The structural study of membrane proteins by electron crystallography

Adv.

THE STRUCTURAL PROTEINS BY ELECTRON

Biophys.,

Vol.

35, pp. 25-80

(1998)

STUDY OF MEMBRANE CRYSTALLOGRAPHY

YOSHINORI FUJIYOSHI Department of Biophysics, Faculty of Science, Kyoto versity, Sakyo-Ku, Kyoto 606-8502, Japan

Uni-

Over the last few years, structural biology has grown into one of the most important fields in modern biology. While molecular biology is continually revealing many new proteins indispensable for life on earth, only their atomic structure can help us to really understand how these proteins accomplish their important biological functions. The information which can be drawn from a structure analysis of a protein strongly depends on the resolution achieved in the study. The higher the resolution of the determined structure can give the more information on the relationship between structure and function of a biological macromolecule. Electron microscopy provides useful biological information over a wide range of resolutions and this is summarized in Table I. However, the ultimate goal of every structural study of a protein will always be to determine its atomic structure, and from X-ray crystallographic and electron microscopic studies we know that a resolution of about 3.0 A is sufficient to interpret structural detail on the level of amino acid side chains. The foundation for crystallography was laid in 1913 by W. H. Bragg and his son W. L. Bragg (I). Due to methodological innovations, especially the development of the isomorphous replacement technique pioneered by Perutz (Z), dramatic progress in recombinant techniques in molecular biology, and the massive progression 25

26

Y.

TABLE I Mark Points Resolution 20.0 7.0 3.5 3.0 2.5 1.5 1.0

of Resolution (A)

in Structural Structural

Information

features

Overall

shape

of Proteins

observable

of the

on a good

of strong intensity with some ambiguities partly resolved

Side

resolved;

well

Atoms

Atoms located to about f 0.01 Anisotropic structure of atomic Atoms

located

to about

map

molecule

Helices as rods The main chain The side chains chain

FUJIYOSHI

located

to about

f 0.04

potential

f 0.004

of computer technology, X-ray crystallography is now well established and has become the most powerful technique to determine the atomic structure of most biological molecules. However, membrane proteins, which are essential for the exchange of information, materials, and energy between cells and their environment, have so far largely eluded structural analysis by X-ray crystallography. Despite the innovative work by Deisenhofer et al. (3), who determined the first atomic structure of an integral membrane protein, a detergent-solubilized bacterial photosynthetic reaction center, by X-ray crystallography, the last decade produced atomic structures of only about 10 kinds of membrane proteins. This is due to the difficulty in obtaining large, well-ordered three-dimensional (3 D) crystals from membrane proteins, a prerequisite for X-ray crystallographic structure determination. It has for some time been speculated that electrons, which interact with matter about 10,000 times stronger than X-rays, could be used for the structural analyses of regular arrays formed by a single layer of protein molecules, so-called two-dimensional (2D) crystals, at atomic resolution. If this could routinely be achieved in practice, one might then imagine that this electron crystallography could strongly complement X-ray crystallography in the determination of atomic structures, especially in the case of membrane proteins which form 2D crystals much more readily than 3D crystals. Moreover, electrons can be focused and lenses have already been developed and provide the means to construct a microscope which directly produces a magnified image of the structure under investigation. Electron microscopy plays an important role for structural

STRUCTURAL

STUDY

OF

MEMBRANE

PROTEINS

27

studies in various fields of basic sciences such as physics, chemistry, and biology, because it can provide structural information from a very small region of a specimen and especially because it does not require a periodic structure of the specimen. The high sensitivity of the electron microscope is due to the strong interaction of electrons with all matter. However, this strong interaction also results in two inevitable problems, namely, damage to the specimen by the irradiating electrons and the necessity of a vacuum for the electron beam. In 1973, when I began to work in electron microscopy, the performance of electron microscopes was generally considered to be sufficient to yield atomic resolution for most inorganic materials which are stable in the electron beam. However, it was uncertain whether electron microscopy was actually suitable for the direct visualization of atoms because irradiation with high energy electrons was thought to lead to vibrations of the atoms in the specimen and thus would cause images recorded by an electron microscope to be blurred. One pessimistic theory had even argued that multiple scattering of the electrons by the specimen (dynamical scattering) would distort the electron microscopic images to a degree that it would prevent their interpretation. Therefore, I still remember the knee-shaking sensation when I saw our initial successful micrograph of the molecular structure of chlorinated copper phthalocyanine. This image clearly resolved for the first time individual atoms (Fig. 1) (4) and convinced me of the potential of electron microscopy for structure analysis at atomic resolution. However, when we subsequently tried to use electron microscopy for the analysis of organic molecules, we encountered severe difficulties. Organic molecules appeared to be too sensitive to withstand the electron beam without significant damage even for a very short time, although we had developed a minimal dose imaging system to record high-resolution images without excess illumination of electrons (5). The principle of the minimal dose imaging system and an image recorded in this way are shown in Figs. 2 and 3, respectively. In the case of biological molecules, the resolution of an image is not limited by the resolution of the microscope, which nowadays achieves a resolution of better than 2.0 A, but by the damage to the specimen caused by the electron beam. Because it was found that this irradiation damage can be significantly reduced by lowering the temperature of the specimen and because we wanted to realize elec-

28

Y. FUJIYOSHI

Fig. 1. Electron microscopical image of chlorinated molecules. The chemical structure of this compound high-resolution image. This micrograph was the first croscopy is capable of visualizing individual atoms in

copper phthalocyanine is superimposed on the proof that electron mian organic molecule.

tron microscopic structure analysis of proteins to a resolution of 3.0 A or better, we set out to develop an electron cryo-microscope equipped with a specimen stage cooled by liquid helium. The 3rd version of such an electron cryo-microscope, which is equipped with a superfluid helium stage as well as a cryo-transfer device for efficient specimen exchange, was installed in 1994 in the International Institute for Advanced Research at the central research laboratories of Matsushita, Ltd. and displays an instrumental resolution of better than 2.0 A. An electron microscopic image yields only projection (2D) information. Therefore, to produce a complete 3D density map of a protein based on electron crystallography, many images of untilted and tilted specimens must be recorded, processed, and combined. This approach was used for the first time by Henderson and Unwin who determined the 3D structure of bacteriorhodopsin (bR) to a

STRUCTUR.4L

STUDY

OF

MEMBRANE

29

PROTEINS

7

Fig. 2. Schematic ing the three modes Point A is located screen. In the new, no longer required.

diagrams of the ‘minimum dose system’ (MDS) illustratof operation: search mode, focus mode, and photo mode. on the main axis and point B on an additional focusing more sophisticated MDS, the additional focusing screen is The area irradiated by the electron beam is shaded.

resolution of 7.0 A (6). After publication of this 7 A map, another 15 years were needed for further development of the methodology for electron crystallographic structure determination before Henderson and cy-workers could present the structure of bR at a resolution of 3.5 A (7). Guided by the amino acid sequence of bR, the 30 density map made it possible to identify the coordinates for individual atoms. Because bR forms very stable and well-ordered 2D crystals in situ, so-called the purple membranes in the plasmamembrane of Hu~o~ac~e~iu~ s~l~na~~u~, this membrane protein has proven an ideal specimen for electron crystallographic structure determination, and thus the progress of this technique is closely linked to the study of the structure of bR. Therefore, in this review I will primarily refer to bR as an example to discuss the current state and the future prospects of electron crystallography. Another example of the successful determination of an atomic structure by electron crystallography is the plant light-harvesting

30

Y.

Fig. 3. High-resolution charge transfer compound, image model

FUJIYOSHI

image of Ag-TCNQ complex, which is an organic recorded with a 200 kV electron microscope. An

of this compound could structure is superimposed

only be obtained on the image.

using

MDS.

A hypothetical

complex II (LHC-II). The atomic model for this membrane protein was published in 1994 by Kiihlbrandt et al. (8). So far, only these two membrane proteins, bR and LHC-II, have been analyzed to atomic resolution by high-resolution electron microscopy. However, many more membrane proteins have already been crystallized in two dimensions and the 2D crystals of several of these proteins will eventually allow determination of their atomic structure by electron crystallography. Tubular crystals, a different type of 2D crystals, enabled Nigel Unwin to analyze the structure of the nicotinic acetylcholine receptor to 9 A resolution in its closed state (9). Moreover, timeresolved electron microscopy performed on the tubular crystals yielded the structure of the open channel providing invaluable insights into the relationship between structure and function of this

STRUCTURAL

STUDY

OF MEMBRANE

31

PROTEINS

ligand-gated transmembrane channel (10). Although tubular crystals constitute a very important field in electron crystallography because many proteins have already been crystallized into tubular crystals, in this review I will only discuss electron crystallography of 2D crystals. Unwin and his collaborators are currently using the tubular crystals to collect electron microscopic data for a 3D map of the acetylcholine receptor at near atomic resolution and therefore we should await the results of their exciting work. Because many other membrane proteins have been crystallized in two dimensions and their structures have been analyzed at a relatively low resolution, I will present aquaporin-1 (AQPl ) (II), a mammalian water channel, as an example of a membrane protein where the 3D structure has so far only been determined to a medium resolution of 6 A.

I.

DEVELOPMENT

OF A HIGH-RESOLUTION

ELECTRON

CRYO-MICRO-

SCOPE

The resolution in an image of a biological macromolecule is usually limited to a value much worse than 3 .O A. This limitation is not due to the resolution of the instrument but is caused by damage of the specimen by the electron beam. Thus, irradiation damage is the most serious problem for electron microscopical structure analysis of a biological specimen at atomic resolution.

1.

Radiation

Damage

The mechanism of radiation damage of an organic molecule is thought to be a two step process. First, deposition of energy on the specimen by inelastic scattering events causes breakage of chemical bonds and subsequently the produced radicals, ions and/or molecular fragments diffuse away, leading to a complete collapse of the initial structure. After many attempts to find solutions to overcome this problem, only cooling of the specimen appears to reduce the effects of radiation damage by limiting the diffusion of the newly formed components. The irradiation damage to both tRNA and catalase crystals was found to be reduced to about 1 /lO and l/20 of the value at room temperature when the specimen was cooled below 20K and 8K, respectively (Fig. 4). At very low temperatures, the lattice heat capacity of all materials becomes almost zero, because its decrease is

32

Y. FUJIYOSHI

;m>,o;,; ,__ -_ ____ ______ _____-__-_------.-.--__ --------------. Cotolose 0.1

.

