The discovery of a new antibody for BRIL-fused GPCR structure determination

G-protein-coupled receptors (GPCRs)—the largest family of cell-surface membrane proteins—mediate the intracellular signal transduction of many external ligands. Thus, GPCRs have become important drug targets. X-ray crystal structures of GPCRs are very useful for structure-based drug design (SBDD). Herein, we produced a new antibody (SRP2070) targeting the thermostabilised apocytochrome b562 from Escherichia coli M7W/H102I/R106L (BRIL). We found that a fragment of this antibody (SRP2070Fab) facilitated the crystallisation of the BRIL-tagged, ligand bound GPCRs, 5HT1B and AT2R. Furthermore, the electron densities of the ligands were resolved, suggesting that SPR2070Fab is versatile and adaptable for GPCR SBDD. We anticipate that this new tool will significantly accelerate structure determination of other GPCRs and the design of small molecular drugs targeting them.

Scientific RepoRtS | (2020) 10:11669 | https://doi.org/10.1038/s41598-020-68355-x www.nature.com/scientificreports/ receptor. However, screening mutation sites within the target GPCR is still necessary to determine those that stabilise the active or inactive state before the construct can be used for crystallisation experiments. As this approach is also very strenuous and time consuming, there is an urgent need for optimisation methods that are more efficient and can be widely applied to various GPCR targets for crystallography.
Here, we demonstrate that antibodies against fusion proteins can enhance the crystallisation of GPCRs with ligands in active or inactive conformational states. Based on the results presented herein, this approach may be widely applicable for various GPCRs, which would vastly streamline drug development efforts.
Structure determination of 5HT 1B -BRIL/ERG/SRP2070Fab and AT 2 R-BRIL/s-Ang II/SRP-2070Fab. The structures of both complexes were solved by molecular replacement and then refined. Following refinement, the R work /R free was 22.2%/28.0% for 5HT 1B -BRIL/ERG/SRP2070 and 22.6%/27.8% for AT 2 R-BRIL/s-Ang II/SRP2070Fab (Table 1). Figure 1 shows the overall structures of both complexes. SRP-2070Fab perpendicularly bound to the BRIL helices. We also observed that helices III and IV of BRIL served as epitopes for SRP2070Fab. In particular, helix III was recognised by both CDR-L3 and CDR-H3, suggesting that Table 1. Data collection and refinement statistics. Data were collected from 144 and 118 crystals of AT 2 R-SRP2070Fab and 5HT 1B -SRP2070Fab complexes, respectively. a Values in parentheses are for the highestresolution shell.   (Fig. 2). These findings indicate that SRP2070 recognises the tertiary structure of BRIL, which may be an important feature of SRP2070Fab as a crystallisation enhancer.
To assess the effect of SRP2070Fab on crystallisation, the crystal packing interactions were investigated. As shown in Fig. 3, crystals of 5HT 1B -BRIL/ERG/SRP2070Fab and AT 2 R-BRIL/s-Ang II/SRP2070Fab were formed by alternative stacking of the aqueous layers composed of SRP2070Fab and BRIL and the lipophilic layers composed of the receptor region. Crystal packing contacts were observed in the aqueous layers between SRP2070Fab molecules and also within the receptor region in the lipophilic layers. The crystal packing calculations for Overall structure of 5HT 1B -BRIL/ERG/SRP2070Fab. ERG is shown as sticks. Each molecule is coloured as follows: 5HT 1B (gray), BRIL (light brown), SRP2070Fab (yellow). (B) Overall structure of AT 2 R-BRIL/s-Ang II/SRP2070Fab. s-Ang II is shown as sticks. Each molecule is coloured as follows: AT 2 R (purple), BRIL (light brown), SRP2070Fab (yellow).   Fig. 5S). When the crystal packing contribution resulting from the head-to-tail dimerisation was excluded, it was clear that SRP2070Fab was the most dominant crystallisation promoter. These were also observed for AT 2 R-BRIL/s-Ang II/SRP2070Fab. The buried ASAs were 1,329.7 Å 2 (AT 2 R), 510.1 Å 2 (BRIL), and 2,293.8 Å 2 (SRP2070Fab), suggesting that the crystal contacts among SRP2070Fab are the most dominant factor for the crystal formation. Unambiguous electron densities corresponding to the ligands, ERG for 5HT 1B and s-Ang II peptide for AT 2 R, were also observed within the orthosteric binding pockets of the receptors (Fig. 4A,B).

