The last twenty years have seen remarkable changes both in the applications of synchrotron radiation to the life sciences, but also in the attitudes of the scientists working in that field. Back in the early 1980's synchrotron radiation was considered by many to be a very expensive and rather unreliable method of producing high intensity radiation and the question of whether the money involved could be spent in a more productive way was often at the forefront of debate. It is very fortunate that scientists with foresight and perseverance gave strong support to the physicists developing synchrotron radiation technology and it is now the case that this technology permeates many areas of the life sciences, including macromolecular crystallography, non-crystalline diffraction and biological spectroscopy, and is considered an absolutely essential tool. There is also an increasing and important interest in the medical applications of synchrotron radiation for basic research, diagnostic imaging and therapy. The current situation is that the life sciences account for some 20-25% of the usage of the ESRF and the pressure for beam time is increasing, particularly in the macromolecular crystallographic and non-crystalline diffraction areas, as more scientists become aware of the potential of applying synchrotron radiation to investigate more and more complex biological systems. These problems are aimed at understanding the living cell and this will undoubtedly lead to important medical and social developments of benefit to all. This pressure is unlikely to decrease over the next decade or two as projects such as the Human Genome Project, which alone is expected to produce some tens of thousands of new proteins, whose structures will need to be investigated using X-ray methods in order to determine structure-function relationships, continue to evolve. In turn, the ESRF is responding to this pressure by constructing new beamlines for protein crystallography and continually reviewing the usage of others. The life science highlights for 1996/7 have not attempted to cover all the exciting new science that is being undertaken at the ESRF, but rather to stress the power of the methodology and the possibilities for future research.



Macromolecular crystallography



Small, weakly diffracting crystals

Macromolecular crystal analysis provides structural information, often close to atomic resolution, which serves to underpin much of modern structural biology. Synchrotron radiation with its characteristics of highly collimated, high intensity, wavelength tuneable X-rays, plays a key role in such analysis, enabling data of very high quality and resolution to be collected even from very small crystals of weakly diffracting macromolecules. At the ESRF a Microfocus beamline, ID13, permits good quality data to be obtained on previously unusably small crystals only a few tens of micrometers in size [1]. The current experimental arrangement is based on a 30 micrometer focused undulator beam, but beams down to a few microns in size are feasible. A major problem concerns sample alignment and much effort has been invested in this area. In addition radiation damage makes the use of cryogenic techniques mandatory and fast data collection is essential through the use of CCD-based detectors. A recent highlight has involved Bacteriorhodopsin. This protein is found in the purple membrane of Halobacterium halobium and acts as a light-driven proton pump across the cell membrane to convert light into chemical energy. Present structural knowledge is based on electron diffraction studies [R. Henderson et al.], but these are limited in resolution to about 3.0 Å. At ID13 diffraction patterns of three-dimensional Bacteriorhodopsin crystals of about 30 µm in size have been, for the first time, measured to high resolution giving data to 2.0 Å. The structure is currently refined to 2.5 Å [2].

Proteins associated with cell membranes, either partially or fully integrated into the membrane, provide a special challenge to macromolecular crystallography. Not only are these important proteins difficult to crystallise, but the crystals are mostly small and give weak diffraction patterns making the use of synchrotron radiation mandatory.

A typical example recently studied at the ESRF by W. Saenger and his colleagues [3,4] is Photosystem I whose structure will provide firm data for understanding oxygen-producing photosynthesis. Photosynthesis is the process whereby solar light energy is converted into chemical energy which is then readily available to living systems. In oxygenic photosynthesis (occuring in plants, algae and cyanobacteria) two separate reaction centre (RC) complexes catalyse this reaction and primarily function in series; they are called respectively, Photosystems I and II (PSI and PSII). While PSII produces a highly positive redox potential to oxidise water, PSI produces a highly negative potential to reduce NADP+ to NADPH. Both have been characterised extensively by biochemical, spectroscopic and partly by structural methods.

The large antenna system of PSI, with approximately 100 chlorophyll a units (Chla), serve to capture excitation energy and channel it to the primary electron donor P700, (a Chla-dimer), located near the lumenal side of the membrane. On excitation the primary donor causes a charge separation and an electron is transferred via intermediate cofactors to the terminal acceptors (two 4Fe-4S clusters) located near the stromal side of the thylakoid membrane. On the basis of data collected on the High Brilliance beamline, ID2, to 3.8 Å resolution, recently extended to 3.5 Å, the protein and the geometry of the cofactors both of the antenna and the electron transfer system can be derived with improved accuracy (Figures 99 and 100). The fine collimation of the beam proved essential in resolving the broad Bragg peaks which result from the high crystal mosaicity.

