WO2003080901A1 - Use of dynamic light scattering (dls) in a method for producing macromolecular crystals - Google Patents

Abstract

A method for producing macromolecular crystals comprising the steps of: (i) dispensing a sample of a solution of macromolecule and crystallising agent; (ii) incubating the sample for a chosen period of time under chosen conditions; (iii) assessing the sample using Dynamic Light Scattering (DLS) (or dedicated volume element in the sample); (iv) on the basis of the assessment, either reiterating steps (ii) and (iii); or (v) incubating the sample for a second chosen period of time under conditions which differ from the conditions of step (ii). A method for producing macromolecular crystals comprising the steps: (i) dispensing a first sample and a second sample of a solution of macromolecule and crystallising agent; (ii) incubating the samples for a chosen period of time under the same chosen conditions; (iii) assessing the first sample (or dedicated volume element in the sample) using Dynamic Light Scattering; (iv) on the basis of the assessment, either reiterating steps (ii) and (iii); or (v) incubating the second sample for a second chosen period of time under conditions which differ from the conditions of step (ii), wherein assessment of the first sample using Dynamic Light Scattering is performed on a scattering volume of in the order of 50 µm x 50 µm, observed at an angle of between 80 and 100°, preferably 90° to the incident light, which is preferably of wavelength 689.5 nm, and/or step (v) is chosen when the assessment indicates that the distribution of calculated hydrodynamic radii has changed from at least one previous iteration of step (iii) so that a chosen proportion of the particles fall within a distinct sub-population which has a higher calculated hydrodynamic radius mode (aggregate mode) than the calculated hydrodynamic radius mode (for the population as a whole) present during at least one previous iteration of step (iii) (monomer mode); or so that there is a chosen increase in the relative number of particles falling within the said distinct sub-population when compared with the number of particles falling within the said distinct sub-population during at least one previous iteration of step (iii).

Description

USE OF DYNAMIC LIGHT SCATTERING (DLS) IN A METHOD FOR PRODUCING MACROMOLECULAR CRYSTALS.

The present invention relates to techniques useful in optimising crystallisation of macromolecules, and their application to automated and high throughput systems.

The subject of crystallisation, especially protein crystallisation, has gained a new strategic relevance in the next phase of the genome project in which X-ray crystallography will play a major role. There have been major advances in the automation of protein preparation and also in the X-ray analysis and bio-informatics stages once diffraction quality crystals are available. But these advances have not yet been matched by equally good methods for the crystallisation process itself. Automation is crucial for high throughput crystallisation as well as for the other phases of structural genomics since the search for good crystals requires the testing of many different crystallisation conditions. In the area of crystallisation, the main effort and resources are currently being invested into the automation of screening procedures to identify crystallisation conditions. However, in spite of the ability to generate numerous trials, so far only a small percentage of the proteins produced have led to structure determinations. This is because screening in itself is not usually enough; it has to be complemented by an equally important procedure in crystal production, namely crystal optimisation.

The real stumbling block in structural genomics has become apparent from various pilot projects which are currently under way. These show that the success rate of getting from cloned protein to structure determination is only about 5-10%. For example, figures taken from the Human Proteome Structural Genomics pilot project [Brookhaven National Laboratory, The Rockefeller University and Albert Einstein College of Medicine: http://proteome.bnl.gov/progress.html] show that, out of 124 proteins which were cloned, 62 were purified. Of these 62, 33 yielded some crystals of some sort. However, only 16 of these crystals were of good enough quality to be useful for structure determination and only 12 have been solved to date (Fig. 1). Clearly this is highlighting a general problem where, even if proteins can be cloned, expressed, solubilised and purified, and even if crystallisation trials do yield some crystals, this usually does not guarantee that the crystals will be good enough for the structure to be solved. For structural genomics to be productive, it is essential that this problem is addressed.

Recently major advances have been made both in the automation of protein expression and purification methods [1] and in X-ray analysis [2] and modelling [3]. In the case of crystallisation, now that commercial screening kits and computer algorithms for designing arrays of potential conditions are readily accessible, it is no longer a major problem to dispense trials automatically [4,5]. Automatic generation of high throughput screening crystallisation trials is also underway [2,6].

The first semi-high-throughput experiments for both screening and optimisation were designed in 1990 as microbatch trials under oil [4]. Microbatch trials consisting of 0.7 - 2 μl drops of a mixture of protein and crystallising agents are generated by an automated liquid dispensing system and are dispensed and incubated under oil in order to prevent evaporation. The automated liquid dispensing system may have two modes of action: one to automatically screen numerous potential crystallisation conditions and the other for optimisation of the most promising screening conditions using a matrix survey [9,10]. The microbatch method has established a unique way of crystallising macromolecules, and many target proteins have been successfully crystallised using it [e.g. 11,12,13]. In its current state, a single automated liquid dispensing system machine can generate about 2000 trials per day. Because batch is mechanically the simplest crystallisation method, this procedure lends itself for adaptation to high-throughput crystallisation. The microbatch method has already been adapted for high throughput screening experiments in the USA using a large bank of syringes dispensing 0.4μl volumes into 1536-well micro-assay plates [Luft, J.R., et al. J. Cryst. Growth 232, 591-595 (2001)].

However, there are a number of issues that still require attention. Some proteins will surely crystallise during this initial screening, but most trials are likely to yield microcrystals or low-ordered crystals. The conversion of such crystals into useful ones requires intellectual input and individualised optimisation techniques. Such techniques do not lend themselves readily to automation and they have yet to be adapted to cope with the huge volume of experiments required by Genome Projects. Consequently, the subject of optimisation has been rather neglected, apart from the obvious first step of merely changing the concentrations or pH around the conditions of interest.

The ultimate way to control the crystallisation process is to separate the phases of nucleation and growth, i.e. to start the process at conditions which induce nucleation and then transfer the system to metastable conditions, which promote optimal growth. Methods to achieve this other than seeding involve changing the temperature [Rosenberger, Lesley] or diluting microbatch drops after incubating them for a given time at spontaneous nucleation conditions [Saridakis et al (1994) Ada Cryst D50, 293-297]. Dilution of microbatch drops showed that the optimum time for dilution was long before the appearance of the first visible microcrystals [Saridakis et al (1994) Acta Cryst D50, 293-297]. This method has recently been adapted by Saridakis & Chayen, 2000) to vapour diffusion, with similar results. As in the case of changing temperature, these techniques yielded improved crystals, but were very time-consuming, since many processes of trial and error were required to determine the right time at which to dilute (the time scale could only be guessed at by reference to the time which it took to see the first crystals). Consequently this method has not yet been adopted for routine use.

