2.1. Offline facility (Cryobench)

The Cryobench is currently located next to the MAD beamline ID29 (de Sanctis et al., 2012; Fig. 1a). The microspectrophoto­meter setup comprises a motorized single-axis goniometer (one rotation, one translation along the same horizontal axis) holding a Huber goniometer head (Rimsting, Germany) which allows two additional degrees of translation, on which a sample, either a crystal or nanolitres of solution on a SPINE standard sample holder (Cipriani et al., 2006), is mounted. Three objectives, mounted in a 90° geometry relative to each other (and a vertical camera) are 5 cm (8 cm) away from the position of the sample (Figs. 1b and 1c). The temperature and humidity of the sample position can be controlled with a liquid-nitrogen open-flow cooler (Cryostream Series 700, Oxford Cryosystems, Oxford, England) for temperatures ranging from 100 to 270 K or with the EMBL/ESRF dehumidifier HC1 (Sanchez-Weatherby et al., 2009) for room-temperature measurements.

Several technical choices have been made in order to maximize the compatibility between the environments of the Cryobench and the ESRF macromolecular crystallography beamlines. In particular, the analogue camera previously used for sample visualization has been replaced by the same digital Prosilica camera as used on the beamlines. New control modules for video acquisition, goniometer operation, laser triggering and cryosystem temperature control have been written with the BLISS FrameWork, an ESRF software-development platform, and run on Linux OS. In this way, experiment-control modules can be taken from, or ported to, the beamline-control software MxCuBE (Gabadinho et al., 2010). All Cryobench experiment-control modules have been grouped together in a main control software called SpeCuBE (Fig. 2).

The UV–vis absorption mode of operation consists of using the two objectives at 180° from each other (Fig. 1d). White light from a balanced deuterium–halogen lamp (Mikropack DH2000-BAL, Ocean Optics) is connected to one objective via an optical fibre, while a fixed-grating spectrophotometer with a CCD detector (model HR2000+ or QE65Pro, Ocean Optics) is connected to the other objective. Alignment of the spectrophotometer consists of a series of micrometric translations parallel or perpendicular to the light path that align the focal volumes of the objectives and maximize light transmission through a 30–100 µm wide pinhole (Melles-Griot) placed at the sample position. The diameter of the focal volume is one quarter of the diameter of the optical fibre (available from 50 to 1000 µm) connected to the objective.

Absorption and fluorescence emission spectra are measured with the commercial SpectraSuite software (Ocean Optics) running on Linux. In order to measure an absorption spectrum A(λ), three different spectra need to be recorded: the dark reference I_dark(λ) (the spectrum in the absence of light), the light reference I_ref(λ) (the lamp spectrum in the absence of a sample) and the sample spectrum I_sample(λ) (the light transmitted through the sample):

Fluorescence emission spectra (Bourgeois et al., 2002) and fluorescence lifetime histograms (Royant et al., 2007) are measured using laser lines as excitation light with two objectives at 90° from each other in order to minimize excitation laser scattering to the CCD detector (Fig. 1e). Because fluorophores are usually photoactivatable or photobleachable, laser irradiation needs to be kept to a minimum by synchronizing laser triggering pulses (typically of milliseconds to hundreds of milliseconds in duration) with the recording of spectra via TTL pulses generated by an ESRF-developed board (OPIOM type). Laser lines can be chosen from more than ten typical wavelengths ranging from 266 nm (deep UV) to 671 nm (far red), with a maximum power of 10–100 mW at the sample position.

Raman spectra are measured in backscattering mode using a Renishaw InVia spectrometer (Carpentier et al., 2007), the probe of which has been modified by the addition of a 45° mirror inserted before the microscope objective (Fig. 1f). Spectra are recorded with the commercial software Wire3 running on a Windows PC. Raman spectra exhibit many more peaks than UV–vis absorption or fluorescence emission spectra; hence, their interpretation relies on a comparison with reference spectra taken from other protein crystals (Fig. 3).

