Among purinergic P2X receptor (P2XR) channels, the P2X7R exhibits the most complex gating kinetics; the binding of orthosteric agonists at the ectodomain induces a conformational change in the receptor complex that favors a gating transition from closed to open and dilated states. Bath Ca2+ affects P2X7R gating through a still uncharacterized mechanism: it could act by reducing the adenosine triphosphate4− (ATP4−) concentration (a form proposed to be the P2X7R orthosteric agonist), as an allosteric modulator, and/or by directly altering the selectivity of pore to cations. In this study, we combined biophysical and mathematical approaches to clarify the role of calcium in P2X7R gating. In naive receptors, bath calcium affected the activation permeability dynamics indirectly by decreasing the potency of orthosteric agonists in a concentration-dependent manner and independently of the concentrations of the free acid form of agonists and status of pannexin-1 (Panx1) channels. Bath calcium also facilitated the rates of receptor deactivation in a concentration-dependent manner but did not affect a progressive delay in receptor deactivation caused by repetitive agonist application. The effects of calcium on the kinetics of receptor deactivation were rapid and reversible. A438079, a potent orthosteric competitive antagonist, protected the rebinding effect of 2’(3′)-O-4-benzoylbenzoyl)ATP on the kinetics of current decay during the washout period, but in the presence of A438079, calcium also increased the rate of receptor deactivation. The corresponding kinetic (Markov state) model indicated that the decrease in binding affinity leads to a decrease in current amplitudes and facilitation of receptor deactivation, both in an extracellular calcium concentration–dependent manner expressed as a Hill function. The results indicate that calcium in physiological concentrations acts as a negative allosteric modulator of P2X7R by decreasing the affinity of receptors for orthosteric ligand agonists, but not antagonists, and not by affecting the permeability dynamics directly or indirectly through Panx1 channels. We expect these results to generalize to other P2XRs.
Purinergic P2X receptors (P2XRs) are ATP-gated cation channels. Seven mammalian receptor subunits, denoted P2X1 through P2X7, and several spliced forms of these subunits have been identified. Each subunit has only two transmembrane (TM) domains, the N and C termini facing the cytoplasm, and a large extracellular loop. These subunits assemble together as homo- or heterotrimers to make functional receptors (North, 2002). P2XRs likely have three intersubunit orthosteric binding sites located in the ectodomain, and their full occupancy appears to be required for conformational changes in the TM channel gate, leading to facilitation of cation influx through the channel pore (Marquez-Klaka et al., 2007; Browne et al., 2010). All P2XRs are permeable to Na+, K+, and Ca2+, and some are also permeable to Cl− (Egan et al., 2006). Positive and negative allosteric modulators of P2XRs interact with binding sites that are topologically distinct from the orthosteric sites recognized by the receptor endogenous agonist ATP, causing conformational changes that profoundly influence the gating of P2XRs (Coddou et al., 2011). In general, it is difficult to explore experimentally the interactions between binding of orthosteric and allosteric ligands and gating because simultaneous assessment of ligand binding and channel gating is not possible. Therefore, understanding ligand-gated systems predominantly depends on development of a kinetic model (Colquhoun, 1998).
The P2X7R is an unusual member of this family of channels. Structurally, the P2X7 subunit is distinguished from other subunits by its long intracellular C-terminal tail containing multiple protein and lipid interaction motifs and a cysteine-rich 18–amino acid segment (Surprenant et al., 1996) and by its inability to make stable heteromeric complexes (Torres et al., 1999; Nicke, 2008). It also appears that this channel exhibits different permeability states (Yan et al., 2008), which further complicates the understanding of coupling of conformational changes in the orthosteric binding domains with the corresponding changes at the TM channel gate. The binding of Ca2+ and Mg2+ to β- and γ-phosphate groups of ATP is a well-established phenomenon (Jahngen and Rossomando, 1983), but at the present time it is unknown which form of ATP (ATP4− and/or Mg/Ca-ATP) acts as an orthosteric agonist for this receptor (Jiang, 2009). It has also not been clarified whether these two macroelements affect the receptor function by acting as allosteric modulators, as earlier data suggest (Virginio et al., 1997), and/or by altering the permeability dynamics, as shown for P2X4R (Khakh et al., 1999). Furthermore, it is unknown how allosteric modulators affect P2X7R gating by changing the affinity of receptors for ATP and/or the strength of gating, i.e., the transduction of signals from orthosteric binding domains to the receptor gate. Finally, the interaction between allosteric agonists/antagonists and orthosteric antagonists has not been examined previously for P2X7R.
In this study, we combined biophysical and mathematical approaches to address these questions. To examine the influence of bath Ca2+ on the permeability dynamics, we performed full concentration response studies using ATP and 2’(3′)-O-4-benzoylbenzoyl)ATP (BzATP) as orthosteric agonists in naive cells (not previously exposed to orthosteric agonists) bathed in Ca2+-containing and Ca2+-deficient medium. To clarify whether Ca2+ affects P2X7R by reducing the free acid concentration of agonist, we manipulated total agonist and Ca2+ concentration, resulting in either fixed or progressively increased free acid agonist concentration. We also studied the influence of Ca2+ on receptor behavior by analyzing the rates of decay of current after washout of agonists or application of a competitive orthosteric antagonist, termed current or receptor deactivation. Based on experimental data, we revised a previously published Markov state model (Yan et al., 2010) to describe regulation of P2X7R by bath calcium, which provides a better understanding of the dynamics of these receptors under orthosteric and allosteric regulation.
