Determination of key residues in MRGPRX2 to enhance pseudo-allergic reactions induced by fluoroquinolones

MAS-related G protein-coupled receptor X2 (MRGPRX2), expressed in human mast cells, is associated with drug-induced pseudo-allergic reactions. Dogs are highly sensitive to the anaphylactoid reactions induced by certain drugs including fluoroquinolones. Recently, dog MRGPRX2 was identified as a functional ortholog of human MRGPRX2, with dog MRGPRX2 being particularly sensitive to fluoroquinolones. The aim of this study was to determine key residues responsible for the enhanced activity of fluoroquinolone-induced histamine release associated with MRGPRX2. Firstly, a structure model of human and dog MRGPRX2 was built by homology modeling, and docking simulations with fluoroquinolones were conducted. This model indicated that E164 and D184, conserved between human and dog, are essential for the binding to fluoroquinolones. In contrast, F78 (dog: Y) and M109 (dog: W) are unconserved residues, to which the species difference in fluoroquinolone sensitivity is attributable. Intracellular calcium mobilisation assay with human MRGPRX2 mutants, in which residues at positions 78 and 109 were substituted to those of dog MRGPRX2, revealed that M109 and F78 of human MRGPRX2 are crucial residues for enhancing the fluoroquinolone-induced histamine release. In conclusion, these key residues have important clinical implications for revealing the mechanisms and predicting the risks of fluoroquinolone-mediated pseudo-allergic reactions in humans.

www.nature.com/scientificreports/ the risks or investigate the mechanisms of drug-induced anaphylactoid reactions. Recently, we identified that dog MRGPRX2 is a functional ortholog of human MRGPRX2, with dog MRGPRX2 being highly sensitive to fluoroquinolones 20 . However, the mechanism underlying its enhanced activity in fluoroquinolone-induced histamine release has not been clarified. The present study was designed to identify key residues associated with fluoroquinolone-induced histamine release inducing the above-mentioned species difference in fluoroquinolone sensitivity. Firstly, a docking simulation with fluoroquinolones was conducted using a structure model of human and dog MRGPRX2 established by homology modeling. Comparison of amino acids around the predicted ligand binding pocket between human and dog indicated that two residues, F78 and M109, might play an important role in fluoroquinolone sensitivity. Then, HEK293 cells transfected with human MRGPRX2 mutants, in which amino acids at positions 78 and 109 were replaced by those in dog MRGPRX2 (F78Y, M109W, and F78Y/M109W), were treated with fluoroquinolones [ciprofloxacin (CPFX), gatifloxacin (GFLX), levofloxacin (LVFX), and pazufloxacin (PZFX)] and subjected to an intracellular calcium mobilisation assay, to assess Gq-coupled receptor activity. Determination of key residues associated with fluoroquinolone-induced histamine release would be clinically valuable for elucidating the mechanisms and predicting the risks of drug-induced pseudo-allergic reactions to fluoroquinolones in humans.

Results
Homology modeling of human MRGPRX2 and docking of fluoroquinolones. To perform docking simulations with fluoroquinolones, a homology model of human MRGPRX2 was built because no crystal structure information was available. As one of the closest homologs in the Protein Data Bank (https:// www. rcsb. org), human KOR (PDB accession id: 6B73, agonist-bound form) was used as a template for the modeling (Fig. 1a). Next, a docking simulation incorporating the induced fit effect with fluoroquinolones was performed. In this simulation, fluoroquinolones binding to human MRGPRX2 were simulated at the same site of the agonist in PDB 6B73. The top-ranked pose is shown in Fig. 1b. In the docking model with CPFX, a fluoroquinolone that induces histamine release by mast cells 14,21 , characteristic salt bridges were found between a basic substituent at the 7 position of CPFX and acidic side chains of E164 and D184 in human MRGPRX2. On the other hand, in the model with PZFX, a fluoroquinolone that does not induce histamine release by mast cells 22 , a side chain at the 7 position was located at a more distant location from E164 and D184 than in CPFX because of the difference in position of the terminal nitrogen between these two fluoroquinolones. Thus, E164 and D184 in human MRG-PRX2 were considered as essential residues involved in molecular interactions associated with the activation of MRGPRX2 by fluoroquinolones.
Amino acid sequence alignment and comparison of human and dog MRGPRX2. The result of aligning human and dog MRGPRX2 is shown in Fig. 2a. Typical motifs of class A GPCRs such as DRY motif and cysteine in transmembrane (TM)-3 were conserved in KOR, while they were not conserved in both human and dog MRGPRX2. E164 (dog: 270) and D184 (dog: 290), which were thought to be involved in key molecular interactions with fluoroquinolones as mentioned above, were shared between human and dog MRGPRX2. Next, a dog MRGPRX2 structure model was built by swapping amino acids of human MRGPRX2 constructed as described above. To identify key residues associated with a species difference, residues around the predicted binding pocket of fluoroquinolones were compared between human and dog MRGPRX2. Among the 21 residues within 5 Å around CPFX docked in human and dog MRGPRX2, only three amino acids (F78, M109, and A189 in human MRGPRX2) were unconserved residues between human and dog (Fig. 2b); the remaining residues (86%) were conserved. Among these three residues, F78 (dog: Y185) and M109 (dog: W215) in human MRGPRX2 were predicted to play some roles in the interaction with its ligands because the side chains of these two residues are oriented towards CPFX, and likely to affect it. Therefore, we selected the residues F78 and M109, located in TM2 and TM3, and constructed human MRGPRX2 mutants in which "dog-type" mutations were introduced (Fig. 2c).