2.2. Online UV–vis absorption microspectrophotometer (`online microspec')

The first apparatus used to measure UV–vis absorption spectra directly on macromolecular crystallography beamlines was designed in the context of a collaborative effort between the University of Oxford, the EMBL Grenoble and the ESRF (Nanao & Ravelli, 2006). A more compact version was then developed by the EMBL and commercialized by MAATEL (Voreppe, France; McGeehan et al., 2009). This apparatus was designed to monitor the buildup and decay of various X-ray-induced phenomena such as solvated electrons or disulfide-bond radicals with spectroscopic signatures in the UV–visible range (Weik et al., 2002; McGeehan et al., 2009). It can also be used to identify X-ray-excited optical luminescence (XEOL) signals from protein crystals, which could serve as another tool to study specific radiation damage (Owen et al., 2012). The online microspec has since been used to monitor the X-ray-induced spectroscopic changes of many coloured proteins and in the design of optimal strategies for diffraction data collection (i.e. the collection of composite data sets from one or several crystals; Adam et al., 2004; Berglund et al., 2002; Bourgeois et al., 2009; Pearson et al., 2004) in order to elucidate crystal structures unaffected by radiation damage; for instance, metalloproteins containing metal centres in the correct oxidation state.

The fluorescence option was implemented by adapting the Cryobench fluorescence setup (Royant et al., 2007). A National Instruments board was added to a laptop station which can generate series of TTL signals (at 0.1–10 Hz repetition rates) that are used to synchronize laser irradiation and spectrophotometer recording via a LabVIEW (National Instruments) application running on a Windows laptop. Because there is no third objective in this setup, either a split fibre has to be connected to one objective (one fibre for excitation light; the other one being connected to the spectrophotometer) or the laser and the spectrophotometer are connected to the two opposing objectives. In the latter case, a filter rejecting the laser line is required to prevent saturation of and damage to the CCD detector.

2.3. Online Raman spectrometer

Online Raman spectroscopy has been made possible on beamlines by successive rounds of improvement. The original experiments were performed by connecting the Raman spectrometer, mounted on a rolling table, to the beamline experimental hutch using a 20 m fibre. The experimental hutch ID29-EH1 is now permanently connected to the Raman spectrometer of the Cryobench via a set of 50 m optical fibres (red trace in Fig. 1a), and a Raman head support has been designed to fit the MD2 diffractometer (Maatel, Voreppe, France). The support itself will be described elsewhere. In the near future, the Cryobench will be similarly connected to an experimental hutch of the new MASSIF suite of endstations situated on the straight section ID30 next to ID29 (Theveneau et al., 2013). The use of Raman spectroscopy in conjunction with X-ray crystallography has been named `Raman-assisted X-ray crystallography' (RaX) and its use has already been extensively reviewed (Bourgeois et al., 2009; McGeehan et al., 2011).

3.1. Alignment issues

The most important issue to consider when aligning a sample is the fact that the optical path has to cross various materials with distinct refraction indices: air (n = 1.00), mother liquor [1.33 (pure water) ≤ n ≤ 1.5 for salts and PEGs] and crystal (n ≃ 1.34 at high protein concentrations; Cole et al., 1995). The alignment of a two-objective microspec consists of matching, in three dimensions, the two focal volumes in the absence of a sample (i.e. when the whole optical path is in air). If the sample is large or if the mother liquor has not been removed and forms a bulk around the sample, the focal volumes will be significantly displaced. Fig. 4 shows a simple situation in which the beam path is perpendicular to the surface of a plate-shaped crystal. The two objectives have been aligned in air with their focal volumes matched at the position of the black dot. Insertion of the protein crystal with a higher refractive index at the sample position shifts both focal volumes in opposite directions. As a consequence of the mismatch between the focal volumes, the transmitted light is not optimally focused onto the entrance of the optical fibre leading to the spectrophotometer, thus decreasing the signal-to-noise ratio of the measurement. In other words, the reference light signal I_ref (equation 1) recorded in the absence of a sample differs from the hypothetical I_ref that would be recorded in the presence of a noncoloured sample of the same shape. However, the resulting absorption spectrum only differs from the hypothetical spectrum by an offset of the baseline, which can be easily corrected for. This situation is aggravated by increases in the size of the sample and the tilt of its position versus the light axis. The effect is even more severe for nonplanar crystalline samples, for example a crystal in a frozen drop of liquid, which results in a lens effect. In this case not only are the focal volumes displaced, but the direction of the transmitted light is also altered. A partial remedy for this issue consists of translating the second objective with the sample in place while monitoring the increase in the collected light until a maximum is achieved.

