Plasma membrane PI(4,5)P2 regulates human PAC1 and PAC2 channel gatingTo investigate the regulatory effects of plasma membrane PI(4,5)P2 on PAC channel activity, the amino acid sequences of two splice variants of human PAC, hPAC1 and hPAC2, and one mouse PAC (mPAC) channels were compared (Supplementary Fig. 1). hPAC1 and hPAC2 exhibited a difference in the N-terminal sequence, with 61 more amino acids in hPAC1. We verified that PAC cDNA extracted from the human embryonic kidney 293 cell line with a plasmid encoding a temperature-sensitive mutant of the SV40 large T antigen (HEK293T) is the shorter hPAC2 isoform. Thus, we produced an hPAC1 channel by inserting 61 amino acids into the N-terminus of hPAC2 cDNA through inverse PCR cloning. We also made five PAC knockout (KO) HEK293T cell lines, KO1 to KO5, via CRISPR/Cas9 (Supplementary Fig. 2a) and confirmed significantly reduced endogenous currents in all KO cell lines at pH 4.6. Among them, KO3 produced almost negligible PAC channel activity (Fig. 1a and Supplementary Fig. 2b).Fig. 1: PI(4,5)P2 depletion inhibits human PAC channels.a Voltage-dependent currents of hPAC channels in extracellular pH 7.4 or 4.6 solutions in wild-type (WT) HEK293T cells expressing PAC channels endogenously and PAC CRISPR knockout KO3 cells. b (left) Representative confocal images of KO3 cells expressing hPAC1-GFP (top) or hPAC2-GFP (bottom) with the plasma membrane marker Lyn-mCherry (Lyn-mCh). The yellow arrow indicates the location and line scanning direction in merged images. Scale bar: 10 μm. (right) Line-scan analyzes of Lyn-mCh and hPAC1-GFP (top) or hPAC2-GFP (bottom). *, peak cytosolic fluorescence intensity. Each experiment was repeated 3 times independently with similar results. c Current density–voltage relationship for hPACs at various pH conditions. n = 5 for pH 7.4, 6.3, 5.3, 5.0; n = 10 for pH 5.5; n = 7 for pH 4.6 in hPAC1; n = 2 for pH 7.4; n = 5 for pH 6.3, 5.5; n = 3 for pH 5.3, 5.0; n = 4 for pH 4.6 in hPAC2. Data are mean ± SEM. d (left) Scheme for PI(4,5)P2 depletion by Dr-VSP. (right) ΔFRETr changes in control (top) or cells transfected with Dr-VSP (bottom) during +100-mV/10-s depolarization pulse. Cells were transfected with the PI(4,5)P2 FRET probes, PHPLCδ1-CFP and PHPLCδ1-YFP. e pH 5.5-induced PAC current traces during test pulse in KO3 cells (top) and transfected with hPAC1 (middle) or hPAC2 (bottom), without (control) or with Dr-VSP. f Current density measured by analyzing the steady-state current at the end of each +100-mV depolarization step in (e). KO3: n = 8; n = 8. hPAC1: n = 6; n = 7. hPAC2: n = 9; n = 6 for −Dr-VSP and +Dr-VSP, respectively. Dots indicate the individual data points for each cell. Data are mean ± SEM. g Average percent inhibition of current densities by PI(4,5)P2 depletion in different pH solutions. h hPAC1 current recovery from Dr-VSP-induced inhibition in cells intracellularly perfused with (top) or without (bottom) ATP. PAC currents were measured at the indicated times after a + 120-mV/10-s depolarization pulse. Dotted lines indicate zero, and dashed lines indicate PAC current amplitude (control) before the depolarizing step. i Average current recovery with (open) or without (filled) of intracellular ATP (n = 6, +ATP; n = 4, -ATP). Data are mean ± SEM. Source data are provided as Source Data file.First, to examine PAC channel cellular distribution, KO3 cells were co-transfected with each hPAC channel labeled with a green fluorescent protein (GFP), hPAC1-GFP, or hPAC2-GFP, and the plasma membrane marker Lyn11-mCherry (Lyn-mCh). We confirmed that the hPAC1 was predominantly colocalized with Lyn-mCh at the plasma membrane (as evidenced by overlapped GFP and mCh line scanning results). In contrast, hPAC2 was primarily distributed in cytosolic organelles and the plasma membrane (Fig. 1b and Supplementary Fig. 3b, c). To further clarify the cytosolic hPAC2 localization, we visualized colocalization with the early endosome marker Rab5 and the lysosomal marker LAMP1. The hPAC2 