If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Key Laboratory of Environmental Biotechnology, Fujian Province University, Xiamen University of Technology, Xiamen, 361024 ChinaDepartment of Pharmaceutical Sciences, Northeastern University School of Pharmacy, Bouve College of Health Sciences, Boston, Massachusetts 02115
G protein–gated inwardly rectifying K+ (GIRK) channels belong to the inward-rectifier K+ (Kir) family, are abundantly expressed in the heart and the brain, and require that phosphatidylinositol bisphosphate is present so that intracellular channel-gating regulators such as Gβγ and Na+ ions can maintain the channel-open state. However, despite high-resolution structures (GIRK2) and a large number of functional studies, we do not have a coherent picture of how Gβγ and Na+ ions control gating of GIRK2 channels. Here, we utilized computational modeling and all-atom microsecond-scale molecular dynamics simulations to determine which gates are controlled by Na+ and Gβγ and how each regulator uses the channel domain movements to control gate transitions. We found that Na+ ions control the cytosolic gate of the channel through an anti-clockwise rotation, whereas Gβγ stabilizes the transmembrane gate in the open state through a rocking movement of the cytosolic domain. Both effects alter the way in which the channel interacts with phosphatidylinositol bisphosphate and thereby stabilizes the open states of the respective gates. These studies of GIRK channel dynamics present for the first time a comprehensive structural model that is consistent with the great body of literature on GIRK channel function.
). They are expressed abundantly in the heart (GIRK1 and GIRK4) and in the brain (all members), among other tissues, where they function both as homotetramers and/or heterotetramers with each other. As their name implies, physiologically they are directly gated by G proteins, specifically the Gβγ dimer of GTP-binding (G) proteins (Gβγ) (
), ones that associate with pertussis toxin–sensitive Gα subunits. GIRK channels are dependent on phosphoinositides, especially PIP2, which via direct interactions stabilize their gates in the open state (
Activation of the atrial KACh channel by the βγ subunits of G proteins or intracellular Na+ ions depends on the presence of phosphatidylinositol phosphates.
Proc. Natl. Acad. Sci. U.S.A.1998; 95 (9448327): 1307-1312
). Another physiological regulator of GIRK activity is intracellular Na+ that is in part coordinated by a specific Asp residue found in the GIRK2 and GIRK4 channel members (
). Both Na+ and Gβγ have been shown to work by increasing the affinity of the channel to PIP2 and to synergize via this mechanism in activating the channel, whereas PIP2 itself is unable to gate efficiently the channel on its own, unlike other Kir channel family members (
). GIRK activation inhibits excitability, slowing the rate of pacemaker and atrial cell firing in the heart, while inhibiting transmitter release by presynaptic neurons or opposing excitation of postsynaptic neurons. GIRK inhibition through hydrolysis or dephosphorylation of PIP2 causes channel desensitization that lasts as long as it takes to resynthesize PIP2 (
). GIRK channels are thought to be good therapeutic targets for multiple conditions that call for a decrease in excitability or arrhythmogenesis, such as epilepsy in the brain and atrial fibrillation in the heart (
Structural studies using crystallography or computational modeling have produced 3D models of GIRK channels in complex with each and all of these physiological regulators, PIP2, Gβγ, and Na+ (
). Atomic resolution crystal structures of GIRK channels have been limited to a truncated GIRK2 channel construct that reproduces functional characteristics of the full-length channel (
). The GIRK2 structures have provided direct evidence for the multiple gate hypothesis of these channels. These structures point to three constrictions along the permeation pathway: the selectivity filter, where K+ ions are selected over other ions; the helix bundle crossing (HBC) gate, also referred in the literature as the inner helix gate, located near the inner leaflet of the plasma membrane bilayer; and a cytosolic gate that is unique to Kir channels, referred to as the G-loop gate (Fig. S1). The HBC gate is comprised by the side chains of Phe-192, whereas the G-loop gate is comprised by two methionine pairs at the top and bottom of the G loop (Fig. S2B).
The GIRK2 structures in complex with the various regulators, obtained mainly from the MacKinnon lab, have provided a wealth of structural information. However, no satisfying structural gating model has resulted that is consistent with the physiological regulation of the channel. None of the structures with physiological regulators have captured the channel gates in the open state. Even the channel bound to all its gating molecules, PIP2, Na+, and Gβγ, which in electrophysiological experiments has been found to activate the channel the most (
), did not capture the channel gates in the open state. The structure of a low activity point mutant, GIRK2 (R201A), has come the closest to capturing the gates open, where two of the subunits bound to PIP2 depicted open conformations, but the remaining two that were not PIP2 bound had their gates closed (
). It was hypothesized that the crystal packing prevented PIP2 from binding to all four subunits, leaving two of the four gates in the closed state. Based on this structure, models were created in an effort to understand what structural changes could be predicted, if all subunits would reach the open state, as expected. The binding of Gβγ was predicted to cause an anti-clockwise movement of ∼4° of both the cytoplasmic and transmembrane (TM) domains that put the channel in a “pre-open” state. The R201A mutant channel showed an additional ∼4° twisting movement to cause full gate opening (
). Table S1 and Fig. S3 show all the crystal structures determined under different conditions with both minimal and α-carbon (Cα) distances between the gates of opposite-facing subunits. This heroic work has answered many questions, but at the same time has left many other questions unanswered. It has also generated puzzling conclusions, perhaps the most troubling of which is the question: does the low activity R201A mutant of the GIRK2 channel represent a valid model of a constitutively active channel that mimics the natural open channel state? Critical mechanistic questions remain, such as which gates are controlled by the two different activators (Na+versus Gβγ or both) and how the movements caused by these regulators tie into the regulation of the gates by PIP2. Given the nature of these dynamic questions, we pursued them using computational modeling and MD simulations based on the pre-open crystal structure of GIRK2.