The structure of PSI is the first representative of the group of iron-sulphur type RCs. Despite insignificant sequence homology it shows some structural resemblance to the RC of purple bacteria, which in turn is related to PSII, both being quinone-type RCs. Thus, apart from clarifying the molecular basis of light conversion in PSI, the X-ray structure has important implications both in terms of the evolution of photosynthetic RC and as regards the structure of RC which have not been amenable to structural analysis.



Structure of the nucleosome core particle

The problem of how DNA is packaged into chromosomes has been a challenge to biologists for many years. In the 1970s it was discovered that all the eukaryotic chromosomes consist of regularly repeating complexes of protein and DNA, called nucleosomes. In turn, each nucleosome consists of a protein octamer, comprising two copies each of four different histones, which together with a fifth histone organise approximately 200 base pairs of DNA.

The structure of the nucleosome core particle at a low resolution of 7 Å was determined in 1984, but now Tim Richmond and his colleagues have used beamline ID13 at the ESRF and a new preparation of crystals for which all components were made in bacteria and assembled after purification, to establish the structure at 2.8 Å resolution. Indeed, the crystals containing DNA of a defined sequence and histones lacking post-translational modifications diffract to around 2.0 Å and an even higher resolution structure should be available in the future.

In the current structure an entire 146 base pair DNA molecule has been identified together with over 80% of the eight histone chains. The DNA is wrapped around the histone octamer in 1.65 turns of a flat, left-handed superhelix, but its path is distorted by bends at several positions. The structure provides a clear template for the interpretation of genetic and biochemical data and will help understanding how the nucleosomes assemble into the higher-order chromatin structures [5].



Anomalous dispersion

One of the major problems facing macromolecular crystallography is the determination of the phases of the diffracted beams. The amplitudes, or at least the intensities, can be measured directly, but both phase and amplitude information is required before an image of the diffracting object can be constructed. However, the tuneable nature of synchrotron sources has led to an increasing use of anomalous dispersion techniques to surmount this problem. The Materials Science beamline, ID11, has been used to undertake studies at an energy 50 eV above the Xe K-edge ( = 0.358 Å) using the ESRF image intensifying CCD detector [6]. Porcine pancreatic elastase, a 26 kDa protein, was used for the collection of two data sets. For the native protein, data were collected to a resolution of 1.4 Å, the crystal was then pressurised with Xe at 16 bars and data collected to a resolution of 1.8 Å; all data collection was performed at ambient temperatures, but the radiation damage appeared to be relatively small. Although the expected anomalous variation in the Bijvoet intensity differences is only 3.4% assuming a single fully occupied Xe site per molecule, isomorphous and anomalous difference Patterson syntheses readily yielded the location of the Xe. Subsequent phasing of the structure factors was easily performed giving electron density maps of excellent quality and demonstrating the feasibility of using ultra-short wavelengths around the Xe absorption edge to solve protein crystal structures.

The structure analysis of a cytoplasmic domain recently solved by R. van Montfort (University of Groningen) typifies the Multi-wavelength Anomalous Dispersion (MAD) method and marks one of the highlights of the year on BM14 since the anomalous signal is very weak with only one Selenium atom per 16 kDa (one Se per 7 kDa molecular weight is a typical concentration for Seleno-methionine analogues). The bacterial phosphoenolpyruvate-dependent phosphotransferase system (PTS) mediates the energy-driven uptake and subsequent phosphorylation of carbohydrates. The carbohydrate-specific part of the mannitol-specific PTS (EIImtl) is composed of a membrane embedded transporter domain, IICmtl and two cytoplasmic domains, IIAmtl and IIBmtl, which sequentially transfer a phosphoryl group to the transported mannitol.

