Structural determinants of the direct inhibition of GIRK channels by Sigma-1 receptor antagonist

G-protein-gated inward rectifier K+ (GIRK) channels play a critical role in the regulation of the excitability of cardiomyocytes and neurons and include GIRK1, GIRK2, GIRK3 and GIRK4 subfamily members. BD1047 dihydrobromide (BD1047) is one of the representative antagonists of the multifunctional Sigma-1 receptor (S1R). In the analysis of the effect of BD1047 on the regulation of Gi-coupled receptors by S1R using GIRK channel as an effector, we observed that BD1047, as well as BD1063, directly inhibited GIRK currents even in the absence of S1R and in a voltage-independent manner. Thus, we aimed to clarify the effect of BD1047 on GIRK channels and identify the structural determinants. By electrophysiological recordings in Xenopus oocytes, we observed that BD1047 directly inhibited GIRK channel currents, producing a much stronger inhibition of GIRK4 compared to GIRK2. It also inhibited ACh-induced native GIRK current in isolated rat atrial myocytes. Chimeric and mutagenesis studies of GIRK2 and GIRK4 combined with molecular docking analysis demonstrated the importance of Leu77 and Leu84 within the cytoplasmic, proximal N-terminal region and Glu147 within the pore-forming region of GIRK4 for inhibition by BD1047. The activator of GIRK channels, ivermectin, competed with BD1047 at Leu77 on GIRK4. This study provides us with a novel inhibitor of GIRK channels and information for developing pharmacological treatments for GIRK4-associated diseases.

There are four members of the GIRK family (GIRK1-4) which can form functional homo-and hetero-tetramers and are expressed in a variety of native tissues.For example, heteromeric GIRK1/2 channels were identified in the brain, whereas, GIRK1/4 and homomeric GIRK4 channels are mainly expressed in atrial myocytes (16)(17)(18)(19).Previous studies have suggested that the GIRK4 channel might be a promising therapeutic target for atrial fibrillation (AF).The induction of AF is inhibited in GIRK4 knock-out mice (20), and the acetylcholine (ACh)-regulated potassium current, GIRK1/4, constitutively activates in atrial myocytes from AF patients (21).Thus, the identification of GIRK4 inhibitors may provide information relevant to drug design targeting GIRK4-related cardiac arrhythmia.
The sigma-1 receptor (S1R) is a multimodal chaperone protein located chiefly at the mitochondrion-associated endoplasmic reticulum (ER) membrane (MAM) at rest (22).In response to cellular stress or the application of its ligands, it can translocate to other regions of the cell including the ER-PM junction and plasma membrane (PM) (22).S1R has been implicated in a diverse array of pathophysiological conditions including drug addiction, Parkinson's disease, Alzheimer's disease, and Amyotrophic Lateral Sclerosis (22,23).BD1047 dihydrobromide (BD1047) is one of the representative antagonists of S1R (24) and in recent animal experiments was shown to be effective for the treatment of neuropathic pain by affecting the expression of S1Rs in the ipsilateral spinal cord dorsal of rats after chronic constriction injury (25,26).BD1047 was also shown to attenuate ethanolinduced neurotoxicity by modulating the function of both S1R and the inositol triphosphate (IP 3 ) receptor within the rat hippocampus, suggesting its possible use in alcohol-related disorders (27).S1R colocalizes with G i/o -coupled M2 muscarinic receptors on the soma of motoneurons (28) suggesting that it plays a role in the regulation of M2 receptor function and/or expression.
To investigate this further, we used GIRK channels as the effector of the G i/o protein-coupled receptors and BD1047 to antagonize S1R activity.We made the unexpected observation that BD1047 and its analog BD1063, directly inhibit GIRK channel currents in the absence of S1R expression and in a voltage-independent manner.Since this is a novel inhibitor of GIRK channels, we set out to identify the structural determinants involved in this binding interaction.Electrophysiological recordings were carried out in both Xenopus oocytes and isolated rat atrial myocytes, together with mutagenesis and molecular docking analysis.We identified BD1047 as a more potent inhibitor of GIRK4 compared to GIRK2 channels and also showed that it inhibits ACh-induced native GIRK currents in rat atrial myocytes.Leu77 and Leu84 within the proximal N-terminus (N-ter) of GIRK4 and Glu147 with the poreforming region are critical for this inhibition, and ivermectin (IVM), a known activator of GIRK channels, competes with BD1047 at Leu77.