Functional assay of human MRGPRX2 mutants.
To determine whether the dog-type human MRG-PRX2 mutants were associated with increased responses, intracellular calcium mobilisation against fluoroquinolones was evaluated. For this evaluation, three types of expression vector, F78Y and M109W (single mutations), and F78Y/M109W (double mutation), were constructed. HEK293 cells were transiently transfected with these human MRGPRX2 mutants and evaluated for reactivity to compound 48/80, a typical histamine releasing agent, and fluoroquinolones (CPFX, GFLX, LVFX, and PZFX). In the mutant F78Y, slightly increased reactivity was observed against treatment with CPFX and LVFX compared with the case for wild type (WT) human MRG-PRX2 (Fig. 3a). On the other hand, in the mutants M109W and F78Y/M109W, significantly increased responses to compound 48/80, CPFX, GFLX, and LVFX were observed from lower concentrations than for WT (Fig. 3a,b). In particular, the reactivity of the double mutant F78Y/M109W against CPFX was equal to or greater than that of dog MRGPRX2. The EC 50 of CPFX in F78Y/M109W was ca. one-eighth that of the WT ( Table 1). The order of the EC 50 values was compound 48/80 < CPFX < LVFX < GFLX, which was the same for all of the mutants. PZFX, which does not induce histamine release by mast cells 22 , did not induce intracellular calcium mobilisation in any of the cells.