Crystals of the subcloned IIAmtl grow as thin plates with typical dimensions of 0.6 x 0.15 x 0.03 mm. The space group is P21 and the asymmetric unit contains four protein molecules. An extensive search for suitable heavy atom derivatives was unsuccessful and consequently a single methionine residue in IIAmmtl was replaced with seleno-methionine. A three-wavelength MAD experiment was carried out at BM14 using a seleno-IIAmtl crystal. The data were collected using the image intensified CCD detector developed at the ESRF, and the ease with which the small anomalous signal was detected demonstrates clearly the stability and uniformity (after calibration) of the detector. The data were of good quality with RSym around 4%, and although there is only one selenium atom per 16.3 kDa, the selenium positions could easily be located in both anomalous and dispersive Patterson maps.

An electron density map calculated with the resulting phase information showed the positions of the four molecules in the asymmetric unit, but tracing of the polypeptide chain was not unambiguous. However, a readily interpretable electron density map was obtained after fourfold non-crystallographic symmetry averaging. The structure of IIAmmtl, is composed of a central five-stranded mixed, -sheet flanked by a-helices on both sides of the -sheet (Figure 101). The phosphorylation site is located at the C-terminal end of one of the -strands and forms the centre of a shallow depression at the surface of the molecule.



Experimental triplet-phase determination from crystals of small macromolecular structures by three-beam diffraction

As indicated above, the phase of a structure factor F(h) to the reciprocal lattice vector (rlv) h of a crystal structure is not accessible directly as the measured intensity depends on F(h)F*(h). Moreover, the phase of a single structure factor depends on the origin of the unit cell and is therefore not a measurable quantity. However, the phase of the product of structure factors of reflections whose rlv's join to form a closed polygon is independent of the origin. In case of three reflections (e.g. g, h-g and -h) this is called a triplet phase which can be determined by a three-beam interference experiment. During such an experiment in addition to the primary reflection h a secondary reflection g will be brought into its diffraction position by an azimuthal -scan about h. In this case the two wavefields in direction K(h) and K(g) will interfere with each other via the rlv's ± (h-g). This interference will modulate the intensities of the wavefields in the K(h) and K(g) directions. The characteristic intensity change in the K(h) direction during the -scan around h depends on the phase difference of the reflections involved, namely on

is the triplet phase and the resonance phase shift of the wavefield in direction of K(g) when g passes through the Ewald sphere. A schematical view of a three-beam situation is shown in Figure 102.

In order to observe these interference effects crystals of low mosaicity are required. However, due to the very small divergence of synchrotron radiation it is often possible to select a single large nearly perfect block from a mosaic crystal. The measured interference profile is a convolution of the intrinsic profile with the angular and wavelength spread of the incident beam. Therefore, a well collimated synchrotron-radiation beam is extremely advantageous. The overlap of interference profiles of adjacent three-beam cases can be minimised by tuning the wavelength properly for each three-beam case.

At the Swiss-Norwegian CRG bending magnet beamline BM1, a group led by E. Weckert and K. Hummer, have undertaken three-beam interference experiments with various crystals of different macromolecular compounds using a specially designed Y-six-circle diffractometer [7,8]. The unit cells ranged from 26,177 Å3 (symmetry P1) to 504,000 Å3 (symmetry P43212). However, in order to develop the method the tetragonal form of hen-egg lysozyme was used for which some triplet phase information had already been measured. In order to demonstrate a possible structure solution by the three-beam interference method, this work was extended with improved measuring methods. Currently about 850 triplet phases have been measured comprising some 750 individual reflections. Their resolution ranges up to 2.5 Å with a maximum of the distribution at about 4 Å. An example of a three-beam interference profile is given in Figure 103 which also shows that in protein crystals large dynamical interference effects may occur. The average intensity change due to the three-beam interference is in the range of 5 to 10% for good quality crystals. At an ESRF bending magnet (2/3 filling mode) depending on the size and the quality of the crystal about four to six triplet phases per hour can be measured. It was possible to determine about 80 to 150 triplet phases from a single crystal before the interference effects were reduced to about 50% of their original value due to the influence of radiation damage.

After selecting origin fixing reflections, new single reflection phases can be obtained by a suitable connection of triplet phases. However, with this method not all reflections contained in the measured triplet set can be assigned a reliable phase. In a similar manner to the Direct Methods multi-solution procedure, the phases of some reflections have to be permuted. For the lysozyme data this is the case for the phases of three reflections. Using a method based on entropy maximisation and likelihood ranking of the permuted sets, reliable estimates of the missing phases can be obtained. The final phase set shows a mean phase difference of about 16° compared to the phases calculated from entry 1LSE of the Brookhaven Protein Data Bank [Bernstein et al., J. Mol. Biol., 272, 741-769 (1997)]. The electron-density map calculated from these reflections seems to be reasonably complete, an example is given in Figure 104, demonstrating that it is indeed possible to obtain a reasonable electron-density map of a small macromolecule using directly measured triplet phases as the only phase information.

