Introduction

In meso crystallization of membrane proteins (MPs), a powerful technique for MP structure determination, critically relies on the choice of the lipids that form the LCP around room temperature1,2. In the method, a solubilized target MP is homogenized with an LCP-forming lipid, to uniformly reconstitute the MP into the native biomembrane mimetic LCP lipid bilayer. The added crystallization solution triggers the nucleation and crystal growth of the MP within the lipid bilayer environment, where the crystallization process essentially relies on the three-dimensional bi-continuous LCP bilayer architecture2. Moreover, the recent development of viscous media injectors3 has facilitated the direct use of a native crystal growth LCP-matrix as a crystal delivery medium in serial femtosecond or millisecond crystallography (SFX or SMX), known as LCP-SFX and LCP-SMX, respectively4,5.

In spite of the increasing functional roles of the LCP, the currently available lipid species have been restricted by the limited choice of polar lipids. In fact, almost all of the past crystallization trials and LCP-SFX data collections have been performed by using monoacylglycerols (MAGs) at room temperature, with MO being the most widely used lipid6. However, MO is not ideal. For instance, according to the equilibrium phase diagram of the MO/water system, the MO-LCP gives way to the LC phase below about 17 °C7. As LCP is prone to supercooling8, the MO-LCP may remain supercooled below 17 °C. However, as the MO-LCP is metastable below 17 °C, the risk always remains that the LCP → LC phase transition could occur at any time during the experiment. The LCP → LC phase transition incurs numerous difficulties. For example, the LC phase does not support MP crystallization, and in LCP-SFX, when microcrystals randomly dispersed in the MO-LCP are injected into an evacuated X-ray diffraction chamber at 20 °C, evaporative cooling causes transformation into the MO-LC phase. This leads to strong MO-LC Bragg diffractions from crystallized hydrocarbon chains and three-dimensional orders in the LC molecular packing, which interfere with the diffraction spots from MP crystals3. As background reduction is crucial in serial microcrystallography9, there is a need to develop new lipid matrices that support MP crystallization and do not undergo the LCP → LC phase transition in LCP-SFX4. We now report that one of the LCP-forming isoprenoid-chained lipids (IPCLs), EROCOC17+4 with the acyl chain of 18 carbon atoms-long10,11 (see Fig. 1 for more details), serves as a new matrix lipid that satisfies the above requirements.

Figure 1
figure 1

Chemical structure of 1-O-(5, 9, 13, 17-tetramethyloctadecanoyl)-rac-erythritol (A) and 2-O-(5, 9, 13, 17-tetramethyloctadecanoyl)-rac-erythritol (B). The lipid used in this study was a mixture of 92% 1-O- and 8% 2-O-isomers, which is abbreviated to EROCOC17+4. In the A2AR crystallization trials, a 9:1 (w/w) mixture of EROCOC17+4 and cholesterol was employed as a crystallization matrix, which is referred to as an EROCOC17+4-matrix (see Methods Crystallization section for more details). The isoprenoid chain is abbreviated as Cp+q, where p and q stand for the number of carbon atoms in the longest unbranched carbon chain in the molecule (excluding the carbonyl carbon) and the number of methyl branches along it, respectively. Thus, EROCOC17+4 represents a lipid molecule, in which the C17+4 chain is linked to the erythritol head group (ER) via an ester linkage (OCO). (C) Chemical structure of 1-(cis-9-octadecanoyl)-rac-glycerol (MO). It is noted that EROCOC17+4 and MO both possess the acyl chain of 18 carbon atoms-long.

Full size image

The recently developed LCP-forming IPCLs10,11,12,13,14, are characterized by their regularly methyl-branched chain structure (oligomers of a –(CH2)2CHCH3CH2– unit with a terminal –(CH2)2CHCH3CH3 unit, see Fig. 1), which is remarkably different from that of MAGs, consisting of different lengths of linear hydrocarbon chains with a cis-double bond at various chain positions (see Fig. 1)6. Owing to the regularly branched chain structure, IPCL-LCPs are characterized by the low LC → LCP phase transition temperature (TK): most of the TK values are close to or below 0 °C, as shown in Supplementary Table 110,11,12,13,14. Thus, they are attractive as a new crystallization matrix lipid for temperature-sensitive proteins and as an LC phase formation-free crystal delivery medium in LCP-SFX. However, to the best of our knowledge, only two structural analyses of bacteriorhodopsin (bR) with IPCL-matrices have been reported so far. One used an ether-type β-XylOC16+413, which yielded crystals at 2 Å resolution with a reduced crystallographic twin ratio15, and the other used an amide-type GlyNCOC15+4, where the successful crystallization was performed at both 20 and 4 °C12. However, the utility of IPCLs as a possible crystal delivery medium in LCP-SFX has not been evaluated.

To assess EROCOC17+4 as a possible matrix lipid for MP crystallization and a crystal delivery medium in LCP-SFX, we employed the G protein-coupled receptor A2AR as a model MP, considering that MO-matrix, a 9:1 (w/w) mixture of MO and cholesterol, grown high-resolution crystal structures with ordered lipids have been reported16,17,18,19. We now demonstrate that EROCOC17+4-matrix, a 9:1 (w/w) mixture of EROCOC17+4 and cholesterol (see Methods Crystallization section for more details), supports the crystallization of A2AR at 20 °C, yielding a high-resolution crystal structure comparable to that of the MO-matrix grown crystals at 1.8 Å resolution (PDB ID: 4EIY, hereafter referred to as the 4EIY model), and that EROCOC17+4-matrix allows us to perform the LCP-SFX data collection with a sample kept at 4 °C, without the LC phase formation.

