3.1. Automation

The automation revolution created for MX by institutes like the ESRF in partnership with the EMBL-Grenoble, by Stanford Synchrotron Radiation Laboratory, and by the more recent synchrotrons such as Diamond (Didcot, UK), SOLEIL (Saclay, France) and the SLS (Swiss Light Source, Switzerland) is a key point in the continued rise of synchrotron crystallography. These robotic developments and smart software pipelines, with their database and data-delivery frameworks, are having a major impact. Reliable sample changers and automated intelligent software are routine at every synchrotron MX beamline. In Europe, the SPINE hardware standard gives consistent sample-mounting pins and bases. A new generation of pins is being developed under the banner of another European project, BioStructX, for the transport of a larger number of samples per cryo-dewar. This is needed as sample cycle times are now down to as little as a few minutes, meaning efficient users, locally or in remote-access mode, can process hundreds of samples per shift. This represents many cryo-dewars of frozen crystals needing to be transported. The higher number density is required to allow the new generation of mail-in dedicated automated beamlines like ESRF's MASSIF (massively automated sample selection integrated facility) to have tenable logistics.

Within Europe, the ISPYB and EDNA software permit sample, data and results tracking, and protected data delivery to users (Delagenière et al., 2011). Each sample is characterized and a data-collection strategy proposed to the user for approval. The system then supervises the data collection and launches automated data processing as the data arrive.

In addition to robotics for cryo-cooled samples, room-temperature screening is possible at a number of beamlines across synchrotron facilities, for example the French FIP beamline at ESRF (la Maire et al., 2011) and the X06DA station at the SLS (Bingel-Erlenmeyer et al., 2011) are able to use their robotics to manipulate crystallization plates and allow crystals to be tested for diffraction quality in situ without the need for harvesting a sample. This permits more efficient feedback on crystallization condition refinement and, sometimes, full data collection. Indeed, Diamond is constructing a facility, VMXi, dedicated to such experiments for collection from in situ crystallization experiments. Automation is also extending backwards from beamlines into crystal harvesting (`fishing') – a considerable bottleneck and still largely done by hand. This may be changing with new laser-cutting based harvesting. The CrystalDirect system developed at EMBL-Grenoble (Cipriani et al., 2012) uses a smooth roboticized procedure with laser photo-ablation to excise crystals on thin films of polyimide, which are attached by the robot to the standard SPINE pin and then cryocooled ready for data collection. Together with the automation of crystallization, and of diffraction experiments on beamlines, systems like CrystalDirect will open the route to a fully hands-free procedure.

Automation allows users to exploit SR MX facilities efficiently. In the last few years this has grown to be a popular mode of access to SR MX beamlines, and it is particularly attractive (cost effective) for industry. As an illustration, the approach taken by SERCAT (Rose et al., 2014), the South East Region (of the USA) Collaborative Access Team of the APS, is to have `outstanding staff; a stable beamline; a good end-station goniometer; a reliable automounter; fast reliable detectors; a single intuitive interface; a secure and robust web client and provide online training resources'. Across many beamlines, remote access is now the way to use the facility, minimizing travel time and costs and the ecological footprint of jet travel by users. At the APS, over 95% of SERCAT's data are collected in this way (Rose et al., 2014). With modern detectors the time-limiting step has become crystal mounting and alignment.

This change to very high throughput has also ushered in a move from multi-shift visits (virtual or real) to the synchrotron to single-shift `pop-ins' where users, and again especially industrial users, have shorter individual visits to ensure a continual data-stream delivery into projects.

3.2. Ever smaller crystal volumes

The current and still developing frontier of synchrotron microcrystallography has a long and distinguished history. An early question was whether smaller samples studied with higher X-ray intensities were feasible. The widespread implementation of mini- and microbeams and high-precision microdiffractometers to measure data from such crystals has had a wide impact in structural biology [and has, for example, been highlighted by Nobel Prize winner Brian Kobilka in Nature recently (Azouz, 2014)].

