display a broad correlation that moves from 83 A in 2% sucrose Ringer’s solution to 93 d in 0.6 diluted Ringer’s solution.

0

I

I

I

I

I

I

I

1 25

I 50

1 75

I 100

I 125

I 150

I 175

Dislance ( 8, ) FIG. 9. Patterson maps calculated from the first 4 reflections for neutron diffraction experiments on intact retinas equilibrated in D,O Ringer’s solution of varying osmolarity. The broad eorrelation at 83 if shifts to longer distances as the osmolarity decreases. The profiles for all experiments have been normalized to the magnitude of their origin peaks.

NEUTRON

DIFFRACTION

OF

INTACT

RETINAS

3.73

The vectors at distances less than 30 A should be dominated by intramembrane correlations (Caspar & Kirschner, 1971). The similarity of the Patterson maps at distances less than 30 A suggests that there is no rearrangement of matter within the membrane as the tonicity is varied.

(b) Pha.se determkation Electron microscope images of disc membranes display mirror planes in the middle of the intradisc and extradisc regions. For a centrosymmetric crystal the structure factors are real, and the phase problem reduces to assigning a + or - sign to the square root of the corrected intensities. Plots of the st’ructure factor moduli verSu$ percentage D,O (Fig. 7) show that the phases of the first four orders do not change until the D,O concentration is less than 30%. The phases for the first four orders were det,ermined by interpretation of the Patterson maps in Figures 8 and 9. It should be stressed that Patterson maps do not and uniquely define the structure and hence the phases. However, the D,O-Ha0 osmotic Patterson maps do provide several pieces of independent evidence that favor the model described below. The broad correlation at 84 A in Figure 8 is interpreted as arising from correlations between two regions of density spaced 84 A apart, whose neutron scattering densities relative to Da0 must have the same sign. This interpretation is compatible with electron microscope images of rod outer segment disc membranes that show two densely stained regions in the unit cell. The D,O-H,O Patt~erson maps in Figure 8 show that the contrast within the structure decreases as the D,O concentration decreases. At 75 A resolution, this decrease in contrast must, be due to the decreased scattering density in the aqueous regions of the structure and not due to contrast, fluctuations within the membrane interior, since all the Patterson maps have the same general modulations and differ only in the amplitude of the modulations. Thus the regions giving rise to the Patterson peak must have neutron scattering densities less than D,O. The shift of the 83 A correlation to 93 A (Fig. 9) suggests that the regions 83 A apart in 2% sucrose Ringer’s solution separate with decreasing osmolarity. Since the 83 A peak shows little or no increase in width as it shifts to longer distances, the widths of the regions 83 A apart remain constant as the regions separate. As the volume of the aqueous compartments increases with decreasing osmolarity, the unit cell size increases. It is clear, therefore, that a model for the disc equivalent to the neutron scattering density profile of the lipid bilayer model for the disc membrane shown in Figure 1 is compatible with the scattering data. The broad Patterson peak at 84 A would represent the separation of centers of the disc membrane pairs. Fourier transformation of this model gives the phases r, 0, 0, 0 for the first four orders for the osmotic experiment’s in D,O Ringer’s solution. Knowing that there is no phase change between 100% and 30% D,O these must be the appropriate phases for the reflections at all the D,O concentrations tested. It should be noted that of all the possible phase combinations, only the phases n, 0, 0, 0 yield Fourier syntheses for the disc membrane in D,O that can be interpreted in terms of two membranes in the unit cell with a higher neutron scattering density in the aqueous regions compared with the membranes.

x14

11.

(c) Neutron One-dimensional

YEACER

ET

.3 1,

scattering density ivrojiles

neut.ron scattering

density

profiles

were calculated

from :

Zvrhx p(x) = 5 i s(h) F(h) cos d, h 1

(11)

where p(x) is the neutron scattering density at a real space distance x, and s(h) is the cosine of the phase of reflection h. Centers of symmetry are located at x = M/2, where k is an integer. Neutron scattering density profiles for experiments in different mixtures of D,O and H,O were calculated on the same relative scale to 75 A resolution. The Fourier maps in Figure 10 show that as the D,O concentration decreases, the overall contrast between the low density troughs at. &44 A and the remainder of the structure diminishes The highest density at all D,O concentrations is at the origin. between The profiles also display shoulders at * 115 A, with a density intermediate that at the origin and the troughs. 2%

I

I

I

I

I

I

I

I 100

I 150

-

0 g=z1

r 48 m E zl 2= B Y) E 2 o2 z

30% I -150

I -100

1 -50

I 0

I 50

D20

Dlstonce I H) FIG. 10. Neutron scattering densit’y profiles calculated from the data in Table 1 for the first 4 orders of diffraction from intact retinas equilibrated in isotonic Ringer’s solution containing different mixtures of D,O and H,O. The Fourier syntheses from top to bottom are experiments in lOOoh, SO%, SO%, 40% and 30% D,O. The absolute neutron scattering density scale was assigned by assuming that (1) the highest scattering density in the Fourier synthesis in 100% D,O Ringer’s solution (at 0 A) has the neutron scattering density of 100% D,O Ringer’s solution. (2) Structures in different D,O/H,O mixtures are isomorphous. (3) Series termination errors are small. (4) The neutron scattering density at 1 44 .k, obtained by extrapolat,ion of the contrast at + 44 A to zero in Fig. 12, is independent of D,O concentration.