BD1047 directly inhibits GIRK channel currents in Xenopus oocytes
To examine the effects of BD1047 (Fig. 1A) on GIRK channels, Xenopus oocytes were used as an in vitro expression system, and two-electrode voltage clamp experiments were performed.We tested the effects of 100 μM BD1047 on the different subtypes of GIRK channels including homomeric GIRK2 and GIRK4 and heteromeric GIRK1/2 and GIRK1/4.Since homomeric GIRK4 channels show a small basal current (3,17,29), it was co-expressed with its physiological activator, G βγ subunits, to enhance current amplitudes.The application of BD1047 decreased the current amplitude of GIRK1/2, GIRK2, GIRK1/4, and GIRK4 by 51.0 ± 1.8%, 37.7 ± 9.8%, 47.4 ± 4.9%, and 79.9 ± 4.6%, respectively, showing that it directly inhibits GIRK channel currents and is a stronger inhibitor of GIRK4 compared to GIRK2 (Fig. 1, B-F).Similar results were obtained when all of the GIRK channels were coexpressed with G βγ subunits, indicating that the coexpression of G βγ did not affect the inhibition of GIRKs by BD1047 (Fig. 1G).It is noted that the overall difference between Figure 1, F and G is due to a difference in the batch of oocytes, as the values of GIRK4 co-expressed with G βγ also show a difference.We tested the effect of another, structurallyrelated antagonist of S1R, BD1063 dihydrochloride (Fig. 1A) (100 μM) and this also showed a 2-fold greater inhibition of GIRK4 compared to GIRK2 (61.4 ± 2.6% versus 31.1±12.0%, respectively) (Fig. 1H).
The voltage dependence of BD1047 inhibition of GIRK4 was examined.It was observed that the current decreased in 96K + , with 10 μM or 100 μM BD1047 bath solutions (Fig. 2, A and B) with a similar ratio at each applied voltage, showing that BD1047 acts in a voltage-independent manner (Fig. 2C).The concentration-inhibition relationships for the effects of BD1047 on GIRK2 and GIRK4 were measured, and GIRK2 was also co-expressed with G βγ subunits to compare the properties in the same condition (Fig. 2, E and F).The IC 50 values were 75.4 ± 9.7 μM for GIRK2 and 17.4 ± 3.7 μM for GIRK4 (Fig. 2G).The Hill coefficient for GIRK4 was 1.4 ± 0.1 (Fig. 2G), suggesting that the binding of one BD1047 molecule to GIRK4 is sufficient to inhibit the channel.
BD1047 inhibits the ACh-induced GIRK current in rat atrial myocytes GIRK1/4 channels are expressed in mammalian atrial myocytes (3) and are activated by ACh via the stimulation of the M 2 muscarinic receptors to slow the heart rate (3,17,30).We examined the effects of BD1047 on native GIRK currents in isolated rat atrial myocytes.The application of 1 μM ACh activated GIRK1/4 currents which slowly desensitized, a wellknown phenomenon of ACh-induced GIRK current (31,32) (Fig. 3B).A specific blocker of GIRK channels, TPN-Q (33), was used to confirm that the ACh-induced current was carried by GIRK channels.3 μM TPN-Q completely suppressed the ACh-induced current indicating that it was carried by GIRK channels and this effect was almost irreversible (Fig. 3C).The ACh-induced GIRK current was also significantly inhibited by 10 μM and 100 μM BD1047 (Fig. 3, D, and E), showing that native GIRK currents are also sensitive to this inhibitor.In the analysis in Figure 3F, the maximal amplitude of ACh-induced GIRK current at −100 mV was normalized as 1, and the normalized current in the absence or presence of an inhibitor was picked up at the same time point.The value of control is less than 1.0, due to a current decrease by slow desensitization.The negative value in the presence of 100 μM BD1047 is interpreted to be due to the block of the basal GIRK current which exists in the absence of ACh.The dose-inhibition relationship for the effects of BD1047 on GIRK in atrial myocytes was analyzed and the IC 50 value is 7.5 ± 1.0 μM (Fig. 3G).

Identification of the structural determinants for inhibition of GIRK4 channels by BD1047
Taking advantage of the different sensitivities of GIRK2 and GIRK4 to BD1047, chimeras were constructed to identify structural determinants of BD1047 inhibition.Five regions of GIRK2 including the N-terminal region (N-ter), transmembrane domain 1 (TM1), the pore-loop region, TM2, and C-terminal region (C-ter), were substituted for the equivalent regions of GIRK4 to generate six chimeras, GIRK42222, GIRK24222, GIRK22422, GIRK22242, GIRK22224 and GIRK24242 (Fig. 4, A and B).The percentage inhibition by BD1047 for GIRK42222, GIRK22422, and GIRK22224 were 61.8 ± 9.9%, 37.1 ± 14.5%, and 49.2 ± 2.5%, respectively, compared to 19.7 ± 3.9% for GIRK2 (Fig. 4C).There was some variability in the extent of BD1047 inhibition of GIRK2 between different batches of oocytes (37.7% in Fig. 1F and 19.7% in Fig. 4C) hence appropriate controls were included in all experiments.These results demonstrate the importance of the N-ter, pore-forming region, and C-ter of GIRK4 for the inhibition by BD1047.
As the chimera GIRK42222 showed the strongest inhibition, a further chimera was generated in which only the most distal  .Dose inhibition relationships of BD1047 on GIRK2 and GIRK4 current.A, the pulse protocol for recording and the recorded current traces in 96K + , with 10 μM BD1047 or 100 μM BD1047.B, IV-relationship of GIRK4 in an extracellular solution containing 96K + , with 10 μM BD1047 or 100 μM BD1047.C, the normalized current at −120 mV, −100 mV, −80 mV and −60 mV in different extracellular solution (Black: 96K + , blue: 10 μM BD1047 or magenta: 100 μM BD1047).The current amplitude of 10 μM BD1047 or 100 μM BD1047 relative to 96K + (I 10 μM BD /I 96K+ or I 100 μM BD /I 96K+ ) at each potential were calculated.Data are mean ± SD (n = 3) for each plot.D, the pulse protocol for recording in (E) and (F).E and F, left: representative current recordings in Xenopus oocytes evoked by the voltage protocol shown above.BD1047 ranging in concentration from 0.1 μM to 500 μM was applied to GIRK2 (E) and GIRK4 (F) channels.Right: the time courses of the current changes in various concentrations of BD1047 in oocytes expressing (E) GIRK2 or (F) GIRK4 channels.Orange dots indicate the recorded current amplitudes at −100 mV and gray dots indicate those at +40 mV.G, dose-inhibition relationships of BD1047 on GIRK2 (blue) and GIRK4 channel (black).Data are mean ± SD (n = 4-5) for each plot.IC 50 is 75.4 ± 9.7 μM (GIRK2) and 17.4 ± 3.7 μM (GIRK4), respectively.GIRK1/4 or (E) GIRK4+G βγ channels.BD1047 was applied and washed out by constant bath perfusion using a peristaltic pump.F, inhibition by 100 μM BD1047, of GIRK1/2, GIRK2, GIRK1/4, and GIRK4+G βγ channel currents.G, inhibition by 100 μM BD1047, of GIRK1/2, GIRK2, GIRK1/4 and GIRK4 channel currents when G βγ subunits were co-expresssed.F and G, data are mean ± SD (n = 5 for each); 58 amino acids (out of 86 within the N-terminus) were substituted (GIRK2 4(1-58)2222) (Fig. 4D).Inhibition by BD1047 of the chimera containing the distal region of N-ter (1-58) was only 20.6 ± 1.0% (which is close to that of GIRK2 WT, 12.9 ± 5.7%), while the inhibition ratio of the chimera containing the whole N-ter was 54.1 ± 13.4%, close to that of GIRK4 WT, 63.6 ± 9.5% (Fig. 4E).These data show that the distal N-ter of GIRK4 is not involved in the inhibitory effect by BD1047 and highlights the importance of the proximal N-ter (amino acids 59-86) of GIRK4.
Single point mutations were made in GIRK2 and GIRK4 to further narrow down the critical binding site of BD1047 within the proximal N-ter.These two channels differ by only three amino acids within the proximal N-terminus (Fig. 5A). .The effects of BD1047 on GIRK2/4 chimeras.A, chimeric constructs were generated in which 5 regions of GIRK2 (N-ter, TM1, the pore-forming loop between TM1 and TM2, TM2 and C-ter) were substituted with the corresponding regions of GIRK4.B, schematic drawing of six chimeras, GIRK42222, GIRK24222, GIRK22422, GIRK22242, GIRK22224, and GIRK24242.Red letters "4" indicate those swapped to the corresponding part from GIRK4.C, comparison of the inhibition percentages before and after the application of 100 μM BD1047 to GIRK42222, GIRK24222, GIRK22422, GIRK22242, GIRK22224 and GIRK24242 chimeras.Gβγ subunits are co-expressed in all cases.D, schematic drawings of GIRK2/4 N-ter chimeras.GIRK2 4(1-86)2222 includes the whole 86 amino acids of N-ter from GIRK4, and GIRK2 4(1-58)2222 includes the distal 58 amino acids of N-ter from GIRK4.E, comparison of the inhibition percentages before and after the application of 100 μM BD1047.Data are mean ± SD (n = 4-5 for each); one-way ANOVA followed by Tukey's test, * indicates p < 0.05, ** indicates p < 0.01.
For Arg73, Thr80, and Ile82 in GIRK2, the equivalent residues in GIRK4 are Gln, Ser, and Leu, respectively.These were individually mutated in GIRK2 and only GIRK2 I82L showed an increase in BD1047 sensitivity similar to GIRK4 (61.0 ± 1.9% compared to 58.9 ± 3.0%) (Fig. 5B).Consistent with this result, the reverse mutation in GIRK4 reduced inhibition by BD1047 (35.2 ± 6.6%) (Fig. 5C).Taken together, these results highlight the importance of Leu77 in GIRK4 in determining the inhibitory effect of BD1047 on GIRK channels.