was mainly colocalized with Rab5 but not LAMP1 (Supplementary Fig. 3d, e). The hPAC2 N-terminus has the endosomal targeting motif YXXφ, composed of 10YQEL13 sequences, where tyrosine (Y) or leucine (L) substitution by alanine (A) causes a relocalization from the vesicle to the cell surface8. Similarly, since the extra 61 hPAC1 amino acids were inserted into the endosomal targeting sequence, hPAC1 was prominently expressed at the cell surface (Supplementary Fig. 3a). hPAC1 and hPAC2 current densities increased under lower extracellular pH. Under pH 5.3, hPAC1 displayed a greater channel activity than hPAC2, coinciding with Y10A or L13A hPAC2 mutant results in a previous study (Fig. 1c)8.We examined whether the plasma membrane phospholipid PI(4,5)P2 regulated hPAC1 and hPAC2 using the zebrafish voltage-sensing phosphatase (Dr-VSP). This phosphatase dephosphorylates the 5ʹ-phosphate of PI(4,5)P2 at the plasma membrane’s inner leaflet when stimulated by a depolarizing voltage (Fig. 1d, left)21,22. To examine the regulatory effect of the inner leaflet PI(4,5)P2 on PAC activity, a 10-s depolarizing step protocol was applied at +100 mV to elicit PAC Cl- currents and deplete PI(4,5)P2. Using the Förster resonance energy transfer (FRET) approach, we examined the time course of Dr-VSP-mediated PI(4,5)P2 depletion by applying a + 100 mV/10 s depolarizing pulse in cells co-transfected with ECFP- and EYFP-tagged PH domains derived from PLCδ1 (PHPLCδ1-CFP and PHPLCδ1-YFP). Control cells did not exhibit a ΔFRETr signal change during the depolarizing pulse. However, the ΔFRETr fell with a 0.25 ± 0.03 s time constant (τ) in Dr-VSP-expressing cells and remained at lower levels during the depolarizing pulse. This observation suggests that Dr-VSP activation reduced PI(4,5)P2 concentration at the plasma membrane (Fig. 1d, right).With the 10-s depolarizing step protocol, we tested whether endogenous PAC channel activity was sensitive to membrane PI(4,5)P2 in wild-type (WT) HEK293T cells. Perfusion of the pH 5.5 solution evoked endogenous PAC currents larger than that in the neutral environment of control cells (Supplementary Fig. 4). However, Dr-VSP co-expression almost blocked channel activity at pH 5.5 entirely, suggesting that PI(4,5)P2 depletion by Dr-VSP inhibits the PAC channel. This feature was further examined using PIP5-kinase type-1γ (PIPKIγ), which elevates PI(4,5)P2 levels by phosphorylating phosphatidylinositol 4-phosphate [PI(4)P]. Comparatively, PIPKIγ co-expression slightly increased current density. The current was significantly decreased during Dr-VSP activation, likely due to the inner leaflet’s rapid PI(4,5)P2 resynthesis. We further evaluated the PI(4,5)P2 sensitivity of PAC channels in KO3 cells transfected with hPAC1 or hPAC2. Heterologous hPAC1 and hPAC2 indicated robust currents in control cells without Dr-VSP at pH 5.5 (Fig. 1e). However, Dr-VSP co-expression inhibited the elevation of hPAC1 and hPAC2 currents. The initial weak peak current may occur due to the slow PI(4,5)P2 depletion by Dr-VSP during the depolarizing pulse (Fig. 1d). hPAC1 and hPAC2 current densities at +100 mV were remarkably reduced in cells co-expressing Dr-VSP (Fig. 1f). The Dr-VSP-induced current inhibition of hPAC1 and hPAC2 was also apparent in the pH 5.0 or 4.6 extracellular solution (Fig. 1g).Next, we tested whether PI(4,5)P2 resynthesis after depletion is needed to recover PAC currents from Dr-VSP-induced suppression. Current recovery was monitored by applying +50 mV/0.5 s test pulses with successively longer delays after Dr-VSP activation starting at 0.5 s, as indicated above the traces23. In the presence of intracellular ATP (3 mM), the hPAC1 current reduced by the depolarizing pulse was restored to initial levels at a recovery time constant τ of 3.7 s (Fig. 1h, i). Since PI(4,5)P2 synthesis from PI(4)P requires intracellular ATP, we attenuated PI 5-kinase activity by dialyzing an ATP-free pipette solution into the cell for 10 min. Without ATP in the pipette solution, the maximum current inhibition from depolarization was not affected, but PAC current recovery was slowed (Fig. 1h, i). These findings together support that PI(4,5)P2 is needed for normal PAC channel activity.Phosphoinositide regulation of PAC channel responds specifically to PI(4,5)P2
