nature structural biology • synchrotron supplement • august 1998 627

synchrotron supplement

In the early days, the low intensity of con-ventional radiation sources and theabsence of efficient detectors for record-ing the full three-dimensional diffractionpatterns posed severe problems for thosewishing to collect diffraction data fromprotein crystals. Synchrotron radiationbegan to play a role in data collection forproteins only in the late 1970s, buildingon pioneering work a few years before(see the article by Kenneth Holmes in thissupplement). Yet, synchrotron radiationhas already had an enormous impact onthe field of crystallography1,2.

The problems of carrying out X-raycrystallographic analyses of macromole-cules arise from the nature of the mole-cules themselves and of the way they packinto crystals. The fact that they are largemeans the repeating unit of the crystal,the unit cell, is large. Hence there aremany weak reflections to be recorded.

This leads to a low signal-to-noise ratio,which can be alleviated by a high intensi-ty source and efficient detector. Anothercomplicating factor is that the moleculesare held together in the crystal by weakinteractions between the proteins, whilea large fraction of the crystal, on averageabout 50%, is disordered solvent. Thusthe molecules in the crystals display sub-stantial disorder, which means that thehigh resolution data are weak or indeedabsent.

Advantages of SRSeveral properties of synchrotron radia-tion make it attractive for studying pro-teins. Its high intensity makes the weakdiffraction from macromolecules possi-ble to record on a tractable time scale.Highly parallel rays that can be finely col-limated to produce a focused beam,which are provided by synchrotron radi-

ation, are essential for large unit cellswith closely spaced reflections. An exam-ple of the resolution possible on a sourcesuch as the European SynchrotronRadiation Facility (ESRF) is shown inFig. 1.

Unlike the radiation from conventionalX-ray sources, synchrotron radiation pro-vides a broad spectrum and the mono-chromatic wavelength is thereforetuneable. First and foremost this allowsthe use of short wavelengths, avoiding theeffects of absorption and associated radi-ation damage, which can be severe withCuKα radiation at 1.54 Å. The spectralregion of choice is 0.8–1.0 Å for most sin-gle wavelength experiments, as this mini-mizes the effects of absorption whileretaining a good diffraction intensity. Thereduction of the effects of secondary radi-ation damage are especially important forexperiments at room temperature.

Illuminating crystallography Keith S. Wilson

Synchrotron radiation is an established tool in macromolecular crystallograpghy. Its high intensity andtunability play crucial roles in many structural analysis.

Fig. 1 A diffraction pattern from a crystal of the core particle of blue-tongue virus, BTV 10, recorded on the high brilliance beamline ID2 at the ESRF.The cell dimensions of the crystal are a = b = 1,120 Å, c = 1,592 Å. Data have been processed to 7.0 Å, the oscillation range was 0.3° with 10–20 s expo-sures. The current data set was collected from over 50 crystals. The detector was a 30 cm Mar-Research image plate system. The photograph was pro-vided courtesy of D. Stuart.

synchrotron supplement

628 nature structural biology • synchrotron supplement • august 1998

next step in time resolved studies is theuse of a single bunch of particles in thestorage ring for recording an X-rayimage, to shorten the time window of thestate studied to that of the bunch length,often at the 100 picosecond level.However, this will pose further problemsbecause of the current limits of detectorsand sample stability .

DetectorsThe use of synchrotron radiation for pro-tein crystallography requires efficientdetection of the diffracted X-ray pattern.Single counting diffractometers are clear-ly inappropriate. The first routine pro-tein crystallography experiments usedfilm as a detector. This was reasonable interms of spatial resolution, but very lim-ited in dynamic range and suffered fromsignificant noise level in the backgroundas a result of the chemical processesinvolved. Some attempts were made touse television or wire chamber gas-filleddetectors for protein crystallography, butthese met with limited success for a vari-ety of reasons such as saturation of thecount ratio and spatial resolution.

At the end of the 1980s, the situationwas transformed rapidly by the develop-ment of imaging plate detectors. The firstautomated on-line read-out system forroutine work on a beamline was devel-oped by Jules Hendrix and Arno Lentferat European Molecular BiologyLaboratory (EMBL) Hamburg, Germanyin 1990 (building on experience fromgroups in Japan). Within a few years,imaging plates became the synchrotronradiation detector of choice (with a com-

mercial system installed on most beam-lines) and have remained so for the lastdecade. With the new imaging plate tech-nology, for the first time the user couldrecord full diffraction images directly oncomputer disk. Within a short space oftime data reduction software packagessuch as MOSFLM3, DENZO4 and XDS5

had advanced such that images could beanalyzed on-site, and users could returnto their home laboratory with aprocessed and analyzed set of X-rayintensities. Data collection was thus rev-olutionized, and the throughput of exist-ing beamlines advanced apace.