Since the amplitudes of the structure factors for the different D,O/H,O mixtures (Table 1) are on the same relative scale, the relative amplitudes of the density fluctuations within the disc structures at different D,O concentrations are depicted correctly. In order to compare these profiles properly and to interpret them chemically, it is essential to know the absolute values for neutron scattering density at each point in every profile. In order to obtain this information from scattering data, the zero-order scattering would have to be measured and the number of discs contributing to coherent scattering determined for each data set. Both quantities would be extremely difficult to measure. A chemically reasonable argument can be made, however, which

NEUTRON

DIFFRACTION

OF

INTACT

RETINAS

335

permits assignment of approximate values for neutron scattering densities within these profiles. It is undoubtedly true that the low density troughs at &44 A in all profiles represent the hydrocarbon-rich interior of the lipid bilayers of the disc membranes. This region of the structure is the least accessible to solvent and should contain the fewest solvent-exchangeable hydrogens per unit volume. Its absolute neutron scattering density should be nearly independent of solvent D content. For purposes of interpretation, we superimpose the profiles so as to align the low density troughs, as in Figure 11, assuming thereby that there are no exchangeable hydrogens in the lipid bilayer region.

I

I

I

,

-150

-100

-50

0

I

50

I 100

I 150

Dlstonce ( & 1 FIG. 11. Neutron scattering density profiles calculated from the first 4 orders of diffraction from intact retinas equilibrated in D,O Ringer’s solution of varying osmolarity. (a) 2% sucrose Ringer’s Ringer’s solution, solution, d = 294 ip; (b) isotonic Ringer’s solution, d == 298 p\; (c) 0.8 diluted tl = 308 A; (d) 0.6 diluted Ringer’s solution, d = 328 A. The separation between the low density troughs increases from 85 .& in 2% sucrose Ringer’s solution t,o 95 A in 0.6 diluted Ringer’s solution. The Fourier syntheses have been scaled to the same vert)ical peak-to-trough distance.

The neutron scattering densities of regions accessible to solvent should depend linearly on solvent D,O content. Those regions richest in solvent (or more strictly, having the highest number of exchangeable protons per unit volume) will increase in density most rapidly as solvent D,O content rises. It is clear from Figure 11 that the intradisc space (centered at 0 A) is the region whose density changes most rapidly with solvent D,O content. The density of the extradisc space also varies, but at only roughly 70% of the rate of the density of the intradisc space. This difference indicates that there must be substantial amounts of material in the extradisc space, occupying roughly 30% of the volume of that space, in excess of whatever non-aqueous material there may be in the intradisc space. In order to assign approximate values for neutron scattering density to these profiles, a second assumption must be made. A value for the neutron scattering density must be assigned to one point in each profile. It is clear from the argument above that

ET' AL.

M. YEAGER

336

the intradisc space is the portion of the disc structure richest in solvent. For t,he purposes of interpretation. we will assume that the density at x = 0 A in each profile is that of the bulk solvent used for the experiment in question. This gives us values at x = 0 A of 6.35 X 10-14, 4.97 X lo- 14, 3.59 x 10-14. 2.20 x lo-l4 and 1.51 x lo-l4 cm/A3 for loo%, SOY& BOO/b,40% and 30% D,U, respectively. Based on these two assumptions, the scaling of the profiles shown in Figure 10 is obtained and further interpretation can be made. By examining the dependence of the contrast at distance x in the structure, C(x), on the scattering density of the solvent, pw, the scattering density at the center of the low densit,y troughs at f44 A can be obtained. The scattering density at distance x: p(x), in a one-dimensional hydrated structure can be broken down into contributions from the aqueous (w) and non-aqueous (s) regions: P(X) = (1 - a% where X is the volume

fraction

(1.2)

+ &b

of water at x. C(x) is then given by: C(x) = Pw - P(X).

Substitution

of equation

(8) into equation

(13)

(9) yields

C(x) = pw (1 - X) - ps (1 -

X).

(14)

Since the contrast, C(x), depends linearly on the scattering density of the aqueous solvent, pw, the scattering density, ps, of the non-aqueous constituents in the unit cell at distance x can be obtained by extrapolating a plot of C(x) versus pW to C(x) = 0. (Note that this extrapolation is independent of t,he absolute density scale of the Fourier syntheses and only depends on the existence of an aqueous region in the unit cell.) If the contrast values, C(x), are on an absolute scattering density scale, the slope of the contrast plot, (1 - X), yields the volume fraction of water at distance x (Harrison, 1969). By extrapolating a plot of C( & 44 A) versus pW to C( &44 8) = 0, the neutron scattering density x lo- 14, corresponding at the center of the low density troughs is found to be O-05( f0.29) to 9&4% D,O. Neutron scattering density profiles for experiments in D,O Ringer’s solution of varying osmolarity are shown in Figure 11. The profile in 2% sucrose Ringer’s solution (Fig. 11(a)) displays two low density troughs at h42.5 A, which move to k47.5 A in O-6 diluted Ringer’s solution (Fig. 1 l(d)). The repeat spacing also increases from 294 b to 328 b in proceeding from 2% sucrose to 0.6 diluted Ringer’s solution. The highest density in the profiles occurs between the troughs at the origin. The profiles also display density shoulders that move from &115 A in 2% sucrose Ringer’s solution to + 130 A in 0.6 diluted Ringer’s solution.

5. Discussion (a) Biluyer

arrangement

of disc membrane lipids

The molecular interpretation of the observed neutron scattering density profiles (Figs 10 and 11) is shown in Figure 12. The continuous neutron scattering density profile is the Fourier synthesis at 75 d resolution for an experiment in isotonic D,O Ringer’s solution. The high density regions between -18 and + 18 A and between

NEUTRON

-150

-100

DIFFRACTION

-50

0

OF INTACT

50 100 Dstonce (8,

150

RETINAS

200

337

250

GO

FIG. 12. Neutron scattering density profiles at 75 A resolution of rod outer segment disc membranes in D,O. Fourier synthesis aalculated for the pure lipid bilayer model depicted at the top of the Figure (. . . . .). The experimental Fourier synthesis was calculated from data recorded in D,O Ringer’s solution (-----). The experimental Fourier synthesis was saaled to the calculated Fourier synthesis of the lipid bilayer model by assuming that the neutron scattering density in the center of the intradisc space (0 A) is that of pure D,O (6.35 x lo-i4 cm/As) and the scattering density at the center of the low density troughs is -0.02~ IO-l4 cm/A3 (Fig. 1) for the lipid bilayer profile and 0.34 x 10-i* cm/A” (Fig. 12) for the experimental profile. The higher density at 144 A in the experimental profile compared with the bilayer model protie would be due to protein in the hydrocarbon region of the membrane. The lower density in the extradisc cytoplasmic region (70 A to 230 A) in the experimental profile would be due principally to non-rhodopsin hydrogenated material as well as rhodopsin extending into the cytoplasmic space.