' 1

.

' 2

' 10

.

' 20

5.5A 8

' 3

(4 '

1 RNA9A 0.1

'

*

' 30

30) 0 4

(6 -

' 5

'

' 6

'

' 7

.

c 8

4 9

10) ' 40

'

1 50

'

' 60

m

' 70

'

1 60

'

1 90

Fig. 4. Irradiation damage of catalase and yeast phenylalanine tRNA crystals. The fading of the intensities of the diffraction spots (4, 30) for the catalase and (6, 10) for the tRNA crystal is plotted as a function of the accumulated electron dose. These plots demonstrate that a further reduction of the specimen temperature from liquid nitrogen to liquid helium temperature results in an improvement of the cryo-protection factor by a factor of 2 or more.

directly proportional to (T/100)3, in which T is the absolute temperature. Furthermore, the heat conductivity of non-metals such as carbon is dramatically reduced at very low temperatures. Due to these effects, the temperature of the specimen can be strongly increased in the area which is irradiated by electrons. Therefore, the actual temperature in an irradiated region of the specimen was estimated by recording the electron diffraction patterns of solid N2 and Ne, which indicated that the specimen temperature must have been lower than 20K and 8K, respectively, even under the condition used to record the electron diffraction patterns (Fig. Sa, b). The vapor pressures of various gases are shown in Fig. SC. From these studies, the cryoprotection factors in relation to the specimen temperature could be precisely determined and the results are summarized in Fig. 6. Upon illumination of a film of solid nitrogen covering an ice layer formed on a platinum coated holey carbon film, bubbles initially appeared only on the metallic frame which has a higher stopping power for the electron beam compared to the ice layer (Fig. 7a). This suggests that upon illumination the temperature of the

STRUCTURAL

STUDY

Fig.

5.

OF

Electron

MEMBRANE

diffraction

33

PROTEINS

patterns

of a solid

nitrogen

film

neon film (b) condensated on the surface of ice-embedded radiation damage measurements, various gases were introduced where they condensated on the surface of the specimen. The fraction rings from the solidified gases allowed the estimation temperature. Panel (c) depicts sure for various gases.

the temperature

dependence

(a) and

a solid

specimens. For into the stage presence of difof the specimen of the

vapor

pres-

34

Y.

FUJIYOSHI

Fig. 6. Plot of the cryoprotection factor as a function of the specimen temperature. The values were determined for catalase and tRNA crystals at specimen temperatures lower than lOOK, 20K, and 8K. The cryoprotection factor for room temperature is defined as 1.

metallic frame rises more than the temperature of the ice film, even though the two are illuminated in the same way. Moreover, in images of ice-embedded tRNA crystals (Fig. 7b), bubbles indicating beam damage were observed much more frequently in areas close to the thick gold frame (area B in Fig. 7b: higher temperature region) than in areas some distance from the edge of the frame (area A: lower temperature region). The image shown in Fig. 7 is, thus, a direct indication that lowering the temperature even below the liquid nitrogen temperature still helps to reduce the effects of radiation damage to biological molecules. Based on these observations, we decided to develop an electron cryo-microscope equipped with a specimen stage cooled by liquid helium. However, mechanical vibrations caused by the boiling of the coolant as well as specimen drift induced by temperature changes in the specimen stage presented difficulties which had to be overcome in the design of a high-resolution electron cryo-microscope.

STRUCTURAL

STUDY

OF MEMBRANE

PROTEISS

35

Fig. 7. Electron micrograph of a solid nitrogen film covering a film of amorphous ice. (a) Bubbles or holes in the nitrogen film (marked by arrowheads) appear first on the platinum coated frame where the stopping power for the electron beam is much higher and where the temperature increases more than in the open areas. (b) Image of a tRNA crystal embedded in amorphous ice. The lattice lines of the crystal can clearly be seen in area A but not in area B close to the gold coated frame of the holey carbon film. This is clear proof that lowering the temperature below the liquid nitrogen temperature still reduces the radiation damage of the tRNA crystal.

36

Y.

FUJIYOSHI

2. Helium-cooled Specimen Stage for High-resolution Electron Microscopy We earlier developed a superfluid helium stage which achieved a resolution of 2.6 A and a paper describing details of design and construction of that stage was published in 1991 (12). However, for structure analysis of biological molecules at atomic resolution, an instrument yielding a resolution better than 2.6 A would be highly beneficial, because biological molecules consist mainly of light atoms which have very small values for their atomic scattering factors in the high-resolution range. Thus, the deterioration of the contrast transfer function strongly affects the signal-to-noise (S/N) ratio of the Fourier components of the images, especially in the high resolution range. The achievable resolution in the structure analysis of biological molecules is therefore limited by the image quality (S/N ratio). Better phase information can only be obtained by evaluation of higher resolution images which, in turn, can only be recorded by a higher resolution electron cryo-microscope. Furthermore, resolutions higher than 2.0 A enable us to observe the lattice lines of gold crystals which are helpful for the calibration of the instrument. Therefore, the specimen stage and the pole pieces of the objective lens of our initial helium-cooled electron microscope were modified to reduce the spherical aberration of the lens from 2.6 mm to 1.9 mm at an acceleration voltage of 400 kV. These modifications improved Scherzer’s limit (13) of the 2nd version of our electron cryo-microscope from 2.6 A to 2.0 A. High-resolution structure analysis of proteins based on electron microscopy requires high-resolution images, including many images from highly tilted specimens. So far, the success rate in collecting high quality images from highly tilted 2D crystals is relatively low. Often the image quality is deteriorated by image drift and/or shift (very rapid drift) which are attributed to the electric charge-up of the irradiated area of the specimen. The low yield of high-resolution images forces us to spend many days at the microscope and to look at many specimens. Therefore, ease of operation of an electron cryo-microscope is a tremendous help in the tedious task of electron microscopic data collection and thus very important for the 3D structure analysis of biological molecules. The superfluid helium stage of our electron cryo-microscope enabled us to collect high resolution images at a stage temperature

STRUCTURAL

STUDY

OF

MEMBRANE

37

PROTEINS

PlJtiP

Capi

Fig.

8.

Principle

the specimen trols the flow be chosen tween the lent helium system.

1 I ary

of the

superfluid

helium

is connected to the helium of liquid helium to the pot.

for optimal helium tank flow

which

cooling.

The

helium

reservoir via a capillary The impedance of the

pot which capillary

cooling concan

performance. A heater is attached to the capillary beand the BeCu-wire to eliminate the problem of turbucaused

vibrations

in

a former

version

of the

cooling

of 1.5K and from 4.2K up to room temperature. However, as described previously (IZ), at a temperature of 1.5K we were confronted with a serious problem of high frequency vibration of the stage. The main cause for the stage vibration was overcome by cutting off the turbulent flow of liquid helium through the capillary by heating a part of the capillary just above the BeCu wire inside the capillary near the helium tank as shown in Fig. 8. This modification stabilized the stage at 1. SK and enabled us to record high-resolution images at this temperature. To collect images at 4.2K, the evacuation of the pot through the exhaust pipe had to be stopped but a fully charged pot of liquid helium was sufficient to work for 15 min; during this time high-resolution images could reproducibly be obtained. However, this method required time to manipulate two valves in order to cool down the pot and to allow the pressure to again rise to atmospheric pressure after evacuation of the pot. This inconvenience was resolved by optimizing the helium flow

38

Y.

FUJIYOSHI

through the capillary which enables us now to record high resolution images at 4.2K without further manipulation for up to 5 hr. A cross section through the 3rd version of our superfluid helium stage is shown in Photo 1. The cryo-stage consists of 4 major parts: the liquid nitrogen tank, the liquid helium tank, the 1 SK pot, and the specimen stage. The yellowish brown part is cooled by liquid nitrogen and kept below a temperature of about 95K. The specimen stage is connected to the liquid nitrogen tank by flexible copper braids. The green parts represent the helium tank which is kept at 4.2K. The blue and dark green areas depict the 1.5K pot and the specimen holder, respectively. The capillary and exhaust pipe are connected to the pot. Both pot and specimen holder are cooled down to 1.5K when the helium in the pot is evacuated through the exhaust pipe by means of a big rotary pump, because the liquid helium in the pot changes to the superfluid phase when it is cooled below 2.17K under a saturated vapor pressure of 4.16 x lo3 Pa. There is no temperature gradient across the liquid helium under a finite heat input. Evaporation of helium occurs without formation of bubbles, because the effective thermal conductivity of superfluid helium is extraordinarily high. Due to the convective type of heat conduction, it may become about 1,000 times higher than that of copper at room temperature. The pot is mounted on a thermal insulator made of fiber-reinforced plastic (FRP) which is fixed on the specimen stage. The specimen stage is also made of FRP and is firmly connected to the lower part of the thermal shield which is cooled with liquid nitrogen by flexible copper braids. While the stage is thermally very strongly coupled to the nitrogen tank, the mechanical coupling is only very weak. This mechanical uncoupling of our specimen stage from vibration sources such as the nitrogen and helium tanks is a very important feature of the stage. It should be noted that we constructed and arranged all parts of our cryo-stage, namely, the nitrogen tank, helium tank, pot, pot support, upper and lower thermal shields, and flexible copper braids, with an axial symmetry in order to minimize the drift induced by any temperature change in an individual part of the stage. This is one of the main reasons why a top-entry design was chosen for our specimen stage instead of a more commonly used side-entry design. Only a top-entry stage can be constructed symmetrical

STRUCTURAL

STUDY

OF MEMBRANE

39

PROTEINS

about the electron-optical axis and this symmetrical construction of the cryo-stage enables us now to record easily and reproducibly images of gold crystals displaying a lattice resolution better than 1 A at a stage temperature of 4.2K. The low consumption rates of liquid nitrogen and liquid helium, which are 170 ml/hr and 140 ml/hr, respectively, is another important feature of our stage. After refilling the two coolants, the working time of our cryo-stage is about 5 hr, during which no maintenance is required. From our experimental perspective, these low consumption rates are very important for efficient and successful high-resolution electron cryo-microscopy.