Results
Na+ or Gβγ couple to distinct gates but when present simultaneously show synergism in gating the channel
Five modeled systems ranging from 160,000 to 400,000 atoms were simulated in the absence and presence of an electric field (EF) for a total of 1.0–1.5 μs, respectively (Table S2). The largest system with all the components is illustrated in Fig. S2A. K+ ion permeation is taken as the criterion to designate a channel conformational state under specific conditions as “closed” or “open.” In the GIRK2–Apo system, no ion permeation was seen regardless of whether an EF was applied or not. In contrast, in all other systems of the GIRK2 channel, such as with PIP2 (GIRK2–PIP2 or GIRK2), with PIP2 + Na+ (GIRK2–Na+), with PIP2 + Gβγ (GIRK2–Gβγ), and with PIP2 + Gβγ + Na+ (GIRK2–Gβγ–Na+), K+ ions permeated, albeit inefficiently, even in the presence of EF across the membrane (0.06 V/nm or Vm = −200 mV). However, in the GIRK2–Gβγ–Na+ system, an almost 10-fold greater number of ions permeated under the transmembrane EF (Table 1). The relative permeation seen in these simulations agrees with electrophysiological results (
). Moreover, in the absence of an external electric field, the MD simulations also showed an outwardly directed flow of K+ (intracellular to extracellular) (data not shown).
Table 1The number of K+ for each system that got through the GIRK2 channel and the channel state without or with EF in the last snapshot of MD simulations
To better understand the gating mechanism, we set the production stage in the presence of an EF, because gating proved more efficient under these conditions (Table 1). All systems were equilibrated by 150 ns of simulation as seen in the root-mean-square deviation plots (Fig. S4). The average distance required to occlude permeation was 5.69 Å and was derived from the GIRK2–Na+ system, which has been shown experimentally to be the least conductive (
). When both the HBC and G-loop gates exceeded the minimal distance of 5.69 Å, the channel was considered to be in the open state. As designated by vertical lines in Fig. 1, this representation of activity would correspond to the open probability (Po) of the channel in the nanosecond time scale, even though not every time the channel was permeable did an ion in fact permeate. From this simplified analysis, it became apparent that Na+ or Gβγ subunits alone could open GIRK2 and could do so in a synergistic manner when applied together (Fig. 1). In fact, the synergism between the Gβγ subunits and Na+ can be seen in records of single channel activity (cell-attached recordings) of the highly related GIRK4 channel. The gating behavior of certain GIRK4 mutants that mimic the Na+ (i.e. I229L), Gβγ (i.e. S176P), or Na+ + Gβγ (i.e. I229L + S176P) conditions (Fig. 2) (
) bears remarkable pattern similarity to the simulated records of Po, even though the time scales separating these records are at least 6 orders of magnitude apart.
Figure 1Overall GIRK2 channel burst characteristics (each vertical line per ns) over simulation time. MD simulations were run on five systems: GIRK2–Apo, GIRK2–PIP2 (GIRK2), GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+. Channels with their HBC and G-loop gates larger than a minimum distance (5.69 Å) that is required for permeation were considered to be open. This minimum distance estimate was derived from the least conductive system GIRK2–Na+ compared with the other two systems studied that have been shown experimentally to display measurable currents (i.e. GIRK2–Gβγ and GIRK2–Gβγ–Na+).
Figure 2Unitary activity of GIRK4* mutants (using as control the highly active GIRK4-S143T). GIRK4* mutants mimic gating by Na+ (I229L), Gβγ (S176P), and Gβγ/Na+ (S176P and I229L) in compressed (A) and more expanded (B) time scales. Open probability of S176P (0.072 ± 0.019; n = 6), I229L (0.059 ± 0.014; n = 5), and S176P/I229L (0.322 ± 0.093; n = 7) are shown. These data obtained from mutants representing the endogenous gating molecules (PIP2, Na+, Gβγ, and Na+/Gβγ) are consistent with prior reports (
Na+ ions predominantly stabilize the G-loop gate in the open state
The GIRK2–PIP2 system revealed that although the HBC gate distances were distributed around the minimal distance for permeation, the G-loop gate distances became limiting (Fig. 3), explaining the inefficient open probability seen in Fig. 1. The open probability of each of the two gates individually (HBC and G loop) is shown in Fig. 4. Once Na+ was included in the GIRK2–PIP2 system, there was a clear right shift of the G-loop gate minimum distributions to the open state (Fig. 3). This effect on the G-loop gate was in sharp contrast to the HBC gate, whose minimal distance distributions were not affected by Na+, remaining centered around the nonpermissive 5.69 Å distance (Fig. 3). The G-loop gate Po increased by more than 7-fold, whereas the HBC gate Po, if anything, was somewhat decreased (Table 2 and Fig. 4). Overall, the Po of both gates was simultaneously increased by ∼6-fold (Fig. 1 and Table 2). As a result, the HBC became the limiting gate to ion flow in the presence of intracellular Na+ ions.
Figure 3Minimal distances for HBC and G-loop gate opening. Histograms of distributions of minimum distances of the HBC gate (left column) and the G-loop gate (right column) based on the simulation (per ns) of the MD trajectory. The same five systems as in Fig. 1 were analyzed, namely GIRK2–Apo, GIRK2–PIP2 (GIRK2), GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+. The red dashed line indicates the cutoff distance under which no K+ permeation took place.
Figure 4GIRK2 channel burst characteristics for individual gates (each vertical line per ns) over simulation time. MD simulations were run on five systems: GIRK2–Apo, GIRK2–PIP2 (GIRK2), GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+. Channels with their HBC (A) and G-loop (B) gates larger than a minimum distance (5.69 Å) that is required for permeation were considered to be open. The minimum distance estimate was derived as in Fig. 1 from the least conductive system GIRK2–Na+ compared with the other two systems studied that have been shown experimentally to display measurable currents (i.e. GIRK2–Gβγ and GIRK2–Gβγ–Na+).