Molecular docking of BD1047 to GIRK4 channel
Molecular docking is a useful approach to computationally predict the preferred binding site and docking orientation of a small molecule with a target protein.Thus, computational docking using the software SwissDock online was conducted to identify the possible binding sites of BD1047 on GIRK4.As the structure of GIRK4 has not been solved yet, a homology structural model of GIRK4 was made based on the available structure of GIRK2 (PDB ID: 6XIT) using SWISS-MODEL online.Then, the obtained structure of GIRK4 and the chemical structure of BD1047 were uploaded to the SwissDock site.
The computational docking result of GIRK4 and BD1047 showed that BD1047 (cyan clusters) can bind to GIRK4 channel including the N-ter, pore-forming region and C-ter (Fig. 6A).Besides demonstrating the possible binding positions, the docking result also predicts all the possible binding directions and orientations of BD1047 at each predicted position so that BD1047 is shown in cluster form.Therefore, when the intensity of the clusters at a position is higher, the possibility that BD1047 can bind to that position is judged to be higher.BD1047 clearly showed the highest intensity of clusters at the N-ter, indicating the highest chance that BD1047 interacts with GIRK4 at the N-ter (Fig. 6A).Highlighting the Leu77 residue on GIRK4 (Fig. 6A) clearly demonstrates the importance of the Leu77 residue on GIRK4 for the binding with BD1047 which is consistent with the mutational analysis (Fig. 5C).On the contrary, the BD1047 cluster does not appear on the N-ter in the docking of BD1047 with the GIRK2 channel (Fig. 6B).Interestingly, the docking of BD1047 on N-ter of GIRK4 is completely lost when Leu77 is mutated to Ile which is the corresponding amino acid on GIRK2 (Fig. 6C), whereas the cluster of BD1047 on the N-ter shows up in GIRK2 I82L mutation (Fig. 6D).Taken together, these results demonstrate the significant role of GIRK4 Leu77 in the binding of BD1047.
The docking results suggest that other amino acids adjacent to Leu77 in GIRK4, such as Tyr73, Leu74, Thr80, leu81, and Leu84, might also play a critical role in binding BD1047 (Fig. 7A).These amino acids could not be identified by comparing the sequence alignment between GIRK2 and GIRK4, because they are conserved in both channels.To test this, alanine was individually substituted for Leu74, Leu81, and Leu84 and inhibition by BD1047 was reduced for all three mutants (Fig. 7B), thus validating the computational docking results.It should be noted that although Leu84 is conserved in both GIRK2 and GIRK4, it plays a critical role in the inhibition by BD1047 as the Leu84Ala mutation completely abolished the inhibition effect and even slightly increased the current amplitude (Fig. 7B).
The pore-forming region and C-terminus are also involved in the binding of BD1047 to GIRK4 The analysis of the chimeras (Fig. 4C) and molecular docking analysis (Figs. 6A and 7A) indicated that not only the N-ter but also the pore-forming region and C-ter are involved in the inhibition of the GIRK4 channel by BD1047.The residues Glu147 and Thr149 residues within the pore-forming region (Figs.5A and 8A) and Glu289 and Glu300 within the C-ter (Figs. 5A and 8B) were selected as candidates for further analysis since they showed a relatively high probability of interaction with BD1047 from the docking results.Alanine substitutions were made at each of the four positions.GIRK4 T149A and GIRK4 E300A channels appeared to be nonfunctional (Fig. 8C), whereas GIRK4 E298A was functional but there was no change in sensitivity to BD1047 compared to the wild-type channel (Fig. 8D).In contrast, GIRK4 E147A showed a significant reduction in sensitivity to BD1047 (from 68.5 ± 7.2% to 24.7 ± 7.7%) (Fig. 8D).To characterize the role of the side chain of Glu147, it was also mutated to Gln which has a similar size as Glu, but with a neutralized charge (Fig. 8E).It was observed that GIRK4 E147Q maintained the strong inhibition as WT, indicating that the size rather than the charge of the side chain of Glu147 is important for the inhibition by BD1047 (Fig. 8E).These data suggest that the Glu147 residue within the pore-forming region of the GIRK4 channel is also involved and the size of the amino acid residue at 147 position is important for the inhibition by BD1047.