II.

STRIJCTURE

OF THE

PLANT

LIGHT-HARVESTING

COMPLEX

II

Before I describe the methodological details of electron crystallography, I would like to present an example of the sort of results which can actually be achieved by this technique. I. Light-harvesting Antenna The light-harvesting chlorophyll a/b-protein complex associated with photosystem II (LHC-II) is composed of a polypeptide chain of 25 kDa, less than 14 non-covalently bound chlorophyll molecules (Chl) a and 6, and 2 or 3 carotenoides (xanthophylls) (14). LHC-II functions as the major collector of solar energy in chloroplast membranes and, accordingly, roughly half of all pigment molecules involved in plant photosynthesis are bound to LHC-II molecules. The exact mechanism of the solar energy transfer process can only be understood when the structure of the participating pigmentprotein complexes such as LHC-II are known in detail. LI-IC-II isolated from pea leaves was initially used for 3D crystallization trials and W. Kiihlbrandt succeeded in growing 3D crystals of detergent-solubilized protein. However, the crystals did not yield X-ray diffraction patterns suitable for high-resolution structure analysis. Alternatively, Kiihlbrandt also tried to crystallize LHC-II in two dimensions and obtained 2D crystals of the LHC-II trimer in lipid bilayers which were ideally suited for electron crystallographic structure analysis (1.5). Using the LHC-II 2D crystals and an electron cryo-microscope equipped with a nitrogen stage as well as a field emission gun, Kiihlbrandt and Wang could

40

Y.

FUJIYOSHI

analyze the LHC-II structure to a resolution of 6 A (16). In fact, the resolution of their structure was not limited by the crystallinity of the 2D crystals but by the radiation damage. Therefore, the 2nd version of our electron cryo-microscope which we developed and constructed in the Protein Engineering Research Institute (PERI, Osaka) was used for data collection to higher resolution. At the very low temperatures used for data collection, the effects of radiation damage were reduced by a factor of more than two compared to the data collection at liquid nitrogen temperature. Therefore, more electrons could be used to record the images to improve the S/N ratio. The phases of the structure factors extracted from the images collected under these conditions were significantly better in terms of resolution as well as accuracy.

2.

Atomic

of LHC-II The amplitude and phase data extracted from electron diffraction patterns and images recorded with our high performance electron

TABLE

Structure

II

Crystallographic

Data

of LHC

2D crystals Layer Lattice

~321

group

a=b=129.5&c=lOOA

constants

(assumed)

Electron Number

diffraction of diffraction

Resolution

limit

patterns

for merging

Number of independent R-merge Completeness

reflections

83 3.2 A 17,920 27.6% 85.4%

(O”-90°)

Images Number Resolution

of images limit for merging

Number of independent Phase residual Completeness

Crystallographic R-factor Free R-factor

reflections

79 3.4A 15,124 27.7’ 87.4%(0°-900)

rejinement 33.0% 37.9%

STRUCTURAL

STUDY

OF

MEMBRANE

PROTEINS

41

microscope enabled Kiihlbrandt and Wang to calculate a 3D density map of LHC-II at a resolution of 3.4 A (8). A typical region of this density map is shown in Photo 2. Due to the high quality of the phase data (Table II) which led to a very low noise level in the density map, an atomic model of LHC-II could be fitted into the 3D map which included more than 80% of the polypeptide chain, 12 chlorophylls, and 2 carotenoids. Using phase and amplitude constraints, a crystallographic R-factor of 33.0% and a free R-factor of 37.9% could be achieved. An end-on view and a side view of the trimeric LHC-II structure found in the thylakoid membranes are shown in Photo 3a and b, respectively. A schematic view of the whole LHC-II is presented in Photo 4 to illustrate the polypeptide folding and the arrangement of the Chls. Helices A and B as well as the two lutein molecules are related to each other by a local pseudo two-fold symmetry. The charged residues Glu65 and Arg185 as well as Arg70 and Glu180 form ion pairs within the hydrophobic interior of the complex. These two ion pairs presumably provide a strong attractive force between the two helices (A and B) and are likely to play a major part in stabilizing the protein in the membrane. This distinguishes LHC-II from most other integral membrane proteins, where the structure is stabilized by the packing of several helices into a helical bundle. The Chls in LHC-II are attached to the polypeptide chain by coordination of the central Mg ion in the chlorin ring to polar amino acid side chains or to main chain carbonyls in the hydrophobic interior of the complex. For 8 Chls the side chain ligands have been identified as shown in Fig. 9, and a typical example, Chl a4 which is bound to Glu65, is shown in Photo 5a. At least three of the four remaining Chls (61, a6, and ~7) are likely to be bound by coordination of the Mg ion to main chain carbonyls as illustrated by the example of Chl a6 in Photo 5b. In this case, the distance of the central Mg atom from the polypeptide seems too long for direct binding. Therefore, Chl a6 might be fixed in its position via intervening water molecules. However, no water molecules could be directly observed in the LHC-II density map at 3.4 A resolution. Comparison of the crystallographic data of this analysis (Table II) with typical data obtained by X-ray crystallography shows that the amplitude data with an R-merge of 27.6% is quite poor compared to X-ray data where the value of R-merge is usually better than 10%.

42

Y.

FUJIYOSHI

stroma

PNGLYDhGG

N

NAWSVATNFVPGK 710 130

Fig. 9. Arrangement of the chlorophyll molecules in respect to the LHC-II polypeptide. The white letters on black background indicate amino acids involved in the binding of chlorophyll molecules (Nature, 367, 620 (1994)).

On the other hand, the phase data with a phase residual of only 27.7” are extremely good compared to X-ray crystallography where the value for the phase residual usually exceeds 50”. To be able to observe water molecules by electron crystallography, however, we need to improve not only the resolution but also the quality of the amplitude data, even though the high quality of the phase data can compensate for the poor amplitudes and allowed the atomic structure of LHC-II to be determined.

III.

ELECTRON

CRYSTALLOGRAPHY

The basic principle of electron crystallography is illustrated in the schematic drawing in Photo 6, in which the gourd-shape represents a protein molecule. A prerequisite for electron crystallography is 2D crystals of the protein of interest, which can be produced by various techniques. 2D crystallization is the least controllable step in electron crystallography and thus it is the most difficult part in the structure determination of a biological macromolecule by this method. For further information on 2D crystallization, the reader is referred to a very good and comprehensive review by Kiihlbrandt

STRUCTCRAL

STUDY

OF MEMBRANE

PROTEINS

43

(15). For high resolution data collection, 2D crystals are embedded in a thin layer of amorphous ice and/or sugar solution (details of specimen preparation will be discussed in section V of this review) and the sample is mounted in the cryo-stage of the electron microscope. During this transfer the specimen is kept below the critical temperature of around 140K at which the water undergoes a phase transition from amorphous ice into cubic ice. Low dose images (Photo 6-2a) and also electron diffraction patterns (Photo 6-2b) of 2D crystals have to be taken at many different tilt angles. Only a single image or electron diffraction pattern can be recorded from one crystal, because the recording of a single image or electron diffraction pattern already causes serious damage to the protein crystal. Therefore, a full electron microscopic data set has to be collected from many 2D crystals. A low dose image has a very poor S/N ratio because the irradiation damage limits the electron dose which can be used for the recording. This makes it impossible to see structural features in an image of a molecule which includes data to a resolution better than 3 A (Photo 6-2a). Accordingly, Fourier transformation of such a noisy image produces a Fourier pattern with an equally poor S/N ratio, but the diffraction spots can still clearly been seen as shown in Photo 6-3. By applying a filter which only allows the information concentrated in the diffraction spots to pass while the statistically distributed noise in the Fourier transform is masked off, it is possible to calculate a noise-filtered (or averaged) image of the structure as shown in Photo 6-5. In the case of a biological sample, electrons can cause deformation of the illuminated crystalline array. Generally, 2D crystals are not perfect and always show some distortion, and also adsorption to the grid causes some distortions. Therefore, Henderson et al. developed a procedure to correct for such lattice distortions which is based on correlation analysis of the crystal. A successful crystal unbending can improve the achievable resolution of a given image by a factor of almost two (Photo 6-4) (I 7). Crystallographic phase as well as amplitude data are extracted from images by Fourier analysis. However, amplitudes extracted from images are generally less accurate than those determined by electron diffraction experiments. Thus, whenever possible, the amplitudes used for the final structure analysis are calculated from

44

Y. FUJIYOSHI

electron diffraction patterns. Both phase and amplitude data from many tilted images and electron diffraction patterns are merged to produce a consistent 3D data set as shown in Photo 6-6a and 6b. Finally, the merged data set is used to calculate a 3D density map in the same way as is routinely done in X-ray crystallography (Photo 6-7). IV. REFINEMENT CROSCOPE

OF THE

HIGH-RESOLUTION

ELECTRON

CRYO-MI-

The electron microscopic analysis of the LHC-II structure (described in section II) demonstrated that amplitude data determined by electron crystallography are not as good as those determined by X-ray crystallography. Electron micrographs are important for the determination of the phases, but electron diffraction patterns are equally or even more important for the structure analysis, because current methods to refine the final model, which were originally developed for X-ray crystallography, are based on the amplitude data from the diffraction patterns. 1. Slow Scan CCD Camera A slow scan charge-coupled device (CCD) camera, developed by Krivanek and Moony (18), displays a higher dynamic range than electron microscopic film (EM-film) as illustrated in Fig. 10a. Figure lO(b and c) shows Wilson plots calculated from an electron diffraction pattern of bR taken by a lk x lk slow scan CCD camera and one recorded on EM-film. The resolution limit of the recorded data can be estimated from the linear region of the Wilson plot. In the case of the lk x lk CCD camera, the resolution was limited to about 2.8 A, even for electron diffraction pattern recorded from a 60” tilted specimen, while the resolution of the pattern recorded on EM-film was limited to only 3.2 A. However, the unscattered beam heavily saturates the CCD cells in the center of an electron diffraction pattern and causes overflows to adjacent cells of the CCD array (Fig. 11). Therefore, we routinely use a beam stopper of an optimized size, which is the minimum size sufficient to stop the central beam of electron diffraction pattern, to avoid these overflow effects. While the values recorded by the CCD cells correspond linearly to the electron dose, the sensitivity of each cell in a CCD array

STRUCTURAL

STUDY

OF

MEMBRANE

45

PROTEINS

16ml measured data 0 3.4141 +m.* ----.