To further assess whether PAC channel activity is regulated by the plasma membrane’s PI(4,5)P2, we applied the rapamycin-inducible translocation of phosphoinositide phosphatases from cytosol to the plasma membrane’s inner surface. As shown in Fig. 2a, we employed the enzymatic fusion proteins of mRFP-FKBP (RF) to manipulate inositol lipids of the plasma membrane as follows: (1) Sac1, PI(4)P 4’-phosphatase; (2) inositol polyphosphate-5-phosphatase E (INPP5E); (3) pseudojanin (PJ), which possesses both INPP5E and Sac1. Using confocal microscopy, we validated whether lipid phosphatases can adequately function at the intracellular side by co-expressing PHOsh1-GFP and PHPLCδ1-GFP as PI(4)P and PI(4,5)P2 biomarkers, respectively (Supplementary Fig. 5).Fig. 2: Phosphoinositide specificity for regulating hPAC1 currents.All experiments were executed at extracellular pH 5.5 conditions. a Schematic of rapamycin (Rapa)-induced poly-phosphoinositide phosphatase translocation on PAC currents. (left) Metabolic pathway of PI(4,5)P2 synthesis by lipid kinases (4-K and 5-K) and lipid phosphatase breakdown (4-P and 5-P). PI, phosphatidylinositol; PI4P, phosphatidylinositol 4-phosphate. (right) Schematic of rapamycin-induced dimerization of FRB and FKBP proteins. This dimerization recruits poly-PI-metabolizing enzymes to the plasma membrane. RF-Dead is a translocatable construct with inactive mutant Sac1 and INPP5E enzymes. b Representative current traces of hPAC1 from cells expressing RF-Dead, RF-Sac1, RF-INPP5E, or RF-PJ before and after rapamycin. Voltages were stepped from –80 to +80 mV in 20-mV intervals. c Summary of whole-cell current vs. voltage relationship. (black) Before rapamycin; (red) after rapamycin. RF-Dead, n = 6; RF-Sac1, n = 6; RF-INPP5E, n = 5; RF-PJ, n = 8. Data are mean ± SEM. d (left) Time-dependent hPAC1 current amplitude changes by rapamycin were measured at +80 mV every 5 s from cells expressing RF-Dead, RF-Sac1, RF-INPP5E, or RF-PJ. Black lines indicate the addition of 100 nM rapamycin for 3 min. (right) hPAC1 current comparisons before (black) and after (red) rapamycin. e Voltage-independence of rapamycin-induced PAC1 channel inhibition (n = 5 for INPP5E; n = 8 for PJ). Student’s two-tailed unpaired t-test: P values are 0.3042 for INPP5E and 0.4410 for PJ. Data are mean ± SEM. f Summary of hPAC1 current inhibition (%) by rapamycin-induced translocation of RF-Dead, RF-Sac1, RF-INPP5E, or RF-PJ to the plasma membrane. Dots indicate the individual data points for each experiment. Bars are mean ± SEM. ns, not significant, compared to RF-Dead. One-way ANOVA followed by Sidak’s multiple comparisons test: P values are 0.5676 for Sac1, <0.0001 for INPP5E, <0.0001 for PJ, and 0.4044 for INPP5E vs. PJ. Source data are provided as Source Data file.Rapamycin treatment recruited all the translocatable enzymes from the cytosol to the plasma membrane. This system operates via rapamycin-bound FKBP (FK506 binding protein), forming a complex with FRB (the rapamycin-binding domain of mTOR)24. This complex is anchored to the membrane through myristoylation and palmitoylation sites of Lyn kinase’s 11 N-terminal amino acid residues (Lyn11). No PHOsh1 or PHPLCδ1 translocation change was observed in the overexpression of RF-Dead without Sac1 and INPP5E enzyme activity (Supplementary Fig. 5). Recruiting RF-Sac1 or RF-INPP5E increased PHOsh1 and PHPLCδ1 fluorescence intensity in the cytosol, dephosphorylating 4’-phosphate from PI(4)P and 5’-phosphate from PI(4,5)P2, respectively. We also verified that recreating RF-PJ on the plasma membrane released PHOsh1 or PHPLCδ1 into the cytosol.We applied these rapamycin-inducible translocation systems to pH 5.5-induced hPAC1 currents to