Monitoring Preparation of Derivative Protein Crystals via Raman Microscopy

This on X-ray Crystallography is a compilation of current trends in the use of X-ray crystallography and related structural determination methods in various fields. The methods covered here include single crystal small-molecule X-ray crystallography, macromolecular (protein) single crystal X-ray crystallography, and scattering and spectroscopic complimentary methods. The fields range from simple organic compounds, metal complexes to proteins, and also cover the meta-analyses of the database for weak interactions.

plastic support above the surface of the reservoir (sitting drop). Batch crystallization is a method where the sample is mixed with precipitant and additives creating a homogenous crystallization medium. In the free interface diffusion the protein sample is stratified over the precipitant solution; over time the sample and precipitant diffuse into one another and crystallization may occur at the interface. In a microdialysis experiment the protein solution is equilibrated through a membrane against the precipitant solution over time in a stepwise manner. In some cases, to improve the X-ray diffraction properties the crystallization is performed in a gel medium. Agarose, agar and silica have been successfully used as gel materials to obtain protein crystals (Chayen, 1998;Vergara et al., 2003;Vergara et al., 2009). It is a frequent task to produce modified protein crystals in order to study the structural modifications undergoing a chemical treatment. These derivative crystals can be prepared via co-crystallization, or via soaking. Co-crystallization means that crystallization of the chemically modified biopolymers (protein and additive) is conducted from solution phase. Alternatively, protein crystals are first grown and then they are chemically modified via diffusion of the additives that are soaked into the solvent channels. Both these preparative procedures can be supported by the application of Raman monitoring, particularly by difference Raman spectra (in co-crystallization) or by a time-resolved Raman microscopy (for soaking procedure).

Principles of Raman spectroscopy 2.1 Raman effect
Raman spectroscopy is a vibrational spectroscopy based on the anelastic scattering of a monochromatic light (frequency ) due to the interaction with the sample. The polarizability tensor of the sample oscillating with a normal frequency 0 , that is associated to the omni present molecular vibrations, interacts with the electric field of the laser light. This produces an induced electric dipole oscillating (then irradiating) with three distinct frequency ( , + 0 and -0 ). The strong elastic scattering at frequency 0 is the Rayleigh scattering that usually is cut in a Raman experiment. The two minor inelastic components are Raman Stokes bands ( -0 ) and Raman antiStokes ( + 0 ). At room temperature only Raman stokes bands are intense enough to be recorded. The difference between the wavenumber of the incident light and the scattered Raman band is the Raman shift (the x-axis of any Raman spectrum). The yaxis is usually reported in arbitrary units, so that accurate quantitative analysis can be based only on ratio of two distinct bands or by using some internal standard. In the Stokes Raman region (at wavelength lower than the excitation line) also fluorescence is collected. Therefore the counts read in the y-axis is always the sum of Raman and fluorescence, that should be kept as low as possible.

Off-resonance and resonance Raman
We should point out that two major kinds of Raman spectra can be collected, depending on the laser line (off-resonant Raman and resonant Raman spectra). In off-resonance Raman spectra no relation exists between excitation line and electronic absorption condition. On the contrary, resonance Raman spectra are collected when a particular excitation wavelength is used, namely within one electronic absorption band of the sample. Resonance Raman spectra are much more intense (depending on the extinction coefficients of the electronic transition) and selective (only normal mode that couples with the vibronic transition can be www.intechopen.com enhanced) than off-resonance Raman spectra. A more extensive presentation of the quantum mechanics of the Raman effect is elsewhere reported (Long, 2002).

Handling of protein crystals for Raman microscopy experiments
Raman microscopy is ideal for the monitoring of bioprocesses as it is non-destructive, inexpensive, rapid and quantitative. Its confocal nature makes it possible to focus through transparent capillary or directly on crystals kept in their crystallization reactor allowing straight analysis on the sample. In co-crystallization experiments two independent Raman spectra are collected on native and derivative crystals, using the same mother liquor with the exception of the reactant. In this case, the comparative analysis of the two spectra can provide differences in a) number, b) position and c) intensity of protein Raman bands. Manipulation of crystals for Raman measurements requires standard handling.

