|Home | About | Journals | Submit | Contact Us | Français|
G protein-coupled receptors (GPCRs) constitute a highly diverse and ubiquitous family of integral membrane proteins, transmitting signals inside the cells in response to an assortment of disparate extracellular stimuli. Their strategic location on the cell surface and their involvement in crucial cellular and physiological processes turn these receptors into highly important pharmaceutical targets. Recent technological developments aimed at stabilization and crystallization of these receptors have led to significant breakthroughs in GPCR structure determination efforts. One of the successful approaches involved receptor stabilization with the help of a fusion partner combined with crystallization in lipidic cubic phase (LCP). The success of using LCP matrix for crystallization is generally attributed to the creation of a more native, membrane-like stabilizing environment for GPCRs just prior to nucleation and to the formation of type I crystal lattices, thus, generating highly ordered and strongly diffracting crystals. Here we describe protocols for reconstituting purified GPCRs in LCP, performing pre-crystallization assays, setting up crystallization trials in manual mode, detecting crystallization hits, optimizing crystallization conditions, harvesting, and collecting crystallographic data The protocols provide a sensible framework for approaching crystallization of stabilized GPCRs in LCP, however, as in any crystallization experiment extensive screening and optimization of crystallization conditions as well as optimization of protein construct and purification steps are required. The process remains risky and these protocols do not necessarily guarantee success.
G protein-coupled receptors (GPCRs) comprise the largest family of integral membrane proteins in the human genome, with ~800 GPCR family members identified (1). These receptors communicate signals across the plasma membrane in response to a variety of extracellular stimuli, ranging from photons, ions and small molecules to peptides and proteins. The signals are amplified and transmitted to downstream effectors inside the cells primarily through coupling to heterotrimeric guanidine nucleotide binding proteins (G proteins). Upon activation, receptors trigger complex cascades of reactions controlling crucial physiological and cellular processes. As such, GPCRs are implicated in multitude of diseases, making them important pharmaceutical targets. In fact, about 40% of drugs on the market act on the G protein-coupled receptors (2).
GPCR family members share common architecture of a 7 transmembrane α-helical bundle (3). It remains intriguing how such a relatively simple scaffold has evolved to selectively bind to thousands of diverse ligands transmitting signals to dozens of different effectors. The primary questions in the GPCR field include understanding mechanisms of signal transduction and aspects of ligand selectivity and specificity. These and other essential questions can only be answered when a variety of biophysical and biochemical data are combined with high-resolution structures of family members stabilized in different functional states.
Despite the great importance of GPCRs and immense worldwide efforts aimed at obtaining their structures, very little progress has been achieved until recently. Indeed, until late 2007 a high resolution structure of only one highly atypical GPCR member, visual rhodopsin, was known (4). Rhodopsin is unusual in that it is highly abundant in the eye's retina and contains a covalently bound ligand, 11-cis retinal, keeping the receptor conformationally stable in a fully inactive state. However, all other receptors are expressed in miniscule amounts in native tissues and activated by diffusible ligands. For structural studies, it has therefore become necessary that these proteins be heterologously overexpressed, solubilized, and purified. Human integral membrane proteins, including GPCRs, are notoriously difficult to express and they are highly unstable when solubilized by detergents. What makes GPCRs even more challenging is their observed conformational heterogeneity highlighting dynamic equilibrium between multiple functional states (5), obviously vital for signal transduction; however, a formidable roadblock for structural studies. For example, most GPCRs possess a certain level of basal activity in the apo-, or unliganted, state. Ligands that increase activity are called agonists, those that bind to the receptor and block it without changing its activity are called antagonists, and ligands that decrease the activity below basal are called inverse agonists. Typically there is a whole range of intermediate states between completely inactive and fully active receptor, and ligands that induce or stabilize them are called partial inverse agonist and partial agonists. In general, receptors are more stable and less conformationally heterogeneous in the inactive states and become progressively less stable and more heterogeneous as they are converted into more active states.
