Identification of a unique Ca2+-binding site in rat acid-sensing ion channel 3

Acid-sensing ion channels (ASICs) evolved to sense changes in extracellular acidity with the divalent cation calcium (Ca2+) as an allosteric modulator and channel blocker. The channel-blocking activity is most apparent in ASIC3, as removing Ca2+ results in channel opening, with the site’s location remaining unresolved. Here we show that a ring of rat ASIC3 (rASIC3) glutamates (Glu435), located above the channel gate, modulates proton sensitivity and contributes to the formation of the elusive Ca2+ block site. Mutation of this residue to glycine, the equivalent residue in chicken ASIC1, diminished the rASIC3 Ca2+ block effect. Atomistic molecular dynamic simulations corroborate the involvement of this acidic residue in forming a high-affinity Ca2+ site atop the channel pore. Furthermore, the reported observations provide clarity for past controversies regarding ASIC channel gating. Our findings enhance understanding of ASIC gating mechanisms and provide structural and energetic insights into this unique calcium-binding site.

and mASIC3, 82582993), human (hASIC1a, 21536349; hASIC2a, 1280439; hASIC2b, 21739677; hASIC3, 3747101; and hASIC4, 8346834), shark (sASIC1a, 63054896; and sASIC1b, 63054898) and lamprey (lASIC1, 63054894) were aligned using CLUSTALW 1 . The alignment was visualized by ESPript 2 . Conserved residues are shown as white text on a red background, and similar residues are as red text on a white background; the cASIC1 Arg65 and Glu426 and the corresponding residues from other members are additionally highlighted in yellow background. The secondary structure of cASIC1 derived from the crystal structure (PDB code: 4NYK 3 ) is marked on the top of the sequence alignment.

Supplementary Figure 2 | Extracellular view of the transmembrane (TM) domain
showing the rotation of TM helices (upper panels) and side view of the TM domain illustrating the water density profile along the channel pore (lower panels). Panels a and b illustrate the results for the mutant simulations with or without Ca 2+ bound; Panels c and d correspond to the wild type simulations with or without Ca 2+ bound. In the upper panels, the pink and blue cylinders are with the two crystal structures in the desensitized and open states (PDB codes: 4NYK and 4NTW), respectively, while the magenta ones denote the representative conformations from the simulations. Note that for TM2, only the extracellular portion (TM2a) is displayed for clarity. The rotation direction and magnitude are indicated by a curved arrow. The water density maps shown in the lower panels are colored yellow. The Ca 2+ and Na + ions are depicted as green and cyan spheres, respectively; note that in the Ca 2+ -free mutant system (panel b), the Na + ions residing within the extracellular vestibule are masked by the contour map. In cationic dummy atom model (or multi-site ion model) (a), the metal core is covalently connected to a certain number of cationic dummy atoms with a fractional charge of +δ (colored white) and the total charge is equivalent to the formal charge carried by the single-atom ion. For a perfect octahedral coordination, the angles Θ/Φ (12 in total) and Ω (3 in total) are equal to 90º and 180º, respectively, while the pentagonal bi-pyramidal geometry includes three distinct angles, Θ (5 in total), Φ (10 in total) and Ω (1 in total), with an ideal value of 72º, 90º and 180º, respectively.

Supplementary Note 1: Comparison of the two Ca 2+ models
In this study, two distinct Ca 2+ models 4,9 were used in the simulations of Ca 2+ -bound channel systems within the context of the additive CHARMM or AMBER force fields (Methods). One is the most common point charge model, which describes the ion as a formal point charge of +2 that interacts with the ligands through non-bonded interactions.
The other one is the more complicated cationic dummy atom model (originally developed by Åqvist and Warshel 9,10 ), which represents partially covalent and partially electrostatic nature of the coordinative bond by locally splitting up the space between the metal ion and the ligand into a covalent bond (between the metal core and the cationic dummy atom) and an electrostatic interaction (between the cationic dummy atom and the partially negatively charged ligand). Despite the different development strategies, our simulation results showed that the two Ca 2+ models produced high structural homogeneity at the channel binding site. In the case of the mutant system, the two differently represented ions exploited the same numbers of the carboxylate oxygens and the water oxygens for an optimal coordination (Figs. 3c-3d; Supplementary Table 2), though the specific spatial arrangements of these ligating oxygens show a difference. In the wild system, the two models achieved essentially the same coordination patterns inside the channel pore (Figs. Table 2). Moreover, comparable pore openings were observed with the two models in both the mutant and wild type channels (Supplementary Figs. 8b   and 8d).

3c-3d; Supplementary
Yet, our MM-GBSA calculations displayed a difference of ~160 kcal/mol between the Ca 2+ binding free energies estimated from the two models for either the mutant and wild type systems. The discrepancy is mainly caused by the distinct electrostatic description used in the two models. It should be noted that, herein we are most interested in the overall trend of the Ca 2+ binding affinity changes as a result of varied protonation state within the same system as well as in the relative change in binding affinity due to mutation, rather than seeking the ideal absolute values. With the dummy atom model, the simulation results demonstrated an overall progressive reduction in the Ca 2+ binding strength with increasing protons at the block site, supporting the previous argument that Ca 2+ and proton compete for the binding site 11 . In addition, both the Ca 2+ models showed an enhanced Ca 2+ binding affinity through the G429E mutation, as we expected. The dummy model gave an increase of ~20% (Supplementary Fig. 6), as compared to that of ~40% with the point charge model (data not depicted). Again, we reason that this difference is related to the different electrostatic treatments between the two models. Yet, it does not qualitatively influence our conclusions.
The Ca 2+ dummy model used herein was developed against the most common pentagonal bi-pyramidal geometry in biological systems 5 , involving a coordination number (CN) of 7 (Supplementary Fig. 4a). In our simulations, interestingly, we discovered that this model is also capable of forming the 6-lignad or 8-ligand coordination ( Fig. 3c; Supplementary Fig. 8a). The observations thus highlight the geometric flexibility of the Ca 2+ dummy model and its ability in simulating the process involving CN changes, as also indicated in other studies 12,13 .