Consequently, the procedure was also applied to an unknown protein structure which had not been solved by other methods (space group P2l2l2l, V ~ 105,000 Å3). Within three days of beam time (16-bunch mode) more than 200 triplet phases were measured from two crystals. Considering the size of the molecule and the experience gained with tetragonal lysozyme, this is already slightly less than half of the amount of triplet phases which will be necessary to lead to a first structural model; further measurements will be undertaken.



Time-resolved macromolecular crystallography

Beamline ID9 at the ESRF has been specifically designed to undertake time-resolved macromolecular crystallography, so that reaction mechanisms can be studied in situ. Proteins and enzymes normally undergo structural changes whilst carrying out their biological and catalytic functions. For example, the binding of ligands to cell receptors induces conformational changes that are the key to the signal transduction mechanism. Whilst the "before" and "after" structures are known for some proteins, the pathways connecting these limiting structures are largely unknown and until recently, largely unexplored. The ability to watch macromolecular structural changes as they occur has been developed at the ESRF on beamline ID9 [9,10,11] and to test this facility the ligand binding haem proteins, myoglobin and haemoglobin, have been selected as test cases.

Haemoglobin, a protein found in red blood cells, transports oxygen from the lungs to the tissues whilst myoglobin, a protein found in muscle, stores the oxygen for conversion into energy as required. When Sir John Kendrew determined the structure of myoglobin in 1960, it immediately posed a problem: there was no obvious way for oxygen to get in and out of the molecule as indicated in Figure 105. Clearly, the protein structure is not static, but dynamic with channels opening and closing to permit access to and from the active binding site. Where are these channels ? How quickly does the protein structure respond to ligand binding and dissociation ? To address these questions, K. Moffat and his colleagues (University of Chicago) have continued their experiments of triggering the dissociation of carbon monoxide from MbCO crystal with a 10 nsec optical pulse, and measured Laue diffraction patterns with 150 psec polychromatic X-ray pulses (nsec and psec are 10-9 and 10-12 seconds respectively). CO was used as a surrogate for O2 because it is easier to photo-detach from Mb and rebinds to the protein more slowly. Structures were determined at various times controlled by the time interval between photolysis and the X-ray pulses. The intensity of the X-ray pulse was sufficient to record a Laue pattern with a single pulse, making it possible to accumulate a complete set of Laue patterns at a sequence of times with a single MbCO crystal (see Figure 106). The sequence of time-resolved structures produces a short "movie" of a protein in action. Simplified snapshots from this movie are shown in Figure 107. The first picture shows the initial state before the optical pulse where the CO is bound to the iron of the haem moiety. The second picture, taken 4 nsec after photo-dissociation, shows an intermediate state in which the CO is rotated by 90o and displaced some 4 Å from the iron. It stays in this configuration for around 350 nsec. The function of this "docking" site is to prevent CO from recoiling back to the chemically attractive iron. The last picture, taken 1 µsec after the optical pulse shows that the CO has left the binding pocket completely. In this time regime the CO diffuses about in the outer part of the protein for a fraction of a millisecond before eventually recombining with the iron through random collisions with the fluctuating haem and thereby closing the photocycle. The time evolution of these features shows for the first time how the protein responds structurally to ligand dissociation and rebinding under ambient temperature conditions [12]. The time resolution of the technique will soon be improved by nearly a factor of 100, with even further improvements possible. Preliminary work has also been undertaken on haemoglobin and this revolutionary capability can be extended to many other photo-biological systems such as bacteriorhodopsin, photo-reactive yellow protein (see below) and the photosynthetic reaction centre.