MATERIALS AND METHODS
Cell culture and transfection
Human embryonic kidney 293 (HEK293) cells were used for the expression of wild-type and mutant P2X7Rs, as described previously (He et al., 2003; Yan et al., 2006). These cells were obtained from the American Type Culture Collection and were routinely maintained in Dulbecco’s modified Eagle’s medium containing 10% (vol/vol) fetal bovine serum (Invitrogen) and 1% (vol/vol) penicillin–streptomycin liquid (Invitrogen) in a tissue culture incubator. For electrophysiological measurements, cells were grown on 35-mm dishes at a density of 0.5 × 106 cells per dish. Transfection was conducted 24 h after plating the cells using 2 µg DNA and 5 µl Lipofectamine 2000 reagent (Invitrogen) in 2 ml of serum-free Opti-MEM (Invitrogen). After 4.5 h of incubation, the transfection mixture was replaced with normal culture medium, and cells were cultured for an additional 24–48 h. Transfected cells were mechanically dispersed and recultured on 35-mm dishes for 2–10 h.
Whole-cell patch-clamp recordings were performed with single cells at room temperature using an Axopatch 200B amplifier (Axon Instruments) as described previously (Yan et al., 2005). Patch electrodes, fabricated from borosilicate glass (type 1B150F-3; World Precision Instruments, Inc.) using a Flaming Brown horizontal puller (P-97; Sutter Instrument), were heat polished to a final tip resistance of 3.5–5.5 Mohm. All current recordings were programmed, captured, and stored using the pClamp 8.0 software packages in conjunction with a converter (Digidata 1322A A/D; Axon Instruments). Experiments were performed on single cells with a mean capacitance of 10 pF. Unless otherwise stated, membrane potential was held at −60 mV. Current–voltage relations were used to estimate changes in reversal potential during agonist application and were obtained by voltage ramps from −80 to 80 mV delivered twice per second during 50 s. Patch electrodes were filled with solution containing 145 mM NaCl, 10 mM EGTA, and 10 mM HEPES; the pH was adjusted with 10 M NaOH to 7.35. The osmolarity of this solution was 305 mosM. Krebs-Ringer–like bath buffer contained 147 mM NaCl, 3 mM KCl, 1 mM MgCl2, 0–10 mM CaCl2, 10 mM glucose, and 10 mM HEPES; the pH was adjusted to 7.35 with 10 M NaOH. In some experiments, ATP and BzATP solutions were prepared daily in bath buffer with pH properly readjusted and were applied using either an Ultrafast Solution-Switching System (LSS-3200; EXFO Burleigh Products Group Inc.) that was simultaneously program controlled by pClamp 8.0 software through a PZ-150M amplifier (Yan et al., 2006) or an RSC-200 Rapid Solution Changer (Biological). Cells with enhanced green fluorescent protein fluorescence were identified before immersing the electrode in bath solution for a gigaohm seal.
Measurements of free calcium concentrations in bath medium
Measurements of free calcium ion concentration in bath medium with or without ATP and BzATP were made with a plastic membrane combination calcium ion selective electrode (ionplus Sure-Flow; Thermo Fisher Scientific). ATP (Sigma-Aldrich) and BzATP (Sigma-Aldrich) were weighed to prepare the 10-mM stock solution in deionized water according to the free acid formula weight on the product label. The Mg2+-free sodium background buffer consisted of 147 mM NaCl, 3 mM KCl, 10 mM glucose, and 10 mM HEPES, and the pH was properly adjusted to 7.35 with NaOH. The 0.95-mM Ca2+ buffer was prepared by addition of 0.1 M Ca2+ standard (Thermo Fisher Scientific) to the background buffer. 320-, 100-, 32-, 10-, 3.2-, 1-, and 0-µM ATP or BzATP solutions were prepared by addition of the same 480-µl sum volume of various combinations of pure water and 10-mM ATP or BzATP stock solution into the 19.6 ml of 0.95-mM Ca2+ buffer. To generate a standard curve, a real-time recording of the millivolt reading of the calcium ion selective electrode was performed with an Orion 4-Star reader (Thermo Fisher Scientific), and the standard curve was calculated by semilogarithmic fitting using KaleidaGraph v 4.1 (Synergy Software). After the standard curve recording, the millivolt readings of different ATP or BzATP solutions were collected, and the test-free Ca2+ concentrations were derived from the standard curve.
Whenever appropriate, the data were presented as mean ± SEM values. Significant differences, with P < 0.01, were determined by Mann-Whitney test using InStat 3.05 (GraphPad Software). Nonlinear curve fitting of deactivation currents was performed with the Clampfit 10.0 (Molecular Devices) predefined biexponential function . ATP4− and BzATP4− concentrations were calculated using Theo Schoenmakers’ program Chelator, which is freely available online. Numerical simulations of the Markov state model were performed using MATLAB (Mathworks). Parameter fitting of f to the simulations of the Markov state model over the interval [a,b] was performed by minimizing the error given by the L2 norm
A Markov state model consisting of eight states (Ci, closed states and Qi, open states, i = 1–4) describing the sequence of binding and unbinding of agonist to P2X7R was used (Fig. S2). Details of the scheme are available in Yan et al. (2010). In brief, each state corresponds to the fraction of receptors that are either unsensitized or sensitized (Fig. S2, top and bottom row, respectively) with a given number of agonist binding sites occupied. Negative cooperativity in agonist binding (i.e., a decrease in binding affinity with the occupation of each site) is assumed to occur only to unsensitized receptors (Fig. S2, top row). However, receptor sensitization (Fig. S2, bottom row) is assumed to restore symmetry of agonist binding and lead to dilation of the channel associated with each receptor. The allosteric binding of divalent extracellular Ca2+ to the P2X7R has been added to the model as a new feature. This is done by assuming that the forward rates (k2, k4, and k6) decrease by a factor F and the backward rates (k1, k3, and k5) increase by a factor (2 − F) when bath extracellular calcium concentration ([Ca2+]e) is increased. When F = 1, the model reduces to the previously published version, and this happens when [Ca2+]e ≈ 2 mM. F is a decreasing Hill function (Fig. S3) and is given by
where α is the maximum level of allosteric regulation (or inhibition) and β is the half-maximum level of divalent cations ([DC]e) required for such regulation.