Competition between inhibitor and activator at Leu77 on GIRK4
IVM is an activator of GIRK channels, and a critical determinant of GIRK2 sensitivity to IVM is Ile82, which is equivalent to Leu77 in GIRK4 (34).Because this residue in involved in binding both IVM and BD1047 (Fig. 9A), the activator and inhibitor are likely to compete. .Computational molecular docking of GIRK4 and BD 1047 showing the contribution of amino acid residues adjacent to Leu77.A, the predicted protein structure is the homology model of GIRK4 based upon the structure of GIRK2 (6XIT).Cyan clusters of BD1047 indicate the predicted dockings in various directions and orientations of BD1047.Tyr73, Leu74, Leu77, Thr80, Leu81, and Lue84 are highlighted in red, magenta, pink, purple, orange and yellow colors, respectively.Right panels are enlarged images of Leu77 and adjacent amino acid residues for BD interaction from the left panel.B, the effects of BD1047 on GIRK4 and its mutants of the predicted docking sites.Comparison of the inhibition percentages before and after the application of 100 μM BD1047 on oocytes expressing GIRK4, GIRK4 L77I, GIRK4 L74A, GIRK4 L81A, and GIRK4 L84A.Data are mean ± SD (n = 4 for each); one-way ANOVA followed by Tukey's test, ** indicates p < 0.01.To test this possibility, we compared both the concentration-inhibition relationship and inhibition kinetics for BD1047 at GIRK4 in the absence and presence of 100 μM IVM (Fig. 9, B-D).In the presence of IVM, there was a clear rightward shift in this relationship, with the IC 50 of BD1047 in the presence of IVM of 62.9 ± 6.0 μM compared to 19.6 ± 0.5 μM in its absence (Fig. 9B), showing a weaker sensitivity of GIRK4 to BD1047 in the presence of IVM.The presence of IVM decelerated the BD1047 0 inhibition kinetics (Fig. 9, C-E) and the time constant of inhibition in the absence and presence of IVM were 11.8 ± 1.1 s and 22.4 ± 1.9 s respectively (Fig. 9F).Taken together, these results show that IVM affects the binding of BD1047 and competes with BD1047, presumably at Leu77 on GIRK4.