Results and discussion

Crystallization of A2AR in the EROCOC17+4-matrix

In the screening of crystallization conditions of A2AR as a model MP with the EROCOC17+4-matrix, we first noted that the optimal crystallization conditions for the MO-matrix, i.e., 100 mM sodium citrate, pH 5.0, 25–28% PEG400, 40–60 mM NaSCN, and 2% 2,5-hexanediol16, supported micro-crystal (≤ ~ 1 µm) formation in the EROCOC17+4-matrix at 20 °C. However, these crystals were too small for high-resolution structural analyses. To identify the optimum crystallizing conditions for the EROCOC17+4-matrix, we screened the effects of the pH (4.5–6.5 with 25–300 mM sodium citrate), and the concentrations of PEG400 or PEG300 (17.5–47.5%), NaCl (0–1000 mM), and the additives NaSCN (0–60 mM) and 2,5-hexanediol (0–2%), on the crystallization behavior, and noted that the most effective parameter was the PEG400 (or equivalently PEG300) concentration. The size and density of EROCOC17+4-matrix grown A2AR crystals as a function of the PEG400 concentration (CPEG400), in an otherwise fixed solution composition of 50 mM NaSCN, 2% 2,5-hexanediol, and 100 mM Na citrate, pH 5.50, are shown in Fig. 2A,B

Figure 2
figure 2

Effect of PEG concentration on size and density of A2AR crystals grown in the EROCOC17+4-matrix. (A) Size (black) and density (magenta) of EROCOC17+4-matrix grown A2AR crystals as a function of the PEG400 concentration, in an otherwise fixed solution composed of 50 mM NaSCN, 2% 2,5-hexanediol, and 100 mM Na citrate, pH 5.50. (B) Cross polarized (upper) and second harmonic generation (lower) images in a 96-well glass screening plate, showing the birefringence and chirality-derived signals of the protein crystals, respectively. (C) Crystals used for SS-ROX under optimized conditions: 38% PEG400, 50 mM NaSCN, 2% 2,5-hexanediol, and 100 mM Na citrate, pH 5.75. Upper: cross polarized, lower: bright field.

Full size image

The crystal size tended to increase as \({C}_{PEG400}\) increased; 3–10 µm at 30–35%, and eventually reaching ~ 15 μm sized-crystals at 35–40% PEG400, where the first crystals appeared at 1–2 days and continued to grow over ~ 1 week to reach the maximum size of ~ 25 μm (Fig. 2C), which is suitable for data collection by serial synchrotron rotation crystallography (SS-ROX)20.

On the other hand, the crystal density decreased as CPEG400 increased, from 107–108 crystals/mL around 30% down to 106–107 crystals/mL around 40%. The \({C}_{PEG400}\)-dependent crystal size/density relation shown in Fig. 2A served as a useful guide for preparing ~ 50 μL of LCP densely packed with small crystals (\(\ge {10}^{7}\) crystals/mL and 5–10 μm) needed for complete data collection by LCP-SFX. The final crystallization condition for the EROCOC17+4-matrix based LCP-SFX was 100 mM sodium citrate, pH 5.5, 37% PEG300, 50 mM NaSCN, and 2% 2,5-hexanediol. Thus, except for the PEG concentrations, the optimum crystallization conditions for A2AR in the EROCOC17+4-matrix did not differ significantly from those for the MO-matrix: the EROCOC17+4-matrix required 35–40% PEG400/PEG300, as compared to 25–28% PEG400 for the MO-matrix.

Crystallographic analyses by SS-ROX and LCP-SFX

The crystallographic data of the EROCOC17+4-matrix grown A2AR crystals are summarized in Table 1. In the present SS-ROX and LCP-SFX experiments, we employed the A2AR crystals prepared at 20 °C. The samples for SS-ROX were flash frozen in SS-ROX_cryo, and those for the LCP-SFX loaded into the injector were kept at 4 or 20 °C during the measurements in LCP-SFX_4 °C and LCP-SFX_20 °C, respectively. The crystallographic space group of these EROCOC17+4-matrix grown crystals was the same as that for the MO-matrix grown crystals (PDB ID 4EIY), C2221. The structures of the EROCOC17+4-matrix grown A2AR crystals were determined at 2.0 Å resolution by SS-ROX at the cryogenic temperature, and at 1.8 Å by LCP-SFX_4 °C and LCP-SFX_20 °C, respectively. As in the cases described in the references21,22, no significant differences were observed in the electron density maps and the models obtained at the three temperatures, except for the increasing B factors and the decreasing number of water molecules at the higher temperatures (Table 1). The resolution of the LCP-SFX data was slightly better than that of the SS-ROX data for the present measurements, although the LCP-SFX and SS-ROX data are difficult to compare because the crystals and X-ray sources are different. On the other hand, comparing the LCP-SFX_4 °C and LCP-SFX_20 °C data obtained under exactly the same conditions, except for the sample temperatures, the LCP-SFX_4 °C model seems to be more reliable. The reliability factors, signal-to-noise ratio (SNR) values, and average B values for LCP-SFX_4 °C are slightly better than those for LCP-SFX_20 °C (Table 1). For example, the Rsplit/work/free values for LCP-SFX_4 °C are 6.6/17.5/20.7% whereas those of LCP-SFX_20 °C are 7.1/17.7/20.9%. Since the diffraction data could be improved by lowering the sample temperature, as described above, the LCP-SFX_4 °C model is used for further structure comparisons and analyses. The root mean square deviation (RMSD) value for the corresponding 299 Cα atoms between the LCP-SFX_4 °C and the MO-matrix based 4EIY model is 0.23 Å RMSD, and those for the inter Na+, a conserved sodium ion bound in A2AR, and ZM241385, a strong inverse agonist for A2AR, are 0.20 and 0.18 Å, respectively (Fig. 3A).