For both the incremental dilution and temperature ramp cycling approaches the most effective moment to intervene with a crystallisation experiment is soon after the formation of the first critical size nuclei which will eventually form the crystal. Although the resolution of a light microscope with oil immersion method can reach approx. some 420nm when illuminating with 550nm (e.g. Leitz planachromat NPL oil), by the time nuclei or crystals can be observed under the microscopes used in practise they have already reached a size of approximately 2-5 microns, and by then it is too late to act, since too many nuclei are likely to have already formed.

We have developed optimisation methods which are suitable for use in automated and high-throughput crystallisation trials. We show that Dynamic Light Scattering (DLS) can be applied in a very practical way to determine the time at which to intervene with a crystallisation experiment and lead it from nucleation to growth conditions. We show that the time at which DLS shows a significant change in the size-distribution profile of species in solution corresponds to the time at which the solution can be effectively transferred to metastable conditions, for optimal growth. The techniques presented here identifies the most likely time for nucleation- growth decoupling to be performed successfully, thus opening the scope for this technique to become routine and high-throughput.

A first aspect of the invention provides a method for producing macromolecular crystals comprising the steps

(i) dispensing a sample of a solution of macromolecule and crystallising agent (ii) incubating the sample for a chosen period of time under chosen conditions (iii) assessing the sample (or selected volume element in the sample) using Dynamic Light Scattering (iv) on the basis of the assessment, either reiterating steps (ii) and (iii) or

(v) incubating the sample for a second chosen period of time under conditions which differ from the conditions of step (ii).

Assessment by Dynamic Light Scattering (DLS) is performed by providing spatial and time resolution during dynamic experiments. A dedicated volume element in the sample may be analysed by DLS; it is not necessary for the entire sample volume to be assessed by DLS. DLS is used to assess the progress of crystallisation in the sample, so that crystallisation conditions can be changed at an appropriate time to promote formation of high quality crystals. Thus, the conditions of step

(ii) are preferably nucleation-promoting conditions. The conditions of step (v) are preferably growth-promoting conditions (ie metastable conditions). The conditions of step (ii) and step (v) may differ in relation to one or more of temperature, concentration of the sample (for example, the sample may be diluted between steps (ii) and (v)), concentration of the solution in the equilibration reservoir (using vapour diffusion methods), or the effectiveness of a barrier to diffusion from the sample (for example changing the thickness or composition of an oil layer over the sample or over the equilibration reservoir).

The DLS assessment may be performed essentially continuously, for example with no or only short breaks (for example 1 to 10 seconds) between periods of measurement (which may be, for example, of about 5 to 40 seconds, for example 20 seconds, duration, as described in Example 1); thus much of the incubation time of step (ii) (during which the crystallisation sample "matures") may also be used for DLS assessment.

It may be desirable to perform DLS measurements on more than one volume element within the sample, and to decide when to change the conditions depending upon the overall aggregation observed. Alternatively, it may be desirable to consider individual or adjacent groups of volume elements separately, and to change the conditions when and/or where aggregation is seen in one or adjacent volume elements even if aggregation is not seen in other volume elements within the same sample. Aggregation is a dynamic process and measurements made in one volume element may not be representative of other volume elements within the same sample. For example, two (or more) solutions may diffuse into each other. In such a case, it is possible to tell using DLS whereabouts in the sample crystallisation is occurring. This position may change with time, as the diffusion continues. It may be desirable to change the sample conditions by moving crystallisation nuclei within the sample, for example relative to an travelling solution front, or by changing conditions locally where crystallisation is occurring, for example by applying a stimulus (such as a change in temperature or electric field) locally. Thus, the invention provides a method for producing macromolecular crystals comprising the steps

(a)dispensing a sample of a solution of macromolecule and crystallising agent

(b)incubating the sample for a chosen period of time under chosen conditions and/or (c)incubating the sample within a chosen volume element under chosen conditions (d)assessing the sample using Dynamic Light Scattering (by providing spatial and time resolution during dynamic experiments) (e)on the basis of the assessment, either reiterating steps (b) (c) and (d) or (fjincubating the sample for a second chosen period of time or at a chosen volume element under conditions which differ from the conditions of step (b).

Spatial sample manipulation may be equivalent to time-determined manipulation for dynamic experiments, when, for example, one specimen diffuses into another one as for macromolecular crystal growth. Conditions may vary spatially and temporally in similar manners; for example concentrations may be different at different times at a given volume element, and/or may be different at different volume elements at the same time.

Dynamic light scattering (DLS) offers a size resolution of "particles" in optical transparent aqueous samples some three orders of magnitude below an optical microscope, and consequently forms a useful tool for an early, non-invasive, insitu observation of a crystallisation event, before it becomes visible with a light microscope. The terms "Photon correlation spectroscopy" (PCS), quasi-elastic light scattering (QELS) and DLS can be considered synonyms. A laser is focussed onto the protein solution as the aggregation and nucleation processes are occurring and the light scattered by the particles within the solution, such as protein molecules or aggregates, is collected. The nucleation in super saturation in the sample bulk is the reason that events recorded in the scattering volume of approx. 50μm times 500μm are bulk representative. The time constant(s) of the second order Auto-Correlation Function (ACF) of the scattered light intensity delivers the diffusion coefficient(s), and hence the hydrodynamic radii of the particles present. DLS is sensitive to variations in particle size (in the range of approx. >lnm) and interactions of protein molecules in solution [Schmitz, S.K. (1990) "An Introduction to Dynamic Light Scattering by Macromolecules", Academic Press, New York; Budayova- Spano et al. J. cryst Growth 235 (2002) 547-554]. The above value of >lnm can be much better for hard spheres. The value depends on the molecule and the sample system. For macromolecules, when using ultra clean samples and not a dynamic experiment, i.e. quiescent solution, a reproducibility of +/- 0.02nm (0.2 Angstroem) has been achieved for a specimen with a radius of 2.78nm (1998, STS-95 Data evaluation, Dieckmann and Dierks unpublished and communicated to Garcia-Ruiz). The size sensitivity is affected by variables of the macro-molecule such as pH, ionic strength, etc. The known x-ray data of the molecule size may also be considered.