3.2. Self-absorption effect

The effect of strongly absorbing crystals on the shape of fluorescence emission spectra has been reported for light-harvesting complex II (LHCII; Barros et al., 2009). LHCII is a photosynthetic protein that contains many different coloured cofactors (carotenoids and chlorophylls) and is one of the proteins with the most coloured cofactors per amino-acid residue. Consequently, even very thin LHCII crystals (10 µm) have an OD above 2.5 and act as bandpass filters in the blue and red regions of the spectra.

This so-called `self-absorption' (or `inner filtering') is modulated by the geometry of sample illumination. If the excitation and emission objectives are focused on the top surface of a crystal of the fluorescent protein Cerulean (Lelimousin et al., 2009; Fig. 5a), one is able to measure the genuine emission spectrum of the protein (the red trace in Fig. 5c). However, if one focuses on the bottom surface, such that both the excitation and the emission light travels through the crystal (Fig. 5b), the significant OD of the crystal (see the absorption spectrum depicted in black in Fig. 5c) will absorb most of the low-wavelength emitted photons, resulting in an apparent red shift of the left emission peak (orange, green and blue traces in Fig. 5c).

3.3. Negative absorption owing to fluorescence

During the recording of UV–vis absorption spectra from fluorescent samples, the incident white light also produces some detectable fluorescence. Depending on the geometry of the two focal volumes (see above), a significant amount of this unwanted fluorescence will be recorded by the collecting objective, which leads to a dip in the resulting absorption spectrum (black trace in Fig. 5d) in the region of maximum fluorescence in the emission spectrum (the red trace in Fig. 5d), formally constituting a negative absorption signal.

4.1. Enzymes with coloured cofactors

Many enzymes rely on cofactors that absorb in the near-UV/blue part of the light spectrum. Usually, the light-absorption properties of the cofactor are not instrumental to the catalytic mechanism. Measurement of the fluorescence emission spectrum from crystals of NADH-containing malate dehydrogenase was one of the first experiments carried out at the Cryobench, identifying the redox state of NADH in the crystal (Bourgeois et al., 2002). Crystals of a Baeyer–Villiger monooxygenase, containing NADPH and FAD as the electron donor and oxygen acceptor, respectively, were studied both at the Cryobench and with the online microspec in order to monitor the reversible oxidation of the flavin by air (at room temperature) (Figs. 6a and 7a) or reduction by either dithionite or X-rays (Orru et al., 2011). Finally, a vitamin B₆-dependent phosphoserine aminotransferase has been shown using online UV–vis absorption spectroscopy to be sensitive to X-rays at the location of its lysine–pyridoxal-5′-phosphate (lysine–PLP) Schiff-base linkage (Dubnovitsky et al., 2005).

The only two known enzymes for which light-induced activation of cofactors is necessary for the catalysed reaction have both been studied at the Cryobench. DNA photolyase uses two cofactors, the flavin FAD and a light-harvesting chromophore which is either 8-hydroxy-5-deazaflavin or 5,10-methenyltetrahydrofolate, to repair DNA lesions caused by UV radiation in bacteria. Online UV–vis absorption spectroscopy was used to monitor light-induced or X-ray-induced reduction of the cofactors (Kiontke et al., 2011; Kort, Komori et al., 2004; Moldt et al., 2009). Light-dependent protochlorophyllide oxidoreductase (POR) catalyses the penultimate step of chlorophyll biosynthesis. Temperature-dependent fluorescence spectroscopy was used on solution samples to study the influence of solvent dynamics on the formation of the first two reaction-intermediate states (Durin et al., 2009).