We are now in the middle of the nextstep forward. Imaging plates suffer froma major limitation. Reading out theimage took a substantial time, originallytwo to four minutes, now reduced to20–40 s for the latest systems.Nevertheless, this step became rate limit-ing on the beamlines at ESRF and APSwhere exposure times are down to theorder of seconds (or sometimes less).Charge coupled devices (CCDs) haveovercome this problem. After a long ges-tation period when they were beingdeveloped by a number of academic andcommercial groups6, they have at lastreached maturity. The problems withCCDs were always one of size — theywere too small — and, related to this,expense. The new devices are alreadyinstalled on beamlines at a number of sites(for more information, see the article byRobert Sweet). The cost per device liesroughly in the range $250,000–$500,000,a small price to pay for the expected gainin effective beamline usage. The qualityof data potentially available at high speedis shown in Fig. 2.

Cryogenic freezingThe development of cryogenic freezingtechniques has been critical to recentadvances in exploiting the potential ofsynchrotron radiation. Already as long asten years ago, the stronger beamlineswere causing substantial radiation dam-age to the specimens, even with the use ofwavelengths below 1 Å and despite therapid data collection times. In the early1990s there was much debate aboutwhether protein crystals were in generalamenable to flash freezing, but happilyflash freezing is now an established tech-nique7,8 and cryogenic facilities are (orshould be) available on all beamlines.Today probably more than three quartersof synchrotron radiation experiments usefrozen samples, with little increase in

The tunability of synchrotron radia-tion can be further exploited by using X-ray absorption edges where the anom-alous scattering effects provide experi-mental phase information. Manyexperiments involve a single wavelengthwith optimization of the scattering at theedge of an isomorphous heavy atomderivative. There have been greatadvances in techniques for recordingmultiwavelength anomalous dispersion(MAD) data on metalloproteins, andmore importantly on crystals of proteinsor nucleic acids into which seleno-methionine or bromine respectively, hasbeen incorporated (see the article byCraig Ogata).

Synchrotron radiation can be appliedas ‘white’ radiation, where the wholespectrum, or a substantial range of thespectrum, is used simultaneously to illu-minate the crystal. This allows very shortexposures, in principle capable of deter-mining the structures of short lived statesin the crystal through time-resolvedstudies (see the article by Keith Moffat).Severe problems arise here in the conflictbetween Bragg’s and Boltzmann’s laws: itis difficult to find effective triggers forenzyme catalysis, or other changes inproteins, that allow a large population ofmolecules to begin the reaction simulta-neously. These are presently restricted tolasers with chromogenic systems.However even if an effective trigger isfound, all the molecules in the crystal willnot, in general, proceed through the reac-tion profile in an orderly manner. Withthe available techniques, time resolvedstudies are limited to a few systems. The

Fig. 2 Harker section v = 1/2 of the anomalous difference Patterson synthesis for a mannanaseplatinum derivative at 1.65 Å resolution. The peak is ~20 σ. A total of 180° of data were recordedin just 35 min on station 9.6 at the SRS, Daresbury Laboratory using an ADSC (San Diego) CCDdevice during a test period. Thanks are due to Miroslav Papiz (SRS) who helped with this work. Thefigure was provided by Gideon Davies. This emphasises the power of such detectors even on sec-ond generation sources.

nature structural biology • synchrotron supplement • august 1998 629

synchrotron supplement

mosaic spread, usually some extension inresolution, a substantial increase in effec-tive crystal life time and, therefore, a hugeincrease in the quality of the data. A morerecent technical development is the trans-port of prefrozen crystals to synchrotronradiation sources, the diffraction qualityhaving been checked in the home labora-tory prior to the visit. This is leading tomuch more efficient use of beam time.However, at the most intense sources,problems of radiation damage are re-emerging, especially for small crystals andwith microfocus lines. Further lowering ofthe collection temperature to liquid heli-um levels may alleviate this, although ithas been known for many years that bio-logical samples can indeed only tolerate afinite radiation dose9.