Ultra-rapid sampling `cartography' (Bowler et al., 2010) of micro-volumes in protein crystal samples to locate the best hot-spot for diffraction is becoming routine and is a key functionality within the ESRF's new multi-station automated MASSIF beamline. The X-ray source technologies of both XFELs (see below) and upgraded synchrotron X-ray storage rings are reaching a similar protein crystal sample size range of microns and sub-microns. For a recent overview of SR MX microcrystal diffraction, see Evans et al. (2011). Could electrons be used instead to measure such diffraction data on MX microcrystals and solve these structures? There are renewed developments in this area now, succeeding in electron crystal structure analysis of proteins by molecular replacement, thus taking advantage of the greater scattering efficiency of electrons by matter compared with X-rays (see e.g. Nannenga et al., 2014).

The simple question of how to manipulate ever smaller single-crystal samples leads to quite different technical approaches. One method developed for application at XFELs uses serial delivery of tiny crystals via a jet injector system at room temperature and so does away with the classical goniostat of the crystallographer (Chapman et al., 2011). This serial method leads to a myriad of still diffraction patterns, but with the advent of XFELs new software has been developed (e.g. CRYSTFEL; White et al., 2012) to handle these data efficiently and produce complete diffraction data sets. The serial delivery method was also recently proven for the synchrotron using lysozyme as the model system, with over 40 000 crystals exposed and merged to a data set of 2.1 Å resolution (Stellato et al., 2014). Taking this concept further, recent work has used the ESRF's ID13 micro/nanofocus diffraction beamline with a sample injector system, again originally developed for use at XFELs but specifically designed for lipidic cubic phase crystals. Another method aims to use many crystals cryo-cooled in one sample support and then raster-scan to locate crystals, collect diffraction data and combine the partial diffraction data sets to form the full data set. A proof-of-concept has been very elegantly performed at the PETRA III facility, where in vivo grown microcrystals were exposed on the P14 microfocus beamline, with data from 80 crystals being combined into a 3.0 Å data set (Gati et al., 2014). All of these developments look ahead to the advent of very low-emittance storage rings with management of the impact of radiation damage.

Fig. 1 shows some key developmental steps in producing ever smaller X-ray focal spot sizes and their optics, which are highly relevant to the above discussion. Table 1 shows the level of X-ray intensities at the sample that are feasible at current national SR facilities for MX (the Canadian Light Source is highlighted as an example).

Figure 1 The ESRF beamline ID13 has been and remains at the cutting edge of how small an X-ray microfocus beam can be and undertakes a wide variety of microdiffraction studies including MX. Shown here are the different optical means of providing different sized very small focal spots. (Image reproduced from https://www.esrf.eu/files/live/sites/www/files/UsersAndScience/Experiments/SoftMatter/ID13/poster/esrf_um_2005.jpg with the permission of Dr Christian Riekel of the ESRF.)

As samples become smaller, they yield fewer diffraction data before radiation damage renders the sample of no use. Which partial data sets can be realistically combined depends on how similar the samples are. Hierarchical cluster analysis, a method introduced by Wayne Hendrickson and co-workers (Liu et al., 2011), allows this to be checked and is also now applied at ESRF, as described by Giordano et al. (2012). Much more accurate anomalous signal measurement and greater success in substructure determination can be obtained by merging data from multiple crystals preselected according to the results of cluster analysis. Advances in phasing methods in MX including harnessing weak anomalous signals were discussed at the January 2015 CCP4 Study Weekend (https://www.ccp4.ac.uk/events/CCP4_2015/programme.html ) and the proceedings will be published in Acta Crystallographica Section D.

3.3. Non-conventional X-ray wavelengths

Most MX beamlines that exist or are under design and construction continue to be optimized for the 2 to 0.8 Å core wavelength range, despite the much wider range of photon wavelengths of around ∼5 to 0.2 Å which has been discussed for use in MX at synchrotron facilities. The core wavelength range covers most of the anomalous absorption edges used, particularly selenium. However, SR MX users are steadily adopting an ever widening practical range of X-ray wavelengths.