70 A and 230 A are interpreted as being predominantly aqueous, since the scattering density in these regions decreases with decreasing D,O concentration (Fig. 10). The narrower 36 A aqueous region between the low density troughs in Figure 12 is interpreted as the intradisc region, and the wider 160 A aqueous region is interpreted as the extradiso region. The increases in the unit cell dimension and the size of the intradisc space as the osmolarity of the Ringer’s solution decreases (Fig. 11) indicate that the volume of water in the intradisc and extradisc compartments increases as the osmolarity of the Ringer’s solution decreases. The contrast-match point for the troughs centered at f44 d is 0.05 (fO.29) x lo-l4 cm/A”. Th’ IS 1ow scattering density (Fig. 1) and the -50 A width of the troughs suggest that a major portion of the lipid molecules are arranged in a bilayer configuration. In isotonic Ringer’s solution the bilayers are separated by 88 A, and the separation between the bilayers increases as the Ringer’s solution is made more hypotonic (Fig. 11). (b) Iderpretation

of the contrast-match

point a.t the center of the lipid

b&layers

The contrast-match point for the center of the disc membrane lipid bilayers at -&44 A contains information about the chemical composition of the hydrocarbon region. To justify a quantitative interpretation of the contrast-match point, the effect 14

M. YEAGER

338

X7’

rl I,.

of resolution had to be examined. Neutron scattering density profiles were calculated for disc membrane models with varying amounts of protein in t’he hydrocarbon region (section (d), below). Model Fourier syntheses in different D&/H,0 mixtures were calculated at 75 A resolution, and the contrast at *44 A, C(+44 A), given hy equation (10) was plotted versus D,O concentration. The error due to the resolution was assessed by comparing the exact contrast-match point defined by the model, and the calculated contrast-match point at 75 A resolution. These calculations revealed that the value of C(&44 A) at 75 A resolution differs by only -1% D,O from the exact value defined by the models. Furthermore, the contrast at &44 A is fairly representative of the entire hydrocarbon region of the lipid bilayer, because small changes in the neutron scattering density level through the hydrocarbon region are averaged due to the resolution of 75 A. Knowing that the contrast-match value at &44 A should be reliable, the chemical composition of the hydrocarbon region of the disc membrane can be estimated. The protein to lipid weight ratio of disc membranes is 50+10/50flO (Daemen, 1973); the molecular weight of rhodopsin is 38,000&3000 (Robinson et al., 1972; Heitzman, 1972; Daemen et al., 1972; Lewis et al., 1974; Yeager, 1975a); rhodopsin comprises SO*lO% of the disc membrane protein (Papermaster & Dreyer, 1974); and the molecular weight of the average rod outer segment lipid is 800 (Daemen, 1973). From these values we obtain a lipid to rhodopsin stoichiometry of 6Of 15. Assuming that the hydrocarbon region of the disc membrane lipid bilayer is occupied exclusively by anhydrous protein and hydrocarbon, the neutron scattering density of the hydrocarbon region, P,,,, is given by: Pm- -

XHPH + X,PF?

(15)

where X, is the volume fraction of protein in the hydrocarbon fraction of hydrocarbon, and X, + X, = 1. From equation pH = -0.02 x 10-14 cm/A3, equation (15) becomes : Pm=

-0.02~10-~~

+ X,(1.92~10-~~

region, X, is the volume (2) and using a value of

+ 1.27~10-~~&),

where /3 is the fraction of Da0 in the solvent and y is the fraction gens in rhodopsin that exchange. X, is given by:

x, =

(16)

of the labile hydro-

pm + 0.02 x 10-14 1.92 x lo- l4 + 1.27 x 10-14/3~y’

(17)

For a lipid to rhodopsin stoichiometry of 60& 15, the total volume of lipid hydrocarbon is (60&15) x (982 A3/2 hydrocarbon chains) = 58,920-&14,730 A3. (The volume of 982 b for two average rod outer segment hydrocarbon chains was calculated from volumes of the atomic nuclei (Traubc, 1899).) Th e calculated anhydrous volumes of rhodopsin and non-rhodopsin protein are 47,000 A3 and 11,750 A3. By setting pm = pw (given by eqn (1)) and substituting the maximum experimental value of p = 0.13 into equation (17) with y = 0, the maximum volume fraction of anhydrous protein residing in the hydrocarbon region is ~15~/~. If all non-rhodopsin protein resides outside the membrane interior then the estimated maximum fraction of dwdopsin in the 30 A thick hydrocarbon region is only -2O%, a low value for an intrinsic membrane protein.