Thus, it is now generally accepted that in order to crystallize a GPCR, it has to be stabilized in a single conformational state. Last year an explosion of technological breakthroughs have made it possible to stabilize the conformation and determine the high-resolution structures of several members of the GPCR family. One of the successful approaches involved developing a monoclonal antibody toward a structural epitope on the third intracellular loop of the human β2-adrenergic receptor (β2AR) (6). The antibody/receptor complex was crystallized in bicelles and the structure was determined at a moderate resolution of 3.4–3.7 Å (7). In the second approach a highly flexible intracellular loop between helices 5 and 6 of the human β2AR was replaced by a compact and stable protein, lysozyme from T4 bacteriophage (T4L) (8). The chimeric receptor (β2AR-T4L) was crystallized in lipidic cubic phase (LCP) and the structure solved to a resolution of 2.4 Å (9). The same approach was recently used to obtain the structure of the human adenosine A2A receptor at 2.6 Å resolution (10). Finally, the third approach toward receptor stabilization involved a systematic mutagenesis strategy (11, 12). This strategy helped to identify 6 mutations in turkey β1-adrenergic receptor that increased its melting temperature in detergent solution by 21 °C. The thermally stabilized receptor was crystallized by a traditional vapor diffusion technique and the structure was solved to 2.8 Å (13). All of these studies were done on receptors with bound ligands, typically inverse agonists or antagonists, locking them in a stable conformation. Additional stabilization was achieved by crystallizing β2AR/Fab complex in bicelles (7), and both T4L fusion receptors, the β2AR-T4L and A2AR-T4L, in the lipidic cubic phase (9, 10, 14).
Comparison of these new structures highlights important details related to the plasticity of the GPCR 7TM core and the involvement of the extracellular loops in the ligand binding (15, 16). In addition, the new structures provided superior templates for homology-based modeling of other GPCRs and for performing virtual ligand screening and structure based drug design (17). While new approaches have enabled GPCR structure determination, more remains to be done to address crucial questions centering on mechanisms of signal transduction upon receptor activation. To gain insights into these mechanisms, it is essential to obtain a high-resolution structure of an activated receptor with an agonist bound. Recently a crystal structure of the bovine opsin bound to a transducin fragment shed some light on the receptor activation mechanism (18). However, the atypical mode of the rhodopsin activation and the lack of an agonist in the structure prevent generalizations applicable to other GPCRs responding to diffusible ligands. It is conceivable that a similar approach of stabilizing an activated receptor by a G protein, Gα subunit or by a fragment of Gα will yield a high resolution structure in the near future.
Two of the best resolution GPCR structures to date, not including rhodopsin, have been obtained using crystallization in LCP, also referred to as in meso crystallization (19). The success of in meso crystallization can be attributed primarily to two factors: stabilization of highly flexible receptors in a membrane-like LCP environment and formation of type I crystal lattice leading to highly ordered and strongly diffracting crystals. Here we describe methods related to reconstitution of purified receptors in LCP, performing pre-crystallization assays, setting up crystallization trials, harvesting crystals and acquiring crystallographic data. We recommend the reader to get familiar with more general and detailed protocols for in meso crystallization recently published in ref. 20. In this chapter the emphasis is on working with highly unstable GPCRs, which require pre-crystallization analysis of multiple constructs, extensive crystallization optimization and crystallographic data collection on multiple microcrystals using a highly collimated synchrotron beam.
The procedure described here can be used to supplement monoolein or any other LCP host lipid with lipophilic additives. Such additives can modulate properties of LCP or can preferentially interact with proteins reconstituted in LCP and therefore are useful for optimization of crystallization conditions. Cholesterol was found to be the best lipid additive significantly improving the β2AR-T4L crystal size (9, 14). Cholesterol was also used for crystallization of adenosine A2A receptor (10). Compatibility of monoolein-based LCP with several lipid-like molecules was reported in ref. 22.