verify phosphoinositide regulation. Supplementing 100 nM rapamycin did not affect hPAC1 activity in cells expressing RF-Dead or RF-Sac1 (Fig. 2b, c). However, recruiting RF-INPP5E or RF-PJ significantly inhibited hPAC1 current density. As a result of observing the real-time changes of hPAC1 currents in a single cell expressing each enzyme, adding rapamycin inhibited pH 5.5-evoked hPAC1 currents in cells expressing RF-INPP5E or RF-PJ by 72 ± 4% and 82 ± 5%, respectively (Fig. 2d, f). This observation proposes that only PI(4,5)P2, but not PI or PI(4)P, is required to activate hPAC1 channels. Furthermore, these results also show that rapamycin-induced current inhibition of hPAC1 was independent of voltages between +40 and +80 mV (Fig. 2e).To examine the effects of PI(3,4,5)P3 on PAC channels, we applied the chimeric enzyme Ci-VSTPEN15, where the Ci-VSP catalytic domain was replaced with full-length PTEN (a phosphatase and tensin homolog). PTEN dephosphorylates 3’-phosphate of PI(3,4,5)P3. We measured Ci-VSPTEN activity using the FRET approach in cells co-transfected with PH domains derived from Bruton’s tyrosine kinase tagged with ECFP or EYFP (PHBtk-CFP and PHBtk-YFP) (Supplementary Fig. 6a). During the +100 mV/10 s depolarizing pulse, PI(3,4,5)P3 dephosphorylation was proven in cells expressing Ci-VSPTEN, but not in the control group (Supplementary Fig. 6b). With the 10-s depolarizing step protocol, we tested whether the hPAC1 channel is sensitive to membrane PI(3,4,5)P3 depletion. However, no difference in pH 5.5-induced activation of the hPAC1 channel was observed between control and co-expressing Ci-VSTPEN cells (Supplementary Fig. 6c, d). Collectively, we concluded that among the inner leaflet PIs, only PI(4,5)P2 can activate the hPAC1 channel.Intracellular application of PI(4,5)P2 potentiates PAC channels in excised inside-out membrane patchesTo determine whether PI(4,5)P2 alone is sufficient to potentiate PAC channels on the inner side of membranes, we examined the activities of hPAC1 and hPAC2 channels in excised inside-out membrane patches. Both hPAC1 and hPAC2 channels were activated by depolarizing voltage stimulations of more than 50 mV in acidic extracellular pH. Representative current traces evoked by different voltage stimulations from a single excised membrane patch of either hPAC1 or hPAC2 channel are shown in Fig. 3a. However, we were unable to detect single-channel activity from the excised inside-out patches of untransfected PAC channel knockout cells (Supplementary Fig. 7a). The single channel openings of both hPAC1 and hPAC2 channels were highly voltage-dependent, as observed in the macroscopic current traces. Larger depolarization-induced significant increases in the channel open probability (NPO) (Fig. 3b). Like as the whole-cell patch clamp recording of a putative PI(4,5)P2-binding site mutant hPAC1 (R396A), we could not measure voltage-dependent currents of R396A mutant hPAC1 from the excised inside-out patches (three independent patches, Supplementary Fig. 7b). These results suggest that the outwardly rectifying I-V relationships of macroscopic hPAC1 and hPAC2 currents are due to the voltage-dependent gating of channels rather than pore blockade, as cytosolic cellular components were removed in the excised inside-out patch-clamp experiments. Gaussian fitting of channel gating levels (closing and openings) allowed us to assign a unitary channel current at specific membrane voltages. The estimated unitary slope conductance of the hPAC1 and hPAC2 channels were approximately 46.06 pS (n = 5) and 30.23 pS (n = 4), respectively (Fig. 3c). Previously, the chord conductance of a single endogenous PAC channel (PAC2) in HEK293 cells was reported to be 10 pS at +43 mV, 13 pS at +86 mV, and 20 pS at +126 mV25. The extrapolated value of slope conductance from the previous report was approximately 25 pS, which aligned well with our measurements. To confirm that the recorded single-channel activities were from acid-activated Cl− current