Table 1 Data collection and refinement statistics at three sample temperatures.
Full size table
Figure 3
figure 3

Comparison of EROCOC17+4- and MO-matrix grown A2AR crystal structures. (A) Crystal structures of LCP-SFX_4 °C (magenta) and 4EIY (cyan). The bRIL moiety inserted in ICL3 is omitted from the coordinates (magenta dashes). Sodium ions (green and dark grey balls for LCP-SFX_4 °C and 4EIY, respectively) and ZM241385 (red and dark grey sticks for LCP-SFX_4 °C and 4EIY, respectively) are also shown. Black lines approximately correspond to the hydrophobic/hydrophilic boundary of the lipid bilayer. IC and EC are intracellular and extracellular sides, respectively. Two well characterized EROCOC17+4 (red; E1 and E2) and three cholesterol (orange; C1, C2 and C3) models are displayed as sticks. (B,C) Molecular surface of LCP-SFX_4 °C with well characterized lipids. The 2mFo-DFc electron density maps around the lipid models are contoured at 1.0 σ (pink mesh). The carbon atoms of the EROCOC17+4 and cholesterol stick models are shown in magenta and orange, respectively. The oxygen atoms of both lipids are shown in red. This figure is prepared by PyMOL Ver. 1.8 (https://pymol.org).

Full size image

Lipid molecules on the A2AR surface

After the refinement of the A2AR structure, the electron density maps showed elongated extra electron densities surrounding the protein, which are usually assigned to lipid molecules. The locations of the extra electron densities observed in this work did not significantly depend on the sample temperatures. Moreover, the positions of the extra electron densities on the present A2AR surface nearly overlapped with those of the electron densities corresponding to lipids in the MO-matrix based 4EIY model, suggesting that the lipid binding positions remain nearly unaltered in the two different matrix lipid environments.

Although the majority of the extra electron densities are rather obscure, we noted that five extra electron densities on the extracellular half of the A2AR surface can be assigned to two different lipid species, cholesterol and EROCOC17+4. The cholesterols are probably derived from the LCP crystallization matrix, whereas the possibility that it is derived from insect cell membrane cannot be ruled out. As shown in Fig. 3A, there are three cholesterol molecules, C1, C2, and C3, at the same locations as those of the three cholesterol molecules identified in the 4EIY model. Two EROCOC17+4 molecules, E1 and E2, are also depicted in Fig. 3A.

The electron density for E1, which is located near cholesterol C1, is well defined for the full EROCOC17+4 molecule: an isoprenoid chain with four methyl branches at the 5th, 9th, 13th, and 17th carbons and a hydrophilic erythritol connected via an ester bond (Fig. 3B and Sup. Figure 1A). Since IPCLs are not native lipids in the insect cells where the A2AR proteins were produced, the electron density for E1 can be unambiguously modeled as EROCOC17+4. Another EROCOC17+4 modeled with relatively clear electron densities for the methyl branches (E2 in Fig. 3A,C and Sup. Figure 1B) exists between helices I and VII. E2 may stabilize the N-terminal segment of helix I, as proposed for the corresponding lipid (OLA11) in the 4EIY model16. It is interesting that this lipid binding site corresponds to the antagonist/ligand binding site of the prostaglandin E receptor EP4, which may be on the ligand entry pathway23. In the 4EIY model, the lipids corresponding to E1 and E2 are MO and oleic acid, respectively. The electron densities for the three terminal carbons of the MO and the two terminal carbons of the oleic acid are weak, possibly due to the flexibility of the terminal regions of the oleic lipids.

LCP-SFX data collection at 4 and 20 °C sample temperatures

As described in the preceding section, the reliability factors of the data were improved by lowering the sample temperature. Table 1 also shows that the LCP-SFX data collection statistics depended on the sample temperature. For instance, the “crystal hit rate” (Number of hit images/Number of collected images) × 100 was as high as 46.0% for the LCP-SFX_4 °C, while it dropped to 20.2% for the LCP-SFX_20 °C, where the hit image is defined as an image with 20 or more diffraction spots, among all images collected at a 30 Hz frequency. The “% indexed images” (Number of indexed images/Number of collected images) × 1004 likewise displayed a similar trend, 32.8% and 18.5% for the LCP-SFX_4 °C and the LCP-SFX_20 °C, respectively. As the present LCP-SFX experiments were performed with the identical A2AR crystal-dispersed LCP, and except for the sample temperatures, were performed under the identical LCP-SFX experimental conditions (e.g., the injector nozzle diameter, the XFEL beam size, the pulse intensity, etc.), these results were most presumably ascribable to the different sample temperatures.

