Historically, the use of neutrons in protein crystallography has been a minority interest due to the need for large crystals to overcome the low flux of neutron sources and the weak interaction of neutrons with matter. For solving the structure of a protein a typical crystal size for neutron single crystal diffraction has been 1 mm³ (i.e. 1 x 1 x 1 mm³). Meanwhile, neutrons have been developed as a powerful tool for solving the structures of small molecules (e.g. drugs and superconductors) by powder and single crystal diffraction, complemented by their magnetic stucture retrieved using polarized neutrons and muon sources. For biological systems the interest in neutron radiation for structural studies arises, as usual, from the anomalously strong and positive interaction between neutrons and deuterium nuclei compared with hydrogen nuclei. ²H₂O can easily be exchanged for ¹H₂O in the crystal mother liquor (and thereby ²H for ¹H at labile sites on the protein) without adversely affecting crystal stability.

For medium- to high-resolution data (1-3 Å), the strong scatter of neutrons from ²H nuclei means that whole water molecules can be observed more frequently in nuclear density maps obtained using neutrons than they are in electron density maps obtained using X-rays. With X-rays complete water molecules are only observed when well-ordered and with data of ~1 Å resolution. More often, a single density peak is observed in X-ray data, providing only locational and not orientational information and therefore only a partial picture of the electrostatic nature of the protein surface and cavities. In any case, water molecules beyond the first hydration layer of a macromolecule will in general not be detected using X-rays since components of outer shells are less well-ordered in location or orientation so that their density is averaged out. With neutrons the strong scatter from ²H₂O means that the less well-ordered waters are still observed in the density map as single peaks. Single density peaks resulting from ²H₂O may arise from a single molecule with some orientational or positional disorder; or from several molecules averaged together. Neutron diffraction also provides a way of determining the positions of polypeptide ¹H atoms from their negative density; however the strong incoherent scatter from ¹H introduces significant noise into the data, leading to interference between e.g. the negative density associated with hydrogen and the positive density of an associated carbon atom. This problem can be circumvented by production of fully-deuterated samples (see below).

At low resolution (10-20 Å) neutron diffraction can provide information concerning the structure of detergent or lipid components of a crystal or the shape of the solvent envelope of the protein. Due to dynamical disorder it will usually only be possible to determine the detergent or lipid structure of a crystal at such resolutions. At 10-20 Å resolution the signal in a diffraction experiment depends on the contrast of the molecules in the crystal against the (solvent) background. With neutron diffraction the contrast of different components of the crystal matrix may easily be selectively enhanced by varying ²H₂O/¹H₂O ratio in the crystal. Contrast variation is also the route for determining a solvent envelope for a protein within the crystal, which may be an important tool in the phasing of diffraction data. For these kinds of low resolution experiments much smaller crystals (of about the same size as those regularly used by X-ray crystallographers) are useful compared to high resolution studies such as those mentioned above. Hence low-resolution neutron diffraction should be a useful weapon in the crystallographer's armoury when contending with the phasing and interpretation of X-ray data.

An excellent recent review on neutron crystallography is Gutberlet et al. (2001) Acta Cryst. D57, 349-354. Link to PDF.

Historically, crystallographic studies with neutrons have been confined to monochromatic beamlines. DB21 at the ILL is the diffractometer which has been used for most studies of this nature. D19 at the ILL will soon be updated with a much increased detector size, providing a 20-30 times increase in the effiency with which data can be collected due to the ability to measure a large portion of it simultaneously. This will open up new areas of research in crystallography of small proteins.

Case study: the detergent structures of porin crystals
The OmpF porin from Eschericia coli is an integral membrane channel-forming protein which spans the outer membrane in Gram-negative bacteria. It is a member of a family of large beta-barrel proteins, structures for a number of which have been determined crystallographically. The structure of OmpF (PDB ref 1OPF) is shown below.

OmpF was crystallized in the presence of detergents, mimicing the hydrophobic environment of the membrane. The low-resolution neutron crystal structures of both tetragonal and trigonal crystal forms revealed the detergent structures in each case. This information can be interpreted in terms of the regions of the protein in contact with the membrane in vivo and provide information on the crystallization mechanism of the protein in each case and how the differential ordering of events leads to differing crystal forms. The work was carried out by E. Pebay-Peyroula, S. Penel, R. M. Garavito, J. P. Rosenbusch, M. Zulauf, G. Rummel, T. Schirmer and P. A. Timmins ((1995) Structure 3, 1051-1059; (1998) Biochimie 80, 543-551). This figure and all the other molecular structures shown in these pages was drawn using BOBSCRIPT (Kraulis, P. J. (1991) J. Appl. Crystallogr.24, 946-950; Esnouf, R. M. (1997) J. Mol. Graph. 15, 132-134) and rendered in RASTER3D (Merritt, E. A. and Murphy, M. E. P. (1994) Acta Crystallogr. D50, 869-873).
 