Table 2The open probability (%) of the constrictions HBC, G-loop gate, and both for the last 350 ns that is in the equilibrium stage of MD in each simulation system
Gβγ dimer acts to predominantly stabilize the HBC gate in the open state
A clear right shift of the HBC gate minimum distributions to the open direction was seen when Gβγ was included in the GIRK2–PIP2 system (Fig. 3). The effect of Gβγ on the G-loop gate also showed a right shift (compared with GIRK2–PIP2), but there was a significant component of distance distributions that Gβγ was not able to affect (Figure 3, Figure 4 and Table 2). Opening of the G-loop gate may become a rate-limiting step to the conductive state when channels get activated by Gβγ. Because this limitation was less than that imposed by the HBC gate in the GIRK2–Na+ system, a higher overall Po of the channel was achieved (Fig. 1 and Table 2). This observation is consistent with experimental results suggesting that in comparison, gating by Gβγ is greater than that by intracellular Na+ (
Na+ and Gβγ subunits work together to activate the channel the most
When both Gβγ and intracellular Na+ ions were included in the simulations, the distance distributions of both gates were predominantly in the open state (Fig. 3), resulting in the greatest Po (Fig. 4 and Table 2). In the presence of an EF, this condition produced the most efficient permeation of K+ ions (Table 1). These data are in good agreement with experimental results (Fig. 2) (
Rocking, tilting, and rotation movements of GIRK2 domains lead to channel activation
As discussed earlier, the only GIRK2 crystal structure depicting both the HBC and G-loop gates in the open state has been the GIRK2(R201A) + PIP2 (3SYQ) structure (
), casting doubt regarding whether this mutant provides a meaningful model of the channel open structure. Principal component analysis (PCA) was performed on our simulated trajectories to discern channel domain movements linked to gating. This analysis suggests that the HBC and G-loop gates move in opposite directions as the channel gates (Movie S1). This kind of movement is in contrast to the proposal based on the 3SYQ open structure (R201A-PIP2), in which the two gates move in the same direction. In fact, when we performed PCA on the crystal structures excluding the 3SYQ structure, similar results to our simulations were obtained, supporting a movement of the two gates in opposite directions (Movie S2). When PCA was repeated including the 3SYQ (GIRK2-R201A-PIP2), the two gates were shown to move in the same direction as previously suggested (
). These results as a whole have suggested that the GIRK2(R201A) + PIP2 mutant involves perhaps a distinct reaction path and mechanism leading to the conformational state depicted by the crystal structure and may not represent a good model of the channel open state achieved by natural gating molecules.
Movements controlling HBC gating
Once the channel was gated from a pre-open to the open state, rocking movements of the CTD and tilting of the TM2 were observed (Movie S1). To quantitatively depict these movements two angles were defined: (a) a dihedral angle (TM2–CTD) formed by the Cα atoms of Ser-196 with Ile-244 and Ile-281 on one hand and with Gln-287 on the other hand and (b) a TM2 tilt angle formed by the TM2 terminal and the vertical axis of the TM domain (Fig. S5A). The larger the TM2–CTD dihedral angles the more the CTD rocks outwardly, away from the vertical axis of the TM domain. Comparison of the closed and open structures reveal an outward CTD rocking movement during opening, an outward TM2 helix tilting and local twists in the Gβγ binding βL–βM and βD2–βE1 loops (Fig. 5). 3D plots were then employed to compare the relationship between the dihedral angle, the tilt angle, and the HBC Cα distance. Fig. 6A shows clearly that as the TM2–CTD dihedral angles and the TM2 tilt angles increase (move outwards), so do the HBC Cα distances. Generally speaking, increases in both the TM2–CTD dihedral and the TM2 tilt angles translate into opening the HBC gate.
Figure 5GIRK2 activation mechanism schemes: the closed and open structures are colored yellow and green, respectively, with both HBC and G-loop gates labeled. CTD rocking, TM2 helix tilting, and the local twists of the βL–βM and βD2–βE1 loops are indicated by the black/red arrows. A, view of two opposing subunits with the front and back subunits removed for clarity. B, 90° counter-clockwise rotation showing the two loops (βL–βM and βD2–βE1) that form the cleft where one Gβγ subunit binds each channel subunit individually.
Figure 6Parameters describing HBC and G-loop gate opening.A, 3D plots of TM2 helix (Val-188–Ser-196) tilt angles, TM2–CTD dihedral angles (Cα atoms of Ser-196, Ile-244, Ile-281, and Gln-287), and average HBC Cα distances. B, 3D plots of TM2–CTD dihedral angles, relative rotation angles of TM2 and CTD, and average G-loop gate minimal distances. The arrow indicates the direction of increasing rotation angles and decreasing TM2–CTD dihedral angles that result in increasing G-loop minimal gate distances, best exemplified by the GIRK2–Na+ system.
It has been proposed that a four-degree anti-clockwise rotation (of the CTD relative to the TMD and as viewed from the membrane) will allow transition of the closed (3SYA or GIRK2–PIP2 + Na+) to the preopen (4KFM or GIRK2–PIP2 + Gβγ + Na+) conformation (
). We used the geometry center of the TMD and CTD excluding the N and C termini, as mentioned under “Experimental Procedures,” to calculate this angle and estimated it to be ∼3.68° (Fig. 7A), consistent with the previously proposed angle of 4° (
). However, no correlation was found between rotation angles and HBC Cα distances in the conductive systems (GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+) that showed progressive opening of the HBC gate (Fig. 7B). In other words, we were unable to correlate the HBC gate opening to an anti-clockwise rotation activation mechanism alone.
Figure 7Movements during the opening of the HBC gate.A, the calculated anti-clockwise rotation angle of CTD of the X-ray crystal structures 3SYA (closed) and 4KFM (pre-open). The estimated 3.68° is consistent with the previously proposed angle of 4° (
). B, the average HBC Cα distances versus the relative anti-clockwise rotation angles of TMD and CTD domains by snapshot (per ns) of MD simulations shown for four systems: GIRK2–Apo and PIP2 containing GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+. The system containing all gating molecules best exemplifies the most open HBC gate.
). In contrast, Gβγ binds GIRK channels at the intersubunit interface between the βD2–βE1 loop of one subunit and the βL–βM loop of its adjacent subunit (
). Because the common edge between the two planes that defined the TM–CTD dihedral angle is at the βD2–βE1 loop level, we aimed to assess relative contributions of Gβγ and Na+ to the local movements of both loops and further to the rocking motion described above.
Changes in the local conformations of the βD2–βE1 and βL–βM loops that comprise the Gβγ-binding site were examined first. Two critical residues to the binding and activation by Gβγ, Phe-254 in the βD2–βE1 loop and Leu-344 in the βL–βM loop (
NMR analyses of the Gβγ binding and conformational rearrangements of the cytoplasmic pore of G protein-activated inwardly rectifying potassium channel 1 (GIRK1).