Structure comparison.
To confirm whether the structures of GPCR portion are affected by SRP2070 binding to BRIL, the structures of 5HT 1B -BRIL/ERG/SRP2070Fab and AT 2 R-BRIL/s-Ang II/SRP2070Fab were compared to the previously reported structures of 5HT 1B -BRIL/ERG (PDB ID: 4IAR) 17 and AT 2 R-BRIL/s-Ang II in complex with an anti-AT 2 R antibody (PDB ID: 5XJM) 18 . Superimpositions of these structures showed that SRP-2070Fab bound to BRIL had very little influence on the GPCR structures; i.e., the root mean square deviations (RMSDs) within the transmembrane regions of the GPCRs were 0.55 Å for 5HT 1B -BRIL/ERG and 0.67 Å for AT 2 R-BRIL/s-Ang II between equivalent Cα atoms. Furthermore, no significant differences were found around the ligand binding sites in both AT 2 R-BRIL/s-Ang II and 5HT 1B -BRIL/ERG as a result of SRP2070Fab binding (Fig. 4C,D). The three activation motifs-called microswitches-PIF 19 , NPxxY 20 , and DRY 20 , were also compared. It was confirmed that the three motifs showed very similar conformations in both comparisons (Supplementary Fig. 1S). Thus, the binding of SRP2070Fab to ICL BRIL fused GPCR has no significant effect on microswitch conformations. And also, it was clearly demonstrated that the activation motifs of AT 2 R-BRIL/s-Ang II in complex with an anti-AT 2 R antibody show an active state in the previous report 18 . Therefore, it is thought that AT 2 R-BRIL/s-Ang II/SRP2070Fab is also in an active form even though BRIL is fused to ICL3. The exam- www.nature.com/scientificreports/ ple of the structure that an agonist binds to the ICL3 BRIL fused GPCR has been already reported. The crystal structure of an active-state human angiotensin II (AngII) type I receptor (AT1R) bound to an AngII analog with partial agonist activity was determined even if BRIL was fused to ICL3 21 . Hence, it is possible that the structure of AT 2 R-BRIL/s-AngII/SRP2070Fab shows an active form. It means that the SRP2070Fab is applicable to active state GPCR structure determination. On the other hand, it is very difficult to state clearly whether the 5HT 1B -BRIL/ERG/SRP2070Fab and 5HT 1B -BRIL/ERG show an active or an inactive form. Although the orientation of PIF motif was coincide with that of 5HT 2B -BRIL structure in complex with ERG (PDB ID: 5TUD) 22 whose structure was clearly demonstrated as an active form by comparing with the structure of β 2 adrenergic receptor (β 2 AR)-Gs complex (PDB ID: 3SN6) 22 , the NPxxY and DRY motifs were different from 5HT 2B -BRIL/ ERG ( Supplementary Fig. 2SA), which means 5HT 1B -BRIL/ERG/SRP2070Fab structure does not show an active form. Furthermore, the three motifs of 5HT 1B -BRIL/ERG/SRP2070Fab structure were not coincide with those of www.nature.com/scientificreports/ β 2 AR bound to the partial inverse agonist carazolol (PDB ID: 2RH1) 23 that was known as an inactive form (Supplementary Fig. 2SB), which suggests that 5HT 1B -BRIL/ERG/SRP2070Fab structure is not a complete inactive form. However, the orientation of TM6 helixes of 5HT 1B -BRIL/ERG/SRP2070Fab showed more similar to that of carazolol bound β 2 AR than that of 5HT 2B -BRIL/ERG ( Supplementary Fig. 3S). Based on these comparisons, the 5HT 1B -BRIL/ERG/SRP2070Fab structure adopts an intermediate active state similar to that of the reported SRP2070Fab-free 5HT 1B -BRIL/ERG structure. This is especially apparent based on the orientation of this intracellular region of TM6, which adopts a conformation observed in the inactive structure rather than active 5HT 2B structure. Outward displacement of intracellular region of TM6 is known to be a structural hallmark of the inactive to active state transition, which results in formation of the cavity that recruits G proteins 13 . Thus, the orientation of fused-BRIL between TM5 and TM6 corresponding to the position of ICL3 is directly affected by the active or inactive state of the GPCR.