NEUTRON

DIFFRACTION

OF INTACT

RETINAS

339

The calculation of the contrast-match point relied on the assumption that the intradisc space is occupied by pure Ringer’s solution. If the intradisc space contains non-aqueous material with a contrast-match value greater than 9&4% D,O (for example protein or carbohydrate), then the contrast-match value at *44 A would be greater than 9% D,O, and the estimate of the amount of protein in the hydrocarbon region would be increased. (c) Asymmetry

of the neutron scattering density pro$les

The dotted profile in Figure 12 is the calculated neutron scattering density profile in D,O at 75 .L%resolution for the lipid bilayer model shown at the top of the Figure. The curvature in the troughs and the modulation in the extradisc space are due to series termination error in the calculated profile. As expected, the density levels of the intradisc and extradisc spaces in the calculated Fourier synthesis for a lipid bilayer model are almost the same, except for a slight difference due to series termination error. In contrast, the experimental Fourier synthesis (continuous curve) is strikingly asymmetric : the neutron scattering density levels in the in.tradisc and extradisc spaces are markedly different. The calculated Fourier synthesis for the lipid bilayer demonstrates that this difference cannot be accounted for by series termination error. The lower density of the extradisc space compared to the intradisc space in the experimental Fourier synthesis in D,O indicates that there is a substantial amount of hydrogenated material in the extradisc region. To investigate the effect of resolution on the neutron scattering density profiles, Fourier syntheses including six and eight reflections (50 A and 38 A resolution, respectively) were calculated using the observed diffraction data in D,O Ringer’s solution. The phases for the first eight orders (n, 0, 0, 0, 0, T, 0, 0) were obtained by Fourier transformation of the lipid bilayer model shown in Figure 12. This method of phasing relies on the assumption that the lipid bilayer dominates the phases, since the greatest contrast in the structure in D,O is between the water and hydrocarbon regions. The reduced extradisc neutron scattering density compared with the density in the intro&c space is preserved in the 50 d and 38 A resolution Fourier syntheses in D,O Ringer’s solution. It should be stressed that by using phases calculated from the lipid bilayer model in which the intradisc and extradisc neutron scattering densities are equal, the experimental Fourier syntheses are biased toward symmetric structures. Therefore, the retention of the asymmetry in these higher resolution Fouriers strongly supports the existence of asymmetry between the intradisc and extradisc neutron scattering density levels. (d) .4 model-building approach to interpret the asymmetry neutron scattering density pro$les

of the

There are three possible explanations for the lower neutron scattering density level in the extradisc aqueous space compared with the intradisc space : (1) that the asymmetry is due to rhodopsin extending into the extradisc region; (2) that the asymmetry is not due to rhodopsin but to other non-aqueous, hydrogenated material; and (3) that the asymmetry is due both to rhodopsin and other non-aqueous hydrogenated material. A model-building approach was used to examine these three possible interpretations.

M

340

YE A c: E R E !l’ A I,.

The following parameters defined the models : (1) the anhydrous molecular volume? of rhodopsin calculated from its amino acid composition (Robinson et al.: 1972 ; Heitzmann, 1972) and the partial specific volumes of amino acid residues (Cohn $ Edsall, 1943) is 47,000 A3; (2) the volume of two average rod outer segment hydrocarbon chains is 982 A” (see section (b), above); (3) the volume of the average lipid headgroup is 640 A3 with a 64 A” cross-sectional area (Luzzati, 1968: Johnson Pt al., 1971; Engelman, 1971; Demel et al., 1972) and a 10 A length (Engelman, 1971); (4) the lipid to rhodopsin stoichiometry is 60 (see section (b), above) ; (5) the neutron scattering densities of water and rhodopsin are given by equations (1) and (2), respectively, and the scattering densities for the lipid headgroup and hydrocarbon chains are shown in Figure 1; (6) the disc membrane lipids are packed in 50 A thick lipid bilayers as shown in Figure 12, with intradisc and extradisc aqueous spaces of 36 A and 160 A, respectively; (7) the region of the lipid bilayers distorted by the presence of rhodopsin has the same neutron scattering density as the undistorted bilayer; (8) the rhodopsin molecule was assumed to be a cylinder with its axis perpendicular to the plane of the disc membrane. Three general shapes were considered: elongated with a length of 80 A and a diameter of 27 A (model a), roughly symmetric with a length of 40 A and a diameter of 38 A (model b) and flattened with a length of 20 A and a diameter of 54 A (model c) Rhodopsin molecules were placed centrosymmetrically at different positions in the 300 A unit cell, and the one-dimensional projection of the neutron scattering density was determined. The Fourier transform of the model scattering density profile was calculated by equation (3), and the Fourier synthesis by equation (4). A qualitative assessment of the model was made by visually comparing the calculated model Fourier synthesis to the experimental Fourier synthesis. The calculated structure factor amplitudes, Fcalc(h), were scaled to the observed structure factor amplitudes, F,,,(h), by setting ,z, (Fcadh))2 The experimental Fourier scattering density scale by calculated from the model. on the choice of phases was

= 5 (Fobs(4)2. h=l

synthesis could then be placed on an absolute neutron including Fcslc(0) in equation (4) and using the phases A quantitative assessment of the model that did not rely made by calculating the residual, R :

where

K2 = 2 VLc@N2/~ (Fo&))2. h=l h=l The results of these calculations are summarized in Table 3. The distance values in the horizontal direction indicate the position of the centroid of the rhodopsin molecule with respect to the center of the intradisc space. In the vertical direction the neutron scattering density of the cytoplasmic extradisc aqueous space is varied from 6.35 x