Photoactive yellow protein (PYP), (see Figure 108), is a light receptor that allows its host, the bacterium Ectothiorhodospira halophila, to detect and then swim away from dangerously high levels of blue light. The chromophore in PYP is a 4-hydroxycinnamic acid molecule bound to the protein's only cysteine residue via a thioester linkage (Figure 109). When this chromophore absorbs a photon, it isomerises from its trans to its cis form and triggers a series of structural rearrangements that lead to a biologically active signalling intermediate. Even though many of the biological and structural questions related to this transformation have been addressed crystallographically, many more remain open. For example, it is not yet known how the protein primes the chromophore for the initial photoreaction to achieve the unusually high quantum yield of 64% for the trans to cis isomerisation. How the chromophore is able to undergo photo-isomerisation in the tightly packed protein interior also remains a mystery. To answer these and many other questions about the initial phase of the light cycle requires the time-resolution currently available only at the ESRF. The research teams of K. Moffat and E. Getzoff have independently collaborated with ESRF scientists to determine X-ray structures of photo-excited PYP with nanosecond and microsecond time resolution, respectively [13]. Preliminary results from high quality Laue patterns obtained with a single electron bunch (see Figure 110) are encouraging. The last millisecond intermediate I2 is shown in Figure 111.




Non-crystalline diffraction


High-resolution X-ray diffraction of contracting muscles

There is a tendency to take for granted the complex mechanisms, both physical and chemical, which enable us to enjoy everyday life, but there are many aspects of the human body which are poorly understood, or even completely unknown. One such example is the mechanism by which our muscles work and research on this topic has been pursued on ID2 at the ESRF by J. Bordas and A. Svensson [14].

X-ray diffraction from whole, live, contracting muscle is currently the only technique available which provides direct structural information about the molecular events responsible for the conversion of chemical energy into the production of force and motion in totally unmodified muscle tissues. One of the central assumptions in practically all the existing models of muscular contraction is that all the elasticity resides in the myosin heads attached to the thick myosin filaments and that force is produced by a distortion in the myosin heads coupled to the use of ATP. More specifically, when an isometrically contracting muscle is submitted to a sudden length release, there is a fall in tension proportional to the length change (T1 phase). This has been traditionally interpreted as due to the discharging of the elastic strain in the myosin heads attached to the actin filament. The high brilliance available from the ESRF has allowed the recording of changes in the spacing and intensities of all the features in the muscle diffraction patterns with a resolution ranging from around 1000 to 2.0 nm as shown in Figure 112. In addition to showing that complex interference phenomena, hitherto unreported, must be taken into account when interpreting diffraction diagrams from contracting live muscle, the data clearly shows that the filaments themselves, rather than the myosin heads, have sufficient elasticity to explain the T1 phase.

Other significant results emerging from this ongoing work concern the realisation that the thin filaments (i.e. actin and associated regulatory proteins) are much more structurally active than so far anticipated. The combined results suggest that a substantial revision of at least some models of contraction may be now necessary.



X-ray absorption spectroscopy (XAS)

Crystallographic techniques provide unique insights into the overall assembly of macromolecules and enable detailed structure-function studies to be undertaken. The routine use of cryogenic methods can lead to data close to atomic resolution, but of course the success of the methodology is totally dependent on the growth of suitable single crystals. On the other hand EXAFS and XANES can be used to probe the local environments of metal sites in biological macromolecules both in solution (often at very low concentrations of the metal) and the solid state. It is also possible to study changes in the local environment of the metal as the macromolecule undergoes chemical changes such as during a catalytic cycle. A combination of crystallographic and X-ray absorption measurements in a complementary fashion can be very useful in structural biology, but in cases where crystallisation has proved problematic or only very small amounts of the protein are available, then spectroscopic methods can provide a powerful tool for structural studies.



Characterisation of the metal sites of the key regulatory FUR protein for iron uptake in bacteria using X-ray absorption spectroscopy

Iron is essential to all living organisms and Gram-negative bacteria have developed sophisticated iron uptake systems which are under the regulatory control of a key protein called FUR, for Ferric Uptake Regulation (Figure 113). FUR is a DNA-binding protein which needs iron, in vivo, in order to be activated to bind specific DNA sequences called iron boxes and then repress the iron uptake. FUR is the iron sensor of the bacteria; in the presence of an excess of iron, FUR is activated and represses the expression of the proteins involved in the iron uptake, but when the iron concentration drops, FUR becomes inactivated and the iron uptake can start again. Understanding the structure and the mechanism of action of such a protein at a molecular level may ultimately lead to the design of specific bactericides.