Beamlines, insertion devices andsourcesThe number of sources and protein cryst-allography beamlines has increased sub-stantially in the last decade. In brief, thebeamlines fall into two general cate-gories: (i) lines with horizontal focusingprovided by a bent single crystal mono-chromator followed by a bent (often seg-

mented) mirror and (ii) lines with dou-ble crystal monochromators and toroidalmirrors for focusing. The former are notreadily tunable and tend to be operated atfixed wavelengths, usually around0.8–0.9 Å. This is below the Pt, Hg, Au, Seand Br absorption edges, providing goodmeasurements of the anomolous scatter-ing for the multiple isomorphousreplacement phasing. The latter are muchmore readily tunable and ideal for MADexperiments but the double crystal set upreduces the overall intensity of the X-raybeam, making some of these beamlines,especially on the lower energy synchro-trons, less than optimal for analysis ofsmall crystals or large unit cells.

Many lines of the first type were builton bending or wiggler magnets on theearlier machines (see the articles by JohnHelliwell and Robert Sweet for adescription of these terms), as were thefirst MAD lines. This was largely due tothe uncertainty of how well undulatormagnets would perform in practice.Experience at the ESRF and elsewhereindicates that undulators are indeedtunable over the energy ranges requiredfor protein crystallography MAD work,

as the physical gap size attainable in theundulator magnets has been rapidlyreduced. One can anticipate that onthird generation synchrotron radiationsources with energies of 2.5 GeV andabove, protein crystallography lines maygenerally be of the second type with anundulator insertion device.

Atomic resolutionOne notable development has been theextension of the tractable resolution foran increasing number of proteins toatomic resolution, roughly 1.2 Å or bet-ter10. This is allowing the extension ofthe standard techniques of small mole-cule crystallography to such proteins,including refinement using fullanisotropic models, and improves thelevel of accuracy towards that requiredfor a full understanding of the biochem-ical behaviour of the protein. Two exam-ples of the level of detail that can beobtained from a second and a third gen-eration source are shown in Fig. 3. Abinitio phase determination is possiblewith such data and the structure of atleast one small protein has been solvedin this manner11.

Fig. 3 a, The structure of a complex between cellulase CelA from Clostridium thermocellum and cellopentaose has been refined to 0.94 Å reso-lution (D.M.A. Guerin & P. Alzari, manuscript in preparation). Extended stacking interactions on both sides of the scissile glycosidic linkage andseveral hydrogen bonding contacts promote a significant distortion of the sugar ring at subsite -1 which displays the skewed conformationshown here. The data were recorded on beamline BW7B at EMBL Hamburg, Germany. Figure kindly provided by Pedro Alzari. b, Stereo view ofa representative region of the 0.97 Å resolution 3Fo - 2Fc Fourier map contoured at 4σ (2.86 e Å–3) for tetragonal hen egg white lysozyme. Datawere collected on the SBC undulator beamline at the Advanced Photon Source (APS). The flux was 1.71 x 1012 photons s–1 100 mA–1. The crystalsize was 0.12 × 0.08 × 0.07 mm3 with a 1 s exposure per 0.5° frame for the high resolution data and a total data collection time of 25 min. Figureproduced with the program BOBSCRIPT12 and kindly provided by Andrzej Joachimiak and Martin Walsh.

ba

synchrotron supplement

630 nature structural biology • synchrotron supplement • august 1998

Large unit cells and cellular mechanicsJonathan M. Grimes and David I. Stuart

Developments in synchrotron radiation mean that the methodological and technological tools are in place todetermine the structures of large multi-component macromolecular machines.

tunities opened up by microfocus beam-lines, cryogenic freezing, the speed ofdata collection and MAD phase determi-nation, synchrotron radiation X-raymethods have become and will remainthe work horse of structural biology forat least the next decade. Most important-ly, however, synchrotron radiation — byhelping to ease the process of structuredetermination — has brought extraimpetus to the expansion of X-ray crys-tallography into biology, allowing us tovisualize many more structures of themolecules that govern chemical processeswithin the cell.

AcknowledgmentsThis brief overview is somewhat subjective andbased largely on my own experience at theEuropean Molecular Biology Laboratory HamburgOutstation. I am greatly indebted to D. Blow whofirst suggested my move to Hamburg and to L.Phillipson who so strongly supported the activitiesthere.

Keith S. Wilson is in the Department ofChemistry, University of York, York YO15DD, UK. email: keith@yorvic.york.ac.uk

1. Helliwell, J.R. Macromlecular crystallographywith synchrotron radiation. (CambridgeUniversity Press, UK; 1992).

2. Ealick, S.E. & Walter, R.L. Synchrotron beamlinesfor macromolecular crystallography. Curr. Op.Struct. Biol. 3, 725–736 (1993).