Discussion
GIRK channels are involved in several different physiological processes in cardiac and neuronal systems and there exists a limited number of selective-isoform modulators.Here we demonstrate that BD1047 directly inhibits GIRK channels, showing a more potent inhibition of GIRK4 compared to GIRK2, and we have identified the structural determinant using a combination of mutagenesis and computational docking analysis.
BD1047 is an antagonist not only to the Sigma-1 receptor but also to GIRK channels BD1047 was identified as an antagonist of S1R by binding assays and behavioral studies in rats.In radioligand binding studies using rat liver and guinea pig brain, BD1047 bound S1R with high affinity.In rats, the microinjection of BD1047 into the red nucleus decreased dystonia induced by haloperidol, a known S1R agonist (35).BD1047 was shown to block intracellular calcium responses in cultured astrocytes and to alleviate S1R-induced pain hypersensitivity in mice (36).Thus, BD1047 has been accepted as a selective antagonist to S1R.Our studies newly show that BD1047 can also affect other molecules such as GIRK channels, in addition to S1R.
Previously, a possible usage of BD1047 was suggested in alcohol use disorder, because BD1047 could attenuate ethanolinduced intracellular Ca 2+ in rat hippocampus through the modulation of the function of S1R and ER-bound IP 3 (27).It is known that GIRK channels can be activated by alcohol (6) and the usage of GIRK inhibitors such as ifenprodil could block the alcohol-induced GIRK currents (37).Thus, it is possible that BD1047 might play a therapeutic role in the alcohol use disorder also by inhibiting GIRK currents, besides the inhibition of the function of S1R.
The mechanism of binding of BD1047 to GIRK4 channel Mutagenesis highlighted Leu77 as playing an important role in BD1047-mediated inhibition of GIRK4; however, BD1047 is still more potent at GIRK4 L77I compared to GIRK2 WT (Fig. 5C) suggesting that other region(s) of GIRK4 contribute to the binding of BD1047.In the analysis of concentrationinhibition relationship of BD1047, the voltage-dependent slow activation upon hyperpolarization of GIRK2 channel change was observed only in the presence of a high concentration of BD1047 (Fig. 2E).The change in the kinetics of the voltage-dependent activation may represent multiple binding sites of BD1047 to GIRK2 channels.For example, BD1047 only binds to a high affinity binding site on GIRK2 when the concentration is low, while it also binds to low affinity binding site to induce slow activation when the concentration is high (Fig. 2E).In addition, GIRK2/4 chimera studies also showed a possible involvement of the pore-forming region and the C-ter of GIRK4 (Fig. 4C), which are consistent with the results of the molecular docking study (Fig. 6A).Meanwhile, docking data indicated some amino acids residues which are conserved between GIRK2 and GIRK4 on the proximal N-ter also have high chance to interact with BD1047 (Fig. 7A).By performing mutagenesis studies on the N-ter, pore-forming region, and Cter of GIRK4, we identified furthermore amino acid residues, Leu74, Leu81, Leu84 in the N-ter and Glu147 in the pore region, which is also involved in the inhibition by BD1047 (Figs 7B and 8D).These results might explain why there was only a partial decrease in the BD1047's inhibition when just one single amino acid, Leu77, was mutated on GIRK4.It appears that the Leu84 conserved in GIRK2 and GIRK4 is very important for the inhibition by BD1047 and it was surprising that the GIRK4 L84A mutation not only abolished the inhibition but also slightly increased the current by BD1047 application.However, the underlying mechanism remains to be elucidated (Fig. 7B).
Interestingly, the docking of GIRK4 L77I and BD1047 showed a loss of the N-ter cluster of BD1047, whereas GIRK2 I82L showed an acquisition of it.Leucine and isoleucine are both hydrophobic amino acids in the same category.The structural difference between them is that the leucine contains an isobutyl side chain with two methyl in the end to form a mermaid-tail-like structure (Fig. 10A), while isoleucine contains a secbutyl side chain with one methyl in the end (Fig. 10B).By superposing the homology model of GIRK4 L77I with the structure of GIRK2 WT, it was confirmed that the side chain position and direction of Ile77 substituted into GIRK4 is the same as Ile82 in GIRK2 (Fig. 10C) and vice versa for GIRK2 I82L and GIRK4 (GIRK4 Leu77) (Fig. 10D).This comparison may explain the loss and gain of the docking of BD1047 after mutation on the N-ter relationships of BD1047 on GIRK4 channel in the absence of IVM (black) and presence of IVM (blue).Data are mean ± SD (n = 4-5) for each plot.IC 50 is 19.6 ± 0.5 (μM) (-IVM) and 62.9 ± 0.5 (μM) (+IVM), respectively.C, the timelapse changes of the current amplitudes at −100 mV (orange plots) in 96K +, ND96 and 96K + with 100 μM BD1047 solution in oocytes only expressing GIRK4+G βγ channels.D, the timelapse changes of the current amplitudes at −100 mV (orange plots) in 96K + , ND96 (black bar), 96K + with 100 μM IVM (green bar), 96K + with 100 μM BD1047 (red bar) and IVM solution (green bar).E, effects of IVM on the inhibition kinetics of BD1047 on GIRK4 current.The inhibition kinetics were calculated from the current amplitude at −100 mV after the application of BD1047 in the absence or presence of IVM.The maximal reduction of current after the application of BD1047 was normalized as 1.F, Tau of BD' inhibition calculated from (E).Data are mean ± SD (n = 4 for each); unpaired t test, *** indicates p < 0.001.
Direct inhibition of GIRK by S1R antagonist of GIRK2 or GIRK4.Meanwhile, it seems that the mermaidtail-like structure formed by two methyl in the isobutyl side chain of leucine is very critical to forming the binding pocket for BD1047.
The results above show that computational docking data indeed help the confirmation of the incomplete mutagenesis study and provide quite comprehensive and accurate information.However, as its trustability is not necessarily guaranteed, the combined usage with electrophysiological analysis of chimeras and mutants performed in this study is indispensable to drawing a solid conclusion.For example, docking data also demonstrated binding of BD1047 within the pore region and C-ter of GIRK4 which is consistent with chimera results, suggesting there might be some other essential amino acid residues for the inhibition.We tried some more mutagenesis study according to the docking results as well (Fig. 8, A and B).Glu147 and Thr149 on pore-forming region as well as Glu298 and Glu300 on the C-ter has been mutated to Ala respectively, but the attenuation of the inhibition of BD1047 was only observed in GIRK4 E147A, not others (Fig. 8, C and D).Leu77 on GIRK4 is a shared hot spot for drug binding Our group reported previously that the antiparasitic drug, IVM, can activate GIRK channels, and the critical amino acid residue for the activation of the GIRK2 by IVM is Ile82.Ile82 is located in the slide helix between the TM1 and the N-terminal cytoplasmic tail domain (CTD) (34).In our experiments, Leu77 of GIRK4, corresponding to Ile82 of GIRK2, was identified as the critical amino acid residue for the inhibition effect by BD1047.Furthermore, another group recently showed a role for GIRK4 Leu77 (corresponding to GIRK2 Ile82) in the activation of GIRK channels by a novel activator, 3hi2one-G4.The 3hi2one-G4 selectively activates the GIRK4 channel via Leu77 and the effects of 3hi2one-G4 is significantly decreased when Leu77 is mutated to Ile77 (38).Taken together, both activator and inhibitor with different chemical structures bind to this slide helix region between TM1 and CTD of the GIRK channel, suggesting that this position might be a shared hot spot for drug binding.Thus, it could be used as a therapeutic target to develop new medicines for GIRK channel-associated diseases.
The identification of an inhibitor that shows selectivity for GIRK4 over GIRK2 has the potential to be of therapeutic use.For example, there is evidence to suggest that GIRK4 plays an important role in heart arrhythmias because GIRK1/4 channels are constitutively activated in the atrial myocytes of chronic AF patients (21) and GIRK4 knockout mice, ACh failed to induce AF (20).Various inhibitors of GIRK channels such as tertiapin-Q, clozapine, fluoxetine or terfenadine have been identified, but they do not show selectivity towards homomeric GIRK4 channels (11)(12)(13)(14)(15).Our results demonstrating that BD1047 is more potent at GIRK4 may help to optimize the development of novel therapies for GIRK4-associated arrhythmia with reduced side effects caused by actions at other GIRK channels.