The FUR protein contains two metal binding sites. In addition to the iron binding site, a new site has recently been discovered which tightly binds zinc. The characterisation of these metal sites is a first step in the understanding of the iron uptake by FUR. In vitro, di-cations such as Co2+ or Mn2+ are able to activate FUR as well as Fe2+ and certainly bind to the same site. A team of scientists at CNRS and CEA laboratories in Grenoble [15] has used these dications as spectroscopic probes for a study of the iron binding site because of the redox instability of the iron-substituted FUR protein.

XAS studies on the manganese, and cobalt substituted FUR proteins from E. coli at the Zn, Co and Mn K-edges have been undertaken at the ESRF, on the CRG-IF BM32 beamline, in order to obtain information about the properties of the two metal sites and their ligands. The high flux available at the ESRF allows 1) a reduction in the data acquisition time and therefore the problems of sample degradation and 2) the collection of reliable results at low concentrations of protein samples (2 mM). The absorption spectra recorded are shown in Figure 114. Both the XANES pattern and EXAFS oscillations enable details of the metal co-ordination to be determined to give a schematic representation of the metal as shown in Figure 115. The manganese and cobalt environments are similar and comprise 6 N/O donors ligands, at 2.21 Å and 2.12 Å, respectively. Probably, 2 or 3 of those ligands are imidazole groups of histidine residues. The zinc environment is tetrahedral with two sulphur donor ligands at 2.32 Å from cysteines and two N/O donor ligands at 2.03 Å, at least one of which is a histidine. A small change in the conformation of the zinc site is observed upon cobalt incorporation into the iron site. This observation is in accordance with the earlier finding that the affinity of the protein for the DNA is increased by a conformational modification induced by metal binding.




Applications of synchrotron radiation to medicine

The applications of synchrotron radiation to medicine cover a number of areas including basic research, diagnostics and therapy. The Medical beamline, ID17, obviously plays a key role in these activities, but other beamlines such as the High-Energy beamline, ID15, are also highly relevant for feasibility and in vitro studies.



Microbeam radiation therapy

The Medical beamline of the ESRF is now in an advanced state of commissioning, and since the spring of 1997 has been open for users with in vitro experiments. One of the three core programs on the beamline is preclinical microbeam radiation therapy [MRT] studies, a program proposed by The European Microbeam Collaborative Research Group (EMCRG), consisting of researchers from University of Bern, Kantonspital Luzern, Paul Scherrer Institute, Villigen, and the ESRF [16]. Following their proposal, a small animal irradiation facility has been built as an integral part of the Medical beamline.

MRT was proposed as a potential technique for the treatment of brain tumours, since it involves radiotherapy with less concomitant damage to normal tissues surrounding the tumour than is the case for conventional radiotherapy with seamless radiation fields. This could be particularly valuable for the treatment of pediatric brain tumours, where side effects in many cases prohibit effective radiotherapy. MRT is performed by sequentially cross-firing a lesion with an array of parallel X-ray beams, each beam in the array having a cross section that is microscopic (25-50 micrometres) in at least one dimension, and separated from neighbouring beams by a 75-200 micrometres wide space with no primary radiation. This creates a highly spatially micro-fractionated absorbed dose distribution in the tissues traversed by the X-ray beams, except in the target volume, where the cross-firing reduces or eliminates the spatial micro-fractionation. The spatial micro-fractionation is central for the MRT concept; it was previously hypothesised by Slatkin et al. and experimentally confirmed at the ESRF that the normal rat brain shows great resistance to microbeam irradiation, even when the absorbed dose delivered by each beam is 10-100 times higher than the maximum non-necrotising doses delivered to the brain by spatially uninterrupted photon radiation.

Synchrotron X-ray sources are the only sources presently available, where microbeam arrays with photon fluence rates high enough to be suitable for MRT irradiation, can be produced. Independent verification of the basic radio-biological observations underlying the MRT concept have now been performed by the EMCRG at the High-Energy beamline, ID15 [16], prior to the commissioning of the Medical beamline. In the first MRT experiments to be performed at the ESRF, the right hemisphere of the normal rat brain was irradiated in vivo with microbeam arrays to absorbed doses in the range 100 - 4000 gray. Each array consisted of 12 microplanar beams, each 3 mm high and 25 micrometres wide, and with a centre to centre spacing of 200 micrometres. The entrance absorbed dose delivered by the three outermost microbeams on each side of the array was 2076 gray, a dose where cellular but no tissue necrosis could be observed. These irradiated slices served as &laqno;markers», to assure that the positioning of the brain during irradiation were correct. In between the markers the microbeams delivered "test" doses in the range 139 - 4000 gray and for each animal the test dose was the same for all six beams between the markers. The absorbed doses were determined by theoretical dosimetry calculations using the synchrotron X-ray spectrum, thermoluminescence dosimetry using LiF dosimeters, and ionisation chamber measurements.