Crystal sampling
Confocal apparatus allows to get rid of any influence of the cover slip (as in vapour diffusion) or container wells (as in FID or batch crystallization supports). Measurement into drops (both hanging and sitting drop) is particularly feasible. In order to avoid significant scattering from mother liquor, thus reduction in spectral quality, it is better to keep the drop as small as possible, and to use a minimal depth into the crystal (particularly for resonance Raman spectra). Water is a weak Raman scatterer, with a small contribution at 1640 cm −1 . Precipitating agents, especially when at high content (PEG, MPD, alcohol) may interfere or not, depending on the spectral region of interest. Agarose and silica gel matrix do not interfere significantly, so crystals grown in gel medium can be used as well. Despite we deal with solid state we will not consider frequencies of the lattice vibrational modes, that for protein crystals are very low (below the Rayleigh cut). For lysozyme crystals it is 25 cm -1 . We will focus on two distinct experimental setups, namely for in-situ and ex-situ Raman spectroscopy, referred to simultaneous or not to X-ray diffraction experiments (also reported as on-line and off-line (McGeehan et al, 2011)). Below we present separately these two setups and applications.

Ex situ Raman experiments
Most of the analysis to monitor chemical modification of protein crystals can be performed by ex situ Raman experiments, which involve distinct acquisition of Raman and X-ray diffraction data. The experiments are, indeed, carried out on a Raman microscope that is physically separated from the X-ray diffractometer. This kind of experiments can be easily performed on a commercial or home-built Raman microscope, and it aims to the definition of the experimental conditions (eg soaking time and reactant/protein molar ratio) for the preparation of derivative protein crystals. Raman spectra on very small drops can be recorded also at low temperatures by using a dedicated autostage (Linkham Co), though care must be taken during cooling to ensure high transparency of the drop and to avoid crystal movement within the drop. When flash freezing is adopted, these impediments are overcome. Crystals monitored at room temperature via ex situ apparatus can even be frozen and taken to the X-ray diffractometer for the data collection.

Raman microscopy apparatus
The Raman confocal microscope (Jasco, NRS-3100) currently available for ex-situ experiments at the Department of Chemistry of the University of Naples 'Federico II' is depicted in Figure 1. One of the 458, 488 and 514-nm lines of an air-cooled Ar + laser (Melles Griot, 35 LAP 431-220), or 406, 413 and 647 nm lines of a water-cooled Kr + laser (Coherent, Innova 320) can be injected into an integrated Olympus confocal microscope and focused to a spot size of approximately 2 m by a 100x or 20x objective. The laser power at the sample depends on the wavelength ranging from 5 to 100 mW. A holographic notch filter is used to reject the excitation laser line. Raman back-scattering is dispersed through a monochromator (2400 or 1200 grooves/mm grating) and collected by a Peltier-cooled 1024 x 128 pixel CCD photon detector (Andor DU401BVI). Frequency shifts are calibrated by using indene, cyclohexane or CCl 4 as standard, depending on the spectral range of interest. An acceptable spectral resolution is usually considered 4 cm -1 . Actually, since Raman shift for off-resonant spectra is independent of the excitation wavelength, in principle also IR or UV excitation could be used, though at least different optics and detector (a different apparatus) is required. Fig. 1. Scheme of Raman apparatus available at the University of Naples "Ferderico II" (Dept of Chemistry) to collect spectra from protein single crystals. Adapted after .