DLS is routinely used in many labs to assess sample mono-dispersity using dilute protein samples [DArcy, A. Acta Cryst D 50, 469-471 (1994); Bergfors, T. Ch. 4 in "Crystallisation of Proteins: Techniques, Strategies and Tips" Bergfors, T. ed (International University Line, USA) 1999, 27-40; Ferre-DAmare and Burly S. K. Structure 1994 2, 357-359]. It has also been used successfully with lysozyme to show an increase in hydrodynamic radius as supersaturation proceeds [Mikol et al (1990) J Mol Biol 213, 187-195; Georgalis & Saenger (1999) 82(4), 271-294; Peters et al (1998) Acta Cryst D54, 873-877]. However, there is no suggestion that the technique can be used in choosing when to change crystallisation conditions.

The term DLS is not considered to cover turbidimetric measurements of the type described in Rosenberger et al (1993) J Crystal Growth 129, 1- 12. The methods described determine only the general extinction properties of the sample and have poor temporal and spatial resolution. Particle sizes cannot be calculated using the methods described in Rosenberger et al. DLS is characterised by detection and analysis of the temporal behaviour of the scattered light intensity fluctuations, as discussed in, for example, Georgalis & Saenger (1999) Science Progress 82(4), 271-294. Fluctuations over time periods of between about 500ns to minutes may be detected. DLS allows particle sizes to be calculated.

We have found that changes in the aggregation profile of a supersaturated protein sample at nucleation conditions as a function of time as determined by DLS can be used as an indicator of the induction time for nucleation, i.e. the time at which post-critical nuclei start their existence

[Malkin A.& McPherson, A. J. Cryst growth 1993 128, 1232, Ataka, M.

Chapter 6 in Crystallization Processes ed Ohtaki, H. Wiley and sons,

1998)].

Preferred characteristics of the DLS assessment are as described below in relation to the second aspect of the invention.

A second aspect of the invention provides a method for producing macromolecular crystals comprising the steps (i) dispensing a first sample and a second sample of a solution of macromolecule and crystallising agent

(ii) incubating the samples for a chosen period of time under the same chosen conditions (iii) assessing the first sample (or dedicated volume element in the sample) using Dynamic Light Scattering

(iv) on the basis of the assessment, either reiterating steps (ii) and (iii) or

(v) incubating the second sample for a second chosen period of time under conditions which differ from the conditions of step (ii), wherein assessment of the first sample using Dynamic Light Scattering is performed on a scattering volume of in the order of 50 μm x 50μm, observed at an angle of between 80 and 100°_, preferably 90° to the incident light, which is preferably of wavelength 689.5 nm, and/or step (v) is chosen when the assessment indicates that the distribution of calculated hydrodynamic radii has changed from at least one previous iteration of step (iii) so that a chosen proportion of the particles fall within a distinct sub-population which has a higher calculated hydrodynamic radius mode (aggregate mode) than the calculated hydrodynamic radius mode (for the population as a whole) present during at least one previous iteration of step (iii) (monomer mode); or so that there is a chosen increase in the relative number of particles falling within the said distinct sub-population when compared with the number of particles falling within the said distinct sub-population during at least one previous iteration of step (iii).

The scattering volume, scattering angle and wavelength indicated above (and further as discussed in Example 1) have been demonstrated by the inventors to be particularly useful or convenient in selecting when to change crystallography conditions. However, it is considered that other scattering volumes, scattering angles or wavelength may also be used. The scattering volume, for example, is chosen partly depending on the scattering angle. If the spatial resolution of DLS is used for the incubation strategy of the sample (ie if it is important to analyse a very small volume, for example if samples are very close together, or if one part of the sample, for example adjacent to a surface from which solvent evaporation is taking place, is to be monitored), then it is desirable for the scattering volume to be small. However, one (radius) limit is given by Rayleigh at around r= 26 μm in air, which gives a minimum of about 52μm for the diameter. The length dimension is dependant on the observation angle.

The ACF may be analysed by inverse Laplace transformation, which delivers the critical relaxation time, which equals the inverse product of the scattering vector and the diffusion constant and a factor 2. The diffusion coefficient(s) of the specie(s) in solution (calculated from the ACF decay) may then be replaced in the Stokes-Einstein equation to give the hydrodynamic radii (rh), if the viscosity is known, or the viscosity (Eta) if the hydrodynamic radius is known. These calculations are well known to those skilled in the art (see, for example, Dierkes et al (1998) Graefe 's Arch Clin Exp Ophthalmol 236, 18-23 and references cited therein). For deciding when to change conditions, the diffusion coefficients are sufficient, but it is a straightforward step to go from them to the radii and therefore preferred to use the calculated radii, which are considered easier to interpret. It is possible to use the ACF itself in assessing when to change the conditions but it is preferred to use the inverse Laplace transformation result, calculated hydrodynamic radii or diffusion coefficients as it is considered easier to interpret these measures.

It is preferred that the chosen proportion of particles (ie proportion of particles falling within a distinct sub-population which has the higher calculated hydrodynamic radius mode (aggregate mode)) is at least 20, 30, 40, 50, 60, 70, 80 or 90% of the total particles (by number).

It is preferred that the chosen increase in the relative number of particles falling within the said distinct sub-population which has the higher calculated hydrodynamic radius mode (aggregate mode) is at least 1.5- fold, preferably, 2, 3, 5, 10, 15, 20, 30, 40, 50, 80, 100, 200, 500 or 1000- fold, for example relative to the number falling within the said subpopulation when the sample has calmed following set-up. It is not considered necessary to determine the absolute number of particles within the subpopulation either following set-up or at later time.

For optimum results it may be desirable to perform dilutions (or other manipulation of the crystallography conditions) at different such proportions in order to determine the optimum proportion or range of proportions or relative increase (as appropriate) at which to change the crystallography conditions. For subsequent crystallisation samples for the same or similar macromolecule, crystallisation conditions may be changed when the optimum proportion or a proportion in the range of optimum proportions or optimum relative increase is reached. Macromolecules may be "similar" when they share features such as overall conformation (eg globular or extended), surface charge, isoelectric point or solubility characteristics. Similar macromolecules may preferably share extensive sequence homology (for example at least 40, 50, 60, 70, 80, or 90% amino acid or nucleotide (as appropriate) identity), but this is not essential for macromolecules to be considered similar in terms of the optimum proportion of particles, or optimum relative change in number of particles, in the higher hydrodynamic radii population when crystallisation conditions are changed. Crystal formation may be assessed by methods well known to those skilled in the art, for example microscopy or diffraction studies. The process of performing and analysing trials for determining optimum proportions as discussed above may be automated and a knowledge-based system developed to predict optimum proportions for further macromolecules.