The mathematical model associated with this Markov state model (shown schematically in Fig. S2) is thus given by
where A is the applied agonist concentration and Li, i = 1–3, are the rates of sensitization/loss of sensitization. The whole-cell current is then given by
where g12 is the conductance of the Q1 and Q2 states and g34 is the conductance of the Q3 and Q4 states (g12 < g34), V is the holding potential, and E is the reversal potential (see Table S1 for parameter values). Although it is possible to model the allosteric binding of Ca2+ to P2X7R by adding new states to the scheme, each representing the fraction of receptors with a given number of occupied allosteric Ca2+ binding sites (as well as occupied orthosteric agonist binding sites), the total number of these allosteric binding sites is not known, and the complexity of the scheme would become unnecessarily large. For this reason, we model the effects of divalent cations using the aforementioned phenomenological approach.
Online supplemental material
Fig. S1 shows the effects of a P2X7R-specific antagonist on BzATP-induced receptor activation and deactivation. Fig. S2 shows a Markov state model describing the binding and unbinding of BzATP to the P2X7R. Fig. S3 illustrates Hill functions describing allosteric regulation of P2X7R by divalent cations, whereas Fig. S4 illustrates the effects of divalent cations on receptor deactivation. Table S1 shows parameter values and distributions used in modeling of P2X7R gating.
Extracellular calcium dependence of orthosteric agonist potency
To clarify whether bath Ca2+ plays a role in the P2X7R permeability kinetics, we studied the concentration dependence of ATP and BzATP on currents using the wild-type rat P2X7R expressed in HEK293 cells bathed in 2-mM Ca2+/1-mM Mg2+–containing buffer or in Ca2+-deficient/1-mM Mg2+–containing buffer. All experiments were performed with naive receptors (not previously exposed to agonist), and all recordings were performed in one cell per dish during a single agonist application/withdrawal. Fig. 1 illustrates the patterns of current responses in cells stimulated with ATP (A and B) and BzATP (C and D) for 40 s or 2 min in 2-mM Ca2+–containing (A and C) and Ca2+-deficient (B and D) medium. In cells bathed in Ca2+–containing medium, monophasic, slowly developing currents of small amplitude were observed in response to 0.1 and 0.32 mM ATP (Fig. 1 A) and 3.2 and 10 µM BzATP (Fig. 1 C). As described previously (Yan et al., 2008, 2010), at higher agonist concentrations (1–10 mM ATP and 32–320 µM BzATP), biphasic currents were observed, with a rapid initial rise in current (I1) followed by a slower secondary rise (I2) that we interpret as dilation of the pore. The rate of dilation increased with elevation in agonist concentrations until, at the highest agonist concentrations, the channels passed almost immediately to the dilated state, and the current consisted mostly of I1, as it did for the lowest agonist concentrations. The same pattern of response was also observed in cells bathed in Ca2+-deficient medium, but with a leftward shift in agonist concentration of ∼0.5 log units (Fig. 1, B and D; and Fig. 2 A). The “noise” current observed in response to 1 and 3.2 mM ATP does not reflect low quality recording but temporally coincides with extensive blebbing. Others have observed similar effects (Roger et al., 2008). Because the development of biphasic current reflects a transition from an open to a dilated state, these results indicate that the pore does not depend on bath Ca2+.
The calculated values of ATP4− concentrations in Ca2+-containing and -deficient media are shown in Fig. 1 (A and B; numbers in parentheses above traces). Using the same program (the Maxc program for a two-chelator, two-metal calculation), we also calculated the concentration of the free acid form of BzATP in solution (Fig. 1, C and D). Comparing the values of the calculated free acid forms of the two agonists in Ca2+-containing and Ca2+-deficient medium indicates practically identical concentration responses in the absence and presence of Ca2+. This is consistent with a hypothesis that ATP4− rather than the divalent-bound ATP is the native agonist for this receptor subtype. However, these results are also consistent with the hypothesis that bath Ca2+ acts as an allosteric inhibitor and that 2 mM Ca2+ causes a rightward shift in the potency of agonists of ∼0.5 log units independent of the concentration of the free acid form of agonists. In the next section, we consider experiments to distinguish between the two hypotheses.
Concentration dependence of agonist potency on extracellular calcium
In further experiments, we varied bath Ca2+ concentrations (0.5, 1, 2, 5, and 10 mM), keeping identical (100 µM) total BzATP concentrations and 1 mM Mg2+ in buffer solution. Fig. 3 A shows that the calculated free acid concentrations of BzATP progressively decreased with an increase in bath Ca2+ concentration. Representative traces of current responses during the initial agonist application are shown in Fig. 3 B. Application of media containing 100 µM BzATP and 0.5 mM Ca2+ resulted in high amplitude biphasic currents, whereas the current amplitude decreased with an increase in bath Ca2+ concentrations to 5 and 10 mM. Fig. 3 C shows the correlation between divalent cation concentrations and amplitude of current response (left) and between free BzATP concentration and amplitude of current response (right) using mean ± SEM values. A similar correlation coefficient was observed in both cases, indicating that these experiments could not distinguish between the two hypotheses for the action of bath Ca2+ as either an allosteric modulator (Fig. 3 C, left) or a factor determining free agonist concentration (Fig. 3 C, right).