NEUTRON

DIFFRACTION

OF INTACT

RETINAS

341

10-l* cm/A3, corresponding to pure D,O Ringer’s solution, to lower values, corresponding to increasing amounts of non-rhodopsin hydrogenated material in the extradisc region. The contours in Table 3 surround the regions with R values less than 0.20 and indicate the models in best agreement between the observed and calculated structure factor amplitudes. Two conclusions can be drawn from Table 3. (1) The best models have neutron scattering density levels in the extradisc space less than pure Da0 Ringer’s solution, suggesting that a signi$cant amount of non-rhodopsin hydrogenated material resides in the extradisc region. (2) The models in best agreement with the experimental structure factor amplitudes orient rhodopsin asymmetrically on the cytoplasmic face of the disc mem,brane. Model calculations using the structure factors in SO%, 60%, 40% and 30% D,O also supported these conclusions. For models a and b, in which rhodopsin has a length of 80 A and 40 A, respectively, the best models are those in which rhodopsin extends into the extradisc aqueous space. These structures are consistent with the interpretation of freeze-fracture electron micrographs of rod outer segments (Corless et al.. 1976) and the accessibility of rhodopsin to labelling and proteolysis by water-soluble macromolecular probes (Steinemann & Stryer, 1973; Yariv et al., 1974; Trayhurn et al., 1974a,b; Saari, 1974; Van Breugel et al., 1975: Pober & Stryer, 1975; Pober et al., 1978). For model c, in which rhodopsin is a flattened cylinder with a length of 20 A, there are two classes of structures in good agreement with the experimental data; however, the structures are implausible. In the first class, rhodopsin is buried in the lipid bilayer on the cytoplasmic face of the disc membrane. These structures are not consistent with the accessibility of rhodopsin to labelling and proteolysis. The second class of structures are unreasonable, since they position rhodopsin in the extradisc space, completely detached from the disc membrane. Model a is in best agreement with the evidence that rhodopsin has an elongated shape (Wu & Stryer, 1972; Yeager, 1975a; Sardet et al., 1976; Osborne et al., 1978; Pober et al., 1978). (e) Non-rho&p&n

extradisc solids

The exact amount of material one must place in the extradisc space in order to account for the data is strongly model dependent, and the errors in the data allow a wide range of choice. Nevertheless, the data do imply the existence of a significant amount of material in that space. If there is a lot of material in the extradisc space, it may be detectable by X-ray diffraction. Model a refined to fit the neutron diffraction data in section (d), above, suggests that the phases of the X-ray structure factors for orders 1 to 10 should be -, -, +, +, f, -, -, -, -, +, respectively. These can be compared with the experimental values for orders 1 to 10 of 57, -10, 35, 59, -47, -143, -143, -69, 69, 50 (Webb, 1972). The phases calculated from the model are identical to the experimental phases of the strongest reflections, except for the first reflection, which has a calculated phase of n instead of 0. The X-ray Fourier synthesis calculated using experimental structure factors and a phase of - 1 for the first reflection (Yeager, 1978) is still dominated by the electron-dense phosphate headgroups, but the extradisc X-ray scattering density level is strikingly higher than the density level in the intradisc space. Since protein and carbohydrate have X-ray scattering densit)ies larger than water, this increased density level in the extradisc space may

0.35 0.32 0.29 0.25 0.27 0.20 0.22 0.25 0.29 0.33 0.37

0.35 0.32 0.29 0.26 0.23 0.24 0.27 0.31 0.35 0.38 0.39

0.38 0.36 0.33 0.31 0.27 0.24 0.27 0.31 0.35 0.40 0.44

b. 6.35 6.05 5.75 5.45 5.15 4.85 4.55 4.25 3.95 3.65 3.35

0.36 0.35 0.33 0.31 0.29 0.27 0.24 0.21 0.24 0.28 0.32

0.23 0.25 0.29 0.33 0.36

50

0.21 0.22 0.27 0.31 0.36

40 0.34

30 0.39

20

Lipid bilayer

0.30 0.27 0.23 0.20 0.15 0.13 O-14 0.19 0.24 0.29 0.34

0.25 0.28 0.32 0.35 0.38

0.28

60

0.26 0.23 0.20 0.17 0.14 0.14 0.16 0.21 0.26 0.30 0.35

0.23 0.27 0.30 0.33 0.36

0.24

70

Distance (A)

0.25 0.23 0.22 0.20 0.18 0.19 0.21 0.24 0.27 0.31 0.38

0.22 0.24 0.27 0.31 0.35

0.22

80

0.26 0.29 0.31 0.34 0.36

0.24

100

0.24 0.21 0.21 IO.19 0.19 0.18 0.16 0.18 0.16 0.19 0.18 0.20 0.21 0.24 0.25 0.27 0.29 0.30 0.32 0.33 0.36 0.37

0.24 0.27 0.29 0.31 0.35

0.23

90

0.30 0.28 0.27 0.26 0.26 0.27 0.29 0.31 0.34 0.36 0.37

0.26 0.29 0.32 0.34 0.37

0.23

110

the experimental structure factor amplitudes in D,O Ringer’s calculated from disc membrane models

a. 6.35 6.05 5.75 5.45 5.15 4.85 4.55 4.25 3.95 3.65 3.35

center of intradisc space $ 0 10

values comparing

Cytoplasmic neutron scattering density (x 1014cm/Aq

Residual

TABLE 3

0.35 0.36 0.37 0.39 0.38 0.37 0.37 0.36 0.36 0.37 0.38

120

solution

0.44 0.43 0.43 0.42 0.41 0.40 0.39 0.37 0.36 0.36 0.38

130

center of extradisc spacr 4 140 150

to structure factors

0.21

0.43

0.33

0.30

3.35

0.49

0.44 0.48

0.40

0.30 0.35

0.26

0.22

0.23

0.26

0.28

0.38

0.44

0.44 0.48

0.41

0.33 0.37

0.30

0.31

0.34

0.36

0.29

0.32

0.21 0.27

0.21

0.25

0.31

Rhodopsin Length (.A) 40 80 20

0.36

0.32

0.30

0.30

0.31

0.33

0.35

0.31

0.27

. .

0.36

0.31

0.27

..