14oLw 12ml-

,/

./’

1C000-

...’

,,I. _..’

._...~.

/k

Bouo6wJw 4ooo-

,... . ..B

,..’ 2lxyl - /..++ ,.,. 0:0 500

_/

,/’

,..’

_,.I

loo0

,..’

,..I

1500

2cuul

2500

3

electrons/pixel

b SSCCD

,1

0 I

”

0.00

I

” 3.8A

0.00

Fig.

10.

and

EM-film.

I

(2SinsA~

3.6A

”

0.1

0.1 3% 3.1A (2sinen)z

Collection of electron diffraction data (a) The plot of the value measured

camera as a function of the electron ing that a CCD camera is suitable Wilson plots of electron diffraction by a CCD camera (b) and EM-film of the curve, arrows). The film.

I

3.OA 22.A

with a slow-scan CCD by a lk x lk slow-scan

camera CCD

dose shows a linear behavior demonstratto record electron diffraction patterns. patterns of 60” tilted bR crystals recorded (c). From the end point of the linear range

it is possible to estimate the resolution values are 2.8 A for the CCD camera and

limit only

of the data (open 3.2 A for the EM-

46

Y.

Fig. 11. slow-scan

Electron CCD

center leads tical streak.

diffraction camera. The

to overflows

pattern strong

to adjacent

of a bR intensity cells

Fig. 12. Electron diffraction patterns 2k x 2k and (b)a lk x lk slow-scan CCD stopper

was

CCD camera a significantly lk x lk CCD

used

to

avoid

overflow

crystal recorded of the incident

of the

of bR camera. effects.

enables us to use a larger camera smaller area of the diffraction camera.

CCD

array

crystals In both However, length, pattern

FUJIYOSHI

with a lk x lk beam in the producing

recorded with cases the same because

the

a ver-

(a) a beam 2k x 2k

the beam stopper covers than in the case of a

varies. However, the individual cells can be calibrated by recording an evenly exposed as well as a non-exposed frame, which are then used to determine the gain of each cell. Using a CCD camera, we were able to collect a high quality 3D amplitude data set for bR, which included more than 200 electron diffraction patterns from

STRUCTURAL

STUDY

TABLE III Rfrrede~ and kne,

for Non-tilted

~____ (a)

OF

MEMBRANE

Data

47

PROTEINS

of bR

Number of diffraction patterns

(b) Cc)

EM-film CCD lkx lk CCD 2k x 2k

” Figures

in parentheses

27 36 30 indicate

standard

12.7(3.2) 13.3(2.3)” 5.6(0.9)

17.6(7.5) 16.5(2.8) 8.9(1.1)

deviations.

specimens tilted up to 70”, within only a few weeks. Compared to the use of EM-film to record electron diffraction patterns, the use of a CCD camera saves a significant amount of time because the diffraction patterns are already recorded in a digitized form. On the other hand, diffraction patterns collected on EM-film have to be scanned which requires several hours for each film. More importantly, the CCD camera enables us to check the quality of each diffraction pattern related to a shape of 2D crystal, and is a great help in taking diffraction patterns very effectively. A comparison of electron diffraction data recorded with a 2k x 2k slow-scan CCD camera and data recorded with a lk x lk slow-scan CCD camera is shown in Fig. 12. It is obvious that the 2k x 2k camera produces much better electron diffraction data as illustrated by the much smaller R-Friedel value of 5.6% compared to the Ik x lk camera which yielded an R-Friedel value of 13% (Table III). A ccordingly, our electron cryo-microscopes after the 3rd version were equipped with a new slow-scan CCD camera with a 2k x 2k chip. Quantitative analysis of electron diffraction data revealed that the 2k x 2k camera produced much better R-Friedel values, especially in the higher resolution range, than the lk x lk camera or EM-film. The R-merge value of our 3D amplitude data set determined from electron diffraction patterns of bR 2D crystals recorded with a 2k x 2k CCD camera is now of similarly good quality as the R-merge value of a typical X-ray crystallographic data set of comparable resolution.

2.

Fiec!d Emission

Gun

As discussed earlier, the side chains of amino acids are well resolved when the structure can be analyzed at a resolution of 3.0 A or better. The deterioration of the contrast transfer function (CTF)

48

Y. FUJIYOSHI

strongly affects the S/N ratio of the Fourier components of images, especially in the high resolution range. The 2nd version of our microscope was equipped with a LaB6-filament and the images were clearly affected by the attenuation of the intensity of the CTF, especially in images which were recorded at a high defocus value as shown in Fig. 13b. The 3rd and also the 4th version of our electron cryo-microscope, the latter enabling us to observe a larger area of the specimen than the former, are equipped with a 300 kV field emission gun (FEG) which gives less attenuation of the contrast transfer function at high-spatial frequency and is therefore more suitable for highresolution imaging of biological samples. As described earlier, a biological sample has generally low scattering amplitudes in the high spatial frequency range, because it is composed of light atoms such as hydrogen, carbon, nitrogen, and oxygen. Therefore, biological samples require a coherent electron beam such as the one provided by the FEG, which leads to a reduced attenuation of the CTF at high-spatial frequencies even under highly defocused imaging conditions as shown in Fig. 13a. 3. EfJicient Cryo-transfer Device Optimizing the specimen preparation is another very important task remaining in electron cryo-microscopy. Ice-embedding has proven to preserve the specimen structure even under the highvacuum conditions of an electron microscope. However, specimen preparation is a trial and error procedure as there are no generally applicable rules. For example, optimizing the thickness of the ice film in which the specimen is embedded is purely empirical. In 1986, we developed a special cryo-transfer device for our top-entry microscope. However, it remained an instrumental inconvenience in that a long time was required to exchange the specimen and the handling was difficult. For example, a common problem was that the specimen fell off the transfer device during the specimen exchange procedure. An instrument suitable for highresolution data collection for electron crystallographic structure determination, however, must provide a fast, smooth, and reliable specimen exchange mechanism. From the user’s point of view, the cryo-transfer device is actually one of the most important parts of an electron cryo-microscope, therefore, we spent a lot of time trying

STRCCTIJRAL

STUDY

OF

MEMBRANE

b

a 3oOkV.

Cs1.6,

C~2.2,

49

PROTEINS

@=1x1@

4ookV,

Cs=1.9,

Cc=2.2,

@3x10-4

1 I

Fig.

f:8CCO

II

13.

Squared

CTF

under

three

6,000 partial

A, and 8,000 A). The attenuation coherency is much less pronounced

pared

to a LaB6

filament

f:8000

/

different

defocus

conditions

(4,000

A,

of the CTF due to energy spread and for a field emission gun (a) com-

(b).

to improve the cryo-transfer device. The volume of a liquid nitrogen vessel of a specimen transfer holder was optimized by changing the construction of the specimen transfer device from a bottom setting to a side setting which was shown in Fig. 8 of ref. 22, because the side setting enabled us to minimize the cooling liquid nitrogen for the specimen holder. The design of our present cryotransfer device is shown in Photo 7a and b as cross sections perpendicular and parallel to the path of the electron beam, respectively. The yellowish brown parts in this drawing are cooled by liquid nitrogen to a temperature of about 95K. Our current procedure to place a new specimen in our electron microscope is as follows: first of all, the specimen is plunged into

50

Y.

FUJIYOSHI

liquid ethane to embed it in a thin film of amorphous ice. The grid is placed into the specimen holder in a nitrogen gas atmosphere at a temperature of about -160°C. The specimen holder is dipped into liquid nitrogen and placed into the transfer holder which is pre-cooled by liquid nitrogen and contains a small amount of liquid nitrogen. The transfer holder is brought into the pre-evacuation chamber of the cryo-transfer device, where the nitrogen is removed by a rotary pump. Subsequently, the specimen holder is moved from the pre-evacuation chamber to the specimen exchange chamber by opening a valve between the two chambers and is seized by a chuck at the transfer rod with a caterpillar. By opening the valve between the exchange chamber and the electron microscope column, the specimen is finally introduced into the stage through the guide for the transferring caterpillar, which was designed for changing direction of specimen holder from horizontal to vertical in the specimen chamber. In this way, a new specimen can be placed into the microscope and observed less than 10 min after observation of the previous specimen has been completed. 4. Images of Ice Embedded Specimens Our cryo-transfer device enabled us to study many biological samples which were prepared by embedding in vitreous ice (19), an improved technique based on the technique of frozen-hydrated preparation (20). Figure 14 summarizes some of our results, namely, high resolution images of a DNA-RecA complex (Fig. 14a and b), double stranded DNA (Fig. 14c), influenza viruses (Fig. 14d) and haemagglutinin (Fig. 14e). More importantly, from electron diffraction patterns of vitrified catalase crystals holding about 70% water, we could confirm that the intensities of the diffraction spots up to a resolution of 2.0 A are very similar to those obtained by X-ray diffraction (Fig. 15). This indicates that the ice embedding technique preserves the intact structure of a protein to a resolution of 2.0 A. For optimal performance, the 3rd version of our electron cryo-microscope equipped with a high-resolution helium stage, a cryo-transfer device, and an FEG is set up in a room with soundproof walls as shown in Fig. 16.

STRUCTURAL

STUDY

OF ~~~MBR,~~E

PROTEINS

Fig. 14. Images of ice-embedded specimens obtained with our microscope using the new cryo-transfer device. {a) image of complexes and (b) optical diffraction pattern from a straight filament shown in (a). images of(c) double stranded DNA, (d) partictes, and (e) a trimer of haemaggiutinin molecules.

s1

electron cryoRet A / DNA Ret A 1 DNA influenza virus

52

Y. FUJIYOSHI

Fig. 15. ous ice.