In situ Raman experiments
Whether the crystal is on the diffractometer or away from it (in situ or ex situ), Raman spectra are not affected by it, but of course only in-situ analysis reveals radiation damage effects (that could be even reversible (Adam et al, 2009)). That is the reason for many efforts in the last decades to make Raman microscopes available at synchrotron beamlines: the European Synchrotron Radiation Facility (ESRF)  dedicates one of their beamlines to in situ Raman-assisted X-ray biocrystallography (see below). A recent review on the microRaman apparatus available at synchrotron facility is reported in (McGeehan et al., 2011). For in situ setup we intend that the X-ray diffractometer is integrated with a laser (guided by an optical fiber) that is backscattered to a spectrograph. To this end, crystals are mounted on an usual crystallographic head kept into a cryo-loop under a nitrogen flux (typically at 100 K). After the crystal is centered for X-ray diffraction data collection experiments, the optimal orientation for the collection of the best Raman spectra has to be determined, typically by manually re-orienting the crystal. Once the crystal orientation that results in the best signal-to-noise ratio is found, Raman spectra are acquired prior and after the X-ray diffraction data collection, at the same crystal orientation. This allow to check if any photo-chemistry is induced by the X-ray beam exposure. For offresonance Raman spectra it is particularly important to keep the drop volume low and to avoid interference from the loop (nylon loops should not intersect the Raman laser beam path (Carpentier et al., 2011)). When Raman spectra are collected as a function of the crystal orientation (Raman crystallography), information can be obtained on the orientation order and the variability of the relative intensity of Raman bands can be estimated (Tsuboi et al., 2007). Indeed, both off-resonance (Kudryavtsev et al., 1998) and resonance spectra (Smulevich & Spiro 1990) are mildly affected by crystal orientation. In this case general physical chemistry rather than structural information can be extracted by Raman crystallography.

Interpretation of Raman spectra of derivative protein crystals 4.1 Raman spectra analysis
A Raman spectrum provides a lot of information: energy shift, intensity and polarization of the scattered light and width of the peak. We will mostly focus on the Raman shift and intensity features. Each Raman shift corresponds to a permitted vibrational transition that relates to the strength constant k of the harmonic oscillator via where µ is the reduced mass. Therefore high frequency corresponds to either stretching or bending with a high strength constant or involving light atoms, such as hydrogen. The presence of the reduced mass at the denominator in equation 1 justifies the wide use of isotopic replacement (eg H/D) for band assignment. Raman transition corresponds, most likely, to a fundamental transition (from the vibrational ground state to the first excited state). Indeed, overtone and combination bands in Raman spectroscopy are much more unlikely than in Infrared spectroscopy (FT-IR). Furthermore, compared to FT-IR, Raman spectroscopy has very different selection rules, and usually bands that are strong in the IR absorption, are not allowed in Raman scattering. Indeed, the great advantage of Raman when studying biological samples is the very weak contribution from water to Raman spectra, compared to IR spectra. Once Raman spectrum has been recorded, reduction for the temperature correction can be performed (Pernice et al., in press). Unless strictly necessary, as for a variable fluorescence coming from background, no baseline correction should be performed. The position of the Raman band can be assigned according to literature or basing on theoretical calculations, isotopic substitution or symmetry considerations (Long, 2002). But usually, the last operations do require experienced spectroscopists (see par 5.1.1) . The comparison of native protein crystal and chemically modified crystal can be performed via difference spectra in case of off-resonance spectrum or just via a simple comparison of the spectra for resonance Raman (Carey, 2006).

Examples of Raman assistance in chemical modifications of protein crystals
Application of complementary techniques that probe different features is an added value to the understanding of the relationship between structure, function and dynamics. Raman microscopy proved to be a valuable support to protein crystallography in all the steps of the 3D structure determination, from the preparation of the derivative crystals up to the interpretation of the electron density maps. We will first present examples of chemical composition changes (cfr 5.1), and then we will move to possible secondary (cfr. 5.2) and tertiary structure (cfr 5.3) modifications occurring upon a physico-chemical treatment.
In some cases crystallography can assist Raman spectroscopy in defining frequency assignments. In this case we speak about crystallography-assisted Raman spectroscopy.

Raman helping phase problem solution
Phase problem in crystallography can be solved for new proteins via multiple wavelength anomalous diffraction (MAD) or by multiple isomorphous replacement (MIR). Both these approaches usually require a chemical modification of the protein crystals.