LCP-FRAP assay is designed to measure diffusion properties of membrane proteins reconstituted in LCP at a variety of different screening conditions (23). The long-range diffusion of membrane proteins in LCP is essential for successful crystallization, however, the microstructure of LCP imposes spatial constrains on diffusion of large proteins or oligomeric protein aggregates. Our data indicate that one of the primary reasons for failure of the in meso crystallization trials with GPCRs is due to a fast nonspecific protein aggregation. While this event is equivalent to massive precipitation, and is easily recognizable in crystallization trials in solution, it does not produce any visual feedback in LCP, since the size of the non-diffusing stuck protein oligomeric aggregates is well below the optical resolution. It has been found that the aggregation behavior of a protein depends on the particular protein construct, host lipid and additives employed in crystallization trials.
The LCP-FRAP protocol, described below, involves labeling the protein of interest with a fluorescent dye, reconstituting the labeled protein into LCP, loading a 96-well glass sandwich plate with LCP and screening solutions, performing a FRAP data acquisition and analyzing the data. Work on automation of data acquisition and data analysis is in progress. The LCP-FRAP assay is useful for prescreening multiple protein constructs, for assessing the role of ligands and lipid additives and for identifying precipitants that are non-conducive to protein diffusion. The latter are then excluded from the subsequent crystallization screens.
The expected labeling percentage is between 2 – 10 %.
The protocol describes data collection using a custom build FRAP station controlled by an ImagePro Advance Microscopy program suite (Fig. 3A). Most of the steps are automated using the scripting language of the ImagePro. Similar protocol can be implemented on any FRAP capable microscope system.
FRAP data analysis steps described here are implemented using ImagePro (Media Cybernetics), Microsoft Excel and Prizm (GraphPad), however other image analysis and curve fitting packages can be employed to perform similar tasks.
Recording a complete fluorescence recovery curve is time consuming, taking 10 – 30 min per sample depending on the protein diffusion coefficient. This limits the number of samples that can be processed in a reasonable time. We noticed that in the case of GPCRs majority of screens result in no measurable recovery of signal. Thus, when a 96-well plate is used to set up FRAP samples, it is possible to sequentially bleach all the samples first and then to measure the total recovered intensity for each sample after 30 – 60 min incubation. This procedure provides protein mobile fractions for all 96 samples within ~2 hrs. Full recovery curves are then recorded only for the samples with significant protein mobility.
In meso crystallization trials can be performed manually or robotically. Protocols for manual crystallization setup are provided in this section; using robotics for in meso crystallization is extensively discussed in refs. 24, 26. It is preferable to perform in meso crystallization trials in glass sandwich plates, described in refs. 24, 27. These plates have excellent optical properties for detection of very small colorless protein crystals growing in LCP matrix. Currently the glass sandwich plates are not available commercially and must be assembled from separately ordered glass slides and perforated spacers (section 3.3.1.). Alternatively, most commercial micro-batch, sitting or hanging drop plates could be used for the manual crystallization setup with the caveat that one may obtain less than optimal conditions for detection of small colorless crystals, primarily due to the scattering of light from a rough boundary between the LCP bolus and precipitant solution (Fig. 4A–C). The problem can be circumvented by sandwiching the LCP bolus with a 5 mm diameter glass coverslip (Warner Instruments, cat.# W2 64-0700) as shown in Fig. 4D–E and used in ref. 28.
Crystallization setup starts with mixing protein solution with lipid as described in section 3.1.2. It is advisable to have plates and screening solutions ready before starting protein reconstitution in LCP, and to proceed to crystallization setup (section 3.3.2.) immediately after forming protein-laden LCP, because some proteins may not be stable in LCP without added precipitant solution. The whole process of manual setting up a 96-well plate including mixing of protein and lipid takes about 1 hr. In addition to the description of the manual in meso crystallization setup, this section contains protocols for crystal detection, optimization and harvesting.