Discussion
In this study, we demonstrate that specific binders, such as antibodies, against soluble fusion partners of GPCRs can facilitate GPCR crystallisation. We discovered a new anti-BRIL antibody (SRP2070Fab), which promoted crystallisation of ICL BRIL-fused GPCRs. Three ICL BRIL-fused GPCR structures were solved in complex with SRP2070Fab. Two of them, 5HT 1B -BRIL with ERG and AT 2 R-BRIL with s-Ang II, are described in this study; the other, ChemR23-BRIL apo form whose structure has not been reported yet, will be reported in a future manuscript. These results suggest that SRP2070Fab is a versatile tool for promoting the crystallisation and structure determination of various ICL BRIL-fused GPCRs. Importantly, this method of using ICL BRIL-fused proteins in complex with SRP2070Fab could also be used to solve the structures of previously unsolved GPCRs.
Based on our results, SRP2070Fab appears to promote crystallisation by affecting crystal packing. For 5HT 1B -BRIL/ERG/SRP2070Fab, the overlapping layers of SRP2070Fab formed strong crystal packing by interacting at multiple sites (Fig. 3A). This was also observed in the crystal structure of AT 2 R-BRIL/s-Ang II SRP2070Fab (Fig. 3B). The crystal contacts between SRP2070Fab molecules were the most dominant contacts in both structures, suggesting that the range of the crystallisation conditions for ICL BRIL-fused GPCRs with SRP2070Fab is likely narrow. Thus, we predict that this system will increase the success rate of crystallisation and reduce the sample volume for crystallisation trials. Although we think the SRP2070Fab could be used for N-or C-terminal BRIL fused GPCR, further studies will be needed to demonstrate that.
Co-crystallisation with SRP2070Fab not only improves the success rate of crystallisation, but it is also advantageous for the structure determination. The phasing power of molecular replacement generally depends on the molecular mass of known search models (BRIL/SRP2070Fab in this study) relative to that of the protein of interest. Thus, the phasing power increases as a result of forming a complex with SRP2070Fab (~ 50 kDa) compared to performing phasing with BRIL (~ 12 kDa) alone. For both 5HT 1B -BRIL/ERG/SRP2070Fab and AT 2 R-BRIL/s-Ang II/SRP2070Fab, phasing by molecular replacement with BRIL/SRP2070Fab resulted in electron densities that were interpretable even in the GPCR portion.
To use the SRP2070Fab for structure-based drug design (SBDD), it is very important to confirm that the SRP2070Fab does not adversely affect the electron density of the ligand. As demonstrated in a previous study, antibodies recognising ICLs are not suitable for SBDD, because the electron density of the ligand tends to be unclear 24 . BRIL is often inserted in place of ICL3 to stabilise GPCRs 11 , and thus SRP2070Fab also binds to the intracellular portion. To determine whether SRP2070Fab is available for SBDD, the B-factor and the ligand density map of structures in this study were compared with the previously reported structures, 5HT 1B -BRIL/ ERG (4IAR) and AT 2 R-BRIL/s-Ang II in complex with anti AT 2 R antibody (5XJM) (Fig. 4). For 5HT 1B -BRIL/ ERG/SRP2070Fab, although the B-factor was higher than that observed in 5HT 1B -BRIL/ERG (4IAR) (Fig. 5A), the ligand electron density in the OMIT map 25,26 was clearly recognised and was similar to that of 5HT 1B -BRIL/ ERG (4IAR) (Fig. 4A). Regarding AT 2 R-BRIL/s-Ang II/SRP2070Fab, the B-factors showed a little bit higher in the GPCR part (Fig. 5B), and the ligand electron density in the OMIT map was as clear as that of AT 2 R-BRIL/s-Ang II in complex with anti AT 2 R antibody (5XJM) (Fig. 4B). Thus, the ligand electron density maps between the structures with SRP2070Fab (intra-cellular recognition) and the antibody recognising extra-cellular domain (5XJM) were in good agreement (Fig. 5C). The B-factor of the GPCR parts of our structures are higher than the previously reported structures (Figs. 5A,B). It is because the interaction between SRP2070Fabs has the important role in the crystal packing and the interaction between GPCRs is relatively weak as shown in Fig. 3. However, we think this property is necessary to the versatility of SRP2070Fab. In short, although the SRP2070Fab do not give effects on GPCR structures, it can improve the crystal packing.