Table 2 lists the D4 °C/D20 °C ratios estimated from the real-time LCP stream images, in the absence and presence of the XFEL beam, where D4 °C and D20 °C stand for the EROCOC17+4-LCP stream diameters at 4 and 20 °C at the XFEL beam intersection position, respectively. For reference, individual values of D4 °C and D20 °C, are also listed. These values were estimated by assuming that the average LCP stream diameter at 20 °C in the absence of the XFEL beam was equal to the injector nozzle diameter d; i.e., D20 °C (beam off)average ≡ 75 μm. It is first noted that the ratio in the absence of the beam, D4 °C (beam off)/D4 °C (beam off) ≈1.3. This suggests that the rheological properties of the EROCOC17+4-LCP are temperature dependent. In fact, we noted that the EROCOC17+4-LCP stream at 20 °C displayed “viscous fluid-like” behavior, whereas at 4 °C, “stiff or solid-like” features became conspicuous. This appears consistent with our empirical observation that the stress required to homogenize the EROCOC17+4-LCP progressively increases as the temperature is lowered. Moreover, the ratio in the presence of the beam, D4 °C (beam on)/D20 °C (beam on) ≈ 1.8, predicts that the “crystal hit rate” at 4 °C will be doubled as compared to that at 20 °C, in line with the observed “crystal hit rate”. This ratio also predicts the increased probability of multiple hits at 4 °C. The data in support of this prediction are actually seen in the “crystal index rate” (Number of indexed images/Number of hit images) × 100, with values of 71.3% at 4 °C and 91.5% at 20 °C.

Table 2 The LCP stream diameter ratio D4 °C/D20 °C in the absence and presence of the XFEL beam.
Full size table

Although the LCP stream was smooth and continuous in the absence of the XFEL beam, the beam-LCP stream intersection caused the LCP stream to exhibit quite dynamic and temperature-dependent behaviors in the presence of the XFEL beam. At 20 °C, the “viscous fluid-like” LCP stream displayed frequent fission and blowing-off of the stream at regular intervals, without significantly altering the D20 °C (beam on) as compared to D20 °C (beam off). On the other hand, at 4 °C the fission/blowing-off frequency of the “stiff or solid-like” LCP stream was reduced and occurred less regularly, tending to form an LCP lump with a larger D4 °C (beam on) as compared to D4 °C (beam off). Thus, the temperature-dependent LCP-SFX data collection statistics were most presumably ascribable to the modified response of the “viscous fluid-like” versus the “stiff or solid-like” LCP stream to the beam intersection.

To summarize, the LCP-SFX data collection at the 4 °C sample temperature could be performed with the higher “% indexed images” (32.8%) as compared to that at the 20 °C sample temperature (18.5%), and yielded more reliable diffraction data. The signal-to-noise ratio was also improved for the LCP-SFX_4 °C as compared to that for the LCP-SFX_20 °C; i.e., 8.3 and 7.6 at 4 and 20 °C, respectively, suggesting that the enhanced signal at 4 °C outweighed the presumed increase in the lipid-derived background diffraction from the LCP-stream with an increased D value. Therefore, the use of low temperature LCP samples could be a useful option for the LCP-SFX data collection.

EROCOC17+4 and the EROCOC17+4-matrix are resistant to adopting the crystalline LC phase

Lipid/water systems generally exhibit a transition from a high temperature liquid crystalline to a low temperature crystalline LC phase at a lipid-specific temperature, TK. Above TK, the lipid chains are in a fluid-like disordered conformation (a disordered chain state), with X-ray diffraction profiles characterized by a broad band around 4.6 Å, similar to a band observed from liquid paraffines24. Below TK, the lipid adopts a crystalline LC phase, where the chains are fully extended and aligned parallel to each other, which is characterized by multiple sharp diffractions at all spacings from the small angle X-ray scattering (SAXS) to wide angle X-ray scattering (WAXS) regimes, reflecting the presence of long- and short-range three-dimensional orders in the molecular packing24.

A typical example of an X-ray diffraction profile from an LC phase is shown in Fig. 4A-(3), where the profile from the 60% (w/w) MO/water system at 1 °C is plotted. Note that the equilibrium MO/water system adopts the crystalline LC phase below 17 °C7. Six representative peaks in the larger \(q\) region are denoted by black dots in the profile and white arrows in the 2D image. The three sharp diffractions in the SAXS regime, denoted by the green dots in the profile and green arrows in the 2D-image in Fig. 4A-(3), are due to the 2nd, 3rd, and 4th diffractions of the MO-LC phase with a lattice constant of 49.6 Å, which reasonably agrees with the reported value of 49.2 Å25.

Figure 4
figure 4

(A) X-ray diffraction profiles of 62.5% (w/w) EROCOC17+4/water (1) and dry EROCOC17+4 (2) at − 20 °C, together with the 2D-diffraction images used to construct each diffraction profile. The white and red arrows denote diffractions from disordered chains on the lipid and from ice, respectively. Profile (3) is from the MO-LC phase [60% (w/w) MO/water] at 1 °C. The humps observed in profiles (1) and (2) below q ~ 0.35 Å−1 are due to x-ray leaks around the beam stopper. \(q=\left(4\pi sin\theta \right)/\lambda \), where 2θ is the scattering angle and the wavelength λ = 1.54 Å. (B) (1) A diffraction image of the EROCOC17+4-matrix, 60% 9:1 (w/w) EROCOC17+4/cholesterol/water, at − 180 °C (measured under an evaporated liquid nitrogen flow). The broad band around 4.6 Å (white arrow) indicates that EROCOC17+4 remained in a disordered chain state at − 180 °C. The four red arrows denote hexagonal ice diffractions. (2) A diffraction image from the A2AR crystallization EROCOC17+4-matrix during the LCP-SFX_4°C data collection in a helium gas environment.