), are close to one another in the absence of Gβγ and farther apart as Gβγ occupies the cleft between the two loops. Fig. S6 shows that, consistent with the 4KFM crystal structure, both Gβγ-present systems (GIRK2–Gβγ/GIRK2–Gβγ–Na+) show a larger Phe-254–Leu-344 Cα distance compared with the systems in the absence of Gβγ (GIRK2/GIRK2–Na+). Addition of Gβγ that interacts with both βD2–βE1 and βL–βM loops causes clear changes in the local conformations (Fig. 8 and Fig. S5, C and D). The reason for the local twists was assessed by the ψ–ф angles as a function of the gating molecules. Significant differences in the ψ–ф distribution of the systems in the presence of Gβγ from those in its absence take place in both loops (Fig. S7). Gln-248 was reported to form contacts with Asn-75, Ser-98, and Trp-99 on the Gβγ subunit, and mutations at Ser-98 and Trp-99 diminish Gβ activation of GIRK (
). Our simulation results are consistent with these experimental findings. In fact, with the exception of Ser-250, which did not display differences in the ψ–ф distribution of the systems in the presence of Gβγ, all other residues have been reported to interact with Gβγ (
). Similarly, the ψ–ф distributions are consistent with the experimental work, showing that Gβγ rather than Na+ could induce strong interactions with both loops except residue Ser-250. As a result, both the βD2–βE1 and the βL–βM loops seem to act like a crank when Gβγ subunits are present. The stronger (GIRK2–Gβγ–Na+ and GIRK2–Gβγ > GIRK2–Na+) the twist rotations are, the more the CTD will rock.
Figure 8Changes in local conformations of the Gβγ-binding loops. The dihedral angles (°) of βL–βM (left column) and βD2–βE1 (right column) loop from trajectories with the largest distributions in GIRK2–Na+ labeled by dashed lines for comparison: the dihedral angle was defined by the Cα atoms of Tyr-349, Asp-346, Glu-345, and Glu-350 for βL–βM loop and Gln-248, Ser-250, Glu-251, and Gly-252 for the βD2–βE1 loop.
For each GIRK2 subunit, only one Na+ binds to the βC-βD1 loop, whereas two Gβγ subunits bind to the βD2–βE1 and βL–βM loops, respectively (Fig. S2A). Because of its distant binding location, Na+ seems unable to induce strong rotations of both loops. Collectively, two Gβγ subunits help one CTD rock more on both sides compared with the binding of a single Na+ ion. These differences between the extent of rocking movements caused by Gβγ versus Na+ are consistent with the greater channel activation seen by the Gβγ subunits versus Na+ (
We also examined the relationship of the TM2–CTD dihedral angle, the anti-clockwise rotation angle between TM and CTD, and the G-loop gate minimal distances in a 3D plot. In Fig. 6B, as the rotation angles increase and the TM2–CTD dihedral angles decrease (as shown by the arrow), the G-loop gate distances increase. In other words, an anti-clockwise rotation and/or an inwardly CTD rocking movement enlarge the G-loop gate.
PIP2 interactions as a result of gating by Na+ and/or Gβγ
PIP2 has been shown to be a necessary cofactor for activating GIRK channels (
Activation of the atrial KACh channel by the βγ subunits of G proteins or intracellular Na+ ions depends on the presence of phosphatidylinositol phosphates.
Proc. Natl. Acad. Sci. U.S.A.1998; 95 (9448327): 1307-1312
). In contrast to its ability to gate other Kir channels, PIP2 is unable to gate efficiently GIRK channels on its own but requires the presence of Na+/Gβγ (
). Thus, we proceeded to examine the interactions between residues comprising the GIRK2-binding site for PIP2 and phosphates 4′ (P4) and 5′ (P5) atoms on its inositol ring.
Pull on the TM2 helix gets the HBC gate opened
Gβγ relative to Na+ stabilizes key interactions of the positively charged Lys-194 residue with the negatively charged phosphates P5 and P4 of PIP2 (Fig. 9, A and B). Meanwhile, Lys-200 in the GIRK2–Gβγ–Na+ system loses its interaction with PIP2 (especially with the P4 phosphate), stabilizing the salt bridge interactions of the Lys-199 with both phosphates of PIP2 (Fig. 9, A and B). The CTD rocking and its anticlockwise rotation are likely to be driving the stronger binding of Lys-199 to the P5/P4 atoms and the concomitant unbinding of Lys-200. This interaction rearrangement may underlie the pull on the TM2 helix to cause its tilting movement (Fig. 5) and as a consequence causing the HBC gate to open. Unlike the induction by Gβγ, Na+ in the GIRK2–Na+ system does not exert much influence on Lys-194, Lys-199, or Lys-200, and it is thus unable to influence the HBC gate (Fig. 9B). Snapshots of GIRK2 with each of its gating particles as they influence interactions with PIP2 are also shown (PIP2, Fig. 9C; PIP2-Na+, Fig. 9D; PIP2-Gβγ, Fig. 9E; and PIP2-Na+-Gβγ, Fig. 9F).
Figure 9Changes in specific residue–PIP2 interactions leading to HBC gating.A, distances between the N atom of the proposed key Lys residues (C) and the P4 (left column) or P5 (right column) atoms of PIP2 taking the crystal structure (PDB code 4KFM) as a reference in red. B, PIP2-Lys pairwise Gibbs free energy change with a standard deviation bar. C–F, binding patterns between the key Lys that are involved in regulating HBC by the PIP2 identified in the crystal structure (PDB code 4KFM), GIRK2–Na+, GIRK2–Gβγ, and GIRK2–Gβγ–Na+. The major contributors to binding are highlighted in cyan.
The presence of PIP2 alone with the channel in our study resulted in ion permeation, albeit low, indicating that GIRK2–PIP2 could be active (Table 1). Although electrophysiological experiments have shown that PIP2 on its own is not efficient to stabilize the channel gate in the open conformation (
Activation of the atrial KACh channel by the βγ subunits of G proteins or intracellular Na+ ions depends on the presence of phosphatidylinositol phosphates.