rather than cationic current, we measured the single-channel activities in either a normal intracellular solution (CsCl and MgCl2) or a cation-replaced solution (NMDG-Cl). The single-channel current amplitudes of both hPAC1 and hPAC2 channels did not significantly differ between the solutions, but the open probabilities decreased under the NMDG-replaced condition, similar to the previous report on the endogenous PAC channel (Supplementary Fig. 8)25.Fig. 3: Single channel activities of human PAC channels and PI(4,5)P2-dependent potentiation of hPAC1 currents in excised inside-out membrane patches.a Single-channel currents and all-point histograms of hPAC1 (left) and hPAC2 (right) channels at the indicated voltage stimulations and pH (external or pipette solution) recorded in inside-out patch clamp configurations. The dashed red lines indicate current levels of closed or open states, and ‘C’ and ‘O’ represent closed and open states, respectively. The number next to ‘O’ indicates the number of open channels. The scales represent 10 pA and 1 s. All-point histograms of single-channel recordings were obtained from 1-minute recordings at the indicated voltage and pH. The voltage-channel open probability (NPo) relationships (b) and unitary current-voltage plot (c) for hPAC1 (open circle, n = 5) and hPAC2 (filled circle, n = 4) channels at pH 5.5. Data are represented as mean ± SEM. d Time-dependent plot of hPAC1 currents in inside-out configuration (left) at pH 5.5. The internal or bath application of 10 μM diC8-PI(4,5)P2 induced a potentiation of PAC1 currents, but the potentiation was inhibited by a PAC channel inhibitor, 4,4’-diisothiocyano-2,2’-stilbenedisulfonic acid (DIDS). hPAC1 currents were elicited by a + 100 mV step pulse for 10 s from holding potential of 0 mV, and the mean currents were obtained from the average values of the last 10 s. The currents were recorded every 0.5 s. Right, representative current traces (numbers in parentheses, right) show hPAC1 currents at arrow-pointing spots. Source data are provided as Source Data file.We then examined the effect of intracellular PI(4,5)P2 on the hPAC1 channel by applying diC8-PI(4,5)P2 to the internal side of excised inside-out membrane patches (Fig. 3d, e). After forming an inside-out patch, only a few hPAC1 channel openings were observed at the single-channel level, but the application of 10 μM diC8-PI(4,5)P2 significantly increased ionic currents to a macroscopic current level of ~30 pA. The treatment of a PAC channel inhibitor, 4,4’-diisothiocyano-2,2’-stilbenedisulfonic acid (DIDS)25, abolished PAC current potentiation even in the presence of 10 μM diC8-PI(4,5)P2 (Fig. 3d). The initial low-level activity of the hPAC1 channel might be due to the rapid rundown of the hPAC1 channel caused by dissociation or degradation of PI(4,5)P2 in the isolated membrane patch.Identification of residues affecting PI(4,5)P2 sensitivity in the intracellular side of PAC channelsStructural analysis has revealed that fundamental residues like lysine and arginine around the membrane-cytosol interface usually interact with PI(4,5)P2 in functionally active ion channels16,17,26. Thus, we conducted alanine-scanning mutagenesis of positively charged residues, including K120 and K124 in the N-terminus and K390, K394, R396, K397, R398, K401, R402, and R403 located in the TM2 cytoplasmic interface of hAPC1 (Fig. 4a). All hPAC1 mutant channels, including the WT hPAC1, were co-localized with Lyn-mCh at the plasma membrane, as indicated by yellow in the merged images (Fig. 4b). Quantifying the co-localization of PAC channels with Lyn-mCh showed non-significant difference in Pearson’s coefficient between WT and mutant channels (Fig. 4d)27. Of the ten single amino acid mutant forms, the pH 5.5-induced current amplitude at +100 mV was dramatically reduced in K394A, R396A, and K397A compared to WT without Dr-VSP expression (control) (Fig. 4c, e). Similarly, we could not measure any significant voltage-dependent currents of R396A