Based on these results, SRP2070Fab does not appear to affect ligand binding, indicating that SRP2070Fab can be used for SBDD. Considering the versatility of SRP2070Fab, we anticipate that this antibody will be beneficial for performing SBDD applications. Because SRP2070Fab recognises only BRIL, it is thought that it has no effect on the GPCR; this would suggest that BRIL may absorb any structure fluctuations caused by binding of SRP2070Fab. While 5HT 1B /ERG and BRIL were linearly aligned through TM5 and TM6 and SRP2070Fab was vertically bound to BRIL, AT 2 R/s-Ang II and BRIL were unaligned by ~ 30°, and SRP2070Fab was vertically bound to BRIL (Fig. 1). This result suggests that there is some structural flexibility between GPCR and BRIL when SRP2070Fab functions as a crystallisation enhancer. Because of this property and the fact that SRP2070Fab is applicable to structure solution of 5HT 1B in intermediate active state, which is rather similar to inactive state structure, as well as active AT 2 R, it is reasonable to hypothesize that SRP2070Fab can be used to study both active and inactive states. This is another important feature of SRP2070 as is applicable to both of agonist and antagonist drug developments.
In recent years, the resolution of cryo-electron microscopy has improved, and several structures of GPCRs have been determined by this technique 27 . We anticipate that SRP2070Fab could also be used for cryo-EM Scientific RepoRtS | (2020) 10:11669 | https://doi.org/10.1038/s41598-020-68355-x www.nature.com/scientificreports/ analyses. By attaching SRP2070Fab to BRIL-fused GPCRs, the molecular weight can be increased and the SRP-2070Fab shape may be useful as a marker for 2D image classification and averaging. To our knowledge, the use of an antibody like SRP2070Fab to promote crystallisation has not been previously reported. This method may facilitate the structural study of many types of GPCRs. Considering the results described herein, SRP2070Fab is a promising structural biology tool for studying GPCRs and could be useful for developing drugs that target GPCRs.  Table 2. The mutations, L93V and F133W, were added to improve stabilisation. The additional N-and C-terminal residues retained after cleavage with tobacco etch virus (TEV) cleavage are italicised.
The subclass of the SRP2070 was a mouse IgG2a. The sequence of the variable region of the light chain is provided in Table 2; CDR-L1, CDR-L2, and CDR-L3 are underlined.
The sequence of the variable region of the heavy chain is provided in Table 2; CDR-H1, CDR-H2, and CDR-H3 are underlined.
Codon-optimised cDNA encoding H. sapiens AT 2 R-BRIL and 5HT 1B -BRIL was cloned into pFastBac1 (Thermo Fisher Scientific, Waltham, MA, USA). For AT 2 R-BRIL, the sequences encoding haemagglutinin (HA) and FLAG tags followed by a TEV cleavage site were added to the N-terminus, and a TEV cleavage site followed by a His 8 tag were added to the C-terminus. For 5HT 1B -BRIL, haemagglutinin (HA), FLAG, and His 8 tags followed by a TEV cleavage site were added to the N-terminus. culture and membrane preparation. AT 2 R-BRIL was expressed in Spodoptera frugiperda (Sf9) cells using the baculovirus-based expression system (cell density = 2.0-4.0 × 10 6 cells/mL; MOI = 0.5; expression time = 60 h). The culture was harvested by centrifugation (7,000 × g, 10 min, 4 °C), and cell pellets were flash frozen at − 80 °C. Frozen cells were thawed and suspended in the hypotonic buffer (10 mM HEPES pH 7.5, 20 mM KCl, 10 mM MgCl 2 , 1 × protein inhibitor cocktail tablet). The suspension was centrifuged, and the pellet was resuspended in high osmotic buffer [10 mM HEPES pH 7.5, 10 mM MgCl 2 , 20 mM KCl, 1 M NaCl, 1 × protease inhibitor cocktail (Sigma Aldrich, Saint Louis, MO, USA)]. The cells were lysed using a dounce homogeniser, and the lysate was centrifuged (100,000 × g, 30 min, 4 °C). The pellet was resuspended in high osmotic buffer containing 40% glycerol and stored at − 80 °C. Plasma membranes expressing 5HT 1B -BRIL were prepared using the same method as for AT 2 R-BRIL.

