Добавил:
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:

Neutron Scattering in Biology - Fitter Gutberlet and Katsaras

.pdf
Скачиваний:
62
Добавлен:
10.08.2013
Размер:
10.95 Mб
Скачать

70 M. Blakeley et al.

 

100

 

 

 

 

 

 

 

 

 

 

 

 

 

 

concanavalin A

 

 

 

 

 

 

 

cubic structure

3

 

 

 

 

Haemoglobin

 

 

 

 

 

 

 

 

 

 

)/mm

10

 

 

 

HEWL

concanavalinA

 

 

 

 

 

 

 

 

 

 

TEWL

3

 

 

Vs

β-cyclodextrin

B12

5 mm crystalvol.

 

 

 

(

 

 

 

x

 

 

volume

 

 

 

 

γ-crystallin

Phasiolinvulgaris

 

 

 

 

 

 

 

 

1

 

 

 

Parvalbumin

concanavalinA

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

specimen

 

 

 

 

 

Porin

PRCviridis

TBSV

 

 

 

 

 

 

 

 

 

 

 

 

nucleosome

 

0.1

 

 

 

 

PRC

 

 

 

 

 

 

 

spheroid matrix STNV

t-RNA

 

 

 

 

 

 

porin

 

 

0.01

 

 

 

 

 

 

 

 

 

103

104

105

106

107

108

 

 

 

 

 

 

3

β-crustacyanin

 

 

 

 

 

cell volume (Vc )/Å

0.005mm3

 

Fig. 4.3. Scatter plot of the examples of protein crystals investigated at ILL on the monochromatic neutron instrument D19 (filled circles) and DB21 (filled squares). The line through the origin shows the empirical limit and so the open circles show where LADI is capable. The three case studies of this paper are highlighted as marked. DB21 o ers data collection at very low resolution (e.g., the TBSV experiment marked involved data collected to 16 ˚Aresolution) in order to map out crystal packing details whereas LADI is used for high resolution neutron protein crystallography i.e. 1.5 to 2.5 ˚Aresolution (as was D19 before it)

Thus the discovery of new pharmaceuticals and of enhanced e ciency compounds would then accelerate. Also, because neutrons are nondestructive, unlike X-rays, room temperature structure and vibration data can be provided, and which is the relevant temperature physiologically. Our most recent work however also shows that nPX data can be measured down to 15 K with benefits in the clarity of the nuclear density seen. Thus future approaches appear to be able to include structures being determined and refined with nPX from 15 K up to room temperature including freeze trapped protein crystal structures. Most recently Hazemann et al. [18] report their use of a radically smaller crystal volume of 0.15 mm3 using fully deuterated aldose reductase protein in a 2.2 ˚A high resolution analysis. This was achieved due to a stronger crystal scattering e ciency for neutrons and also eliminating the hydrogen incoherent scatter. Clearly this is an exciting development. A conceptual design report (CDR) for the new MANDI SNS di ractometer for protein crystallography is now in press [19], and includes an estimate of a few days only to measure a complete high resolution data set from a fully deuterated protein crystal of 0.125 mm3 total volume, and also the CDR for the new JPARC protein crystal di ractometer in Japan is recently published [20]. These two CDRs continue a fascinating new wave of instrumentation for this field of research.

4 Neutron Protein Crystallography

71

+

+ +

Fig. 4.4. The large molecular weight frontier. LADI di raction recorded from a crystal of cubic concanavalin A where there is 50 kDa in the asymmetric unit of the crystal (from [6]). The di raction extends to 3.5 ˚A thus new sources, as well as perdeuteration opportunities will extend the technical capability to reach highresolution structure analyzes even in such cases. The actual LADI pattern is shown (top) and the predicted Laue pattern (bottom) using the Daresbury Laue software analysis package [16] modified for neutrons and a cylindrical detector geometry [17]

Acknowledgements

JRH is very grateful to his past and present colleagues and students in the University of Manchester Structural Chemistry Research Laboratory, which is devoted to biological and chemical crystallography and includes extensive crystallographic methods development. The research grant support of BBSRC, The Wellcome Trust and The Leverhulme Trust and of EPSRC,

72 M. Blakeley et al.

BBSRC, British Council, the UK/Israel Fund and the Institut Laue Langevin for PhD studentship support and the EU (for Host Institute, Network and Marie Curie Training Centre awards) are all also gratefully acknowledged. The CHESS SR Facility Cornell, USA, ESRF Grenoble and SRS Daresbury and the Institut Laue Langevin in Grenoble are thanked for SR and neutron beamtime, respectively, for the studies described herein.