Full size image

Before discussing the phase behavior of the EROCOC17+4/water system at − 20 and around − 40 ~  to − 60 °C, let us briefly outline the phase behavior of the EROCOC17+4/water system above 0 °C10. The EROCOC17+4-LCP is identified in the lipid concentration \({W}_{Lipid}(=100-{W}_{Water})\) range from 60 to 80% (w/w) and over a temperature range from ca. 0 to ca. 55 °C [10, Sup. Table 1]. Thus, \({W}_{Lipid}\) = 60% (w/w) represents the water saturated EROCOC17+4-LCP composition. Once \({W}_{Lipid}\) is decreased below 60% (w/w), the excess water that can no longer be incorporated within the LCP separates out to adopt the “excess water + water saturated LCP” two-phase coexisting state. In the \({W}_{Lipid}\) range from 80 to ca. 92% (w/w), the system adopts a lamellar (Lα) phase. In the highest \({W}_{Lipid}\) regime above ca. 93% (w/w), the phase structures have not yet been identified, due to the slow rate toward reaching the equilibrium state.

The X-ray diffraction profiles from the 62.5% (w/w) EROCOC17+4/water and dry EROCOC17+4 at − 20 °C are shown in Fig. 4A-(1) and (2), respectively. Both profiles are quite distinct from that of the MO-LC phase in Fig. 4A-(3). They are characterized by a broad band around 4.7 Å, [\(q(\equiv 2\pi /d)\) ~1.3 Å−1, white arrow in the 2D images], indicative of the disordered hydrocarbon chain of the lipid. The large peak at \(q\)= 1.62 Å−1 is the (100) diffraction of hexagonal ice (red arrows in the profile and the 2D image in Fig. 4A-(1))26, which disappeared above 0 °C (data not shown). No ice peak was observed in the dry EROCOC17+4 profile in Fig. 4A-(2), as expected. These results indicate that EROCOC17+4 remained in the disordered chain state (hence no LC phase formation) at least down to − 20 °C, over a wide range of lipid hydration values (\({W}_{Water}\)) from fully hydrated 40% down to nearly 0% (w/w) water.

The adoption of the disordered chain state well below 0 °C is further supported by a differential scanning calorimetry study of the EROCOC17+4/water system10, which describes the effort in developing the LC phase by cooling (ca. − 3 °C/min) the EROCOC17+4/water system from ca. 25 to − 60 °C, followed by a 3 h isothermal incubation at − 60 °C prior to the initiation of the heating run at a rate of 0.3 °C/min. As seen in the heating thermogram of 63.0% (w/w) EROCOC17+4 measured over a temperature range from − 40 to 15 °C (Sup. Figure 2A), only a single endothermic peak associated with melting of ice is present, indicating that EROCOC17+4 did not adopt the LC phase at least down to − 40 °C10.

Let us next examine a possible phase state of the EROCOC17+4-matrix at a cryogenic temperature. To avoid excessively rapid cooling, the EROCOC17+4-matrix was first cooled slowly (ca. − 2 °C/min) from room temperature to − 20 °C, followed by 2–3 days of isothermal incubation prior to the incubation in liquid nitrogen (see X-ray scattering measurements section for details). We examined the diffraction profiles of twelve EROCOC17+4-matrix specimens at − 180 °C and confirmed that all of the specimens gave disordered chain profiles, as seen in Fig. 4B-(1), where a typical diffraction image of the EROCOC17+4-matrix (without A2AR) at − 180 °C, measured after a one-week incubation in liquid nitrogen, is shown. It displayed a broad band profile around 4.6 Å (white arrow), indicating that the EROCOC17+4-matrix did not adopt a crystalline LC phase in liquid nitrogen, at least under the present experimental conditions. The four additional sharp diffractions at 3.87, 3.70, 3.44, and 2.66 Å (red arrows) are hexagonal ice diffractions26.

The possible phase structure of the EROCOC17+4/water system below 0 °C was inferred from the SAXS profiles of the 62.5% (w/w) EROCOC17+4 at − 20 °C and the 63% (w/w) EROCOC17+4 at − 60 °C. They both gave two peaks in a 2:1 ratio, 44.5 and 22.5 Å at − 20 °C and 46.8 and 23.4 Å at − 60 °C, respectively, suggesting the formation of an Lα phase with lattice constants of \({\xi }_{{L}_{\alpha }}\left(-20^\circ C\right)\) = 44.5 Å and \({\xi }_{{L}_{\alpha }}\left(-60^\circ C\right)\) = 46.8 Å, respectively. At − 180 °C, the four weak diffractions at 16.2, 12.2, 9.7, and 8.1 Å, observed in the SAXS region of Fig. 4B-(1), could also be compared with the 3rd, 4th, 5th and 6th diffractions of the Lα phase with the lattice constant of \({\xi }_{La}\left(-180^\circ C\right)\) = 48.0 Å. These results indicate that the EROCOC17+4/water and the EROCOC17+4-matrix/water systems most presumably adopted the Lα phase below 0 °C. For comparison, MO adopts the LC phase at 1 °C (Fig. 4A-(3)), and eleven MO-matrixes subjected to the same cryo-cooling procedure displayed strong LC diffractions at − 180 °C without exception, as shown in Sup. Figure 2.