Production of monoclonal antibody fragment (SRP2070Fab).
Purified GPCR-BRIL proteins were reconstituted into liposomes consisting of 20:1 (w/w) egg phosphatidylcholine (Sigma-Aldrich)/monophosphoryl lipid A (Sigma-Aldrich). MRL/lpr mice were immunised with liposomal GPCR-BRIL. After immunisation, the mouse spleens were removed, and spleen cells were fused with myeloma cells. Hybridoma cells were screened by liposome-ELISA, denatured dot blot analysis, and fluorescence size exclusion chromatography (FSEC). To screen antibodies that specifically bind to the antigen with correct structure, but not the denatured one, liposome-ELISA positive and denatured dot blot clones were pooled, and then we confirmed the binding capacity in aqueous solution by FSEC. Hybridoma cells producing SRP2070 were intraperitoneally administered to mice, and ascites were collected. SRP2070 was purified by ammonium sulphate precipitation and protein G column chromatography. The subclass of SRP2070 was mouse IgG2a. After cleaving SRP2070 with papain, SRP2070Fab was purified by protein A column chromatography to remove the Fc portion and further purified by size exclusion chromatography (Superdex200 10/300 GL, GE Healthcare) by the changing buffer into PBS.  www.nature.com/scientificreports/ Purification of 5HT 1B -BRIL/ERG/SRP2070Fab. The construct for structural analysis was expressed in accordance with a previous study 17 . The amino acids from Leu240 to Met305 were replaced with BRIL. In addition, the glycosylation site was removed by deleting 32 amino acid residues at the N-terminus. The HA signal sequence, FLAG tag, His 10 -tag, and TEV protease recognition site were added. Furthermore, the L138W mutation was introduced to increase thermal stability. The DNA coding sequence was inserted into pFastBac1 using BamHI and HindIII.
The 5HT 1B -BRIL/ERG/SRP2070Fab complex was prepared by mixing 5HT 1B -BRIL/ERG and SRP2070Fab at a molar ration of 1:1.5 for 1 h on ice. The complex was purified by SEC (Superdex 200 increase 200, GE healthcare). Peak fractions containing 5HT 1B -BRIL/ERG/SRP2070Fab were collected and concentrated to 30 mg/mL using a 100-kDa MWCO Amicon concentrator and then used for crystallisation experiments.
Crystallisation and data collection of 5HT 1B -BRIL/ERG/SRP2070Fab complex. 5HT 1B -BRIL/ ERG/SRP2070Fab were reconstituted into LCP by mixing with host lipids (monoolein:cholesterol = 9:1) at a protein:lipid ratio (v:w) of 2:3 using a mixer consisting of two 100 μL gas tight syringes (Hamilton Company). Crystal trays (96 well glass sandwich plates), containing 30 nL LCP sample covered by 800 nL mother liquor per well, were set up using a Gryphon LCP crystallisation robot (Art Robbins Instruments) and then incubated at 20 °C. Droplets were imaged chronologically by RockImager 1,000 (FORMULATRIX, Bedford, MA). Crystals were obtained from precipitant conditions containing 0.4 M KSCN, 0.1 M NaOAc pH 5.5, and 30% PEG 400. Plate-shaped crystals grew to a maximum size of 30 × 20 × 5 μm 3 in 3-4 days, were harvested using MiTeGen MicroMounts (MiTeGen), and flash frozen in liquid nitrogen. X-ray data were collected at the BL32XU beamline at SPring-8 (Japan Synchrotron Radiation Research Institute). Most crystals diffracted to ~ 3.0 Å resolution (1.0 s exposure, 1.0° oscillation, 10 μm beam attenuated by a 500-1,200 μm aluminium sheet). To reduce radiation damage, crystals were exchanged after every 5°-10° of data collection. A complete data set at 3.0 Å resolution was obtained by merging the individual data sets collected from 144 crystals using KAMO 28 .