Figure 116 shows a stained histological section of a microbeam irradiated rat brain cut horizontally along the beam propagation direction. The position of the marker triplets with absorbed doses of 2076 gray are clearly seen as triplets of lighter stripes, caused by the disappearance of neuron cell nuclei. In between the markers, six evenly spaced microbeams have delivered 342 gray each. These test doses cannot be distinguished at all by eye, demonstrating that the threshold absorbed dose for cellular necrosis is more than an order of magnitude higher after microbeam irradiation than the threshold dose for tissue necrosis after spatially uninterrupted photon irradiation.

The results of the experiment verify not only a very high threshold absorbed dose for cellular necrosis, but also a considerably higher threshold absorbed dose for tissue necrosis. Both these radiobiological properties are fundamental for the MRT concept, and a strong motivation to pursue further preclinical studies as a step towards the long term goal, i.e. clinical validation of MRT for therapy of pediatric brain tumours.

Microbeam irradiation opens up not only the possibility of MRT for tumours, but also the creation of a new tool for radiobiological studies, since radiation effects on tissues can now be studied in vivo at unprecedented doses. In an effort to find a cell system that could be used as an alternative to mammals in the optimisation of some physical irradiation parameters to be used for MRT, experiments on Drosophila melanogaster, performed with a collaborator from the Institute für Medizinische Radiobiologie, Zurich [17] have shown promise. These experiments again indicate the high resistance of biological systems to microbeam irradiation and further supporting the basic ideas behind MRT.



Computed tomography of brain phantoms

A joint group of scientists from the RSRM (Rayonnement Synchrotron et Recherche Médicale), Grenoble, and the ESRF have been using the Medical beamline, ID17, to investigate the potential use of computed tomography techniques for brain pathology studies.

Both medical imaging and computing techniques have dramatically improved in the past twenty years. Signal processing and image reconstruction techniques are now available which allow very accurate studies on anatomic slices of the human body. Nowadays, efforts are aimed at extending the usefulness of these techniques beyond anatomic mapping to obtain functional or metabolic information (cerebral blood flow, oxygen consumption, glucose consumption, uptake and behaviour of particular metabolites such as neuromediators or drugs). This kind of information appears essential, especially in the field of neuroscience, for an understanding of physio-pathology and for guiding therapeutic decisions. At present, physicians often have to use invasive techniques such as intracerebral electrode implants to access such information (cerebral micro-dialysis, stereo-electro-ence-phalography).

Currently magnetic resonance imaging (MRI) appears to be one of the most useful techniques, but the information obtained is usually only qualitative and as a consequence, the feasibility of using monochromatic synchrotron X-ray beams is under investigation. The CT (Computed Tomography) scan project at the Medical beamline ID17 represents such an approach and is particularly focused on brain pathologies (tumours, ischemia and epilepsy). The predominant aim of the project is to answer the following questions:

a. Is it possible to obtain selectively, concentration measurements of contrast agents (such as gadolinium and iodine) at clinical concentrations by using a synchrotron radiation CT scanner? Accessing concentration measurements of intravenous contrast agents, i.e. accessing blood distribution is one of the central purposes of a program to characterise blood brain barrier rupture phenomena, endothelial permeability and the tumour angiogenesis process.

b. Is it possible to access concentration measurements of endogenous elements such as potassium and to focus on tissue ionic modifications? These modifications are the keys in understanding and predicting metabolic comportment after epilepsy crisis or ischemic injuries.

Two energy-selective imaging methods, DPA-CT (Dual Photon Absorptiometry) and KES (K-Edge Subtraction) will be implemented in the clinical system to improve contrast resolution and provide quantitative images. DPA-CT is to be used to detect brain lesions of altered elemental concentrations, such as potassium. It is based on the dependency of the Compton and photo-electric part of the attenuation coefficient on the energy and the atomic number Z. The tissues are mainly constituted with two pools of elements, low Z and intermediate Z. X-ray transmission is measured at two widely separated energies. The data are processed to produce two quantitative images of the effective average density of each element group, with an improved contrast resolution compared with a single energy CT image. The KES technique is mainly dedicated to imaging small cerebral tumours and ischemia analysis. This technique uses the sharp rise in the photoelectric component of the attenuation coefficient of gadolinium and iodine at the binding energy of the K electron. Two images are taken in turn, immediately above and below the K-edge. The resulting subtracted image is quantitative for the contrast agent and at much higher contrast resolution than a conventional CT image.