It is therefore clear that EROCOC17+4 is far more resistant to the adaptation of the LC phase as compared to MO. This allowed us to readily perform the LCP-SFX data collection at a 4 °C sample temperature in a helium gas atmosphere without any LC phase formation. Figure 4B-(2) shows a typical in situ X-ray diffraction image of the EROCOC17+4-matrix captured during the LCP-SFX_4 °C data collection, in which the broad diffraction around 4.6 Å (white arrow) indicative of a fluid-like disordered chain state was seen, in addition to the diffraction spots from the A2AR crystals.

EROCOC17+4 as a possible LC formation-free crystal delivery medium

In LCP-SFX, the viscous media injector streams the microcrystal-dispersed LCP inside the diffraction chamber, in either vacuum or helium gas conditions. In the case of a vacuum chamber at 20 °C, the MAG-matrix must be prepared from a short chain MAG, such as 9.7 MAG or 7.9 MAG, which are less prone to LC phase formation than MO3,4,6; for instance, TK of 7.9 MAG is 6 °C27. For crystals that only grow in the MO-LCP, it must be doped with one of the short chain MAGs just before loading into the injector to avoid the LCP → LC phase transition upon evaporative cooling3,4.

Thus, the short chain 9.7 MAG and 7.9 MAG with the acyl chains of 16 carbon atoms-long serve as an LC formation-free crystal delivery medium, whereas MO (9.9 MAG with the acyl chain of 18 carbon atoms-long, which yielded the most successful outcomes for a variety of MPs2, Fig. 1) does not. However, the measurements without extra matrix manipulation just before loading are more favorable. Moreover, it is well known that crystal growth and quality critically rely on the hydrocarbon chain length of the matrix lipid; e.g., according to Li et al., “a chain length disparity of just one carbon (in the 14- vs 15-carbon homologues) meant either getting or not getting crystals of outer membrane sugar transporter”28. Thus, the short chain MAGs may not be applicable to all MPs. The lipid with the acyl chain of 18 carbon atoms-long that serves as an LC formation-free crystal delivery medium is useful.

It is often stated rather vaguely that ICPLs remain in the fluid state at low temperatures. However, as shown in Sup. Table 1, the TK values (the lower temperature limit of the fluid state) and the temperature range over which the LCP is stable are lipid dependent. This is also true for the maximum hydration level of LCP (\({W}_{Water}^{max}\)) that determines the maximum protein buffer loading capacity of the LCP-matrix. Thus, IPCLs with these parameters favorable for the crystallization and crystal delivery medium should be explored.

EROCOC17+4 is akin to MO, in the sense that they both contain the acyl chain of 18 carbon atoms-long (Fig. 1), and the \({W}_{Water}^{max}\) value of 40% (w/w) (Sup. Table 1) is comparable to that of MO. Moreover, EROCOC17+4 forms the stable LCP over a temperature range from ca. 0 to ca. 55 °C [10, Sup. Table 1], and remains in the fluid state at least down to − 20 °C, over a wide range of hydration levels, \({W}_{Water}\) from 40% (w/w) to nearly 0% (w/w) dry state. Thus, EROCOC17+4 represents the first lipid with the acyl chain of 18 carbon atoms-long that meets the requirements for a matrix for MPs crystallization and an LC formation-free crystal delivery medium without extra matrix manipulations.

Concluding remarks

This study highlights the use of EROCOC17+4 as a new matrix lipid for A2AR crystallization and an A2AR crystal delivery medium in LCP-SFX. The results are summarized as follows.

  1. (1)

    The EROCOC17+4-matrix supports the growth of X-ray diffraction quality A2AR crystals and serves as a crystal delivery medium in LCP-SFX.

  2. (2)

    The optimum crystallization conditions for A2AR crystals in the EROCOC17+4-matrix did not differ significantly from those in the MO-matrix, and the EROCOC17+4-matrix could be manipulated by essentially the same experimental procedures as employed for the MAG-matrices. Thus, EROCOC17+4 can readily be accommodated within the current in meso crystallography experiments.

  3. (3)

    One noteworthy feature of EROCOC17+4 and the EROCOC17+4-matrix is the resistance to the adoption of the LC phase. This makes it valuable for use in lower temperature crystallizations of MPs. In preliminary crystallization trials at 4 °C, we confirmed that the EROCOC17+4-matrix yielded A2AR crystals that diffracted to ca. 3 Å (not optimized), suggesting that EROCOC17+4 has potential for the in meso crystallization of thermally unstable MPs.

  4. (4)

    In LCP-SFX, EROCOC17+4 is the first lipid with the acyl chain of 18 carbon atoms-long that can serve as an LC formation-free crystal delivery medium without extra matrix manipulations for low temperature data collection, and probably under vacuum chamber conditions as well. These features appear best suited for the lower temperature data collection of thermally unstable MP crystals, for which improved LCP-SFX diffraction data may also be expected.

  5. (5)

    As EROCOC17+4 is a member of the LCP-forming IPCL family10,11,12,13,14 (Sup. Table 1), IPCLs together with MAGs will expand the reliability/efficiency of in meso crystallization and serial microcrystallography. More details about the phase behavior and physical characteristics of EROCOC17+4 will be published elsewhere.