One way to overcome this limitation is to have variable Ca2+ concentrations and identical free agonist concentrations. As estimated by the Maxc program calculations, these types of experiments are practically impossible to do with ATP because it acts as an orthosteric agonist for P2X7R in the submillimolar to millimolar concentration range. Specifically, binding of divalent cations to ATP in this concentration range affects not only the free agonist concentration but also the free divalent cation concentrations. Thus, to have minimum impact on divalent cation concentrations, an agonist should be at least in a micromolar concentration range, which is the case with BzATP. As shown in Fig. 3 D, calculations indicated that the free acid form of BzATP could be fixed at 1.6 µM by varying the total BzATP concentration from 17 to 90 µM and the total Ca2+ concentration from 0 to 10 mM in the presence of constant (1 mM) Mg2+.
However, this analysis is based on the assumption that BzATP binds calcium with a similar potency as ATP. Others have also estimated free BzATP concentration using an ATP binding program (Virginio et al., 1997). To test the validity of the calculation method for free BzATP concentration, we used a calcium ion selective electrode to estimate free bath Ca2+ concentrations in the presence of variable concentrations of ATP and BzATP. To exclude the impact of Mg2+ on calcium measurements, we used Mg2+-free buffer with Ca2+ fixed at 950 µM and with ATP and BzATP concentrations of 0, 1, 3.2, 10, 32, 100, and 320 µM. Fig. 2 B shows highly comparable concentration-dependent effects of ATP on free Ca2+ concentration obtained by calculations and by measurements, confirming the validity of this method. At 0–100-µM concentration, there was no significant difference between values of free Ca2+ in the presence of ATP and BzATP, whereas at higher agonist concentrations, free Ca2+ concentrations were 2–3% higher in the presence of BzATP than ATP. These results indicate that BzATP at the concentrations used in our experiments binds Ca2+ with a comparable capacity to ATP.
Fig. 3 (E and F) summarizes the results of stimulation of cells with BzATP and bath Ca2+ in concentrations shown in Fig. 3 D. Under such experimental conditions, the relationship between total agonist concentration and the pattern of current response was lost. Practically, the lowest total BzATP concentration (17 µM) was most effective, and the highest concentration (90 µM) was the least effective. Furthermore, although the free acid from of BzATP was identical in all four media, we still observed the full concentration-dependent inhibitory effect of Ca2+ on the peak amplitude of current response. The finding that bath Ca2+ affects P2X7R gating independently of the concentrations of the free acid form of agonists indicates the allosteric nature of action of this cation.
Others have shown that the rP2X7R-H130A mutant is resistant to Mg2+ inhibition (Acuña-Castillo et al., 2007). To examine whether this residue also accounts for allosteric action of Ca2+, we stimulated cells with 100 µM BzATP in Ca2+-deficient and 10-mM Ca2+–containing medium. Like the wild-type receptor, the mutant receptor also responded to agonist application with a delay in receptor activation and 15-fold decrease in the peak amplitude of current when bathed in 10 mM Ca2+ (Fig. 3, B and G). This indicates that different residues account for the allosteric action of Ca2+ and Mg2+.
Cells expressing wild-type receptors and clamped at 60 mV also responded to agonist application with biphasic currents, but the current was outward (Fig. 3 H). Furthermore, Ca2+ inhibited receptor activation in a concentration-dependent manner (Fig. 3, H and I), which is highly comparable with that observed in cells clamped at −60 mV (Fig. 3, E and F). As Ca2+ influx would be negligible in this condition, the results suggest that Ca2+ probably does not act near the channel pore.
Calcium dependence of receptor deactivation
If bath Ca2+ allosterically affects the potency of orthosteric ligands at P2X7R, it should also affect the rates of receptor deactivation. In general, the washout of ATP was accompanied by a relatively rapid decline of current (deactivation), whereas the decline of the current after washout of the more potent BzATP was slower (Fig. 1). In concentrations that caused biphasic responses when cells were bathed in Ca2+-containing medium, the best approximation for decay of current after the washout of agonist was achieved using biexponential fitting, and the averaged decay constants τ1 and τ2 were about sevenfold larger for BzATP than for ATP (τ1 = 2.8 ± 0.5 vs. 0.4 ± 0.2 s; τ2 = 14.9 ± 1.1 vs. 2.1 ± 0.2 s, respectively).
To study the effects of bath Ca2+ on the kinetics of receptor deactivation, we used several experimental approaches. In the first experiment, the same cells were repetitively stimulated with 100 µM BzATP for 1 s separated by washout periods of 2 min in 2-mM Ca2+–containing medium (Fig. 4 A, left) or 4 min in Ca2+-deficient medium (Fig. 4 A, right). Superimposed records from these experiments show two phenomena. First, deactivation of receptors was dramatically slowed in cells bathed in Ca2+-deficient medium, indicating that bath Ca2+ reduces the potency of agonist. Second, repetitive stimulation slowed receptor deactivation independently of the status of bath Ca2+ (Fig. 4 B). Both phenomena were also observed during repetitive application of 100 µM BzATP for 40 s (Fig. 4 C).
The effect of bath Ca2+ on receptor deactivation was concentration dependent. Fig. 4 D (left) shows representative traces (gray) with fitted biexponential curves (black) from cells bathed in media containing 0.5, 5, and 10 mM Ca2+. From the fitting curves, the τ2s were derived, and the mean ± SEM values were generated (Fig. 4 D, inset). To study the time course of Ca2+ action, we replaced 2 mM Ca2+–containing medium with Ca2+-deficient medium (Fig. 4 E, horizontal bars) during washout of agonists. The addition of Ca2+ rapidly (0.5 s) and reversibly affected the rate of receptor deactivation. The same effect was also observed in cells expressing P2X7R-H130A mutant (Fig. 4 F). Thus, bath Ca2+ rapidly, and in a concentration-dependent manner, facilitates receptor deactivation.