Proc. Natl. Acad. Sci. U.S.A.1998; 95 (9448327): 1307-1312
). The low permeation in the GIRK2–PIP2 system simulations could also be due to the pre-open conformation as the initial structure for MD, which has overcome part of the energy barrier required for the CTD rocking movement. When we took an HBC closed conformation as our initial model, for example the last snapshot of the GIRK2–Apo system, PIP2 alone could not activate the channel even when the simulation time scale was extended to 1 μs. In other words, PIP2 alone could not overcome the CTD rocking energy barrier in this time scale. Furthermore, the HBC Po of GIRK2–PIP2 (44.86%) seems larger than that of GIRK2–Na+ (35.14%) as shown in Fig. 4A and Table 2. However, in an additional 0.5-μs simulation, the corresponding Po of GIRK2–PIP2 and GIRK2–Na+ showed an approximate 7% decrease and 34% increase, respectively (data not shown), which suggests that PIP2 alone is inefficient in stabilizing the HBC gate in the open state.
G-loop gate is stabilized by the interactions with the CD loop/βI strand
The G-loop gate of GIRK2 consists of seven residues (MVEATGM). Other than the terminal methionines with longer side chains, it is difficult for other residues to induce strong short-range van der Waals interactions. Instead, the long-range electrostatic interactions elicited by the only charged G-loop residue Glu-315 become critical for its movement. The G-loop gate has been found in a GIRK1 chimera to be stabilized in the open state by the equivalent Glu residue through a hydrogen bond of Glu-304 with His-222 of the CD loop (
). A similar hydrogen bond also exists in GIRK2. For example, in the GIRK2–Na+ system, the hydrogen bond between E315(N) and H233(O) is observed with significant occupancy 10.40% (Table 3). This bond is lost in GIRK2–Gβγ/GIRK2–Gβγ–Na+ systems because of longer distances (Fig. 10). The other G-loop gate stabilization in the closed state of the GIRK1 chimera was proposed to be the intersubunit Glu-304–Arg-313 salt bridge (
). However, in GIRK2 this interaction seems to favor G-loop gate opening. Glu-315 (G loop)–Arg-324 (adjacent βI strand) was formed in the open state of the GIRK2–Na+/GIRK2–Gβγ systems (Fig. 10). During channel activation by Na+, this salt bridge causes Glu-315 to be pulled outward (Fig. 11A), which facilitates the opening movement of G-loop gate via an anti-clockwise rotation (Fig. 6).
Table 3The key hydrogen bonds of cytosolic domain proposed to facilitate the G-loop gating
Figure 10Changes in specific residue–PIP2 interactions leading to G-loop gating. Shown are the average nonbonded interactions with an error bar and the distances between the major residues involved in G loop gating over simulation time. The crystal structure (PDB code 4KFM) is taken as a distance reference in red. SC, side chain; MC, main chain.
Figure 11Stabilization of the G-loop gate in the open state. Network of the nonbonded interactions among G loop, CD loop, slide helix, LM loop, N terminus, and βI strand for GIRK2–Na+ (A) and GIRK2–Gβγ–Na+ (B). Two adjacent subunits (yellow and green) are shown with the front PIP2 removed for clarity.
PIP2-mediated Arg-230–Arg-77/Asp-81 interactions affect stabilization of the G-loop gate
Simulations with the GIRK1 chimera have revealed that the N-terminal Arg-52 (Lys-64 in GIRK2)–PIP2 salt bridge interactions in the G-loop gate stabilize the closed G-loop gate structure (
). Similarly, in the GIRK2–Na+ system open G-loop gate simulated structure, the Arg-77–PIP2 salt bridge was established. The carbonyl oxygen atom of residue Arg-77 was positioned to interact with Arg-230, which is two positions away from the key Asp-228 residue that coordinates Na+ and Asp-81 in a Arg-230–Arg-77–Asp-81 triad interaction pattern (Fig. 11A). In contrast to the GIRK2–Na+ system, in the GIRK2–Gβγ/GIRK2–Gβγ–Na+ systems the R77-PIP2 interactions are almost lost, which instead introduces a larger repulsion between Arg-230 and Arg-77 and consequently limits Arg-230 to interact only with residue Asp-81 (Fig. 11B). Considering the CTD anti-clockwise rotation, the outwardly pointing orientation of Arg-230 (GIRK2–Na+) is preferred to the inwardly pointing orientation (GIRK2–Gβγ/GIRK2–Gβγ–Na+), and it is more likely to stabilize the rotated conformation (Figure 11, Figure 12). As the rotation angles decrease (GIRK2–Na+ > GIRK2–Gβγ > GIRK2–Gβγ–Na+), the Glu-315–His-233 and Glu-315–Arg-324 stabilization is lost step by step. It is thus inferred that the PIP2 mediated Arg-230–Arg-77/Asp-81 interactions probably affect the stabilization of the G-loop gate in view of the CTD anti-clockwise rotation.
Figure 12Cartoon of the PIP2-mediated gating mechanism of the G-loop gate.A, GIRK2–Na+. B, GIRK2–Gβγ–Na+ and the synergism effect caused by the C-linker rotation. C, GIRK2–Gβγ. D, GIRK2–Gβγ–Na+. The facing subunits are colored in green (A–D) with an adjacent in orange (A and B). One PIP2 only is shown in each panel for clarity. The sodium ion is shown as a purple ball. The positive and negative groups are designated blue and red, respectively, in Fischer projection style. Once the Arg-230–Arg-77–Asp-81 triad interaction pattern (A) other than the Arg-230–Asp-81 (B) is induced, the opening movement of G-loop gate is facilitated by a larger anti-clockwise rotation with the aid of Glu-315–His-233 and Glu-315–Arg-324 hydrogen bond interactions. The Lys-200–PIP2 interaction (C) is disrupted by a 7.74° rotation of the C-linker, which is induced by the shorter Asp-228–Arg-201 distance (D).