Methods

Expression and purification of A2AR

A2AR harboring a bRIL, apocytochrome b562RIL from Escherichia coli (M7W, H102I, and R106L) fused in the third intracellular loop (ICL3), was prepared according to the method of Liu et al.16. Briefly, A2AR was expressed in Sf9 insect cells, using a baculovirus expression system. The A2AR was solubilized from the membrane fraction in 1% n-dodecyl-β-D-maltopyranoside (DDM, Anatrace) and 0.2% cholesteryl hemisuccinate (CHS, Sigma), with 4 mM theophylline (Sigma). The theophylline was substituted with 25 µM ZM241385 (Tocris and Sigma-Aldrich) on TALON cobalt-chelating resin (Clontech), in 20 mM HEPES buffer (pH 7.5) including 800 mM NaCl, 15 mM imidazole, 0.02% DDM, 0.004% CHS, and 10% glycerol, during affinity purification. The purified protein, in 20 mM HEPES buffer (pH 7.5) including 800 mM NaCl, 250 mM imidazole, 0.02% DDM, 0.004% CHS, 10% glycerol, and 25 µM ZM241385, was concentrated to 60 mg mL-1 (Amicon Ultra-4 100 kDa MWCO, Merck). The purity and monodispersity of the samples were assessed by analytical size exclusion chromatography (Sup. Figure 3).

Crystallization

We performed ‘‘standard’’ in meso crystallization trials2,29 at 20 °C, with a 9:1 (w/w) mixture of EROCOC17+4 and cholesterol as the crystallization matrix. As described in Fig. 1, EROCOC17+4 is a mixture of 92% 1-O- and 8% 2-O-isomers. Typically, 9 mg of A2AR in buffer was homogenized with 13.5 mg EROCOC17+4-matrix [A2AR: EROCOC17+4-matrix = 40:60 ratio (w/w)] by a home-made mixing device, to obtain the viscous and transparent A2AR-LCP30.

A 40 nL portion of the homogenized A2AR-LCP was dispensed onto a siliconized glass plate with a 96-hole double stick sheet with a 135 µm thick spacer (Kajixx Co., Kawasaki, Japan)30, using a home-build robotic dispenser (Hato et al., to be published). The dispensed A2AR-LCP boluses were then covered with a glass coverslip after the addition of a series of 1 μL crystallization solutions. The 96-well glass plates were incubated at 20 °C in a RockImager 1000 (FORMULATRIX). The long axes of 10 crystals for the 30 and 32.5% PEG conditions (< 5 μm in size), 30 crystals for the 35% and 37.5% PEG conditions (~ 10 μm in size), and 50 crystals for the 40% PEG conditions (10 ~ 15 μm in size) were measured manually in bright field images (Sup. Figure 4), and the averages and standard deviations were determined. Crystal densities were also estimated by manually counting the separated and bright spots observed within a selected region of the whole (~ 40 nL) LCP bolus of cross polarized images: the whole area for the 40% PEG conditions, ~ 1/4 of the area for the 32.5–37.5% PEG conditions, and ~ 1/8 of the area for the 30% PEG conditions of corresponding LCP boluses (Sup. Figure 4). As the density values inevitably have large experimental uncertainties, they should be considered as rough estimates. The crystals for LCP-SFX were produced as described in the reference31, except that Ito MS-GAN025 (250 μL) microsyringes were used without mixing with any other lipid before the LCP-SFX measurement. As judged from the diffraction ability, the crystals were stable between 4–25 °C for at least a few weeks.

Serial synchrotron rotation crystallography (SS-ROX)

SS-ROX data were collected at 100 K at SPring-8 BL32XU32. After breaking the cover glass with a glass cutter29, small crystals (~ 25 µm) in LCP were harvested with 200 µm cryo-loops (MiTeGen) and flash-cooled without additional cryo-protectant. Using a 5 × 5 µm2 beam, an entire cryo-loop was scanned with 0.5° rotation and 0.45 s per image with a photon flux of 6.7 × 1011 photons/s at a wavelength of 1.0000 Å, using an EIGER X 9 M detector (DECTRIS)20. The experiment was performed automatically by the ZOO system33. Fourteen cryo-loops provided 6346 hits with more than 20 spots in an image, identified using the Cheetah program34 adapted for EIGER, and 3701 hits were indexed using kamo.single_images_integration20 with XDS35. Among the indexed images, 2100 images were selected based on the highest average I/σ(I). Intensities were merged using kamo.merge_single_images_integrated20 in a Monte-Carlo fashion, as in CrystFEL36, by discarding low-partiality reflections.

Lipidic cubic phase serial femtosecond crystallography (LCP-SFX)

LCP-SFX data were collected using the 20 and 4 °C samples at SACLA37 BL338 EH4c (LCP-SFX_4 °C and _20 °C, Table 1). For each measurement, a ~ 25 µL A2AR-LCP sample was transferred into a dedicated sample injector with a 75 µm ϕ nozzle. The sample temperature was controlled by a temperature variable liquid circulator (OHM Electric Co., Ltd.) connected to the sample injector. For the data collection using the 4 °C sample, the sample injector and the circulating hose were covered with urethane and rubber jackets, respectively, to maintain the temperature and prevent condensation. The sample was injected at a rate of 0.24 µL min−1 into a 30 Hz X-ray laser pulse, to achieve a 30 µm inter-pulse distance at the injected sample. To achieve a smooth flow, a room temperature helium sheath gas (0.6 L min−1) was applied around the injected sample, and aspirated (0.8 L min−1) just below the sample. The X-ray energy was 10.1 keV (1.23 Å wavelength), the beam intensity was ~ 560 µJ per pulse, and the data were collected with a multiport CCD39, Phase III, with improved quantum efficiency at a higher energy X-ray region around 12 keV with a thicker (300 µm) sensor. Less than 100 min were needed to obtain a complete data set with the 25 µL A2AR-LCP samples (31,312 and 46,877 indexed images for 20 and 4 °C, respectively). Data collection was guided by the data processing pipeline for SACLA40, based on Cheetah34 and CrystFEL 0.6.236. Images with 20 or more spots detected by Cheetah were considered as hits and processed by CrystFEL. DirAx 1.1641 was used for indexing. Integrated intensities were merged by process_hkl in the CrystFEL suite, with linear scaling and the per-image resolution cutoff (the “–push-res 1.5” option).