Effects of A438079 and carbenoxolone (CBX) on receptor activation and deactivation
Facilitation of receptor deactivation by bath Ca2+ could be explained either by inhibition of the rebound effect of BzATP on current or by facilitated dissociation of residual bound agonist from binding sites, the latter being consistent with allosteric action of this cation. To clarify this issue, in further studies we used A438079 (Torcis), a human P2X7R-specific competitive antagonist (Nelson et al., 2006; McGaraughty et al., 2007). In cells expressing rat P2X7R, A438079 applied alone did not generate any current (unpublished data). When applied together with 3.2 mM ATP, it reduced the peak amplitude of current in a concentration-dependent manner (Fig. S1 A), with an estimated IC50 value of ∼1 µM. A438079 also concentration dependently inhibited 100-µM BzATP–induced current amplitude (Fig. S1 B) with a similar IC50 value. Furthermore, in cells with developed biphasic current caused by the application of 100 µM BzATP, A438079 rapidly and in a dose-dependent manner decreased the peak amplitude of current with an IC50 value of ∼0.3 µM (Fig. S1 C). When cells were stimulated with 32 µM BzATP, A438079 also inhibited current with an IC50 of ∼0.3 µM. Together, these data show that A438079 also acts as a typical competitive orthosteric antagonist to rat receptor, with ∼3,000-fold higher potency than ATP and 200-fold higher potency than BzATP, and could be used to study effects of calcium on the kinetics of receptor deactivation.
Fig. 5 A shows that A438079 facilitated deactivation of the wild-type receptor in both Ca2+-containing (left) and Ca2+-deficient medium (right) and that this effect was rapid (∼0.8 s). To further clarify this issue, we used the T15E-P2X7R mutant, which is more sensitive to BzATP and, consequently, deactivates slowly (Yan et al., 2010). When applied during the washout of free BzATP, A438079 also facilitated the mutant receptor deactivation. The effect of A438079 of increasing the rate of receptor deactivation was observed in cells bathed in both Ca2+-containing and -deficient medium (Fig. 5 B). These results indicate that BzATP is partially rebound during the washout period and is replaced by a more potent ligand molecule. When deactivation of receptors was initiated by washing the cells with Ca2+-deficient medium in the presence of A438079, the subsequent addition of 2 mM Ca2+ facilitated receptor deactivation (Fig. 5 C). These results indicate that the rapid effect of bath Ca2+ on the current decay during the washout period is not dependent on the rebinding effect of orthosteric agonists on P2X7R gating.
In our experimental conditions, HEK293 cells express mRNA transcripts for pannexin-1 (Panx1), but not Panx2 and Panx3. The expression of Panx1 protein transcripts was below detection by Western blot analysis (Li et al., 2011). The inhibitory effect of Ca2+ on receptor activation was preserved in cells treated with 32 µM CBX, an inhibitor of Panx1 channels. Fig. 5 D shows the inhibitory effect of 10 mM Ca2+ on 100-µM BzATP–induced current in cells expressing the wild-type receptor and perfused with Ca2+-deficient medium containing 32 µM CBX. These data indicate that even if Panx1 channels are expressed in HEK293 cells in sufficient numbers, their interaction with P2X7R does not account for inhibition of receptor function by Ca2+. Together with the findings that bath Ca2+ affects receptor activation in a concentration-dependent manner and independently of free agonist concentration, these experiments definitively show that Ca2+ in physiological concentrations acts as a negative allosteric modulator.
Kinetic model for calcium-dependent allosteric regulation
To have a better understanding of the effects of bath Ca2+ on P2X7R channel gating, we used a Markov state model consistent with the model presented in Yan et al. (2010), except for the inclusion of the allosteric regulation exerted by [Ca2+]e on the P2X7Rs, which was suggested by the experimental results presented in Figs. 1–4. This was done by assuming that [Ca2+]e increases the backward rates (k1, k3, and k5) by the fraction 2 − F and decreases the forward rates (k2, k4, and k6) by the fraction F. In Fig. 6, we show current simulations (obtained using Eq. 9) generated by stimulating model cells (described by Eqs. 1–8) with increasing doses of agonist (BzATP) concentrations at two different levels of [Ca2+]e: 2 and 0 mM. At 2-mM [Ca2+]e, current simulations were similar to those shown in Fig. 1 C (and identical to the ones displayed in Yan et al.  under the same conditions). To briefly summarize our results, we found that the model exhibited monophasic slow current responses at low BzATP concentrations (3.2 and 10 µM) and biphasic current responses that included fast and slow components at high BzATP concentrations (32, 100, and 320 µM) during both the activation and deactivation phases of the current. We suggested in Yan et al. (2010) that the slow component of the current exhibited at high agonist concentrations is a result of the sensitization of receptors (i.e., dilation of the channel pores associated with these receptors). This corresponds to a shift in the fraction of receptors in the top row to the bottom row of Fig. S2. The fast component of the activation phases gradually dominated over the slow component as the agonist concentration was increased. With the additional feature included in the model, subjecting the model cells to Ca2+-deficient medium ([Ca2+]e = 0 mM), as was done in Fig. 6 B, increased the potency of agonist binding (as a result of an increase in the value of the ratio F/[2 − F]). A reduction in agonist concentration by 0.5 log units generated the same outcomes obtained in Fig. 6 A, consistent with the cell recordings shown in Fig. 1 (C and D).