Electron

diffraction

pattern

of a catalase

crystal

embedded

in vitre-

STRUCTURAL

V.

STUDY

SPECIMEN

OF MEMBRANE

PREPARATION

53

PROTEINS

TECHNIQUES

Electron crystallography requires 2D crystals which are easily deformed by very weak mechanical forces. The deformation causes blurring and/or extinguishing of diffraction spots particularly in the case of highly tilted specimens. Both images and electron diffraction patterns from highly tilted specimens are essential for highresolution structure analysis. This means that a specimen support which minimizes the deformation of the specimen is of primary importance for data collection in high-resolution electron crystallography.

1.

New Molybdenum

Grids

As stated earlier, an electron cryo-microscope is required for the observation of proteins at high-resolution to reduce beam-induced radiation damage and to overcome the problem of specimen dehydration in the vacuum of the electron microscope by embedding the specimen in amorphous ice. However, cooling of the specimen leads to cryo-crinkling of the carbon film which is caused by different shrinkage of the carbon film and the supporting electron microscope grid when the temperature is lowered. This presents a serious problem when specimens are cooled to low temperatures. Booy and

Fig. 16. Electron microscopic system tional Institute for Advanced Research version of our electron cryo-microscope

as currently installed in the Interna(IIAR, Matsushita, Ltd.). The 3rd is equipped with a super-fluid helium

stage, a cryo-transfer scan system.

a slow

device,

an FEG,

scan

CCD

camera,

and

a spot

54

Y. FUJIYOSHI

Pawley found that grids made of titanium, molybdenum, or tungsten dramatically reduce the cryo-crinkling effect (21), and more recently the use of molybdenum grids has become standard in electron cryo-microscopy. However, the surfaces of commercially available molybdenum grids are not flat and smooth, and the rough surface of these grids can also induce the carbon film to crinkle as shown in Fig. 17a and b. We designed various types of molybdenum grids specifically for high-resolution electron crystallography. All the newly designed grids were 3 mm in diameter and had a thickness of 20 pm. While one grid design had circular holes 50 m in diameter, a different design featured circular holes 100 p in diameter. Holes with hexagonal shape were also created to increase the ratio of observable area, and various arrangements of the holes were tested. All grids were carefully manufactured by photochemical etching to preserve a smooth surface especially around the hole edges. After the etching, the grids were washed with sulfuric acid and distilled water. Ultrasonication in chloroform was the final step in cleaning of the grids. All these grids were checked in a scanning electron microscope (a typical example is shown in Fig. 17~) and the carbon film supported by such grids displayed no crinkling as shown in Fig. 17d.

2.

Carbon Films by Spark-less Evaporation

While high quality molybdenum grids are important for electron crystallography, the flatness of the carbon film itself is even more important. Only an atomically flat surface of the specimen support film can be expected to preserve the planarity of the 2D crystals to the highest degree. A high quality carbon film can also reinforce the 2D crystals which adhere to it and reduce beam induced movement. Butt et al. examined the flatness of carbon films by atomic force microscopy and developed a multiple evaporation technique to prepare very flat carbon films (22). Although in our laboratory this multiple evaporation technique sometimes yielded very flat carbon films, the results were not very reproducible. In our procedure to produce flat carbon films, a pure carbon rod (a spectroscopic carbon electrode 3 mm in diameter with a purity of 99.9999%) is thinned to about 1.5 mm diameter and mounted in a JEE400 vacuum evaporator. The most important

STRUCTLFRAL

STUDY

OF

Fig.

17.

ning num

electron microscope, grid appears quite

carbon

Carbon

MEMBRANE

film

chemically no crinkling

deposited

film

on molybdenum the rough. on such

etched molybdenum of the carbon film.

55

PROTEISS

grids.

(a) When

imaged

surface of a commercially (b) The optical micrograph a grid grids

shows

crinkles.

is much

(c) The

smoother

with

a scan-

available molybdeillustrates that and

surface (d)

a

of photothey

induce

requirement for the preparation of flat carbon films is the spark-less evaporation of the carbon, and the most reliable technique to achieve this is a pre-evaporation of the carbon rod before preparation of the actual carbon film on mica. For this purpose, we thin an approximately 7 mm long stretch at one end of the carbon rod. While the freshly cleaved mica is covered, we evaporate carbon until the carbon stops sparkling during evaporation. Then, the cover over the mica is removed and the carbon deposited very slowly and carefully on the mica surface, avoiding any sparks. Spar-

56

Y.

FUJIYOSHI

kling of the rod during the actual evaporation of the carbon is probably prevented by the carbon crust which formed during the preevaporation (see schematic drawing in Fig. 18) and presumably acts as a kind of filter for large carbon particles to which sparks are attributed. The vacuum is kept at better than 2 x 10e6 Torr even during the actual carbon evaporation. However, the vacuum condition generally deteriorates by carbon evaporation because the high temperature causes the release of gases from the carbon rods. The pre-evaporation minimizes the out-gassing effect because the carbon rods are essentially baked out. 3.

Trehalose Embedding

Technique

For trehalose embedding, a carbon film was floated on an air-waterinterface and picked up with one of the newly designed molybdenum grids. If the molybdenum grid had been previously cleaned with chloroform and was too hydrophobic, glow discharging was used to render it more hydrophilic. bR 2D crystals, suspended in citrate buffer (pH 5.5) containing 3% (w/v) trehalose, were then applied to the molybdenum grid with the carbon film and the grid mounted on a plunger (Reichert KF 80) with tweezers. Within 5 set after blotting off the excess bR suspension with a filter paper, the specimen was plunged into liquid ethane of about -160°C (Fig. 19).

The frozen specimen was transferred into our FEG electron cryo-microscope and observed at an acceleration voltage of 300 kV. Images were taken at a magnification of 50,000 x with an electron dose of 15 electrons/A2 using a minimum dose system (5) and an exposure time of 2 sec. The EM-films (Kodak SO 163) were developed for 14 min in a D-19 developer at 20°C. Micrographs were digitized with a scanner (Leaf Scan-45) (23) and analyzed with

Fig. 18. evaporator. evaporation probably

Schematic drawing of the setting of the The carbon crust on the tip of the right which is needed for spark-less carbon prevents the deposition of carbon clusters

carbon rods in the carbon rod forms during the preevaporation. This crust onto the mica surface.

STRUCTURAL

STUDY

OF

MEMBRANE

57

PROTEINS

modified versions of the MRC image processing programs. Images and electron diffraction patterns recorded from specimens prepared by this trehalose-embedding technique showed a resolution beyond 3.0 A, even for specimens tilted by 60” in the direction perpendicular to the tilt axis. The number of diffraction spots with IQ values ranging from 1 to 4 to a resolution of 4 A was increased by 50% for images recorded at tilt angles of 20”, 45”, and 60”; also, the electron diffraction data was greatly improved. Merging of 17 electron diffraction patterns recorded with a 1k x 1k CCD camera from trehalose-embedded and partially hydrated bR 2D crystals resulted in an R-merge value of 9.3%. This value is considerably better than the R-merge value of 12.8% obtained by merging the same number of diffraction patterns recorded from glucose embedded bR 2D crystals. Thus, using our photochemically etched molybdenum grids, very flat carbon film, and the new specimen preparation technique, we were able to collect images and electron diffraction patterns of bR displaying a resolution better than 3.0 A. Some problems still remain for efficient data collection, however, especially in the case of highly tilted specimens. Our newly designed molybdenum grids are thicker than conventional EM grids and when they are tilted, the frame covers a substantial area near the edges of the holes. This presents a problem because 2D crystals tend to stick close to the

Fig. 19. croscopy

Schematic by a rapid

drawing freezing

of specimen technique.

preparation

for

electron

cryo-mi-

58

Y.

FUJIYOSHI

edges of the holes of the EM grid. For highly tilted specimens, this grid must be used because at a tilt angle of 70”, the 50 pm holes are reduced to ellipses with a short axis of only 17 pm, which is relatively small to record high quality diffraction patterns. We therefore used molybdenum grids with a hole size of 100 pm which enabled us to record diffraction patterns of bR 2D crystals at a tilt angle of 70”. However, the flatness of the bR 2D crystals prepared on these grids was slightly inferior to the preparations on the grids with a hole size of 50 pm.

VI.

STRUCTURE

1.

Proton Pump

OF BACTERIORHODOPSIN

bR is a membrane protein found in H. salinarium (24) with a bound retinal molecule which serves to absorb light energy. bR functions as a very efficient proton pump and absorption of a single photon by the retinal suffices to transport a proton from the inside of the cell to the outer medium. Many studies have been performed on bR and revealed the functional intermediates of the bR photocycle (25). These studies laid the basis for the proposed mechanism of the proton pumping by bR (7). However, details of the proton pumping mechanism by which a proton is actually transported from the inside to the outside are not known. It is not clear how protons are efficiently introduced into the channel of bR, how they are carried along the channel, or how the protons are released from the other side of the protein. To address these challenging questions, highresolution electron crystallography has been recognized as one of the most promising tools. Continuing on from their pioneering work, Unwin and Henderson (6), they were able to determine an electron crystallographic density map of bR at a resolution of 3.5 A. Based on this 3D map, Henderson and his coworkers built an atomic model of the protein including the retinal molecule (7) and deposited the atomic co-ordinates in the Brookhaven Protein Data Bank (PDB). They thereafter improved their atomic model by crystallographic refinement with electron diffraction data corrected for diffuse scattering and additional phase information calculated from 30 new images of tilted specimens (26). The atomic coordinates of the resulting new atomic model of bR have also been registered in the PDB (2BRD).