Longer wavelengths up to ∼5 Å are under active development for MX at both Diamond and NSLS II. The long-wavelength (1.5–4 Å) Diamond I23 project will be just such a user facility to enhance and optimize the anomalous signals from low atomic number elements. These include sulfur in proteins and/or phosphorus in RNA/DNA crystals, which are needed where protein labelling to introduce anomalous scatterers, such as that involving selenomethionine via molecular biology gene expression or heavy-atom chemical derivatization, is not feasible. In addition, the wavelength range of I23 will provide access to the M edges of elements, of uranium for example, with larger anomalous signals (Fig. 2). The idea here is to use the M edge f′′ maximum values of up to 100 electrons when `white lines' are present (Liu et al., 2001). The anomalous differences from such an f′′ will thus be attractive for measurements of, say, a crystalline large molecular machine complex for MAD structure determination. For such high-angle diffraction data experiments, a large semi-cylindrical vacuum-compatible pixel detector has been designed specifically by DECTRIS (Baden, Switzerland) to capture the diffraction data whilst operating in vacuo. To obtain high-quality data, this will be coupled with X-ray computed tomography to obtain the crystal sample shape and volume for an analytical sample absorption correction.

Figure 2 The resonant scattering coefficients f′ and f′′ for the M edges of uranium. Reproduced from the website set up by Dr Ethan Merritt, who is gratefully acknowledged.

At very short (∼0.5 Å) and ultra-short (∼0.3 Å) wavelengths, high storage-ring energies yield a copious flux output. To date, these have been used for applications such as high-pressure MX, where the restricted aperture of the diamond anvil cell is less of a limitation with shorter wavelengths (Fourme et al., 2011), and high-energy MAD such as at the holmium K edge (see e.g. Jakoncic et al., 2006). Another application of high photon energies involves minimizing radiation damage. Nave & Hill (2005) have cogently argued that crystals smaller than 10 µm may encounter reduced radiation damage as the photoelectrons can escape from the crystal lattice when using short-wavelength X-rays. Building on this, Finfrock et al. (2013) presented evidence in favour of the use of a 1 µm-wide line focus even for data collection from crystals of around 100 µm, which could also help in the management of radiation damage.

3.4. Time-resolved studies

Time-resolved Laue protein crystallography at the ESRF has opened up a whole new field of sub-nanosecond crystal structure analyses. `Fast time-resolved' biomolecular science examples include carbon monoxy myoglobin and phospho yellow protein (PYP); for a review, see Ren et al. (1999). More complex, but by nature slower, cases include following the enzyme reaction of hydroxymethylbilane synthase in the crystal (Helliwell et al., 1998). There are only a limited number of such time-resolved studies in the literature, for which there are several reasons. Firstly, crystal lattice interactions can block the necessary structural changes for a given biochemical reaction to proceed. Secondly, crystal size determines the scattering strength of a sample and thereby the required exposure time, which clearly increases as a sample gets smaller. This can obviously be at odds with the intrinsic time resolution required to monitor a given molecular structural change. Different measuring protocols exist which try to surmount this challenge, such as the Hadamard measuring sequence (Yorke et al., 2014) or the simpler approach of crystal-to-crystal averaging at equivalent time slices (Helliwell et al., 1998). Meanwhile, XFELs now provide femtosecond duration pulses, typically 10 to ∼50 fs. Their use is attractive for the fastest time-resolved protein crystallography studies. It has been proposed that even single molecules could be studied (Neutze et al., 2000), which would free us from the crystal lattice restrictions referred to above. A recent comprehensive compilation of XFEL science applied to structural biology, including various time-resolved structural studies, is given by Spence & Chapman (2014).

3.5. Non-traditional and other applications in SR MX

There are several topics which, in the last few years, have continued to attract attention and development. Room-temperature crystallography is a growing biological crystallography research activity and is a reminder that cryo-derived MX structures do show structural differences. Structural changes occur mostly in the dynamics, as shown by the increased proportion of split-occupancy side-chains at cryo-temperatures. Radiation damage at room temperature used to be the norm for MX in the pre-ribosome crystallography days and damage was mitigated by modest cooling to, typically, 4°C. As mentioned above, serial femtosecond crystallography is generally undertaken at room temperature in any case and `the diffraction outruns the damage'. Neutron MX is damage-free and room temperature is routinely employed. However, cryotechniques have other advantages than simply radiation-damage mitigation, namely, under well chosen conditions, improved order and freeze trapping of structural intermediates, provided they are longer lived than the freezing time.