When the precision of the DLS measurements is very good, for example due to a very clean sample, it may be possible to detect even small changes in aggregation or the very first occurrence of aggregates, and this may be interpreted as an indication to change conditions. But when the systems show considerable aggregation from the beginning, it may be necessary to set the criterion somewhat higher, for example require a higher proportion of the particles to be in the higher hydrodynamic radius population.

It is will be appreciated that the calculated hydrodynamic radius either for "monomers" or "aggregates" will depend upon the size of the macromolecule, so the relevant calculated hydrodynamic radii will vary depending upon the nature of the macromolecule to be crystallised. There are monomers of say 0.4nm diameter and there are aggregates at 2000nm and there may be, for example, four size distribution classes in between. "Monomer" and "aggregate" sizes may be determined in "trial runs", for example using a machine as briefly described above. A typical "monomer" hydrodynamic radius mode for a polypeptide of about 14 to 25 kDa may be about 2nm, whilst a typical "aggregate" hydrodynamic radius mode could be about 5 to 200 nm. For example, lysozyme (14.5 kDa) has a 2.1nm monomer, the dimer of which is 2.65nm (Georgalis et al.(1995) Adv. Coll. Interf. Sci. 58, 57-86). For thaumatin dimers (2x22 kDa) we have ca. 3nm hydrodynamic radius (Juarez-Martinez et al.(2001) J.Cryst.Growth, 232, 119-131) whereas an assembly of nine dimers of Human Transferrin Receptor (i.e. 18x70kDa) is 16nm (Schueler et al. (1999) Biophys. J. 77, 1117-1125).

Typically the "aggregate" modal hydrodynamic radius will be at least 1.5, 2, preferably, 5, 10, 20, 50, 100, 1000, 10000 or more times the "monomer" modal hydrodynamic radius. The hydrodynamic radius of a crystal may be around 1000 000 000 000 times the monomer radius in size, but it is likely that the optimum aggregate size at which crystallisation conditions are changed would be smaller than this.

Preferably the change of conditions is made within 1, 10, 20, 30, 40 or 50 minutes, or 1, 2 or 3 hours, of the measurement (or last of a group of measurements) which is interpreted as showing that the chosen proportion of particles being in the "aggregate" population has been reached. Still more preferably the change is made within 2 hours, most preferably within 30 minutes, still more preferably within 10, 5, 2 or 1 minutes of the relevant measurement.

In the second aspect of the invention, measurements are made on one sample are used to determine when to change the conditions of further (parallel) samples. The monitored sample and parallel samples may differ only in relation to volume and the way in which they are contained. Thus, it is preferred that the monitored sample and parallel sample are substantially identical, with the possible exception of volume and container. For example, the monitored sample and parallel sample have the same concentrations of macromolecule and crystallising agent.

Although it is preferred that DLS is performed directly on the sample from which it is hoped to obtain crystals, it may be advantageous to use parallel samples, only one of which is monitored using DLS. This may be useful in detailed optimisation of the time for changing conditions. For example, as described further in Example 1, multiple further samples may be used, with different samples being transferred to different conditions, and/or at different times based on the DLS results. Successful crystallisation protocols may then be replicated to obtain further crystals. The samples may be observed by an automated microscope and image analysis for input to the "feed-back" or knowledge-based system.

It will be appreciated that DLS measurements may be made on several different volumes within one crystallisation sample. Thus, measurements may be made on two or more 50μm-sampled volumes separated by, for example, lOOμm. Parallel sampling (for example using two lasers of different wavelengths) may be used, as discussed further below. This may have the desirable effect of increasing statistical significance/reproducibility, which is needed.

By diluting (for example), at various times after set up, batch drops set at the same (nucleation) conditions as the solution from which DLS data are being recorded, the optimum DLS result for dilution may be determined. Indeed, if the dilutions result in the solution being brought to metastable levels of super saturation, solutions that contain post-critical nuclei at the time of dilution will sustain their growth into visible crystals, whilst pre- critical ones will be dissolved. Example 1 describes experiments in which we have monitored the end results of microbatch crystallisation experiments, in which the solution is diluted to metastable conditions at various times after set up, and the results compared with changes in the size-distribution time-profiles as resolved by DLS from an identical solution, at corresponding times. Preferably the sample assessed by DLS is contained in a glass or plastic cuvette. The container needs to be highly transparent for the radiation used (at least for the portions of the container through which the radiation is required to pass) and preferably has substantially static properties in time, and temperature gradients for refractive index. Plastics and glasses generally are suitable materials. The sample may be in the form of a hanging drop, in which case a fibre-optic connector may be useful in ensuring that the sample is illuminated by the laser(s). Techniques described in EP 1 022 549 may be particularly useful when using birefringent materials (which includes many plastics) for the container. The techniques may allow the effects or contributions from the container to be eliminated. The techniques may also allow multiple volume elements, which may be in multiple samples (for example in a single multiwell plate) or within a single crystallisation sample, to be analysed in parallel, yielding output (for example calculated hydrodynamic radius data) as a function of location on the plate. Thus, individual volume elements or individual sample maturity may be assessed and individual samples ready for manipulation identified from analysis of multiple samples in parallel. The techniques make use of light of at least two wavelengths.

In relation to both preceding aspects of the invention, it is preferred that the DLS-apparatus is able to handle a crystallisation sample in standard cuvettes of only 20 μl or less. For example, the apparatus used in Example 1 may be used with sample volumes of less than about 7 μl. It is preferred that the sample volume is of less than 0.5, 1, 2, 3, 4, 5, 6, 7, 10, 15, 20 or 30 μl. The smaller volumes may be more economical with possibly scarce macromolecules, whilst larger volumes may give more reliable data. Handling difficulties may mean that volumes of at least 1 μl are preferred._There is a theoretical limit, when the Brownian motion of the particles is disturbed by the container wall. Previous experiments indicate that this is the case when the dimension of the container reaches 10 or 20 times the particle diameter. Strange effects may happen then. The surface tension of the oil (when used) may also induce currents which may make measurements difficult on small samples. For example, surface driven phenomena (Oil/Droplet) may be seen in the bulk of oil-covered droplets at a volume of 2μl which may be detected as oscillating ACF's. This may impose a limit, which may be overcome by using a precise automated san unit ■ operated at the droplet, but it may be difficult using smaller volumes in conjunction with oil.

The DLS apparatus termed DIMINIGON-A (Dierks and Partner, Hamburg, Germany) may be particularly suitable for performing methods of the invention.