0.17

0.17 m

0.15 0.15

0.20

0.18 0.17

x

0.26

0.36 0.38

0.38 0.38

0.37 0.37

0.39

0.30 0.32 0.34

0.37

0.29

0.33 0.35

0.33 0.31

0.27 0.29 0.31

0.32 0.32 0.32

0.36 0.38

0.37 0.36

0.39

0.40

0.41

0.41

0.39 0.41

0.38

0.32 0.36

0.36 0.33

0.38

0.40

0.43

0.45

0.46 0.47

0.45

27 38 54

dimensions Diameter

(A)

to the plane of the disc membrane. The distance values in the horizontal with respect to the center of thp intradisc space. In the vertical direction The contours surround the regions with residual values less than 0.20 and structure factor amplitudes. See the text for details.

0.35

0.32 0.35

0,28

0.27

0.29

0.31

0.29

0.24

0.27

0.30

Cylindrical rhodopsin molecules are oriented with their axes perpendicular direction indicate the position of the ccntroid of t’he rhodopsin molecules the neutron scattering density of the cytoplasmic extradisc space is varied. indicate the models in best agreement between the observed and calculated

3.65

0.24 0.28

0.39 0.35

3.95

4.55 4.25

0.24 0.20

0.48 0.45

4.85

0.30

0.27

0.52

0.50

0.32

0.53

5.15

0.35

6.05 5.75

5.45

0.37

0.55

0.54

o. 6.35

hl. PEAGER

344

ET

ill,

represent non-aqueous material residing in this region, consistent with our interpretation of the asymmetry of the neutron scattering density profiles. Osmotic shrinking and swelling data (Blaurock & Wilkins, 1972; Blaurock, 1972) have been used to assign a phase of +I for the first reflection. However, two of the four best phasing combinations in the analysis reported by Schwartz et al. (1975) had phases of -1 for the first reflection. The concentration of solids in the rod outer segment obtained by refractometry is 44110 g/100 cm3 (Sidman, 1957; Blaurock & Wilkins, 1969; Webb, 1972). Based on the rod outer segment composition and size stated above, this solids concentration predicts that the extradisc space has a solids content of 30%. The solids content of the extradisc space based on birefringence studies (Liebman et al., 1974) is 16%. Liebman (personal communication) has stated that the birefringence data are consistent with an extradisc solids concentration ranging from IO to 30%. Furthermore, a source of uncertainty in the measurement of the index of refraction and the static birefringence is whether the isolated rod outer segments were osmotically intact. Although these measurements with visible light predict a solids concentration in the extradisc region lower than our data indicate, the conclusion remains that there appears to be a large amount of solute material in the extradisc space. (f) Comparison

with the data ofSa;bil

et al. (1976)

Table 4 compares our neutron diffraction experiments on intact retinas with similar experiments on isolated rod outer segments oriented in a magnetic field (Chabre et al., 1975; Saibil et al., 1976). The diffraction patterns in D,O Ringer’s solution are qualitatively very similar. The first four orders dominate the pattern with the observed intensity of reflection 1 strongest, followed by reflections 3, 2 and 4, respectively. The Fourier syntheses at 75 A resolution are also very similar, in that they both display low density troughs corresponding to the lipid bilayers and high density regions in the aqueous intradisc and extradisc spaces. The notable difference between the Fourier syntheses is that the neutron scattering density level in the extradisc region is lower in our Fourier synthesis, thereby predicting a greater amount of nonaqueous material in the extradisc space. Both analyses, however, lead to the conclusion that there is more dissolved material in the extradisc space than in the intradisc space. Since the Fourier syntheses were calculated using t’he same phases, these density differences in the extradisc region must be due to differences in the structure factor amplitudes. Table 4 shows that the amplitudes for orders 2, 3 and 4 are larger in our experiment, causing the lower density level in the extradisc space in our Fourier synthesis. We obtained a somewhat lower contrast-match point for the center of the lipid bilayer than did Saibil et al. (1976). In addition, we assume a higher protein to lipid ratio than they did, leading us to predict a smaller fraction of rhodopsin in the membrane interior. We are left with a difference between our results and those of Saibil et al. (1976) which must be examined. An assumption in our analysis is that the background is appropriately measured in the direction perpendicular to the rod axes. If we make the most extreme alternative assumption, that the background is represented by the observed minimum between diffraction orders, the discrepancy in the amount of hydrogenated material in the interdisc space is reduced by a factor of two, but a

NEUTRON

DIFFRACTION

OF TABLE

Comparison

RETINAS

Ssmplc

work

Intact of rods

Detector

frog

345

4

et al. (1976)

of OUT results with those obtained by Saibil This

Orientation

INTACT

Saibil

retinas

et nl. (1976)

Isolated

Natural

Magnetic

Soller slit step-scanning diffractometer

Area

frog rod outer

segments

field

detector

(AX/X

= 0.08)

(Ail/h = 0.025) Time of data collection orders 1 to 4

for

Period

- 70 min

29515

Structure in D,O

a

- 10 min

295 A

factor amplitudes Ringer’s solution F(l)

F(2) F(3) F(4) F(5) F(6) F(7) Phases

t 353*25 2461 I9 434113 383519 186&21 17618

$ 353 229 383 235 126 136 85

n, 0, 0, 0

a, 0, 0, 0, ?, 37, 57 -36

Width

of intradisc

space

-36

Width

of extradisc

space

-160

Center-to-center separation of lipid bilayers

-88

I!