mutant PAC1 from the excised inside-out patches (Supplementary Fig. 7b). Triple mutation of three effective basic residues (K394A/R396A/K397A) did not affect the channel’s plasma membrane distribution but predominantly abolished the hPAC1 currents in acidic pH ranges down to 5.0 (Fig. 4f, g). With Dr-VSP, there was minor suppression of currents in the triple mutant form; however, the additional suppression of whole-cell current by Dr-VSP was even found in PAC-KO cells (see Fig. 1f). We also conducted alanine substitution mutations of the positively charged residues K59, K63, K329, K333, R335, K336, R337, K340, R341, and R342 in the hPAC2 channel, corresponding to hPAC1 residues (Supplementary Fig. 9a), and identified cellular distribution and channel activity. Mutant hPAC2 construct distribution was consistent with the WT hPAC2 expression pattern (Supplementary Fig. 9b). Similar to hPAC1, K333, R335, and K336 alanine mutations of hPAC2 substantially impaired channel activity (Supplementary Fig. 8c, d). Moreover, triple mutation (K333A/R335A/K336A) of hPAC2 channel retrained the current generation and Dr-VSP suppression without affecting the cellular distribution. These results suggest that those three basic amino acids in the TM2 domain’s distal end are important for regulating the PI(4,5)P2 regulation in both hPAC1 and hPAC2 channels.Fig. 4: Identification of residues critical for hPAC1 PI(4,5)P2 regulation through alanine scanning.a Sequence of positively charged residues clustering in hPAC1. Purple lines indicate N- and C-terminal sections. Green lines indicate transmembrane domain (TM) 1 and 2 sections. Positively charged residues indicated in blue are selectively mutated to alanine (A). b Representative confocal images of KO3 cells expressing wild-type (WT) or alanine-substituted hPAC1 mutants with the plasma membrane marker Lyn-mCh. Scale bar: 10 μm. Scatter plots in right panels display a 2D intensity histogram between hPAC1-GFP and Lyn-mCh pixels in the confocal image. Each experiment was repeated 3 times independently. c Representative current traces of WT or mutant channels elicited at +100 mV in the control (left) and cells expressing Dr-VSP (right). d Pearson’s correlation coefficients (R) for the colocalization between hPAC1-GFP and Lyn-mCh confocal images. n = 6 for WT; n = 10 for K120A, K390A, K397A, R398A, K401A; n = 9 for K124A, K394A, R396A, K394/R396A/K397A, K402A, K403A. Dots indicate the individual data points for each cell. Data are mean ± SEM. e hPAC1 current densities at +100 mV for control and cells expressing Dr-VSP in pH 5.5. The dashed gray line indicates the standard WT level. n = 6 for WT, K394A, K402A; n = 7 for K120A; n = 4 for K124A; n = 5 for K390A, R396A, K397A, K394/R396A/K397A, R398A, K403A; n = 3 for K401A. Data are mean ± SEM. ns, not significant compared to WT. One-way ANOVA followed by Dunnett’s multiple comparisons test. P values: 0.9993, K120A; 0.1938, K124A; 0.9994, K390A; 0.0009, K394A (***P < 0.001); <0.0001, K396A (****P < 0.0001); 0.0103, K397A (*P < 0.05); <0.0001, K394/R396/K397A (****P < 0.0001); 0.8065, K398A; 0.9938, K401A; 0.9999, R402A; 0.9249, R403A. f Families of hPAC1 current of WT (left) and K394A/R396A/K397A mutant (right) elicited by the voltage steps shown at the top under different extracellular pH condition. g pH-dependent hPAC1 current measured at +100 mV in cells expressing hPAC1 WT (n = 6 for pH 7.4, 5.8, 5.5; n = 5 for pH 6.3; n = 7 for pH 5.3, 5.0) or K394A/R396A/K397A (n = 5 for pH 7.4, 6.3, 5.8, 5.5, 5.3, 5.0). Data are mean ± SEM. Source data are provided as Source Data file.Verification of PI(4,5)P2-PAC interaction from structural modelingThe PAC channel’s cryo-EM structure can be divided into resting, activated, and desensitized states10. Since this structure has been identified in the PAC2 channel, we investigated K333, R335, and K336 residue positions in the hPAC2 channel. Through the simulation using I-TASSER and the cryo-EM structure, we obtained the resting state conformation