4.2. Metalloproteins

Haem proteins, with an Fe atom bound to a porphyrin ring, are gas or electron transporters or enzymes. Early offline studies attempted to establish redox states before and after X-ray data collection (Dias et al., 2004; Hersleth et al., 2008; Marcaida et al., 2006; Williams et al., 2006) and showed their high X-ray susceptibility, the explanation for which must reside in their redox activity (Figs. 6b and 7b). Later studies used the online microspec to monitor the rate of haem photoreduction, a phenomenon which was eventually used to mimic the physiological redox reaction (Gumiero et al., 2011; Hersleth & Andersson, 2011; Macedo et al., 2009). Besides these more complex experiments, offline studies were used to characterize the functional state of haems in protein crystals (Rajagopal et al., 2011; Wahlgren et al., 2012).

Some nonhaem iron enzymes have their catalytic iron coordinated by at least four amino-acid residues of the protein. The superoxide reductase (SOR) from the microaerophilic bacterium Desulfoarculus baarsii falls into this category. SOR acts as a defence against oxidative stress and presents the advantage over superoxide dismutase of not releasing dioxygen upon superoxide reduction. A composite data-collection strategy (Berglund et al., 2002; Pearson et al., 2004) taking advantage of online UV–vis microspectrophotometry allowed comparison of its reduced and oxidized states (Adam et al., 2004). Subsequently, in order to prove the presence in the crystals of a peroxo adduct on the iron with an end-on geometry, Raman spectroscopy was used directly on crystals soaked with either H₂¹⁶O₂ or H₂¹⁸O₂ to show the presence of an Fe—O covalent bond (Katona et al., 2007; Figs. 6c and 7c). Other nonhaem iron proteins contain geometric arrangements of Fe and S atoms: iron–sulfur cluster proteins. UV–vis absorption spectroscopy has been used to investigate the redox state of Fe–S clusters (Aragão et al., 2003; Sainz et al., 2006) or of catalytic Fe atoms (Nicolet et al., 2001; Figs. 6d and 7d).

The X-ray-induced photoreduction of the mononuclear copper electron-transfer protein azurin and of multi-copper oxidases (MCOs) has been investigated using online microspectrophotometry, either directly in crystals (Ferraroni et al., 2012; Macedo et al., 2009; Figs. 6e and 7e) or indirectly by monitoring the buildup of solvated electrons in protein solutions (De la Mora et al., 2012). Another MCO had its spectroscopic signature in crystals compared with that in solution (Bento et al., 2010). Finally, crystals of CnrX, the metal-sensing domain of a membrane protein complex implicated in bacterial resistance to nickel and cobalt, was shown to bind one divalent cobalt cation in a high-spin state by UV–vis absorption spectroscopy (Trepreau et al., 2011).

4.3. Photoactive proteins

Photoactive proteins contain chromophores, the light-induced isomerization of which triggers structural rearrangements of the protein leading to the fulfilment of its function.

One such family of proteins studied at the Cryobench is that of the retinal proteins, which exhibit very distinct functions in various organisms. One early intermediate state of bacterio­rhodopsin, a proton pump from the halophilic archaeon Halobacterium salinarum, has been characterized using either excitation by a green (Royant et al., 2000) or a red laser (Edman et al., 2004) at low temperature. A similar study using a blue-cyan laser was performed on sensory rhodopsin II, an archaeal photoreceptor implicated in negative phototaxis (Edman et al., 2002). Various intermediate states in the photocycle of bovine rhodopsin, the protein responsible for vision, were similarly characterized upon excitation at low temperature (Okada et al., 2002) and also after reconstitution with the isomerized chromophore at room temperature (Choe et al., 2011) (Figs. 6f and 7f).

Photoactive yellow protein (PYP), thought to be implicated in bacterial phototaxis, has been the topic of compelling structural studies using Laue crystallography of intermediate states on the picosecond-to-second timescale (Jung et al., 2013; Ihee et al., 2005). PYP was studied with low-temperature monochromatic crystallography, using UV–vis light absorption as a complementary method to probe the altered kinetics of the photocycle in the crystalline state compared with those observed in solution (Kort, Hellingwerf et al., 2004; Kort et al., 2003).