The macromolecule may be any macromolecule, but it is preferred if it is a biological macromolecule. The biological macromolecule may be any biological macromolecule including nucleic acids, complex polysaccharides and viruses. Preferably, the biological macromolecules are polypeptides. A polypeptide comprises at least one chain of amino acid residues which are covalently joined by peptides bonds. A polypeptide chain may have any number of amino acid residues, preferably at least two, more preferably at least 100, 500, 1000 or 2000. The polypeptide chain may have more than 2000 residues. A polypeptide may contain residues in the chain which are unusual or artificial, and may comprise non-peptide bonds such as disulphide bonds. The residues may be further modified, for example to include a phosphate group or a sugar chain (eg an oligosaccharide) or a lipid moiety. The term polypeptide includes glycoproteins and lipoproteins as well as other post- translationally modified polypeptides. A polypeptide may comprise more than one chain (for example, two chains linked by a disulphide bond between the sulphur in the side chain of cysteine residues), and may further comprise inorganic or organic co-factors or groups. Such modifications and additions are included within the term "polypeptide".

Crystallisation agents are known in the art and the composition may be optimised according to the nature of the macromolecule to be crystallised. Typical crystallisation agents include salt(s) and buff er(s).

The crystallisation agent and macromolecule may be dispensed at the same time, or may be dispensed separately or sequentially in any order. The crystallisation agent and macromolecule may be dispensed under oil or may subsequently be covered by oil (for example light paraffin oil; BDH, UK). When parallel samples are used, both the monitored sample and the parallel sample(s) may be covered by (ie dispensed under or subsequently covered by) oil, so that the rate of change of concentration of the monitored sample and the parallel sample(s) are effectively equal (ie within about 20, 10, or 5%).

The rate of change may be negligible. By "negligible" we mean the evaporation from the sample drop under the oil is undetectable over a period of at least a day, preferably over a period of at least 2 days, or 5 days or a week. Preferably, negligible evaporation is a loss of water from a solution which is sufficiently small that it cannot be detected after a period of at least two weeks or 1 month or 2 months or 3 months. Evaporation from the sample drop may be judged by any suitable means, including by assessment of the size of the sample drop, or by the appearance of dryness. The oil may be any suitable liquid oil. Advantageously, the oil is of a density lower than that of the macromolecule/crystallisation agent solution. This is because the oil is preferably one which is used to overlay a crystallisation drop. If the oil were denser than the liquid in the crystallisation drop, it would fail to "sit" on top of the drop. Preferably, the oil has a density of around 0.84 g cm^"3. Furthermore it is preferred if the oil dispensed by the system is one which can act as an inert sealant and does not interact with crystallisation trials, for example, one that does not cause precipitation. Hence, it is preferred if the oil consists of or comprises paraffin. More preferably the oil consists of paraffin light, a purified mixture of liquid saturated hydrocarbons obtained from petroleum. A suitable paraffin is one such as is available from Hampton Research, CA 92677-3913 USA under catalogue no HR3-411.

Preferably, for example when the sample is covered with a thin layer of oil which permits evaporation from the sample, the concentration of the macromolecule solution prior at the start of the incubation of step (ii) is undersaturated or metastable. In other words, the concentration of the solution is outside the nucleation zone of the phase diagram of that solution. It is preferred that during the incubation of step (ii) the concentration of the macromolecule solution increases (for example by evaporation of the solvent from the solution) so that the concentration of the solution reaches a concentration within the nucleation zone.

DLS is used to detect when nucleation has occurred. Once nucleation has occurred, the conditions under which the solution is incubated are changed.

For example, the thickness of the layer of oil covering the crystallisation sample may be increased, as described in GB 0108289.0 and Chayen, Ada Crst D, submitted. Alternatively, a layer of oil on the reservoir (for hanging drop or related techniques) may be added or thickened, as described in Chayen (1997) J Appl Cryst 30, 198-202 or D'Arcy et al. (1996) J. Crystal Growth 168, 175-180.

If the paraffin oil layer is less than about 3.5 mm (for example between 0.7 to 1.2 mm), the layer will allow evaporation of the solvent (for example water) from the solution of the macromolecule (or from the reservoir, as appropriate) because the thickness of the oil layer is insufficient to render the evaporation negligible. Typically, the thickness of the layer of oil at which evaporation ceases to be negligible is less than 3.5mm.

Preferences in relation to oil layers and changes to the thickness of oil layers are as discussed in GB 0108289.0, Chayen (1997) J Appl Ciyst 30, 198-202 or D'Arcy et al. (1996) J. Crystal Growth 168, 175-180.

Such changes may result in the sample remaining within the nucleation zone of the solution phase diagram, but may slow progress further into the nucleation zone or any other unwanted position in the phase diagram not contributing to delivery of crystals. An example of a crystallisation phase diagram is shown in Chayen et al (1996) Q Reviews of Biophysics 29:227- 278.

Alternatively, changes may be made which move the sample from the nucleation to the metastable zone, so that existing nuclei may grow, but further nucleation is limited. For example, the sample may be diluted by the addition of further liquid, for example solvent, for example water or aqueous buffer solution, which may also comprise further crystallisation agent, but which preferably does not comprise further macromolecule, to the sample. Alternatively, the sample may be transferred to conditions under which further solvent enters the sample by vapour diffusion. Dilution by the addition of a volume of between about 5 and 15% of the sample volume prior to the dilution may be appropriate.

A reduction in the rate of evaporation may be suitable when the initial sample (ie as set up) is in the metastable zone. When the initial sample is in the supersaturated zone (ie as set up), dilution of the sample is preferred.

Alternative or further changes in conditions may include addition of material, for example the same or other macromolecules, or application of physical parameters such as voltages, temperture cycling. It may also be desirable to make such changes when stagnation of nucleation or crystal growth is detected.

Setting up and/or dilution of the samples and/or addition of oil may be performed using an automated liquid dispensing system. The automated liquid dispensing system may be any suitable dispensing system, including the IMP AX and Oryx 6 systems available from Douglas Instruments, Hungerford, Berks, UK, for example adapted to be capable of dispensing oil.

Preferably the automated liquid dispensing system and DLS measuring apparatus (and preferably also a microscope for examining the samples (for example using image analysis techniques) are integrated such that DLS measurements and addition of liquid (for example solvent and/or oil) may be performed by an integrated machine. For example, a whole microtitre plate (or multiple microtitre plates) of samples may be read, followed by solvent and/or oil being added to samples which had reached the necessary degree of nucleation. The microtitre plase may be continuously monitered by a combined optical/micro-dispenser unit.

A third aspect of the invention provides a use of a Dynamic Light Scattering measuring apparatus in the method of the first or second aspect of the invention. The DLS measuring apparatus is preferably a DIMINIGON-A apparatus. The DLS measuring apparatus may preferably comprise two or more lasers suitable for producing light of different wavelengths, as discussed in EP 1 022 549.