As a whole, the G loop is likely to be gated as follows: the larger Lys-64–PIP2 distance fluctuations facilitate the formation of an intersubunit His-68 (N terminus)–Val-351 (LM loop) hydrogen bond (Fig. 11 and Table 3). Based on the intersubunit Arg-77–PIP2 interactions, Arg-230 in the CD loop interacts with Asp-81 (GIRK2–Gβγ/GIRK2–Gβγ–Na+) or Asp-81 + Arg-77 (GIRK2–Na+). Different binding patterns determine whether the G-loop gate could be stabilized by the adjacent βI strand or the βI + CD loop in view of CTD anti-clockwise rotation.
Discussion
In the past decade, great advances have been made in obtaining structural snapshots of a truncated GIRK2 channel that is in contact with its physiological intracellular regulators, PIP2, Na+, and Gβγ. However, no coherent structural dynamic model has emerged that is consistent with over three decades of functional data on the gating of GIRK channels. Here, we used all-atom microsecond-scale MD simulations to discern the large movements leading to key molecular interactions that underlie channel gating by Na+ and Gβγ. In the conceptual model that has emerged, Na+ binding to the CD loop causes it to interact intramolecularly with the G-loop gate (Glu-315–His-233) and stabilize it in the open state, with the aid of an intermolecular interaction of the G loop with the adjacent βI strand (Glu-315–Arg-324). Gβγ binds to the βD2–βE1 and βL–βM cleft and causes a large change in the TM2–CTD dihedral angle, which serves as a crank to initiate a rocking movement of the CTD and, as a consequence, the tilting of TM2 helix and the opening of the HBC gate.
These movements alter the interactions of GIRK2 with PIP2 that serve to stabilize the gates in the appropriate state causing channel gating to take place. In the case of the G-loop gate, the pivotal role of the CD loop interactions with the G loop is regulated on one hand by Asp-228 and His-233 that are utilized for Na+ ion coordination (
) and on the other hand by Arg-230 that is influenced by a slide helix PIP2-interacting residue, Arg-77. Based on the intersubunit Arg-77–PIP2 interaction, Arg-230 in the CD loop interacts with Asp-81 (GIRK2–Gβγ/GIRK2–Gβγ–Na+) or Asp-81 + Arg-77 (GIRK2–Na+), which determines whether the G-loop gate can be stabilized by the adjacent βI strand or βI + CD loop (Fig. 11). On the other hand, the Arg-77–PIP2 interaction seems to act as the connection of the G-loop gate to the HBC. The negatively charged head of PIP2 is located between the positively charged Arg-77 and Lys-194, experiencing competitive interactions (Fig. 12). Once the Arg-77–PIP2 interaction takes place, the Lys-194–PIP2 interaction is impaired (GIRK2–Na+). Otherwise, the stronger Lys-194–PIP2 interaction in the GIRK2–Gβγ/GIRK2–Gβγ–Na+ systems pulls the bottom of TM2 to open the HBC gate (Figs. 5, 9, and 12). The key amino acid residues predicted by our simulations to be involved in the stabilization of channel–PIP2 interactions, such as Lys-194, Arg-230, Asp-81, Glu-315, and Arg-324, are highly conserved among Kir channels (Fig. S8), further underscoring their important role proposed by our study. The importance of these residues proposed by our computational studies ought to be validated experimentally in future studies.
Our observations from both the HBC and G-loop gates do account for the movement of the two gates in opposite directions, as indicated by the PCA. In the presence of only Na+ ions, the HBC of the GIRK2–Na+ system is barely affected because of the limited CTD rocking movement; the G-loop gate is enlarged by an anti-clockwise rotation with the aid of the Glu-315–His-233 and Glu-315–Arg-324 interactions. On the other hand, in the presence of only Gβγ, the HBC in the GIRK2–Gβγ system is widened by the induced larger CTD rocking; in contrast, the G-loop gate is somewhat destabilized by the loss of the Glu-315–His-233 hydrogen bond. Lastly, in the presence of Na+ and Gβγ, the HBC in the GIRK2–Gβγ–Na+ system assumes the largest size, but both the Glu-315–His-233 and Glu-315–Arg-324 stabilization interactions of the G-loop gate are lost. In other words, a larger outward CTD rocking movement enlarges the HBC but destabilizes the G-loop gate. In fact, none of the published GIRK2 WT crystal structures existed with both gates closed, based on the minimum distance (5.69 Å) required for permeation.
The synergism between Na+ and Gβγ in gating GIRK channels (
) was also observed in our simulations. The Po of the HBC gate is the highest in the GIRK2–Gβγ–Na+ system, whereas that of the G-loop gate is similar in the GIRK2–Na+ and GIRK2–Gβγ systems (Table 2). Thus, the synergism is likely to correlate with the control of the HBC gate. Asp-228 acts as a key residue in the coordination of Na+. Na+ in the GIRK2–Gβγ–Na+ system causes a decrease in the distance of Asp-228–Arg-201 compared with the GIRK2–Gβγ system (Fig. 12, C and D). A rotation by 7.74° of the C-linker is induced, which disrupts the Lys-200–PIP2 interaction that was present in the GIRK2–Gβγ system. As a result, the Lys-194–PIP2 interaction is enhanced, and a larger and more stabilized HBC gate results in the GIRK2–Gβγ–Na+ system. Using both experimental and MD simulation approaches, Lys-200 was previously identified as a key player in GIRK2 channel gating by Gβγ or ethanol (
). Disruption of the Lys-200–PIP2 interaction through neutralization mutations caused channel activation and enhancement of channel–PIP2 interactions, a result consistent with the conclusions from the present simulation studies.