To investigate the effect of extracellular Ca2+ (or divalent cations in general) on current amplitudes, we stimulated a single model cell with 100 µM BzATP and only varied [Ca2+]e. Fig. 7 A shows that the simulated current amplitudes decreased in proportions similar to those observed experimentally when [Ca2+]e was gradually increased from 1.5 to 3 and then to 11 mM (compare with Fig. 3 B). Extending this result to a heterogeneous population of model cells (randomly generated from the distributions listed in Table S1) stimulated by the same level of BzATP concentration (100 µM) and subjected to various levels of [Ca2+]e (between 0.5 and 10 mM) with fixed (1 mM) Mg2+ concentration revealed that the Ca2+-dependent dose–response curve for the current amplitude generated by the whole population, shown in Fig. 7 (B, left), is best fitted by a decreasing Hill function, with a Hill coefficient of n = 4 and half-maximum inhibition of IC50 = 4.1 mM. The best fit to the BzATP4−-dependent dose–response curve, shown in Fig. 7 (B, right), was an increasing Hill function with n = 4 and EC50 = 3.9 µM (where BzATP4− at each data point was calculated in a manner similar to that described in the Materials and methods section). It should be mentioned here that even if the effect of extracellular Ca2+ was modeled differently, e.g., by assuming that F is a linearly decreasing function of [Ca2+]e (considered experimentally unrealistic), we would still get a Hill-like function for both Ca2+- and free BzATP–dependent dose–response curves (unpublished data). It is possible that the fit in Fig. 3 C is linear because of the small number of data points; with data at a greater number of concentrations, the experimental curve might turn out to be sigmoidal, as suggested here.
The P2X7R inhibition observed in Fig. 7 (A and B) is clearly not caused by a decrease in free agonist concentration (as a result of Ca2+ binding) because F and 2 − F do not reflect this feature but rather the allosteric regulation. To confirm this, we repeated the same experimental procedure conducted in Fig. 3 (D–F) by initially stimulating a single model cell with the same level of free BzATP (obtained by varying both extracellular Ca2+ and BzATP concentrations in a manner similar to Fig. 3 D). Fig. 7 C shows that the simulated current amplitudes decreased with the increase in [Ca2+]e, which is compatible with an allosteric regulation of the receptor by Ca2+. Extending this idea to a heterogeneous population of model cells, as in Fig. 7 D, revealed that a negative correlation exists between current amplitudes and [Ca2+]e, even though free BzATP in each case is the same, an outcome consistent with the experimental results in Fig. 3 E.
The main advantage of using kinetic models of this nature is the ability to extract quantitative information about the fraction of receptors in each state. This may provide insight into how extracellular Ca2+ allosterically inhibits receptor activation and facilitates receptor deactivation. For this reason, we have simulated in Fig. 8 the time evolution of the two closed states, C1 (black) and C4 (gray) in A, and the two open states, Q1 + Q2 (black) and Q3 + Q4 (gray) in B, to illustrate how maintaining the same level of free BzATP (by varying [Ca2+]e and BzATP concentration identified by the [DC]e values and gray bars in Fig. 8, respectively) could produce the effects observed in Figs. 3 and 7. Fig. 8 A shows that even though the increase in [Ca2+]e only decreased the activation of naive states slightly, the accumulation of receptors in Q1 + Q2 was significantly reduced (Fig. 8 B), leading to a significant reduction in the sensitization of receptors quantified by the fraction Q3 + Q4 (during stimulation) and C4 (during washout). This explains the progressive loss of the slow component (I2) of the activation phase of the current observed in Figs. 3 E and 7 C. The loss of the sensitized receptor component is also responsible for the loss of the slow deactivation component at high agonist concentration (Fig. S4). This explains the increase in the deactivation rate observed in Fig. 7 C.
The dependency of the deactivation phase of current on extracellular Ca2+ observed in Fig. 3 was also reproduced by the model, as shown in Fig. 9. Activation of a naive model cell repeatedly with 100 µM BzATP for 1 s (Fig. 9 A) in the presence of 2 mM Ca2+ (Fig. 9 A, left) and the absence of extracellular Ca2+ (Fig. 9 A, right) showed that it takes longer for the current to go back to baseline in the latter case (55 s) than in the former (11 s). This is confirmed in Fig. 9 B, where the simulated currents obtained in each BzATP pulse, normalized by their maximum values, exhibited a faster deactivation phase in Ca2+-containing medium (black) than in Ca2+ -deficient medium (gray). Furthermore, subsequent pulses showed slower deactivation than the former pulses in both the presence and absence of external Ca2+ as a result of receptor sensitization, as was previously described in Yan et al. (2010). As illustrated in Fig. 8, the allosteric inhibition exerted by Ca2+ led to a reduction in receptor accumulation in state Q2, causing a decrease in the fraction of sensitized receptors (i.e., a decline in the number of P2X7Rs transitioning from the top row to the bottom row of Fig. S2). Consequently, during washout, the deactivation phase at higher [Ca2+]e is primarily governed by the backward rates (k1, k3, and k5) of the unsensitized states (Fig. S2, top row), which are faster than those (k1) of the sensitized states (Fig. S2, bottom row), in contrast to the behavior in Ca2+-deficient medium (see Fig. S4 for the breakdown into sensitized and unsensitized receptors). Increasing the duration of stimulation to 40 s in Fig. 9 C resulted in similar outcomes; namely, slower deactivation phases in the absence of extracellular Ca2+ for both first (left) and second pulses (right). In other words, such behavior was independent of how long the model cells were stimulated for.