References

1.F. Shu, V. Ramakrishnan, B.P. Schoenborn, PNAS (USA) 97, 3872–3877 (2001)

2.D.W.J. Cruickshank, J.R. Helliwell, K. Mo at, Acta Crytallogr. A43, 656–674 (1987)

3.J.R. Helliwell, C. Wilkinson, X-ray and neutron Laue di raction, in Neutron and Synchrotron Radiation for Condensed Matter Studies: Applications to Soft Condensed Matter and Biology (Springer Verlag, Berlin, New York, 1994)

4.S. Arzt, J.W. Campbell, M.M. Harding, Q. Hao, J.R. Helliwell, J. Appl. Crystallogr. 32, 554–562 (1999)

5.N. Niimura, T. Chatake, K. Kurihara, M. Maeda, Cell Biochem. Biophys. 40, 351–370 (2004)

6.A.J. Kalb (Gilboa), D.A.A. Myles, J. Habash, J. Raftery, J.R. Helliwell, J. Appl. Crystallogr. 34, 454–457 (2001)

7.J. Habash, J. Raftery, R. Nuttall, H.J. Price, C. Wilkinson, A.J. Kalb (Gilboa), J.R. Helliwell, Acta Crystallogr. D56, 541–550 (2000)

8.A. Deacon, T. Gleichmann, A.J. Kalb (Gilboa), H.J. Price, J. Raftery, G. Bradbrook, J. Yariv, J.R. Helliwell, J. Chem. Soc. Faraday Trans. 24, 4305–4312 (1997)

9.D.W.J. Cruickshank, J.R. Helliwell, L.N. Johnson, Time-Resolved Macromolecular Crystallography (OUP, Oxford, 1992)

10.M. Blakeley, A.J. Kalb (Gilboa), J.R. Helliwell, D.A.A. Myles, Proc. Natl. Acad. Sci. USA 23, 16405–16410 (2004)

11.J.R. Helliwell, Time-Resolved Chemistry, Faraday Trans. 122 (2003)

12.P. Langan, G. Greene, B.P. Schoenborn, J. Appl. Crystallogr. 37, 24–31 (2004)

13.The European Spallation Source: Technical and Science Case (ESS Project Books, J¨ulich, 2002)

14.T. Bayerl, O. Byron, J.R. Helliwell, D. Svergun, J.-C. Thierry, J. Zaccai, in The European Spallation Source: Science Case (ESS Project Books, J¨ulich, 2002)

15.M. Cianci, P.J. Rizkallah, A. Olczak, J. Raftery, N.E. Chayen, J.R. Helliwell, PNAS (USA) 99, 9795–9800 (2002)

16.J.R. Helliwell, J. Habash, D.W.J. Cruickshank, M.M. Harding, T.J. Greenhough, J.W. Campbell, I.J. Clifton, M. Elder, P.A. Machin, M.Z. Papiz, S. Zurek, J. Appl. Crystallogr. 22, 483–497 (1989)

17.J.W. Campbell, Q. Hao, M.M. Harding, N.D. Nguti, C. Wilkinson, J. Appl. Crystallogr. 31, 496 (1998)

18.I. Hazemann, M.T. Dauvergne, M.P. Blakeley, F. Meilleur, M. Haertlein, A. Van Dorsselaer, A. Mitschler, D.A.A. Myles, A.D. Podjarny, Acta Crystollogr. D 61, 1413–1417 (2005).

19.A.J. Schultz, P. Thiyagarajan, J.P. Hodges, C. Rehm, D.A.A. Myles, P. Langan, A.D. Mesecar, J. Appl. Cryst., in press.

20.I. Tanaka, T. Ozeki, T. Ohara, K. Kurihara, N. Niimura, J. Neutr. Res. 13 (2005), 49–54.

5

Detergent Binding

in Membrane Protein Crystals by Neutron Crystallography

P. Timmins

5.1 Introduction

The structures of membranes and membrane proteins are taking on ever increasing importance since the observation that they represent perhaps >30% of the proteins encoded by the human genome but much less than 1% of the structures determined to close to atomic resolution by X-ray crystallography. The di culties encountered in X-ray crystallographic structure determination have been so serious that the first structure of a membrane protein to be solved at even medium resolution was by electron di raction. This was the famous case of the purple membrane from Halobacterium salinarum where the structure of the bacteriorhodopsin embedded in its natural membrane was solved. Due to its unique crystallinity in the natural membrane, it is still the only whole membrane structure to have been solved.