A fourth aspect of the invention provides a use of an automated liquid dispensing system in a method of the first or second aspect of the invention. The method preferably comprises the step of, on the basis of the assessment, adding liquid and/or oil to the sample using the automated liquid dispensing system. The automated liquid dispensing system is preferably an Oryx 6 or IMP AX system.

The invention will now be described in more detail with the aid of the following Figures and Examples:

Figure Legends

Figure 1. DLS measurements with Trypsin. The size of the species is displayed on the Y axis and succesive measurements in the series are spread along the X axis. Increasing brightness represents the relative change of each species contained in the solution as a function of time. The brightness scale is "normalised" with the most numerous species at a given time being assigned the maximum value of 1. A) After mixing; B) t=3h to 3h20min; C) t=4h to 5h; D) t=5h to 6h; E) t=6h to 7h30min; F) t=7h30min. Figure 2. DLS measurements with Thaumatin. Curves show the relative numbers versus size of the species in solution are plotted separately for each measurement of the series.

Example 1.

In this Example we have monitored by DLS the crystallisation of proteins mixed with their crystallising agents, so as to get an indication as to when to dilute the trial in order to lead it out of the nucleation phase and into the growth phase. This was achieved using a DLS-apparatus, which is able to handle a crystallisation sample in standard cuvettes of only 20 μl. This volume can be decreased by using specific containments significantly down to below 7 μl, if needed.

We have tested our hypothesis that changes in the aggregation profile of a supersaturated protein sample at nucleation conditions as a function of time can be used as an indicator of the induction time for nucleation, i.e. the time at which post-critical nuclei start their existence Malkin & Pherson, Ataka). By diluting, at various times after set up, batch drops set at the same (nucleation) conditions as the solution from which DLS data are being recorded, this hypothesis can be qualitatively verified. Indeed, if the dilutions result in the solution being brought to metastable levels of super saturation, solutions that contain post-critical nuclei at the time of dilution will sustain their growth into visible crystals, whilst pre-critical ones will be dissolved. We have therefore monitored the end results of microbatch crystallisation experiments, where the solution was diluted to (known) metastable conditions at various times after set up, and compared these results with changes in the size-distribution time-profiles as resolved by DLS from an identical solution, at corresponding times. Materials and Methods

A DLS apparatus (DIMINIGON-A , Dierks and Partner, Hamburg, Germany) and software package also commercially available from the same entity was used following the DLS-system description as briefly given in Dierks et al (1999) Grafe 's Arch Gin Exp Ophthamol 236, 18-23 and the references mentioned therein. For easy reference the main DLS- system characteristics are briefly given. The solid state laser with 30mW, lasing at 689.5nm is temperature controlled at room temperature with an accuracy of approx. +/-0.05 ° C and illuminates a scattering volume spanned by receiver and transmitter collimators to yield a 50μm x 50μm sized scattering volume observed for data presented here at an angle of 90°. The mono-mode receiver is fed into a photo-multiplier tube with a quantum efficiency ranging from 5 to 7 % at this wavelength, which delivers via pulse shaper the signals to a real-time correlator with a sampling time structure starting at 800ns and a sampling time structure spanning 9 decades. A computer system accomplishes all operations such as temperature stabilization of cuvette holders, data evaluation and selection of measurement times, which are mostly 20 to 80s for the data presented here. The error estimates on the data quality are fully compliant with accuracies delivered in Ref. Dieckmann & Dierks (2000) Conf Mat & Nanotec Instrumentation for material synthesis and near field optical microscopy SPIE 4098, 11-25. The instrument is designed to take reliable measurements from as little as 10 or 20-30μl of solution, contained in a glass cuvette which can be covered with light paraffin oil, thus closely mimicking the conditions of a microbatch experiment. The software uses CONTIN (e.g. Provencher (1982) Computer Phys Comm 21, 229-242) to analyse the Auto-Correlation Function (ACF) of the intensity fluctuation, and is equipped with a variety of possibilities for displaying and evaluating data, some of which are shown in the Results.

Two proteins were used in this study: porcine pancreatic trypsin type IX (cat. no. T-0134) and thaumatin from Thaumatococcus danielii (T-7638) purchased from Sigma (Steinheim, Germany) and used without further purification. Furthermore, hen egg-white lysozyme (Sigma L-6876) was used to compare the obtained hydrodynamic radii with the ones published to allow for confirmation on appropriate values for the variables such refractive index and viscosity for quiescent systems, and to assess the reliability of the size estimates given by the apparatus. As outlined above, the reliability, accuracy and reproducibility was good if the supersaturated solution was filtered through 0.1 μm mesh size filters or smaller. All salts and buffers used were also purchased from Sigma.

The solutions, from which DLS- measurements were taken, were dispensed (30μl) in the glass cuvettes that are furnished with the apparatus, whereas parallel microbatch experiments were set up in Terasaki-type plates (Nunc, Denmark) covered with light paraffin oil (BDH, U.K.). Many identical 5μl crystallisation drops were set up for each protein, at conditions known to promote nucleation and the rapid growth of fairly small crystals. The same conditions were used for the DLS measurements. These conditions were: a) for trypsin: 20 mg/ml protein, 34 %(sat) ammonium sulphate, lOOmM

b) for thaumatin: 32 mg/ml protein, 0.45M sodium potassium tartrate, 50mM PIPES pH 6.7.

For each protein, lOOμl of solution at the above conditions was prepared and filtered through 0.1 μl mesh size micro-centrifuge filters (Ultrafree- MC, Millipore, Bedford, USA). The solution was then distributed between the DLS cuvette and the wells of a Terasaki plate.

A series of auto-piloted DLS measurements, with 20sec ACF acquisition time for each measurement and Isec stand-by between measurements was recorded, starting as soon as possible (approx. 5 to 15 minutes) after the mixing and filtering of the ingredients, when the solution had calmed. The stand-by time may need to be longer than Is due to the time the PC needs for the data analysis and for starting a new measurement (for example 5 to 10s). One could also have no stand by time if the inverse Laplace transformation is performed off-line. Further series were recorded at regular time intervals thereafter, as detailed below. Each series consisted of at least 20 measurements. This ensured both that distorted ACFs or ones with overflows could be identified and dismissed.