Typically crystals start appearing in the time frame between a few hours and a few weeks after setup. Majority of GPCR crystals are first detected between 12 hours and 7 days of incubation. Crystallization wells should be inspected periodically either under a microscope or with an automated imager. We use the RockImager line of incubator/imagers from Formulatrix which is compatible with imaging the glass sandwich plates. In this section we describe inspection of wells and detection of crystals using a microscope equipped with a polarizer and a rotating analyzer. An automatic imager adds a convenience of taking images of wells at scheduled time points and storing them in a database for later viewing. In difficult cases, however, it is useful to take the plate out of the imager and to inspect it under a microscope using a higher magnification than what is normally achievable with the imager.
Initial crystal hits in LCP often appears as showers of very small crystal and crystallization conditions have to be optimized. In general, in meso crystal optimization strategies are similar to optimization strategies for crystallization of macromolecules in solution. One difference is that along with tweaking the precipitant solutions composition and pH, one has the ability to optimize the LCP host lipid and lipid additives. When planning for optimization of in meso trials it is important to understand how LCP responds to changes in concentration of a given component. For example, salts at high concentration can transform LCP into a hexagonal phase, which is not conducive to crystallization, while some small organic molecules, like 2-methyl-2,4-pentanediol (MPD), can completely dissolve the lipids, usually inducing massive protein precipitation. Effects of soluble and lipid-like components on LCP are described in refs. 22, 30, 31. Other parameters that we found to be critical for optimization of crystal growth in meso are the size and the shape of LCP bolus and the protein concentration. In this section we describe steps that have proven to be useful for optimization of β2AR-T4L crystals (9).
Due to a strong adhesion of the double sticky spacer to glass slides it is impossible to peel off the cover slip to expose a single well for crystal harvesting. Therefore a piece of the cover glass should be cut and removed first. When crystals are obtained in commercial plastic trays, in which well opening is a trivial procedure, skip Section 220.127.116.11. and proceed directly to crystal harvesting section 18.104.22.168.
The protocol describes data collection strategies from small LCP grown GPCR crystals using a 10 μm minibeam on the GM/CA CAT beamlines at the Advance Photon Source (Argonne, IL). In most parts the protocol can be directly translated to data collection on any other microcrystallography beamline with beam size of 10 μm or less.
This part of the protocol was originally designed to align invisible in a frozen opaque lipidic mesophase crystals of β2AR-T4L with a 10 μm minibeam at the GM/CA CAT beamlines. Recently an automatic rastering was implemented at the GM/CA CAT to facilitate crystal centering, however, the original procedure described here can be useful at other microfocus beamlines.
Radiation damage limits high resolution data collection from small crystals. Collecting and merging data from multiple crystals is necessary to acquire a full data set. It is essential to optimize data collection strategy to obtain the best possible results. Here we describe the protocol used to collect high quality data from ~ 30×15×5 μm3 β2AR-T4L crystals and ~ 60×10×5 μm3 crystals of A2AR-T4L.
The in meso crystallization method was introduced more than a decade ago (34), and since that time it has proven successful in obtaining high resolution structures of difficult membrane proteins, such as human GPCRs (9, 10, 14). The method, however, has been in limited use mostly due to the difficulties in handling of sticky and viscous, gel-like LCP material. The protocols in this chapter provide directions for approaching in meso crystallization of GPCRs or any other membrane proteins. As with any crystallization experiment, the in meso crystallization approach requires substantial screening and optimization efforts, without being guaranteed success. Additionally, for many membrane proteins, and especially in the case of GPCRs, extensive protein engineering aimed at protein stabilization and elimination of highly flexible and disordered parts is an absolute pre-requisite before entering into crystallization trials. Lastly, although crystallography provides atomic resolution 3-d structures, these static snapshots of preferred stable conformations preclude full understanding of the dynamic nature of these molecules. With increases in the number of GPCR crystal structures other complimentary approaches, such as NMR and computer modeling, also covered in this volume, are becoming essential for determining the structural basis for ligand specificity and for deciphering the mechanisms of signal transduction.
This work was supported in part by the NIH Roadmap Initiative grant P50 GM073197 (JCIMPT), Protein Structure Initiative grant U54 GM074961 (ATCG3D), and NIH grant R21 RR025336. The authors acknowledge contributions from colleagues Michael A. Hanson, Wei Liu, Jeffrey Liu, Mark Griffith, Ellen Chien, Veli-Pekka Jaakola, Chris Roth and Peter Kuhn.