Case study: structures of whale myoglobin
The structures of whale myoglobin in its native form and as a carbonmonoxy (reversibly-inhibited) complex have been solved by neutron crystallography. The two structures are shown below and reveal locational and orientational information for water molecules in the first hydration shell of the protein. The native structure was solved for a fully deuterated sample, which permitted the location of polypeptide hydrogen atoms (as deuterium atoms) allowing e.g. determination of the ionization state of histidines. Perdeuteration of the sample was essential in this aspect of the structure since buried histidines are restricted in their capacity for exchange with mother liquor ²H₂O. A molecular dynamics simulation  (Biophys. J. (2000) 79, 2966-2974) of myoglobin appeared to locate 294 hydration sites that could be shown to agree reasonably well with the positions determined from neutron and X-ray crystallography. Their residence time distribution was analyzed, suggesting that particularly long water residence times on the protein are found only in its cavities and clefts. This suggests that factors like hydrogen bonding and the hydrophobicity of residue sidechains play a relatively minor role in determining the residence times of long-lived hydration sites.

Neutron crystallographic structures of sperm whale myoglobin as a carbonmonoxy complex at a resolution of 1.8 Å (left PDB ref 2MB5) and the native structure at 2.0 Å (right PDB ref 1CQ2). The alpha-carbon backbone is shown as a ribbon with the haem groups in blue and red respectively. The ²H₂O molecules resolved in the structures are shown with grey deuterium atoms and yellow oxygen atoms. The carbonmonoxymyoglobin structure was solved by X. Cheng and B. P. Schoenborn ((1990) Acta Crystallogr. B46, 196 et seq.). The native structure was solved by F. Shu, V. Ramakrishnan and B. P. Schoenborn ((2000) Proc. Nat'l Acad. Sci. USA 97, 3872-3877). In each case the data were obtained using monochromatic neutron radiation.

Case study: an important neutron structure of lysozyme
Much more information on the water arrangement around a protein has been obtained for lysozyme using Laue diffraction.

Neutron crystallographic structure of hen egg-white lysozyme (PDB ref 1LZN) obtained at 1.7 Å resolution using the Laue neutron diffractometer LADI at the ILL in Grenoble. Orthogonal views are shown with the alpha-carbon backbone coloured cyan, oxygen atoms red and deuterium atoms yellow.
 

The use of Laue radiation much increases the effective flux at the crystal and hence the signal-to-noise ratio of the diffraction data. However, the interpretation of the data is simultaneously complicated by the measurement of reflections from neutrons of varying wavelengths. The Laue diffraction was measured using a novel purpose-designed neutron imaging plate (for review see (1999) Curr. Op. Struct. Biol. 9, 602-608) and its interpretation was greatly assisted by being able to solve the structure by molecular replacement. This permitted the water structure of the crystallized protein to be solved, including the whole first hydration layer. 214 out of 268 labile hydrogen atoms were substituted with deuterium atoms at the protein surface, including a single aliphatic hydrogen (His51 C^E1). 115 water molecules were completely observed in the density map and 129 were incompletely observed; complete molecules were modelled as ²HO²H, incomplete molecules as a single O. Eight water clusters extending away from the protein surface but anchored by single, tightly bound ²H₂O molecules were observed. In addition, strings of alternating ordered and disordered hydration sites extended over the protein surface, suggesting the presence of a dynamic hydration layer with ionic "flip-flop" occurring between bound waters. This suggestion is  supported by quasi-elastic neutron scattering (QINS) measurements, which indicated that 75 +/- 5 % of the water molecules in the unit cell existed in a water network formed by fluctuating hydrogen bond patterns. These results support models for the interaction of water with proteins based on simulation studies (Proteins: Struct. Funct. Genet. 18, 133-147). The innermost hydration layer is more densely packed than bulk solvent, which agrees with SANS measurements made by Zaccai, Svergun and colleagues ((1998) Proc. Nat'l Acad. Sci. USA95, 2267-2272). Furthermore, the first hydration layer of residues with apolar sidechains adopted a cage-like structure raised some 4-5 Å above the surface of the protein. The existence of such structures has also been inferred from SANS measurements by Pertsemlidis and colleagues ((1999) Proc. Nat'l Acad. Sci. USA 96, 481-486). The lysozyme structure is described in a fine and detailed paper by C. Bon, M. S. Lehman and C. Wilkinson ((1999) Acta Crystallogr. D55. 978-987).