Phytochromes are red/far-red photochromic biliprotein photoreceptors which regulate many light-affected processes in plants (notably the germination of seeds and flowering) and certain bacteria and fungi. Isomerization of their bilin chromophore leads to a conformational change which triggers downstream signal transduction. UV–vis absorption and Raman spectroscopy were used to compare the crystalline protein with its solution state and showed that in the former photoconversion to the far-red-absorbing state was significantly hampered but was not completely inhibited (Anders et al., 2013; Essen et al., 2008; Mailliet et al., 2009, 2011). This suggests that phytochromes may be functional in the crystalline state but that crystal contacts prevent the large-scale domain movements necessary for full functionality.

Photosynthetic proteins are a special case of photoactive proteins where chromophore isomerization is not required for function. Their functional role is architectural: they hold in place a multitude of chromophores covering the whole visible-light spectrum in order to harvest sunlight optimally and transfer energy quanta to the active site of photosystems. Crystals of such a protein, light-harvesting complex II from pea, were thoroughly investigated by UV–vis absorption and emission fluorescence spectroscopies and fluorescent lifetime measurements to show that the protein was in the active, light-transmitting state (Barros et al., 2009; Figs. 6g and 7g). α-Phycoerythrocyanin (PEC) is a bilin photoreceptor that is part of the light-harvesting complex, the phycobilisome, of certain cyanobacteria. Phycobilisomes absorb light and transfer its energy to the photosynthetic reaction centres, acting as a light-harvesting antenna. Phycoviolobilin, the chromophore of PEC, isomerizes between two stable configurations upon light irradiation. UV–vis absorption was used to monitor the efficiency of photoconversion, and the structures of the two isomeric forms were eventually obtained to provide the structural basis of PEC photochemistry (Schmidt et al., 2007).

4.4. Fluorescent proteins

While it was rapidly realised that the chromophores of fluorescent proteins, as well as the surrounding residues, were particularly sensitive to radiation damage by X-rays (Adam et al., 2009; Royant & Noirclerc-Savoye, 2011; Figs. 6h and 7h) and thus that special care is required during data collection and structural analysis, the Cryobench has become an essential instrument to study the structure and function of proteins homologous to green fluorescent protein (GFP; Tsien, 1998). The most common application of the Cryobench is to compare the spectroscopic signatures of different proteins in crystals with those in solution. While these are generally very similar (Carpentier et al., 2009; Goedhart et al., 2012; Lelimousin et al., 2009; Violot et al., 2009; von Stetten, Noirclerc-Savoye et al., 2012), fluorescence lifetimes are shorter in crystals, which is related to the higher refractive index of the crystal mother liquor (Royant et al., 2007; Suhling et al., 2002), and the fact that diffusion of small molecules close to the chromophore can induce subtle rearrangements of side chains and subsequent displacement of absorption and emission peaks (von Stetten, Batot et al., 2012). Moreover, the Cryobench is adapted to the elucidation of various reversible or irreversible mechanisms of fluorescent proteins: photoactivation, photobleaching and photoswitching. These studies take advantage of the various available laser lines available at the Cryobench as well as the possibility of adjusting the temperature at which measurements are made (Adam et al., 2008; de Rosny & Carpentier, 2012; Duan et al., 2013; Faro et al., 2010, 2011).

4.5. Noncoloured proteins or DNA

All of the examples given above highlight studies of proteins that are coloured. Optical spectroscopy can, in some cases, also be applied to noncoloured samples which contain specific chemical bonds. Disulfide bonds bridging two cysteines result in a distinct Raman stretching mode because of the greater mass of S atoms and their relative scarcity within proteins. Raman spectroscopy has thus been used to study the presence, or absence, of disulfide bonds in the active sites of various mutants of the thiol–disulfide oxidoreductase DsbA (Lafaye et al., 2009; Figs. 6i and 7i). Disulfide bonds are known for their X-ray sensitivity (Weik et al., 2000). UV–vis absorption can be used to identify the presence of radiation damage by monitoring the formation of a disulfide radical, an intermediate state in bond reduction, in various protein crystals (McGeehan et al., 2009; Murray & Garman, 2002; Weik et al., 2002) and online Raman spectroscopy can be used to monitor the kinetics of X-ray-induced disulfide-bond photoreduction (Carpentier et al., 2010; McGeehan et al., 2011). Similarly, X-ray-induced debromination of brominated DNA crystals can also be monitored using Raman spectroscopy (McGeehan et al., 2007).