Raman detection of Se-Met incorporation into protein crystals
Se-Met derivatives are expressed for phase determination using multiwavelength anomalous dispersion (MAD) experiments (Cassetta et al., 1999). Se-Met inclusion is a widespread approach for MAD experiments since Met residues are present with an average occurrence of 1 per 59 residues (Hendrikson et al., 1990). Once crystals of a Se-Met derivative protein have been obtained, only then the absorption spectrum can be used to reveal the presence of selenium for determination of the crystal structure. Se-Met derivative crystals should be stored in a reducing environment (usually adding DTT) and diffraction experiments should be carried out using fresh crystals. It has been reported that two-week storage of Se-Met protein crystals can produce significant deterioration of the crystal diffraction power (Doublié, 1997). Since synchrotron beam-time is not always available immediately after growth of the crystals, a tool to check the Se-Met status in stored crystals is beneficial. A Raman microscopy study was conducted on isomorphous wild-type and Se-Met crystals of the βγ-crystallin from Geodia cydonium (geodin) . Geodin is a protein of unknown function, whose structural characterization could provide information on how homologous βγ-crystallin monomeric proteins evolved. Table 1 summarizes the most relevant Raman bands observed in off-resonant Raman spectra for the SeMet-derivative crystals of geodin. Spectra at low frequency region (400-1200 cm -1 ) are reported in Figures 2. Usually, the Raman spectrum of a polypeptide is subdivided into three main regions of interest: 1) the range between 870-1150 cm −1 , associated with the www.intechopen.com vibrations of the backbone C -C and C -N bonds; 2) the range between 1230-1350 cm −1 , containing the amide III region vibrations, associated with normal modes of various combinations of the C -H and N-H deformations together with the C -C and C -N stretchings (Asher et al., 2001); 3) the range between 1630-1700 cm− 1 , associated with C=O stretching modes, defined as the amide I region (Ngarize et al., 2004). Furthermore, the lower and higher regions can also be informative: i) the conformation and detection of disulphide bridges can be investigated in the low frequency (500-540 cm -1 ) region (Kudryatsev et al., 1998); ii) hydrophobic interactions can be investigated by analysing the C-H stretching region (2800-3200 cm -1 ) (Chourpa et al., 2006), and the S-H stretching region 2550-2600 cm -1 can serve as a valuable probe of local dynamics (Thomas, 1999). The Raman spectra collected on the isomorphous crystals of wild-type and Se-Met geodin crystals reveal the same secondary structure features (Amide I and III). These findings suggest that the presence of the Se-Met does not alter the structure of geodin, as observed for most of the proteins. The main difference between the two spectra (excluding the slightly different intensity of mother liquor signals) is in one narrow region at low frequency. Indeed, the bands in the 570-600 cm -1 region are present only in the Se-Met geodin crystal, and not in the spectra of the mother liquor or in the wild-type geodin crystal (Table 1). This www.intechopen.com spectral feature can be confidently assigned to the C-Se stretching, in agreement with previous Raman studies on selenium-containing organic compounds (Hamada & Morishima, 1977;Paetzold et al., 1967), and on the selenomethionine aminoacid (Zainal & Wolf, 1995;Lopez et al., 1981). Using these assignments, SeMet Raman peaks have been observed also in the SeMet derivative of protein SOUL (Rossi et al., 2009). In these crystals, a quantitative evaluation of the relative amount of SeMet replacement was also achieved by comparative analysis. In principle, Raman microspectra could be also used to reveal post translational modifications, such as phosphorylation, acetylation, trimethylation, ubiquitination and www.intechopen.com glycosilation, as already done by comparing spectra of native and derivative proteins in solution (Sundararajan, et al. 2006;Brewster et al., 2011).

Raman detection of heavy atom derivatives
As previously stated, heavy atoms can be incorporated into protein crystals for the purposes of phasing. The simplest way to soak a heavy atom into protein crystals is to immerse the crystal straight into a drop containing the heavy atom at the final concentration for a soaking time ranging from 10 min up to several days. Different salts can be used for this purpose: successful experiences have been reported using K 2 PtCl 4 , KAu(CN) 2 , K 2 HgI 4 , UO 2 (C 2 H 3 O 2 ) 2 , HgCl 2 , K 3 UO 2 F 5 and para-chloromercurybenzoic sulfate (PCMBS) (Boggon & Shapiro, 2000). During this procedure, typical question are: did actually the ion soak into the crystal? Is the time of soaking long enough for allowing the heavy atom diffusion into the crystal? Raman microscopy can help answering these questions without waiting for a Patterson map. Reports of Raman-based direct Pt incorporation and indirect Eu 3+ soaking are available in literature (Carpenter et al., 2007). Indeed, the Pt-Cl stretching appears at 330 cm -1 that can be compared to the Trp band at 760 cm -1 . From the time-evolution of the relative intensity I330/I760 we can follow Pt 2+ soaking into the crystal. Authors have recently defined a protocol to follow Hg 2+ binding to free Cys residues (unpublished results), that is dependent on the number of Cys residues involved in the binding (1 or 2).