The authors acknowledge the support of Janet Smith, Robert Fischetti, and the GM/CA-CAT team at the Advanced Photon Source, for assistance in development and use of the minibeam and beamtime. The GM/CA-CAT beamline (23-ID) is supported by the National Cancer Institute (Y1-CO-1020) and the National Institute of General Medical Sciences (Y1-GM-1104).
1Protein prepared for crystallization trials or pre-crystallization assays should run on an analytical size exclusion chromatography column as a single sharp monomer peak without significant contribution from dimers or larger oligomeric aggregates. Protein solution should not contain excessive amounts of detergent as high concentration of detergents could destroy the lipidic cubic phase (35, 36). If the high detergent concentration creates a problem, try reducing the detergent content as low as possible during purification steps. For example, for His-tagged proteins use saturated binding to a Ni sepharose resin and elution with a minimal amount of buffer to concentrate the protein without increasing the detergent content, and finally use the largest possible cutoff centrifugal concentrators.
2Protein can be labeled with a variety of fluorescent dyes, readily available from Invitrogen, GE Healthcare and other companies. We have tried Fluorescein, Rhodamine 6G, Tetramethylrhodamine and Cy3. Our preferable choice is Cy3 due to its good solubility, low environmental sensitivity and suitable bleaching properties. There are two common conjugation chemistries for protein labeling: thiol-reactive and amino-reactive. The advantage of using amino-reactive dyes is that virtually any protein can be labeled and that these dyes impose a minimal disturbance to the protein when attached predominantly to the N-terminus. The drawback of the amino-labeling is that co-purified with the protein free amino group-containing lipids, such as phosphatidylethanolamines, are also labeled. This introduces unwanted background signal, which is difficult to separate from the protein signal. The thiol-reactive dyes do not label lipids, however, they rely on availability of free cysteins exposed on the protein surface. Additionally we have found that cystein labeling often destabilizes GPCRs. In this protocol we describe universal protein labeling at its N-terminus with succinimidyl ester derivatives of fluorescent dyes. We also recommend trying cystein labeling if it does not disturb the properties of the protein. Labeling with thiol-reactive dyes can be performed according to manufacturer's instructions.
3The 96-well plate is designed primarily for compatibility with robotic crystallization setup and imaging (Fig. 4H). In case of manual operations it is more convenient to make glass sandwich plates of standard 25 × 75 mm2 microscope slides, spacers with punched holes arranged in 8 rows and 2 columns (Fig. 4F) or 9 rows and 3 columns (Fig. 4G), and 18 × 18 mm2 (Fig. 4F and H) or 25 × 25 mm2 (Fig. 4G) glass cover slips depending on the wells configuration.
4Screening solutions for LCP-FRAP should be selected to represent a range of common precipitants. We found that PEG-ion-pH and ion-pH screens with 30 %v/v PEG 400 or 20 %w/v PEG 4000, variety of salts and pH ranging from 5 to 8 are good starting points in screening for diffusion of GPCRs with and without fused lysozyme.
5To perform FRAP measurements we have used a simple and affordable system consisting of a fluorescent microscope with an attached laser. Alternatively, any dedicated commercial FRAP systems and most confocal microscopes can be used to run LCP-FRAP assays that are described in this protocol.
6The Hamilton syringe dispenser can be modified as described in ref. 37 to decrease the dispensing volume ~3 times, from 200 nL to 70 nL.
7It is convenient to use commercial sparse matrix screens for initial screening. On average, however, about 30 % of commercial screen conditions are not compatible with LCP crystallization, since they transform LCP into lamellar or hexagonal phase or completely dissolve the lipid (30). Such conditions can be diluted 2 times with water to increase the compatibility percentage. Additionally, more specific grid or sparse matrix screens can be prepared following results of the LCP-FRAP assays described in section 3.2.