At the same regular intervals, microbatch drops were also diluted (two drops per interval) with filtered buffer solution, to metastable conditions. Metastable conditions had been determined beforehand for each protein; by establishing the super solubility curve around published conditions (Chayen et al (2001) J. Mol Biol. 312, 591-595). A super solubility curve separates the spontaneous nucleation zone of a crystallisation phase diagram from the zone at which the solution remains clear (metastable + unsaturated). The metastable zone is then the area of conditions below the super solubility curve, where nuclei transferred from the spontaneous nucleation zone (e.g. by dilution) will continue to grow. The method is described in more detail in Saridakis et al. (1994) and information on the proteins studied here was available to us from previous work (Chayen et al (2001) J. Mol Biol. 312, 591-595). In this case, one metastable condition for each protein was sufficient. These were: a) for trypsin: 18 mg/ml protein, 30%(sat) ammonium sulphate, lOOmM

b) for thaumatin: 20 mg/ml protein, 0.28M sodium potassium tartrate, 50mM PIPES pH 6.7.

Terasaki plates were examined by a light microscope with above said "standard low" resolution a few days as well as more than one month after each experiment, in order to determine differences in crystal sizes and optical quality as a function of the time after set up at which the drops had been diluted. The results were then compared with the corresponding series of DLS data showing the size distribution profile of different species in solution at the given time. Such "full-scale" experiments were performed twice for each protein. In addition, various incomplete experiments performed at other times (e.g. consisting of dilutions only, or for checking the repeatability of the DLS measurements) were in generally good agreement with the "full-scale" experiments.

Results

Trypsin

The results representative to the data obtained for these sample systems for a run of DLS- measurements with trypsin are presented in Figure 1. This mode of presentation displays the size of the species on the y-axis and successive measurements in the series spread along the x-axis. A brightness scale represents the relative change of each species contained in the solution as a function of time. However, without further assumptions, derived from specific tests, it is not possible to assign a specific scale (e.g. concentration or number of particles) to this brightness representation. The brightness scale is 'normalised' with the most numerous species being always assigned the maximum value of 1. The more precise values for the peak centres and widths that are stated in the text have been read from histograms (not shown here) obtained directly from CONTIN.

Data collected very shortly after mixing and after sample calm down as described above (Fig.^" la) show a strong majority of size distribution for a small components possibly corresponding to monomers (peak slightly spread around 2nm) and a much smaller population of large size distributions (a peak that extends between 90 and 120nm). The aggregate peak vanishes after a short time. This situation remains practically unchanged until the series ending at t = 3h20' (only data for t = 3h - 3h20' are shown here; Fig. lb). In the series collected after 4 hours (Fig. lc) the weak peak corresponding to the larger size distribution has become stronger and more spread towards larger size (80-160nm), and in the following series, starting at 5 hours after set up (Fig. Id), the aggregate peak shifts to slightly larger size (100-180nm).

With the series starting 6 hours after set up (Fig. Ie), some disruption of both peaks starts to be taking place. The detected species are distributed in bimodal size distributions centred at 2nm and 175nm. The next series, starting at t = 7h30 (Fig. If) presents a completely different picture, with the virtual disappearance of the two peaks that had hitherto been representing the great majority of species in solution, and the appearance of many peaks above approx. 200nm. The situation remains relatively stable afterwards.

In the microbatch drops, a dilution time of 6 to 7 hours after set up reproducibly yielded larger crystals than the undiluted controls. Dilutions performed later usually resulted in crystals no better (by visual inspection) than the controls, whereas those performed earlier generally resulted in clear drops.

Thaumatin

The discontinuous change is less pronounced in the case of thaumatin, so a more sensitive method was used to display and assess the results (Fig. 2 for one of the runs). Curves showing the relative numbers versus size of the species in solution were plotted separately for each measurement of the series, and compared. Again, one should be cautious when interpreting the numbers as relative abundancies. For a series of measurements starting at 10*, 30', lh30*. 2h30', 3h30' and 4h40' after set up, almost every measurement in the series yields a curve very close to the typical ones shown in Fig. 2a-c (for clarity, three curves per series only are shown). These correspond to two kinds of species in solution, the smaller one possibly consisting of monomeres. (peak centred at ca. 2nm) and a smaller population of larger aggregates that slowly shifts to larger size.

In the series measured at t = 5h30', various extra peaks make their appearance and approximately half the measurements in the series now yield very different size-distribution profiles, the original peaks remaining nevertheless clearly dominant (Fig. 2d). The profiles remain fairly unchanged thereafter, with a slow progression of the small component resolved species towards larger sizes.

The microbatch drops set up in identical conditions remained clear when diluted between 30' and 2h30' after set up and yielded single crystals larger than those in the undiluted controls when diluted between 4 and 5h30' after set up. Those diluted later than 6 hours after set up yielded small crystals, no better (often worse) than those in the controls.

Conclusion

This study has focused on separating the nucleation and growth phases by diluting crystallising solutions from nucleation to metastable conditions. Dilution is a good method to use either when a protein is not temperature- sensitive or to avoid handling crystal seeds.

The results demonstrate that Dynamic Light Scattering can be applied in a very practical way to determine the time at which to intervene with a crystallisation experiment and lead it from nucleation to growth conditions. We have shown, for two model proteins, that the time at which DLS showed a significant change in the size-distribution profile of species in solution, corresponded to the time at which the solution can be effectively transferred to metastable conditions, for optimal growth. Shortly after the nucleation induction time, the solution will contain a small number of post-critical nuclei that can then, by dilution of the solution that they are in, as was done here, or by other methods (e.g. temperature shift), grow in the metastable zone of conditions. The method presented here will therefore pinpoint the most likely time for such nucleation-growth decoupling to be performed successfully in a given system.

References

1. Stevens, R.C. (2000). Design of high-throughput methods of protein production for structural biology. Structure 8, R177-R185. 2. Abola, E., Kuhn, P., Earnest, T. & Stevens, R.C. (2000). Automation of X-ray crystallography. Nature Struct. Biol. 7, 973-977.

3. Sanchez, R., Pieper, U., Melo, F., Eswar, N., Marti-Renom, M.A., Madhusudhan, M.S., Mirkovic, N. & Sali, A. (2000). Protein structure modelling for structural genomics. Nature Struct. Biol. 7, 986-990.

4. Chayen, N.E., Shaw Stewart, P.D., Maeder, D.L. & Blow, D.M. (1990). An automated system for microbatch protein crystallisation and screening. J. Appl. Cryst. 23, 297-302.