Raman detection of chemical modifications perturbing secondary structure
In some rare cases chemical modifications, such as pH change or Cys-Cys reduction, can significantly change even secondary structure. By taking advantage from the sensitivity of Amide I band to conformational variations, Thomas revealed a β-sheet to α-helix transition upon TCEP [tris(2-carboxyethyl)phosphine] addition to bovine insulin crystals (Zheng et al., 2004). Unfortunately, upon chemical reduction crystals do not diffract anymore, and no crystallographic counterpart is available. An analogous conformational transition occurred upon pH decrease of 5S subunit of transcarboxylate (Zheng et al., 2004), where modifications in the Amide III region (Mikhonin et al., 2006) suggested a significant modification into the crystal matrixes. Also hydration can affect deeply Amide I, and eventually secondary structure, as observed for a collagen-like polypeptide, for which Raman spectrum of the lyophilized powder showed very different Amide I frequencies when compared to single crystals (Figure 3, from Merlino at al., 2008a). In particular, in this case, spectra of (PPG) 10 powders are characterized by three distinct well-resolved amide I bands (1638, 1655, 1690 cm −1 ), that do not match the frequencies observed in the spectra of (PPG) 10 crystals (1629, 1645 and 1669 cm −1 ). The three amide I bands, and especially the well-resolved band at 1690 cm −1 , do not match the frequencies reported by FT-IR solution studies for the unfolded (PPG) 10 (maximum at 1633 and a shoulder at 1665 cm −1 ) or for the unfolded polyproline (one band at 1621 cm −1 ) (Bryan et al., 2007). The amide I bands of (PPG) 10 crystals and powders are significantly different, not only for the intensity distribution of the crystal spectrum (depending on the orientation of the polarization of the exciting laser beam with respect to the main axes of the Raman tensor in the respective unit cell), but also for the frequency of the Raman peaks (After Merlino et al., 2008a). Fig. 3. Medium frequency (A) and low frequency (B) Raman spectra of (PPG) 10 powders, crystals, and mother liquor from which crystals grew up. In spectra (A) and (B) the signals attributed to the mother liquor are tagged by a star. Spectral resolution is 4 cm −1 . After (Merlino et al. 2008a).

Raman detection of chemical modification perturbing tertiary structure
Certainly, the most common structural modifications are related to tertiary structure. The amino-acidic modification can be followed even by off-resonance Raman spectra, whereas modification at metal center can only be detected via resonance Raman spectra. www.intechopen.com

Off-resonance Raman spectra showing tertiary structure modifications
Many Raman markers can provide information on the modified microenvironment of side chains (particularly free Cys and Tyr residues), disulfide bridges conformation or ligands entry, or metal center. The relative intensity of Fermi resonance doublet from Tyr (I856/I830) is strictly dependent on the strength of the H-bond formed by tyrosines with the adjacent groups (1.25 exposed and 0.5 for strong H-bond donation of the OH). Sulfidryl groups exposure of Cys residues can be investigate following the kinetics of H/D exchange vs temperature. Also the pKa of Cys can be extracted by titrating S-H stretching at 2550-70 cm -1 (mercaptoethanol 2580 cm -1 ) compared to the envelope at 2800-3000 cm -1 (Thomas, 1999). When ligand soaking is followed, difference Raman spectra can provide indication on both bound ligands and related conformation changes of the proteins. Carey reported a review of Raman-microscopy applications to follow ligand binding , and more recently he has applied this technique in kinetic crystallography to monitor time evolution oflactamases binding with clinical inhibitors (Carey, 2011). These spectroscopic evidences can be even used as restrains during crystallographic model refinement.