The results obtained by Bon et al. can be compared very interestingly with the work of Nakasako on beta-trypsin ((1999) J. Mol. Biol. 289, 547-564). Nakasako performed X-ray diffraction experiments at 293 and 100 K and observed much of the first hydration layer of trypsin but nothing beyond it. All the waters were, furthermore, represented by single density peaks and therefore orientations could not be modelled for them. More waters were visible at 100 K compared to 293 K, which was as expected given the reduced thermal energy present at liquid nitrogen temperatures. Concerning the more fundamental insights discussed above, Nakasako also observed chains of water molecules extending over the protein surface and cages around hydrophobic residues. Nakasako also discusses the work of Zaccai, Svergun and colleagues on the nature of the first hydration layer and the signal derived from it in diffraction and scattering experiments. A later molecular dynamics study of hydrated lysozyme produced results in silico in excellent agreement with those obtained experimentally, again indicating that hydration water is more dense than bulk water; partly due to the geometrically different environment of a protein surface and partly due to conformational changes in a shortening of the average water O-O distance and an increase in coordination number. Denser water on average was found in cavities on the protein surface, which is a finding also in agreement with experimental data. Proc. Nat'l Acad. Sci. USA (2002) 99, 5378-5383.

The structure of lysozyme determined using neutron Laue diffraction was intended to demonstrate the kind of experiments now possible using neutron crystallography. A Laue source speeded data capture up by an order of magnitude . The data were collected from a single crystal, which was subseqently used for X-ray diffraction to complement the neutron data. However, in this case the protein crystal used was very large. Studies providing a similar amount of information to that obtained in the lysozyme work will continue to rely on large crystals (and smallish unit cells of ~70 x 70 x 70 Å) unless the fluxes of neutron sources can be considerably increased. The technical advances manifest in this result nevertheless open the way to neutron studies of larger systems where a whole structure may not be retrievable but where useful information on hydration and ionic exchange can now be obtained. Such a study on concanavalin A has been published (J. Chem. Soc. Faraday Trans. (1999) 93, 4313-4317). Another example of the use of the Laue neutron diffraction approach has been the determination of the structure of the coenzyme cob(II)alamin in parallel at 0.90 Å resolution using X-rays, 1.0 Å resolution using monochromatic neutrons and 1.43 Å using neutron Laue radiation (Acta Crystallogr. (1999) D55, 51-59).

Other studies described:
J. B. Cooper and D. A. A. Myles (Acta Crystallogr. (2000) D56, 246-248). The structure of the aspartic proteinase endothiapepsin in complex with a substrate analogue analyzed to 2.2  Å resolution using Laue neutron diffraction.

J. Habash et al. and J. R. Helliwell (Acta. Crystallogr (2000) D56, 541-550). Structures of concanavalin A were determined in tandem using neutrons and X-rays. Laue neutron radiation was used on station LADI of the ILL. 148 water molecules were investigated. The location of waters was more accurate with neutrons than X-rays; more interestingly the hydrogen positions assigned by XPLOR for the X-ray model were initially not within the nuclear density derived from the neutron diffraction. However refinement of these positions against the neutron data brought the modelled hydrogens within the nuclear density with what the authors descibe as obviously correct fits. Details of 2 waters within the manganese coordination sphere were also described based on the nuclear density map, with the D atoms of the water molecules pointing away from Mn.

A. Ostermann et al. (Biophys. Chem. (2002) 95, 183-193). The structure of myoglobin by neutron crystallography at 1.5 Å resolution was determined, with hydrogen-deuterium exchange to allow location of water molecules within the protein and on its surface and, for example, the positioning of protons on the haem group to be determined. H-D exchange was also correlated with structural distribution and flexibility. The data collection was relatively "rapid" (24 days) and was carried out using a research reactor at the JAERI.

L. Coates et al. (Biochemistry (2001) 40, 13149-13157). The authors describe the largest neutron crystal structure of a protein determined to date, that of endothiapepsin. Data were collected using the LADI neutron difractometer at the ILL. The aim of the study was to define the protonation state of key catalytic groups and thereby shed important light on its mechanism. In the presence of a transition state analogue inhibitor of the enzyme the data show that Asp 215 of this structure is protonated while Asp 32 is not and therefore represents the negatively-charged residue in the transition state complex.
 

Back to the introductory page