Discussion
In the present study, we identified M109 and F78 in human MRGPRX2 as key residues associated with fluoroquinolone-induced histamine release inducing the difference in fluoroquinolone sensitivity between human and dog. Introducing dog-type mutations into human MRGPRX2 markedly enhanced responses to fluoroquinolones. Our results provide important insights into the mechanism behind fluoroquinolone-induced histamine www.nature.com/scientificreports/ release in dogs, which are highly sensitive to these drugs. Moreover, the results suggest the mechanisms behind MRGPRX2-related hypersensitivity in humans. Reddy et al. and Lansu et al. identified E164 and D184 of human MRGPRX2 as essential residues for the activation of MRGPRX2 by substance P and opioids 5,6 . These two amino acids are negatively charged residues that were found to interact with cationic opioid ligands 6 . Additionally, it was demonstrated that replacing E164 with a positively charged residue resulted in the loss of responses to substance P 6 . In contrast, the response to LL-37 or dynorphin was not lost even in the same mutant, suggesting that different ligands might interact with different amino acids around the predicted ligand-binding pocket 6 . In this study, our docking model suggested that E164 and D184 would be essential for MRGPRX2 activation associated with fluoroquinolones because of the characteristic interactions between CPFX and E164/D184. We previously reported that a basic substituent at the 7 position of the fluoroquinolone ring may be associated with histamine release 14 . Consistent with that report, a terminal basic nitrogen of piperazine at the 7 position of the fluoroquinolone ring, which is present in the three fluoroquinolones used in this study (CPFX, GFLX, and LVFX), was predicted to form interactions such as a salt bridge or a hydrogen bond with acidic side chains of E164 and D184. On the other hand, in PZFX, which does not induce histamine release by mast cells 22 , no salt bridge or hydrogen bond was formed because of the greater distance between the nitrogen and E164/D184. The difference between CPFX and PZFX in our MRGPRX2 www.nature.com/scientificreports/ docking model was consistent with their distinct histamine release potential in vivo/vitro 17,22 . However, these residues were not considered to be a key factor contributing to the enhanced activity to fluoroquinolones because they were conserved between human and dog. Human MRGPRX2 is considered as "an atypical opioid receptor" 5 ; many of the motifs conserved in class A GPCRs are not found in human and dog MRGPRX2. While the homology of amino acid sequences between human and dog MRGPRX2 was 62% 20 , residues around the predicted ligand-binding site were found to be well conserved; only three amino acids differed. In the functional assay with "dog-type" human MRGPRX2 mutants, M109W and F78W/M109W showed markedly increased responses compared with the WT. In particular, the double mutant F78Y/M109W appeared to demonstrate a greater response than dog MRGPRX2 against CPFX. These results clearly suggest that M109 (dog: W215) and F78 (dog: Y185) of human MRGPRX2 are key residues contributing to the enhanced responses to fluoroquinolones.
Introducing a mutation into a GPCR often affects its basal activity or ligand sensitivity 23,24 . In this study, human MRGPRX2 mutants did not exhibit changes of basal activity compared with WT in the calcium mobilisation assay, suggesting that these mutations did not affect the constitutive activity of MRGPRX2. In contrast, the mutants M109W and F78Y/M109W demonstrated a left shift of the dose-response curves and decreased In contrast, the EC 50 of LVFX was similar to or lower than that of GFLX in dog MRGPRX2 WT, suggesting that the ligand selectivity to these fluoroquinolones differs between human and dog MRGPRX2. Because the order did not change even in the mutants constructed in this study, the difference in ligand selectivity may arise from regions other than F78 and M109. Because the residues around the ligand binding pocket of MRGPRX2 are highly conserved between human and dog, comparison between human and dog MRGPRX2 including the site responsible for G-protein coupling, such as intracellular loops, may be needed to clarify the mechanism behind the different ligand selectivity between human and dog.
Non-synonymous single-nucleotide polymorphisms (SNPs) would affect the responses to ligands, associating with hypersensitivity against drugs in some patients 30,31 . Although a number of naturally occurring missense variants of human MRGPRX2 have been reported, most of the variants showed similar or decreased responses compared with WT 32 . On the other hand, Chompunud Na Ayudhya et al. have reported that human MRGPRX2 mutants at the carbonyl terminus, which are associated with receptor phosphorylation and desensitisation, showed enhanced responses to substance P 33 . In the present study, we constructed human MRGPRX2 mutants at the TM domain, which is associated with ligand binding, and these mutants demonstrated enhanced responses to fluoroquinolones compared with WT. Our results indicate that missense mutations in F78 or M109 of human MRGPRX2 would induce enhanced responses to certain fluoroquinolones including CPFX in humans. With regard to human MRGPRX2 variants, F78L and V108A are found in gnomAD as naturally occurring missense mutations located at or near positions F78 and M109, respectively. The mutant F78L was reported not to alter the activity against hemokinin-1, substance P, icatibant, and human β-defensin-3 7 , generally consistent with the results of this study. However, it should be considered that the mutation at F78 may induce a greatly enhanced response when it is accompanied by mutation at M109. No reports have been published with regard to V108A's function. Further analysis of SNPs located around F78 or M109 would provide important information to investigate MRGPRX2-related hypersensitivity.
In summary, we focused on dog MRGPRX2, which is sensitive to fluoroquinolones compared with human MRGPRX2, and identified key residues associated with fluoroquinolone-induced histamine release that explain this species difference. Our results have important clinical implications for revealing the mechanism behind fluoroquinolone-mediated pseudo-allergic reactions and assessing the risk of them in humans. Thermo Fisher Scientific Inc.) supplemented with 20 mM hydroxyethylpiperazine-N′-2 ethanesulfonic acid (HEPES; Sigma-Aldrich Co. LLC) and 0.05 vol% bovine serum albumin (BSA; Sigma-Aldrich Co. LLC). The highest concentration of fluoroquinolones was set at 1000 µg/mL based on previous reports, at which the test substances induced marked intracellular calcium mobilisation in MRGPRX2-expressing HEK293 cells 10,20 or caused histamine release in rat or human mast cells 14,17,21 . Intracellular calcium levels were analysed using Calcium Kit II-iCellux (Dojindo Molecular Technologies, Inc., Kumamoto, Japan), in accordance with the manufacturer's instructions. HEK293 cells (1.75 × 10 4 cells/well) were loaded with 1.25 mM probenecid and calcium probe for 45 min at 25 °C. Changes in fluorescence intensities between before and after addition of the test articles were measured over time using FLIPR Tetra (Molecular Devices, LLC, Sunnyvale, CA) with excitation at 470-495 nm and emission at 515-575 nm. The test articles were added 10 s after beginning the measurements. The data were analysed using ScreenWorks (Molecular Devices, LLC, Version 3.2.0.14) to determine the difference between maximal and minimal fluorescence intensity (max-min). As CPFX at 333 and 1000 µg/ mL induced nonspecific increases in intracellular calcium levels in untransfected cells, these data were excluded from the analysis. All experiments were performed in quadruplicate.
Statistical analysis. Data are presented as the mean ± S.D. for calcium mobilisation. Half-maximal effective concentration (EC 50 ) of each test article used in the calcium mobilisation assay was calculated from individual Emax and E0 for each variant using the four-parameter sigmoidal model. Statistical significance (P < 0.05) was determined by unpaired t-test. These analyses were performed using GraphPad Prism 7.03 (GraphPad Software, La Jolla, CA).

Data availability
All data generated or analysed during this study are included in this published article.