The first membrane proteins to be crystallized were the photo-reaction center from Rhodopseudomonas viridis and OmpF porin from E. coli in the early 1980s. Since then a number of other membrane proteins have been crystallized and their structures solved by X-ray crystallography but these remain much less than 1% of the total of soluble proteins. A particularity of practically all membrane protein structures is that they are crystallized from detergent solubilized proteins and the detergent becomes an integral part of the crystal. Although the protein itself is well ordered and its structure can be obtained at high resolution, the detergent used to solubilize the protein is fluid and disordered and hence invisible in X-ray maps. Neutron crystallography, although up to now unable to locate high resolution features due to insu ciently large crystals, is ideally suited to locate the detergent phase and hence provide information on crystal packing and also on detergent–protein interactions analogous to lipid–protein interactions in the real membrane.

5.2 Advantages of Neutrons

Membranes can be studied either in their native state where the principal components are protein and lipid or in a solubilized state where the components

74 P. Timmins

are protein and lipid or protein and detergent. Each of these components has a characteristic scattering length density for neutrons and each may in principle be deuterium labelled allowing that scattering length density to be varied. Replacement of water by heavy water allows a kind of isomorphous replacement which facilitates phasing in one dimensional di raction.

In fact the whole concept of contrast variation originated from experiments by Bragg and Perutz [1] to try and determine the overall shape of haemoglobin using salt solutions to modify the electron density of the crystal solvent with respect to the protein. The natural contrast between protein and water is very low for X-rays; the electron density of RNAse for example is 0.432 e˚A3 and for pure water 0.335 e˚A3. The addition of salt to water raises the electron density even closer to that of protein. Small molecules such as sucrose or indeed high salt concentrations may be used to vary the contrast but the chemical e ects of such additives are often such as to destabilize the crystal or even the complex itself or alter the conformations of the component molecules. The exchange of deuterium for hydrogen is a much less perturbative change. In addition it is not trivial to collect low resolution X-ray di raction data from crystals of biological macromolecules although a number of workers have built specialized beam-lines to do this [2, 3].

Manipulation of contrast in neutron scattering is particularly straightforward and has been heavily exploited for many years in neutron small angle scattering. The presence of a well-defined solvent (water) phase in crystals of macromolecular complexes means that the same principles may be applied in neutron crystallography – hence the term small-angle neutron crystallography. Figure 5.1 shows the neutron scattering length density for a number of chemically distinct components of biological molecules. These are calculated from the atomic composition of average proteins, nucleic acids etc. and do not vary greatly between for example di erent proteins. The values shown for detergents and lipids may vary considerably, however, depending particularly on the balance between hydrophilic head and hydrophobic tail.

At low resolution the scattering is due to the contrast between solvent and macromolecule, i.e., the di erence in scattering length density. Thus for example proteins have a zero scattering length density di erence with respect to water in a solution containing 40% D2O/60% H2O. This concentration at which a particle has zero contrast is known as the isopicnic point. It should be noted that the concept of zero contrast applies only to the average scattering length density and hence only to scattering at Q = 0. At all other Q-values there will be some contributions from internal scattering length density fluctuations within the molecule such that there are always some parts of the molecule that have a positive or negative contrast. Hence the Bragg scattering is a minimum at the isopicnic point but is not zero. A striking point of Fig. 5.1 is that the scattering length densities of H2O and D2O encompass those of all other natural components of biological macromolecules. The only exception to this is the case of fully deuterated protein or nucleic acid which have a scattering length density greater than that of pure D2O.

)

2 cm

10

lengthScattering (10density

5 Membrane Protein Crystals by Neutron Crystallography

75

9.0

8.0

D-DNA

7.0

D-protein

6.0

5.0

DNA

4.0

RNA

3.0

2.0

Protein

water

1.0

detergent or lipid

0.0

1.0

0

10

20

30

40

50

60

70

80

90

100

 

 

 

 

 

% D2O

 

 

 

 

 

Fig. 5.1. Neutron scattering length densities of biological macromolecules as a function of the deuterium content of the solvent water

5.3 Instrumentation and Data Reduction

One of the di culties encountered in low resolution protein crystallography is the problem of measuring data at very low resolution – including for example the first order of a 700 ˚A unit cell. Hence, cold neutrons and H/D contrast variation can be combined to measure neutron di raction data in a rather simple way. A dedicated di ractometer was constructed at ILL some years ago as collaboration between ILL and EMBL. This instrument, DB21, is a four-circle di ractometer with multidetector and is described in some detail in Roth et al. [4, 5]. A schematic representation of the instrument is shown in Fig. 5.2