The initial structures were solved by molecular replacement as follows: as MR search models, the BRIL/ SRP2070 complex (solved in-house) structure was separated into BRIL/Fv region and Fc region, and the receptor region of 5HT 1B was clipped from 5HT 1B -BRIL structure (PDB ID: 4IAR). The MR search was performed with MOLREP 29 using three search models. Rigid-body and restrained refinements were performed with REFMAC5 30 . The ligand was placed in the electron density and the model was corrected with COOT 31 . The polder ligand omit map 26 was calculated with PHENIX 32 .
Purification of AT 2 R-BRIL/s-Ang II/SRP2070Fab. Residues 1-34 and 347-363 of human AT 2 R were deleted, and BRIL was inserted between N242 and K246 of AT 2 R. The HA signal sequence, FLAG tag, and TEV protease recognition site were added to the N-terminus, and TEV protease recognition site and His 8 tag were added to the C-terminus of AT 2 R. The membrane fraction purified from the Sf9 cell pellet was solubilised and incubated (1 h, 4 °C) with solubilisation buffer (50 mM HEPES pH 7.5, 800 mM NaCl, 10% glycerol, 1% DDM, 0.2% CHS, and 15 mM imidazole) supplemented with 100 mg/mL iodoacetamide (Wako Pure Chemical Industries, Ltd.) and 200 μM s-Ang II (Peptide Institute Inc.). After centrifugation (100,000 × g for 30 min, 4 °C), the supernatant was incubated with TALON Superflow Metal Affinity Resin (Clontech, 1-2 mL of resin per 500 mL of original culture volume was used) overnight at 4 °C with gentle agitation. After incubation, the resin was washed with twenty column volumes of wash buffer (50 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol, 0.1% DDM, 0.02% CHS, 15 mM imidazole, and 200 μM s-Ang II), then eluted by five column volumes of elution buffer (50 mM HEPES pH 7.5, 200 mM NaCl, 10% glycerol, 0.03% DDM, 0.006% CHS, 250 mM imidazole, and 200 μM s-Ang II). The eluted protein sample was concentrated using a 100-kDa MWCO Amicon concentrator, and the buffer was exchanged into wash buffer without imidazole.
The AT 2 R-BRIL/s-Ang II/SRP2070Fab complex was prepared by mixing AT 2 R-BRIL/s-Ang II and SRP-2070Fab at a molar ratio of 1:1.5 for 1 h on ice. His-tagged TEV protease was added to the sample and incubated 4 °C overnight to remove the N-terminal FLAG tag and C-terminal His 8 -tag. The His 8 -tagged TEV protease and cleaved C-terminal fragment were removed by incubating with 2 mL of Ni-NTA fast flow resin (GE Healthcare) at 4 °C for 1 h. AT 2 R-BRIL/s-Ang II/SRP2070Fab complex was collected as the Ni-NTA column flow-through fraction and concentrated using a 100-kDa MWCO Amicon concentrator. The final complex sample was purified by SEC (Superdex 200 10/300, GE Healthcare). Peak fractions containing AT 2 R-BRIL/s-Ang II/SRP2070Fab were collected and concentrated to ~ 30 mg/mL using a 100-kDa MWCO Amicon concentrator.
The initial structure was determined by molecular replacement with PHASER using BRIL/SRP2070Fv, SRP2070Fv, and the receptor region of AT 1 R-BRIL (PDB ID: 4ZUD) as search models. Rigid-body and restrained refinements were performed with REFMAC5 30 . The ligand was placed in the electron density and the model was corrected with COOT 31 . The polder ligand omit map 26 was calculated with PHENIX 32 .

Data availability
Atomic coordinates and structure factors are deposited in the Protein Data Bank under accession codes 7C61 for the 5HT 1B -BRIL/ERG/SRP2070Fab and 7C6A for the AT 2 R-BRIL/s-Ang II/SRP2070Fab. All of the data and information about constructs used in this study are available upon reasonable request from the corresponding author.