A recent computational study examined conduction through the GIRK2 channel pore but also gating in the absence on the gating molecules Gβγ or Na+ (
). By comparing these two systems, the authors found that PIP2 enlarged the HBC gate, whereas the G-loop gate was found to be narrower than the HBC gate. However, the PIP2-bound GIRK2 channels were conductive, albeit inefficiently. These results are consistent with the results of our study. Moreover, because our study focused on gating, including the effects of the gating particles Gβγ and Na+, it elucidated the relative efficiencies of gating under different conditions (i.e. Na+, Gβγ, or Na+ + Gβγ). Our results are in complete agreement with experimental findings. Bernsteiner et al. (
) as providing evidence that PIP2 alone could activate GIRK1/4 channels. This is in fact not the case, because Na+ was also present in the solutions used by Huang et al. to result in efficient GIRK channel gating. Thus, the experimental literature agrees that Na+ or Gβγ alone or together enhance the ability of PIP2 to gate GIRK channels to different levels. In fact, a single point mutant has been shown to strengthen channel–PIP2 interactions sufficiently, such that PIP2 could now gate alone (without Na+ or Gβγ) (
An interesting residue in the PIP2 binding pocket is Trp-91, which also showed a dependence of its distance from Lys-194 based on the level of opening of the HBC gate. Fig. S9 shows that Trp-91 was 9.0 Å away from Lys-194 in GIRK2–Gβγ but reached closer to Lys-194 in GIRK2–Gβγ–Na+, approximately within 7–8 Å. Even though this is still not close enough for Trp-91 and Lys-194 to engage in cation–π interactions, these subtle conformational differences in the Na+ + Gβγ system over the Gβγ demonstrate the sensitivity of dynamic experiments (MD simulations) over static ones (X-ray crystal structures).
In our simulations, the GIRK2 system showed a pre-open to HBC-closed process, which suggested that the HBC gate closing comes ahead of the G-loop gate. Consistent with our previous results (
), it can be concluded that the open/closed movements of the HBC and G-loop gates are sequential but not simultaneous. From closed to open, the G-loop gate opening precedes that of the HBC gate (Ref.
, GIRK1 chimera), whereas from open to close, the HBC closing precedes the G-loop gate (present study, GIRK2). Differences between GIRK2 and other GIRK homomeric (e.g. GIRK4) or heteromeric channels (e.g. GIRK1/2) remain to be explored to offer a molecular understanding of how intracellular channel modulators may differentially regulate the activity of GIRK channels.
In conclusion, the GIRK2 channel is activated by the intracellular regulators PIP2, Na+ ions, and/or Gβγ subunits via rocking and anti-clockwise rotation movements of the CTD. Because of the lack of direct or allosteric interactions in either the βD2–βE1 or the βL–βM loops, the Na+ ion action is limited to the control of the G-loop gate through the same subunit Glu-315–His-233 and adjacent subunit Glu-315–Arg-324 hydrogen bonds, thus inducing an anti-clockwise rotation. In contrast, binding of the Gβγ subunits in the cleft created by the βD2–βE1 and the βL–βM loops of adjacent subunits twists both the loops in a crank-like manner to rock the CTD and pull the HBC gate to open with the aid of Lys-194–PIP2 salt bridge interactions. The interactions of Arg-77 (slide helix)–PIP2 seems to be what connects the G-loop to the HBC gate. Na+ enlarges the G-loop gate and drives formation of the Arg-77–PIP2 interaction that in turn destabilizes the Lys-194–PIP2 interaction and the conductive state of HBC. The synergism of Na+ + Gβγ is likely to be attributed to a 7.74° rotation of the C-linker that as a result disrupts the Lys-200–PIP2 and enhances the Lys-194–PIP2 attraction.
Experimental procedures
Setting up GIRK2 model systems for MD simulations
The crystal structure of GIKR2 channel (PDB code 4KFM), which is considered to be in the pre-open state, was used to build the initial models by adding different endogenous regulators, such as PIP2, Na+, and Gβγ. To characterize the role of each regulator alone or in combination, five simulation systems in total were constructed, which are termed GIRK2, GIRK2–PIP2, GIRK2–PIP2-Na+ (GIRK2–Na+), GIRK2–PIP2-Gβγ (GIRK2–Gβγ), and GIRK2–PIP2-Gβγ–Na+ (GIRK2–Gβγ–Na+), respectively (Table S2). All the model systems were built by taking the following steps: (a) All the missing side chains of residues, Ile-55, Arg-73, Glu-127, Phe-141, Lys-165, Lys-301, and Glu-303 in GIRK2; Arg-42 and Arg-214 in the Gβ subunit; and Glu-58 and Glu-63–Phe-67 in the Gγ subunit (PDB code 4KFM) were added by Discovery Studio 2017 software. (b) All the hydrogen atoms were added using the H++ website server (http://biophysics.cs.vt.edu/)
). The protonated states of the titratable residues were determined by pKa calculations at neutral physiological conditions (pH 7.0). (c) The PIP2 molecule was built based on the DiC1PIP2 in the crystal structure (PDB code 4KFM) by adding alkyl tails using Maestro (Schrödinger, LLC). The geometry of the PIP2 structure was optimized by ab initio quantum chemistry at the Hartree–Fock/6–31G* level, followed by restrained electrostatic potential charge calculations using the GAUSSIAN 09 program (
). The antechamber module of AmberTools17 was employed to generate the required force field parameters for PIP2 based on the derived restrained electrostatic potential charges and the GAFF2 force field (
). (d) The GIRK2 channel and complexes were immersed in explicit lipid bilayer of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylethanolamine, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylserine, and cholesterol with molecular ratio of 25:5:5:1 (
) and a water box (116 × 116 × 150 Å3 for GIRK2, GIRK2–PIP2, and GIRK2–Na+ and 162 × 162 × 170 Å3 for GIRK2–Gβγ and GIRK2–Gβγ–Na+, a box edge of at least 14 Å from the protein periphery in each dimension using periodic boundary conditions) by using the CHARMM-GUI Membrane Builder webserver (http://www.charmm-gui.org/?doc=input/membrane)5 (
). 150 mm KCl was added into the system with the K+ ions, and water molecules from the crystal structure retained. (e) The tleap module of Ambertools17 was used to neutralize the complexes by adding additional K+ or Cl− ions prior to generation of the topology and coordinates files. The FF14SB, LIPID17, and GAFF2 force fields were chosen for protein, mixed lipid membrane, and PIP2, respectively. The simulation systems obtained contain a range from 160,000 to 400,000 atoms (Table S2).