Conclusion
In this study we investigated the mechanism of action of BD1047 at GIRK channels and showed (1) BD1047 directly inhibits GIRK channel currents independent of the sigma-1 receptor and produces a greater inhibition of GIRK4 compared to GIRK2; (2) It also inhibits ACh-induced native GIRK current in isolated rat atrial myocytes; (3) Switching the proximal cytoplasmic N-terminal regions of GIRK2 and GIRK4 is sufficient to switch sensitivity to BD1047 and within this region of GIRK4, Leu77, and Leu84 are a critical determinant of this inhibition; Mutagenesis study on the pore-forming region of GIRK4 also indicates the importance of the Glu147; (4) Molecular docking analysis similarly suggests that Leu77 and Leu84 as well as Glu147 contribute to the BD1047 binding site on GIRK4; (5) An activator of GIRK channels, IVM, competes with BD1047 at Leu77 on GIRK4.These results provide us with a novel inhibitor of the GIRK channel and information for developing pharmacological treatments to GIRK-associated diseases.

Ethical approval
All animal experiments in this study were approved by the Animal Care Committee of the National Institutes of Natural Sciences (an umbrella institution of National Institute for Physiological Sciences), and were performed in accordance with its guidelines.

Preparation of Xenopus laevis oocytes
Oocytes were isolated from X. laevis (purchased from Hamamatsu Seibutsu Kyouzai, Hamamatsu, Japan) under anesthesia of 0.15% tricaine (Sigma-Aldrich) by surgery.The surgical operation on the frog was performed on ice and an incision was made in the frog's abdomen to remove the oocytes.Isolated oocytes were incubated in 2 mg mL −1 collagenase type I (Sigma-Aldrich) for 6 h to remove the follicular membrane and incubated in frog Ringer's solution containing 88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO 3 , 0.3 mM Ca(NO 3 ) 2 , 0.41 mM CaCl 2 , 0.82 mM MgSO 4 and 15 mM HEPES with 0.1% penicillin-streptomycin at 17 C.After the injection of 50 nl of cRNA, oocytes were incubated in frog Ringer's solution for a further 3 to 4 days prior to electrophysiological recordings.

Isolation of rat atrial myocytes
Eight-weeks female adult Wistar rats (180-200 g) were purchased from Japan SLC, Inc and housed in square cages with wood shavings, two rats per cage, on a light-dark cycle (12:12 h) at 23 C ± 1 deg.C and fed standard rat chow in the Animal Facility of the National Institute for Physiological Sciences (NIPS, Japan).Isolation of atrial myocytes was performed as previously described (15,39).Rats were anesthetized after the intraperitoneal (I.P.) injection of 10 mg kg −1 xylazine hydrochloride and 100 mg kg −1 thiopental sodium for 6 to 8 min following i.p. injection of 1000 U kg −1 heparin.Then successful anaesthesia was judged by the full disappearance of the pedal withdrawal reflex.The rats were sacrificed by taking out the heart from the thoracic cavity under anesthesia and the heart was connected to a modified Langendorff perfusion system via the aorta.A cannula was inserted into the aorta with the continuously retrograde perfusion of the Ca 2+ -free Tyrode's solution (137 mM NaCl, 5.4 mM KCl, 1 mM MgCl 2 , 10 mM glucose, and 10 mM HEPES, pH 7.4 with NaOH) containing 0.6 mg ml −1 collagenase type II (Worthington) and 0.2 mg ml −1 protease type XIV (Sigma-Aldrich) at 37 C for 30 to 35 min through coronary arteries via aorta.Then the KB solution (10 mM taurine, 10 mM oxalic acid, 70 mM K glutamate, 25 mM KCl, 10 mM KH 2 PO 4 , 11 mM glucose, 0.5 mM EGTA, and 10 mM HEPES, pH 7.3 with KOH) was used to wash the digested tissue.The atria were cut into pieces with scissors and transferred to a culture dish.Atrial myocytes were dissociated by gently shaking the cut tissue pieces using a tweezer.Dissociated atrial myocytes were incubated in the KB solution and myocytes were re-seeded onto Poly-L-Lysine (PLL) coated glasses and used for electrophysiological recordings on the same day.

Mutagenesis and cDNA and cRNA preparations
For experiments in Xenopus oocytes, cDNAs of mouse GIRK2 and rat GIRK4 were subcloned into pGEMHE.
Mutations in GIRK2 and GIRK4 were introduced using the PfuUltra II Fusion HS DNA Polymerase kit (Agilent Technologies) and verified by DNA sequencing.After linearization of cDNA by restriction enzymes, complementary RNAs were transcribed using mMessage mMachine kit (Ambion).The amount of each cRNA injected per oocyte was as follows: for rat GIRK1 (8.3 ng), mouse GIRK2 (4.2 ng), rat GIRK4 (8.3 ng), bovine G-protein β1 (8.3 ng) and bovine G-protein γ2 (8.3 ng).

Electrophysiological recording
Two-electrodes voltage clamp recordings were made from Xenopus oocytes 3 to 4 days post-injection and data were acquired using an oocyte clamp amplifier (OC-725C, Warner Instruments), a digital analog converter (Digidata 1550A, Molecular Devices) and pCLAMP 10.5 software (Molecular Devices).Recording microelectrodes had a resistance of 0.1 to 0.5 MΩ when filled with 3 M potassium acetate with 10 mM KCl.The 96K + extracellular (EC) recording solution contained 96 mM KCl, 3 mM MgCl 2 , and 5 mM HEPES (pH 7.5 with KOH) and the ND96 solution contained 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 1.8 mM CaCl 2 and 5 mM HEPES (pH 7.5 with NaOH).All experiments were performed at 25 to 28 C with continuous perfusion using a peristaltic pump (AC-2110 II, ATTA) and BD1047 was applied to the whole bath.
In the whole cell patch clamp recording from isolated rat atrial myocytes, data were acquired using a patch clamp amplifier (AXOPATCH 200B), a digital analog converter (Digidata 1440A, Molecular Devices) and pCLAMP 10.7 software (Molecular Devices).Myocytes were attached to PLL coated glass coverslips 1 to 3 h before recording and placed in the recording chamber, and membrane currents were recorded under a whole cell patch clamp using a glass micropipette with an access resistance of 3 to 5 MΩ when filled with 130 mM KCl, 5 mM Na 2 ATP, 4 mM MgCl 2 , 0.1 mM CaCl 2 , 3 mM EGTA, 0.2 mM GTP, and 10 mM HEPES (pH 7.4 with KOH).A Ca 2+ -free Tyrode solution was used until the whole cell patch clamp configuration was achieved to avoid a twitching motion of the cardiac myocytes.The solution was then switched to a 20K + EC solution containing 115 mM NaCl, 20 mM KCl, 0.53 mM MgCl 2 , 1.8 mM CaCl 2 , 5.5 mM glucose, and 5.5 mM HEPES (pH 7.4 with NaOH).All experiments were performed at 25 to 28 C and ligands (1 μM ACh, 3 μM tertiapin-Q (TPN-Q), BD1047) dissolved in 20K + solution were applied by gravity flow using a muti-valve controller system (VC-8 Valve controller, Warner instruments) combined with SF-77B perfusion fast step (Warner Instruments) for inlet and washed out via a suction pipette by pressure for outlet.