The beam is monochromated by a potassium intercalated graphite crystal giving a neutron wavelength of 7.56 ˚A and a wavelength spread of 2% (FWHM) or a pyrolytic graphite crystal giving a wavelength of 4.6 ˚A. The beam is collimated using LiF pinholes and a graphite collimator after passing a series of graphite filters for eliminating λ/2, λ/3, and λ/4 contamination. The detector is of the Anger camera type allowing a resolution of 1.75 ×1.53 mm. Due to the γ-sensitivity of this detector no cadmium is used in beam defining apertures

 

Multidetector

 

 

 

Beam stop

 

 

 

Sample

Filter

 

 

 

 

 

 

Neutron beam

 

 

 

Monitor

 

 

Alignement telescope

Pyrex window

 

76

P. Timmins

Shielding

 

 

Argon filled

 

 

 

flightpath

Bi single crystal

Monochromator

Eulerian cradle (sample orientation mechanism)

Sapphire window

Secondary shutter

Fig. 5.2. Schematic representation of the DB21 instrument

that are instead made of LiF backed with B4C. A single crystal of Bi removes in-beam γ-radiation.

Data acquisition is by φ- scans or ω-scans. In low symmetry space groups reorientation of the crystal may be required to obtain a full data step due to the rather restricted reciprocal space coverage of the current detector. Determination of the crystal orientation may be di cult at low resolution and many of the standard autoindexing algorithms have failed. A program based on the graphics package ‘O’ has been written to allow manual rotation of the known reciprocal cell into the reciprocal space di raction pattern in order to determine the crystal orientation [6]. Data reduction is carried out using a modified version of XDS [7]. Another problem that arises is the scaling of data measured from crystals of di erent H2O/D2O content. This is done by using the parabolic relationship between intensities or the linear relationship between structure amplitudes in centric zones [8] as described in Eq. 5.2. This of course requires consistent indexing which in certain enantiomorphic space groups may be a problem. For example in space group P312 it is impossible to distinguish between hkl and khl reflections. In high resolution data this problem can be resolved when the final structure is known but in the neutron low resolution case it is absolutely necessary to carry out a correct scaling. In practise it may be necessary to perform the scaling using all possible combinations of index and to select that having the best scaling for all reflections.

5.3.1 The Crystallographic Phase Problem

The variation of the crystallographic structure factor as a function of contrast can be expressed as:

5 Membrane Protein Crystals by Neutron Crystallography

77

F (h, X) = F (h, 0) + XF (h)HD,

(5.1)

where h is the reciprocal lattice point, X is the mole fraction of [D2O]/[D2O]+ [H2O] in the crystal, F (h)HD is the vector di erence between the structure factor in H2O and that in D2O. Multiplying by the complex conjugate we obtain the di racted intensity:

I(h, X) = F (h, 0)2 + 2X cos φF (h, 0)FHD(h) + X2FHD2 (h),

(5.2)

where φ is the phase angle between F (h, 0) and F (h)HD.

This relationship has several important consequences for low resolution crystallography including, as mentioned above, the possibility of scaling together data from di erent contrasts and the interpolation of missing data [8]. In terms of structure solution it is of fundamental importance as it means that the phase di erence, φ between any two contrasts, of a reflection h, may be determined except for the sign ±, if the amplitudes at 3 contrasts are known. Therefore if the structure is known at any one contrast then the phases at that contrast may be calculated and then determined at any other contrast except for knowledge of the sign. In the particular case of centrosymmetric reflections where φ = 0 or π then there is of course no ambiguity and the phase may be calculated at any contrast [9].

In most studies carried out to date the structure of one component of the macromolecular complex has been determined by X-rays or could be modelled from other information. Hence structure factors calculated from the known part of the structure at a contrast where the other component is visible provide starting phases for the determination of the structure at any contrast. This is illustrated in Fig. 5.3 which demonstrates the vector relationships between structure factors at four di erent contrasts. The two triangles bounded by F0, FHD, and FD are the two possible relationships which can be constructed through knowledge of the structure factor amplitudes alone following Eq. 5.1, and corresponding to the two possible signs of φ. This particular figure illustrates the case of, for example, a protein/detergent complex where data would be measured at 40% D2O where the protein is invisible, 10% D2O where the detergent is invisible and three other contrasts, 0%, 70%, and 100% D2O. In this case we imagine that the protein structure is known and that the detergent structure is to be determined. We may therefore calculate the phase of the structure factor in 10% D2O and thus determine the orientation of the phase triangle with just the ambiguity of sign corresponding to the two triangles shown. This is very closely analogous to the situation of single isomorphous replacement [10, 11] in X-ray protein crystallography. Once this (ambiguous) phase has been determined then the ambiguity may be resolved and an approach to the true phase may be made using density modification based on constraints such as the invariability of the known part of the structure, solvent flattening or noncrystallographic symmetry averaging [12, 13].