3. Leslie, A.G.W. Recent changes to the MOSFLMpackage for processing film and image platedata. CCP4 and ESF-EACMB Newsletter on ProteinCrystallography, Number 26 (SERC Laboratory,Daresbury, Warrington WA 4AD, England; 1992).

4. Otwinowski, Z. & Minor, W. Processing of X-raydiffraction data collected in oscillation mode.Meth. Enz. 276, 307–326 (1997).

5. Kabsch, W. Automatic processing of rotationdiffraction data from crystals of initially unknownsymmetry and cell constants. J. Appl. Crystallogr.26, 795–800 (1993).

6. Gruner, S.M. & Ealick, S.E. Charge coupled deviceX-ray detectors for macromolecularcrystallography. Structure 3, 13–15 (1997).

7. Watenpaugh, K.D. Macromolecularcrystallography at cryogenic temperatures. Curr.Opin. Struct Biol. 1, 1012–1015 (1991).

8. Garman, E.F. & Schneider, T.R. MacromolecularCryocrystallography. J. Appl. Crystallogr. 30,211–237 (1997).

9. Henderson, R Cryo protection of protein crystalsagainst radiation damage in electron and X-raydiffraction. Proc. Roy Soc. London Ser. B 241, 6–8(1990).

10. Dauter, Z., Lamzin, V.S. & Wilson, K.S. Thebenefits of atomic resolution. Curr. Opin. Struct.Biol. 7, 681–688 (1997).

11. Frazao, C. et al. Ab initio determination of thecrystal structure of cytochrome c6 andcomparison with plastocyanin. Structure 3,1159–1169 (1995).

12. Esnouf, R. M. An extensively modified version ofmolscript that includes generally enhancedcolouring capabilities. J. Mol. Graph. Model 15,132–134 (1997).

SummarySychrotron radiation facilities every-where have become increasingly userfriendly with easy-to-operate beamlines,detectors with straightforward acquisi-tion, data handling software and cryo-genic freezing. The ease of use andautomation of the beamlines themselvesis continually evolving, and the introduc-tion of graphical user interfaces (GUIs) isgreatly simplifying the lot of the user.This means that use of synchrotron radi-ation has moved away from a techniquebeing used only by a small number ofspecialists to one that is now routinelyapplied by the large majority of proteincrystallography groups world-wide.Estimates of what proportion of newstructures are determined using synchro-tron radiation range from ~40% to>70%, but the proportion is certainlysubstantial and growing. All the currentbeamlines are considerably oversub-scribed by at least a factor of two — andthis is likely to be an underestimate of thereal demand for synchrotron radiation.

With the advances in cloning, expres-sion and crystallization, and the oppor-

Protein crystallography has had a rollercoaster ride since the structure of myo-globin was revealed approaching 40years ago. One result is that we mayalready have collected structures for amajor portion of the unique proteindomain folds in the human genome1.This invites more radical visions of the future. One is the working throughof genome projects to ‘structuralgenomics’ where an industrialization ofthe crystallographic process could leadto the production of a massive num-ber of structures (see the article bySung-Hou Kim in this supplement).Counterpoised to this is a perhaps moreintellectually challenging focus: proteincomplexes.

Many of the key activities of the cellare now thought to involve multi-com-ponent macromolecular machines. Thatthese assemblies occupy center stage incell biology may reflect a deep change inthe way we think about the cell and howit works, as argued by Alberts2. A seriesof landmark structures have been deter-mined recently that exemplify what canbe already be achieved (for example, seerefs 3–11). In the future structural stud-ies will proceed alongside biologicalstudies, so that as the contents and stoi-chiometry of a functional complex areuncovered, expression, assembly andcrystallization strategies will be put inplace. There is ample evidence that crys-tallography is an appropriate technique

for the study of weakly interacting com-plexes, for instance the micromolaraffinity MHC Class I complexes withTCR and CD8, (for example, refs12–14). Some of these new targets willrepresent cutting-edge challenges, withproblems such as small crystals, weakdiffraction, sensitivity to X-rays andlarge unit cells cropping up increasinglyoften.

As Fig. 1 shows, macromolecular crys-tals already span a considerable range ofsizes. The routine availability of syn-chrotron radiation is a necessary under-pinning for the analysis of the largerunit cells. Characteristics such as highbrilliance and low beam divergenceopen up to the structural biologist prob-