Molecular docking
The homology model of GIRK4 was generated based on the structure of GIRK2 (6XIT, RCSB protein data bank) using SWISS-MODEL (https://swissmodel.expasy.org/).The dimer form of GIRK2 was directly built using PyMOL software and the dimer form GIRK4 was made based on the homology model of GIRK4 using PyMOL software.The tetrameric structure of mutants, GIRK2 I82L and GIRK4 L77I, were built using SWISS-MODEL and then modified to dimer form by PyMOL.The chemical structure of BD1047 was downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/compound/BD-1047-Dihydrobromide).The dimer form of GIRK2, GIRK4, GIRK2 I82L or GIRK4 L77I with the structure of BD1047 were uploaded to SwissDock (http://www.swissdock.ch/) to obtain the docking results respectively.The computational docking results were presented with color marks using Chimera 1.16 software (https://www.cgl.ucsf.edu/chimera/olddownload.html).

Chemicals
Oxotremorine-M (OXO-M) (Sigma-Aldrich) was dissolved in distilled water to make a 50 mM stock solution and further diluted in the bath solution for the final concentration 50 μM.BD1047 and BD1063 (Tocris Bioscience) were dissolved in distilled water to make a 50 mM stock solution and further diluted in the bath solution for the final concentrations.IVM (Sigma-Aldrich) was dissolved in DMSO to make a 10 mM stock solution and further diluted in the bath solution to 100 μM.ACh (Sigma-Aldrich) was dissolved in distilled water to make a stock solution and then diluted in the bath solution to 1 μM.Sources of other materials are described in the relevant methods.

Data analyses
Two-electrode voltage clamp data were analyzed by Clampfit 10.7 (Molecular Devices) and Igor Pro 5.0 (Wave-Metrics) and are shown as mean ± SD from n single oocytes.The current amplitude of GIRK channel was recorded at −100 mV before and after the application of BD1047.The inhibition percentage of GIRK current by BD1047 was calculated by the decrease in the current amplitude in the presence of 100 μM BD1047 from the current before its application at −100 mV (I GIRK basal -I GIRK BD1047 (+) /I GIRK basal ), and the basal current amplitude after subtraction of the current in ND96 solution was normalized as 100%.The dose-inhibition relationship was analyzed by sequential application of various concentrations of BD1047.Using Igor, the data were fitted to a Hill equation: y = I min + (I max − I min )/(1 + (x/IC 50 ) ^(Hillslope)).To analyze the effects of IVM on the rate of BD1047's inhibition, the maximal inhibited current amplitude was normalized as 1, and the normalized inhibited current in temporal variation was calculated from the beginning to the end of the application of BD1047 (I normalized BD inhibited current = (I BD -I BD begin )/(I BD final -I BD begin )).Then data were fitted to the exponential formula: f ðtÞ ¼ P n¼1 i¼1 Ai e −t=τi + C and tau of the inhibition were calculated by the fitting by Clampfit 10.7 (Molecular Devices).
In the patch-clamp experiments, all data were analyzed by Clampfit 10.7 (Molecular Devices) and Igor Pro 5.0 (Wave-Metrics) and are shown as mean ± SD from n single atrial myocytes.To eliminate the influence of the desensitization of ACh-induced current, the maximal amplitude of ACh-induced GIRK current at −100 mV was normalized as 1 after the Direct inhibition of GIRK by S1R antagonist subtraction of the basal current and the normalized current in the absence or presence of inhibitor (BD1047 or TPN-Q) was measured at the same time point (after application of TPN-Q or BD1047 for 27 s) and then calculated by I normalized current = (I ACh-GIRK inhibitor (+) (27s) -I basal )/(I ACh-GIRK(max) -I basal ).The normalized current of 0 means there was no remaining AChinduced GIRK current after the application of inhibitors.
All experiments were performed with the number of n ≥4 for each group.Statistical significances of differences were evaluated using Tukey's multiple comparison tests following one-way ANOVA or unpaired t test.The values of p <0.05 were judged to be statistically significant.

Figure 1 .
Figure 1.The effects of BD1047 and BD1063 on GIRK channels.A, left: the pulse protocol for recording; Middle: chemical structure of BD1047; Right: chemical structure of BD1063.B-E, left: representative current traces measured in Xenopus oocytes; Right: The time-lapse changes of the current amplitudes at −100 mV (orange plots) and at +40 mV (gray dots) in ND96, 96K + and 96K + with 100 μM BD1047 solution in oocytes expressing (B) GIRK1/2, (C) GIRK2, (D)

Figure 3 .Figure 4
Figure 3. Effects of BD1047 on ACh-induced GIRK channel in rat atrial myocytes.A, the voltage recording protocol used for patch-clamp recording from atrial myocytes.B and E, the timelapse change of whole-cell currents recorded from the rat atrial myocytes in 20 mM K + solution with (B) 1 μM ACh alone, (C) ACh with 3 μM TPN-Q (green bar), (D) ACh with 10 μM BD1047 (red bar) or (E) 100 μM of BD1047 (red bar).F, comparison of the normalized current after the application of inhibitor for 27 s.The control group indicates the desensitization of ACh-induced current which is calculated at the same time point as other groups.Data are mean ± SD (n = 5-7 for each); One way ANOVA followed by Tukey's test, ** indicates p < 0.01.G, dose-inhibition relationships of BD1047 on ACh-induced GIRK channel.Data are mean ± SD (n = 2-5) for each plot.IC 50 is 7.5 ± 1.0 μM.