An MX diffraction pattern can have many features which we do not usually seek to explain, namely diffuse scattering, excluding the obvious solvent ring. That this might offer specific information on protein structural dynamics, if it can be teased out from the lattice dynamics, is a long-standing topic. Recently, a short summary of a conference was published (Wall, Adams et al., 2014), and the book by Peter Moore (Moore, 2012) nicely summarizes the mathematics of structural and lattice dynamics. Both of these works indicate a renewed determination to use this diffraction information to provide a more complete model and interpretation. In physical crystallography there have been extensive developments, and the so-called full profile analysis is a promising approach for biological crystallography too. The retention of raw diffraction data could provide helpful and much more extensive data set case studies. The dissection of the respective dynamics components, mentioned above, would benefit from the growing trend of measuring MX data at both cryogenic and room temperatures. Interpretation of the diffuse scattering can be achieved via molecular dynamics simulations of protein vibrations, which now extend over time periods as long as 1 µs (Wall, van Benschoten et al., 2014).

The technique of SFX has focused attention on whether micro- and nanocrystals are better quality (typically, a lower mosaicity is referred to) than `routine-sized' crystals (see, for example, the volume edited by Spence & Chapman, 2014). The techniques described in detail in the book by Chayen et al. (2011) can be applied to the systematic evaluation of crystal perfection as a function of sample size. In chemical microcrystallography, Andrews et al. (1988) stated that, where crystals have a high mosaicity, they will not grow larger. The corollary of this is that if microcrystals have a low mosaicity then they can grow larger. Certainly, at ICCBM15 (15th International Conference on the Crystallization of Biological Macromolecules, 17–20 September 2014, Hamburg, Germany) there was a very healthy interest in growing large enough crystals for neutron MX where research into the structure and function of a molecule warrants it. It is therefore very important that we do not give up on knowing how to grow larger crystals, nor on acquiring further knowledge of growing them.

4.1. SR MX leads the way for commercial industry access

Industrial use of the synchrotron research infrastructure is a core mission of most such facilities. At the SRS Daresbury, commercial access to MX was the largest share of the Daresbury Analytical Research Technical Service (DARTS) (Maclean et al., 2006). Nowadays, most synchrotrons have an industry or business development office, managing and developing links with industry and creating economic value and impact from industrial access [see, for example, Cutler (2014), Shotton et al. (2014) and Mitchell et al. (2011)]. Drug discovery is still a significant income generator for all facilities, yielding millions of euros per year in industrially derived income flowing into the facilities.

Industry requirements have driven aspects of SR MX automation and, in particular, the development of metadata models and database systems to track data collections and processed results and to make all the information available via a secure web link for remote access and download of results. Malbet-Monaco et al. (2013) explored the impact of this `reverse spin-off' benefit back to the facilities. Industrial needs have also driven the development of `mail-in' crystallography, with the services described in Nature in 2003 (Schmidt, 2003) for the NSLS and the ESRF being prominent examples.

Spin-off data-collection service companies have been created, such as Expose GmbH from the SLS which provides access to SR MX facilities. Other businesses, all using synchrotron access, deliver rapid structure solution and structural biology services, such as Saromics (Sweden) which delivers kinase structural data, Bio-Xtal (France), Shamrock Structures (US) and VivaBiotech (China), amongst many others. The wholesale uptake and acceptance of structural biology, and thus of SR MX, for drug discovery by the pharma­ceutical industry, and the maturation of automation, have led to such enterprises having tenable business cases. High-throughput SR MX has assisted in the viability of the fragment-based drug discovery industry (Badger, 2012), which relies on a steady stream of biophysical and structural data, in particular from X-ray crystallography [see Chilingaryan et al. (2012) for a recent review].

SR facilities are now seeking to develop in a similar manner to other industry sectors, with national governments and the EU seeing industrial activity as one of the metrics of a facility's success (whether synchrotrons, neutron sources, lasers or other research infrastructure).

The exploitation of SR MX facilities for economic value is achieved not only via industrial use and the occasional patent or university spin-off based on structural data, but also in the transfer of technology and knowledge to instrumentation suppliers who then sell the MX instrumentation worldwide. Examples of this include diffractometers, sample-changer robots and sample pins.