All-atom microsecond-scale MD simulations
A two-stage energy minimization protocol, steepest descent algorithm (10,000 steps) and conjugate gradient (10,000 steps) was performed for each model system. The systems were heated from 0 to 300 K using Langevin thermostat algorithm with a 0.5-fs time step to avert internal disturbance. In the heating stage, the protein and lipid bilayer were initially fixed to remove any potential steric clashes from K+ or Cl− ions, and water molecules; followed by gradually reduced position restraints on the protein and membrane (10 to 0.1 kcal/mol·Å2 in 10 steps of total 20 ns). 0.5-μs MD simulations were conducted using the constant-temperature, constant-pressure ensemble (NPT) without electric field, followed by the same time scale using the constant-temperature, constant-volume ensemble (NVT). To accelerate the permeation events in the limited time scale of simulations, an external voltage 0.06 V/nm (
) was employed, under which the secondary structures of all four subunits of GIRK2 were well-maintained. However, higher electric fields resulted in structural instability. The PMEMD.CUDA program in AMBER16 was used to conduct the simulations. Long-range electrostatics were calculated using the particle mesh Ewald method with a 10-Å cutoff. A 4-fs time step by employing hydrogen mass repartition algorithm for system solutes (
) was used to accelerate the MD simulations. The SHAKE algorithm was applied on the solvent molecules.
Analysis of MD simulation results
All the analysis was done on the trajectories with external electric field, unless otherwise mentioned. Geometry analysis (distances and dihedral angles), PCA, and generalized Born surface area binding free-energy calculations were implemented by Amber16 and Ambertools17 (
). The channel-gating mechanism was studied by PCA based on the concatenated trajectories of GIRK2 (pre-open to closed) and GIRK2–Gβγ–Na+ (pre-open to open). For comparison, the PCA was also performed on available GIRK2 crystal structures (PDB codes 3SYA, 3SYC, 3SYO, 3SYP, and 4KFM) with and without the “open” conformation R201A with PIP2 (PDB code 3SYQ). To evaluate the TM–CTD rotation angle, larger-displacement residues in the N and C termini were excluded. To make the angle more reasonable, the relative tilting of TM domain to CTD occurred during simulations was removed by our program in Python 2.7.5, implemented in MDAnalysis 0.18 (
). All the calculations were conducted on servers equipped with NVIDIA Tesla K80, K40, and GeForce GTX-1080 graphical cards. The executables were built under the Community Enterprise Operating System (CentOS) version 7 and NVIDIA Compute Unified Device Architecture (CUDA) toolkit version 8.0.
Ion channel expression
Various ion channel cDNAs were generated by introducing point mutations on the background of the channel GIRK4 (S143T) (
) using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). cRNAs were in turn generated by in vitro transcription using the Message Machine kit (Ambion, Austin, TX). The cRNA concentrations were estimated by comparing the fluorescence intensity of diluted cRNA with cRNA marker (Gibco) which were electrophoresed in parallel on formaldehyde gels. Ion channel proteins were expressed in Xenopus oocytes by injecting the desired amount of cRNA into the oocytes. The desirable expression level for single channel recording was achieved by incubating the oocytes at 18 °C for 1–2 days after injecting 0.01–0.1 ng cRNA/oocyte. Oocytes were isolated and microinjected as previously described (
). The institutional animal care and use committee–approved protocol for use of Xenopus laevis frogs to isolate oocytes at Northeastern University was last approved on December 2018 by protocol 17–0102R-A1.
Single-channel recording and analysis
The single-channel activity was recorded using an Axopatch 200A amplifier (Axon Instruments). The pipette solution contained 96 mm KCl, 1 mm MgCl2, and 10 mm HEPES (pH 7.40). The bath solution contained 96 mm KCl, 5 mm EGTA, and 10 mm HEPES (pH 7.40). 100 μm gadolinium was routinely included in the pipette solution to suppress native stretch channel activity in the oocyte membrane. Chemicals were purchased from Sigma. Single-channel currents were filtered at 1–2 kHz with a six-pole low-pass Bessel filter, sampled at 5–10 kHz, and stored directly into the computer’s hard disk through the DIGIDATA 1200 interface (Axon Instruments). Single-channel analysis was carried out with pClamp8 (Axon Instruments). The open probability of a single-channel in a recording was calculated by dividing the sum of the durations that the channel dwelling at the open state by the total length of recording. The length of the recordings used for computing the channel open probability ranged from 100.0 to 1804.8 s (573.8 ± 499.9 s, n = 18).
Author contributions
D. L. and D. E. L. conceptualization; D. L., D. G., M. C., and D. E. L. resources; D. L., T. J., D. G., and M. C. data curation; D. L., T. J., D. G., M. C., and D. E. L. software; D. L. and T. J. formal analysis; D. L. and D. E. L. supervision; D. L. and T. J. validation; D. L., T. J., and D. E. L. investigation; D. L. and T. J. visualization; D. L., T. J., M. C., and D. E. L. methodology; D. L., T. J., and D. E. L. writing-original draft; D. L. and D. E. L. project administration; T. J. and D. G. writing-review and editing.
Acknowledgments
We are grateful to Profs. Hong Xing Zhang and Leigh Plant for feedback on the manuscript. We are grateful to the Engineering Research Center for Medical Data Mining and Application of Fujian Province (Xiamen University of Technology) for the graphics processing unit–based calculation platforms. The computations were also supported by the ITS (Information Technology Services) Research Computing at Northeastern University.
Activation of the atrial KACh channel by the βγ subunits of G proteins or intracellular Na+ ions depends on the presence of phosphatidylinositol phosphates.
Proc. Natl. Acad. Sci. U.S.A.1998; 95 (9448327): 1307-1312
NMR analyses of the Gβγ binding and conformational rearrangements of the cytoplasmic pore of G protein-activated inwardly rectifying potassium channel 1 (GIRK1).
This work was supported by Foundation of Fujian Educational Committee Grant JA13236, Xiamen University of Technology Grant E201300100), and National Institutes of Health Grant NIH R01 HL059949-22. The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
30441-5&id=gr1.jpg){kind=link}
30441-5&id=gr2.jpg){kind=link}
30441-5&id=gr3.jpg){kind=link}
30441-5&id=gr4.jpg){kind=link}
30441-5&id=gr5.jpg){kind=link}
30441-5&id=gr6.jpg){kind=link}
30441-5&id=gr7.jpg){kind=link}
30441-5&id=gr8.jpg){kind=link}
30441-5&id=gr9.jpg){kind=link}
30441-5&id=gr10.jpg){kind=link}
30441-5&id=gr11.jpg){kind=link}
30441-5&id=gr12.jpg){kind=link}