8During in meso crystallization the soluble fraction of LCP is diluted ~50 times, due to vast excess of precipitant (1 μL) over the LCP bolus (50 nL), thus significantly depleting the ligand in the receptor vicinity. Therefore, depending on the ligand binding affinity, solubility and off-rate, it may be necessary to supplement crystallization screens with the ligand.
9We prefer MiTeGen MicroMounts over standard nylon loops for harvesting crystals from LCP due to their rigidity and thinner profile, which allow for easier penetration into LCP and for picking minimal amount of lipids along with the crystal.
10Do not exceed 100 – 200 mg of total lipid mixture per vial, since complete removal of solvent will be difficult. Use several vials or larger volume vial for making larger volume stocks of mixed lipids.
11Melting temperature of monoolein is 37 °C. Different lipids can melt at higher or lower temperature. Adjust the incubation temperature to be few degrees above the lipid melting temperature. Incubation time should be no longer than necessary to melt the lipid in order to avoid possible degradation.
12After taking the molten lipid from incubator it remains liquid at room temperature for few minutes before it solidifies allowing for transfers with a pipette.
13The syringe mixer can warm up due to a friction between the lipidic mesophase and the syringe coupler needle. To avoid heating up the sample, limit the mixing rate to ~1 stroke / s. A useful trick is to slightly cool down the mixer by putting it for few seconds on ice or in a refrigerator, which speeds up the process of achieving a homogeneous cubic phase. Do not overcool it, however, as below 18 °C monoolein-based LCP is unstable and can convert into a lamellar crystalline phase damaging the protein.
14We provide the protocol for labeling and cleaning proteins with C-terminal His-tag. Procedure for labeling untagged proteins should be modified accordingly to include either ion-exchange or size exclusion chromatography for removing the unreacted dye.
15The pH of the buffer should be between 7 and 7.5 to label predominately the free N-terminus of the protein. The buffer should not contain free amino groups.
16Number of laser pulses depends on the dye, filters, laser power, sample thickness, etc. Adjust the number of pulses to achieve ~30–50 % bleaching.
17If the experimental recovery data do not follow Eq.1, it is likely that more than one population of diffusing molecules is present in the sample. The extra signals can originate from different protein oligomeric states or from labeled lipids. If two populations of molecules have substantially (more than an order of magnitude) different diffusion rates, then it is possible to fit the experimental recovery curve with a two component equation and extract both contributions. In practice, when amino-labeling is used, even after an extensive wash in an attempt to remove labeled lipids, there is a residual recovery signal of 5 to 15 % coming from the labeled lipids. The protein diffusion signal can in most cases be extracted using a two component diffusion equation with a fixed characteristic diffusion time for lipid molecules determined from a separate FRAP measurement on a sample containing only labeled lipids (23).
18If the needle is too far from the glass surface the cubic phase bolus coming out of the needle as a tube curls back and sticks to the needle. If the needle is too close to the glass surface the cubic phase bolus balls up and sticks to the needle. Both of these cases result in no or incomplete delivery. The right distance and feeling for accurate dispensing come with practice and is relatively easy to achieve.
19Crystallization plates can be incubated at different temperatures, however, monoolein-based cubic phase is stable only at temperatures above 18 °C (38). Recently, several new lipids were developed specifically for low temperature in meso crystallization (39, 40). It is important to avoid temperature fluctuations during plate incubation and imaging, because fluctuations by just few degrees can induce formation of liquid droplets inside the cubic phase. These droplets scatter light, making it difficult to detect crystals.
20Appearance of similar to original crystals in such control trials can confirm that something else apart from the protein is crystallizing. Control trials resulting in no crystals, however, do not prove that the original crystal was from the protein, since it is difficult to exactly reproduce protein-free conditions.
21Be aware that sometimes even low percentage labeling can prevent protein from crystallizing. Try different dyes and conjugation chemistries as discussed in Note 2.