of the hPAC2 channel. The Ramachandran plot of the simulation was strikingly similar to that of hPAC2’s cryo-EM structure (Supplementary Fig. 10a). The model revealed that the three fundamental residues (K333, R335, and K336) and their side chains were only located in the cytosolic region (Supplementary Fig. 10b). However, we determined that the three basic residues were translocated and resided at the membrane–cytosolic interface in both activated and desensitized conformations (Fig. 5a).Fig. 5: Identification of PI(4,5)P2 coordination in the hPAC2 channel through structural modeling.a Cryo-EM structure models of resting, activated, and desensitized PAC2 channel conformations (Protein Data Bank ID: resting, 7SQG; activated, 7SQF; desensitized, 7SQH). Two black lines indicate the outer and inner plasma membrane surfaces. Blue, red, and green boxes indicate the location of putative PI(4,5)P2-binding residues K333, R335, and K336, respectively. b Molecular docking between K333, R335, and K336 of the activated chain A conformation and PI(4,5)P2 (top) or PI(4)P (bottom). (left) Autodock Vina and UCSF Chimera were used for GIRK PI(4,5)P2 or PI(4)P to the cryo-EM structure of activated PAC2. (right) Nine conformation superpositions of PI(4,5)P2 or PI(4)P. c Docking between K333, R335, and K335 of the desensitized chain A conformation and PI(4,5)P2. (top) Autodock Vina and UCSF Chimera were used for GIRK PI(4,5)P2 to the cryo-EM structure. (bottom) Nine conformation superpositions of PI(4,5)P2. d Hypothetical model of a cytosolic PI(4,5)P2 binding site depending on resting, activated, or desensitized PAC channel conformation.During the structural rearrangement from the resting to the activated state, the TM2 domain is raised in the extracellular direction by 10Å10. This conformational change in the activated form translocates K333, R335, and K336A to the plasma membrane’s inner surface. With this structural feature, we assessed the PI(4,5)P2 coupling of K333, R335, and K336 residues in activated and desensitized hPAC2 channels via molecular docking simulation. Using the PI(4,5)P2 molecule extracted from the G protein-gated inwardly rectifying potassium (GIRK) K+ channel28, we identified three salt bridges between the docked PI(4,5)P2 molecule and the putative PI(4,5)P2 binding sites in activated form: 1) R335 side chains interact with 5’-phosphate of PI(4,5)P2 through two salt bridges, and 2) 4’-phosphate of PI(4,5)P2 coordinates the K336 residue through another salt bridge (Fig. 5b top). For PI(4)P, the charge–charge interaction between 5’-phosphate and R335 disappeared, whereas the interaction between 4’-phosphate and K336 remained (Fig. 5b bottom).Although PI(4,5)P2 sensitivity was not confirmed in the desensitized PAC channel, we simulated molecular docking to examine PI(4,5)P2 binding to the three positively charged residues in a desensitized state. The docked PI(4,5)P2 has two salt bridges with the R335 side-chain in a desensitized conformation (Fig. 5c). In the hPAC2 resting conformation at natural pH, even the basic residues were located in the cytosol away from the plasma membrane. Our docking simulation indicated that PI(4,5)P2 does not interact with the K333, R335, or K336 side-chain (Supplementary Fig. 10c). Together, these results suggest that the hPAC2 channel in the resting state does not interact with PI(4,5)P2, but can couple with the cytoplasmic leaflet PI(4,5)P2 when the TM2 domain is outwardly shifted after activation (Fig. 5d). The PI(4,5)P2 coupling of the TM2 domain remains stable during the desensitized state.The PAC channel is modulated by Gq-coupled muscarinic receptorsThe plasma membrane PI(4,5)P2 level is dynamically regulated by signaling pathways, and Gq-GPCR activation reduces PI(4,5)P2 by hydrolyzing it into DAG and IP3 via PLCβ. First, we assessed whether Gq-GPCR can function in a pH 5.5 acidic environment by monitoring PHPLCδ1-GFP translocation, a PI(4,5)P2 probe in KO3 cells co-transfected with the M1 muscarinic receptor (M1R). As shown in Fig. 6a, M1R stimulation with oxotremorine-M (Oxo-M) translocated PHPLCδ1-GFP from the plasma membrane to the cytosol at pH 7.4 and 5.5. However, PHPLCδ1-GFP’s elevated cytosolic intensity at pH 5.5 was slightly lower than at pH 7.4 (Fig. 6b), likely due to the low agonist binding affinity to muscarinic receptors in acidic pH conditions29,30. To test whether receptor activation can modulate PAC channel activity, KO3 cells expressing hPAC1 or hPAC2 with M1R were stimulated with Oxo-M, which considerably and reversibly reduced pH 5.5-induced hPAC1 and hPAC2 currents (Fig. 6c). The maximum inhibitions by M1R activation were 84 ± 4 % and 87 ± 3 % with the time constants τ = 6 ± 2 s and τ = 5 ± 1 s for hPAC1 and hPAC2, respectively (Fig. 6d, e).Fig. 6: Muscarinic modulation of PAC channels in HEK293T or SH-SY5Y cells.a Live confocal images of PHPLCδ1-GFP in HEK293T cells transfected with M1R before (control) and during ( + Oxo-M) application of 50 μM Oxo-M at pH 7.4 (black) or 5.0 (red). Scale bars are 10 μm. b Time-dependent changes in cytosolic fluorescence intensity by PHPLCδ1–GFP translocation upon M1R activation (n = 10, pH 7.4; n = 12, pH 5.5). Data are mean ± SEM. c (left) PAC current modulation by M1R activation with a 15-s Oxo-M application (50 μM) in cells expressing hPAC1 (top) or hPAC2 (bottom). Current amplitudes were measured at +100 mV every 5 s. (right) Representative current traces at different time points in time course recording. d Percent inhibition of pH 5.5-induced PAC currents by 50 μM Oxo-M (n = 6 for hPAC1; n = 5 for hPAC2). Data are mean ± SEM. e Time constants (τ) of current inhibition by muscarinic simulation in hPAC1 (n = 6) and hPAC2 (n = 5). Dots indicate the individual data points for each cell. Data are mean ± SEM. Student’s two-tailed unpaired t-test: P value, 0.5674. f Identifying PAC gene expression via RT-PCR analysis with RNA extracted from wild-type HEK293T or SH-SY5Y cell lines. The experiment was repeated 4 times independently with similar results. g Representative voltage-dependent whole-cell currents of endogenous PAC channels in SH-SY5Y cells under pH 7.4, pH 5.5, and pH 5.0 solutions. h Current density-voltage relationship of endogenous PAC in SH-SY5Y cells at the indicated pH values (n = 5, pH 7.4; n = 5, pH 5.5; n = 4, pH 5.0). Data are mean ± SEM. i, Endogenous PAC current modulation by muscarinic stimulation with a 15-s Oxo-M application (30 μM) in SH-SY5Y cells at pH 5.5 (blue) or 5.0 (red). PAC current amplitude was measured at +100 mV every 5 s. j Percent (%) inhibition of pH 5.5- or 5.0-induced endogenous PAC currents by 30 μM Oxo-M (n = 8 for pH 5.5; n = 6 for pH 5.0). Dots indicate the individual data points for each cell. Data are mean ± SEM. Student’s two-tailed unpaired t-test: P value, 0.0013 (**P < 0.01). Source data are provided as Source Data file.PAC channels are ubiquitously expressed in all tissues and cell types11,31,32,33. According to the Human Protein Atlas (https://www.proteinatlas.org/), RNA expression of the PAC channel is the highest in the SH-SY5Y cell line, a human neuroblastoma (Supplementary Fig. 11). We compared hPAC1 and hPAC2 gene expressions in HEK293T and SH-SY5Y through RT-PCR analysis using sequence-specific designed primers (Supplementary Table 1). Both cell types expressed hPAC2 (but not hPAC1) with a relatively higher density in SH-SY5Y cells (Fig. 6f). We then examined innate PAC2 channel regulation by the endogenous M3 muscarinic receptor (M3R) in SH-SY5Y cells34. There were potent Cl- currents in the acidic extracellular solution with outward rectification at positive membrane potentials (Fig. 6g, h). Oxo-M treatment at pH 5.5 and pH 5.0 diminished endogenous PAC currents by 79 ± 1% and 68 ± 2%, respectively (Fig. 6i, j). These results demonstrate that PAC channels can be modulated by activating endogenous Gq-GPCRs in native cells.
The plasma membrane inner leaflet PI(4,5)P2 is essential for the activation of proton-activated chloride channels