STRUCTURAL

STUDY

OF MEMBRANE

PROTEINS

59

However, because X-ray crystallography has already achieved structure analysis at much higher resolutions than 3.5 A, even with membrane proteins, we set out to analyze the structure of bR at a higher resolution, independent of the structure analysis by Henderson and co-workers. Moreover, confirmation of the reliability of structure analysis by electron crystallography is still thought to be useful for this method. While Henderson et al. proposed a model for the proton transport mechanism within the bR channel (7), we were also interested in the mechanisms by which protons are efficiently guided to the opening of the bR channel and by which the protons are released from the other protein surface. Another goal of our work was to confirm the possibility that electron crystallography could detect the ionization state of charged amino acid residues. We earlier attempted to detect charge transfer on chlorinated copper ghthalocyanine based on quantitative analysis of the images at 1.8 A resolution (27). However, the results were ambiguous because the lattice constant of this organic crystal is not large enough to yield information in the low resolution range of the electron scattering factors. If the ionization conditions could actually be detected, electron crystallography would be a very promising technique to trace the proton movement along essential amino acids in the bR channel. Detection of the ionization states would tell us exactly the localization of the amino acids to which the proton is bound in the various intermediates of the bR photocycle. The discrimination of ionization conditions is also important to understand the detailed relationship between structure and function of proteins in general.

2.

Data Collection

Large 2D crystals were prepared by fusion of purple membrane patches (28) which were purified from H. salinarium JW 5, a retinal deficient mutant strain (29). To reconstitute functional bR molecules, the retinal was added to the culture medium in the dark. Specimens were prepared for electron cryo-microscopy as described in section V and electron diffraction patterns were recorded with a slow scan CCD camera as described in section IV. While electron diffraction patterns up to a tilt angle of 60” showed reflection spots to a resolution of 2.5 A, the resolution of diffraction patterns recorded at a tilt angle of 70” was limited to about 3.0 A.

60

Y.

Fig. 20. angles (O’, which are spots. Tbe and 2.4 ip.

Untilted

200

45”

60”

Calculated diffraction patterns of bR images recorded at 20”, 45”, and 60”). The b oxed numbers represent the a measure of the signal/noise ratio of the individual circles indicate a resolution of 80 .&, 20 A, 10 A, 7 A,

FUJIYOSHI

various tilt IQ values diffraction 5 A, 3.5 A,

High-resolution images were recorded mainly with the 3rd version of our FEG electron cryo-microscope which was operated at an acceleration voltage of 300 kV while the specimen was kept at a temperature of 4.2K. Most images were recorded as the conditions stated in section V. Computational image analysis which included crystal unbending was used to extract the Fourier components and the IQ values are a measure of the S/N ratio of each Fourier component (I 7) (Fig. 20a, b, c, and d at tilting angles of O”, 20*, 45”, and 60”, respectively). The IQ values indicated that the phase data extracted from our images were of high quality to a resolution of better than 3.0 A as shown in Fig. 20.

STRUCTURAL

STUDY

OF

MEMBRANE

61

PROTEINS

5’. Structure Analysis An atomic model of bR including all surface loops was built into the experimental 3D density map which was calculated from 366 electron diffraction patterns and 129 electron micrographs. The resulting R-merge and phase residual were 15.5% and 26.7”, respectively, and a complete crystallographic table is given in Table IV. The resolution of 3.0 A was only achieved by using our electron microscope equipped with a liquid helium stage by which the effects of radiation damage are reduced by a factor of about two and 10 compared to liquid nitrogen temperature and room temperature, respectively (as discussed in section I). Using our electron cryomicroscope, we could collect high resolution electron crystallographic data from bR 2D crystals up to a tilt angle of 70”. In practice, it is very difficult to collect data from higher tilt angles and thus a cone-shaped region in the reciprocal space is not sampled. In electron crystallography this is known as the missing cone problem. However, because we could collect data up to a tilt angle of 70”, the

‘TABLE

I\’

Crystallographic ~____

Data

of bR

20 crystals Layer Lattice

group constants

P3 a=b=62.45A,c=lOOA (assumed)

--___~.-

Electron Number Resolution Number

diffraction of diffraction patterns limit for merging of independent reflections

366 2.8 A

R-merge

9,531 15.6%

Completeness

89%

(OO-90’)

Images Number

of images

Resolution limit for Number of independent Phase residual Completeness -~____

Crystallographic R-facto1 Free R-factor

merging reflections

129 3.oA 7,129 26.7” 81 .O%(OO-900)

refinement 24.0% 33.4%

62

Y. FUJIYOSHI C-Termjm

-

N-l

Fig.

21.

retinal in anti-parallel

Overall the

center P-sheet

view

of a bR

monomer

of the molecule, structure between

depicting

and the helices

the

seven

a-helices,

loop regions which include B and C on the extracellular

the an

side.

residual missing cone was reduced to less than 5% of the entire Fourier space. The reduction of the missing cone dramatically improved the quality of the 3D map, especially in the direction perpendicular to the membrane plane. This is illustrated in Photo 8a, which shows the region of our map around the retinal molecule and Photo 8b which shows two aromatic residues, Tyr18.5 and Phe208. These two regions of the 3D map were shown in the previous paper by Henderson and co-workers (7). The side-chain densities for Tyr185 and Phe208 with a center-to-center distance between the two aromatic rings of about 5 A (the distance between the two closest atoms is only about 3 A) are clearly resolved as shown in Photo 8b. The high quality of our 3D map enabled us not only to trace the seven transmembrane a-helices but also to interpret all loops in terms of the bR amino acid sequence. The loops displayed distinct structures, such as an anti-parallel P-sheet and quick turns at the end of extended a-helices (Fig. 21), which were notOvisualized by a recent X-ray crystallographic analysis of bR at 2.5 A (30). 4. Lipid Molecules Our experimental 3D map revealed eight lipid molecules

related to

STRUCTURAL

STUDY

OF

MEMBRANE

PROTEINS

63

one bR molecule. In the crystallographic refinement these lipid molecules were modeled as phosphatidyl glycerophosphate monomethyl ester with dihydrophytol chains (Photo 9) (31), which is a major phospholipid found in purple membranes (32). The arrangement of the lipid molecules with respect to the bR trimer is very similar to the results obtained by Grigorieff et al. (26). Interactions between the molecules in the bR trimer are mainly formed by hydrophobic-hydrophobic contacts of the transmembrane helices (Photo ‘lOa), while in the case of LHC-II, the trimer is stabilized primarily by protein interactions just outside of the lipid bilayer as shown in Photo 11. In contrast to LHC-II, the bR trimer is stabilized by the interaction of the B-helix in one molecule with helices D and El of an adjacent bR molecule in the hydrophobic core of the bilayer. Interestingly, one lipid molecule between helices B and E is only present in the cytoplasmic leaflet and has no counterpart in the extracellular leaflet of the bilayer as shown in Photo lob. LHC-II showed no helix-helix or lipid-protein interactions within the membrane stabilizing the trimer (Photo 11). 5. Surface Structure of bR Analysis of the potential map calculated from our atomic model revealed a characteristic distribution of charged amino acids on both surfaces of the protein. This characteristic distribution of acidic and basic amino acid residues enabled us to propose hypothetical mechanisms for efficient guidance of protons to the channel opening on one side of the protein and for the efficient release of the protons from the other side. The program GRASP was used to visualize the structural features of the cytoplasmic (Photo 12a) and the extracellular surface (Photo 12b) of the bR trimer, highlighting the charge distribution. Because of the illumination direction used to shade the structure in Photo 12, it is difficult to see the three-fold symmetry of the bR trimer in this figure. Therefore, only one channel entrance (indicated by an arrowhead) can be seen in the trimer on the cytoplasmic surface. The entrance is located in a negatively charged plane (red area) surrounded by mountain-like protrusions which include positive charges (shown in blue). At the boundaries of the bR molecules, especially in the inner area of the trimer, negatively charged grooves can be observed and the lipid molecules, which are caught

64

Y. FUJIYOSHI

in the outer area of these grooves (as previously described; Photo lob), have a head group which is also negatively charged. Because the structure analyzed by electron crystallography represents an intact trimer, this characteristic charge distribution on the cytoplasmic surface of the bR trimer suggests a very attractive mechanism for efficient channeling of the protons to the cytoplasmic entrance of the transmembrane channel. In our proposed mechanism, the positively charged protons are initially adsorbed on the negatively charged surfaces of the lipid bilayer, inside as well as outside of the bR trimer. Protons adsorbed on the lipid surfaces outside of the bR trimer would migrate along the grooves formed between the individual bR monomers into the center of the bR trimer. Protons which have initially adsorbed to the lipid surface inside the trimer as well as those which were guided into the center of the trimer are trapped there until they finally enter the channel because, as described earlier, positively charged mountain-like protrusions enclose the plane in which the channel entrance is located. On the extracellular surface of the bR trimer, a completely different surface structure and charge distribution is observed. The center of the trimer is dominated by positive charges so that protons leaving the negatively charged channel (red area in Photo 12b) tend to be released from the surface of the bR trimer. Thus, our electron crystallographic analysis of the intact trimer structure of bR led to a very attractive explanation for the mechanisms governing the efficient accumulation of protons on the cytoplasmic side and the release of protons from the extracellular side.

6. Detection of the Ionization State While the scattering factor of a neutral and a charged oxygen atom are very similar for X-rays (Fig. 22a), they differ greatly for electrons in the low resolution range. The electron scattering factor of a negatively charged oxygen atom even has a negative value in the low resolution range as shown in Fig. 22a. Therefore, the density of a negatively charged atom in the simulation of a potential map appears weak or even to have a negative contrast, while a neutral oxygen atom creates a clear positive contrast in the potential map as shown in Fig. 22b. However, such effects should disappear in the potential map if the low resolution data is omitted. This is actually

STRUCTURAL

a

STUDY

OF MEMBRANE

65

PROTEINS

b

fW 3 1

0.2 2.5

03 1.67

.

Fig.

22.