22Before attempting UV imaging check that material, from which crystallization plates are made, is sufficiently transparent to UV light. For example, for glass sandwich plates we use 1 mm slides made of electroverre glass (Erie Scientific) with transmittance of 40 % at 280 nm, allowing for efficient excitation of protein's tryptophans. The coverslips used in these plates are made of 0.2 mm borosilicate glass, which is essentially non-transparent to UV light.
23When the described optimization procedure was applied to initial crystal hits of β2AR-T4L, improvements were achieved almost at each step. Salt in the initial hit, lithium sulfate, was replaced by sodium sulfate as a result of step 2. Initial Hepes buffer was replaced by bis tris propane, which consistently gave better crystal size and shape, on step 3. Additions of cholesterol and dioleoylphosphatidylethanolamine have improved crystal size on step 5. The optimum concentration of cholesterol was found to be 10 %(wt/wt). The best soluble additive was identified as 5% of 1,4-butanediol on step 6. We found that increasing concentration of β2AR-T4L from 20–30 mg/mL, which was used in the initial screening to 50 mg/mL resulted in larger crystal size on step 8. Increasing protein concentration even further to 60 mg/mL destabilized the lipidic cubic phase and abolished crystal growth. Finally, we observed that decreasing the volume of cubic phase from 50 to 20 nL on step 9 improved the crystal size even further.
24The rationale behind adjusting the LCP bolus volumes is that the larger the volume the longer it takes for the precipitants to diffuse in the LCP and establish an equilibrium. The transient gradients forming during this process may affect the nucleation and crystal growth rates. In meso crystallization robot (24) can be used to vary lipidic cubic phase volume between 20 and ~500 nL. With manual setup the cubic phase volume range is limited to 70 – 700 nL when using the modified syringe dispenser coupled to 10 and 100 μL syringes (37).
25Certain precipitants may transform LCP into a sponge phase (41). It is virtually impossible to harvest crystals grown in the sponge phase using the described technique. The sponge phase has a liquid-like properties and is drawn away from the well by the surface tension when the well is opened up. To overcome this problem the sponge phase can be transformed back into the cubic phase by lowering the concentration of the precipitant. The cubic phase has a gel-like consistency and stays in place when well is opened. For example, the common precipitant, PEG 400, at concentrations above 35 %(v/v) can induce sponge phase formation. Lowering PEG 400 concentration to 30 %(v/v) on this step will bring LCP back within few minutes. When lowering PEG 400 concentration keep concentrations of other ingredients unchanged.
26On this step it is possible to dissolve the lipid and free the crystals using enzymatic lipid digestion (42), detergent solutions (43), sponge-inducing compounds (41) or mineral oils. We have observed that any of these treatments were detrimental to the diffraction quality of harvested crystals of β2AR-T4L and A2AR-T4L. These crystals diffracted the best when harvested directly from the cubic phase.
27When harvesting a crystal from the cubic phase, try picking up as little lipid as possible to reduce scattering background from lipids during data collection. If the crystal is located deep inside the cubic phase bolus, use a micro-tool or an empty MiTeGen MicroMount to remove excess of the cubic phase from the top and expose the crystal to the surface. Then use a fresh MicroMount to harvest the crystal.
28Beam attenuation will depend on the flux. Attenuate as much as possible to reduce the radiation damage while still having the ability to observe low resolution diffraction from the crystals.
29On this step it is important to collect a full dataset with a minimal damage to the crystal. Resolution is not critical as long as the dataset is complete and scales well, thus giving reasonable Rsym. Typical datasets we collected at this step had resolution ~5 – 6 Å and Rsym < 10%.
30This method relies on a high isomorphicity between crystals. The in meso grown crystals of β2AR-T4L and A2AR-T4L did have this quality. The data rejection rate due to poor merging was less than 20 %.
31To obtain a full dataset for β2AR-T4L in complex with carazolol we merged data from 27 crystals (9). For β2AR-T4L/timolol (14) and A2AR-T4L/ZM241385 (10) 11 and 13 crystals were used respectively. Fewer crystals were required in the latter two cases because of the higher symmetry space group and slightly lower resolution.