78 P. Timmins

Imaginary

 

 

F 0

F 10

 

 

 

 

 

F 70

F 40

 

F

0

 

 

 

F 100

F 40

Real

F 70

F 100

Fig. 5.3. Vector diagram illustrating the relationship between structure amplitudes and phases at di erent contrasts. The two vector triangles are oriented arbitrarily with the FHD vector parallel to the real axis

5.4 Comparison of Protein Detergent Interactions in Several Membrane Protein Crystals

Because of the very di erent scattering lengths of protein and detergent and in particular the very high contrast obtained between detergent and D2O, complexes of protein and detergent make ideal objects for study by neutron contrast variation. X-ray crystallographic studies have to date succeeded in resolving the structure of a number of membrane proteins and in some cases individual tightly bound detergent or residual lipid molecules have been observed but in no case has this method been able to visualize the solubilizing belt of detergent. Using the known X-ray structure and neutron di raction data measured at a number of H2O/D2O component contrasts it has been possible to calculate neutron scattering length density maps which show only the detergent. The key factor here is the detergent content of the cell. The volume in which the protein is located is known from the X-ray structure but generally speaking the amount of detergent in the crystals is unknown. It is therefore necessary to estimate this and to perform solvent flattening and density modification as a function of the detergent content as a variable parameter [13].

5 Membrane Protein Crystals by Neutron Crystallography

79

5.4.1 Reaction Centers and Light Harvesting Complexes

The first membrane protein/detergent structures to be studied by low resolution neutron crystallography were the photosynthetic reaction centers of Rhodopseudomonas viridis [14] and Rhodobacter sphaeroides [13]. In the reaction center from R. viridis only one molecule of the detergent N ,N ’- dimethyldodecylamine-N -oxide (LDAO) could be visualized in the high resolution X-ray maps. The neutron di raction results show the reaction center to be surrounded by a detergent belt some 25–30 ˚A thick. This corresponds to roughly twice the length of an extended LDAO molecule. The aliphatic chain of the detergent molecule observed in the X-ray maps falls within the neutron density although its polar head is outside. The lack of density corresponding to the head may be due to its small size compared with the resolution of the data and to its rather low scattering density arising from disorder and hydration. As Roth et al. point out it should also be noted that in the case of the reaction center the detergent does not necessarily mimic the biological membrane as in the real membrane the reaction center is surrounded by and makes contact with several light harvesting molecules.

The photoreaction center from R. sphaeroides crystallizes in the presence of n-octyl-β-glucoside (β-OG) and the small amphiphile heptane-1,2,3-triol (HP). Michel [15] was the first to use small amphiphiles in membrane protein crystallization and suggested that they had the e ect of reducing the size of detergent micelles and thereby facilitating crystallization of the membrane protein/detergent complex. This e ect was confirmed in neutron small angle scattering experiments by Timmins et al. [16] which showed that HP was indeed included in micelles at least of LDAO and decreased their radius of gyration. Gast et al. [17] also showed by turbidity measurements that addition of 5% HP decreased by a factor of two the amount of LDAO bound to the reaction center of R. viridis. The most striking observation in the neutron di raction results on the R. sphaeroides reaction center was that the detergent ring formed by the β-OG was almost identical and size and shape to that formed by the LDAO around the R. viridis reaction center. Given the similarity in total length of β-OG and LDAO it would appear that detergent size/geometry plays a key role in the crystallization of membrane proteins.

As mentioned above the reaction centers are usually found in the membrane closely associated with several copies of light harvesting proteins and therefore the protein–detergent interactions observed in the crystal may not always be representative of protein lipid interactions in the membrane. Recently a study has been published on the light harvesting complex LH2 from the photosynthetic purple bacterium Rhodopseudomonas acidophila, a nonameric complex with a central hole which in vivo contains membrane lipids [18]. The complex was crystallized from solutions containing β-OG and experiments were performed using both hydrogenated and tail-deuterated detergent. As well as demonstrating the presence of a detergent ring as with other membrane proteins the experiments also showed the central hole to be