Scattering

factors

trons (a; broken line) and simulated potential maps oxygen

atom

(b;

of a neutral

and

an ionized

for X-rays (a; continuous of a charged oxygen ion

oxygen

atom

for

elec-

line) as well as computer (b; upper) and a neutral

lower).

helpful because it allows discrimination between a negatively charged atom and an atom which is neutral but disordered, which both would appear as weak densities in the map. For example, when we recalculated a map using only data from 7 A to 3 A, the initially weak density of a negatively charged atom became much stronger, while the density of a neutral but disordered atom remained weak. In our 3D map of bR including the low resolution data, we saw no density for Asp85 or Asp212 (Photo 13a); however, when we recalculated the map using only data from 7 A to 3 A, the densities for these two residues became visible (Photo 13b). On the other hand, the densities for Asp96 and Asp1 15 were observed in maps calculated with and without the low resolution data (Photo 13~ for Asp96 and Photo 13d for Asp1 15). By these examinations, we confirmed that electron crystallography is capable of discriminating between charged and uncharged Asp residues, because it was already spectroscopically shown that Asp85 and Asp212 are charged while Asp96 and Asp1 15 are uncharged (33). This result

66

Y.

FUJIYOSHI

shows that high-resolution electron crystallography, taking advantage of the refined electron cryo-microscope (section IV) and the improved specimen preparation technique (section V), can actually be used to detect protons in the structure of a membrane protein. The different appearance of the density of a charged and an uncharged residue could also be observed in the density map of LHC-II which was discussed in section II. The positively charged Arg side chains were very well defined, whereas there was virtually no density for the Glu side chains even though they are probably equally well ordered because they participate in the formation of salt bridges. Thus, the invisibility of the density for the Glu side chains may be due to the negative electron scattering factors of negatively charged groups. Most other acidic side chains had similarly poorly defined densities in0 the map which included the low resolution data of LHC at 3.4 A resolution. The various intermediate states of bR during the photocycle are well characterized (34) and helped Henderson to propose the model for proton translocation across the channel in bR based on the atomic model (7). However, to really understand the proton pumping mechanism in every detail, we must determine the high resolution 3D structure of all the intermediates of the bR photocycle. If the electron crystallographic analysis of the intermediates could really be achieved in practice, we would be able to create a movie of the proton translocation. Since electron crystallography can discriminate between the charged and uncharged state of the key amino acid residues in the transmembrane helices of bR based on the characteristic difference of the electron scattering factors, we were able to follow the proton on its passage through the bR channel. Therefore, high-resolution electron crystallography will be able to provide us with new and exciting insights into the structure and function of bR and many other membrane proteins.

VII.

1.

STRUCTURE

OF AQPl

Wuter Channel

Large volumes of water cross the membranes of specialized cells in vertebrates, invertebrates, and plants. For example, in humans hundreds of liters of water pass through biological membranes in a

STRUCTURAL

STUDY

OF MEMBRAKE

PROTEINS

67

single day. The existence of water channels was already postulated as early as half a century ago (35) because the water permeability of the lipid bilayer itself cannot account for the osmotically driven water flow observed across red blood cell membranes. Recently, the gene for an abundant integral membrane protein from the red cell membrane was cloned (36) and expression of its mRNA in Xenopus oocytes induced a mercury sensitive osmotic swelling of the oocytes (37). This 28 kDa protein is now the archetypal member of a new subfamily of the growing major intrinsic protein (MIP) family, which all function as water channels, and has been termed AQPl. AQPl has been purified and reconstituted into 2D crystals in the presence of phospholipids (38-41). Sequence analysis has suggested that all known proteins of the MIP family contain six transmembrane helices and a highly conserved NPA (Asn-Pro-Ala) motif in each of the two most prominent loops between the 2nd and the 3rd helix as well as between the 5th and the 6th helix ($2). Site-directed mutagenesis on the loops containing the NPA motifs further indicated that these segments are probably close to the water pore because these domains were shown to be essential for the function of the water channel. 2. 30 structure of AQPl The 3D structure of AQPl shown in Photo 14 was calculated from 45019 diffraction intensities and 6854 phases determined from images and electron diffraction patterns collected over a tilt angle from 0” to 60”. The square-shaped AQPl tetramer in the center of the figure is viewed from the extracellular side, while the four adjacent tetramers exhibit their cytosolic surface (Photo 14). A section through an AQPl monomer parallel to the membrane plane reveals six distinct circular densities surrounding an additional, noncircular density (Photo 15). The shape and dimensions of the six peripheral densities suggest that they represent typical membrane spanning a-helices, in agreement with the known helical structure of AQPZ . The complex central density (marked by X) surrounded by the six distinct rods differs from the simple helical rod shape shown in Photo 1.5. This mass is mostly flat and wide, especially on the extracellular side, and is branched. We cannot assign this density to a transmembrane helix because immunological stud.ies of AQPl proved that both termini, the N- and the C-

68

Y.

Fig.

23.

secondary nent loops thought in the map

Transmembrane structure LB and

model prediction LE which

to fold back into the center of the monomer (Nature,

387,

626

of AQPl

based

FUJIYOSHI

on the hydropathy

from the amino acid sequence. contain the highly conserved membrane and might form the seen in the electron crystallographic

The NPA

plot

and

two promimotifs are

complex 3D

density density

(1997)).

terminus, are located in the cytoplasm and this supports a structure based on an even number of transmembrane helices. However, in the hourglass model of AQPl , the extended loops B and E carrying the NPA motifs fold back into the membrane (Fig. 23). Although connections of the central density to the rest of the molecule are not clearly seen, the density X might represent these loops, suggesting that the water channel is located in this region. However, much higher resolution information is required to gain insight into the tantalizing specificity of AQPl for water. Interestingly, the six helices of AQPl form a right-handed helical bundle (Photo 16), although all other known membrane protein structures display left-handed helical bundles. We confirmed the handedness of our AQPl structure by calibrating the structural analysis with the structure of bR which is solved to atomic resolution and where the seven transmembrane helices are well established to form a left-handed helical bundle. The quite unusual right-handed helical bundle structure of AQPl can be explained by helical-helical interactions of the transmembrane helices which are tilted by up to 40” with respect to each adjacent helix and enable the AQPl molecule to form a widely open entrance to the water channel. However, to understand the structural basis for the highly specific and efficient water channel formed by AQPl, a 3D map at

STRUCTURAL

STUDY

OF

MEMBRANE

PROTEINS

69

atomic resolution is required, and this is likely to be achieved in the near future. SUMMARY

A high-resolution electron cryo-microscope equipped with a topentry specimen stage has been refined by modifying a previously described superfluid helium stage. Instruments equipped with such a cryo-stage achieve a resolution of better than 2.0 A and have proved extremely powerful in the high-resolution structure analysis of membrane proteins. Improvement of the electron microscopic system in combination with improved specimen preparation techniques allowed the structure of bR to be analyzed to a resolution of 3.0 A. The 3D structure of bR, especially the surface features, revealed the structural basis for the efficient guidance of protons to the entrance of the transmembrane channel. Based on the characteristic difference of the atomic scattering factors for electrons of ionized atoms versus neutral atoms as well as the data analysis, charged and uncharged amino acid residues could be discriminated. Thus, electron crystallography is providing us with new and exciting insights into the structure of membrane proteins because it not only enables us to determine the structure of a membrane protein, but allows us to study its interaction with the surrounding lipid molecules and to determine its ionization state. Acknowl’edgments These studies were performed in collaboration with W. Kiihlbrandt and D. N. Wang for LHC-II; K. Mitsuoka, T. Hirai, K. Murata, A. Miyazawa, A. Kidera, and Y. Kimura for bR; and T. Walz, T. Hirai, K. Murata, J. B. Heymann, K. Mitsuoka, B. L. Smith, P. Agre, and A. Engel for AQPl. I would like to express sincere thanks to Dr. T. Walz for critical reading of the manuscript. Parts of this work (bR and AQPl) have been supported by the Japan Society for the Promotion of Science (JSPSRFTF9hL00502). I thank Macmillan Magazines, Ltd. (Nature Japan K. K.) for its permission to reproduce figures from papers published in Nature issues.

70

Y. FUJIYOSHI

REFERENCES L. Bragg, Proc. Camb. Phil. Sot., 17, 43 (1913). Green, V. M. Ingram, and M. F. Perutz, Proc. Roy. Sot., A225, 287 (1954). 3 J. Deisenhofer, 0. Epp, K. Miki, R. Huber, and H. Michel, Nature, 318, 618 (1985). 4 N. Uyeda, T. Kobayashi, K. Ishizuka, and Y. Fujiyoshi, Chemica Scripta, 14, 47 (1978). 5 Y. Fujiyoshi, T. Kobayashi, T. Ishizuka, N. Uyeda, Y. Ishida, and Y. Harada, Ultramicroscopy, 5, 459 (1980). 6 R. Henderson and P.N.T. Unwin, Nature, 257, 28(1975); P. N. T. Unwin and R. Henderson,]. Mol. Biol., 94, 425 (1975). 7 R. Henderson, J. M. Baldwin, T. Ceska, F. Zemlin, E. Beckmann, and K. H. Downing,J. Mol. Biol., 213, 899 (1990). 8 W. Kiihlbrandt, D. N. Wang, and Y. Fujiyoshi, Nature, 367, 614 (1994). 9 N. Unwin, J. MoZ Biol., 229, 1101 (1993). 10 N. Unwin, Nature, 373, 37 (1995). 11 T. W&z., T. Hirai, K. Murata et al., Nature, 387, 624 (1997). 12 Y. Fujiyoshi, T. Mizusaki, K. Morikawa et al., Ultramicroscopy, 38, 241 (1991). 13 0. Schemer, 1. Appl. Phys., 20, 20 (1949). 14 D. Siefermann-Harms, Biochim. Biophys. Acta, 811, 325 (1985). 15 W. Kiihlbrandt, Q. Rev. Biophys., 25, 1 (1992). 16 W. Kiihlbrandt and D. N. Wang, Nature, 350, 130 (1991). 17 R. Henderson, J. M. Baldwin, K. H. Downing, J. Lepault, and F. Zemlin, 1 W.

2 D. W.

Ultramicroscopy,

19, 147 (1986).