Structural refinement

The crystal structure of MO-matrix-grown A2AR (PDB ID: 4EIY) was employed as the starting model. Models were rebuilt by iterative manual modifications in COOT42 and refinement by REFMAC543 in CCP444. Structural figures were prepared with PyMOL (Schrödinger, L. L. C. The PyMOL Molecular Graphics System. Version 1.8). The stereochemistry of the protein structures was checked by Rampage45, and no amino acid residues were in the disallowed region of the Ramachandran plot. Cholesterol and EROCOC17+4 models were introduced where the shape and size of the electron density fit well. For the EROCOC17+4 stereochemistry, (2R,3S) was applied to the erythritol moiety, and the 5,9,13,17-tetramethyloctadecanoyl moiety was modeled as the All R (5R,9R,13R) conformation, as in β-XylOC16+415. Linear hydrocarbons were modeled for elongated electron densities that could not be characterized well as a specific lipid. Polder maps46, which exclude the bulk solvent around the individually omitted lipids E1, E2, C1, C2 and C3, were prepared with Phenix47.

X-ray scattering measurements

The X-ray scattering measurements of dry EROCOC17+4, 62.5% (w/w) EROCOC17+4/water at − 20 °C, and 60% (w/w) MO/water at 1 °C were performed with a Rigaku Nano-Viewer, using Ni-filtered CuKα radiation (λ = 1.54 Å) generated by a Rigaku FR-E unit (45 kV, 45 mA) with a triple pinhole collimator (0.4 mmϕ × 0.2 mmϕ × 0.45 mmϕ). Sample-to-detector distances of 92 mm (WAXS) or 448 mm (SAXS), were calibrated by using lead stearate48 and stearic acid49, respectively. The sample temperature was controlled with an UT4040-PF Peltier module (Ampère, Tokyo) at − 20 ± 0.5 °C and 1 ± 0.1 °C, monitored by a Pt-100 thermoresistor (Hayashi Denko, Pt-100-A-1-M, Tokyo). The two-dimensional powder diffractions were recorded by a PILATUS100K-S detector (DECTRIS, Switzerland) and analyzed by a Rigaku 2-DP data processing system. The data collection of 63.0% (w/w) EROCOC17+4/water at −60 °C was performed with Ni-filtered CuKα radiation (λ = 1.54 Å) from a Rigaku RU-200 X-ray generator (40 kV, 100 mA) with a double pinhole collimator (0.5 mmϕ × 0.3 mmϕ), and the sample temperature was controlled with a Mettler FP82HT temperature control-stage10,11.

The samples (except for dry lipid) were homogenized by using a homebuilt mixing device consisting of a pair of MS-GAN025 (250 μl) microsyringes (Ito Co., Fuji, Japan) with a monolithic stainless-steel coupler10,30. The homogenized sample was then transferred into quartz capillaries (1.5 mm in diameter, Glass Berlin) and immediately flame sealed and glued with 5-min epoxy (Konishi Co., Ltd. Osaka). To fully develop a sub-zero degree temperature phase at − 20 °C (the Lα phase), the EROCOC17+4 samples were first cooled (ca. − 2 °C/min) from room temperature to − 20 °C, followed by one day of isothermal incubation at − 20 °C (Medicool, Sanyo Electric Co., Ltd., Osaka) prior to the − 20 °C data collection. The − 20 °C isothermal incubation for up to three weeks did not alter the results. The MO sample was first incubated at − 20 °C to fully develop the LC phase, and then the sample temperature was increased to 1 °C prior to the data collection at 1 °C.

The X-ray scattering profiles at − 180 °C of the EROCOC17+4- and MO-matrixes, 60% 9:1 (w/w) EROCOC17+4/cholesterol/water and 60% 9:1 (w/w) MO/cholesterol/water, were measured by a Rigaku FR-X unit (45 kV, 66 mA, CuKα radiation λ = 1.54 Å) with a sample to PILATUS200K detector distance of 70 mm, calibrated by using lysozyme crystals. The image data were analyzed by the Rigaku Crystal Clear data processing system.

The EROCOC17+4- and MO-matrixes in a 200 µm cryo-loop (MiTeGen) were placed in a hermetically sealed plastic container, which contained a small amount of water at the bottom to ensure that the matrix specimens were maintained in the water saturated phase condition. To avoid excessively rapid cooling, the matrix specimens were first cooled (ca. − 2 °C/min) from room temperature to − 20 °C, followed by 2–3 days of − 20 °C isothermal incubation to fully develop a sub-zero degree temperature phase (the Lα phase), and were then incubated in liquid nitrogen. The − 180 °C measurements were performed under an evaporated liquid nitrogen flow, after 1- and 7-day incubation durations in liquid nitrogen. Both diffraction profiles were indistinguishable.