To further examine the effects of various concentrations of extracellular Ca2+ on the deactivation phase of the current, we subjected a naive model cell, stimulated with 100 µM BzATP for 40 s, to 0.5-, 5-, and 10-mM [Ca2+]e. The deactivation phases of simulated currents, normalized by their maximum values and shown in Fig. 9 (D, left), exhibited similar outcomes to those observed experimentally in Fig. 4 D: the increase in [Ca2+]e led to a progressive increase in the rate of receptor deactivation. Extending this study to a heterogeneous population of 10 model cells subjected to 0-, 0.5-, 2-, and 5-mM [Ca2+]e and stimulated with 100 µM BzATP revealed that the average time constant τoff2 obtained from fitting a biexponential function to the deactivation phases, decreased as [Ca2+]e was increased, a result consistent with the experimentally observed outcomes in Fig. 4 D. As indicated earlier, the decrease is caused by a gradual loss of the sensitized compartment responsible for the slow deactivation (as illustrated in Fig. 8 and Fig. S4). Finally, alternating [Ca2+]e between 2 and 0 mM during the deactivation phase of the simulated current only (as shown in Fig. 9 E) repetitively for 3 s (left) or once for 5 s at the onset of the deactivation phase (right) exhibited similar outcomes to those observed in Fig. 4 E; namely, fast deactivation at high [Ca2+]e, slow deactivation at low [Ca2+]e, and instantaneous change between these two distinct behaviors once the [Ca2+]e is altered.
We have shown here that the agonistic actions of ATP and BzATP on P2X7R were affected by bath Ca2+ concentration in the presence of fixed Mg2+ concentration (1 mM). Consistent with findings by others (Surprenant et al., 1996; Virginio et al., 1997; Klapperstück et al., 2001), bath Ca2+ inhibited agonist-induced current in a concentration-dependent manner, causing a rightward shift in dose–response experiments. In our experimental conditions with naive receptors, the difference was ∼0.5 log units when cells were bathed in Ca2+-deficient and 2-mM Ca2+–containing medium. For P2X4R, it has also been shown that bath Ca2+ affects the permeability state of the receptor; in the presence of bath Ca2+, the channel is only permeable to small cations, whereas in cells bathed in Ca2+-deficient medium, the channel pore is permeable to larger organic cations (Khakh et al., 1999). This finding was confirmed more recently using fast-scanning atomic force microscopy (Shinozaki et al., 2009).
Initially, it was suggested that P2X7R is activated rapidly (within milliseconds) after application of agonists, followed by a gradual (within seconds to minutes) increase in permeability to organic cations and fluorescent dyes (Surprenant, 1996; Chessell et al., 1997; Michel et al., 1999), which parallels the presence of I1 and I2 in our experiments. However, several observations suggest that fluorescent dye entry occurs through a separate pathway than organic cations (Petrou et al., 1997; Klapperstück et al., 2000), and the coupling of P2X7R to Panx1 channels was suggested to account for cellular entry of fluorescent dyes (Pelegrin and Surprenant, 2006). Our recent data on this topic are consistent with the original hypothesis of P2X7R pore dilating during the sustained agonist application, including the finding that I2 current growth temporally coincides with the transition from an open to a dilated state (Yan et al., 2008, 2010). Here, we show that removal of bath Ca2+ was not essential for the development of a secondary current growth (Fig. 1). Dilation of the P2X2R pore also does not require removal of bath Ca2+ (Chaumont and Khakh, 2008). Our data do not exclude the possibility that Panx1 contributes to dye uptake in P2X7R-expressing HEK293 cells and do not argue against the role of Panxs in other purinergic signaling functions. Several recent publications (Dubyak, 2009), including our own work (Li et al., 2011), indicate that Panxs release ATP and that this action may include interactions with P2XRs.
A rightward shift in the potency of two orthosteric agonists in cells bathed in medium containing 2 mM Ca2+ was not visible when the calculated free agonist concentration was used in concentration response relationship analysis. Others have reached similar conclusions in experiments using Ca2+-deficient medium and manipulating bath Mg2+ concentrations (Klapperstück et al., 2001; Liu, M., S.D. Silberberg, and K.J. Swartz. 2011. ATP4− activates P2X receptors. Annual Biophysical Meeting. 1472-Pos.). This is consistent with the hypothesis that ATP4−, rather than the divalent-bound ATP, is the native agonist for this receptor subtype (Dahlquist and Diamant, 1974; Cockcroft and Gomperts, 1979; Di Virgilio et al., 1998).
However, these experiments do not exclude the possibility that 2 mM Ca2+ causes a rightward shift in the potency of agonists of ∼0.5 log units independently of the concentration of the free acid form of agonists. To overcome limitations in the interpretation of such data, we performed experiments with variable agonist and bath Ca2+ concentrations leading to a constant free agonist concentration. Under such experimental conditions, we still observed a concentration-dependent inhibitory effect of bath Ca2+ on receptor functions, clearly indicating that this cation exerts inhibitory effects on gating independently of the concentration of orthosteric agonists. Recently, it was shown that the rP2X7R-H130A mutant is resistant to Mg2+ inhibition (Acuña-Castillo et al., 2007), reinforcing the hypothesis that this metal also acts through an allosteric mechanism. Here, we showed that this residue does not account for the effects of Ca2+ on agonist-induced activation of receptors. We also showed that Ca2+ does not act near the channel pore. Finally, our data argue against the hypothesis that Ca2+ effects on P2X7R activation is through Panx1 interacting with P2X7R. First, the expression of Panx1 protein in HEK293 cells in our experiments is very low, below the detection by Western blot analysis (Li et al., 2011), and P2X7R protein is overexpressed, suggesting that only a small fraction of receptors could be associated with Panx1, in contrast to almost complete inhibition of current by 10 mM Ca2+. Second, in cells bathed in medium containing 32 µM CBX, an inhibitor of Panx1 channels, the inhibitory effect of high Ca2+ on P2X7R activation was not changed.