0. L. Krivanek and P. E. Mooney, Ultramicroscopy, 49, 95 (1993). M. Adrian, J. Dubochet, J. Lepault, and A. W. McDowall, Nature, 308, 32 (1984). K. A. Taylor and R. M. Glaeser, Science, 186, 1036 (1974). F. P. Booy and J. B. Pawley, Ultramicroscopy, 48, 273 (1993). H. J. Butt, D. N. Wang, P. K. Hansma, and W. Kiihlbrandt, Ultramicroscopy, 36, 307 (1991). 23 K. Mitsuoka, K. Murata, Y. Kimura, K. Namba, and Y. Fujiyoshi, Ultramicroscopy, 68, 109 (1997). 24 D. Oesterhelt and W. Stoeckenius, Nature New Biol., 233, 149 (1971). 25 S. 0. Smith, J. A. Pardoen, P. P. J. Mulder, B. Curry, J. Lugtenburg, and R. Math&, Biochemistry, 22, 6141 (1983). 26 N. Grigorieff, T. A. Ceska, K. H. Downing, J. M. Baldwin, and R. Henderson, 1. Mol. Biol., 259, 393 (1996). 27 Y. Fujiyoshi, K. Ishizuka, M. Tsuji, T. Kobayashi, and N. Uyeda, “Proc. 7th Int. Congr. on High Volt. Electron Microsc., Berkeley”, ed. by R.M. Fisher, R. Gronsky, and K.H. Westmacott, National Technical Information Service, Springfield, p. 21 (1983). 28 J. Baldwin and R. Henderson, Ultramicroscopy, 14, 319 (1984). 29 D. Oesterhelt and W. Stoeckenius, Methods Enzymol., 31, 667 (1974). 30 E. P. Peyroula, G. Rummel, J. P. Rosenbush, and E. M. Landau, Nature, 277, 1676 18 19 20 21 22

31 32 33 34

(1997). K. Mitsuoka, T. Hirai, K. Murata et al.,/. Mol. Biol., submitted (1998). M. Kates, N. Moldoveanu, and L. C. Stewart, Biochim. Biophys. Acta, 1169, (1993). M. S. Braiman, T. Mogi, T. Marti, L. J. Stern, H. G. Khorana, and K. J. Rothschild, Biochemistry, 27, 8516 (1988). R. H. Lazier, R. A. Bogomolni, and W. Stoeckenius, Biophys.]., 15, 955 (1975).

46

STUDY

35 36 37 38 39

A. K. Solomon,]. Gen. Physiol., 41, 243 (1957). and P. Agre, Proc. Natl. Acad. Sci. USA, 88, 11110 (1991). T. P. Carroll, W. B. Guggino, and P. Agre, Science, 256, 385 (1992). L. Smith, P. Agre, and A. Engel, EMBOJ., 13, 2985 (1994). Typke, B. L. Smith, P. Agre, and A. Engel, Nature Strut. Biol., 2, 730

40

V. W. Side1 G. M. Preston G. M. Preston, T. Walz, B. T. Walz, D. (1995). A. K. Mitra,

Strut. 41 42

OF

MEMBRANE

71

STRUCTURAL

PROTEINS

and

A. N. van Hoek, M. C. Wiener, 2, 726 (1995). and H. Li,j. Mol. Biol., 251, 413 and M. H. Saier, j. Memb. Biol.,

A. S. Verkman,

B. K. Jap J. H. Park

(1995). 153, 171

Received

EXPLANATION

PHOTO

and

M.

OF

1.

electron components

microscope. of the

through

the

The color stage, indicating

of the

for

(1996). publication

February

colored

superfluid

helium

in the figure progressively light

(purple

stage

of the

3rd

indicates

the

lowest

density

map

Extract

for

LHC-II.

map was sufficient to locate individual atoms of the molecules (Nature, 367, 616 (1994)). Structure of the LHC-II trimer as determined by electron

amino

PHOTO 3. end

on view

different PHOTO

(a) and

colors (by 4. Schematic

crystallographic

the side

courtesy side

view

(b),

protein, which are labeled are colored in green for

yellow

lutein.

for the 5. a4

helix

4. (b)

tide chain, for direct

Interaction is liganded

blue

bands

The

of chlorophyll by Glu65

shortest

the peptide coordination,

chlorophyll 6. Basic

gourd-shape collections fraction

LHC-II

and The acids

the approximate

experithe In

are drawn

monomer.

position

the

crystallography.

monomers

of an LHC-II

red

and

The

A to D, are represented by purple ribbons. chlorophyll a, light green for chlorophyll

indicate

of the lipid

distance carbonyl suggesting

molecule principle

represents a 2D of phase analyzing

molecules with the LHC-II polypeptide. in helix B which is charge-compensated between

the

center

of chlorophyll

of Gly78, is 4.5 .A. This that a water molecule

(by courtesy of electron

of Dr. W. crystallographic

crystallized protein from images and

patterns, respectively. Schematic drawings of the by :superfluid helium. (a) Cross

PHOTO 7. cooled

individual

of our

in

helical The

b, and bilayer

367, 618 (1994)).

P~uro rophyll

of the P~uro

The

the

of Dr. W. Kiihlbrandt). view of the 3D structure

regions of rhe chromophores

(Nature,

1998.

of the various the increase in

temperature

mental chlorophyll

of the electron

version

refers to the temperature higher temperatures with

PHOTO 2. 3D

17,

PHOTOS

Cross-section

the wavelength highest one).

the

Nature

Yeager,

Biol.,

beam and (b) parallel to it. The yellowish about 1OOK by liquid nitrogen. PHOTO 8. Extracts of our 3.0 A density

Kiihlbrandt). structure

and

parts the

the polypep-

determination.

Green and blue calculated from

latest cryo-transfer section perpendicular

map

a6 and

distance seems too long to allow might be involved in the binding

molecule. amplitudes

brown

(a) Chloby Arg185 in

device for the to the path are

atomic

The

refer to data electron dif-

top entry stage of the electron

cooled

to a temperature

model

of bR

map. (a) Density for the retinal which is located almost in the center of the (b) The side-chain densities for Tyr185 and Phe208 with a center-to-center the aromatic rings of 5 A are clearly resolved.

fitted bR

of into

molecule. distance

the of

72

Y.

9.

PHOTO

Density

glycerophosphate lipid viewed at 1.5 (T and analysis

of

a lipid

monomethyl from different 3 (T, respectively.

to be of a good

quality,

even

interactions molecule

helix E of the adjacent hydrophobic helix-helix

bR

lipid

molecule

be seen in the by hydrophobic

PHOTO 12. Surface side of the bR trimer. respectively.

Due

representation can clearly PHOTO 13. Asp1

15. The

molecule. contacts,

map

The however, extracellular in the

and

views red

chains

(Nature,

389,

modeled

unidirectional

in blue

207

of the

as

interactions the cytoplasmic B and

helix

a phosphatidyl

molecules. between between

No

side and

used one

(a, b) the

adjacent

monomer

bilayer. The LHC-II just outside of the lipid

only

helix helix

bR

and (b) negative

is held

of the Schiff

base,

was calculated

using

lines which

was calculated are strongly

using affected

a

can together

the extracellular surface charges, of the

three

lines

are

molecule

interactions

trimer bilayer.

for the calculation

opening retinal

B and B and

between the monomers leaflet of the bilayer

E of the

leaflet. LHC-II trimer.

illumination

contour

lipid

(a) Interface (b) interface

of (a) the cytoplasmic areas indicate positive

3.0 A while the map shown in red contour to 3.0 A, omitting the low-resolution data side

major in helix

the program GRASP, (marked by arrowhead). maps of bR around shown

regions

core of the lipid of the monomers

rendered Blue

to the

with be seen Potential

in the

between

in the interactions

hydrophobic contacts

was

in the bR trimer. in the bR trimer and

is inserted

which has no counterpart PHOTO 11. Monomer

which

ester with dihydrophytol chains. (a) and (b) show the same directions. The blue and red lines represent maps contoured The very similar features seen in the two maps suggest our

PHOTO 10. Monomer helix D of an adjacent

single

molecule,

FUJIYOSHI

proton

surface channels

(c) Asp96,

all data

from

only data from by the charge

and

(d)

54 A to 7.0 A of the

(1997)).

PHOTO

14. Top view on the 6 A 3D density map of AQPl as determined by electron crystallography of reconstituted 2D crystals. The central square shaped AQPl tetramer is seen from its extracellular side while the four adjacent tetramers are viewed from the cytoplasmic

side.

PHOTO

Slice through an AQPl monomer densities surround a central,

tinct cannot

15. rod-like

be assigned to a simple 3D structure and

PHOTO 16.

helices, surrounding E, form a right-handed integral

membrane

the protein

central a-helical (by

parallel branched

to the density

transmembrane a-helix. projection structure of AQPl. density which was assigned bundle. This has never courtesy

of Dr.

A. Engel).

membrane (marked The

plane. Six disby ‘IX”) which

six transmembrane

to the functional been observed

loops before

CZB and for an

STRCCTLRAL

STUIIY

OF

Photo 1 Specimen

------

Chamber

Exhaust

hIEMBRANE

/

window

-\

73

PROTEIUS

-Liquid

uid

rmal Flexible

Shield support Specimen stage Capillary Pole piece

Photo 2

Photo 3

/ // /

I/

Np

tank

He

tank

shield copper

Thermal shield Exhaust pipe -pot support - Pot -Specimen

holder

Y. FUJIYOSHI

74

Photo 4

Photo 5

I

I

STRUCTt:HAL

STUDY

OF

., : ,’ I..’ ,. : .@:’ ,:,’

.::.;

.‘_

MEMRRA\IE

PROTEIXS

75

76

Y. FUJIYOSHI

Photo 7

a

Copper

Specimen

exchange Transfer

bloc

chamb rod

Specimen

chamber

I Guide Pre-evacuation

Valve chamber

Specimen

holder/

Transfer

of cryo-transfer

for rapid-freezing

device technique

Specimen

b

Flexible cooling

Vitlve

holder

Cross-section Plunger

’

copper braid holder chuck

chamber

0

rail

STRUCTLRAI,

Photo 8

S’I’UI)Y

OF

MERIBRATE

PROTEINS

77

78

Photo 9

‘hoto 10 !

Photo 11

Y. FUJIYOSHI

STRUCTIJRAL

Photo

12

STUDY

OF

MEMLIRRANE

PROTEINS

79

80

Y. FGJIYOSHI