5. Stevens, R.C. (2000). High-throughput protein crystallization. Curr. Opin. Struct. Biol. 10, 558-563.

6. Luft, J.R., Wolfley, J., Bianca, M., Weeks, D., Jurisica, I., Rogers, P., Glasgow, J., Fortier, S. & DeTitta, G.T. (2000). Gearing up for ~40K crystallization experiments a day: meeting the needs of HTP structural proteomics projects. The 8^{t l} International Conference on the Crystallization of Biological Macromolecules Oral Presentation

Abstracts, p.29.

7. Mayans, O., van der Ven, P.F.M., Wilm, M., Mues, A., Young, P., Fuerst, D., Wilmanns, M. & Gautel, M. (1998). Structural basis for activation of the titin kinase domain during myofibrillogenesis. Nature 395, 863-869.

8. Normile, D. (1995). Search for better crystals explores inner, outer space. Science 270, 1921-1922.

9. Chayen, N.E., Shaw Stewart, P.D. & Blow, D.M. (1992). Microbatch crystallisation under oil - a new technique allowing many small- volume crystallisation trials. J. Cryst. Growth 122, 176-180.

10. Chayen, N.E., Shaw Stewart, P.D. and Baldock, P. (1994). New developments of the IMP AX small-volume crystallisation system. Ada Cryst. D50, 456-458. 11. Stock, D., Leslie, A.G.W. & Walker, J.E. (1999). Molecular architecture of the rotary motor in ATP synthase. Science 286, 1700- 1705.

12. Barrett, T.E., Sawa, R., Panayotou, G., Barlow, T., Brown, T., Jiricny, J. & Pearl, L.H. (1998). Crystal structure of a G:T/U mismatch- specific DNA glycosylase: mismatch recognition by complementary- strand interaction. Cell 92, 117-119.

13. Chayen, N.E. (1998). Comparative studies of protein crystallisation by vapour diffusion and microbatch. A a Cryst. D54, 8-15.

Claims

1. A method for producing macromolecular crystals comprising the steps of

(i) dispensing a sample of a solution of macromolecule and crystallising agent (ii) incubating the sample for a chosen period of time under chosen conditions (iii) assessing the sample using Dynamic Light Scattering (DLS)

(iv) on the basis of the assessment, either reiterating steps (ii) and (iii) or (v) incubating the sample for a second chosen period of time under conditions which differ from the conditions of step (ii).

2. A method for producing macromolecular crystals comprising the steps (i) dispensing a first sample and a second sample of a solution of macromolecule and crystallising agent

(ii) incubating the samples for a chosen period of time under the same chosen conditions

(iii) assessing the first sample using Dynamic Light Scattering (iv) on the basis of the assessment, either reiterating steps (ii) and (iii) or

(v) incubating the second sample for a second chosen period of time under conditions which differ from the conditions of step (ii), wherein assessment of the first sample using Dynamic Light Scattering is performed on a scattering volume of in the order of 50 μm x 50μm, observed at an angle of between 80 and 100°_> preferably 90° to the incident light, which is preferably of wavelength 689.5 nm, and/or step (v) is chosen when the assessment indicates that the distribution of calculated hydrodynamic radii has changed from at least one previous iteration of step (iii) so that a chosen proportion of the particles fall within a distinct sub-population which has a higher calculated hydrodynamic radius mode (aggregate mode) than the calculated hydrodynamic radius mode (for the population as a whole) present during at least one previous iteration of step (iii) (monomer mode); or so that there is a chosen increase in the relative number of particles falling within the said distinct sub-population when compared with the number of particles falling within the said distinct sub-population during at least one previous iteration of step (iii).

3. The method of claim 1 wherein step (v) is chosen when the distribution of calculated hydrodynamic radii has changed from at least one previous iteration of step (iii) so that there is a chosen increase in the relative number of particles falling within the said distinct sub-population when compared with the number of particles falling within the said distinct sub- population during at least one previous iteration of step (iii).

4. The method of any one of the preceding claims wherein the conditions of step (ii) include nucleati on-promoting conditions.

5. The method of any one of the preceding claims wherein the conditions of step (v) include growth-promoting conditions (ie metastable conditions).

6. The method of any one of the preceding claims wherein the conditions of step (ii) and step (v) differ in relation to one or more of concentration of the sample, temperature or the effectiveness of a barrier to diffusion from the sample.

7. The method of claim 6 wherein the sample is diluted (in relation to the macromolecule) between steps (ii) and (v).

8. The method of claim 6 wherein the concentration of the solution in an equilibration reservoir is changed, for example decreased, for example in relation to a crystallising agent, between steps (ii) and (v).

9. The method of claim 6 wherein the thickness or composition of an oil layer over the sample or over an equilibration reservoir is changed between steps (ii) and (v).

10. The method of any one of claims 2 to 9 wherein the chosen increase in the relative number of particles falling within the said distinct sub- population is at least 1.5-fold, preferably, 2, 3, 5, 10, 15, 20, 30, 40, 50, 80, 100, 200, 500 or 1000-fold, relative to the number falling within the said subpopulation when the sample has calmed following set-up.

11. The method of any one of claims 2 to 10 wherein the "aggregate" modal hydrodynamic radius is at least 2, preferably, 5, 10, 20, 50 or 100 times the "monomer" modal hydrodynamic radius.

12. The method of any one of the preceding claims wherein the DLS measurements are made using of light of at least two wavelengths.

13. The method of any one of the preceding claims wherein the sample assessed by DLS has a volume of less than 0.5, 1, 2, 3, 4, 5, 6, 7, 10, 15, 20 or 30 μl.

14. The method of any one of the preceding claims wherein the conditions are changed using an automated liquid dispensing system.

15. Use of a Dynamic Light Scattering measuring apparatus in the method of any one of the preceding claims.

16. The use of claim 15 wherein the DLS measuring apparatus is a DIMINIGON-A™ apparatus.

17. The use of claim 15 wherein the DLS measuring apparatus comprises two or more lasers suitable for producing light of different wavelengths.

18. The use of an automated liquid dispensing system in a method according to any one of claims 1 to 14.

19. The method or use according to any one of the preceding claims wherein the macromolecule is a biological macromolecule.

20. The method or use according to claim 19 wherein the biological macromolecule is a polypeptide.

21. A method for producing macromolecular crystals comprising the steps (a)dispensing a sample of a solution of macromolecule and crystallising agent (b)incubating the sample for a chosen period of time under chosen conditions and/or (c)incubating the sample within a chosen volume element under chosen conditions (d)assessing the sample using Dynamic Light Scattering (by providing spatial and time resolution during dynamic experiments) (e)on the basis of the assessment, either reiterating steps (b) (c) and (d) or incubating the sample for a second chosen period of time or at a chosen volume element under conditions which differ from the conditions of step (b).