4.2. Public understanding of science, engineering and technology

SR facilities and their research user programmes attract considerable public and political interest (see e.g. https://www.stfc.ac.uk/3388.aspx ). Government funding agencies closely monitor the `high-impact' publications that research produces. MX is a principal contributor to demonstrating success by these metrics. A clearly helpful factor over the years has been the award of Nobel Prizes for structural biology for which the research required SR MX. Thus, there was the Nobel Prize to John Walker (shared) for F1ATPase (SRS); to Rod McKinnon for the potassium ion channel (CHESS, USA); to Roger Kornberg for RNA polymerase (SSRL); to Venki Ramakrishnan, Tom Steitz and Ada Yonath for ribosomal structure studies (featuring many SR facilities and especially involving the NSLS, APS and ESRF); and to Brian Kobilka (shared) for GPCRs (APS and ESRF). These and other crystallography-derived Nobel Prizes are described in the book by Olovsson et al. (2015), including the key research articles.

4.3. An evolving offer to SR MX users

The view of central SR facilities and users towards MX is evolving. For structural biology users, the role of the synchrotron now extends beyond `just' data provision via X-ray crystallography. Protein crystallographers have adopted bioSAXS (biological small-angle X-ray scattering) in recent years. An interesting initiative for a combined MX and bioSAXS beamline at the ALS in Berkeley is described by Classen et al. (2013). The rise of bioSAXS has included automation of the beamline hardware and software pipeline to allow efficient data validation and modelling.

Combined X-ray and neutron structural biology studies [see e.g. the book by Svergun et al. (2013)] are being facilitated at science campuses like that of ESRF and the ILL, with the neutron work making use of deuteration and contrast matching for the study of multi-protein or DNA/RNA/protein complexes. Central facility sites are also actively creating joint services for structural biology, combining X-rays and neutrons with other techniques such as NMR and electron microscopy (EM). At the Rutherford Appleton Laboratory (RAL), which hosts Diamond and ISIS, new state-of-the-art cryo-EM facilities will be treated like a beamline, with users able to request this technique in addition to SR MX. Indeed, most major research infrastructure sites in Europe have or are developing multidisciplinary centres, like the Partnership for Structural Biology (Grenoble), the first of its type, the Centre for Structural Systems Biology (Hamburg), and the Membrane Protein Laboratory and Oxford Protein Production Facility (RAL). In the USA, the APS is building its Advanced Protein Crystallization Facility, which will allow the production, characterization and crystallization of proteins. All these, often multi-facility/multi-laboratory, partnerships stretch facility impact beyond the supply of X-rays and/or neutrons to helping users to prepare proteins, perform quality control and crystallize them, refine conditions, and use complementary structural analysis techniques to provide an overall structural picture of challenging macromolecular systems. Crystallization and the use of SR are closely intertwined (Chayen et al., 2010).

Critical information complementary to SR MX is provided by neutron MX, which is becoming a routine technique for the community. This is largely thanks to the provision of deuterated macromolecules as part of neutron facility programmes, and enhanced neutron beamline performance to reduce data-collection times and enhance diffraction resolution limits. A major review and summary of the field of neutron MX has been published by Blakeley (2009). This includes showing how the limits of high molecular weight and smaller crystal sample, as well as the speed of measurement, have been significantly improved with the neutron Laue method. The increased background noise that inevitably comes from using the broader spectral bandpass of the Laue method has had only a marginal, if any, effect on the diffraction resolution limit achieved. The total elapsed time for taking a data set measurement has also improved significantly. A future development at the ILL is the CYCLOPS (cylindrical CCD Laue octagonal photo scintillator) single-crystal diffractometer. This has a set of area detectors and will further speed up single-crystal diffraction intensity measurements with neutrons. Recently, the ILL beamline D19, with its refurbished detector, has allowed very high-resolution neutron structures to be determined (Cuypers et al., 2013). There are also high-megawatt spallation neutron source MX instruments at the USA's SNS and Japan's JPARC. The steady growth of the field of neutron MX is described in the book by Niimura & Podjarny (2011).