Resonance Raman spectra showing tertiary structure modifications
Metal centers have been well studied via Resonance Raman (RR) spectroscopy for hemoprotein and not-containing heme proteins. A recent example was reported for the major haemoglobin from the sub-Antarctic fish Eleginops maclovinus, for which a variety of coordination, spin state were observed keeping the same oxidation states . In this example, Raman microscopy experiments were conducted on two different carbomonoxy crystals (called Ortho and Hexa) as well as on their ferric and deoxy forms ( Figure 4). The high-frequency region (1300-1700 cm -1 ) of the RR spectrum includes the porphyrin inplane vibrational modes (which are sensitive to the electron density of the macrocycle and the oxidation, coordination and spin state of the iron ion). The deoxy forms of both Ortho and Hexa are pentacoordinated high-spin states (bands at 1355, 1548-1549, 1582 and 1602-1607 cm -1 ). The ferric form contains a hexacoordinated lowspin hemichrome (bands at 1505, 1559, 1588 and 1640 cm -1 ) (Merlino et al., 2011). After long laser exposure times (about 10 min), the Ortho but not the Hexa form appears to be unstable under laser irradiation, and it irreversibly converts to a hexacoordinated low-spin haemochrome state (bands at 1361, 1496 and 1587 cm -1 , respectively). RR microscopy can be also used to identify anomaly in the coordination state of hemoprotein (Merlino et al., 2008b;Vergara et al., 2010). A recent and fine case is the unusual deoxy coordination found in the Antarctic fish hemoglobin from Trematomus newnesi (Hb1Tn) . The crystal structure of deoxy form of this protein reveals distinct coordinations at the α and β hemes, and a high disorder at the EF helices of α heme, hosting the active site. Particularly, the distances His-Fe-His were unusual at the α subunits, raising doubt of hexa-coordination. The medium frequency region ruled out any contribution of bis-histidyl deoxy coordination, confirming a penta coordination. The low frequency regions clearly showed an heterogeneity in the Fe-His stretching, with a broad band attributed to the α/β structural differences in coordination ( Figure 5, after . Fig. 4. Resonance Raman spectra of crystals of Hb1Em in 100 mM Tris-HCl buffer pH 8.0 at room temperature in the carbomonoxy (CO), deoxygenated (Deoxy) and ferric (Met) forms for Hexa (a) and Ortho (b) crystals. The Ortho deoxygenated form (b) converts quickly into haemochrome (Haemo) after 10 min laser exposure. Excitation wavelength, 514.5 nm; laser power at the sample 2 mW for the ferric and deoxy forms and 0.1 mW for the carbomonoxy form. All spectra were an average of at least six spectra with 2 min integration time (After Merlino et al., 2010). www.intechopen.com

Raman-assisted biocrystallography to study chemical mechanisms and Xray radiation damage
Raman assistance can go much beyond the crystal preparation and can provide additional structural information to solve some ambiguities in the electron density map interpretation. Examples include the 1) effect of X-ray damage as a function of the X-ray dose (Garman, 2010). Some examples are breakage of disulphide bonds (Carpentier et al., 2010) and Brominated-DNA photo-dissociation (McGeehan et al., 2007), and even X-ray-induced transient photobleaching in a photoactivatable green fluorescent Protein (Adam et al., 2009). Indeed, third generation synchrotrons brought back the problem of X-ray damage, even at 100 K (Garman, 2010). A complete microspectroscopic picture of the radiation damage is not yet available. The possibility to investigate the different rates of decay for different chemical groups would help crystallographers in making more stable derivative crystals. Authors are currently working on this subject in collaboration with SLS scientists. 2) X-ray induced photo-reduction that may affect metal spin and oxidation state. 3). Identification and freeze-trapping of reaction intermediates (Carpenter et al., 2011). Bourgeois' group was pioneer in this field reporting non-resonance Raman of a trapped iron(III)-(hydro)peroxo species in crystals of superoxide reductase, a nonheme mononuclear iron enzyme that scavenges superoxide radicals (Katona et al., 2007), and further investigations came on RNA polymerase reactions as well (Carey et al., 2011). However this is out of the scope of this chapter, although it represents a formidable challenge for Raman-assisted X-ray biocrystallography.

Aknowledgements
This work was financially supported by PNRA.