At the present time, measurements have been carried out on phantoms to evaluate the detection limits of both techniques. DPA was carried out using cylindrical Plexiglas and aluminum head phantoms of diameters 10 and 18 cm containing solutions of KOH at concentrations from 0.3 to 50 mg/ml. Data were obtained at 40 keV and 80 keV, for different absorbed doses. Similar phantoms filled with Gd and I solutions of different concentrations (0.1 to 3 mg/ml) were imaged at their respective K-edges (50.239 keV for Gd and 33.169 keV for I). The CT configuration used a fixed fan-shaped X-ray beam of 1.5 mm height and 120 mm width and a sample stage rotating around a vertical axis. Attenuation was measured with a linear-array germanium detector with 432 elements each of width 0.35 mm.

The preliminary experiments show that the sensitivity of the system to variations of potassium concentrations is below 10 mg/ml for an absorbed dose acceptable for clinical radiology. Under the same experimental conditions, l mg/ml (~ 10%) of iodine or gadolinium is quantifiable in a 6 mm diameter detail and 2 mg/ml (~ 10%) in a 3 mm diameter detail. Further measurements are needed to test the limits of quantification for the CT configuration, but it is hoped to reach the targets of 3 mg/ml in potassium detectability and 1.5 mg/ml in iodine or gadolinium detectability, so that angiogenesis and blood brain barrier rupture in tumour process can be effectively studied.

At present, this imaging research is at a preclinical technical validation stage, but improvements can be made by optimisation of the experimental parameters. It is supported by a grant from the Région Rhône-Alpes for a research programme focused on neurosciences involving Université Joseph-Fourier (Grenoble), CEA-LETI (Grenoble), INSA (Lyon), the Hospital of Grenoble and the ESRF. Collaborations with scientists at other synchrotron sources, DESY Hamburg and the National Light Source at Brookhaven, have been initiated and further collaborations with biologists and physicians indicate a growing interest in this promising area of medical research.


Wavelength dispersive tomography applied to the Rayleigh-to-Compton method

The Rayleigh-to-Compton (RC) method is a very subtle technique to extract spectroscopic information about the elemental composition in bulky materials. The use of scattered high energy X-ray photons makes it applicable for imaging problems, where conventional X-ray spectroscopy employing the characteristic emission lines of atoms fails because absorption effects are dominant. The method can be used for spectroscopic imaging of biological samples which contain mainly low and medium Z elements.

The RC method has recently been analysed by Harding et al. and it has been shown that the best contrast-to-noise ratio for low Z materials is achieved at low momentum transfer, i.e. in forward scattering geometry. Stimulated by these studies, a novel RC-imaging technique has been developed and tested on the High-Energy beamline, ID15B [18]. This technique allows, due to the high-energy resolution of a bent analyser crystal, the separation of the elastically (Rayleigh) and inelastically (Compton) scattered photons for forward scattering angles. The primary beam which probes the sample creates a line source of Rayleigh and Compton scattered photons which is imaged dispersively by a conical analyser crystal [19] onto two scintillation detectors. The recorded signals are therefore projection data and tomographic reconstruction algorithms can be applied. The ratio, pixel by pixel, of the Rayleigh and Compton image gives the RC image with pixel values proportional to the effective atomic number. For a quantitative analysis the reconstructed RC ratios have to be calibrated with values measured for samples of known elemental composition.

The feasibility of the method was demonstrated by imaging a slice of human vertebra sample (provided by M. Pateyron) embedded in an epoxy glue. The pixel size in the image shown in Figure 117 is 100 µm compared to the typical trabecular bone width of about 300 µm. In this particular image, partial volume effects make an interpretation of changes in the elemental composition difficult, but RC images are free from density effects, e.g. microporosity, which affect quantitative absorption images. In future, the bending scheme of the analyser will be optimised to reduce aberrations and the calibration improved. This novel approach to spectroscopic imaging should provide new possibilities in sample characterisation and medical studies in vitro.