-155

A

.!i A

-9oA

A

Neutron scattering density (x 10’4 cm/A3) (origin at center of intradisc space) P(0 13) p(44-45 A) p(115 A) p(147-150 A) Contrast-match value of lipid bilayers o/o Rhodopsin region

t 1 D,O §

at center

in hydrocarbon

6.35 0.05 ho.29 4.7 5.9

5 6.01 - 7.06 0.48 5.63 - 6.65 5-99 ~ 7,06

9 +4%

D,O

15%

- 20%

maximum

50 + 100% lower limit: pure Ringer’s in aqueous spaces upper limit : 20% carbohydrate in aqueous spaces

D,O

Errors are 1 standard deviation from the mean. Structure factors in D,O Ringer’s solution were obtained by extrapolation plots to 100% DzO. Range of densities is due to different phase choices for reflection 5.

of F(h) versus o/o

346

11. YE i\ GE R E T rl I;.

may be found in rcccnt ol,significant difference remains. X possible esplanat’ion serrations showing that a number of soluble proteins, amounting to 30% of t’ht> t&al prot’ein of the rod out)er segmentj. may be lost if the outer membranr of the rod is not 1979; Schnctkamp et nl., 197!l). intact, (Adams et ul., 1978; Godchaux $ Zimmerman, ,4s a proportion of the rods in arlp isolate will be made permeable by isolation procedures, the preparation of isolated rods used by Saibil et al. (1976) may contain less non-rhodopsin protein than the intact retinas used in our work. Combining the earlier observations with ours leads to bhe suggestion that, the non-rhodopsin proteins are located in the interdisc space.

1Ve tllank MS Helen Saibil, Dr Marc Chabre and Dr David Worcester for communicating and Dr Allan Oseroff for Ilelpfld comment,s on this tllcir results to 11s beforcl publicatiolr. manuscript. We are indebted to Mr Gerald .Johnson and Mr Ed Caruso for expert technical Enson for inr-aluable assistjallce in data processing and mamassistance and MS Shnrcll script preparat,ion. WC also thank Dr Joseph Corless and Dr Paul Licbman for helpful discussions regarding tllc: clremicnl composition of rod out,cr segmwt,s. This work was supported by grants from tllrx National Eye Instit,utc (b>Y-01070 and EY-02005) and thr National Science Folmtlatiou (BMS 75.03809). Tl~c rrscarch at Brookharen National Laboratory was carried out under tllc auspices of tile United States Atomic Energy Commission and tllc United States Energy Researcll and Devolopmcnt Administration. One author (M. Y.) is gratefnl for support from the Medical Scientist Training Program (GM02044) and this work is part of tlrr PllD dissertat,ion of this author.

REFERENCES Adams, A. J ., Tanaka, M. & Shichi, H. (1978). Exp. Eye lies. 27, 595WX2. Arndt, U. W. &Willis, B. T. M. (1966). Single Crystal Difractometry, Cambridge University Press, London. Arnott, S. (1965). Polymer, 6, 478-486. Bacon, G. E. (1962). Xezctron DL;ifSractio~z, Oxford Universit,y Press, London. Blasie, J. K. (1972). Biophys. J. 12, 191-196. Blanrock, A. E. (1972). Advan. Eq. Med. Biol. 24, 53-55. Blaurock, A. E. & Wilkins, M. H. F. (1969). Nature (London), 223, 906--909. Blaurock, A. E. & Wilkins, M. H. F. (1972). lZature (London), 236, 313-314. Bownds, D. & Brodie, A. E. (1975). J. Gen. Physiol. 66, 407-411. Bragg, W. L. & Perntz, M. F. (1952). Acta Crystallogr. 5, 277-283. Brierley, G. P., Floischman, D., Hugllos, S. D., Hunter, G. R. & McConnell, D. G. (1968). Biochim. Biophys. Acta, 163, 117~ -120. Caspar, D. 1,. D. & Kirsclrner, D. A. (1971). Nature Mew Biol. 231, 46-52. Caspar, D. L. D. & Phillips, W. C. (1975). Brookhaven Symp. Biol. 27, VII-107. Cella, R. J., Lee, B. 85 Hughes, R. kS. (1970). Acta Crystallogr. sect. A, 26, 118-124. Chabrc, M. (1975). Rio&m. Biophys. ilcta, 382, 322--335. Chabrc, M. & Cavaggioni, A. (1973). Xature New Biol. 244, 118~121. Chabrr, M. & Cavaggioni, A. (1975). Biochim. Biophys. Acta, 382, 336--343. Cllabre, M., Saibil, H. 8: Worcester, I). L. (1975). Brookhaven, Symp. Biol. 27, 111-77. Clark, A. W. & I3ranton, D. (1968). Z. Zellforsch. 91, 586-594. Cohen, A. I. (1971). J. Cell Biol. 48, 547-565. Cohn, E. J. & Edsall, J. T. (1943). Proteins, Am&o Acids and Peptides, Reinhold Publishing Corp., New York. Corless, J. M. (1972). Nature (London.), 237, 229-231. Corless, J. M., Cobbs, W. H., Costello, M. J. & Robertson, J. D. (1976). Ezp. Eye Res. 23, 295-307. Daemen, F. J. M. (1973). Biochim. Biophys. Acta, 300, 255-265.