Figure 5 .
Figure 5.The effects of BD 1047 on GIRK2 and GIRK4 mutants.A, the amino acid sequence alignment of mouse GIRK2 and rat GIRK4.A red dotted line indicates the start of the proximal N-ter of GIRK2 and GIRK4.Red arrows indicate non-conserved amino acids in the proximal N-ter region.Green arrows indicate the possible binding sites of BD1047 to GIRK4 on the pore-forming region.Blue arrows indicate the possible binding sites between BD1047 and GIRK4 on the C-ter.B, comparison of the inhibition percentages before and after the application of 100 μM BD1047 on GIRK2, GIRK4, GIRK2 4(1-86)2222, GIRK2 R73Q, GIRK2 T80S and GIRK2 I82L.C, comparison of the inhibition percentages before and after the application of 100 μM BD1047 on GIRK2, GIRK2 I82L, GIRK4 and GIRK L77I.Data are mean ± SD (n = 4-5 for each); one-way ANOVA followed by Tukey's test, * indicates p < 0.05; ** indicates p < 0.01.

Figure 6 .
Figure 6.Computational molecular docking of BD1047 to GIRK4 and GIRK2.A and B, the homology structure model of GIRK4 (A) and GIRK4 L77I (B) based on the structure of GIRK2 (6XIT).Cyan clusters of BD1047 indicate the predicted dockings of BD1047 to GIRK4 WT or GIRK4 L77I in various directions and orientations.Leu77 or Ile77 is highlighted in pink color.C and D, the structure of GIRK2 (6XIT) (C) and homology structure model of GIRK2 I82L (D) based on the GIRK2.Cyan clusters of BD1047 indicate the predicted dockings of BD1047 to GIRK2 WT or GIRK2 I82L in various directions and orientations.Ile82 or Leu82 is highlighted in pink color.The right panels are an enlarged image of BD1047 docking on N-ter region in each left panel.

Figure 7
Figure 7. Computational molecular docking of GIRK4 and BD 1047 showing the contribution of amino acid residues adjacent to Leu77.A, the predicted protein structure is the homology model of GIRK4 based upon the structure of GIRK2 (6XIT).Cyan clusters of BD1047 indicate the predicted dockings in various directions and orientations of BD1047.Tyr73, Leu74, Leu77, Thr80, Leu81, and Lue84 are highlighted in red, magenta, pink, purple, orange and yellow colors, respectively.Right panels are enlarged images of Leu77 and adjacent amino acid residues for BD interaction from the left panel.B, the effects of BD1047 on GIRK4 and its mutants of the predicted docking sites.Comparison of the inhibition percentages before and after the application of 100 μM BD1047 on oocytes expressing GIRK4, GIRK4 L77I, GIRK4 L74A, GIRK4 L81A, and GIRK4 L84A.Data are mean ± SD (n = 4 for each); one-way ANOVA followed by Tukey's test, ** indicates p < 0.01.

Figure 8 .
Figure 8. Identification of the binding sites of BD1047 to GIRK4 in the pore-forming region and the C-ter.A and B, using the homology model of GIRK4 based on the structure of GIRK2 (6XIT), cyan clusters of BD1047 indicate the predicted dockings in various directions and orientations of BD1047.A, the possible binding sites of BD1047 on the pore-forming region of GIRK4.Glu174 and Thr149 are highlighted in pink.B, the possible binding sites of BD1047 on the C-ter region of GIRK4.Glu298 and Glu300 are highlighted in pink.C, representative current recordings in ND96, 96K + , 96K + with 100 μM BD1047 of GIRK4 WT, GIRK4 E147A, GIRK4 T149A, GIRK4 E289A, and GIRK4 E300A in Xenopus oocytes evoked by the voltage protocol shown in the figure.G βγ subunits are co-expressed in all five cases.D, comparison of the inhibition percentages before and after the application of 100 μM BD1047 on GIRK4 WT, GIRK4 E147A, and GIRK4 E289A.E, comparison of the inhibition percentages before and after the application of 100 μM BD1047 on GIRK4 WT, GIRK4 E147A, and GIRK4 E147Q.Data are mean ± SD (n = 4-5 for each); one-way ANOVA followed by Tukey's test, ** indicates p < 0.01.

Figure 9 .
Figure 9. Competition between BD1047 and IVM at GIRK4.A, the amino acid sequence alignment of mouse GIRK2 and rat GIRK4.The blue arrow and red arrow indicate the Ile82 of GIRK2 and the corresponding amino acid Leu77 on GIRK4, respectively.B, the effects of 100 μM IVM on the concentrationresponse relationship of BD1047 on the GIRK4 channel.Left and middle: the representative current recordings in Xenopus oocytes.Various concentrations of BD1047 from 0.1 μM to 500 μM were applied to the GIRK4 channel in the absence of IVM (left) and presence of IVM (middle).Right: Dose-inhibition

Figure 10 .
Figure 10.Comparison of the side chain between GIRK4 and GIRK2.A, the chemical structure of leucine.The Isobutyl side chain is highlighted with a red frame.B, the chemical structure of isoleucine.Secbutyl side chain is highlighted with a red frame.C, left: the overlay of the structure of GIRK2 WT (6XIT) (yellow color) and homology model of GIRK4 L77I based on GIRK2 (6XIT) (cyan color).Right: an enlarged image from the left panel.D, left: the overlay of the structure of the homology model of GIRK4 WT based on GIRK2 (6XIT) (gray color) and the homology model of GIRK2 I82L based on GIRK2 (6XIT) (orange color).Right: an enlarged image from the left panel.