In our study, the application of agonists and antagonist and their removal was performed using an ultra-rapid perfusion system with good time resolution (Yan et al., 2006). In general, the time required for washout of agonist inversely correlates with the potency of agonist for binding sites (Rettinger and Schmalzing, 2004). Consistent with this conclusion, we observed that the reduced potency of ATP compared with BzATP to activate P2X7R was accompanied by a rapid decay of current after washout of ATP and a slow decay of current after washout of BzATP. We also observed that removal of Ca2+ in the presence of 1 mM Mg2+ slowed the rates of receptor deactivation in a concentration- and time-dependent manner, further supporting the conclusion that bath Ca2+ affects the potency of agonist. However, during repetitive agonist application, we observed a progressive decrease in the rates of receptor deactivation, which we previously termed receptor sensitization (Yan et al., 2010). Here, we showed that this process occurs in both Ca2+-containing and -deficient medium, indicating that, like pore dilation, sensitization of receptors is not influenced by bath Ca2+.
To study the influence of allosteric ligands on the occupancy of P2X7R orthosteric sites, others have analyzed receptor deactivation in the presence and absence of orthosteric and allosteric agonists and antagonists (Bianchi and Macdonald, 2001). Previous studies indicated that A438079 is a human P2X7R-specific competitive antagonist (Nelson et al., 2006; McGaraughty et al., 2007). Our experiments showed that this compound also acts as a competitive orthosteric antagonist for rP2X7R. The potency of this antagonist was 200-fold higher than BzATP, a critical feature in experiments on the kinetics of receptor deactivation.
When A438079 was applied during the washout period, it facilitated deactivation in Ca2+-containing and Ca2+-deficient medium with similar rates. The same results were obtained with the T15E-P2X7R mutant, which is more sensitive to BzATP and, consequently, deactivates more slowly (Yan et al., 2010). We interpreted this observation as an indication that a fraction of BzATP released from the binding domain during the washout period rebinds, and that process is reduced or abolished in the presence of a more potent competitive antagonist. We also observed that deactivation of receptors in the presence of A438079 is facilitated by the addition of Ca2+ to bath medium, clearly indicating that the effect of Ca2+ on the rate of receptor deactivation is also present under conditions in which the rebinding effect of BzATP on the rate of deactivation was practically eliminated.
Some of these features of P2X7Rs were better understood by using a minimal kinetic Markov state model that incorporated not only the orthosteric regulation exerted by the agonist (which could lead to receptor sensitization) but also the allosteric regulation exerted by extracellular Ca2+. The model, which was previously developed in Yan et al. (2010), integrated the allostery of Ca2+ in a simplified manner by assuming that an increase in [Ca2+]e accelerated agonist unbinding and impeded agonist binding by factors that depended sigmoidally on [Ca2+]e. This model not only preserved the essential properties of current recording associated with agonist stimulation but also captured the diverse effects of [Ca2+]e on the channel. It showed that the loss of the second phase of the current at high [Ca2+]e is caused by a decline in the number of sensitized receptors, which in turn led to an increase in the rate of deactivation, because backward rates of unsensitized receptors are faster than those of sensitized receptors. The model also revealed that Ca2+-dependent and free BzATP–dependent dose–response curves obtained from a heterogeneous population of model cells are (decreasing and increasing, respectively) Hill-like functions with n = 4 and IC50 = 4.1 mM for the former and EC50 = 3.9 µM for the latter. The structures of these response curves were found to be independent of the assumed Ca2+ dependency of the inhibition, suggesting that the experimental dose–response curves obtained here might be affected by the small number of concentrations used.
The effect of extracellular Ca2+ on P2X7R could have been modeled by including additional states characterizing the allosteric binding of Ca2+ to the receptor. The lack of information about the number of these binding sites and the potential increase in complexity of the kinetic model associated with such an approach prompted us to model this effect using Ca2+-dependent Hill functions that directly acted on the binding and unbinding of the orthosteric agonist. Even though the use of this latter approach cannot determine how many allosteric binding sites the receptor may possess, it shows that the diverse effects of extracellular Ca2+ can all be accounted for by a common effect on binding and unbinding rates of agonist.
In summary, our investigations show that the binding of the orthosteric agonists ATP and BzATP to P2X7R causes gating transitions from closed to open to dilated states (Fig. 10 A), resulting in the generation of monophasic currents in intermediated concentration range and biphasic currents at higher agonist concentration (Fig. 10 B). We further show that calcium influx is not required for conformations associated with pore dilation that coincide with the transition from low to high affinity states for orthosteric ligands, causing a leftward shift in the potency of agonists (Fig. 10 C) accompanied by a delay in receptor deactivation (orthosteric facilitation). We show by two independent tests that calcium in physiological concentrations acts as a negative allosteric modulator (or allosteric inhibitor): (1) by activating receptors with fixed free agonist concentrations and variable bath calcium concentrations and (2) by studying receptor deactivation in the presence and absence of P2X7R-specific competitive antagonist. Furthermore, we used a kinetic model to uncover how calcium exerts its allosteric regulation of these receptors. Systematic scanning of ectodomain residues is needed to identify amino acids responsible for allosteric actions of calcium.
We are thankful for constructive discussions with Drs. M. Tomic and C. Coddou and for H130A mutant from Dr. J.P. Huidobro-Toro.
The authors were supported by the Intramural Research Program of the National Institutes of Health, National Institute of Child Health and Human Development (grants to Z. Yan and S.S. Stojilkovic), and National Institute of Diabetes and Digestive and Kidney Diseases (grants to A. Khadra and A. Sherman).
Author contributions: Z. Yan carried out the bulk of the experiments, and A. Khadra did the modeling.
Lawrence G. Palmer served as editor.