NEUTRON

DIFFRACTION

OF

INTACT

RETINAS

347

Daemen, F. J. M., DeGrip, W. J. & Jansen, P. A. A. (1972). Biochim. Biophys. Acta, 271, 419-428. Deas, H. D. (1952). Acta CrystaZZogr. 5, 542-546. Demel, R. A., Geurts Van Kessel, W. S. M. & Van Deenen, L. L. M. (1972). Biochim. Biophys. Acta, 266, 26-40. DoRobortis, E. & Lasansky, A. (1961). The Structure of the Eye (Smelser, G. K., ed.), p. 29, Academic Press, New York. Dowling, J. E. (1967). Molecular Organization and Biological Function (Allen, J. M., ed.). ch. 7, p. 186, Harper and Row, New York. Engelman, D. M. (1971). J. Mol. BioZ. 58, 153-165. Franklin, R. E. & Gosling, R. G. (1953). A& Crystallogr. 6, 678-685. Gibbs, E. L. (1963). Lab. Animal Care, 13, 781-786. Godchanx, W. & Zimmerman, W. F. (1979). Exp. Eye Res. 28, 483-490. Gras, W. J. & Worthington, C. R. (1969). Proc. Nat. Acad. Sci., U.S.A. 63, 233-238. Hapills, W. A. (1972). Annu. Rev. Biophys. Bioeng. 1, 131.-137. Harrison, S. C. (1969). J. Mol. BioZ. 42, 457-483. He&t,, S., Shlaer, S. S. & Pirennc, M. H. (1942). J. Gen. l’hysiol. 25, 819-827. Heitzmann, H. (1972). Ii’ature Ykrew BioZ. 235, 114. Keller, J., Ostwald, T. J. & Bok, D. (1971). ,J. Cell Biol. 48, 633-649. Holmes, K. C. & Barrington Leigh, J. (1974). Acta Crystallogr. sect. A, 30, 635-638. James, R. W. (1965). The Optical Pr’rinciples of the Diffruction of X-rays, Cornell University Press, Ithaca. Johnson, S. M., Bangham, A. D., Hill, M. W. & Kern, E. D. (1971). Biochim. Biophys. Acta, 233, 820-826. Kirschnor, D. A. (1971). Ph.D. thesis, Harvard University, Cambridge. Kirschner, D. A. (1974). Spectroscopy in Biology and Chemistry Neutron, X-ray, Laser, ch. 6, p. 203, Academic Press, Inc., New York. Kirschner, D. A. & Caspar, D. L. D. (1972). Ann. N.I-. Acad. Sci. 195, 309-316. Kirsclmer, D. A., Caspar, D. L. D., Schoenborn, B. P. & Nunes, A. C. (1975). Brookhaven 8ymp. BioZ. 27, 111-68. Korenbrot, J. I., Brown, D. T. & Cone, R. A. (1973). ,J. Cell BioZ. 56, 389-398. Levine, Y. K. (1973). Prog. Surf. Sci. 3, 279-284. Lewis, M. S., Krieg, L. C. & Kirk, W. D. (1974). Ezp. Eye Res. 18, 29-41. Liebman, P. A. (1975). In Photoreceptor Optics (Snyder, A. W. & Menzel, R., eds), p. 199, Springer Verlag, New York. Liebman, P. iz., .Jagger, W. S., Kaplan, M. W. & Barpoot, F. G. (1974). Nature (London), 251, 31-36. Luzzu.ti, V. (1968). In Biological Membranes (Chapman, D., rd.), p. 71, Academic Press. New York. Moor<,, P. B. (1975). Brookhaven Symp. BioZ. 27, VII-67. Moon,, P. B., Ellgelman, D. M. 8: Schoenborn, B. P. (1974). Proc. Nat. Acad. Sci., U.S.A. 71, 172-176. Nate. G. W., Cnlley, D. D., Emmons, M. B., Gibbs, E. L., Hutchison, V. H. & MeKinney, 11. G. ( 1974 ). Guidelines for the Breeding, Care amd Management of Laboratory Animals ; Amphibians, National Academy of Sciences-National Research Council, Washington. Nilssorl. S. E. G. (1965). J. Ultrastruct. Res. 12, 207-214. Nun~s, A. C. (1973). Nucl. In~strum. Methods, 108, 189-l!)O. Osborne, H. B., Sardet, C., Michel-Villaz, M. & Cllabrc,, M. (1978). J. Mol. BioZ. 123, 177-206. Papermaster, D. S. & Dreyer, W. J. (1974). Biochemistry, 13, 2438-2444. Pobcr, J. S. & Stryer, L. (1975). J. Mol. Biol. 95, 477.-481. Pober, J. S., Iwanij, V., Reich, E. & Stryer, L. (1978). Biochemistry, 17, 2163-2169. Robinson, W. E., Gordon-Walker, A. & Bowtrds, D. (1972). Nature New Biol. 235, 112-114. Saari, ,I. C. (1974). J. Cell BioZ. 63, 480-491. Saibil, H.,, Chabre, M. & Worcester, D. (1976). Nature (London), 262, 266-271.

348

M. YEAGER

ET

AL.

Sardet, C., Tardieu, A. & Luzzati, V. (1976). J. &fol. Biol. 105, 383-407. Saxena, A. M. & Sclloenhorn, H. I’. (1977). dcta Crystdlogr. sect. a, 33, 805-813. Schmatz, W., Springer, T., Scheltcn, ,J. & Ibel, K. (1074). J. ilppl. CrystaElogs. 7, 96- 103. Schnetkamp, P. P. M., Klompmakers, A. A. & Daemen, h’. J. M. ( 197CJ). Biochim. Biophys. Acta, 552, 379%389. Schoenborn, 13. P. (1976). Biochim. Biophys. Acta. 457, 41-55. Schwartz, S., Cain, J. E., Drat,z, E. .4. & Blasie, J. K. (1975). Biophys. -7. 15, 1201 -1211. Sidman, R. L. (1957). J. Biophys. Biochem. Cyt. 3, 15.-29. SjGstrand, F. S. (1959). Ergeh. Biol. 21, 128-140. Steinemann, A. & Stryer, L. (1973). Biochemistry, 12, 1499-1502. Stubbs, G. J. (1974). Acta Crystallogr. sect. R, 30, 639-645. Tomita,, T. (1970). Quart. Rev. Biophys. 3, 179-180. Traube, J. (1899). Sam&. Chewa. Chem. Tech. Vort. 4, 255-263. Trayhurn, P., Mandel, P. & Virmaux, N. (1974a). Ezp. Eye Res. 19, 259-267. Trayhurn, P., Mandel, P. & Virmaux, N. (19746). FEBS Letters, 38, 351-358. Van Breugel, P. J. G. M., Daemen, F.